diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index 2a8385df3f934..e416c91ba52d2 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -1831,6 +1831,10 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
   bool hasScratchBaseForwardingHazard() const {
     return GFX1250Insts && getGeneration() == GFX12;
   }
+
+  /// \returns true if the subtarget requires a wait for xcnt before atomic
+  /// flat/global stores & rmw.
+  bool requiresWaitXCntBeforeAtomicStores() const { return GFX1250Insts; }
 };
 
 class GCNUserSGPRUsageInfo {
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.h b/llvm/lib/Target/AMDGPU/SIInstrInfo.h
index f7c7bb509c9ef..bcf8a86effe5a 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.h
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.h
@@ -1051,6 +1051,8 @@ class SIInstrInfo final : public AMDGPUGenInstrInfo {
       return AMDGPU::S_WAIT_DSCNT;
     case AMDGPU::S_WAIT_KMCNT_soft:
       return AMDGPU::S_WAIT_KMCNT;
+    case AMDGPU::S_WAIT_XCNT_soft:
+      return AMDGPU::S_WAIT_XCNT;
     default:
       return Opcode;
     }
diff --git a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
index 53f554eccb1fb..27e2ac5372796 100644
--- a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
@@ -587,7 +587,11 @@ class SIGfx12CacheControl : public SIGfx11CacheControl {
                       SIAtomicScope Scope, SIAtomicAddrSpace AddrSpace) const;
 
 public:
-  SIGfx12CacheControl(const GCNSubtarget &ST) : SIGfx11CacheControl(ST) {}
+  SIGfx12CacheControl(const GCNSubtarget &ST) : SIGfx11CacheControl(ST) {
+    // GFX12.0 and GFX12.5 memory models greatly overlap, and in some cases
+    // the behavior is the same if assuming GFX12.0 in CU mode.
+    assert(ST.hasGFX1250Insts() ? ST.isCuModeEnabled() : true);
+  }
 
   bool insertWait(MachineBasicBlock::iterator &MI, SIAtomicScope Scope,
                   SIAtomicAddrSpace AddrSpace, SIMemOp Op,
@@ -2340,12 +2344,16 @@ bool SIGfx12CacheControl::insertWait(MachineBasicBlock::iterator &MI,
         STORECnt |= true;
       break;
     case SIAtomicScope::WORKGROUP:
-      // In WGP mode the waves of a work-group can be executing on either CU of
-      // the WGP. Therefore need to wait for operations to complete to ensure
-      // they are visible to waves in the other CU as the L0 is per CU.
-      // Otherwise in CU mode and all waves of a work-group are on the same CU
-      // which shares the same L0.
-      if (!ST.isCuModeEnabled()) {
+      // GFX12.0:
+      //   In WGP mode the waves of a work-group can be executing on either CU
+      //   of the WGP. Therefore need to wait for operations to complete to
+      //   ensure they are visible to waves in the other CU as the L0 is per CU.
+      //   Otherwise in CU mode and all waves of a work-group are on the same CU
+      //   which shares the same L0.
+      //
+      // GFX12.5:
+      //   TODO DOCS
+      if (!ST.isCuModeEnabled() || ST.hasGFX1250Insts()) {
         if ((Op & SIMemOp::LOAD) != SIMemOp::NONE)
           LOADCnt |= true;
         if ((Op & SIMemOp::STORE) != SIMemOp::NONE)
@@ -2397,7 +2405,7 @@ bool SIGfx12CacheControl::insertWait(MachineBasicBlock::iterator &MI,
     //
     // This also applies to fences. Fences cannot pair with an instruction
     // tracked with bvh/samplecnt as we don't have any atomics that do that.
-    if (Order != AtomicOrdering::Acquire) {
+    if (Order != AtomicOrdering::Acquire && ST.hasImageInsts()) {
       BuildMI(MBB, MI, DL, TII->get(AMDGPU::S_WAIT_BVHCNT_soft)).addImm(0);
       BuildMI(MBB, MI, DL, TII->get(AMDGPU::S_WAIT_SAMPLECNT_soft)).addImm(0);
     }
@@ -2449,10 +2457,14 @@ bool SIGfx12CacheControl::insertAcquire(MachineBasicBlock::iterator &MI,
     ScopeImm = AMDGPU::CPol::SCOPE_DEV;
     break;
   case SIAtomicScope::WORKGROUP:
-    // In WGP mode the waves of a work-group can be executing on either CU of
-    // the WGP. Therefore we need to invalidate the L0 which is per CU.
-    // Otherwise in CU mode all waves of a work-group are on the same CU, and so
-    // the L0 does not need to be invalidated.
+    // GFX12.0:
+    //  In WGP mode the waves of a work-group can be executing on either CU of
+    //  the WGP. Therefore we need to invalidate the L0 which is per CU.
+    //  Otherwise in CU mode all waves of a work-group are on the same CU, and
+    //  so the L0 does not need to be invalidated.
+    //
+    // GFX12.5
+    //   TODO DOCS
     if (ST.isCuModeEnabled())
       return false;
 
@@ -2497,7 +2509,8 @@ bool SIGfx12CacheControl::insertRelease(MachineBasicBlock::iterator &MI,
   if (Pos == Position::AFTER)
     ++MI;
 
-  // global_wb is only necessary at system scope for gfx120x targets.
+  // global_wb is only necessary at system scope for GFX12.0,
+  // they're also necessary at device scope for GFX12.5.
   //
   // Emitting it for lower scopes is a slow no-op, so we omit it
   // for performance.
@@ -2507,6 +2520,12 @@ bool SIGfx12CacheControl::insertRelease(MachineBasicBlock::iterator &MI,
         .addImm(AMDGPU::CPol::SCOPE_SYS);
     break;
   case SIAtomicScope::AGENT:
+    // TODO DOCS
+    if (ST.hasGFX1250Insts()) {
+      BuildMI(MBB, MI, DL, TII->get(AMDGPU::GLOBAL_WB))
+          .addImm(AMDGPU::CPol::SCOPE_DEV);
+    }
+    break;
   case SIAtomicScope::WORKGROUP:
     // No WB necessary, but we still have to wait.
     break;
@@ -2569,17 +2588,31 @@ bool SIGfx12CacheControl::enableVolatileAndOrNonTemporal(
 }
 
 bool SIGfx12CacheControl::finalizeStore(MachineInstr &MI, bool Atomic) const {
-  MachineOperand *CPol = TII->getNamedOperand(MI, OpName::cpol);
-  if (!CPol)
-    return false;
+  assert(MI.mayStore() && "Not a Store inst");
+  const bool IsRMW = (MI.mayLoad() && MI.mayStore());
+  bool Changed = false;
+
+  // GFX12.5 only: xcnt wait is needed before flat and global atomics stores/rmw
+  if (Atomic && ST.requiresWaitXCntBeforeAtomicStores() && TII->isFLAT(MI)) {
+    MachineBasicBlock &MBB = *MI.getParent();
+    BuildMI(MBB, MI, MI.getDebugLoc(), TII->get(S_WAIT_XCNT_soft)).addImm(0);
+    Changed = true;
+  }
+
+  // Remaining fixes do not apply to RMWs
+  if (IsRMW)
+    return Changed;
 
+  MachineOperand *CPol = TII->getNamedOperand(MI, OpName::cpol);
+  if (!CPol) // Some vmem operations do not have a scope and are not concerned.
+    return Changed;
   const unsigned Scope = CPol->getImm() & CPol::SCOPE;
 
   // GFX12.0 only: Extra waits needed before system scope stores.
   if (!ST.hasGFX1250Insts()) {
     if (!Atomic && Scope == CPol::SCOPE_SYS)
       return insertWaitsBeforeSystemScopeStore(MI);
-    return false;
+    return Changed;
   }
 
   // GFX12.5 only: Require SCOPE_SE on stores that may hit the scratch address
@@ -2589,7 +2622,7 @@ bool SIGfx12CacheControl::finalizeStore(MachineInstr &MI, bool Atomic) const {
       (!ST.hasCUStores() || TII->mayAccessScratchThroughFlat(MI)))
     return setScope(MI, CPol::SCOPE_SE);
 
-  return false;
+  return Changed;
 }
 
 bool SIGfx12CacheControl::setAtomicScope(const MachineBasicBlock::iterator &MI,
@@ -2778,6 +2811,7 @@ bool SIMemoryLegalizer::expandAtomicCmpxchgOrRmw(const SIMemOpInfo &MOI,
   assert(MI->mayLoad() && MI->mayStore());
 
   bool Changed = false;
+  MachineInstr &RMWMI = *MI;
 
   if (MOI.isAtomic()) {
     const AtomicOrdering Order = MOI.getOrdering();
@@ -2812,6 +2846,7 @@ bool SIMemoryLegalizer::expandAtomicCmpxchgOrRmw(const SIMemOpInfo &MOI,
                                    Position::AFTER);
     }
 
+    Changed |= CC->finalizeStore(RMWMI, /*Atomic=*/true);
     return Changed;
   }
 
diff --git a/llvm/lib/Target/AMDGPU/SOPInstructions.td b/llvm/lib/Target/AMDGPU/SOPInstructions.td
index a003a46191a87..8012e9e6bc9bc 100644
--- a/llvm/lib/Target/AMDGPU/SOPInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SOPInstructions.td
@@ -1656,6 +1656,11 @@ let OtherPredicates = [HasImageInsts] in {
   def S_WAIT_KMCNT_soft : SOPP_Pseudo <"s_soft_wait_kmcnt", (ins s16imm:$simm16), "$simm16">;
 }
 
+
+let SubtargetPredicate = HasWaitXcnt in {
+  def S_WAIT_XCNT_soft : SOPP_Pseudo<"s_soft_wait_xcnt", (ins s16imm:$simm16), "$simm16">;
+}
+
 // Represents the point at which a wave must wait for all outstanding direct loads to LDS.
 // Typically inserted by the memory legalizer and consumed by SIInsertWaitcnts.
 
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll
index 481a2540eacb7..e886ea4fc6ac6 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll
@@ -1501,6 +1501,7 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat(ptr addrspace(1) %pt
 ; GFX1250-NEXT:    v_mul_f64_e32 v[0:1], 4.0, v[0:1]
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1571,6 +1572,9 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_agent(ptr addrspace(
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_mul_f64_e32 v[0:1], 4.0, v[0:1]
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1645,6 +1649,7 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_system(ptr addrspace
 ; GFX1250-NEXT:    v_mul_f64_e32 v[0:1], 4.0, v[0:1]
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1715,6 +1720,9 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_flush(ptr addrspace(
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_mul_f64_e32 v[0:1], 4.0, v[0:1]
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1792,6 +1800,7 @@ define double @global_atomic_fadd_f64_rtn_pat_agent(ptr addrspace(1) %ptr, doubl
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_mov_b64_e32 v[2:3], 4.0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v[0:1], v[0:1], v[2:3], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
@@ -1902,6 +1911,9 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_agent_safe(ptr addrs
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_mul_f64_e32 v[0:1], 4.0, v[0:1]
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1947,6 +1959,7 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat(ptr %ptr) #1 {
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
@@ -1987,6 +2000,9 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat_agent(ptr %ptr) #1 {
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], 4.0
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
@@ -2031,6 +2047,7 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat_system(ptr %ptr) #1 {
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -2107,6 +2124,7 @@ define double @flat_atomic_fadd_f64_rtn_pat_agent(ptr %ptr) #1 {
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_mov_b64_e32 v[2:3], 4.0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v[0:1], v[0:1], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
@@ -2190,6 +2208,9 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat_agent_safe(ptr %ptr) {
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], 4.0
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
@@ -2418,6 +2439,7 @@ define double @local_atomic_fadd_f64_rtn_pat(ptr addrspace(3) %ptr, double %data
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_mov_b64_e32 v[2:3], 4.0
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    ds_add_rtn_f64 v[0:1], v0, v[2:3]
 ; GFX1250-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/atomics-system-scope.ll b/llvm/test/CodeGen/AMDGPU/atomics-system-scope.ll
index 5fc9f4a0f8038..4bb2a13d02cc7 100644
--- a/llvm/test/CodeGen/AMDGPU/atomics-system-scope.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomics-system-scope.ll
@@ -364,6 +364,7 @@ define i16 @global_one_as_atomic_min_i16(ptr addrspace(1) %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    global_atomic_cmpswap_b32 v5, v[0:1], v[6:7], off th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
@@ -406,6 +407,7 @@ define i16 @global_one_as_atomic_umin_i16(ptr addrspace(1) %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    global_atomic_cmpswap_b32 v5, v[0:1], v[6:7], off th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
@@ -448,6 +450,7 @@ define i16 @global_one_as_atomic_max_i16(ptr addrspace(1) %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    global_atomic_cmpswap_b32 v5, v[0:1], v[6:7], off th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
@@ -490,6 +493,7 @@ define i16 @global_one_as_atomic_umax_i16(ptr addrspace(1) %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    global_atomic_cmpswap_b32 v5, v[0:1], v[6:7], off th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
@@ -1344,6 +1348,7 @@ define i16 @flat_one_as_atomic_min_i16(ptr %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_cmpswap_b32 v5, v[0:1], v[6:7] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
@@ -1386,6 +1391,7 @@ define i16 @flat_one_as_atomic_umin_i16(ptr %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_cmpswap_b32 v5, v[0:1], v[6:7] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
@@ -1428,6 +1434,7 @@ define i16 @flat_one_as_atomic_max_i16(ptr %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_cmpswap_b32 v5, v[0:1], v[6:7] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
@@ -1470,6 +1477,7 @@ define i16 @flat_one_as_atomic_umax_i16(ptr %ptr, i16 %val) {
 ; GFX1250-NEXT:    v_lshlrev_b32_e32 v5, v3, v5
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_and_or_b32 v6, v7, v4, v5
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_cmpswap_b32 v5, v[0:1], v[6:7] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
diff --git a/llvm/test/CodeGen/AMDGPU/flat-saddr-atomics.ll b/llvm/test/CodeGen/AMDGPU/flat-saddr-atomics.ll
index 004d3c0c1cf53..9e3348bbfdef6 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-saddr-atomics.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-saddr-atomics.ll
@@ -7,6 +7,8 @@
 define amdgpu_ps void @flat_xchg_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xchg_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_swap_b32 v0, v1, s[2:3] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -21,6 +23,8 @@ define amdgpu_ps void @flat_xchg_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 define amdgpu_ps void @flat_xchg_saddr_i32_nortn_offset_2047(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xchg_saddr_i32_nortn_offset_2047:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_swap_b32 v0, v1, s[2:3] offset:2047 scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -36,6 +40,8 @@ define amdgpu_ps void @flat_xchg_saddr_i32_nortn_offset_2047(ptr inreg %sbase, i
 define amdgpu_ps void @flat_xchg_saddr_i32_nortn_offset_neg2048(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xchg_saddr_i32_nortn_offset_neg2048:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_swap_b32 v0, v1, s[2:3] offset:-2048 scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -50,6 +56,8 @@ define amdgpu_ps void @flat_xchg_saddr_i32_nortn_offset_neg2048(ptr inreg %sbase
 define amdgpu_ps float @flat_xchg_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xchg_saddr_i32_rtn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_swap_b32 v0, v0, v1, s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -65,6 +73,8 @@ define amdgpu_ps float @flat_xchg_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset,
 define amdgpu_ps float @flat_xchg_saddr_i32_rtn_2048(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xchg_saddr_i32_rtn_2048:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_swap_b32 v0, v0, v1, s[2:3] offset:2048 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -81,6 +91,8 @@ define amdgpu_ps float @flat_xchg_saddr_i32_rtn_2048(ptr inreg %sbase, i32 %voff
 define amdgpu_ps float @flat_xchg_saddr_i32_rtn_neg2048(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xchg_saddr_i32_rtn_neg2048:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_swap_b32 v0, v0, v1, s[2:3] offset:-2048 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -109,6 +121,8 @@ define amdgpu_ps float @flat_xchg_saddr_uniform_ptr_in_vgprs_rtn(i32 %voffset, i
 ; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b32 v0, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -123,6 +137,8 @@ define amdgpu_ps float @flat_xchg_saddr_uniform_ptr_in_vgprs_rtn(i32 %voffset, i
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b32 v0, v[2:3], v1 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -145,6 +161,8 @@ define amdgpu_ps float @flat_xchg_saddr_uniform_ptr_in_vgprs_rtn_immoffset(i32 %
 ; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b32 v0, v0, v1, s[0:1] offset:42 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -159,6 +177,8 @@ define amdgpu_ps float @flat_xchg_saddr_uniform_ptr_in_vgprs_rtn_immoffset(i32 %
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b32 v0, v[2:3], v1 offset:42 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -182,6 +202,8 @@ define amdgpu_ps void @flat_xchg_saddr_uniform_ptr_in_vgprs_nortn(i32 %voffset,
 ; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -195,6 +217,8 @@ define amdgpu_ps void @flat_xchg_saddr_uniform_ptr_in_vgprs_nortn(i32 %voffset,
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b32 v[2:3], v1 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -215,6 +239,8 @@ define amdgpu_ps void @flat_xchg_saddr_uniform_ptr_in_vgprs_nortn_immoffset(i32
 ; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
 ; GFX1250-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] offset:42 scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -228,6 +254,8 @@ define amdgpu_ps void @flat_xchg_saddr_uniform_ptr_in_vgprs_nortn_immoffset(i32
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v3, null, 0, v3, vcc_lo
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b32 v[2:3], v1 offset:42 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -271,6 +299,8 @@ define amdgpu_ps <2 x float> @flat_xchg_saddr_i64_rtn(ptr inreg %sbase, i32 %vof
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB10_5
 ; GFX1250-SDAG-NEXT:  .LBB10_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -317,6 +347,8 @@ define amdgpu_ps <2 x float> @flat_xchg_saddr_i64_rtn(ptr inreg %sbase, i32 %vof
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB10_5
 ; GFX1250-GISEL-NEXT:  .LBB10_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -371,6 +403,8 @@ define amdgpu_ps <2 x float> @flat_xchg_saddr_i64_rtn_neg128(ptr inreg %sbase, i
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB11_5
 ; GFX1250-SDAG-NEXT:  .LBB11_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -420,6 +454,8 @@ define amdgpu_ps <2 x float> @flat_xchg_saddr_i64_rtn_neg128(ptr inreg %sbase, i
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB11_5
 ; GFX1250-GISEL-NEXT:  .LBB11_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -469,6 +505,8 @@ define amdgpu_ps void @flat_xchg_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:  .LBB12_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB12_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -506,6 +544,8 @@ define amdgpu_ps void @flat_xchg_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:  .LBB12_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB12_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b64 v0, v[4:5], s[2:3] scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -551,6 +591,8 @@ define amdgpu_ps void @flat_xchg_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-SDAG-NEXT:  .LBB13_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB13_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_swap_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -591,6 +633,8 @@ define amdgpu_ps void @flat_xchg_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-GISEL-NEXT:  .LBB13_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB13_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_swap_b64 v0, v[4:5], s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -621,6 +665,8 @@ define amdgpu_ps void @flat_xchg_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %v
 define amdgpu_ps float @flat_add_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_add_saddr_i32_rtn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_u32 v0, v0, v1, s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -636,6 +682,8 @@ define amdgpu_ps float @flat_add_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i
 define amdgpu_ps float @flat_add_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_add_saddr_i32_rtn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_u32 v0, v0, v1, s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -652,6 +700,8 @@ define amdgpu_ps float @flat_add_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %vof
 define amdgpu_ps void @flat_add_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_add_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_u32 v0, v1, s[2:3] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -665,6 +715,8 @@ define amdgpu_ps void @flat_add_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 define amdgpu_ps void @flat_add_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_add_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_u32 v0, v1, s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -699,6 +751,8 @@ define amdgpu_ps <2 x float> @flat_add_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB18_5
 ; GFX1250-SDAG-NEXT:  .LBB18_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_add_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -745,6 +799,8 @@ define amdgpu_ps <2 x float> @flat_add_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB18_5
 ; GFX1250-GISEL-NEXT:  .LBB18_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_add_u64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -799,6 +855,8 @@ define amdgpu_ps <2 x float> @flat_add_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB19_5
 ; GFX1250-SDAG-NEXT:  .LBB19_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_add_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -848,6 +906,8 @@ define amdgpu_ps <2 x float> @flat_add_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB19_5
 ; GFX1250-GISEL-NEXT:  .LBB19_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_add_u64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -897,6 +957,8 @@ define amdgpu_ps void @flat_add_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:  .LBB20_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB20_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_add_u64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -937,6 +999,8 @@ define amdgpu_ps void @flat_add_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:  .LBB20_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB20_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_add_u64 v0, v[4:5], s[2:3] scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -985,6 +1049,8 @@ define amdgpu_ps void @flat_add_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-SDAG-NEXT:  .LBB21_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB21_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_add_u64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1028,6 +1094,8 @@ define amdgpu_ps void @flat_add_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-GISEL-NEXT:  .LBB21_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB21_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_add_u64 v0, v[4:5], s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1061,6 +1129,8 @@ define amdgpu_ps void @flat_add_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 define amdgpu_ps float @flat_sub_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_sub_saddr_i32_rtn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_sub_u32 v0, v0, v1, s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1076,6 +1146,8 @@ define amdgpu_ps float @flat_sub_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i
 define amdgpu_ps float @flat_sub_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_sub_saddr_i32_rtn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_sub_u32 v0, v0, v1, s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1092,6 +1164,8 @@ define amdgpu_ps float @flat_sub_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %vof
 define amdgpu_ps void @flat_sub_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_sub_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_sub_u32 v0, v1, s[2:3] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1105,6 +1179,8 @@ define amdgpu_ps void @flat_sub_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 define amdgpu_ps void @flat_sub_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_sub_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_sub_u32 v0, v1, s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1139,6 +1215,8 @@ define amdgpu_ps <2 x float> @flat_sub_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB26_5
 ; GFX1250-SDAG-NEXT:  .LBB26_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_sub_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1185,6 +1263,8 @@ define amdgpu_ps <2 x float> @flat_sub_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB26_5
 ; GFX1250-GISEL-NEXT:  .LBB26_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_sub_u64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1239,6 +1319,8 @@ define amdgpu_ps <2 x float> @flat_sub_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB27_5
 ; GFX1250-SDAG-NEXT:  .LBB27_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_sub_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1288,6 +1370,8 @@ define amdgpu_ps <2 x float> @flat_sub_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB27_5
 ; GFX1250-GISEL-NEXT:  .LBB27_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_sub_u64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1337,6 +1421,8 @@ define amdgpu_ps void @flat_sub_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:  .LBB28_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB28_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_sub_u64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1377,6 +1463,8 @@ define amdgpu_ps void @flat_sub_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:  .LBB28_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB28_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_sub_u64 v0, v[4:5], s[2:3] scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1425,6 +1513,8 @@ define amdgpu_ps void @flat_sub_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-SDAG-NEXT:  .LBB29_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB29_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_sub_u64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1468,6 +1558,8 @@ define amdgpu_ps void @flat_sub_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-GISEL-NEXT:  .LBB29_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB29_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_sub_u64 v0, v[4:5], s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1501,6 +1593,8 @@ define amdgpu_ps void @flat_sub_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 define amdgpu_ps float @flat_and_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_and_saddr_i32_rtn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_and_b32 v0, v0, v1, s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1516,6 +1610,8 @@ define amdgpu_ps float @flat_and_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i
 define amdgpu_ps float @flat_and_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_and_saddr_i32_rtn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_and_b32 v0, v0, v1, s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1532,6 +1628,8 @@ define amdgpu_ps float @flat_and_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %vof
 define amdgpu_ps void @flat_and_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_and_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_and_b32 v0, v1, s[2:3] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1545,6 +1643,8 @@ define amdgpu_ps void @flat_and_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 define amdgpu_ps void @flat_and_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_and_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_and_b32 v0, v1, s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1579,6 +1679,8 @@ define amdgpu_ps <2 x float> @flat_and_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB34_5
 ; GFX1250-SDAG-NEXT:  .LBB34_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_and_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1626,6 +1728,8 @@ define amdgpu_ps <2 x float> @flat_and_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB34_5
 ; GFX1250-GISEL-NEXT:  .LBB34_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_and_b64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1681,6 +1785,8 @@ define amdgpu_ps <2 x float> @flat_and_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB35_5
 ; GFX1250-SDAG-NEXT:  .LBB35_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_and_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1731,6 +1837,8 @@ define amdgpu_ps <2 x float> @flat_and_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB35_5
 ; GFX1250-GISEL-NEXT:  .LBB35_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_and_b64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1781,6 +1889,8 @@ define amdgpu_ps void @flat_and_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:  .LBB36_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB36_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_and_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1822,6 +1932,8 @@ define amdgpu_ps void @flat_and_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:  .LBB36_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB36_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_and_b64 v0, v[4:5], s[2:3] scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1871,6 +1983,8 @@ define amdgpu_ps void @flat_and_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-SDAG-NEXT:  .LBB37_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB37_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_and_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -1915,6 +2029,8 @@ define amdgpu_ps void @flat_and_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-GISEL-NEXT:  .LBB37_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB37_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_and_b64 v0, v[4:5], s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -1949,6 +2065,8 @@ define amdgpu_ps void @flat_and_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 define amdgpu_ps float @flat_or_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_or_saddr_i32_rtn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_or_b32 v0, v0, v1, s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1964,6 +2082,8 @@ define amdgpu_ps float @flat_or_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i3
 define amdgpu_ps float @flat_or_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_or_saddr_i32_rtn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_or_b32 v0, v0, v1, s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1980,6 +2100,8 @@ define amdgpu_ps float @flat_or_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %voff
 define amdgpu_ps void @flat_or_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_or_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_or_b32 v0, v1, s[2:3] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -1993,6 +2115,8 @@ define amdgpu_ps void @flat_or_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset, i
 define amdgpu_ps void @flat_or_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_or_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_or_b32 v0, v1, s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -2027,6 +2151,8 @@ define amdgpu_ps <2 x float> @flat_or_saddr_i64_rtn(ptr inreg %sbase, i32 %voffs
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB42_5
 ; GFX1250-SDAG-NEXT:  .LBB42_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_or_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2074,6 +2200,8 @@ define amdgpu_ps <2 x float> @flat_or_saddr_i64_rtn(ptr inreg %sbase, i32 %voffs
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB42_5
 ; GFX1250-GISEL-NEXT:  .LBB42_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_or_b64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2129,6 +2257,8 @@ define amdgpu_ps <2 x float> @flat_or_saddr_i64_rtn_neg128(ptr inreg %sbase, i32
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB43_5
 ; GFX1250-SDAG-NEXT:  .LBB43_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_or_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2179,6 +2309,8 @@ define amdgpu_ps <2 x float> @flat_or_saddr_i64_rtn_neg128(ptr inreg %sbase, i32
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB43_5
 ; GFX1250-GISEL-NEXT:  .LBB43_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_or_b64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2229,6 +2361,8 @@ define amdgpu_ps void @flat_or_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset, i
 ; GFX1250-SDAG-NEXT:  .LBB44_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB44_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_or_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2270,6 +2404,8 @@ define amdgpu_ps void @flat_or_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset, i
 ; GFX1250-GISEL-NEXT:  .LBB44_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB44_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_or_b64 v0, v[4:5], s[2:3] scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2319,6 +2455,8 @@ define amdgpu_ps void @flat_or_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vof
 ; GFX1250-SDAG-NEXT:  .LBB45_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB45_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_or_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2363,6 +2501,8 @@ define amdgpu_ps void @flat_or_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vof
 ; GFX1250-GISEL-NEXT:  .LBB45_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB45_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_or_b64 v0, v[4:5], s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2397,6 +2537,8 @@ define amdgpu_ps void @flat_or_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vof
 define amdgpu_ps float @flat_xor_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xor_saddr_i32_rtn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_xor_b32 v0, v0, v1, s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -2412,6 +2554,8 @@ define amdgpu_ps float @flat_xor_saddr_i32_rtn(ptr inreg %sbase, i32 %voffset, i
 define amdgpu_ps float @flat_xor_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xor_saddr_i32_rtn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_xor_b32 v0, v0, v1, s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -2428,6 +2572,8 @@ define amdgpu_ps float @flat_xor_saddr_i32_rtn_neg128(ptr inreg %sbase, i32 %vof
 define amdgpu_ps void @flat_xor_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xor_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_xor_b32 v0, v1, s[2:3] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -2441,6 +2587,8 @@ define amdgpu_ps void @flat_xor_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 define amdgpu_ps void @flat_xor_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %voffset, i32 %data) {
 ; GFX1250-LABEL: flat_xor_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_xor_b32 v0, v1, s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    global_inv scope:SCOPE_DEV
@@ -2475,6 +2623,8 @@ define amdgpu_ps <2 x float> @flat_xor_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB50_5
 ; GFX1250-SDAG-NEXT:  .LBB50_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_xor_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2522,6 +2672,8 @@ define amdgpu_ps <2 x float> @flat_xor_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB50_5
 ; GFX1250-GISEL-NEXT:  .LBB50_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_xor_b64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2577,6 +2729,8 @@ define amdgpu_ps <2 x float> @flat_xor_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB51_5
 ; GFX1250-SDAG-NEXT:  .LBB51_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_xor_b64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2627,6 +2781,8 @@ define amdgpu_ps <2 x float> @flat_xor_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB51_5
 ; GFX1250-GISEL-NEXT:  .LBB51_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_xor_b64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2677,6 +2833,8 @@ define amdgpu_ps void @flat_xor_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:  .LBB52_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB52_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_xor_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2718,6 +2876,8 @@ define amdgpu_ps void @flat_xor_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:  .LBB52_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB52_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_xor_b64 v0, v[4:5], s[2:3] scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2767,6 +2927,8 @@ define amdgpu_ps void @flat_xor_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-SDAG-NEXT:  .LBB53_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB53_3: ; %atomicrmw.global
+; GFX1250-SDAG-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-SDAG-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-SDAG-NEXT:    flat_atomic_xor_b64 v[0:1], v[2:3] scope:SCOPE_DEV
 ; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    global_inv scope:SCOPE_DEV
@@ -2811,6 +2973,8 @@ define amdgpu_ps void @flat_xor_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-GISEL-NEXT:  .LBB53_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB53_3: ; %atomicrmw.global
+; GFX1250-GISEL-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-GISEL-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-GISEL-NEXT:    flat_atomic_xor_b64 v0, v[4:5], s[2:3] offset:-128 scope:SCOPE_DEV
 ; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    global_inv scope:SCOPE_DEV
@@ -2873,7 +3037,7 @@ define amdgpu_ps void @flat_max_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-LABEL: flat_max_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_max_i32 v0, v1, s[2:3]
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -2885,7 +3049,7 @@ define amdgpu_ps void @flat_max_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-LABEL: flat_max_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_max_i32 v0, v1, s[2:3] offset:-128
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -2914,20 +3078,17 @@ define amdgpu_ps <2 x float> @flat_max_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB58_4
 ; GFX1250-SDAG-NEXT:  .LBB58_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB58_5
 ; GFX1250-SDAG-NEXT:  .LBB58_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_i64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB58_2
 ; GFX1250-SDAG-NEXT:  .LBB58_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -2961,20 +3122,17 @@ define amdgpu_ps <2 x float> @flat_max_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB58_4
 ; GFX1250-GISEL-NEXT:  .LBB58_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB58_5
 ; GFX1250-GISEL-NEXT:  .LBB58_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_i64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB58_2
 ; GFX1250-GISEL-NEXT:  .LBB58_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -3016,20 +3174,17 @@ define amdgpu_ps <2 x float> @flat_max_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB59_4
 ; GFX1250-SDAG-NEXT:  .LBB59_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB59_5
 ; GFX1250-SDAG-NEXT:  .LBB59_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_i64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB59_2
 ; GFX1250-SDAG-NEXT:  .LBB59_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -3066,20 +3221,17 @@ define amdgpu_ps <2 x float> @flat_max_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB59_4
 ; GFX1250-GISEL-NEXT:  .LBB59_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB59_5
 ; GFX1250-GISEL-NEXT:  .LBB59_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_i64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB59_2
 ; GFX1250-GISEL-NEXT:  .LBB59_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -3120,9 +3272,9 @@ define amdgpu_ps void @flat_max_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB60_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_i64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB60_2
@@ -3159,9 +3311,9 @@ define amdgpu_ps void @flat_max_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB60_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_i64 v0, v[4:5], s[2:3]
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB60_2
@@ -3206,9 +3358,9 @@ define amdgpu_ps void @flat_max_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB61_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_i64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB61_2
@@ -3248,9 +3400,9 @@ define amdgpu_ps void @flat_max_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB61_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_i64 v0, v[4:5], s[2:3] offset:-128
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB61_2
@@ -3307,7 +3459,7 @@ define amdgpu_ps void @flat_min_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-LABEL: flat_min_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_min_i32 v0, v1, s[2:3]
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -3319,7 +3471,7 @@ define amdgpu_ps void @flat_min_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-LABEL: flat_min_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_min_i32 v0, v1, s[2:3] offset:-128
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -3348,20 +3500,17 @@ define amdgpu_ps <2 x float> @flat_min_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB66_4
 ; GFX1250-SDAG-NEXT:  .LBB66_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB66_5
 ; GFX1250-SDAG-NEXT:  .LBB66_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_i64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB66_2
 ; GFX1250-SDAG-NEXT:  .LBB66_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -3395,20 +3544,17 @@ define amdgpu_ps <2 x float> @flat_min_saddr_i64_rtn(ptr inreg %sbase, i32 %voff
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB66_4
 ; GFX1250-GISEL-NEXT:  .LBB66_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB66_5
 ; GFX1250-GISEL-NEXT:  .LBB66_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_i64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB66_2
 ; GFX1250-GISEL-NEXT:  .LBB66_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -3450,20 +3596,17 @@ define amdgpu_ps <2 x float> @flat_min_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB67_4
 ; GFX1250-SDAG-NEXT:  .LBB67_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB67_5
 ; GFX1250-SDAG-NEXT:  .LBB67_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_i64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB67_2
 ; GFX1250-SDAG-NEXT:  .LBB67_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -3500,20 +3643,17 @@ define amdgpu_ps <2 x float> @flat_min_saddr_i64_rtn_neg128(ptr inreg %sbase, i3
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB67_4
 ; GFX1250-GISEL-NEXT:  .LBB67_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB67_5
 ; GFX1250-GISEL-NEXT:  .LBB67_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_i64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB67_2
 ; GFX1250-GISEL-NEXT:  .LBB67_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -3554,9 +3694,9 @@ define amdgpu_ps void @flat_min_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB68_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_i64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB68_2
@@ -3593,9 +3733,9 @@ define amdgpu_ps void @flat_min_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB68_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_i64 v0, v[4:5], s[2:3]
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB68_2
@@ -3640,9 +3780,9 @@ define amdgpu_ps void @flat_min_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB69_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_i64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB69_2
@@ -3682,9 +3822,9 @@ define amdgpu_ps void @flat_min_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %vo
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB69_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_i64 v0, v[4:5], s[2:3] offset:-128
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB69_2
@@ -3741,7 +3881,7 @@ define amdgpu_ps void @flat_umax_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-LABEL: flat_umax_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_max_u32 v0, v1, s[2:3]
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -3753,7 +3893,7 @@ define amdgpu_ps void @flat_umax_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-LABEL: flat_umax_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_max_u32 v0, v1, s[2:3] offset:-128
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -3782,20 +3922,17 @@ define amdgpu_ps <2 x float> @flat_umax_saddr_i64_rtn(ptr inreg %sbase, i32 %vof
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB74_4
 ; GFX1250-SDAG-NEXT:  .LBB74_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB74_5
 ; GFX1250-SDAG-NEXT:  .LBB74_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB74_2
 ; GFX1250-SDAG-NEXT:  .LBB74_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -3829,20 +3966,17 @@ define amdgpu_ps <2 x float> @flat_umax_saddr_i64_rtn(ptr inreg %sbase, i32 %vof
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB74_4
 ; GFX1250-GISEL-NEXT:  .LBB74_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB74_5
 ; GFX1250-GISEL-NEXT:  .LBB74_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_u64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB74_2
 ; GFX1250-GISEL-NEXT:  .LBB74_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -3884,20 +4018,17 @@ define amdgpu_ps <2 x float> @flat_umax_saddr_i64_rtn_neg128(ptr inreg %sbase, i
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB75_4
 ; GFX1250-SDAG-NEXT:  .LBB75_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB75_5
 ; GFX1250-SDAG-NEXT:  .LBB75_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB75_2
 ; GFX1250-SDAG-NEXT:  .LBB75_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -3934,20 +4065,17 @@ define amdgpu_ps <2 x float> @flat_umax_saddr_i64_rtn_neg128(ptr inreg %sbase, i
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB75_4
 ; GFX1250-GISEL-NEXT:  .LBB75_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB75_5
 ; GFX1250-GISEL-NEXT:  .LBB75_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_u64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB75_2
 ; GFX1250-GISEL-NEXT:  .LBB75_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -3988,9 +4116,9 @@ define amdgpu_ps void @flat_umax_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB76_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_u64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB76_2
@@ -4027,9 +4155,9 @@ define amdgpu_ps void @flat_umax_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB76_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_u64 v0, v[4:5], s[2:3]
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB76_2
@@ -4074,9 +4202,9 @@ define amdgpu_ps void @flat_umax_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB77_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_max_u64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB77_2
@@ -4116,9 +4244,9 @@ define amdgpu_ps void @flat_umax_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB77_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_max_u64 v0, v[4:5], s[2:3] offset:-128
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB77_2
@@ -4175,7 +4303,7 @@ define amdgpu_ps void @flat_umin_saddr_i32_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-LABEL: flat_umin_saddr_i32_nortn:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_min_u32 v0, v1, s[2:3]
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -4187,7 +4315,7 @@ define amdgpu_ps void @flat_umin_saddr_i32_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-LABEL: flat_umin_saddr_i32_nortn_neg128:
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    flat_atomic_min_u32 v0, v1, s[2:3] offset:-128
-; GFX1250-NEXT:    s_wait_dscnt 0x0
+; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-NEXT:    s_endpgm
   %zext.offset = zext i32 %voffset to i64
   %gep0 = getelementptr inbounds i8, ptr %sbase, i64 %zext.offset
@@ -4216,20 +4344,17 @@ define amdgpu_ps <2 x float> @flat_umin_saddr_i64_rtn(ptr inreg %sbase, i32 %vof
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB82_4
 ; GFX1250-SDAG-NEXT:  .LBB82_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB82_5
 ; GFX1250-SDAG-NEXT:  .LBB82_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB82_2
 ; GFX1250-SDAG-NEXT:  .LBB82_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -4263,20 +4388,17 @@ define amdgpu_ps <2 x float> @flat_umin_saddr_i64_rtn(ptr inreg %sbase, i32 %vof
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB82_4
 ; GFX1250-GISEL-NEXT:  .LBB82_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB82_5
 ; GFX1250-GISEL-NEXT:  .LBB82_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_u64 v[0:1], v3, v[4:5], s[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB82_2
 ; GFX1250-GISEL-NEXT:  .LBB82_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -4318,20 +4440,17 @@ define amdgpu_ps <2 x float> @flat_umin_saddr_i64_rtn_neg128(ptr inreg %sbase, i
 ; GFX1250-SDAG-NEXT:    s_cbranch_execnz .LBB83_4
 ; GFX1250-SDAG-NEXT:  .LBB83_2: ; %atomicrmw.phi
 ; GFX1250-SDAG-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_branch .LBB83_5
 ; GFX1250-SDAG-NEXT:  .LBB83_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_u64 v[0:1], v[4:5], v[2:3] th:TH_ATOMIC_RETURN
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB83_2
 ; GFX1250-SDAG-NEXT:  .LBB83_4: ; %atomicrmw.private
 ; GFX1250-SDAG-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-SDAG-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[4:5]
-; GFX1250-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-SDAG-NEXT:    v_subrev_nc_u32_e32 v0, s1, v4
 ; GFX1250-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-SDAG-NEXT:    v_cndmask_b32_e32 v4, -1, v0, vcc_lo
@@ -4368,20 +4487,17 @@ define amdgpu_ps <2 x float> @flat_umin_saddr_i64_rtn_neg128(ptr inreg %sbase, i
 ; GFX1250-GISEL-NEXT:    s_cbranch_execnz .LBB83_4
 ; GFX1250-GISEL-NEXT:  .LBB83_2: ; %atomicrmw.phi
 ; GFX1250-GISEL-NEXT:    s_or_b32 exec_lo, exec_lo, s0
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_branch .LBB83_5
 ; GFX1250-GISEL-NEXT:  .LBB83_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_u64 v[0:1], v3, v[4:5], s[2:3] offset:-128 th:TH_ATOMIC_RETURN
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr6
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
+; GFX1250-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB83_2
 ; GFX1250-GISEL-NEXT:  .LBB83_4: ; %atomicrmw.private
 ; GFX1250-GISEL-NEXT:    s_mov_b32 s1, src_flat_scratch_base_lo
 ; GFX1250-GISEL-NEXT:    v_cmp_ne_u64_e32 vcc_lo, 0, v[6:7]
-; GFX1250-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-GISEL-NEXT:    v_subrev_nc_u32_e32 v0, s1, v6
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_cndmask_b32_e32 v6, -1, v0, vcc_lo
@@ -4422,9 +4538,9 @@ define amdgpu_ps void @flat_umin_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB84_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_u64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB84_2
@@ -4461,9 +4577,9 @@ define amdgpu_ps void @flat_umin_saddr_i64_nortn(ptr inreg %sbase, i32 %voffset,
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB84_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_u64 v0, v[4:5], s[2:3]
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB84_2
@@ -4508,9 +4624,9 @@ define amdgpu_ps void @flat_umin_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-SDAG-NEXT:    s_endpgm
 ; GFX1250-SDAG-NEXT:  .LBB85_3: ; %atomicrmw.global
 ; GFX1250-SDAG-NEXT:    flat_atomic_min_u64 v[0:1], v[2:3]
-; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr0_vgpr1
 ; GFX1250-SDAG-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX1250-SDAG-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-SDAG-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-SDAG-NEXT:    s_cbranch_execz .LBB85_2
@@ -4550,9 +4666,9 @@ define amdgpu_ps void @flat_umin_saddr_i64_nortn_neg128(ptr inreg %sbase, i32 %v
 ; GFX1250-GISEL-NEXT:    s_endpgm
 ; GFX1250-GISEL-NEXT:  .LBB85_3: ; %atomicrmw.global
 ; GFX1250-GISEL-NEXT:    flat_atomic_min_u64 v0, v[4:5], s[2:3] offset:-128
-; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr2
 ; GFX1250-GISEL-NEXT:    ; implicit-def: $vgpr4_vgpr5
+; GFX1250-GISEL-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_and_not1_saveexec_b32 s0, s0
 ; GFX1250-GISEL-NEXT:    s_cbranch_execz .LBB85_2
diff --git a/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll b/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll
index 0cb2b0b7df3d2..c68871e7c84fa 100644
--- a/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll
@@ -1473,6 +1473,7 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat(ptr addrspace(1) %pt
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1513,6 +1514,9 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_agent(ptr addrspace(
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], 4.0
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1557,6 +1561,7 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_system(ptr addrspace
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1597,6 +1602,9 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_flush(ptr addrspace(
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], 4.0
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1673,6 +1681,7 @@ define double @global_atomic_fadd_f64_rtn_pat_agent(ptr addrspace(1) %ptr, doubl
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_mov_b64_e32 v[2:3], 4.0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v[0:1], v[0:1], v[2:3], off th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
@@ -1765,6 +1774,9 @@ define amdgpu_kernel void @global_atomic_fadd_f64_noret_pat_agent_safe(ptr addrs
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], 4.0
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    global_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1809,6 +1821,7 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat(ptr %ptr) #1 {
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
@@ -1849,6 +1862,9 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat_agent(ptr %ptr) #1 {
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], 4.0
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
@@ -1893,6 +1909,7 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat_system(ptr %ptr) #1 {
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX1250-NEXT:    global_wb scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
@@ -1969,6 +1986,7 @@ define double @flat_atomic_fadd_f64_rtn_pat_agent(ptr %ptr) #1 {
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_mov_b64_e32 v[2:3], 4.0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v[0:1], v[0:1], v[2:3] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
@@ -2063,6 +2081,9 @@ define amdgpu_kernel void @flat_atomic_fadd_f64_noret_pat_agent_safe(ptr %ptr) {
 ; GFX1250-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
 ; GFX1250-NEXT:    v_mov_b64_e32 v[0:1], 4.0
 ; GFX1250-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1250-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-NEXT:    s_wait_storecnt 0x0
+; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    flat_atomic_add_f64 v2, v[0:1], s[0:1] scope:SCOPE_DEV
 ; GFX1250-NEXT:    s_wait_storecnt_dscnt 0x0
@@ -2136,6 +2157,7 @@ define double @local_atomic_fadd_f64_rtn(ptr addrspace(3) %ptr, double %data) {
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    ds_add_rtn_f64 v[0:1], v0, v[2:3]
 ; GFX1250-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
@@ -2275,6 +2297,7 @@ define double @local_atomic_fadd_f64_rtn_pat(ptr addrspace(3) %ptr, double %data
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_mov_b64_e32 v[2:3], 4.0
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    ds_add_rtn_f64 v[0:1], v0, v[2:3]
 ; GFX1250-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
@@ -2307,6 +2330,7 @@ define double @local_atomic_fadd_f64_rtn_ieee_unsafe(ptr addrspace(3) %ptr, doub
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    ds_add_rtn_f64 v[0:1], v0, v[2:3]
 ; GFX1250-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
@@ -2339,6 +2363,7 @@ define double @local_atomic_fadd_f64_rtn_ieee_safe(ptr addrspace(3) %ptr, double
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_dual_mov_b32 v3, v2 :: v_dual_mov_b32 v2, v1
+; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    ds_add_rtn_f64 v[0:1], v0, v[2:3]
 ; GFX1250-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-global.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-global.ll
index 80445f793934b..260ad842fee60 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-global.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-global.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @workgroup_acquire_fence() {
 ; GFX6-LABEL: workgroup_acquire_fence:
@@ -78,6 +79,12 @@ define amdgpu_kernel void @workgroup_acquire_fence() {
 ; GFX12-CU-LABEL: workgroup_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") acquire, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -145,6 +152,12 @@ define amdgpu_kernel void @workgroup_release_fence() {
 ; GFX12-CU-LABEL: workgroup_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") release, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -217,6 +230,12 @@ define amdgpu_kernel void @workgroup_acq_rel_fence() {
 ; GFX12-CU-LABEL: workgroup_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -289,6 +308,12 @@ define amdgpu_kernel void @workgroup_seq_cst_fence() {
 ; GFX12-CU-LABEL: workgroup_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -359,6 +384,12 @@ define amdgpu_kernel void @workgroup_one_as_acquire_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") acquire, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -426,6 +457,12 @@ define amdgpu_kernel void @workgroup_one_as_release_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") release, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -498,6 +535,12 @@ define amdgpu_kernel void @workgroup_one_as_acq_rel_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -570,6 +613,12 @@ define amdgpu_kernel void @workgroup_one_as_seq_cst_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -662,6 +711,13 @@ define amdgpu_kernel void @agent_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") acquire, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -744,6 +800,13 @@ define amdgpu_kernel void @agent_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") release, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -842,6 +905,14 @@ define amdgpu_kernel void @agent_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -940,6 +1011,14 @@ define amdgpu_kernel void @agent_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1032,6 +1111,13 @@ define amdgpu_kernel void @agent_one_as_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") acquire, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1114,6 +1200,13 @@ define amdgpu_kernel void @agent_one_as_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") release, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1212,6 +1305,14 @@ define amdgpu_kernel void @agent_one_as_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1310,6 +1411,14 @@ define amdgpu_kernel void @agent_one_as_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1404,6 +1513,13 @@ define amdgpu_kernel void @system_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence acquire, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1490,6 +1606,13 @@ define amdgpu_kernel void @system_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence release, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1594,6 +1717,14 @@ define amdgpu_kernel void @system_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence acq_rel, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1698,6 +1829,14 @@ define amdgpu_kernel void @system_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence seq_cst, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1792,6 +1931,13 @@ define amdgpu_kernel void @system_one_as_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") acquire, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1878,6 +2024,13 @@ define amdgpu_kernel void @system_one_as_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") release, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -1982,6 +2135,14 @@ define amdgpu_kernel void @system_one_as_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
@@ -2086,6 +2247,14 @@ define amdgpu_kernel void @system_one_as_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"global"}
   ret void
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-local.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-local.ll
index 7a419a5031ba9..767dbc1432242 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-local.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence-mmra-local.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @workgroup_acquire_fence() {
 ; GFX6-LABEL: workgroup_acquire_fence:
@@ -76,6 +77,11 @@ define amdgpu_kernel void @workgroup_acquire_fence() {
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") acquire, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -142,6 +148,10 @@ define amdgpu_kernel void @workgroup_release_fence() {
 ; GFX12-CU-LABEL: workgroup_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") release, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -208,6 +218,10 @@ define amdgpu_kernel void @workgroup_acq_rel_fence() {
 ; GFX12-CU-LABEL: workgroup_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -274,6 +288,10 @@ define amdgpu_kernel void @workgroup_seq_cst_fence() {
 ; GFX12-CU-LABEL: workgroup_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -331,6 +349,10 @@ define amdgpu_kernel void @workgroup_one_as_acquire_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") acquire, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -388,6 +410,10 @@ define amdgpu_kernel void @workgroup_one_as_release_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") release, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -445,6 +471,10 @@ define amdgpu_kernel void @workgroup_one_as_acq_rel_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -502,6 +532,10 @@ define amdgpu_kernel void @workgroup_one_as_seq_cst_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -570,6 +604,11 @@ define amdgpu_kernel void @agent_acquire_fence() {
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") acquire, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -636,6 +675,10 @@ define amdgpu_kernel void @agent_release_fence() {
 ; GFX12-CU-LABEL: agent_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") release, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -702,6 +745,10 @@ define amdgpu_kernel void @agent_acq_rel_fence() {
 ; GFX12-CU-LABEL: agent_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -768,6 +815,10 @@ define amdgpu_kernel void @agent_seq_cst_fence() {
 ; GFX12-CU-LABEL: agent_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -825,6 +876,10 @@ define amdgpu_kernel void @agent_one_as_acquire_fence() {
 ; GFX12-CU-LABEL: agent_one_as_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") acquire, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -882,6 +937,10 @@ define amdgpu_kernel void @agent_one_as_release_fence() {
 ; GFX12-CU-LABEL: agent_one_as_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") release, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -939,6 +998,10 @@ define amdgpu_kernel void @agent_one_as_acq_rel_fence() {
 ; GFX12-CU-LABEL: agent_one_as_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -996,6 +1059,10 @@ define amdgpu_kernel void @agent_one_as_seq_cst_fence() {
 ; GFX12-CU-LABEL: agent_one_as_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1064,6 +1131,11 @@ define amdgpu_kernel void @system_acquire_fence() {
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence acquire, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1130,6 +1202,10 @@ define amdgpu_kernel void @system_release_fence() {
 ; GFX12-CU-LABEL: system_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence release, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1196,6 +1272,10 @@ define amdgpu_kernel void @system_acq_rel_fence() {
 ; GFX12-CU-LABEL: system_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence acq_rel, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1262,6 +1342,10 @@ define amdgpu_kernel void @system_seq_cst_fence() {
 ; GFX12-CU-LABEL: system_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence seq_cst, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1319,6 +1403,10 @@ define amdgpu_kernel void @system_one_as_acquire_fence() {
 ; GFX12-CU-LABEL: system_one_as_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") acquire, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1376,6 +1464,10 @@ define amdgpu_kernel void @system_one_as_release_fence() {
 ; GFX12-CU-LABEL: system_one_as_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") release, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1433,6 +1525,10 @@ define amdgpu_kernel void @system_one_as_acq_rel_fence() {
 ; GFX12-CU-LABEL: system_one_as_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") acq_rel, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
@@ -1490,6 +1586,10 @@ define amdgpu_kernel void @system_one_as_seq_cst_fence() {
 ; GFX12-CU-LABEL: system_one_as_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") seq_cst, !mmra !{!"amdgpu-synchronize-as", !"local"}
   ret void
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll
index 0e459ed0f1243..af3b7535c990c 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-fence.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @singlethread_acquire_fence() {
 ; GFX6-LABEL: singlethread_acquire_fence:
@@ -65,6 +66,10 @@ define amdgpu_kernel void @singlethread_acquire_fence() {
 ; GFX12-CU-LABEL: singlethread_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread") acquire
   ret void
@@ -122,6 +127,10 @@ define amdgpu_kernel void @singlethread_release_fence() {
 ; GFX12-CU-LABEL: singlethread_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread") release
   ret void
@@ -179,6 +188,10 @@ define amdgpu_kernel void @singlethread_acq_rel_fence() {
 ; GFX12-CU-LABEL: singlethread_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread") acq_rel
   ret void
@@ -236,6 +249,10 @@ define amdgpu_kernel void @singlethread_seq_cst_fence() {
 ; GFX12-CU-LABEL: singlethread_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread") seq_cst
   ret void
@@ -293,6 +310,10 @@ define amdgpu_kernel void @singlethread_one_as_acquire_fence() {
 ; GFX12-CU-LABEL: singlethread_one_as_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread-one-as") acquire
   ret void
@@ -350,6 +371,10 @@ define amdgpu_kernel void @singlethread_one_as_release_fence() {
 ; GFX12-CU-LABEL: singlethread_one_as_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread-one-as") release
   ret void
@@ -407,6 +432,10 @@ define amdgpu_kernel void @singlethread_one_as_acq_rel_fence() {
 ; GFX12-CU-LABEL: singlethread_one_as_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread-one-as") acq_rel
   ret void
@@ -464,6 +493,10 @@ define amdgpu_kernel void @singlethread_one_as_seq_cst_fence() {
 ; GFX12-CU-LABEL: singlethread_one_as_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: singlethread_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("singlethread-one-as") seq_cst
   ret void
@@ -521,6 +554,10 @@ define amdgpu_kernel void @wavefront_acquire_fence() {
 ; GFX12-CU-LABEL: wavefront_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront") acquire
   ret void
@@ -578,6 +615,10 @@ define amdgpu_kernel void @wavefront_release_fence() {
 ; GFX12-CU-LABEL: wavefront_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront") release
   ret void
@@ -635,6 +676,10 @@ define amdgpu_kernel void @wavefront_acq_rel_fence() {
 ; GFX12-CU-LABEL: wavefront_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront") acq_rel
   ret void
@@ -692,6 +737,10 @@ define amdgpu_kernel void @wavefront_seq_cst_fence() {
 ; GFX12-CU-LABEL: wavefront_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront") seq_cst
   ret void
@@ -749,6 +798,10 @@ define amdgpu_kernel void @wavefront_one_as_acquire_fence() {
 ; GFX12-CU-LABEL: wavefront_one_as_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront-one-as") acquire
   ret void
@@ -806,6 +859,10 @@ define amdgpu_kernel void @wavefront_one_as_release_fence() {
 ; GFX12-CU-LABEL: wavefront_one_as_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront-one-as") release
   ret void
@@ -863,6 +920,10 @@ define amdgpu_kernel void @wavefront_one_as_acq_rel_fence() {
 ; GFX12-CU-LABEL: wavefront_one_as_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront-one-as") acq_rel
   ret void
@@ -920,6 +981,10 @@ define amdgpu_kernel void @wavefront_one_as_seq_cst_fence() {
 ; GFX12-CU-LABEL: wavefront_one_as_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: wavefront_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("wavefront-one-as") seq_cst
   ret void
@@ -998,6 +1063,12 @@ define amdgpu_kernel void @workgroup_acquire_fence() {
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") acquire
   ret void
@@ -1073,6 +1144,12 @@ define amdgpu_kernel void @workgroup_release_fence() {
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") release
   ret void
@@ -1153,6 +1230,12 @@ define amdgpu_kernel void @workgroup_acq_rel_fence() {
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") acq_rel
   ret void
@@ -1233,6 +1316,12 @@ define amdgpu_kernel void @workgroup_seq_cst_fence() {
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup") seq_cst
   ret void
@@ -1303,6 +1392,12 @@ define amdgpu_kernel void @workgroup_one_as_acquire_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_acquire_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") acquire
   ret void
@@ -1370,6 +1465,12 @@ define amdgpu_kernel void @workgroup_one_as_release_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_release_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") release
   ret void
@@ -1442,6 +1543,12 @@ define amdgpu_kernel void @workgroup_one_as_acq_rel_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_acq_rel_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") acq_rel
   ret void
@@ -1514,6 +1621,12 @@ define amdgpu_kernel void @workgroup_one_as_seq_cst_fence() {
 ; GFX12-CU-LABEL: workgroup_one_as_seq_cst_fence:
 ; GFX12-CU:       ; %bb.0: ; %entry
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: workgroup_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("workgroup-one-as") seq_cst
   ret void
@@ -1606,6 +1719,13 @@ define amdgpu_kernel void @agent_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") acquire
   ret void
@@ -1688,6 +1808,13 @@ define amdgpu_kernel void @agent_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") release
   ret void
@@ -1786,6 +1913,14 @@ define amdgpu_kernel void @agent_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") acq_rel
   ret void
@@ -1884,6 +2019,14 @@ define amdgpu_kernel void @agent_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent") seq_cst
   ret void
@@ -1976,6 +2119,13 @@ define amdgpu_kernel void @agent_one_as_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") acquire
   ret void
@@ -2058,6 +2208,13 @@ define amdgpu_kernel void @agent_one_as_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") release
   ret void
@@ -2156,6 +2313,14 @@ define amdgpu_kernel void @agent_one_as_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") acq_rel
   ret void
@@ -2254,6 +2419,14 @@ define amdgpu_kernel void @agent_one_as_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: agent_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("agent-one-as") seq_cst
   ret void
@@ -2348,6 +2521,13 @@ define amdgpu_kernel void @system_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence acquire
   ret void
@@ -2434,6 +2614,13 @@ define amdgpu_kernel void @system_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence release
   ret void
@@ -2538,6 +2725,14 @@ define amdgpu_kernel void @system_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence acq_rel
   ret void
@@ -2642,6 +2837,14 @@ define amdgpu_kernel void @system_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence seq_cst
   ret void
@@ -2736,6 +2939,13 @@ define amdgpu_kernel void @system_one_as_acquire_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_acquire_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") acquire
   ret void
@@ -2822,6 +3032,13 @@ define amdgpu_kernel void @system_one_as_release_fence() {
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_release_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") release
   ret void
@@ -2926,6 +3143,14 @@ define amdgpu_kernel void @system_one_as_acq_rel_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_acq_rel_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") acq_rel
   ret void
@@ -3030,6 +3255,14 @@ define amdgpu_kernel void @system_one_as_seq_cst_fence() {
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: system_one_as_seq_cst_fence:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   fence syncscope("one-as") seq_cst
   ret void
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
index 07ad8cb0c4a3d..6c70f25a10783 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
@@ -11,6 +11,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_agent_unordered_load(
 ; GFX7-LABEL: flat_agent_unordered_load:
@@ -187,6 +188,17 @@ define amdgpu_kernel void @flat_agent_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent") unordered, align 4
@@ -369,6 +381,17 @@ define amdgpu_kernel void @flat_agent_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent") monotonic, align 4
@@ -566,6 +589,18 @@ define amdgpu_kernel void @flat_agent_acquire_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent") acquire, align 4
@@ -789,6 +824,20 @@ define amdgpu_kernel void @flat_agent_seq_cst_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent") seq_cst, align 4
@@ -939,6 +988,17 @@ define amdgpu_kernel void @flat_agent_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent") unordered, align 4
@@ -1088,6 +1148,17 @@ define amdgpu_kernel void @flat_agent_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent") monotonic, align 4
@@ -1261,6 +1332,20 @@ define amdgpu_kernel void @flat_agent_release_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent") release, align 4
@@ -1434,6 +1519,20 @@ define amdgpu_kernel void @flat_agent_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent") seq_cst, align 4
@@ -1583,6 +1682,17 @@ define amdgpu_kernel void @flat_agent_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") monotonic
@@ -1763,6 +1873,19 @@ define amdgpu_kernel void @flat_agent_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") acquire
@@ -1936,6 +2059,20 @@ define amdgpu_kernel void @flat_agent_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") release
@@ -2140,6 +2277,22 @@ define amdgpu_kernel void @flat_agent_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") acq_rel
@@ -2344,6 +2497,22 @@ define amdgpu_kernel void @flat_agent_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") seq_cst
@@ -2552,6 +2721,20 @@ define amdgpu_kernel void @flat_agent_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") acquire
@@ -2789,6 +2972,23 @@ define amdgpu_kernel void @flat_agent_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") acq_rel
@@ -3026,6 +3226,23 @@ define amdgpu_kernel void @flat_agent_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent") seq_cst
@@ -3264,6 +3481,21 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3533,6 +3765,23 @@ define amdgpu_kernel void @flat_agent_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3795,6 +4044,24 @@ define amdgpu_kernel void @flat_agent_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4088,6 +4355,26 @@ define amdgpu_kernel void @flat_agent_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4381,6 +4668,26 @@ define amdgpu_kernel void @flat_agent_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4650,6 +4957,23 @@ define amdgpu_kernel void @flat_agent_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4919,6 +5243,23 @@ define amdgpu_kernel void @flat_agent_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5212,6 +5553,26 @@ define amdgpu_kernel void @flat_agent_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5505,6 +5866,26 @@ define amdgpu_kernel void @flat_agent_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5798,6 +6179,26 @@ define amdgpu_kernel void @flat_agent_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6091,6 +6492,26 @@ define amdgpu_kernel void @flat_agent_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6384,6 +6805,26 @@ define amdgpu_kernel void @flat_agent_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6677,6 +7118,26 @@ define amdgpu_kernel void @flat_agent_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6970,6 +7431,26 @@ define amdgpu_kernel void @flat_agent_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7263,6 +7744,26 @@ define amdgpu_kernel void @flat_agent_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7545,6 +8046,23 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7844,6 +8362,24 @@ define amdgpu_kernel void @flat_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8152,6 +8688,26 @@ define amdgpu_kernel void @flat_agent_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8479,6 +9035,27 @@ define amdgpu_kernel void @flat_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8806,6 +9383,27 @@ define amdgpu_kernel void @flat_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9109,6 +9707,24 @@ define amdgpu_kernel void @flat_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9408,6 +10024,24 @@ define amdgpu_kernel void @flat_agent_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9735,6 +10369,27 @@ define amdgpu_kernel void @flat_agent_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10062,6 +10717,27 @@ define amdgpu_kernel void @flat_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10389,6 +11065,27 @@ define amdgpu_kernel void @flat_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10716,6 +11413,27 @@ define amdgpu_kernel void @flat_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -11039,6 +11757,27 @@ define amdgpu_kernel void @flat_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -11366,6 +12105,27 @@ define amdgpu_kernel void @flat_agent_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -11693,6 +12453,27 @@ define amdgpu_kernel void @flat_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -12020,6 +12801,27 @@ define amdgpu_kernel void @flat_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -12204,6 +13006,17 @@ define amdgpu_kernel void @flat_agent_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent-one-as") unordered, align 4
@@ -12386,6 +13199,17 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent-one-as") monotonic, align 4
@@ -12593,6 +13417,19 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent-one-as") acquire, align 4
@@ -12826,6 +13663,21 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("agent-one-as") seq_cst, align 4
@@ -12976,6 +13828,17 @@ define amdgpu_kernel void @flat_agent_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent-one-as") unordered, align 4
@@ -13125,6 +13988,17 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent-one-as") monotonic, align 4
@@ -13298,6 +14172,20 @@ define amdgpu_kernel void @flat_agent_one_as_release_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent-one-as") release, align 4
@@ -13471,6 +14359,20 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("agent-one-as") seq_cst, align 4
@@ -13620,6 +14522,17 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") monotonic
@@ -13796,6 +14709,19 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") acquire
@@ -13969,6 +14895,20 @@ define amdgpu_kernel void @flat_agent_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") release
@@ -14169,6 +15109,22 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") acq_rel
@@ -14369,6 +15325,22 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") seq_cst
@@ -14587,6 +15559,21 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") acquire
@@ -14834,6 +15821,24 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") acq_rel
@@ -15081,6 +16086,24 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("agent-one-as") seq_cst
@@ -15319,6 +16342,21 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15584,6 +16622,23 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15846,6 +16901,24 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16135,6 +17208,26 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16424,6 +17517,26 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16689,6 +17802,23 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16954,6 +18084,23 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17243,6 +18390,26 @@ define amdgpu_kernel void @flat_agent_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17532,6 +18699,26 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17821,6 +19008,26 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18110,6 +19317,26 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18399,6 +19626,26 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18688,6 +19935,26 @@ define amdgpu_kernel void @flat_agent_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18977,6 +20244,26 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19266,6 +20553,26 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19548,6 +20855,23 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19857,6 +21181,25 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20165,6 +21508,26 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20502,6 +21865,28 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20839,6 +22224,28 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21152,6 +22559,25 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21461,6 +22887,25 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21798,6 +23243,28 @@ define amdgpu_kernel void @flat_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -22135,6 +23602,28 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -22472,6 +23961,28 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -22809,6 +24320,28 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -23142,6 +24675,28 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -23479,6 +25034,28 @@ define amdgpu_kernel void @flat_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -23816,6 +25393,28 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -24153,6 +25752,28 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-lastuse.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-lastuse.ll
index a00af8e5b6582..3bd4d39dbb7a5 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-lastuse.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-lastuse.ll
@@ -1,6 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefix=GFX12 %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefix=GFX12 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_last_use_load_0(ptr %in, ptr %out) {
 ; GFX12-LABEL: flat_last_use_load_0:
@@ -16,6 +17,17 @@ define amdgpu_kernel void @flat_last_use_load_0(ptr %in, ptr %out) {
 ; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_last_use_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] th:TH_LOAD_LU
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %val = load i32, ptr %in, align 4, !amdgpu.last.use !{}
   store i32 %val, ptr %out
@@ -55,6 +67,21 @@ define amdgpu_kernel void @flat_last_use_load_1(ptr %in, ptr %out) {
 ; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_last_use_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s4, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v1, s[2:3] scale_offset th:TH_LOAD_LU
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %val.gep = getelementptr inbounds i32, ptr %in, i32 %tid
@@ -80,6 +107,17 @@ define amdgpu_kernel void @flat_last_use_and_volatile_load(ptr %in, ptr %out) {
 ; GFX12-NEXT:    s_wait_dscnt 0x0
 ; GFX12-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_last_use_and_volatile_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] th:TH_LOAD_BYPASS scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %val = load volatile i32, ptr %in, align 4, !amdgpu.last.use !{}
   store i32 %val, ptr %out
@@ -100,6 +138,17 @@ define amdgpu_kernel void @flat_last_use_and_nontemporal_load(ptr %in, ptr %out)
 ; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_last_use_and_nontemporal_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] th:TH_LOAD_LU
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %val = load i32, ptr %in, align 4, !amdgpu.last.use !{}, !nontemporal !0
   store i32 %val, ptr %out
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll
index 3c24c36ec547d..abc75816f9485 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll
@@ -11,6 +11,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX7-LABEL: flat_nontemporal_load_0:
@@ -187,6 +188,17 @@ define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] th:TH_LOAD_NT
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load i32, ptr %in, align 4, !nontemporal !0
@@ -555,6 +567,21 @@ define amdgpu_kernel void @flat_nontemporal_load_1(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s4, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v1, s[2:3] scale_offset th:TH_LOAD_NT
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -739,6 +766,17 @@ define amdgpu_kernel void @flat_nontemporal_store_0(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 th:TH_STORE_NT
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_store_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] th:TH_STORE_NT scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load i32, ptr %in, align 4
@@ -1095,6 +1133,20 @@ define amdgpu_kernel void @flat_nontemporal_store_1(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 th:TH_STORE_NT
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_store_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v1, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s2, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v0, v0, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scale_offset th:TH_STORE_NT scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -1293,6 +1345,17 @@ define amdgpu_kernel void @flat_nontemporal_volatile_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_volatile_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] th:TH_LOAD_NT scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load volatile i32, ptr %in, align 4, !nontemporal !0
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
index b88a10ab24a98..4cc3046e83bf1 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
@@ -11,6 +11,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_singlethread_unordered_load(
 ; GFX7-LABEL: flat_singlethread_unordered_load:
@@ -187,6 +188,17 @@ define amdgpu_kernel void @flat_singlethread_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread") unordered, align 4
@@ -369,6 +381,17 @@ define amdgpu_kernel void @flat_singlethread_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread") monotonic, align 4
@@ -551,6 +574,17 @@ define amdgpu_kernel void @flat_singlethread_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread") acquire, align 4
@@ -733,6 +767,17 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread") seq_cst, align 4
@@ -883,6 +928,17 @@ define amdgpu_kernel void @flat_singlethread_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread") unordered, align 4
@@ -1032,6 +1088,17 @@ define amdgpu_kernel void @flat_singlethread_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread") monotonic, align 4
@@ -1181,6 +1248,17 @@ define amdgpu_kernel void @flat_singlethread_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread") release, align 4
@@ -1330,6 +1408,17 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread") seq_cst, align 4
@@ -1479,6 +1568,17 @@ define amdgpu_kernel void @flat_singlethread_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") monotonic
@@ -1628,6 +1728,17 @@ define amdgpu_kernel void @flat_singlethread_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") acquire
@@ -1777,6 +1888,17 @@ define amdgpu_kernel void @flat_singlethread_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") release
@@ -1926,6 +2048,17 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") acq_rel
@@ -2075,6 +2208,17 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") seq_cst
@@ -2268,6 +2412,19 @@ define amdgpu_kernel void @flat_singlethread_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") acquire
@@ -2462,6 +2619,19 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") acq_rel
@@ -2656,6 +2826,19 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread") seq_cst
@@ -2894,6 +3077,21 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3132,6 +3330,21 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3370,6 +3583,21 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3608,6 +3836,21 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3846,6 +4089,21 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4084,6 +4342,21 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4322,6 +4595,21 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4560,6 +4848,21 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4798,6 +5101,21 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5036,6 +5354,21 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5274,6 +5607,21 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5512,6 +5860,21 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5750,6 +6113,21 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5988,6 +6366,21 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6226,6 +6619,21 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6508,6 +6916,23 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6792,6 +7217,23 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7076,6 +7518,23 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7360,6 +7819,23 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7644,6 +8120,23 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7928,6 +8421,23 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8212,6 +8722,23 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8496,6 +9023,23 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8780,6 +9324,23 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9064,6 +9625,23 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9348,6 +9926,23 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9632,6 +10227,23 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9916,6 +10528,23 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10200,6 +10829,23 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10484,6 +11130,23 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10668,6 +11331,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread-one-as") unordered, align 4
@@ -10850,6 +11524,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread-one-as") monotonic, align 4
@@ -11032,6 +11717,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread-one-as") acquire, align 4
@@ -11214,6 +11910,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("singlethread-one-as") seq_cst, align 4
@@ -11364,6 +12071,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread-one-as") unordered, align 4
@@ -11513,6 +12231,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread-one-as") monotonic, align 4
@@ -11662,6 +12391,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread-one-as") release, align 4
@@ -11811,6 +12551,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("singlethread-one-as") seq_cst, align 4
@@ -11960,6 +12711,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") monotonic
@@ -12109,6 +12871,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") acquire
@@ -12258,6 +13031,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") release
@@ -12407,6 +13191,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") acq_rel
@@ -12556,6 +13351,17 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") seq_cst
@@ -12749,6 +13555,19 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") acquire
@@ -12943,6 +13762,19 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") acq_rel
@@ -13137,6 +13969,19 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("singlethread-one-as") seq_cst
@@ -13375,6 +14220,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -13613,6 +14473,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -13851,6 +14726,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14089,6 +14979,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14327,6 +15232,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14565,6 +15485,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14803,6 +15738,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15041,6 +15991,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15279,6 +16244,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15517,6 +16497,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15755,6 +16750,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15993,6 +17003,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16231,6 +17256,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16469,6 +17509,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16707,6 +17762,21 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16989,6 +18059,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_ret_cmpx
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17273,6 +18360,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17557,6 +18661,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17841,6 +18962,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18125,6 +19263,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18409,6 +19564,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18693,6 +19865,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18977,6 +20166,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19261,6 +20467,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19545,6 +20768,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19829,6 +21069,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20113,6 +21370,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20397,6 +21671,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20681,6 +21972,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20965,6 +22273,23 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
index 919fc3e8f4e4f..b1d95bf6864e5 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
@@ -11,6 +11,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_system_unordered_load(
 ; GFX7-LABEL: flat_system_unordered_load:
@@ -187,6 +188,17 @@ define amdgpu_kernel void @flat_system_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in unordered, align 4
@@ -369,6 +381,17 @@ define amdgpu_kernel void @flat_system_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in monotonic, align 4
@@ -568,6 +591,18 @@ define amdgpu_kernel void @flat_system_acquire_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in acquire, align 4
@@ -793,6 +828,20 @@ define amdgpu_kernel void @flat_system_seq_cst_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in seq_cst, align 4
@@ -943,6 +992,17 @@ define amdgpu_kernel void @flat_system_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out unordered, align 4
@@ -1092,6 +1152,17 @@ define amdgpu_kernel void @flat_system_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out monotonic, align 4
@@ -1269,6 +1340,20 @@ define amdgpu_kernel void @flat_system_release_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out release, align 4
@@ -1446,6 +1531,20 @@ define amdgpu_kernel void @flat_system_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out seq_cst, align 4
@@ -1595,6 +1694,17 @@ define amdgpu_kernel void @flat_system_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in monotonic
@@ -1777,6 +1887,19 @@ define amdgpu_kernel void @flat_system_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in acquire
@@ -1954,6 +2077,20 @@ define amdgpu_kernel void @flat_system_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in release
@@ -2164,6 +2301,22 @@ define amdgpu_kernel void @flat_system_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in acq_rel
@@ -2374,6 +2527,22 @@ define amdgpu_kernel void @flat_system_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in seq_cst
@@ -2584,6 +2753,20 @@ define amdgpu_kernel void @flat_system_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in acquire
@@ -2827,6 +3010,23 @@ define amdgpu_kernel void @flat_system_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in acq_rel
@@ -3070,6 +3270,23 @@ define amdgpu_kernel void @flat_system_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in seq_cst
@@ -3308,6 +3525,21 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3579,6 +3811,23 @@ define amdgpu_kernel void @flat_system_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3845,6 +4094,24 @@ define amdgpu_kernel void @flat_system_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4144,6 +4411,26 @@ define amdgpu_kernel void @flat_system_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4443,6 +4730,26 @@ define amdgpu_kernel void @flat_system_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4714,6 +5021,23 @@ define amdgpu_kernel void @flat_system_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4985,6 +5309,23 @@ define amdgpu_kernel void @flat_system_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5284,6 +5625,26 @@ define amdgpu_kernel void @flat_system_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5583,6 +5944,26 @@ define amdgpu_kernel void @flat_system_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5882,6 +6263,26 @@ define amdgpu_kernel void @flat_system_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6181,6 +6582,26 @@ define amdgpu_kernel void @flat_system_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6480,6 +6901,26 @@ define amdgpu_kernel void @flat_system_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6779,6 +7220,26 @@ define amdgpu_kernel void @flat_system_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7078,6 +7539,26 @@ define amdgpu_kernel void @flat_system_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7377,6 +7858,26 @@ define amdgpu_kernel void @flat_system_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7659,6 +8160,23 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7960,6 +8478,24 @@ define amdgpu_kernel void @flat_system_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8272,6 +8808,26 @@ define amdgpu_kernel void @flat_system_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8605,6 +9161,27 @@ define amdgpu_kernel void @flat_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8938,6 +9515,27 @@ define amdgpu_kernel void @flat_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9243,6 +9841,24 @@ define amdgpu_kernel void @flat_system_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9544,6 +10160,24 @@ define amdgpu_kernel void @flat_system_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9877,6 +10511,27 @@ define amdgpu_kernel void @flat_system_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10210,6 +10865,27 @@ define amdgpu_kernel void @flat_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10543,6 +11219,27 @@ define amdgpu_kernel void @flat_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10876,6 +11573,27 @@ define amdgpu_kernel void @flat_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -11205,6 +11923,27 @@ define amdgpu_kernel void @flat_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -11538,6 +12277,27 @@ define amdgpu_kernel void @flat_system_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -11871,6 +12631,27 @@ define amdgpu_kernel void @flat_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -12204,6 +12985,27 @@ define amdgpu_kernel void @flat_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -12388,6 +13190,17 @@ define amdgpu_kernel void @flat_system_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("one-as") unordered, align 4
@@ -12570,6 +13383,17 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("one-as") monotonic, align 4
@@ -12779,6 +13603,19 @@ define amdgpu_kernel void @flat_system_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("one-as") acquire, align 4
@@ -13014,6 +13851,21 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("one-as") seq_cst, align 4
@@ -13164,6 +14016,17 @@ define amdgpu_kernel void @flat_system_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("one-as") unordered, align 4
@@ -13313,6 +14176,17 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("one-as") monotonic, align 4
@@ -13490,6 +14364,20 @@ define amdgpu_kernel void @flat_system_one_as_release_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("one-as") release, align 4
@@ -13667,6 +14555,20 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("one-as") seq_cst, align 4
@@ -13816,6 +14718,17 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") monotonic
@@ -13994,6 +14907,19 @@ define amdgpu_kernel void @flat_system_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") acquire
@@ -14171,6 +15097,20 @@ define amdgpu_kernel void @flat_system_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") release
@@ -14377,6 +15317,22 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") acq_rel
@@ -14583,6 +15539,22 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") seq_cst
@@ -14803,6 +15775,21 @@ define amdgpu_kernel void @flat_system_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") acquire
@@ -15056,6 +16043,24 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") acq_rel
@@ -15309,6 +16314,24 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("one-as") seq_cst
@@ -15547,6 +16570,21 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15814,6 +16852,23 @@ define amdgpu_kernel void @flat_system_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16080,6 +17135,24 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16375,6 +17448,26 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16670,6 +17763,26 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16937,6 +18050,23 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17204,6 +18334,23 @@ define amdgpu_kernel void @flat_system_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17499,6 +18646,26 @@ define amdgpu_kernel void @flat_system_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17794,6 +18961,26 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18089,6 +19276,26 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18384,6 +19591,26 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18679,6 +19906,26 @@ define amdgpu_kernel void @flat_system_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18974,6 +20221,26 @@ define amdgpu_kernel void @flat_system_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19269,6 +20536,26 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19564,6 +20851,26 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19846,6 +21153,23 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20157,6 +21481,25 @@ define amdgpu_kernel void @flat_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20469,6 +21812,26 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20812,6 +22175,28 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21155,6 +22540,28 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21470,6 +22877,25 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21781,6 +23207,25 @@ define amdgpu_kernel void @flat_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -22124,6 +23569,28 @@ define amdgpu_kernel void @flat_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -22467,6 +23934,28 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -22810,6 +24299,28 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -23153,6 +24664,28 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -23492,6 +25025,28 @@ define amdgpu_kernel void @flat_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -23835,6 +25390,28 @@ define amdgpu_kernel void @flat_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -24178,6 +25755,28 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -24521,6 +26120,28 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_system_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll
index a88e0e217fdb4..e2739debea36a 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll
@@ -7,6 +7,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX7-LABEL: flat_nontemporal_load_0:
@@ -143,6 +144,17 @@ define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load volatile i32, ptr %in, align 4
@@ -415,6 +427,21 @@ define amdgpu_kernel void @flat_nontemporal_load_1(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s4, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v1, s[2:3] scale_offset scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -563,6 +590,18 @@ define amdgpu_kernel void @flat_nontemporal_store_0(
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_store_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load i32, ptr %in, align 4
@@ -831,6 +870,21 @@ define amdgpu_kernel void @flat_nontemporal_store_1(
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_nontemporal_store_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v1, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s2, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v0, v0, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scale_offset scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -971,6 +1025,17 @@ define amdgpu_kernel void @flat_volatile_workgroup_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_volatile_workgroup_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic volatile i32, ptr %in syncscope("workgroup") acquire, align 4
@@ -1090,6 +1155,19 @@ define amdgpu_kernel void @flat_volatile_workgroup_release_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_volatile_workgroup_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
    i32 %in, ptr %out) {
 entry:
   store atomic volatile i32 %in, ptr %out syncscope("workgroup") release, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
index 7c637a20ab47b..cd12f48468777 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
@@ -11,6 +11,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_wavefront_unordered_load(
 ; GFX7-LABEL: flat_wavefront_unordered_load:
@@ -187,6 +188,17 @@ define amdgpu_kernel void @flat_wavefront_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront") unordered, align 4
@@ -369,6 +381,17 @@ define amdgpu_kernel void @flat_wavefront_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront") monotonic, align 4
@@ -551,6 +574,17 @@ define amdgpu_kernel void @flat_wavefront_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront") acquire, align 4
@@ -733,6 +767,17 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront") seq_cst, align 4
@@ -883,6 +928,17 @@ define amdgpu_kernel void @flat_wavefront_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront") unordered, align 4
@@ -1032,6 +1088,17 @@ define amdgpu_kernel void @flat_wavefront_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront") monotonic, align 4
@@ -1181,6 +1248,17 @@ define amdgpu_kernel void @flat_wavefront_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront") release, align 4
@@ -1330,6 +1408,17 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront") seq_cst, align 4
@@ -1479,6 +1568,17 @@ define amdgpu_kernel void @flat_wavefront_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") monotonic
@@ -1628,6 +1728,17 @@ define amdgpu_kernel void @flat_wavefront_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") acquire
@@ -1777,6 +1888,17 @@ define amdgpu_kernel void @flat_wavefront_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") release
@@ -1926,6 +2048,17 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") acq_rel
@@ -2075,6 +2208,17 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") seq_cst
@@ -2268,6 +2412,19 @@ define amdgpu_kernel void @flat_wavefront_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") acquire
@@ -2462,6 +2619,19 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") acq_rel
@@ -2656,6 +2826,19 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront") seq_cst
@@ -2894,6 +3077,21 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3132,6 +3330,21 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3370,6 +3583,21 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3608,6 +3836,21 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3846,6 +4089,21 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4084,6 +4342,21 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4322,6 +4595,21 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4560,6 +4848,21 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4798,6 +5101,21 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5036,6 +5354,21 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5274,6 +5607,21 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5512,6 +5860,21 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5750,6 +6113,21 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5988,6 +6366,21 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6226,6 +6619,21 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6508,6 +6916,23 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6792,6 +7217,23 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7076,6 +7518,23 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7360,6 +7819,23 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7644,6 +8120,23 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7928,6 +8421,23 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8212,6 +8722,23 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8496,6 +9023,23 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8780,6 +9324,23 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9064,6 +9625,23 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9348,6 +9926,23 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9632,6 +10227,23 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9916,6 +10528,23 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10200,6 +10829,23 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10484,6 +11130,23 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10668,6 +11331,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront-one-as") unordered, align 4
@@ -10850,6 +11524,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront-one-as") monotonic, align 4
@@ -11032,6 +11717,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront-one-as") acquire, align 4
@@ -11214,6 +11910,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("wavefront-one-as") seq_cst, align 4
@@ -11364,6 +12071,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront-one-as") unordered, align 4
@@ -11513,6 +12231,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront-one-as") monotonic, align 4
@@ -11662,6 +12391,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront-one-as") release, align 4
@@ -11811,6 +12551,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("wavefront-one-as") seq_cst, align 4
@@ -11960,6 +12711,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") monotonic
@@ -12109,6 +12871,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") acquire
@@ -12258,6 +13031,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") release
@@ -12407,6 +13191,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") acq_rel
@@ -12556,6 +13351,17 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") seq_cst
@@ -12749,6 +13555,19 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") acquire
@@ -12943,6 +13762,19 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") acq_rel
@@ -13137,6 +13969,19 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("wavefront-one-as") seq_cst
@@ -13375,6 +14220,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -13613,6 +14473,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -13851,6 +14726,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14089,6 +14979,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14327,6 +15232,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14565,6 +15485,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14803,6 +15738,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15041,6 +15991,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15279,6 +16244,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15517,6 +16497,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15755,6 +16750,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15993,6 +17003,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16231,6 +17256,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16469,6 +17509,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16707,6 +17762,21 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16989,6 +18059,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17273,6 +18360,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17557,6 +18661,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17841,6 +18962,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18125,6 +19263,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18409,6 +19564,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18693,6 +19865,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18977,6 +20166,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19261,6 +20467,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19545,6 +20768,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19829,6 +21069,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20113,6 +21370,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20397,6 +21671,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_relc_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_acq_relc_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
    ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20681,6 +21972,23 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
index 0fd4aa4a7a93f..4d22b6abadad7 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
@@ -11,6 +11,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @flat_workgroup_unordered_load(
 ; GFX7-LABEL: flat_workgroup_unordered_load:
@@ -187,6 +188,17 @@ define amdgpu_kernel void @flat_workgroup_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup") unordered, align 4
@@ -369,6 +381,17 @@ define amdgpu_kernel void @flat_workgroup_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup") monotonic, align 4
@@ -563,6 +586,17 @@ define amdgpu_kernel void @flat_workgroup_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup") acquire, align 4
@@ -776,6 +810,19 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup") seq_cst, align 4
@@ -926,6 +973,17 @@ define amdgpu_kernel void @flat_workgroup_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup") unordered, align 4
@@ -1075,6 +1133,17 @@ define amdgpu_kernel void @flat_workgroup_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup") monotonic, align 4
@@ -1241,6 +1310,19 @@ define amdgpu_kernel void @flat_workgroup_release_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup") release, align 4
@@ -1407,6 +1489,19 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup") seq_cst, align 4
@@ -1556,6 +1651,17 @@ define amdgpu_kernel void @flat_workgroup_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") monotonic
@@ -1724,6 +1830,18 @@ define amdgpu_kernel void @flat_workgroup_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") acquire
@@ -1890,6 +2008,19 @@ define amdgpu_kernel void @flat_workgroup_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") release
@@ -2075,6 +2206,20 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") acq_rel
@@ -2260,6 +2405,20 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") seq_cst
@@ -2465,6 +2624,19 @@ define amdgpu_kernel void @flat_workgroup_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") acquire
@@ -2690,6 +2862,21 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") acq_rel
@@ -2915,6 +3102,21 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup") seq_cst
@@ -3153,6 +3355,21 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3410,6 +3627,22 @@ define amdgpu_kernel void @flat_workgroup_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3665,6 +3898,23 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -3939,6 +4189,24 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4213,6 +4481,24 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4470,6 +4756,22 @@ define amdgpu_kernel void @flat_workgroup_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -4727,6 +5029,22 @@ define amdgpu_kernel void @flat_workgroup_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5001,6 +5319,24 @@ define amdgpu_kernel void @flat_workgroup_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5275,6 +5611,24 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5549,6 +5903,24 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -5823,6 +6195,24 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6105,6 +6495,23 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6401,6 +6808,23 @@ define amdgpu_kernel void @flat_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -6702,6 +7126,25 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7017,6 +7460,25 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7332,6 +7794,25 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7630,6 +8111,23 @@ define amdgpu_kernel void @flat_workgroup_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -7926,6 +8424,23 @@ define amdgpu_kernel void @flat_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8241,6 +8756,25 @@ define amdgpu_kernel void @flat_workgroup_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8556,6 +9090,25 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -8871,6 +9424,25 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9186,6 +9758,25 @@ define amdgpu_kernel void @flat_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9499,6 +10090,25 @@ define amdgpu_kernel void @flat_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -9814,6 +10424,25 @@ define amdgpu_kernel void @flat_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10129,6 +10758,25 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10444,6 +11092,25 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -10628,6 +11295,17 @@ define amdgpu_kernel void @flat_workgroup_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup-one-as") unordered, align 4
@@ -10810,6 +11488,17 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup-one-as") monotonic, align 4
@@ -11000,6 +11689,17 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup-one-as") acquire, align 4
@@ -11202,6 +11902,19 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    flat_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %in, ptr %out) {
 entry:
   %val = load atomic i32, ptr %in syncscope("workgroup-one-as") seq_cst, align 4
@@ -11352,6 +12065,17 @@ define amdgpu_kernel void @flat_workgroup_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup-one-as") unordered, align 4
@@ -11501,6 +12225,17 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup-one-as") monotonic, align 4
@@ -11660,6 +12395,19 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup-one-as") release, align 4
@@ -11819,6 +12567,19 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr %out) {
 entry:
   store atomic i32 %in, ptr %out syncscope("workgroup-one-as") seq_cst, align 4
@@ -11968,6 +12729,17 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") monotonic
@@ -12127,6 +12899,18 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") acquire
@@ -12286,6 +13070,19 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") release
@@ -12455,6 +13252,20 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") acq_rel
@@ -12624,6 +13435,20 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") seq_cst
@@ -12825,6 +13650,19 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") acquire
@@ -13039,6 +13877,21 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") acq_rel
@@ -13253,6 +14106,21 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr %out, i32 %in syncscope("workgroup-one-as") seq_cst
@@ -13491,6 +14359,21 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -13739,6 +14622,22 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -13987,6 +14886,23 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14245,6 +15161,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14503,6 +15437,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14751,6 +15703,22 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -14999,6 +15967,22 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15257,6 +16241,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15515,6 +16517,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -15773,6 +16793,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16031,6 +17069,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16289,6 +17345,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16547,6 +17621,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -16805,6 +17897,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17063,6 +18173,24 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX12-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17345,6 +18473,23 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonicmonotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonicmonotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17637,6 +18782,23 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -17931,6 +19093,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18235,6 +19416,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18539,6 +19739,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -18833,6 +20052,23 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19125,6 +20361,23 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19429,6 +20682,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -19733,6 +21005,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20037,6 +21328,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20341,6 +21651,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20643,6 +21972,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -20947,6 +22295,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21251,6 +22618,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
@@ -21555,6 +22941,25 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    flat_store_b32 v[0:1], v2
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: flat_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    flat_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    flat_store_b32 v0, v1, s[0:1] scope:SCOPE_SE
+; GFX1250-CU-NEXT:    s_endpgm
     ptr %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll
index 74a72e04fa4ae..7b8c55aeea80e 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_agent_unordered_load(
 ; GFX6-LABEL: global_agent_unordered_load:
@@ -190,6 +191,17 @@ define amdgpu_kernel void @global_agent_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent") unordered, align 4
@@ -374,6 +386,17 @@ define amdgpu_kernel void @global_agent_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent") monotonic, align 4
@@ -574,6 +597,18 @@ define amdgpu_kernel void @global_agent_acquire_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent") acquire, align 4
@@ -793,6 +828,20 @@ define amdgpu_kernel void @global_agent_seq_cst_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent") seq_cst, align 4
@@ -950,6 +999,17 @@ define amdgpu_kernel void @global_agent_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent") unordered, align 4
@@ -1106,6 +1166,17 @@ define amdgpu_kernel void @global_agent_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent") monotonic, align 4
@@ -1287,6 +1358,20 @@ define amdgpu_kernel void @global_agent_release_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent") release, align 4
@@ -1468,6 +1553,20 @@ define amdgpu_kernel void @global_agent_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent") seq_cst, align 4
@@ -1622,6 +1721,17 @@ define amdgpu_kernel void @global_agent_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") monotonic
@@ -1805,6 +1915,19 @@ define amdgpu_kernel void @global_agent_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") acquire
@@ -1984,6 +2107,20 @@ define amdgpu_kernel void @global_agent_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") release
@@ -2192,6 +2329,22 @@ define amdgpu_kernel void @global_agent_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") acq_rel
@@ -2400,6 +2553,22 @@ define amdgpu_kernel void @global_agent_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") seq_cst
@@ -2598,6 +2767,20 @@ define amdgpu_kernel void @global_agent_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") acquire
@@ -2826,6 +3009,23 @@ define amdgpu_kernel void @global_agent_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") acq_rel
@@ -3054,6 +3254,23 @@ define amdgpu_kernel void @global_agent_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent") seq_cst
@@ -3273,6 +3490,21 @@ define amdgpu_kernel void @global_agent_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3521,6 +3753,23 @@ define amdgpu_kernel void @global_agent_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3765,6 +4014,24 @@ define amdgpu_kernel void @global_agent_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4038,6 +4305,26 @@ define amdgpu_kernel void @global_agent_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4311,6 +4598,26 @@ define amdgpu_kernel void @global_agent_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4559,6 +4866,23 @@ define amdgpu_kernel void @global_agent_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4807,6 +5131,23 @@ define amdgpu_kernel void @global_agent_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5080,6 +5421,26 @@ define amdgpu_kernel void @global_agent_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5353,6 +5714,26 @@ define amdgpu_kernel void @global_agent_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5626,6 +6007,26 @@ define amdgpu_kernel void @global_agent_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5899,6 +6300,26 @@ define amdgpu_kernel void @global_agent_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6172,6 +6593,26 @@ define amdgpu_kernel void @global_agent_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6445,6 +6886,26 @@ define amdgpu_kernel void @global_agent_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6718,6 +7179,26 @@ define amdgpu_kernel void @global_agent_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6991,6 +7472,26 @@ define amdgpu_kernel void @global_agent_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7240,6 +7741,23 @@ define amdgpu_kernel void @global_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7507,6 +8025,24 @@ define amdgpu_kernel void @global_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7783,6 +8319,26 @@ define amdgpu_kernel void @global_agent_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8079,6 +8635,27 @@ define amdgpu_kernel void @global_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8375,6 +8952,27 @@ define amdgpu_kernel void @global_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8646,6 +9244,24 @@ define amdgpu_kernel void @global_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8913,6 +9529,24 @@ define amdgpu_kernel void @global_agent_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9209,6 +9843,27 @@ define amdgpu_kernel void @global_agent_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9505,6 +10160,27 @@ define amdgpu_kernel void @global_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9801,6 +10477,27 @@ define amdgpu_kernel void @global_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10097,6 +10794,27 @@ define amdgpu_kernel void @global_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10389,6 +11107,27 @@ define amdgpu_kernel void @global_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10685,6 +11424,27 @@ define amdgpu_kernel void @global_agent_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10981,6 +11741,27 @@ define amdgpu_kernel void @global_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -11277,6 +12058,27 @@ define amdgpu_kernel void @global_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -11463,6 +12265,17 @@ define amdgpu_kernel void @global_agent_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent-one-as") unordered, align 4
@@ -11647,6 +12460,17 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent-one-as") monotonic, align 4
@@ -11847,6 +12671,18 @@ define amdgpu_kernel void @global_agent_one_as_acquire_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent-one-as") acquire, align 4
@@ -12066,6 +12902,20 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("agent-one-as") seq_cst, align 4
@@ -12223,6 +13073,17 @@ define amdgpu_kernel void @global_agent_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent-one-as") unordered, align 4
@@ -12379,6 +13240,17 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent-one-as") monotonic, align 4
@@ -12560,6 +13432,20 @@ define amdgpu_kernel void @global_agent_one_as_release_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent-one-as") release, align 4
@@ -12741,6 +13627,20 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("agent-one-as") seq_cst, align 4
@@ -12895,6 +13795,17 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") monotonic
@@ -13078,6 +13989,19 @@ define amdgpu_kernel void @global_agent_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") acquire
@@ -13257,6 +14181,20 @@ define amdgpu_kernel void @global_agent_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") release
@@ -13465,6 +14403,22 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") acq_rel
@@ -13673,6 +14627,22 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") seq_cst
@@ -13871,6 +14841,20 @@ define amdgpu_kernel void @global_agent_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") acquire
@@ -14099,6 +15083,23 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") acq_rel
@@ -14327,6 +15328,23 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("agent-one-as") seq_cst
@@ -14546,6 +15564,21 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14794,6 +15827,23 @@ define amdgpu_kernel void @global_agent_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15038,6 +16088,24 @@ define amdgpu_kernel void @global_agent_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15311,6 +16379,26 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15584,6 +16672,26 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15832,6 +16940,23 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16080,6 +17205,23 @@ define amdgpu_kernel void @global_agent_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16353,6 +17495,26 @@ define amdgpu_kernel void @global_agent_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16626,6 +17788,26 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16899,6 +18081,26 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17172,6 +18374,26 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17445,6 +18667,26 @@ define amdgpu_kernel void @global_agent_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17718,6 +18960,26 @@ define amdgpu_kernel void @global_agent_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17991,6 +19253,26 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18264,6 +19546,26 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18513,6 +19815,23 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18780,6 +20099,24 @@ define amdgpu_kernel void @global_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19076,6 +20413,27 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19372,6 +20730,27 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19643,6 +21022,24 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19910,6 +21307,24 @@ define amdgpu_kernel void @global_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20206,6 +21621,27 @@ define amdgpu_kernel void @global_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20502,6 +21938,27 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20798,6 +22255,27 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -21094,6 +22572,27 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -21386,6 +22885,27 @@ define amdgpu_kernel void @global_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -21682,6 +23202,27 @@ define amdgpu_kernel void @global_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -21978,6 +23519,27 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -22274,6 +23836,27 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_DEV
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-lastuse.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-lastuse.ll
index 5f952b98041f3..48f9a35b1a8b7 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-lastuse.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-lastuse.ll
@@ -1,6 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefix=GFX12 %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefix=GFX12 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_last_use_load_0(ptr addrspace(1) %in, ptr addrspace(1) %out) {
 ; GFX12-LABEL: global_last_use_load_0:
@@ -14,6 +15,18 @@ define amdgpu_kernel void @global_last_use_load_0(ptr addrspace(1) %in, ptr addr
 ; GFX12-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_last_use_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[2:3], 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %val = load i32, ptr addrspace(1) %in, align 4, !amdgpu.last.use !{}
   store i32 %val, ptr addrspace(1) %out
@@ -37,6 +50,21 @@ define amdgpu_kernel void @global_last_use_load_1(ptr addrspace(1) %in, ptr addr
 ; GFX12-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_last_use_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s4, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v1, s[2:3] scale_offset th:TH_LOAD_LU
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %val.gep = getelementptr inbounds i32, ptr addrspace(1) %in, i32 %tid
@@ -58,6 +86,17 @@ define amdgpu_kernel void @global_last_use_and_volatile_load(ptr addrspace(1) %i
 ; GFX12-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_last_use_and_volatile_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] th:TH_LOAD_BYPASS scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %val = load volatile i32, ptr addrspace(1) %in, align 4, !amdgpu.last.use !{}
   store i32 %val, ptr addrspace(1) %out
@@ -81,6 +120,21 @@ define amdgpu_kernel void @global_last_use_and_nontemporal_load(ptr addrspace(1)
 ; GFX12-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_last_use_and_nontemporal_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s4, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v1, s[2:3] scale_offset th:TH_LOAD_LU
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %val.gep = getelementptr inbounds i32, ptr addrspace(1) %in, i32 %tid
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll
index 16e55058e4fc8..c9125a5b1bb2a 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_nontemporal_load_0(
 ; GFX6-LABEL: global_nontemporal_load_0:
@@ -189,6 +190,18 @@ define amdgpu_kernel void @global_nontemporal_load_0(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_nontemporal_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[2:3], 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load i32, ptr addrspace(1) %in, align 4, !nontemporal !0
@@ -448,6 +461,21 @@ define amdgpu_kernel void @global_nontemporal_load_1(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_nontemporal_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s4, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v1, s[2:3] scale_offset th:TH_LOAD_NT
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -633,6 +661,18 @@ define amdgpu_kernel void @global_nontemporal_store_0(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] th:TH_STORE_NT
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_nontemporal_store_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[2:3], 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] th:TH_STORE_NT
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load i32, ptr addrspace(1) %in, align 4
@@ -866,6 +906,20 @@ define amdgpu_kernel void @global_nontemporal_store_1(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] th:TH_STORE_NT
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_nontemporal_store_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[2:3], 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s3, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v0, v0, s3
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scale_offset th:TH_STORE_NT
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -1056,6 +1110,17 @@ define amdgpu_kernel void @global_nontemporal_volatile_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_nontemporal_volatile_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] th:TH_LOAD_NT scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load volatile i32, ptr addrspace(1) %in, align 4, !nontemporal !0
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll
index 8042d38716107..859069c91aef3 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_singlethread_unordered_load(
 ; GFX6-LABEL: global_singlethread_unordered_load:
@@ -190,6 +191,17 @@ define amdgpu_kernel void @global_singlethread_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread") unordered, align 4
@@ -374,6 +386,17 @@ define amdgpu_kernel void @global_singlethread_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread") monotonic, align 4
@@ -558,6 +581,17 @@ define amdgpu_kernel void @global_singlethread_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread") acquire, align 4
@@ -742,6 +776,17 @@ define amdgpu_kernel void @global_singlethread_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread") seq_cst, align 4
@@ -899,6 +944,17 @@ define amdgpu_kernel void @global_singlethread_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread") unordered, align 4
@@ -1055,6 +1111,17 @@ define amdgpu_kernel void @global_singlethread_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread") monotonic, align 4
@@ -1211,6 +1278,17 @@ define amdgpu_kernel void @global_singlethread_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread") release, align 4
@@ -1367,6 +1445,17 @@ define amdgpu_kernel void @global_singlethread_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread") seq_cst, align 4
@@ -1521,6 +1610,17 @@ define amdgpu_kernel void @global_singlethread_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") monotonic
@@ -1675,6 +1775,17 @@ define amdgpu_kernel void @global_singlethread_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") acquire
@@ -1829,6 +1940,17 @@ define amdgpu_kernel void @global_singlethread_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") release
@@ -1983,6 +2105,17 @@ define amdgpu_kernel void @global_singlethread_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") acq_rel
@@ -2137,6 +2270,17 @@ define amdgpu_kernel void @global_singlethread_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") seq_cst
@@ -2319,6 +2463,19 @@ define amdgpu_kernel void @global_singlethread_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") acquire
@@ -2502,6 +2659,19 @@ define amdgpu_kernel void @global_singlethread_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") acq_rel
@@ -2685,6 +2855,19 @@ define amdgpu_kernel void @global_singlethread_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread") seq_cst
@@ -2904,6 +3087,21 @@ define amdgpu_kernel void @global_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3123,6 +3321,21 @@ define amdgpu_kernel void @global_singlethread_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3342,6 +3555,21 @@ define amdgpu_kernel void @global_singlethread_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3561,6 +3789,21 @@ define amdgpu_kernel void @global_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3780,6 +4023,21 @@ define amdgpu_kernel void @global_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3999,6 +4257,21 @@ define amdgpu_kernel void @global_singlethread_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4218,6 +4491,21 @@ define amdgpu_kernel void @global_singlethread_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4437,6 +4725,21 @@ define amdgpu_kernel void @global_singlethread_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4656,6 +4959,21 @@ define amdgpu_kernel void @global_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4875,6 +5193,21 @@ define amdgpu_kernel void @global_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5094,6 +5427,21 @@ define amdgpu_kernel void @global_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5313,6 +5661,21 @@ define amdgpu_kernel void @global_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5532,6 +5895,21 @@ define amdgpu_kernel void @global_singlethread_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5751,6 +6129,21 @@ define amdgpu_kernel void @global_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5970,6 +6363,21 @@ define amdgpu_kernel void @global_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6219,6 +6627,23 @@ define amdgpu_kernel void @global_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6470,6 +6895,23 @@ define amdgpu_kernel void @global_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6721,6 +7163,23 @@ define amdgpu_kernel void @global_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6972,6 +7431,23 @@ define amdgpu_kernel void @global_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7223,6 +7699,23 @@ define amdgpu_kernel void @global_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7474,6 +7967,23 @@ define amdgpu_kernel void @global_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7725,6 +8235,23 @@ define amdgpu_kernel void @global_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7976,6 +8503,23 @@ define amdgpu_kernel void @global_singlethread_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8227,6 +8771,23 @@ define amdgpu_kernel void @global_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8478,6 +9039,23 @@ define amdgpu_kernel void @global_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8729,6 +9307,23 @@ define amdgpu_kernel void @global_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8980,6 +9575,23 @@ define amdgpu_kernel void @global_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9231,6 +9843,23 @@ define amdgpu_kernel void @global_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9482,6 +10111,23 @@ define amdgpu_kernel void @global_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9733,6 +10379,23 @@ define amdgpu_kernel void @global_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9919,6 +10582,17 @@ define amdgpu_kernel void @global_singlethread_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread-one-as") unordered, align 4
@@ -10103,6 +10777,17 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread-one-as") monotonic, align 4
@@ -10287,6 +10972,17 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread-one-as") acquire, align 4
@@ -10471,6 +11167,17 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("singlethread-one-as") seq_cst, align 4
@@ -10628,6 +11335,17 @@ define amdgpu_kernel void @global_singlethread_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread-one-as") unordered, align 4
@@ -10784,6 +11502,17 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread-one-as") monotonic, align 4
@@ -10940,6 +11669,17 @@ define amdgpu_kernel void @global_singlethread_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread-one-as") release, align 4
@@ -11096,6 +11836,17 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("singlethread-one-as") seq_cst, align 4
@@ -11250,6 +12001,17 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") monotonic
@@ -11404,6 +12166,17 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") acquire
@@ -11558,6 +12331,17 @@ define amdgpu_kernel void @global_singlethread_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") release
@@ -11712,6 +12496,17 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") acq_rel
@@ -11866,6 +12661,17 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") seq_cst
@@ -12048,6 +12854,19 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") acquire
@@ -12231,6 +13050,19 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") acq_rel
@@ -12414,6 +13246,19 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("singlethread-one-as") seq_cst
@@ -12633,6 +13478,21 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_monotonic_cmpxch
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -12852,6 +13712,21 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13071,6 +13946,21 @@ define amdgpu_kernel void @global_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13290,6 +14180,21 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13509,6 +14414,21 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13728,6 +14648,21 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13947,6 +14882,21 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14166,6 +15116,21 @@ define amdgpu_kernel void @global_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14385,6 +15350,21 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14604,6 +15584,21 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14823,6 +15818,21 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15042,6 +16052,21 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15261,6 +16286,21 @@ define amdgpu_kernel void @global_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15480,6 +16520,21 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15699,6 +16754,21 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15948,6 +17018,23 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_monotonic_ret_cm
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16199,6 +17286,23 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_monotonic_ret_cmpx
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16450,6 +17554,23 @@ define amdgpu_kernel void @global_singlethread_one_as_release_monotonic_ret_cmpx
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16701,6 +17822,23 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_monotonic_ret_cmpx
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16952,6 +18090,23 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_monotonic_ret_cmpx
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17203,6 +18358,23 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_acquire_ret_cmpx
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17454,6 +18626,23 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_acquire_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17705,6 +18894,23 @@ define amdgpu_kernel void @global_singlethread_one_as_release_acquire_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17956,6 +19162,23 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_acquire_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18207,6 +19430,23 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_acquire_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18458,6 +19698,23 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_seq_cst_ret_cmpx
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18709,6 +19966,23 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_seq_cst_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18960,6 +20234,23 @@ define amdgpu_kernel void @global_singlethread_one_as_release_seq_cst_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19211,6 +20502,23 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_seq_cst_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19462,6 +20770,23 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_seq_cst_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll
index be148464c156e..28d179b159974 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_system_unordered_load(
 ; GFX6-LABEL: global_system_unordered_load:
@@ -190,6 +191,17 @@ define amdgpu_kernel void @global_system_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in unordered, align 4
@@ -374,6 +386,17 @@ define amdgpu_kernel void @global_system_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in monotonic, align 4
@@ -576,6 +599,18 @@ define amdgpu_kernel void @global_system_acquire_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in acquire, align 4
@@ -797,6 +832,20 @@ define amdgpu_kernel void @global_system_seq_cst_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in seq_cst, align 4
@@ -954,6 +1003,17 @@ define amdgpu_kernel void @global_system_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out unordered, align 4
@@ -1110,6 +1170,17 @@ define amdgpu_kernel void @global_system_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out monotonic, align 4
@@ -1295,6 +1366,20 @@ define amdgpu_kernel void @global_system_release_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out release, align 4
@@ -1480,6 +1565,20 @@ define amdgpu_kernel void @global_system_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out seq_cst, align 4
@@ -1634,6 +1733,17 @@ define amdgpu_kernel void @global_system_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in monotonic
@@ -1819,6 +1929,19 @@ define amdgpu_kernel void @global_system_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in acquire
@@ -2002,6 +2125,20 @@ define amdgpu_kernel void @global_system_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in release
@@ -2216,6 +2353,22 @@ define amdgpu_kernel void @global_system_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in acq_rel
@@ -2430,6 +2583,22 @@ define amdgpu_kernel void @global_system_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in seq_cst
@@ -2630,6 +2799,20 @@ define amdgpu_kernel void @global_system_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in acquire
@@ -2864,6 +3047,23 @@ define amdgpu_kernel void @global_system_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in acq_rel
@@ -3098,6 +3298,23 @@ define amdgpu_kernel void @global_system_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in seq_cst
@@ -3317,6 +3534,21 @@ define amdgpu_kernel void @global_system_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3567,6 +3799,23 @@ define amdgpu_kernel void @global_system_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3815,6 +4064,24 @@ define amdgpu_kernel void @global_system_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4094,6 +4361,26 @@ define amdgpu_kernel void @global_system_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4373,6 +4660,26 @@ define amdgpu_kernel void @global_system_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4623,6 +4930,23 @@ define amdgpu_kernel void @global_system_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4873,6 +5197,23 @@ define amdgpu_kernel void @global_system_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5152,6 +5493,26 @@ define amdgpu_kernel void @global_system_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5431,6 +5792,26 @@ define amdgpu_kernel void @global_system_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5710,6 +6091,26 @@ define amdgpu_kernel void @global_system_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5989,6 +6390,26 @@ define amdgpu_kernel void @global_system_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6238,6 +6659,23 @@ define amdgpu_kernel void @global_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6507,6 +6945,24 @@ define amdgpu_kernel void @global_system_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6809,6 +7265,27 @@ define amdgpu_kernel void @global_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7111,6 +7588,27 @@ define amdgpu_kernel void @global_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7384,6 +7882,24 @@ define amdgpu_kernel void @global_system_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7653,6 +8169,24 @@ define amdgpu_kernel void @global_system_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7955,6 +8489,27 @@ define amdgpu_kernel void @global_system_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8257,6 +8812,27 @@ define amdgpu_kernel void @global_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8559,6 +9135,27 @@ define amdgpu_kernel void @global_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8861,6 +9458,27 @@ define amdgpu_kernel void @global_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9159,6 +9777,27 @@ define amdgpu_kernel void @global_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9461,6 +10100,27 @@ define amdgpu_kernel void @global_system_relese_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_relese_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9763,6 +10423,27 @@ define amdgpu_kernel void @global_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10065,6 +10746,27 @@ define amdgpu_kernel void @global_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10251,6 +10953,17 @@ define amdgpu_kernel void @global_system_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("one-as") unordered, align 4
@@ -10435,6 +11148,17 @@ define amdgpu_kernel void @global_system_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("one-as") monotonic, align 4
@@ -10637,6 +11361,18 @@ define amdgpu_kernel void @global_system_one_as_acquire_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("one-as") acquire, align 4
@@ -10858,6 +11594,20 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("one-as") seq_cst, align 4
@@ -11015,6 +11765,17 @@ define amdgpu_kernel void @global_system_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("one-as") unordered, align 4
@@ -11171,6 +11932,17 @@ define amdgpu_kernel void @global_system_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("one-as") monotonic, align 4
@@ -11356,6 +12128,20 @@ define amdgpu_kernel void @global_system_one_as_release_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("one-as") release, align 4
@@ -11541,6 +12327,20 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("one-as") seq_cst, align 4
@@ -11695,6 +12495,17 @@ define amdgpu_kernel void @global_system_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") monotonic
@@ -11880,6 +12691,19 @@ define amdgpu_kernel void @global_system_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") acquire
@@ -12063,6 +12887,20 @@ define amdgpu_kernel void @global_system_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") release
@@ -12277,6 +13115,22 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") acq_rel
@@ -12491,6 +13345,22 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") seq_cst
@@ -12691,6 +13561,20 @@ define amdgpu_kernel void @global_system_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") acquire
@@ -12925,6 +13809,23 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") acq_rel
@@ -13159,6 +14060,23 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("one-as") seq_cst
@@ -13378,6 +14296,21 @@ define amdgpu_kernel void @global_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13628,6 +14561,23 @@ define amdgpu_kernel void @global_system_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13876,6 +14826,24 @@ define amdgpu_kernel void @global_system_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16 scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14155,6 +15123,26 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14434,6 +15422,26 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14684,6 +15692,23 @@ define amdgpu_kernel void @global_system_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14934,6 +15959,23 @@ define amdgpu_kernel void @global_system_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15213,6 +16255,26 @@ define amdgpu_kernel void @global_system_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15492,6 +16554,26 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15771,6 +16853,26 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16050,6 +17152,26 @@ define amdgpu_kernel void @global_system_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16329,6 +17451,26 @@ define amdgpu_kernel void @global_system_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16608,6 +17750,26 @@ define amdgpu_kernel void @global_system_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16887,6 +18049,26 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17166,6 +18348,26 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16 scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17415,6 +18617,23 @@ define amdgpu_kernel void @global_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17684,6 +18903,24 @@ define amdgpu_kernel void @global_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17964,6 +19201,26 @@ define amdgpu_kernel void @global_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18266,6 +19523,27 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18568,6 +19846,27 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18841,6 +20140,24 @@ define amdgpu_kernel void @global_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19110,6 +20427,24 @@ define amdgpu_kernel void @global_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19412,6 +20747,27 @@ define amdgpu_kernel void @global_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19714,6 +21070,27 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20016,6 +21393,27 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20318,6 +21716,27 @@ define amdgpu_kernel void @global_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20616,6 +22035,27 @@ define amdgpu_kernel void @global_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20918,6 +22358,27 @@ define amdgpu_kernel void @global_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -21220,6 +22681,27 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -21522,6 +23004,27 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_system_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    global_wb scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_inv scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll
index 8a5c5dda9f79c..e16b40b8606cd 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll
@@ -8,6 +8,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_volatile_load_0(
 ; GFX6-LABEL: global_volatile_load_0:
@@ -146,6 +147,17 @@ define amdgpu_kernel void @global_volatile_load_0(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_volatile_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load volatile i32, ptr addrspace(1) %in, align 4
@@ -345,6 +357,21 @@ define amdgpu_kernel void @global_volatile_load_1(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_volatile_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s4, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v1, s[2:3] scale_offset scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -501,6 +528,19 @@ define amdgpu_kernel void @global_volatile_store_0(
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_volatile_store_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[2:3], 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load i32, ptr addrspace(1) %in, align 4
@@ -693,6 +733,21 @@ define amdgpu_kernel void @global_volatile_store_1(
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scope:SCOPE_SYS
 ; GFX12-CU-NEXT:    s_wait_storecnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_volatile_store_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[2:3], 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s3, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v0, v0, s3
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1] scale_offset scope:SCOPE_SYS
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -838,6 +893,17 @@ define amdgpu_kernel void @global_volatile_workgroup_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_volatile_workgroup_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic volatile i32, ptr addrspace(1) %in syncscope("workgroup") acquire, align 4
@@ -969,6 +1035,19 @@ define amdgpu_kernel void @global_volatile_workgroup_release_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_volatile_workgroup_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
    i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic volatile i32 %in, ptr addrspace(1) %out syncscope("workgroup") release, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll
index 151ba07a0b531..8dd0f53d7c203 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_wavefront_unordered_load(
 ; GFX6-LABEL: global_wavefront_unordered_load:
@@ -190,6 +191,17 @@ define amdgpu_kernel void @global_wavefront_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront") unordered, align 4
@@ -374,6 +386,17 @@ define amdgpu_kernel void @global_wavefront_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront") monotonic, align 4
@@ -558,6 +581,17 @@ define amdgpu_kernel void @global_wavefront_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront") acquire, align 4
@@ -742,6 +776,17 @@ define amdgpu_kernel void @global_wavefront_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront") seq_cst, align 4
@@ -899,6 +944,17 @@ define amdgpu_kernel void @global_wavefront_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront") unordered, align 4
@@ -1055,6 +1111,17 @@ define amdgpu_kernel void @global_wavefront_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront") monotonic, align 4
@@ -1211,6 +1278,17 @@ define amdgpu_kernel void @global_wavefront_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront") release, align 4
@@ -1367,6 +1445,17 @@ define amdgpu_kernel void @global_wavefront_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront") seq_cst, align 4
@@ -1521,6 +1610,17 @@ define amdgpu_kernel void @global_wavefront_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") monotonic
@@ -1675,6 +1775,17 @@ define amdgpu_kernel void @global_wavefront_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") acquire
@@ -1829,6 +1940,17 @@ define amdgpu_kernel void @global_wavefront_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") release
@@ -1983,6 +2105,17 @@ define amdgpu_kernel void @global_wavefront_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") acq_rel
@@ -2137,6 +2270,17 @@ define amdgpu_kernel void @global_wavefront_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") seq_cst
@@ -2319,6 +2463,19 @@ define amdgpu_kernel void @global_wavefront_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") acquire
@@ -2502,6 +2659,19 @@ define amdgpu_kernel void @global_wavefront_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") acq_rel
@@ -2685,6 +2855,19 @@ define amdgpu_kernel void @global_wavefront_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront") seq_cst
@@ -2904,6 +3087,21 @@ define amdgpu_kernel void @global_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3123,6 +3321,21 @@ define amdgpu_kernel void @global_wavefront_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3342,6 +3555,21 @@ define amdgpu_kernel void @global_wavefront_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3561,6 +3789,21 @@ define amdgpu_kernel void @global_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3780,6 +4023,21 @@ define amdgpu_kernel void @global_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3999,6 +4257,21 @@ define amdgpu_kernel void @global_wavefront_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4218,6 +4491,21 @@ define amdgpu_kernel void @global_wavefront_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4437,6 +4725,21 @@ define amdgpu_kernel void @global_wavefront_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4656,6 +4959,21 @@ define amdgpu_kernel void @global_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4875,6 +5193,21 @@ define amdgpu_kernel void @global_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5094,6 +5427,21 @@ define amdgpu_kernel void @global_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5313,6 +5661,21 @@ define amdgpu_kernel void @global_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5532,6 +5895,21 @@ define amdgpu_kernel void @global_wavefront_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5751,6 +6129,21 @@ define amdgpu_kernel void @global_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5970,6 +6363,21 @@ define amdgpu_kernel void @global_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6219,6 +6627,23 @@ define amdgpu_kernel void @global_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
    ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6470,6 +6895,23 @@ define amdgpu_kernel void @global_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6721,6 +7163,23 @@ define amdgpu_kernel void @global_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
    ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6972,6 +7431,23 @@ define amdgpu_kernel void @global_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7223,6 +7699,23 @@ define amdgpu_kernel void @global_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7474,6 +7967,23 @@ define amdgpu_kernel void @global_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7725,6 +8235,23 @@ define amdgpu_kernel void @global_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7976,6 +8503,23 @@ define amdgpu_kernel void @global_wavefront_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8227,6 +8771,23 @@ define amdgpu_kernel void @global_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8478,6 +9039,23 @@ define amdgpu_kernel void @global_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8729,6 +9307,23 @@ define amdgpu_kernel void @global_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8980,6 +9575,23 @@ define amdgpu_kernel void @global_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9231,6 +9843,23 @@ define amdgpu_kernel void @global_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9482,6 +10111,23 @@ define amdgpu_kernel void @global_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9733,6 +10379,23 @@ define amdgpu_kernel void @global_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9919,6 +10582,17 @@ define amdgpu_kernel void @global_wavefront_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront-one-as") unordered, align 4
@@ -10103,6 +10777,17 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront-one-as") monotonic, align 4
@@ -10287,6 +10972,17 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront-one-as") acquire, align 4
@@ -10471,6 +11167,17 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("wavefront-one-as") seq_cst, align 4
@@ -10628,6 +11335,17 @@ define amdgpu_kernel void @global_wavefront_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront-one-as") unordered, align 4
@@ -10784,6 +11502,17 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront-one-as") monotonic, align 4
@@ -10940,6 +11669,17 @@ define amdgpu_kernel void @global_wavefront_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront-one-as") release, align 4
@@ -11096,6 +11836,17 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("wavefront-one-as") seq_cst, align 4
@@ -11250,6 +12001,17 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") monotonic
@@ -11404,6 +12166,17 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") acquire
@@ -11558,6 +12331,17 @@ define amdgpu_kernel void @global_wavefront_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") release
@@ -11712,6 +12496,17 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") acq_rel
@@ -11866,6 +12661,17 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") seq_cst
@@ -12048,6 +12854,19 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") acquire
@@ -12231,6 +13050,19 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") acq_rel
@@ -12414,6 +13246,19 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("wavefront-one-as") seq_cst
@@ -12633,6 +13478,21 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -12852,6 +13712,21 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13071,6 +13946,21 @@ define amdgpu_kernel void @global_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13290,6 +14180,21 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13509,6 +14414,21 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13728,6 +14648,21 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13947,6 +14882,21 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14166,6 +15116,21 @@ define amdgpu_kernel void @global_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14385,6 +15350,21 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14604,6 +15584,21 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14823,6 +15818,21 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15042,6 +16052,21 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15261,6 +16286,21 @@ define amdgpu_kernel void @global_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15480,6 +16520,21 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15699,6 +16754,21 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15948,6 +17018,23 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_monotonic_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16199,6 +17286,23 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16450,6 +17554,23 @@ define amdgpu_kernel void @global_wavefront_one_as_release_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16701,6 +17822,23 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16952,6 +18090,23 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17203,6 +18358,23 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_acquire_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17454,6 +18626,23 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17705,6 +18894,23 @@ define amdgpu_kernel void @global_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17956,6 +19162,23 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18207,6 +19430,23 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18458,6 +19698,23 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18709,6 +19966,23 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18960,6 +20234,23 @@ define amdgpu_kernel void @global_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19211,6 +20502,23 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19462,6 +20770,23 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll
index 69b0c7f93ab0e..36732b509c702 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @global_workgroup_unordered_load(
 ; GFX6-LABEL: global_workgroup_unordered_load:
@@ -190,6 +191,17 @@ define amdgpu_kernel void @global_workgroup_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup") unordered, align 4
@@ -374,6 +386,17 @@ define amdgpu_kernel void @global_workgroup_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup") monotonic, align 4
@@ -563,6 +586,17 @@ define amdgpu_kernel void @global_workgroup_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup") acquire, align 4
@@ -764,6 +798,19 @@ define amdgpu_kernel void @global_workgroup_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup") seq_cst, align 4
@@ -921,6 +968,17 @@ define amdgpu_kernel void @global_workgroup_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup") unordered, align 4
@@ -1077,6 +1135,17 @@ define amdgpu_kernel void @global_workgroup_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup") monotonic, align 4
@@ -1251,6 +1320,19 @@ define amdgpu_kernel void @global_workgroup_release_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup") release, align 4
@@ -1425,6 +1507,19 @@ define amdgpu_kernel void @global_workgroup_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup") seq_cst, align 4
@@ -1579,6 +1674,17 @@ define amdgpu_kernel void @global_workgroup_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") monotonic
@@ -1743,6 +1849,18 @@ define amdgpu_kernel void @global_workgroup_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") acquire
@@ -1915,6 +2033,19 @@ define amdgpu_kernel void @global_workgroup_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") release
@@ -2097,6 +2228,20 @@ define amdgpu_kernel void @global_workgroup_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") acq_rel
@@ -2279,6 +2424,20 @@ define amdgpu_kernel void @global_workgroup_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") seq_cst
@@ -2466,6 +2625,19 @@ define amdgpu_kernel void @global_workgroup_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") acquire
@@ -2674,6 +2846,21 @@ define amdgpu_kernel void @global_workgroup_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") acq_rel
@@ -2882,6 +3069,21 @@ define amdgpu_kernel void @global_workgroup_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup") seq_cst
@@ -3101,6 +3303,21 @@ define amdgpu_kernel void @global_workgroup_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3330,6 +3547,22 @@ define amdgpu_kernel void @global_workgroup_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3567,6 +3800,23 @@ define amdgpu_kernel void @global_workgroup_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -3814,6 +4064,24 @@ define amdgpu_kernel void @global_workgroup_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4061,6 +4329,24 @@ define amdgpu_kernel void @global_workgroup_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4290,6 +4576,22 @@ define amdgpu_kernel void @global_workgroup_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4519,6 +4821,22 @@ define amdgpu_kernel void @global_workgroup_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -4766,6 +5084,24 @@ define amdgpu_kernel void @global_workgroup_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5013,6 +5349,24 @@ define amdgpu_kernel void @global_workgroup_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5260,6 +5614,24 @@ define amdgpu_kernel void @global_workgroup_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5507,6 +5879,24 @@ define amdgpu_kernel void @global_workgroup_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -5754,6 +6144,24 @@ define amdgpu_kernel void @global_workgroup_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6001,6 +6409,24 @@ define amdgpu_kernel void @global_workgroup_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6248,6 +6674,24 @@ define amdgpu_kernel void @global_workgroup_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6495,6 +6939,24 @@ define amdgpu_kernel void @global_workgroup_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -6744,6 +7206,23 @@ define amdgpu_kernel void @global_workgroup_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7000,6 +7479,23 @@ define amdgpu_kernel void @global_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7269,6 +7765,25 @@ define amdgpu_kernel void @global_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7545,6 +8060,25 @@ define amdgpu_kernel void @global_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -7821,6 +8355,25 @@ define amdgpu_kernel void @global_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8079,6 +8632,23 @@ define amdgpu_kernel void @global_workgroup_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8335,6 +8905,23 @@ define amdgpu_kernel void @global_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8611,6 +9198,25 @@ define amdgpu_kernel void @global_workgroup_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -8887,6 +9493,25 @@ define amdgpu_kernel void @global_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9163,6 +9788,25 @@ define amdgpu_kernel void @global_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9439,6 +10083,25 @@ define amdgpu_kernel void @global_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9713,6 +10376,25 @@ define amdgpu_kernel void @global_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -9989,6 +10671,25 @@ define amdgpu_kernel void @global_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10265,6 +10966,25 @@ define amdgpu_kernel void @global_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10541,6 +11261,25 @@ define amdgpu_kernel void @global_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -10727,6 +11466,17 @@ define amdgpu_kernel void @global_workgroup_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup-one-as") unordered, align 4
@@ -10911,6 +11661,17 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup-one-as") monotonic, align 4
@@ -11100,6 +11861,17 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup-one-as") acquire, align 4
@@ -11297,6 +12069,19 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    global_load_b32 v1, v0, s[2:3]
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(1) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(1) %in syncscope("workgroup-one-as") seq_cst, align 4
@@ -11454,6 +12239,17 @@ define amdgpu_kernel void @global_workgroup_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup-one-as") unordered, align 4
@@ -11610,6 +12406,17 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup-one-as") monotonic, align 4
@@ -11776,6 +12583,19 @@ define amdgpu_kernel void @global_workgroup_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup-one-as") release, align 4
@@ -11942,6 +12762,19 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(1) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(1) %out syncscope("workgroup-one-as") seq_cst, align 4
@@ -12096,6 +12929,17 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") monotonic
@@ -12260,6 +13104,18 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") acquire
@@ -12424,6 +13280,19 @@ define amdgpu_kernel void @global_workgroup_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") release
@@ -12598,6 +13467,20 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") acq_rel
@@ -12772,6 +13655,20 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX12-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") seq_cst
@@ -12959,6 +13856,19 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") acquire
@@ -13159,6 +14069,21 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") acq_rel
@@ -13359,6 +14284,21 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(1) %out, i32 %in syncscope("workgroup-one-as") seq_cst
@@ -13578,6 +14518,21 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -13807,6 +14762,22 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14036,6 +15007,23 @@ define amdgpu_kernel void @global_workgroup_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14275,6 +15263,24 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14514,6 +15520,24 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14743,6 +15767,22 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -14972,6 +16012,22 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15211,6 +16267,24 @@ define amdgpu_kernel void @global_workgroup_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15450,6 +16524,24 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15689,6 +16781,24 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -15928,6 +17038,24 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16167,6 +17295,24 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16406,6 +17552,24 @@ define amdgpu_kernel void @global_workgroup_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16645,6 +17809,24 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -16884,6 +18066,24 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, v3
 ; GFX12-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[1:2], s[0:1] offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v0, v[2:3], s[0:1] offset:16
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17133,6 +18333,23 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_monotonic_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17389,6 +18606,23 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17650,6 +18884,25 @@ define amdgpu_kernel void @global_workgroup_one_as_release_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -17918,6 +19171,25 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18186,6 +19458,25 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18444,6 +19735,23 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_acquire_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18700,6 +20008,23 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -18968,6 +20293,25 @@ define amdgpu_kernel void @global_workgroup_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19236,6 +20580,25 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19504,6 +20867,25 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -19772,6 +21154,25 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20038,6 +21439,25 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20306,6 +21726,25 @@ define amdgpu_kernel void @global_workgroup_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20574,6 +22013,25 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
@@ -20842,6 +22300,25 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_loadcnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: global_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0xc
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v3, v1
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    global_atomic_cmpswap_b32 v1, v0, v[2:3], s[0:1] offset:16 th:TH_ATOMIC_RETURN
+; GFX1250-CU-NEXT:    s_wait_loadcnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(1) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-agent.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-agent.ll
index 0467c5047a0be..6dde2d824ecaf 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-agent.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-agent.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @local_agent_unordered_load(
 ; GFX6-LABEL: local_agent_unordered_load:
@@ -177,6 +178,18 @@ define amdgpu_kernel void @local_agent_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent") unordered, align 4
@@ -348,6 +361,18 @@ define amdgpu_kernel void @local_agent_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent") monotonic, align 4
@@ -524,6 +549,18 @@ define amdgpu_kernel void @local_agent_acquire_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent") acquire, align 4
@@ -718,6 +755,20 @@ define amdgpu_kernel void @local_agent_seq_cst_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent") seq_cst, align 4
@@ -859,6 +910,16 @@ define amdgpu_kernel void @local_agent_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent") unordered, align 4
@@ -999,6 +1060,16 @@ define amdgpu_kernel void @local_agent_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent") monotonic, align 4
@@ -1157,6 +1228,18 @@ define amdgpu_kernel void @local_agent_release_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent") release, align 4
@@ -1315,6 +1398,18 @@ define amdgpu_kernel void @local_agent_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent") seq_cst, align 4
@@ -1455,6 +1550,16 @@ define amdgpu_kernel void @local_agent_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") monotonic
@@ -1611,6 +1716,17 @@ define amdgpu_kernel void @local_agent_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") acquire
@@ -1769,6 +1885,18 @@ define amdgpu_kernel void @local_agent_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") release
@@ -1943,6 +2071,19 @@ define amdgpu_kernel void @local_agent_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") acq_rel
@@ -2117,6 +2258,19 @@ define amdgpu_kernel void @local_agent_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") seq_cst
@@ -2304,6 +2458,19 @@ define amdgpu_kernel void @local_agent_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") acquire
@@ -2510,6 +2677,21 @@ define amdgpu_kernel void @local_agent_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") acq_rel
@@ -2716,6 +2898,21 @@ define amdgpu_kernel void @local_agent_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent") seq_cst
@@ -2883,6 +3080,18 @@ define amdgpu_kernel void @local_agent_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3066,6 +3275,19 @@ define amdgpu_kernel void @local_agent_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3251,6 +3473,20 @@ define amdgpu_kernel void @local_agent_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3452,6 +3688,21 @@ define amdgpu_kernel void @local_agent_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3653,6 +3904,21 @@ define amdgpu_kernel void @local_agent_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3836,6 +4102,19 @@ define amdgpu_kernel void @local_agent_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4019,6 +4298,19 @@ define amdgpu_kernel void @local_agent_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4220,6 +4512,21 @@ define amdgpu_kernel void @local_agent_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4421,6 +4728,21 @@ define amdgpu_kernel void @local_agent_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4622,6 +4944,21 @@ define amdgpu_kernel void @local_agent_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4823,6 +5160,21 @@ define amdgpu_kernel void @local_agent_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5024,6 +5376,21 @@ define amdgpu_kernel void @local_agent_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5225,6 +5592,21 @@ define amdgpu_kernel void @local_agent_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5426,6 +5808,21 @@ define amdgpu_kernel void @local_agent_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5627,6 +6024,21 @@ define amdgpu_kernel void @local_agent_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5836,6 +6248,21 @@ define amdgpu_kernel void @local_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6052,6 +6479,21 @@ define amdgpu_kernel void @local_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6281,6 +6723,23 @@ define amdgpu_kernel void @local_agent_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6515,6 +6974,23 @@ define amdgpu_kernel void @local_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6749,6 +7225,23 @@ define amdgpu_kernel void @local_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6965,6 +7458,21 @@ define amdgpu_kernel void @local_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7181,6 +7689,21 @@ define amdgpu_kernel void @local_agent_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7415,6 +7938,23 @@ define amdgpu_kernel void @local_agent_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7649,6 +8189,23 @@ define amdgpu_kernel void @local_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7883,6 +8440,23 @@ define amdgpu_kernel void @local_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8117,6 +8691,23 @@ define amdgpu_kernel void @local_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8351,6 +8942,23 @@ define amdgpu_kernel void @local_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8585,6 +9193,23 @@ define amdgpu_kernel void @local_agent_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8819,6 +9444,23 @@ define amdgpu_kernel void @local_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -9053,6 +9695,23 @@ define amdgpu_kernel void @local_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -9226,6 +9885,18 @@ define amdgpu_kernel void @local_agent_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent-one-as") unordered, align 4
@@ -9397,6 +10068,18 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent-one-as") monotonic, align 4
@@ -9568,6 +10251,18 @@ define amdgpu_kernel void @local_agent_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent-one-as") acquire, align 4
@@ -9739,6 +10434,18 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("agent-one-as") seq_cst, align 4
@@ -9880,6 +10587,16 @@ define amdgpu_kernel void @local_agent_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent-one-as") unordered, align 4
@@ -10020,6 +10737,16 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent-one-as") monotonic, align 4
@@ -10160,6 +10887,16 @@ define amdgpu_kernel void @local_agent_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent-one-as") release, align 4
@@ -10300,6 +11037,16 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("agent-one-as") seq_cst, align 4
@@ -10440,6 +11187,16 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") monotonic
@@ -10580,6 +11337,16 @@ define amdgpu_kernel void @local_agent_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") acquire
@@ -10720,6 +11487,16 @@ define amdgpu_kernel void @local_agent_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") release
@@ -10860,6 +11637,16 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") acq_rel
@@ -11000,6 +11787,16 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") seq_cst
@@ -11182,6 +11979,19 @@ define amdgpu_kernel void @local_agent_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") acquire
@@ -11365,6 +12175,19 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") acq_rel
@@ -11548,6 +12371,19 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("agent-one-as") seq_cst
@@ -11715,6 +12551,18 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11882,6 +12730,18 @@ define amdgpu_kernel void @local_agent_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12049,6 +12909,18 @@ define amdgpu_kernel void @local_agent_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12216,6 +13088,18 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12383,6 +13267,18 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12550,6 +13446,18 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12717,6 +13625,18 @@ define amdgpu_kernel void @local_agent_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12884,6 +13804,18 @@ define amdgpu_kernel void @local_agent_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13051,6 +13983,18 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13218,6 +14162,18 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13385,6 +14341,18 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13552,6 +14520,18 @@ define amdgpu_kernel void @local_agent_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13719,6 +14699,18 @@ define amdgpu_kernel void @local_agent_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13886,6 +14878,18 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14053,6 +15057,18 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14262,6 +15278,21 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14473,6 +15504,21 @@ define amdgpu_kernel void @local_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14684,6 +15730,21 @@ define amdgpu_kernel void @local_agent_one_as_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14895,6 +15956,21 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15106,6 +16182,21 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15317,6 +16408,21 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15528,6 +16634,21 @@ define amdgpu_kernel void @local_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15739,6 +16860,21 @@ define amdgpu_kernel void @local_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15950,6 +17086,21 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16161,6 +17312,21 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16372,6 +17538,21 @@ define amdgpu_kernel void @local_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16583,6 +17764,21 @@ define amdgpu_kernel void @local_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16794,6 +17990,21 @@ define amdgpu_kernel void @local_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -17005,6 +18216,21 @@ define amdgpu_kernel void @local_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -17216,6 +18442,21 @@ define amdgpu_kernel void @local_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll
index 78209ee34cad4..689932469d78d 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @local_nontemporal_load_0(
 ; GFX6-LABEL: local_nontemporal_load_0:
@@ -193,6 +194,18 @@ define amdgpu_kernel void @local_nontemporal_load_0(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_nontemporal_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(1) %out) {
 entry:
   %val = load i32, ptr addrspace(3) %in, align 4, !nontemporal !0
@@ -428,6 +441,22 @@ define amdgpu_kernel void @local_nontemporal_load_1(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_nontemporal_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_mov_b32 s2, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s2
+; GFX1250-CU-NEXT:    s_mov_b32 s2, 2
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_lshl_add_u32 v1, v1, s2, s3
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(1) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -597,6 +626,18 @@ define amdgpu_kernel void @local_nontemporal_store_0(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_nontemporal_store_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[2:3], 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(3) %out) {
 entry:
   %val = load i32, ptr addrspace(1) %in, align 4
@@ -802,6 +843,22 @@ define amdgpu_kernel void @local_nontemporal_store_1(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_nontemporal_store_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[0:1], 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s1, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v0, v0, s1
+; GFX1250-CU-NEXT:    s_mov_b32 s1, 2
+; GFX1250-CU-NEXT:    v_lshl_add_u32 v0, v0, s1, s2
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(3) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -991,6 +1048,18 @@ define amdgpu_kernel void @local_nontemporal_volatile_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_nontemporal_volatile_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(1) %out) {
 entry:
   %val = load volatile i32, ptr addrspace(3) %in, align 4, !nontemporal !0
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-singlethread.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-singlethread.ll
index f84d451f8ecb0..97c80ece2b053 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-singlethread.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-singlethread.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @local_singlethread_unordered_load(
 ; GFX6-LABEL: local_singlethread_unordered_load:
@@ -177,6 +178,18 @@ define amdgpu_kernel void @local_singlethread_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread") unordered, align 4
@@ -348,6 +361,18 @@ define amdgpu_kernel void @local_singlethread_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread") monotonic, align 4
@@ -519,6 +544,18 @@ define amdgpu_kernel void @local_singlethread_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread") acquire, align 4
@@ -690,6 +727,18 @@ define amdgpu_kernel void @local_singlethread_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread") seq_cst, align 4
@@ -831,6 +880,16 @@ define amdgpu_kernel void @local_singlethread_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread") unordered, align 4
@@ -971,6 +1030,16 @@ define amdgpu_kernel void @local_singlethread_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread") monotonic, align 4
@@ -1111,6 +1180,16 @@ define amdgpu_kernel void @local_singlethread_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread") release, align 4
@@ -1251,6 +1330,16 @@ define amdgpu_kernel void @local_singlethread_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread") seq_cst, align 4
@@ -1391,6 +1480,16 @@ define amdgpu_kernel void @local_singlethread_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") monotonic
@@ -1531,6 +1630,16 @@ define amdgpu_kernel void @local_singlethread_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") acquire
@@ -1671,6 +1780,16 @@ define amdgpu_kernel void @local_singlethread_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") release
@@ -1811,6 +1930,16 @@ define amdgpu_kernel void @local_singlethread_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") acq_rel
@@ -1951,6 +2080,16 @@ define amdgpu_kernel void @local_singlethread_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") seq_cst
@@ -2133,6 +2272,19 @@ define amdgpu_kernel void @local_singlethread_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") acquire
@@ -2316,6 +2468,19 @@ define amdgpu_kernel void @local_singlethread_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") acq_rel
@@ -2499,6 +2664,19 @@ define amdgpu_kernel void @local_singlethread_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread") seq_cst
@@ -2666,6 +2844,18 @@ define amdgpu_kernel void @local_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -2833,6 +3023,18 @@ define amdgpu_kernel void @local_singlethread_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3000,6 +3202,18 @@ define amdgpu_kernel void @local_singlethread_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3167,6 +3381,18 @@ define amdgpu_kernel void @local_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3334,6 +3560,18 @@ define amdgpu_kernel void @local_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3501,6 +3739,18 @@ define amdgpu_kernel void @local_singlethread_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3668,6 +3918,18 @@ define amdgpu_kernel void @local_singlethread_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3835,6 +4097,18 @@ define amdgpu_kernel void @local_singlethread_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4002,6 +4276,18 @@ define amdgpu_kernel void @local_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4169,6 +4455,18 @@ define amdgpu_kernel void @local_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4336,6 +4634,18 @@ define amdgpu_kernel void @local_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4503,6 +4813,18 @@ define amdgpu_kernel void @local_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4670,6 +4992,18 @@ define amdgpu_kernel void @local_singlethread_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4837,6 +5171,18 @@ define amdgpu_kernel void @local_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5004,6 +5350,18 @@ define amdgpu_kernel void @local_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5213,6 +5571,21 @@ define amdgpu_kernel void @local_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5424,6 +5797,21 @@ define amdgpu_kernel void @local_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5635,6 +6023,21 @@ define amdgpu_kernel void @local_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5846,6 +6249,21 @@ define amdgpu_kernel void @local_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6057,6 +6475,21 @@ define amdgpu_kernel void @local_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6268,6 +6701,21 @@ define amdgpu_kernel void @local_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6479,6 +6927,21 @@ define amdgpu_kernel void @local_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6690,6 +7153,21 @@ define amdgpu_kernel void @local_singlethread_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6901,6 +7379,21 @@ define amdgpu_kernel void @local_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7112,6 +7605,21 @@ define amdgpu_kernel void @local_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7323,6 +7831,21 @@ define amdgpu_kernel void @local_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7534,6 +8057,21 @@ define amdgpu_kernel void @local_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7745,6 +8283,21 @@ define amdgpu_kernel void @local_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7956,6 +8509,21 @@ define amdgpu_kernel void @local_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8167,6 +8735,21 @@ define amdgpu_kernel void @local_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8340,6 +8923,18 @@ define amdgpu_kernel void @local_singlethread_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread-one-as") unordered, align 4
@@ -8511,6 +9106,18 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread-one-as") monotonic, align 4
@@ -8682,6 +9289,18 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread-one-as") acquire, align 4
@@ -8853,6 +9472,18 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("singlethread-one-as") seq_cst, align 4
@@ -8994,6 +9625,16 @@ define amdgpu_kernel void @local_singlethread_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread-one-as") unordered, align 4
@@ -9134,6 +9775,16 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread-one-as") monotonic, align 4
@@ -9274,6 +9925,16 @@ define amdgpu_kernel void @local_singlethread_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread-one-as") release, align 4
@@ -9414,6 +10075,16 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("singlethread-one-as") seq_cst, align 4
@@ -9554,6 +10225,16 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") monotonic
@@ -9694,6 +10375,16 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") acquire
@@ -9834,6 +10525,16 @@ define amdgpu_kernel void @local_singlethread_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") release
@@ -9974,6 +10675,16 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") acq_rel
@@ -10114,6 +10825,16 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") seq_cst
@@ -10296,6 +11017,19 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") acquire
@@ -10479,6 +11213,19 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") acq_rel
@@ -10662,6 +11409,19 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("singlethread-one-as") seq_cst
@@ -10829,6 +11589,18 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_monotonic_cmpxchg
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -10996,6 +11768,18 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11163,6 +11947,18 @@ define amdgpu_kernel void @local_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11330,6 +12126,18 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11497,6 +12305,18 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11664,6 +12484,18 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11831,6 +12663,18 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11998,6 +12842,18 @@ define amdgpu_kernel void @local_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12165,6 +13021,18 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12332,6 +13200,18 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12499,6 +13379,18 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12666,6 +13558,18 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12833,6 +13737,18 @@ define amdgpu_kernel void @local_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13000,6 +13916,18 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13167,6 +14095,18 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13376,6 +14316,21 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_monotonic_ret_cmp
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13587,6 +14542,21 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_monotonic_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13798,6 +14768,21 @@ define amdgpu_kernel void @local_singlethread_one_as_release_monotonic_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14009,6 +14994,21 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_monotonic_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14220,6 +15220,21 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_monotonic_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14431,6 +15446,21 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_acquire_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14642,6 +15672,21 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_acquire_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14853,6 +15898,21 @@ define amdgpu_kernel void @local_singlethread_one_as_release_acquire_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15064,6 +16124,21 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_acquire_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15275,6 +16350,21 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_acquire_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15486,6 +16576,21 @@ define amdgpu_kernel void @local_singlethread_one_as_monotonic_seq_cst_ret_cmpxc
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15697,6 +16802,21 @@ define amdgpu_kernel void @local_singlethread_one_as_acquire_seq_cst_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15908,6 +17028,21 @@ define amdgpu_kernel void @local_singlethread_one_as_release_seq_cst_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16119,6 +17254,21 @@ define amdgpu_kernel void @local_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16330,6 +17480,21 @@ define amdgpu_kernel void @local_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-system.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-system.ll
index 74a297241d851..2a09fce2474d2 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-system.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-system.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @local_system_unordered_load(
 ; GFX6-LABEL: local_system_unordered_load:
@@ -177,6 +178,18 @@ define amdgpu_kernel void @local_system_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in unordered, align 4
@@ -348,6 +361,18 @@ define amdgpu_kernel void @local_system_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in monotonic, align 4
@@ -524,6 +549,18 @@ define amdgpu_kernel void @local_system_acquire_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in acquire, align 4
@@ -718,6 +755,20 @@ define amdgpu_kernel void @local_system_seq_cst_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in seq_cst, align 4
@@ -859,6 +910,16 @@ define amdgpu_kernel void @local_system_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out unordered, align 4
@@ -999,6 +1060,16 @@ define amdgpu_kernel void @local_system_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out monotonic, align 4
@@ -1157,6 +1228,18 @@ define amdgpu_kernel void @local_system_release_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out release, align 4
@@ -1315,6 +1398,18 @@ define amdgpu_kernel void @local_system_seq_cst_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out seq_cst, align 4
@@ -1455,6 +1550,16 @@ define amdgpu_kernel void @local_system_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in monotonic
@@ -1611,6 +1716,17 @@ define amdgpu_kernel void @local_system_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in acquire
@@ -1769,6 +1885,18 @@ define amdgpu_kernel void @local_system_release_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in release
@@ -1943,6 +2071,19 @@ define amdgpu_kernel void @local_system_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in acq_rel
@@ -2117,6 +2258,19 @@ define amdgpu_kernel void @local_system_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in seq_cst
@@ -2304,6 +2458,19 @@ define amdgpu_kernel void @local_system_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in acquire
@@ -2510,6 +2677,21 @@ define amdgpu_kernel void @local_system_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in acq_rel
@@ -2716,6 +2898,21 @@ define amdgpu_kernel void @local_system_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in seq_cst
@@ -2883,6 +3080,18 @@ define amdgpu_kernel void @local_system_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3066,6 +3275,19 @@ define amdgpu_kernel void @local_system_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3251,6 +3473,20 @@ define amdgpu_kernel void @local_system_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3452,6 +3688,21 @@ define amdgpu_kernel void @local_system_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3653,6 +3904,21 @@ define amdgpu_kernel void @local_system_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3836,6 +4102,19 @@ define amdgpu_kernel void @local_system_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4019,6 +4298,19 @@ define amdgpu_kernel void @local_system_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4220,6 +4512,21 @@ define amdgpu_kernel void @local_system_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4421,6 +4728,21 @@ define amdgpu_kernel void @local_system_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4622,6 +4944,21 @@ define amdgpu_kernel void @local_system_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4823,6 +5160,21 @@ define amdgpu_kernel void @local_system_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5024,6 +5376,21 @@ define amdgpu_kernel void @local_system_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5225,6 +5592,21 @@ define amdgpu_kernel void @local_system_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5426,6 +5808,21 @@ define amdgpu_kernel void @local_system_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5627,6 +6024,21 @@ define amdgpu_kernel void @local_system_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5836,6 +6248,21 @@ define amdgpu_kernel void @local_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6052,6 +6479,21 @@ define amdgpu_kernel void @local_system_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6281,6 +6723,23 @@ define amdgpu_kernel void @local_system_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6515,6 +6974,23 @@ define amdgpu_kernel void @local_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6749,6 +7225,23 @@ define amdgpu_kernel void @local_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6965,6 +7458,21 @@ define amdgpu_kernel void @local_system_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7181,6 +7689,21 @@ define amdgpu_kernel void @local_system_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7415,6 +7938,23 @@ define amdgpu_kernel void @local_system_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7649,6 +8189,23 @@ define amdgpu_kernel void @local_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7883,6 +8440,23 @@ define amdgpu_kernel void @local_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8117,6 +8691,23 @@ define amdgpu_kernel void @local_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8351,6 +8942,23 @@ define amdgpu_kernel void @local_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8585,6 +9193,23 @@ define amdgpu_kernel void @local_system_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8819,6 +9444,23 @@ define amdgpu_kernel void @local_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -9053,6 +9695,23 @@ define amdgpu_kernel void @local_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -9226,6 +9885,18 @@ define amdgpu_kernel void @local_system_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("one-as") unordered, align 4
@@ -9397,6 +10068,18 @@ define amdgpu_kernel void @local_system_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("one-as") monotonic, align 4
@@ -9568,6 +10251,18 @@ define amdgpu_kernel void @local_system_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("one-as") acquire, align 4
@@ -9739,6 +10434,18 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("one-as") seq_cst, align 4
@@ -9880,6 +10587,16 @@ define amdgpu_kernel void @local_system_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("one-as") unordered, align 4
@@ -10020,6 +10737,16 @@ define amdgpu_kernel void @local_system_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("one-as") monotonic, align 4
@@ -10160,6 +10887,16 @@ define amdgpu_kernel void @local_system_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("one-as") release, align 4
@@ -10300,6 +11037,16 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("one-as") seq_cst, align 4
@@ -10440,6 +11187,16 @@ define amdgpu_kernel void @local_system_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") monotonic
@@ -10580,6 +11337,16 @@ define amdgpu_kernel void @local_system_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") acquire
@@ -10720,6 +11487,16 @@ define amdgpu_kernel void @local_system_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") release
@@ -10860,6 +11637,16 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") acq_rel
@@ -11000,6 +11787,16 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") seq_cst
@@ -11182,6 +11979,19 @@ define amdgpu_kernel void @local_system_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") acquire
@@ -11365,6 +12175,19 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") acq_rel
@@ -11548,6 +12371,19 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("one-as") seq_cst
@@ -11715,6 +12551,18 @@ define amdgpu_kernel void @local_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11882,6 +12730,18 @@ define amdgpu_kernel void @local_system_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12049,6 +12909,18 @@ define amdgpu_kernel void @local_system_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12216,6 +13088,18 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12383,6 +13267,18 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12550,6 +13446,18 @@ define amdgpu_kernel void @local_system_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12717,6 +13625,18 @@ define amdgpu_kernel void @local_system_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12884,6 +13804,18 @@ define amdgpu_kernel void @local_system_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13051,6 +13983,18 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13218,6 +14162,18 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13385,6 +14341,18 @@ define amdgpu_kernel void @local_system_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13552,6 +14520,18 @@ define amdgpu_kernel void @local_system_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13719,6 +14699,18 @@ define amdgpu_kernel void @local_system_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13886,6 +14878,18 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14053,6 +15057,18 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14262,6 +15278,21 @@ define amdgpu_kernel void @local_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14473,6 +15504,21 @@ define amdgpu_kernel void @local_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14684,6 +15730,21 @@ define amdgpu_kernel void @local_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14895,6 +15956,21 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15106,6 +16182,21 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15317,6 +16408,21 @@ define amdgpu_kernel void @local_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15528,6 +16634,21 @@ define amdgpu_kernel void @local_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15739,6 +16860,21 @@ define amdgpu_kernel void @local_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15950,6 +17086,21 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16161,6 +17312,21 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16372,6 +17538,21 @@ define amdgpu_kernel void @local_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16583,6 +17764,21 @@ define amdgpu_kernel void @local_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16794,6 +17990,21 @@ define amdgpu_kernel void @local_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -17005,6 +18216,21 @@ define amdgpu_kernel void @local_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -17216,6 +18442,21 @@ define amdgpu_kernel void @local_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_system_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll
index 5e5e3bf83d610..4dc4c51dd2419 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll
@@ -8,6 +8,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @local_volatile_load_0(
 ; GFX6-LABEL: local_volatile_load_0:
@@ -141,6 +142,18 @@ define amdgpu_kernel void @local_volatile_load_0(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_volatile_load_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(1) %out) {
 entry:
   %val = load volatile i32, ptr addrspace(3) %in, align 4
@@ -308,6 +321,22 @@ define amdgpu_kernel void @local_volatile_load_1(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_volatile_load_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, v0
+; GFX1250-CU-NEXT:    s_load_b32 s3, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x8
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1250-CU-NEXT:    s_mov_b32 s2, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v1, v1, s2
+; GFX1250-CU-NEXT:    s_mov_b32 s2, 2
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_lshl_add_u32 v1, v1, s2, s3
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v1
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(1) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -429,6 +458,18 @@ define amdgpu_kernel void @local_volatile_store_0(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_volatile_store_0:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[2:3], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[2:3], 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(3) %out) {
 entry:
   %val = load i32, ptr addrspace(1) %in, align 4
@@ -570,6 +611,22 @@ define amdgpu_kernel void @local_volatile_store_1(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_volatile_store_1:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[0:1], 0x0
+; GFX1250-CU-NEXT:    s_wait_xcnt 0x0
+; GFX1250-CU-NEXT:    s_mov_b32 s1, 0x3ff
+; GFX1250-CU-NEXT:    v_and_b32_e64 v0, v0, s1
+; GFX1250-CU-NEXT:    s_mov_b32 s1, 2
+; GFX1250-CU-NEXT:    v_lshl_add_u32 v0, v0, s1, s2
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(1) %in, ptr addrspace(3) %out) {
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -698,6 +755,18 @@ define amdgpu_kernel void @local_volatile_workgroup_acquire_load(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_volatile_workgroup_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic volatile i32, ptr addrspace(3) %in syncscope("workgroup") acquire, align 4
@@ -813,6 +882,18 @@ define amdgpu_kernel void @local_volatile_workgroup_release_store(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_volatile_workgroup_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
    i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic volatile i32 %in, ptr addrspace(3) %out syncscope("workgroup") release, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-wavefront.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-wavefront.ll
index b24622a48a16b..b8ad75049aff8 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-wavefront.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-wavefront.ll
@@ -12,6 +12,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1100 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX11-CU %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12-WGP %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1200 -mattr=+cumode < %s | FileCheck --check-prefixes=GFX12-CU %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -O0 -mcpu=gfx1250 < %s | FileCheck --check-prefixes=GFX1250-CU %s
 
 define amdgpu_kernel void @local_wavefront_unordered_load(
 ; GFX6-LABEL: local_wavefront_unordered_load:
@@ -177,6 +178,18 @@ define amdgpu_kernel void @local_wavefront_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront") unordered, align 4
@@ -348,6 +361,18 @@ define amdgpu_kernel void @local_wavefront_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront") monotonic, align 4
@@ -519,6 +544,18 @@ define amdgpu_kernel void @local_wavefront_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront") acquire, align 4
@@ -690,6 +727,18 @@ define amdgpu_kernel void @local_wavefront_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront") seq_cst, align 4
@@ -831,6 +880,16 @@ define amdgpu_kernel void @local_wavefront_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront") unordered, align 4
@@ -971,6 +1030,16 @@ define amdgpu_kernel void @local_wavefront_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront") monotonic, align 4
@@ -1111,6 +1180,16 @@ define amdgpu_kernel void @local_wavefront_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront") release, align 4
@@ -1251,6 +1330,16 @@ define amdgpu_kernel void @local_wavefront_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront") seq_cst, align 4
@@ -1391,6 +1480,16 @@ define amdgpu_kernel void @local_wavefront_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") monotonic
@@ -1531,6 +1630,16 @@ define amdgpu_kernel void @local_wavefront_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") acquire
@@ -1671,6 +1780,16 @@ define amdgpu_kernel void @local_wavefront_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") release
@@ -1811,6 +1930,16 @@ define amdgpu_kernel void @local_wavefront_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") acq_rel
@@ -1951,6 +2080,16 @@ define amdgpu_kernel void @local_wavefront_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") seq_cst
@@ -2133,6 +2272,19 @@ define amdgpu_kernel void @local_wavefront_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") acquire
@@ -2316,6 +2468,19 @@ define amdgpu_kernel void @local_wavefront_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") acq_rel
@@ -2499,6 +2664,19 @@ define amdgpu_kernel void @local_wavefront_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront") seq_cst
@@ -2666,6 +2844,18 @@ define amdgpu_kernel void @local_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -2833,6 +3023,18 @@ define amdgpu_kernel void @local_wavefront_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3000,6 +3202,18 @@ define amdgpu_kernel void @local_wavefront_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3167,6 +3381,18 @@ define amdgpu_kernel void @local_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3334,6 +3560,18 @@ define amdgpu_kernel void @local_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3501,6 +3739,18 @@ define amdgpu_kernel void @local_wavefront_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3668,6 +3918,18 @@ define amdgpu_kernel void @local_wavefront_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -3835,6 +4097,18 @@ define amdgpu_kernel void @local_wavefront_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4002,6 +4276,18 @@ define amdgpu_kernel void @local_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4169,6 +4455,18 @@ define amdgpu_kernel void @local_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4336,6 +4634,18 @@ define amdgpu_kernel void @local_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4503,6 +4813,18 @@ define amdgpu_kernel void @local_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4670,6 +4992,18 @@ define amdgpu_kernel void @local_wavefront_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -4837,6 +5171,18 @@ define amdgpu_kernel void @local_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5004,6 +5350,18 @@ define amdgpu_kernel void @local_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5213,6 +5571,21 @@ define amdgpu_kernel void @local_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5424,6 +5797,21 @@ define amdgpu_kernel void @local_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5635,6 +6023,21 @@ define amdgpu_kernel void @local_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -5846,6 +6249,21 @@ define amdgpu_kernel void @local_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6057,6 +6475,21 @@ define amdgpu_kernel void @local_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6268,6 +6701,21 @@ define amdgpu_kernel void @local_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6479,6 +6927,21 @@ define amdgpu_kernel void @local_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6690,6 +7153,21 @@ define amdgpu_kernel void @local_wavefront_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -6901,6 +7379,21 @@ define amdgpu_kernel void @local_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7112,6 +7605,21 @@ define amdgpu_kernel void @local_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7323,6 +7831,21 @@ define amdgpu_kernel void @local_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7534,6 +8057,21 @@ define amdgpu_kernel void @local_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7745,6 +8283,21 @@ define amdgpu_kernel void @local_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -7956,6 +8509,21 @@ define amdgpu_kernel void @local_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8167,6 +8735,21 @@ define amdgpu_kernel void @local_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -8340,6 +8923,18 @@ define amdgpu_kernel void @local_wavefront_one_as_unordered_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_unordered_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront-one-as") unordered, align 4
@@ -8511,6 +9106,18 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront-one-as") monotonic, align 4
@@ -8682,6 +9289,18 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront-one-as") acquire, align 4
@@ -8853,6 +9472,18 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_load(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_load:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %in, ptr addrspace(3) %out) {
 entry:
   %val = load atomic i32, ptr addrspace(3) %in syncscope("wavefront-one-as") seq_cst, align 4
@@ -8994,6 +9625,16 @@ define amdgpu_kernel void @local_wavefront_one_as_unordered_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_unordered_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront-one-as") unordered, align 4
@@ -9134,6 +9775,16 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront-one-as") monotonic, align 4
@@ -9274,6 +9925,16 @@ define amdgpu_kernel void @local_wavefront_one_as_release_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront-one-as") release, align 4
@@ -9414,6 +10075,16 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_store(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_store:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
 entry:
   store atomic i32 %in, ptr addrspace(3) %out syncscope("wavefront-one-as") seq_cst, align 4
@@ -9554,6 +10225,16 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") monotonic
@@ -9694,6 +10375,16 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") acquire
@@ -9834,6 +10525,16 @@ define amdgpu_kernel void @local_wavefront_one_as_release_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") release
@@ -9974,6 +10675,16 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") acq_rel
@@ -10114,6 +10825,16 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_atomicrmw(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX12-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") seq_cst
@@ -10296,6 +11017,19 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") acquire
@@ -10479,6 +11213,19 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") acq_rel
@@ -10662,6 +11409,19 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_ret_atomicrmw(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_ret_atomicrmw:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
 entry:
   %val = atomicrmw volatile xchg ptr addrspace(3) %out, i32 %in syncscope("wavefront-one-as") seq_cst
@@ -10829,6 +11589,18 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -10996,6 +11768,18 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11163,6 +11947,18 @@ define amdgpu_kernel void @local_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11330,6 +12126,18 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11497,6 +12305,18 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_monotonic_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11664,6 +12484,18 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11831,6 +12663,18 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -11998,6 +12842,18 @@ define amdgpu_kernel void @local_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12165,6 +13021,18 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12332,6 +13200,18 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_acquire_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12499,6 +13379,18 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12666,6 +13558,18 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -12833,6 +13737,18 @@ define amdgpu_kernel void @local_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13000,6 +13916,18 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13167,6 +14095,18 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX12-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX12-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_seq_cst_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
+; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13376,6 +14316,21 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_monotonic_ret_cmpxch
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13587,6 +14542,21 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -13798,6 +14768,21 @@ define amdgpu_kernel void @local_wavefront_one_as_release_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14009,6 +14994,21 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14220,6 +15220,21 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14431,6 +15446,21 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14642,6 +15672,21 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -14853,6 +15898,21 @@ define amdgpu_kernel void @local_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15064,6 +16124,21 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15275,6 +16350,21 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_acquire_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15486,6 +16576,21 @@ define amdgpu_kernel void @local_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15697,6 +16802,21 @@ define amdgpu_kernel void @local_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acquire_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -15908,6 +17028,21 @@ define amdgpu_kernel void @local_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_release_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16119,6 +17254,21 @@ define amdgpu_kernel void @local_wavefront_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_acq_rel_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
@@ -16330,6 +17480,21 @@ define amdgpu_kernel void @local_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX12-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX12-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX12-CU-NEXT:    s_endpgm
+;
+; GFX1250-CU-LABEL: local_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg:
+; GFX1250-CU:       ; %bb.0: ; %entry
+; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x0
+; GFX1250-CU-NEXT:    s_load_b32 s2, s[4:5], 0x4
+; GFX1250-CU-NEXT:    s_load_b32 s1, s[4:5], 0x8
+; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
+; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
+; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
+; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
 entry:
   %gep = getelementptr i32, ptr addrspace(3) %out, i32 4
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-workgroup.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-workgroup.ll
index 94f5aab1eb67d..97ece0f94ccc0 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-workgroup.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-workgroup.ll
@@ -762,7 +762,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_load(
 ; GFX1250-CU-NEXT:    s_load_b32 s0, s[4:5], 0x4
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_load_b32 v1, v0
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -1235,7 +1236,8 @@ define amdgpu_kernel void @local_workgroup_release_store(
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
@@ -1404,7 +1406,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_store(
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_store_b32 v0, v1
 ; GFX1250-CU-NEXT:    s_endpgm
     i32 %in, ptr addrspace(3) %out) {
@@ -1890,7 +1893,8 @@ define amdgpu_kernel void @local_workgroup_release_atomicrmw(
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in) {
@@ -2075,7 +2079,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_atomicrmw(
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -2261,7 +2266,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_atomicrmw(
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v0, v0, v1
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -2679,7 +2685,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_ret_atomicrmw(
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -2899,7 +2906,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_ret_atomicrmw(
 ; GFX1250-CU-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_storexchg_rtn_b32 v1, v0, v1
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -3475,7 +3483,8 @@ define amdgpu_kernel void @local_workgroup_release_monotonic_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_endpgm
     ptr addrspace(3) %out, i32 %in, i32 %old) {
@@ -3689,7 +3698,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_monotonic_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -3904,7 +3914,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_monotonic_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -4511,7 +4522,8 @@ define amdgpu_kernel void @local_workgroup_release_acquire_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -4726,7 +4738,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_acquire_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -4941,7 +4954,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_acquire_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -5156,7 +5170,8 @@ define amdgpu_kernel void @local_workgroup_monotonic_seq_cst_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -5371,7 +5386,8 @@ define amdgpu_kernel void @local_workgroup_acquire_seq_cst_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -5586,7 +5602,8 @@ define amdgpu_kernel void @local_workgroup_release_seq_cst_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -5801,7 +5818,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_seq_cst_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -6016,7 +6034,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_seq_cst_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s0
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_b32 v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    s_endpgm
@@ -6714,7 +6733,8 @@ define amdgpu_kernel void @local_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
@@ -6964,7 +6984,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -7214,7 +7235,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -7926,7 +7948,8 @@ define amdgpu_kernel void @local_workgroup_release_acquire_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -8176,7 +8199,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -8426,7 +8450,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -8676,7 +8701,8 @@ define amdgpu_kernel void @local_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -8926,7 +8952,8 @@ define amdgpu_kernel void @local_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -9176,7 +9203,8 @@ define amdgpu_kernel void @local_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -9426,7 +9454,8 @@ define amdgpu_kernel void @local_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
@@ -9676,7 +9705,8 @@ define amdgpu_kernel void @local_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v2, s1
-; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
+; GFX1250-CU-NEXT:    s_wait_storecnt 0x0
+; GFX1250-CU-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-CU-NEXT:    ds_cmpstore_rtn_b32 v1, v0, v1, v2 offset:16
 ; GFX1250-CU-NEXT:    s_wait_dscnt 0x0
 ; GFX1250-CU-NEXT:    v_mov_b32_e32 v0, s0