[clang] Fix OMPT ident flag in combined distribute parallel for pragma (llvm#80987)

Authored-by: Raúl Peñacoba Veigas <rpenacob@bsc.es>

Author: mikaoP

clang/lib/CodeGen/CGOpenMPRuntime.cpp

@@ -2647,6 +2647,9 @@ void CGOpenMPRuntime::emitDistributeStaticInit(
 void CGOpenMPRuntime::emitForStaticFinish(CodeGenFunction &CGF,
                                           SourceLocation Loc,
                                           OpenMPDirectiveKind DKind) {
+  assert(DKind == OMPD_distribute || DKind == OMPD_for ||
+         DKind == OMPD_sections &&
+             "Expected distribute, for, or sections directive kind");
   if (!CGF.HaveInsertPoint())
     return;
   // Call __kmpc_for_static_fini(ident_t *loc, kmp_int32 tid);

clang/lib/CodeGen/CGStmtOpenMP.cpp

@@ -2910,10 +2910,10 @@ void CodeGenFunction::EmitOMPOuterLoop(
   EmitBlock(LoopExit.getBlock());
 
   // Tell the runtime we are done.
-  auto &&CodeGen = [DynamicOrOrdered, &S](CodeGenFunction &CGF) {
+  auto &&CodeGen = [DynamicOrOrdered, &S, &LoopArgs](CodeGenFunction &CGF) {
     if (!DynamicOrOrdered)
       CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getEndLoc(),
-                                                     S.getDirectiveKind());
+                                                     LoopArgs.DKind);
   };
   OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen);
 }
@@ -3019,6 +3019,7 @@ void CodeGenFunction::EmitOMPForOuterLoop(
   OuterLoopArgs.Cond = S.getCond();
   OuterLoopArgs.NextLB = S.getNextLowerBound();
   OuterLoopArgs.NextUB = S.getNextUpperBound();
+  OuterLoopArgs.DKind = LoopArgs.DKind;
   EmitOMPOuterLoop(DynamicOrOrdered, IsMonotonic, S, LoopScope, OuterLoopArgs,
                    emitOMPLoopBodyWithStopPoint, CodeGenOrdered);
 }
@@ -3080,6 +3081,7 @@ void CodeGenFunction::EmitOMPDistributeOuterLoop(
   OuterLoopArgs.NextUB = isOpenMPLoopBoundSharingDirective(S.getDirectiveKind())
                              ? S.getCombinedNextUpperBound()
                              : S.getNextUpperBound();
+  OuterLoopArgs.DKind = OMPD_distribute;
 
   EmitOMPOuterLoop(/* DynamicOrOrdered = */ false, /* IsMonotonic = */ false, S,
                    LoopScope, OuterLoopArgs, CodeGenLoopContent,
@@ -3452,15 +3454,16 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(
         // Tell the runtime we are done.
         auto &&CodeGen = [&S](CodeGenFunction &CGF) {
           CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getEndLoc(),
-                                                         S.getDirectiveKind());
+                                                         OMPD_for);
         };
         OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen);
       } else {
         // Emit the outer loop, which requests its work chunk [LB..UB] from
         // runtime and runs the inner loop to process it.
-        const OMPLoopArguments LoopArguments(
+        OMPLoopArguments LoopArguments(
             LB.getAddress(*this), UB.getAddress(*this), ST.getAddress(*this),
             IL.getAddress(*this), Chunk, EUB);
+        LoopArguments.DKind = OMPD_for;
         EmitOMPForOuterLoop(ScheduleKind, IsMonotonic, S, LoopScope, Ordered,
                             LoopArguments, CGDispatchBounds);
       }
@@ -4082,7 +4085,7 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) {
     // Tell the runtime we are done.
     auto &&CodeGen = [&S](CodeGenFunction &CGF) {
       CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getEndLoc(),
-                                                     S.getDirectiveKind());
+                                                     OMPD_sections);
     };
     CGF.OMPCancelStack.emitExit(CGF, S.getDirectiveKind(), CodeGen);
     CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_parallel);
@@ -5782,7 +5785,7 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
             });
         EmitBlock(LoopExit.getBlock());
         // Tell the runtime we are done.
-        RT.emitForStaticFinish(*this, S.getEndLoc(), S.getDirectiveKind());
+        RT.emitForStaticFinish(*this, S.getEndLoc(), OMPD_distribute);
       } else {
         // Emit the outer loop, which requests its work chunk [LB..UB] from
         // runtime and runs the inner loop to process it.

clang/lib/CodeGen/CodeGenFunction.h

@@ -3831,6 +3831,8 @@ class CodeGenFunction : public CodeGenTypeCache {
     Expr *NextLB = nullptr;
     /// Update of UB after a whole chunk has been executed
     Expr *NextUB = nullptr;
+    /// Distinguish between the for distribute and sections
+    OpenMPDirectiveKind DKind = llvm::omp::OMPD_unknown;
     OMPLoopArguments() = default;
     OMPLoopArguments(Address LB, Address UB, Address ST, Address IL,
                      llvm::Value *Chunk = nullptr, Expr *EUB = nullptr,

clang/test/OpenMP/amdgcn_target_device_vla.cpp

@@ -539,7 +539,7 @@ int main() {
 // CHECK:       omp.loop.exit:
 // CHECK-NEXT:    [[TMP34:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
 // CHECK-NEXT:    [[TMP35:%.*]] = load i32, ptr [[TMP34]], align 4
-// CHECK-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP35]])
+// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP35]])
 // CHECK-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK:       omp.precond.end:
 // CHECK-NEXT:    ret void

clang/test/OpenMP/amdgpu_target_with_aligned_attribute.c

@@ -301,7 +301,7 @@ void write_to_aligned_array(int *a, int N) {
 // CHECK-AMD:       omp.loop.exit:
 // CHECK-AMD-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
 // CHECK-AMD-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
-// CHECK-AMD-NEXT:    call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP18]])
+// CHECK-AMD-NEXT:    call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP18]])
 // CHECK-AMD-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK-AMD:       omp.precond.end:
 // CHECK-AMD-NEXT:    ret void

clang/test/OpenMP/bug60602.cpp

@@ -564,7 +564,7 @@ int kernel_within_loop(int *a, int *b, int N, int num_iters) {
 // CHECK:       omp.loop.exit:
 // CHECK-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
-// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB3]], i32 [[TMP23]])
+// CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
 // CHECK-NEXT:    br label [[OMP_PRECOND_END]]
 // CHECK:       omp.precond.end:
 // CHECK-NEXT:    ret void