[AMDGPU]: Unpack packed instructions overlapped by MFMAs post-RA scheduling (llvm#157968) (llvm#4137)

This is a cleaned up version of PR llvm#151704. These optimizations are now
performed post-RA scheduling.

(cherry picked from commit c256966)

Co-authored-by: Akash Dutta <137309513+akadutta@users.noreply.github.com>

Author: bcahoon

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

@@ -6081,6 +6081,66 @@ bool SIInstrInfo::isOperandLegal(const MachineInstr &MI, unsigned OpIdx,
   return isImmOperandLegal(MI, OpIdx, *MO);
 }
 
+bool SIInstrInfo::isNeverCoissue(MachineInstr &MI) const {
+  bool IsGFX950Only = ST.hasGFX950Insts();
+  bool IsGFX940Only = ST.hasGFX940Insts();
+
+  if (!IsGFX950Only && !IsGFX940Only)
+    return false;
+
+  if (!isVALU(MI))
+    return false;
+
+  // V_COS, V_EXP, V_RCP, etc.
+  if (isTRANS(MI))
+    return true;
+
+  // DOT2, DOT2C, DOT4, etc.
+  if (isDOT(MI))
+    return true;
+
+  // MFMA, SMFMA
+  if (isMFMA(MI))
+    return true;
+
+  unsigned Opcode = MI.getOpcode();
+  switch (Opcode) {
+  case AMDGPU::V_CVT_PK_BF8_F32_e64:
+  case AMDGPU::V_CVT_PK_FP8_F32_e64:
+  case AMDGPU::V_MQSAD_PK_U16_U8_e64:
+  case AMDGPU::V_MQSAD_U32_U8_e64:
+  case AMDGPU::V_PK_ADD_F16:
+  case AMDGPU::V_PK_ADD_F32:
+  case AMDGPU::V_PK_ADD_I16:
+  case AMDGPU::V_PK_ADD_U16:
+  case AMDGPU::V_PK_ASHRREV_I16:
+  case AMDGPU::V_PK_FMA_F16:
+  case AMDGPU::V_PK_FMA_F32:
+  case AMDGPU::V_PK_FMAC_F16_e32:
+  case AMDGPU::V_PK_FMAC_F16_e64:
+  case AMDGPU::V_PK_LSHLREV_B16:
+  case AMDGPU::V_PK_LSHRREV_B16:
+  case AMDGPU::V_PK_MAD_I16:
+  case AMDGPU::V_PK_MAD_U16:
+  case AMDGPU::V_PK_MAX_F16:
+  case AMDGPU::V_PK_MAX_I16:
+  case AMDGPU::V_PK_MAX_U16:
+  case AMDGPU::V_PK_MIN_F16:
+  case AMDGPU::V_PK_MIN_I16:
+  case AMDGPU::V_PK_MIN_U16:
+  case AMDGPU::V_PK_MOV_B32:
+  case AMDGPU::V_PK_MUL_F16:
+  case AMDGPU::V_PK_MUL_F32:
+  case AMDGPU::V_PK_MUL_LO_U16:
+  case AMDGPU::V_PK_SUB_I16:
+  case AMDGPU::V_PK_SUB_U16:
+  case AMDGPU::V_QSAD_PK_U16_U8_e64:
+    return true;
+  default:
+    return false;
+  }
+}
+
 void SIInstrInfo::legalizeOperandsVOP2(MachineRegisterInfo &MRI,
                                        MachineInstr &MI) const {
   unsigned Opc = MI.getOpcode();

llvm/lib/Target/AMDGPU/SIInstrInfo.h

@@ -1139,6 +1139,8 @@ class SIInstrInfo final : public AMDGPUGenInstrInfo {
   bool isImmOperandLegal(const MachineInstr &MI, unsigned OpNo,
                          const MachineOperand &MO) const;
 
+  bool isNeverCoissue(MachineInstr &MI) const;
+
   /// Return true if this 64-bit VALU instruction has a 32-bit encoding.
   /// This function will return false if you pass it a 32-bit instruction.
   bool hasVALU32BitEncoding(unsigned Opcode) const;