[llvm] unpack packed instructions overlapped by MFMAs post-RA scheduling (PR #157968)

[Jeffrey Byrnes via llvm-commits]

================
@@ -417,6 +458,230 @@ bool SIPreEmitPeephole::removeExeczBranch(MachineInstr &MI,
   return true;
 }
 
+bool SIPreEmitPeephole::isUnpackingSupportedInstr(MachineInstr &MI) const {
+  unsigned Opcode = MI.getOpcode();
+  switch (Opcode) {
+  case AMDGPU::V_PK_ADD_F32:
+  case AMDGPU::V_PK_MUL_F32:
+  case AMDGPU::V_PK_FMA_F32:
+    return true;
+  default:
+    return false;
+  }
+  llvm_unreachable("Fully covered switch");
+}
+
+bool SIPreEmitPeephole::hasReadWriteDependencies(const MachineInstr &PredMI,
+                                                 const MachineInstr &SuccMI) {
+  for (const MachineOperand &PredOps : PredMI.operands()) {
+    if (!PredOps.isReg() || !PredOps.isDef())
+      continue;
+    Register PredReg = PredOps.getReg();
+    if (!PredReg.isValid())
+      continue;
+    for (const MachineOperand &SuccOps : SuccMI.operands()) {
+      if (!SuccOps.isReg() || !SuccOps.isDef())
+        continue;
+      Register SuccReg = SuccOps.getReg();
+      if (!SuccReg.isValid())
+        continue;
+      if ((PredReg == SuccReg) || TRI->regsOverlap(PredReg, SuccReg))
+        return true;
+    }
+  }
+  return false;
+}
+
+uint16_t SIPreEmitPeephole::mapToUnpackedOpcode(MachineInstr &I) {
+  unsigned Opcode = I.getOpcode();
+  // Use 64 bit encoding to allow use of VOP3 instructions.
+  // VOP3 instructions allow VOP3P source modifiers to be translated to VOP3
+  // e32 instructions are VOP2 and don't allow source modifiers
+  switch (Opcode) {
+  case AMDGPU::V_PK_ADD_F32:
+    return AMDGPU::V_ADD_F32_e64;
+  case AMDGPU::V_PK_MUL_F32:
+    return AMDGPU::V_MUL_F32_e64;
+  case AMDGPU::V_PK_FMA_F32:
+    return AMDGPU::V_FMA_F32_e64;
+  default:
+    return std::numeric_limits<uint16_t>::max();
+  }
+  llvm_unreachable("Fully covered switch");
+}
+
+void SIPreEmitPeephole::addOperandAndMods(MachineInstrBuilder NewMI,
+                                          unsigned Src_Mods,
+                                          unsigned NegModifier,
+                                          unsigned OpSelModifier,
+                                          MachineOperand &SrcMO) {
+  unsigned NewSrcMods = 0;
+  const TargetRegisterInfo *RI = SrcMO.getParent()
----------------
jrbyrnes wrote:

The TargetRegisterInfo is available through SIPreEmitPeephole

https://github.com/llvm/llvm-project/pull/157968