[X86] Add missing immediate qualifier to the (V)INSERTPS instruction names

Matches (V)BLENDPS etc and makes it easier to algorithmically recreate the instruction name in various analysis scripts I'm working on

Author: RKSimon

llvm/lib/Target/X86/MCTargetDesc/X86InstComments.cpp

@@ -703,14 +703,14 @@ bool llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
     DestName = getRegName(MI->getOperand(0).getReg());
     break;
 
-  case X86::INSERTPSrr:
-  case X86::VINSERTPSrr:
-  case X86::VINSERTPSZrr:
+  case X86::INSERTPSrri:
+  case X86::VINSERTPSrri:
+  case X86::VINSERTPSZrri:
     Src2Name = getRegName(MI->getOperand(2).getReg());
     [[fallthrough]];
-  case X86::INSERTPSrm:
-  case X86::VINSERTPSrm:
-  case X86::VINSERTPSZrm:
+  case X86::INSERTPSrmi:
+  case X86::VINSERTPSrmi:
+  case X86::VINSERTPSZrmi:
     DestName = getRegName(MI->getOperand(0).getReg());
     Src1Name = getRegName(MI->getOperand(1).getReg());
     if (MI->getOperand(NumOperands - 1).isImm())

llvm/lib/Target/X86/X86InstrAVX512.td

@@ -649,12 +649,12 @@ defm : vinsert_for_mask_cast<"VINSERTI64x4Z", v32i8x_info, v64i8_info,
 // vinsertps - insert f32 to XMM
 let ExeDomain = SSEPackedSingle in {
 let isCommutable = 1 in
-def VINSERTPSZrr : AVX512AIi8<0x21, MRMSrcReg, (outs VR128X:$dst),
+def VINSERTPSZrri : AVX512AIi8<0x21, MRMSrcReg, (outs VR128X:$dst),
       (ins VR128X:$src1, VR128X:$src2, u8imm:$src3),
       "vinsertps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
       [(set VR128X:$dst, (X86insertps VR128X:$src1, VR128X:$src2, timm:$src3))]>,
       EVEX, VVVV, Sched<[SchedWriteFShuffle.XMM]>;
-def VINSERTPSZrm: AVX512AIi8<0x21, MRMSrcMem, (outs VR128X:$dst),
+def VINSERTPSZrmi : AVX512AIi8<0x21, MRMSrcMem, (outs VR128X:$dst),
       (ins VR128X:$src1, f32mem:$src2, u8imm:$src3),
       "vinsertps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
       [(set VR128X:$dst, (X86insertps VR128X:$src1,

llvm/lib/Target/X86/X86InstrInfo.cpp

@@ -2424,9 +2424,9 @@ MachineInstr *X86InstrInfo::commuteInstructionImpl(MachineInstr &MI, bool NewMI,
     WorkingMI->getOperand(3).setImm(Mask ^ Imm);
     break;
   }
-  case X86::INSERTPSrr:
-  case X86::VINSERTPSrr:
-  case X86::VINSERTPSZrr: {
+  case X86::INSERTPSrri:
+  case X86::VINSERTPSrri:
+  case X86::VINSERTPSZrri: {
     unsigned Imm = MI.getOperand(MI.getNumOperands() - 1).getImm();
     unsigned ZMask = Imm & 15;
     unsigned DstIdx = (Imm >> 4) & 3;
@@ -7274,9 +7274,9 @@ MachineInstr *X86InstrInfo::foldMemoryOperandCustom(
     ArrayRef<MachineOperand> MOs, MachineBasicBlock::iterator InsertPt,
     unsigned Size, Align Alignment) const {
   switch (MI.getOpcode()) {
-  case X86::INSERTPSrr:
-  case X86::VINSERTPSrr:
-  case X86::VINSERTPSZrr:
+  case X86::INSERTPSrri:
+  case X86::VINSERTPSrri:
+  case X86::VINSERTPSZrri:
     // Attempt to convert the load of inserted vector into a fold load
     // of a single float.
     if (OpNum == 2) {
@@ -7289,13 +7289,13 @@ MachineInstr *X86InstrInfo::foldMemoryOperandCustom(
       const TargetRegisterClass *RC = getRegClass(MI.getDesc(), OpNum, &RI, MF);
       unsigned RCSize = TRI.getRegSizeInBits(*RC) / 8;
       if ((Size == 0 || Size >= 16) && RCSize >= 16 &&
-          (MI.getOpcode() != X86::INSERTPSrr || Alignment >= Align(4))) {
+          (MI.getOpcode() != X86::INSERTPSrri || Alignment >= Align(4))) {
         int PtrOffset = SrcIdx * 4;
         unsigned NewImm = (DstIdx << 4) | ZMask;
         unsigned NewOpCode =
-            (MI.getOpcode() == X86::VINSERTPSZrr)  ? X86::VINSERTPSZrm
-            : (MI.getOpcode() == X86::VINSERTPSrr) ? X86::VINSERTPSrm
-                                                   : X86::INSERTPSrm;
+            (MI.getOpcode() == X86::VINSERTPSZrri)  ? X86::VINSERTPSZrmi
+            : (MI.getOpcode() == X86::VINSERTPSrri) ? X86::VINSERTPSrmi
+                                                    : X86::INSERTPSrmi;
         MachineInstr *NewMI =
             fuseInst(MF, NewOpCode, OpNum, MOs, InsertPt, MI, *this, PtrOffset);
         NewMI->getOperand(NewMI->getNumOperands() - 1).setImm(NewImm);

llvm/lib/Target/X86/X86InstrSSE.td

@@ -5447,7 +5447,7 @@ let Constraints = "$src1 = $dst" in
 // in the target vector.
 multiclass SS41I_insertf32<bits<8> opc, string asm, bit Is2Addr = 1> {
   let isCommutable = 1 in
-  def rr : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
+  def rri : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
       (ins VR128:$src1, VR128:$src2, u8imm:$src3),
       !if(Is2Addr,
         !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
@@ -5456,7 +5456,7 @@ multiclass SS41I_insertf32<bits<8> opc, string asm, bit Is2Addr = 1> {
       [(set VR128:$dst,
         (X86insertps VR128:$src1, VR128:$src2, timm:$src3))]>,
       Sched<[SchedWriteFShuffle.XMM]>;
-  def rm : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
+  def rmi : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
       (ins VR128:$src1, f32mem:$src2, u8imm:$src3),
       !if(Is2Addr,
         !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),