[AVX-512] Add support for decoding shuffle mask from constant pool for masked VPERMILPS/PD.

llvm-svn: 284450

1 files changed, 53 insertions, 24 deletions

diff --git a/llvm/lib/Target/X86/X86MCInstLower.cpp b/llvm/lib/Target/X86/X86MCInstLower.cpp
index 25a9cd0ec2e..df3c24d1e57 100644
--- a/llvm/lib/Target/X86/X86MCInstLower.cpp
+++ b/llvm/lib/Target/X86/X86MCInstLower.cpp
@@ -1527,44 +1527,73 @@ void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) {
     break;
   }
 
+  case X86::VPERMILPSrm:
+  case X86::VPERMILPSYrm:
+  case X86::VPERMILPSZ128rm:
+  case X86::VPERMILPSZ128rmk:
+  case X86::VPERMILPSZ128rmkz:
+  case X86::VPERMILPSZ256rm:
+  case X86::VPERMILPSZ256rmk:
+  case X86::VPERMILPSZ256rmkz:
+  case X86::VPERMILPSZrm:
+  case X86::VPERMILPSZrmk:
+  case X86::VPERMILPSZrmkz:
   case X86::VPERMILPDrm:
   case X86::VPERMILPDYrm:
   case X86::VPERMILPDZ128rm:
+  case X86::VPERMILPDZ128rmk:
+  case X86::VPERMILPDZ128rmkz:
   case X86::VPERMILPDZ256rm:
-  case X86::VPERMILPDZrm: {
+  case X86::VPERMILPDZ256rmk:
+  case X86::VPERMILPDZ256rmkz:
+  case X86::VPERMILPDZrm:
+  case X86::VPERMILPDZrmk:
+  case X86::VPERMILPDZrmkz: {
     if (!OutStreamer->isVerboseAsm())
       break;
-    assert(MI->getNumOperands() >= 6 &&
-           "We should always have at least 6 operands!");
-    const MachineOperand &DstOp = MI->getOperand(0);
-    const MachineOperand &SrcOp = MI->getOperand(1);
-    const MachineOperand &MaskOp = MI->getOperand(5);
-
-    if (auto *C = getConstantFromPool(*MI, MaskOp)) {
-      SmallVector<int, 8> Mask;
-      DecodeVPERMILPMask(C, 64, Mask);
-      if (!Mask.empty())
-        OutStreamer->AddComment(getShuffleComment(DstOp, SrcOp, SrcOp, Mask));
+    unsigned SrcIdx, MaskIdx;
+    unsigned ElSize;
+    switch (MI->getOpcode()) {
+    default: llvm_unreachable("Invalid opcode");
+    case X86::VPERMILPSrm:
+    case X86::VPERMILPSYrm:
+    case X86::VPERMILPSZ128rm:
+    case X86::VPERMILPSZ256rm:
+    case X86::VPERMILPSZrm:
+      SrcIdx = 1; MaskIdx = 5; ElSize = 32; break;
+    case X86::VPERMILPSZ128rmkz:
+    case X86::VPERMILPSZ256rmkz:
+    case X86::VPERMILPSZrmkz:
+      SrcIdx = 2; MaskIdx = 6; ElSize = 32; break;
+    case X86::VPERMILPSZ128rmk:
+    case X86::VPERMILPSZ256rmk:
+    case X86::VPERMILPSZrmk:
+      SrcIdx = 3; MaskIdx = 7; ElSize = 32; break;
+    case X86::VPERMILPDrm:
+    case X86::VPERMILPDYrm:
+    case X86::VPERMILPDZ128rm:
+    case X86::VPERMILPDZ256rm:
+    case X86::VPERMILPDZrm:
+      SrcIdx = 1; MaskIdx = 5; ElSize = 64; break;
+    case X86::VPERMILPDZ128rmkz:
+    case X86::VPERMILPDZ256rmkz:
+    case X86::VPERMILPDZrmkz:
+      SrcIdx = 2; MaskIdx = 6; ElSize = 64; break;
+    case X86::VPERMILPDZ128rmk:
+    case X86::VPERMILPDZ256rmk:
+    case X86::VPERMILPDZrmk:
+      SrcIdx = 3; MaskIdx = 7; ElSize = 64; break;
     }
-    break;
-  }
 
-  case X86::VPERMILPSrm:
-  case X86::VPERMILPSYrm:
-  case X86::VPERMILPSZ128rm:
-  case X86::VPERMILPSZ256rm:
-  case X86::VPERMILPSZrm: {
-    if (!OutStreamer->isVerboseAsm())
-      break;
     assert(MI->getNumOperands() >= 6 &&
            "We should always have at least 6 operands!");
     const MachineOperand &DstOp = MI->getOperand(0);
-    const MachineOperand &SrcOp = MI->getOperand(1);
-    const MachineOperand &MaskOp = MI->getOperand(5);
+    const MachineOperand &SrcOp = MI->getOperand(SrcIdx);
+    const MachineOperand &MaskOp = MI->getOperand(MaskIdx);
 
     if (auto *C = getConstantFromPool(*MI, MaskOp)) {
       SmallVector<int, 16> Mask;
-      DecodeVPERMILPMask(C, 32, Mask);
+      DecodeVPERMILPMask(C, ElSize, Mask);
       if (!Mask.empty())
         OutStreamer->AddComment(getShuffleComment(DstOp, SrcOp, SrcOp, Mask));
     }