[X86][AVX512] Fixed decoding of permd/permpd variable mask shuffles + enabled them for target shuffle combining

Corrected element mask masking to extract the bottom index bits (now matches the perm2 implementation but for unary inputs).

llvm-svn: 274571

3 files changed, 10 insertions, 7 deletions

diff --git a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index 713595dbbc1..1a51aa5858e 100644
--- a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -556,8 +556,9 @@ void DecodeVPERMIL2PMask(MVT VT, unsigned M2Z, ArrayRef<uint64_t> RawMask,
 
 void DecodeVPERMVMask(ArrayRef<uint64_t> RawMask,
                       SmallVectorImpl<int> &ShuffleMask) {
-  for (int i = 0, e = RawMask.size(); i < e; ++i) {
-    uint64_t M = RawMask[i];
+  uint64_t EltMaskSize = RawMask.size() - 1;
+  for (auto M : RawMask) {
+    M &= EltMaskSize;
     ShuffleMask.push_back((int)M);
   }
 }
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 2e30c131cdd..d6a36901520 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -5008,8 +5008,8 @@ static bool getTargetShuffleMask(SDNode *N, MVT VT, bool AllowSentinelZero,
     Ops.push_back(N->getOperand(1));
     SDValue MaskNode = N->getOperand(0);
     SmallVector<uint64_t, 32> RawMask;
-    unsigned MaskLoBits = Log2_64(VT.getVectorNumElements());
-    if (getTargetShuffleMaskIndices(MaskNode, MaskLoBits, RawMask)) {
+    unsigned MaskEltSize = VT.getScalarSizeInBits();
+    if (getTargetShuffleMaskIndices(MaskNode, MaskEltSize, RawMask)) {
       DecodeVPERMVMask(RawMask, Mask);
       break;
     }
@@ -30778,6 +30778,7 @@ SDValue X86TargetLowering::PerformDAGCombine(SDNode *N,
   case X86ISD::MOVSS:
   case X86ISD::MOVSD:
   case X86ISD::VPPERM:
+  case X86ISD::VPERMV:
   case X86ISD::VPERMV3:
   case X86ISD::VPERMIL2:
   case X86ISD::VPERMILPI:
diff --git a/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp b/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp
index 856a2bc6a12..8c700e3f5cd 100644
--- a/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp
+++ b/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp
@@ -300,6 +300,7 @@ void DecodeVPERMVMask(const Constant *C, MVT VT,
   if (MaskTy->isVectorTy()) {
     unsigned NumElements = MaskTy->getVectorNumElements();
     if (NumElements == VT.getVectorNumElements()) {
+      unsigned EltMaskSize = Log2_64(NumElements);
       for (unsigned i = 0; i < NumElements; ++i) {
         Constant *COp = C->getAggregateElement(i);
         if (!COp || (!isa<UndefValue>(COp) && !isa<ConstantInt>(COp))) {
@@ -309,9 +310,9 @@ void DecodeVPERMVMask(const Constant *C, MVT VT,
         if (isa<UndefValue>(COp))
           ShuffleMask.push_back(SM_SentinelUndef);
         else {
-          uint64_t Element = cast<ConstantInt>(COp)->getZExtValue();
-          Element &= (1 << NumElements) - 1;
-          ShuffleMask.push_back(Element);
+          APInt Element = cast<ConstantInt>(COp)->getValue();
+          Element = Element.getLoBits(EltMaskSize);
+          ShuffleMask.push_back(Element.getZExtValue());
         }
       }
     }