[AVX-512] Fix DecodeVPERMV3Mask to handle cases where the constant pool entry has a different type than the shuffle itself.

Summary: This is especially important for 32-bit targets with 64-bit shuffle elements.This is similar to how PSHUFB and VPERMIL handle the same problem.

Reviewers: RKSimon

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D25666

llvm-svn: 284451

1 files changed, 22 insertions, 17 deletions

diff --git a/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp b/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp
index f00c85eda1a..04b00d34d92 100644
--- a/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp
+++ b/llvm/lib/Target/X86/X86ShuffleDecodeConstantPool.cpp
@@ -309,26 +309,31 @@ void DecodeVPERMVMask(const Constant *C, MVT VT,
     ShuffleMask.push_back(Element);
 }
 
-void DecodeVPERMV3Mask(const Constant *C, MVT VT,
+void DecodeVPERMV3Mask(const Constant *C, unsigned ElSize,
                        SmallVectorImpl<int> &ShuffleMask) {
   Type *MaskTy = C->getType();
-  unsigned NumElements = MaskTy->getVectorNumElements();
-  if (NumElements == VT.getVectorNumElements()) {
-    unsigned EltMaskSize = Log2_64(NumElements * 2);
-    for (unsigned i = 0; i < NumElements; ++i) {
-      Constant *COp = C->getAggregateElement(i);
-      if (!COp) {
-        ShuffleMask.clear();
-        return;
-      }
-      if (isa<UndefValue>(COp))
-        ShuffleMask.push_back(SM_SentinelUndef);
-      else {
-        APInt Element = cast<ConstantInt>(COp)->getValue();
-        Element = Element.getLoBits(EltMaskSize);
-        ShuffleMask.push_back(Element.getZExtValue());
-      }
+  unsigned MaskTySize = MaskTy->getPrimitiveSizeInBits();
+  (void)MaskTySize;
+  assert((MaskTySize == 128 || MaskTySize == 256 || MaskTySize == 512) &&
+         "Unexpected vector size.");
+  assert((ElSize == 8 || ElSize == 16 || ElSize == 32 || ElSize == 64) &&
+         "Unexpected vector element size.");
+
+  // The shuffle mask requires elements the same size as the target.
+  SmallBitVector UndefElts;
+  SmallVector<uint64_t, 8> RawMask;
+  if (!extractConstantMask(C, ElSize, UndefElts, RawMask))
+    return;
+
+  unsigned NumElts = RawMask.size();
+
+  for (unsigned i = 0; i != NumElts; ++i) {
+    if (UndefElts[i]) {
+      ShuffleMask.push_back(SM_SentinelUndef);
+      continue;
     }
+    int Index = RawMask[i] & (NumElts*2 - 1);
+    ShuffleMask.push_back(Index);
   }
 }
 } // llvm namespace