[X86][SSE] movddup shuffle mask decodes

Patch to provide shuffle decodes and asm comments for the SSE3/AVX1 movddup double duplication instructions.

llvm-svn: 226705

1 files changed, 20 insertions, 6 deletions

diff --git a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index e4c58d4..399d583 100644
--- a/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/llvm/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -76,12 +76,26 @@ void DecodeMOVSHDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask) {
   for (int i = 0, e = NumElts / 2; i < e; ++i) {
     ShuffleMask.push_back(2 * i + 1);
     ShuffleMask.push_back(2 * i + 1);
-  }
-}
-
-void DecodePSLLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
-  unsigned VectorSizeInBits = VT.getSizeInBits();
-  unsigned NumElts = VectorSizeInBits / 8;
+  }

+}

+

+void DecodeMOVDDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask) {

+  unsigned VectorSizeInBits = VT.getSizeInBits();

+  unsigned ScalarSizeInBits = VT.getScalarSizeInBits();

+  unsigned NumElts = VT.getVectorNumElements();

+  unsigned NumLanes = VectorSizeInBits / 128;

+  unsigned NumLaneElts = NumElts / NumLanes;

+  unsigned NumLaneSubElts = 64 / ScalarSizeInBits;

+

+  for (unsigned l = 0; l < NumElts; l += NumLaneElts)

+    for (unsigned i = 0; i < NumLaneElts; i += NumLaneSubElts)

+      for (unsigned s = 0; s != NumLaneSubElts; s++)

+        ShuffleMask.push_back(l + s);

+}

+

+void DecodePSLLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {

+  unsigned VectorSizeInBits = VT.getSizeInBits();

+  unsigned NumElts = VectorSizeInBits / 8;

   unsigned NumLanes = VectorSizeInBits / 128;
   unsigned NumLaneElts = NumElts / NumLanes;