[X86][SSE] movddup shuffle mask decodes

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

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226705 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/InstPrinter/X86InstComments.cpp b/lib/Target/X86/InstPrinter/X86InstComments.cpp
index a4c0ca8..b7b2ba3 100644 (file)
--- a/lib/Target/X86/InstPrinter/X86InstComments.cpp
+++ b/lib/Target/X86/InstPrinter/X86InstComments.cpp
@@ -200,12 +200,30 @@ bool llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS,
   case X86::MOVSHDUPrm:
   case X86::VMOVSHDUPrm:
     DestName = getRegName(MI->getOperand(0).getReg());
-    DecodeMOVSHDUPMask(MVT::v4f32, ShuffleMask);
-    break;
-
-  case X86::PSLLDQri:
-  case X86::VPSLLDQri:
-    Src1Name = getRegName(MI->getOperand(1).getReg());
+    DecodeMOVSHDUPMask(MVT::v4f32, ShuffleMask);\r
+    break;\r
+\r
+  case X86::VMOVDDUPYrr:\r
+    Src1Name = getRegName(MI->getOperand(1).getReg());\r
+    // FALL THROUGH.\r
+  case X86::VMOVDDUPYrm:\r
+    DestName = getRegName(MI->getOperand(0).getReg());\r
+    DecodeMOVDDUPMask(MVT::v4f64, ShuffleMask);\r
+    break;\r
+\r
+  case X86::MOVDDUPrr:\r
+  case X86::VMOVDDUPrr:\r
+    Src1Name = getRegName(MI->getOperand(1).getReg());\r
+    // FALL THROUGH.\r
+  case X86::MOVDDUPrm:\r
+  case X86::VMOVDDUPrm:\r
+    DestName = getRegName(MI->getOperand(0).getReg());\r
+    DecodeMOVDDUPMask(MVT::v2f64, ShuffleMask);\r
+    break;\r
+\r
+  case X86::PSLLDQri:\r
+  case X86::VPSLLDQri:\r
+    Src1Name = getRegName(MI->getOperand(1).getReg());\r
     DestName = getRegName(MI->getOperand(0).getReg());
     if(MI->getOperand(MI->getNumOperands()-1).isImm())
       DecodePSLLDQMask(MVT::v16i8,

diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index e4c58d4..399d583 100644 (file)
--- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/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;
+  }\r
+}\r
+\r
+void DecodeMOVDDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask) {\r
+  unsigned VectorSizeInBits = VT.getSizeInBits();\r
+  unsigned ScalarSizeInBits = VT.getScalarSizeInBits();\r
+  unsigned NumElts = VT.getVectorNumElements();\r
+  unsigned NumLanes = VectorSizeInBits / 128;\r
+  unsigned NumLaneElts = NumElts / NumLanes;\r
+  unsigned NumLaneSubElts = 64 / ScalarSizeInBits;\r
+\r
+  for (unsigned l = 0; l < NumElts; l += NumLaneElts)\r
+    for (unsigned i = 0; i < NumLaneElts; i += NumLaneSubElts)\r
+      for (unsigned s = 0; s != NumLaneSubElts; s++)\r
+        ShuffleMask.push_back(l + s);\r
+}\r
+\r
+void DecodePSLLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {\r
+  unsigned VectorSizeInBits = VT.getSizeInBits();\r
+  unsigned NumElts = VectorSizeInBits / 8;\r
   unsigned NumLanes = VectorSizeInBits / 128;
   unsigned NumLaneElts = NumElts / NumLanes;
 

diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.h b/lib/Target/X86/Utils/X86ShuffleDecode.h
index 6ba3c64..188ea41 100644 (file)
--- a/lib/Target/X86/Utils/X86ShuffleDecode.h
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.h
@@ -37,12 +37,14 @@ void DecodeMOVHLPSMask(unsigned NElts, SmallVectorImpl<int> &ShuffleMask);
 void DecodeMOVLHPSMask(unsigned NElts, SmallVectorImpl<int> &ShuffleMask);
 
 void DecodeMOVSLDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
-
-void DecodeMOVSHDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
-
-void DecodePSLLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
-
-void DecodePSRLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
+\r
+void DecodeMOVSHDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);\r
+\r
+void DecodeMOVDDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);\r
+\r
+void DecodePSLLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);\r
+\r
+void DecodePSRLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);\r
 
 void DecodePALIGNRMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);