From a7a4b836a38bfdb1c25137d0e473fbe2f9c44406 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Wed, 21 Jan 2015 22:02:30 +0000
Subject: [PATCH] [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
---
 .../X86/InstPrinter/X86InstComments.cpp       | 30 +++++++++++++++----
 lib/Target/X86/Utils/X86ShuffleDecode.cpp     | 26 ++++++++++++----
 lib/Target/X86/Utils/X86ShuffleDecode.h       | 14 +++++----
 3 files changed, 52 insertions(+), 18 deletions(-)

diff --git a/lib/Target/X86/InstPrinter/X86InstComments.cpp b/lib/Target/X86/InstPrinter/X86InstComments.cpp
index a4c0ca882c7..b7b2ba36ce7 100644
--- 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);
+    break;
+
+  case X86::VMOVDDUPYrr:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    // FALL THROUGH.
+  case X86::VMOVDDUPYrm:
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeMOVDDUPMask(MVT::v4f64, ShuffleMask);
+    break;
+
+  case X86::MOVDDUPrr:
+  case X86::VMOVDDUPrr:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
+    // FALL THROUGH.
+  case X86::MOVDDUPrm:
+  case X86::VMOVDDUPrm:
+    DestName = getRegName(MI->getOperand(0).getReg());
+    DecodeMOVDDUPMask(MVT::v2f64, ShuffleMask);
+    break;
+
+  case X86::PSLLDQri:
+  case X86::VPSLLDQri:
+    Src1Name = getRegName(MI->getOperand(1).getReg());
     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 e4c58d42c5d..399d5834188 100644
--- 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;
+  }
+}
+
+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;
 
diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.h b/lib/Target/X86/Utils/X86ShuffleDecode.h
index 6ba3c64f8ec..188ea41bb57 100644
--- 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);
+
+void DecodeMOVSHDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
+
+void DecodeMOVDDUPMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
+
+void DecodePSLLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
+
+void DecodePSRLDQMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
 
 void DecodePALIGNRMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);