Merge with LLVM upstream r146714 (Dec 16th 2011)

Change-Id: Ied458adb08bf9a69250cbcee9b14b44d17e8701a

4 files changed, 48 insertions, 115 deletions

diff --git a/lib/Target/X86/Utils/CMakeLists.txt b/lib/Target/X86/Utils/CMakeLists.txt
index caffd8b..2e72c34 100644
--- a/lib/Target/X86/Utils/CMakeLists.txt
+++ b/lib/Target/X86/Utils/CMakeLists.txt
@@ -4,9 +4,4 @@ add_llvm_library(LLVMX86Utils
   X86ShuffleDecode.cpp
   )
 
-add_llvm_library_dependencies(LLVMX86Utils
-  LLVMCore
-  LLVMSupport
-  )
-
 add_dependencies(LLVMX86Utils X86CommonTableGen)
diff --git a/lib/Target/X86/Utils/LLVMBuild.txt b/lib/Target/X86/Utils/LLVMBuild.txt
index 3ee441e..de0a30f 100644
--- a/lib/Target/X86/Utils/LLVMBuild.txt
+++ b/lib/Target/X86/Utils/LLVMBuild.txt
@@ -21,4 +21,3 @@ name = X86Utils
 parent = X86
 required_libraries = Core Support
 add_to_library_groups = X86
-
diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index f6c9d7b..e7631b6 100644
--- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -95,54 +95,31 @@ void DecodePSHUFLWMask(unsigned Imm,
   ShuffleMask.push_back(7);
 }
 
-void DecodePUNPCKLBWMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i8, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLWDMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i16, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLDQMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i32, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLQDQMask(unsigned NElts,
-                          SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(MVT::getVectorVT(MVT::i64, NElts), ShuffleMask);
-}
-
-void DecodePUNPCKLMask(EVT VT,
-                       SmallVectorImpl<unsigned> &ShuffleMask) {
-  DecodeUNPCKLPMask(VT, ShuffleMask);
-}
+void DecodeSHUFPMask(EVT VT, unsigned Imm,
+                     SmallVectorImpl<unsigned> &ShuffleMask) {
+  unsigned NumElts = VT.getVectorNumElements();
 
-void DecodePUNPCKHMask(unsigned NElts,
-                       SmallVectorImpl<unsigned> &ShuffleMask) {
-  for (unsigned i = 0; i != NElts/2; ++i) {
-    ShuffleMask.push_back(i+NElts/2);
-    ShuffleMask.push_back(i+NElts+NElts/2);
-  }
-}
+  unsigned NumLanes = VT.getSizeInBits() / 128;
+  unsigned NumLaneElts = NumElts / NumLanes;
 
-void DecodeSHUFPSMask(unsigned NElts, unsigned Imm,
-                      SmallVectorImpl<unsigned> &ShuffleMask) {
-  // Part that reads from dest.
-  for (unsigned i = 0; i != NElts/2; ++i) {
-    ShuffleMask.push_back(Imm % NElts);
-    Imm /= NElts;
-  }
-  // Part that reads from src.
-  for (unsigned i = 0; i != NElts/2; ++i) {
-    ShuffleMask.push_back(Imm % NElts + NElts);
-    Imm /= NElts;
+  int NewImm = Imm;
+  for (unsigned l = 0; l < NumLanes; ++l) {
+    unsigned LaneStart = l * NumLaneElts;
+    // Part that reads from dest.
+    for (unsigned i = 0; i != NumLaneElts/2; ++i) {
+      ShuffleMask.push_back(NewImm % NumLaneElts + LaneStart);
+      NewImm /= NumLaneElts;
+    }
+    // Part that reads from src.
+    for (unsigned i = 0; i != NumLaneElts/2; ++i) {
+      ShuffleMask.push_back(NewImm % NumLaneElts + NumElts + LaneStart);
+      NewImm /= NumLaneElts;
+    }
+    if (NumLaneElts == 4) NewImm = Imm; // reload imm
   }
 }
 
-void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate
@@ -161,10 +138,10 @@ void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
   }
 }
 
-/// DecodeUNPCKLPMask - This decodes the shuffle masks for unpcklps/unpcklpd
+/// DecodeUNPCKLMask - This decodes the shuffle masks for unpcklps/unpcklpd
 /// etc.  VT indicates the type of the vector allowing it to handle different
 /// datatypes and vector widths.
-void DecodeUNPCKLPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
+void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate
@@ -183,36 +160,23 @@ void DecodeUNPCKLPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask) {
   }
 }
 
-// DecodeVPERMILPSMask - Decodes VPERMILPS permutes for any 128-bit 32-bit
-// elements. For 256-bit vectors, it's considered as two 128 lanes, the
-// referenced elements can't cross lanes and the mask of the first lane must
-// be the same of the second.
-void DecodeVPERMILPSMask(unsigned NumElts, unsigned Imm,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  unsigned NumLanes = (NumElts*32)/128;
-  unsigned LaneSize = NumElts/NumLanes;
-
-  for (unsigned l = 0; l != NumLanes; ++l) {
-    for (unsigned i = 0; i != LaneSize; ++i) {
-      unsigned Idx = (Imm >> (i*2)) & 0x3 ;
-      ShuffleMask.push_back(Idx+(l*LaneSize));
-    }
-  }
-}
+// DecodeVPERMILPMask - Decodes VPERMILPS/ VPERMILPD permutes for any 128-bit
+// 32-bit or 64-bit elements. For 256-bit vectors, it's considered as two 128
+// lanes. For VPERMILPS, referenced elements can't cross lanes and the mask of
+// the first lane must be the same of the second.
+void DecodeVPERMILPMask(EVT VT, unsigned Imm,
+                        SmallVectorImpl<unsigned> &ShuffleMask) {
+  unsigned NumElts = VT.getVectorNumElements();
 
-// DecodeVPERMILPDMask - Decodes VPERMILPD permutes for any 128-bit 64-bit
-// elements. For 256-bit vectors, it's considered as two 128 lanes, the
-// referenced elements can't cross lanes but the mask of the first lane can
-// be the different of the second (not like VPERMILPS).
-void DecodeVPERMILPDMask(unsigned NumElts, unsigned Imm,
-                         SmallVectorImpl<unsigned> &ShuffleMask) {
-  unsigned NumLanes = (NumElts*64)/128;
-  unsigned LaneSize = NumElts/NumLanes;
+  unsigned NumLanes = VT.getSizeInBits() / 128;
+  unsigned NumLaneElts = NumElts / NumLanes;
 
-  for (unsigned l = 0; l < NumLanes; ++l) {
-    for (unsigned i = l*LaneSize; i < LaneSize*(l+1); ++i) {
-      unsigned Idx = (Imm >> i) & 0x1;
-      ShuffleMask.push_back(Idx+(l*LaneSize));
+  for (unsigned l = 0; l != NumLanes; ++l) {
+    unsigned LaneStart = l*NumLaneElts;
+    for (unsigned i = 0; i != NumLaneElts; ++i) {
+      unsigned Idx = NumLaneElts == 4 ? (Imm >> (i*2)) & 0x3
+                                      : (Imm >> (i+LaneStart)) & 0x1;
+      ShuffleMask.push_back(Idx+LaneStart);
     }
   }
 }
diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.h b/lib/Target/X86/Utils/X86ShuffleDecode.h
index 35f6530..243728f 100644
--- a/lib/Target/X86/Utils/X86ShuffleDecode.h
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.h
@@ -46,50 +46,25 @@ void DecodePSHUFHWMask(unsigned Imm,
 void DecodePSHUFLWMask(unsigned Imm,
                        SmallVectorImpl<unsigned> &ShuffleMask);
 
-void DecodePUNPCKLBWMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLWDMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLDQMask(unsigned NElts,
-                         SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLQDQMask(unsigned NElts,
-                          SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKLMask(EVT VT,
-                       SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodePUNPCKHMask(unsigned NElts,
-                       SmallVectorImpl<unsigned> &ShuffleMask);
-
-void DecodeSHUFPSMask(unsigned NElts, unsigned Imm,
-                      SmallVectorImpl<unsigned> &ShuffleMask);
+void DecodeSHUFPMask(EVT VT, unsigned Imm,
+                     SmallVectorImpl<unsigned> &ShuffleMask);
 
-/// DecodeUNPCKHPMask - This decodes the shuffle masks for unpckhps/unpckhpd
+/// DecodeUNPCKHMask - This decodes the shuffle masks for unpckhps/unpckhpd
 /// etc.  VT indicates the type of the vector allowing it to handle different
 /// datatypes and vector widths.
-void DecodeUNPCKHPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
+void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
 
-/// DecodeUNPCKLPMask - This decodes the shuffle masks for unpcklps/unpcklpd
+/// DecodeUNPCKLMask - This decodes the shuffle masks for unpcklps/unpcklpd
 /// etc.  VT indicates the type of the vector allowing it to handle different
 /// datatypes and vector widths.
-void DecodeUNPCKLPMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
-
+void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<unsigned> &ShuffleMask);
 
-// DecodeVPERMILPSMask - Decodes VPERMILPS permutes for any 128-bit 32-bit
-// elements. For 256-bit vectors, it's considered as two 128 lanes, the
-// referenced elements can't cross lanes and the mask of the first lane must
-// be the same of the second.
-void DecodeVPERMILPSMask(unsigned NElts, unsigned Imm,
-                        SmallVectorImpl<unsigned> &ShuffleMask);
 
-// DecodeVPERMILPDMask - Decodes VPERMILPD permutes for any 128-bit 64-bit
-// elements. For 256-bit vectors, it's considered as two 128 lanes, the
-// referenced elements can't cross lanes but the mask of the first lane can
-// be the different of the second (not like VPERMILPS).
-void DecodeVPERMILPDMask(unsigned NElts, unsigned Imm,
+// DecodeVPERMILPMask - Decodes VPERMILPS/ VPERMILPD permutes for any 128-bit
+// 32-bit or 64-bit elements. For 256-bit vectors, it's considered as two 128
+// lanes. For VPERMILPS, referenced elements can't cross lanes and the mask of
+// the first lane must be the same of the second.
+void DecodeVPERMILPMask(EVT VT, unsigned Imm,
                         SmallVectorImpl<unsigned> &ShuffleMask);
 
 void DecodeVPERM2F128Mask(unsigned Imm,