Merge with LLVM upstream r160668 (Jul 24th 2012)

Conflicts:
	include/llvm/Support/ELF.h
	lib/CodeGen/AsmPrinter/AsmPrinter.cpp
	lib/Support/Memory.cpp
	lib/Transforms/Instrumentation/AddressSanitizer.cpp

Change-Id: Iddd658cf2eadc7165b2805b446d31af2c5c9917f

2 files changed, 63 insertions, 45 deletions

diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index a802333..8b87c1f 100644
--- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -64,13 +64,13 @@ void DecodeMOVLHPSMask(unsigned NElts, SmallVectorImpl<int> &ShuffleMask) {
 /// DecodePSHUFMask - This decodes the shuffle masks for pshufd, and vpermilp*.
 /// VT indicates the type of the vector allowing it to handle different
 /// datatypes and vector widths.
-void DecodePSHUFMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
+void DecodePSHUFMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   unsigned NumLanes = VT.getSizeInBits() / 128;
   unsigned NumLaneElts = NumElts / NumLanes;
 
-  int NewImm = Imm;
+  unsigned NewImm = Imm;
   for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
     for (unsigned i = 0; i != NumLaneElts; ++i) {
       ShuffleMask.push_back(NewImm % NumLaneElts + l);
@@ -80,48 +80,55 @@ void DecodePSHUFMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
   }
 }
 
-void DecodePSHUFHWMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
-  ShuffleMask.push_back(0);
-  ShuffleMask.push_back(1);
-  ShuffleMask.push_back(2);
-  ShuffleMask.push_back(3);
-  for (unsigned i = 0; i != 4; ++i) {
-    ShuffleMask.push_back(4+(Imm & 3));
-    Imm >>= 2;
+void DecodePSHUFHWMask(MVT VT, unsigned Imm,
+                       SmallVectorImpl<int> &ShuffleMask) {
+  unsigned NumElts = VT.getVectorNumElements();
+
+  for (unsigned l = 0; l != NumElts; l += 8) {
+    unsigned NewImm = Imm;
+    for (unsigned i = 0, e = 4; i != e; ++i) {
+      ShuffleMask.push_back(l + i);
+    }
+    for (unsigned i = 4, e = 8; i != e; ++i) {
+      ShuffleMask.push_back(l + 4 + (NewImm & 3));
+      NewImm >>= 2;
+    }
   }
 }
 
-void DecodePSHUFLWMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
-  for (unsigned i = 0; i != 4; ++i) {
-    ShuffleMask.push_back((Imm & 3));
-    Imm >>= 2;
+void DecodePSHUFLWMask(MVT VT, unsigned Imm,
+                       SmallVectorImpl<int> &ShuffleMask) {
+  unsigned NumElts = VT.getVectorNumElements();
+
+  for (unsigned l = 0; l != NumElts; l += 8) {
+    unsigned NewImm = Imm;
+    for (unsigned i = 0, e = 4; i != e; ++i) {
+      ShuffleMask.push_back(l + (NewImm & 3));
+      NewImm >>= 2;
+    }
+    for (unsigned i = 4, e = 8; i != e; ++i) {
+      ShuffleMask.push_back(l + i);
+    }
   }
-  ShuffleMask.push_back(4);
-  ShuffleMask.push_back(5);
-  ShuffleMask.push_back(6);
-  ShuffleMask.push_back(7);
 }
 
 /// DecodeSHUFPMask - This decodes the shuffle masks for shufp*. VT indicates
 /// the type of the vector allowing it to handle different datatypes and vector
 /// widths.
-void DecodeSHUFPMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
+void DecodeSHUFPMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   unsigned NumLanes = VT.getSizeInBits() / 128;
   unsigned NumLaneElts = NumElts / NumLanes;
 
-  int NewImm = Imm;
+  unsigned NewImm = Imm;
   for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
-    // Part that reads from dest.
-    for (unsigned i = 0; i != NumLaneElts/2; ++i) {
-      ShuffleMask.push_back(NewImm % NumLaneElts + l);
-      NewImm /= NumLaneElts;
-    }
-    // Part that reads from src.
-    for (unsigned i = 0; i != NumLaneElts/2; ++i) {
-      ShuffleMask.push_back(NewImm % NumLaneElts + NumElts + l);
-      NewImm /= NumLaneElts;
+    // each half of a lane comes from different source
+    for (unsigned s = 0; s != NumElts*2; s += NumElts) {
+      for (unsigned i = 0; i != NumLaneElts/2; ++i) {
+        ShuffleMask.push_back(NewImm % NumLaneElts + s + l);
+        NewImm /= NumLaneElts;
+      }
     }
     if (NumLaneElts == 4) NewImm = Imm; // reload imm
   }
@@ -130,7 +137,7 @@ void DecodeSHUFPMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
 /// DecodeUNPCKHMask - This decodes the shuffle masks for unpckhps/unpckhpd
 /// and punpckh*. VT indicates the type of the vector allowing it to handle
 /// different datatypes and vector widths.
-void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<int> &ShuffleMask) {
+void DecodeUNPCKHMask(MVT VT, SmallVectorImpl<int> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate
@@ -150,7 +157,7 @@ void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<int> &ShuffleMask) {
 /// DecodeUNPCKLMask - This decodes the shuffle masks for unpcklps/unpcklpd
 /// and punpckl*. VT indicates the type of the vector allowing it to handle
 /// different datatypes and vector widths.
-void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<int> &ShuffleMask) {
+void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask) {
   unsigned NumElts = VT.getVectorNumElements();
 
   // Handle 128 and 256-bit vector lengths. AVX defines UNPCK* to operate
@@ -167,19 +174,26 @@ void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<int> &ShuffleMask) {
   }
 }
 
-void DecodeVPERM2X128Mask(EVT VT, unsigned Imm,
+void DecodeVPERM2X128Mask(MVT VT, unsigned Imm,
                           SmallVectorImpl<int> &ShuffleMask) {
   if (Imm & 0x88)
     return; // Not a shuffle
 
   unsigned HalfSize = VT.getVectorNumElements()/2;
-  unsigned FstHalfBegin = (Imm & 0x3) * HalfSize;
-  unsigned SndHalfBegin = ((Imm >> 4) & 0x3) * HalfSize;
 
-  for (int i = FstHalfBegin, e = FstHalfBegin+HalfSize; i != e; ++i)
-    ShuffleMask.push_back(i);
-  for (int i = SndHalfBegin, e = SndHalfBegin+HalfSize; i != e; ++i)
-    ShuffleMask.push_back(i);
+  for (unsigned l = 0; l != 2; ++l) {
+    unsigned HalfBegin = ((Imm >> (l*4)) & 0x3) * HalfSize;
+    for (unsigned i = HalfBegin, e = HalfBegin+HalfSize; i != e; ++i)
+      ShuffleMask.push_back(i);
+  }
+}
+
+/// DecodeVPERMMask - this decodes the shuffle masks for VPERMQ/VPERMPD.
+/// No VT provided since it only works on 256-bit, 4 element vectors.
+void DecodeVPERMMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask) {
+  for (unsigned i = 0; i != 4; ++i) {
+    ShuffleMask.push_back((Imm >> (2*i)) & 3);
+  }
 }
 
 } // llvm namespace
diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.h b/lib/Target/X86/Utils/X86ShuffleDecode.h
index 5b8c6ef..70d8171 100644
--- a/lib/Target/X86/Utils/X86ShuffleDecode.h
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.h
@@ -35,31 +35,35 @@ void DecodeMOVHLPSMask(unsigned NElts, SmallVectorImpl<int> &ShuffleMask);
 // <0,2> or <0,1,4,5>
 void DecodeMOVLHPSMask(unsigned NElts, SmallVectorImpl<int> &ShuffleMask);
 
-void DecodePSHUFMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
+void DecodePSHUFMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
 
-void DecodePSHUFHWMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
+void DecodePSHUFHWMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
 
-void DecodePSHUFLWMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
+void DecodePSHUFLWMask(MVT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
 
 /// DecodeSHUFPMask - This decodes the shuffle masks for shufp*. VT indicates
 /// the type of the vector allowing it to handle different datatypes and vector
 /// widths.
-void DecodeSHUFPMask(EVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
+void DecodeSHUFPMask(MVT VT, unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
 
 /// DecodeUNPCKHMask - This decodes the shuffle masks for unpckhps/unpckhpd
 /// and punpckh*. VT indicates the type of the vector allowing it to handle
 /// different datatypes and vector widths.
-void DecodeUNPCKHMask(EVT VT, SmallVectorImpl<int> &ShuffleMask);
+void DecodeUNPCKHMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
 
 /// DecodeUNPCKLMask - This decodes the shuffle masks for unpcklps/unpcklpd
 /// and punpckl*. VT indicates the type of the vector allowing it to handle
 /// different datatypes and vector widths.
-void DecodeUNPCKLMask(EVT VT, SmallVectorImpl<int> &ShuffleMask);
+void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
 
 
-void DecodeVPERM2X128Mask(EVT VT, unsigned Imm,
+void DecodeVPERM2X128Mask(MVT VT, unsigned Imm,
                           SmallVectorImpl<int> &ShuffleMask);
 
+/// DecodeVPERMMask - this decodes the shuffle masks for VPERMQ/VPERMPD.
+/// No VT provided since it only works on 256-bit, 4 element vectors.
+void DecodeVPERMMask(unsigned Imm, SmallVectorImpl<int> &ShuffleMask);
+
 } // llvm namespace
 
 #endif