Merge commit 'a662a9862501fc86904e90054f7c1519101d9126'

Conflicts:
	include/llvm/CodeGen/IntrinsicLowering.h
	include/llvm/MC/MCAssembler.h
	include/llvm/MC/MCObjectStreamer.h
	lib/LLVMBuild.txt
	lib/Linker/LinkArchives.cpp
	lib/MC/MCAssembler.cpp
	lib/MC/MCELFStreamer.cpp
	lib/MC/MCParser/AsmParser.cpp
	lib/MC/MCPureStreamer.cpp
	lib/MC/WinCOFFStreamer.cpp
	lib/Makefile
	lib/Support/Unix/Memory.inc
	lib/Support/Unix/Process.inc
	lib/Support/Unix/Program.inc
	lib/Target/ARM/ARM.h
	lib/Target/ARM/ARMFastISel.cpp
	lib/Target/ARM/ARMISelLowering.cpp
	lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
	lib/Target/Mips/MipsInstrFPU.td
	lib/Target/X86/CMakeLists.txt
	lib/Target/X86/X86ISelLowering.cpp
	lib/Target/X86/X86TargetMachine.cpp
	lib/Target/X86/X86TargetObjectFile.cpp
	lib/Transforms/InstCombine/InstCombineCalls.cpp
	test/CodeGen/X86/fast-isel-x86-64.ll
	tools/llc/llc.cpp
	tools/lto/LTOModule.cpp
	utils/TableGen/EDEmitter.cpp

1 files changed, 61 insertions, 45 deletions

diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 29339405aa..344d1447a8 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -19,21 +19,22 @@
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringExtras.h"
+#include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/Assembly/Writer.h"
-#include "llvm/CallingConv.h"
 #include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/CodeGen/MachineConstantPool.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/MachineModuleInfo.h"
-#include "llvm/Constants.h"
-#include "llvm/DataLayout.h"
 #include "llvm/DebugInfo.h"
-#include "llvm/DerivedTypes.h"
-#include "llvm/Function.h"
-#include "llvm/GlobalAlias.h"
-#include "llvm/GlobalVariable.h"
-#include "llvm/Intrinsics.h"
+#include "llvm/IR/CallingConv.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/GlobalAlias.h"
+#include "llvm/IR/GlobalVariable.h"
+#include "llvm/IR/Intrinsics.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -884,15 +885,17 @@ unsigned SelectionDAG::getEVTAlignment(EVT VT) const {
 // EntryNode could meaningfully have debug info if we can find it...
 SelectionDAG::SelectionDAG(const TargetMachine &tm, CodeGenOpt::Level OL)
   : TM(tm), TLI(*tm.getTargetLowering()), TSI(*tm.getSelectionDAGInfo()),
-    OptLevel(OL), EntryNode(ISD::EntryToken, DebugLoc(), getVTList(MVT::Other)),
+    TTI(0), OptLevel(OL), EntryNode(ISD::EntryToken, DebugLoc(),
+                                    getVTList(MVT::Other)),
     Root(getEntryNode()), Ordering(0), UpdateListeners(0) {
   AllNodes.push_back(&EntryNode);
   Ordering = new SDNodeOrdering();
   DbgInfo = new SDDbgInfo();
 }
 
-void SelectionDAG::init(MachineFunction &mf) {
+void SelectionDAG::init(MachineFunction &mf, const TargetTransformInfo *tti) {
   MF = &mf;
+  TTI = tti;
   Context = &mf.getFunction()->getContext();
 }
 
@@ -1074,7 +1077,8 @@ SDValue SelectionDAG::getConstantFP(double Val, EVT VT, bool isTarget) {
     return getConstantFP(APFloat((float)Val), VT, isTarget);
   else if (EltVT==MVT::f64)
     return getConstantFP(APFloat(Val), VT, isTarget);
-  else if (EltVT==MVT::f80 || EltVT==MVT::f128 || EltVT==MVT::f16) {
+  else if (EltVT==MVT::f80 || EltVT==MVT::f128 || EltVT==MVT::ppcf128 ||
+           EltVT==MVT::f16) {
     bool ignored;
     APFloat apf = APFloat(Val);
     apf.convert(*EVTToAPFloatSemantics(EltVT), APFloat::rmNearestTiesToEven,
@@ -3370,10 +3374,11 @@ static SDValue getMemsetStringVal(EVT VT, DebugLoc dl, SelectionDAG &DAG,
   }
 
   assert(!VT.isVector() && "Can't handle vector type here!");
-  unsigned NumVTBytes = VT.getSizeInBits() / 8;
+  unsigned NumVTBits = VT.getSizeInBits();
+  unsigned NumVTBytes = NumVTBits / 8;
   unsigned NumBytes = std::min(NumVTBytes, unsigned(Str.size()));
 
-  APInt Val(NumBytes*8, 0);
+  APInt Val(NumVTBits, 0);
   if (TLI.isLittleEndian()) {
     for (unsigned i = 0; i != NumBytes; ++i)
       Val |= (uint64_t)(unsigned char)Str[i] << i*8;
@@ -3382,7 +3387,10 @@ static SDValue getMemsetStringVal(EVT VT, DebugLoc dl, SelectionDAG &DAG,
       Val |= (uint64_t)(unsigned char)Str[i] << (NumVTBytes-i-1)*8;
   }
 
-  if (TLI.isIntImmLegal(Val, VT))
+  // If the "cost" of materializing the integer immediate is 1 or free, then
+  // it is cost effective to turn the load into the immediate.
+  const TargetTransformInfo *TTI = DAG.getTargetTransformInfo();
+  if (TTI->getIntImmCost(Val, VT.getTypeForEVT(*DAG.getContext())) < 2)
     return DAG.getConstant(Val, VT);
   return SDValue(0, 0);
 }
@@ -3422,7 +3430,8 @@ static bool isMemSrcFromString(SDValue Src, StringRef &Str) {
 static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps,
                                      unsigned Limit, uint64_t Size,
                                      unsigned DstAlign, unsigned SrcAlign,
-                                     bool IsZeroVal,
+                                     bool IsMemset,
+                                     bool ZeroMemset,
                                      bool MemcpyStrSrc,
                                      bool AllowOverlap,
                                      SelectionDAG &DAG,
@@ -3437,7 +3446,7 @@ static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps,
   // 'MemcpyStrSrc' indicates whether the memcpy source is constant so it does
   // not need to be loaded.
   EVT VT = TLI.getOptimalMemOpType(Size, DstAlign, SrcAlign,
-                                   IsZeroVal, MemcpyStrSrc,
+                                   IsMemset, ZeroMemset, MemcpyStrSrc,
                                    DAG.getMachineFunction());
 
   if (VT == MVT::Other) {
@@ -3464,39 +3473,43 @@ static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps,
 
   unsigned NumMemOps = 0;
   while (Size != 0) {
-    if (++NumMemOps > Limit)
-      return false;
-
     unsigned VTSize = VT.getSizeInBits() / 8;
     while (VTSize > Size) {
       // For now, only use non-vector load / store's for the left-over pieces.
-      EVT NewVT;
+      EVT NewVT = VT;
       unsigned NewVTSize;
+
+      bool Found = false;
       if (VT.isVector() || VT.isFloatingPoint()) {
         NewVT = (VT.getSizeInBits() > 64) ? MVT::i64 : MVT::i32;
-        while (!TLI.isOperationLegalOrCustom(ISD::STORE, NewVT)) {
-          if (NewVT == MVT::i64 &&
-              TLI.isOperationLegalOrCustom(ISD::STORE, MVT::f64)) {
-            // i64 is usually not legal on 32-bit targets, but f64 may be.
-            NewVT = MVT::f64;
-            break;
-          }
-          NewVT = (MVT::SimpleValueType)(NewVT.getSimpleVT().SimpleTy - 1);
+        if (TLI.isOperationLegalOrCustom(ISD::STORE, NewVT) &&
+            TLI.isSafeMemOpType(NewVT.getSimpleVT()))
+          Found = true;
+        else if (NewVT == MVT::i64 &&
+                 TLI.isOperationLegalOrCustom(ISD::STORE, MVT::f64) &&
+                 TLI.isSafeMemOpType(MVT::f64)) {
+          // i64 is usually not legal on 32-bit targets, but f64 may be.
+          NewVT = MVT::f64;
+          Found = true;
         }
-        NewVTSize = NewVT.getSizeInBits() / 8;
-      } else {
-        // This can result in a type that is not legal on the target, e.g.
-        // 1 or 2 bytes on PPC.
-        NewVT = (MVT::SimpleValueType)(VT.getSimpleVT().SimpleTy - 1);
-        NewVTSize = VTSize >> 1;
       }
 
+      if (!Found) {
+        do {
+          NewVT = (MVT::SimpleValueType)(NewVT.getSimpleVT().SimpleTy - 1);
+          if (NewVT == MVT::i8)
+            break;
+        } while (!TLI.isSafeMemOpType(NewVT.getSimpleVT()));
+      }
+      NewVTSize = NewVT.getSizeInBits() / 8;
+
       // If the new VT cannot cover all of the remaining bits, then consider
       // issuing a (or a pair of) unaligned and overlapping load / store.
       // FIXME: Only does this for 64-bit or more since we don't have proper
       // cost model for unaligned load / store.
       bool Fast;
-      if (AllowOverlap && VTSize >= 8 && NewVTSize < Size &&
+      if (NumMemOps && AllowOverlap &&
+          VTSize >= 8 && NewVTSize < Size &&
           TLI.allowsUnalignedMemoryAccesses(VT, &Fast) && Fast)
         VTSize = Size;
       else {
@@ -3505,6 +3518,9 @@ static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps,
       }
     }
 
+    if (++NumMemOps > Limit)
+      return false;
+
     MemOps.push_back(VT);
     Size -= VTSize;
   }
@@ -3533,8 +3549,8 @@ static SDValue getMemcpyLoadsAndStores(SelectionDAG &DAG, DebugLoc dl,
   MachineFunction &MF = DAG.getMachineFunction();
   MachineFrameInfo *MFI = MF.getFrameInfo();
   bool OptSize =
-    MF.getFunction()->getFnAttributes().
-      hasAttribute(Attributes::OptimizeForSize);
+    MF.getFunction()->getAttributes().
+      hasAttribute(AttributeSet::FunctionIndex, Attribute::OptimizeForSize);
   FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Dst);
   if (FI && !MFI->isFixedObjectIndex(FI->getIndex()))
     DstAlignCanChange = true;
@@ -3549,7 +3565,7 @@ static SDValue getMemcpyLoadsAndStores(SelectionDAG &DAG, DebugLoc dl,
   if (!FindOptimalMemOpLowering(MemOps, Limit, Size,
                                 (DstAlignCanChange ? 0 : Align),
                                 (isZeroStr ? 0 : SrcAlign),
-                                true, CopyFromStr, true, DAG, TLI))
+                                false, false, CopyFromStr, true, DAG, TLI))
     return SDValue();
 
   if (DstAlignCanChange) {
@@ -3639,8 +3655,8 @@ static SDValue getMemmoveLoadsAndStores(SelectionDAG &DAG, DebugLoc dl,
   bool DstAlignCanChange = false;
   MachineFunction &MF = DAG.getMachineFunction();
   MachineFrameInfo *MFI = MF.getFrameInfo();
-  bool OptSize = MF.getFunction()->getFnAttributes().
-    hasAttribute(Attributes::OptimizeForSize);
+  bool OptSize = MF.getFunction()->getAttributes().
+    hasAttribute(AttributeSet::FunctionIndex, Attribute::OptimizeForSize);
   FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Dst);
   if (FI && !MFI->isFixedObjectIndex(FI->getIndex()))
     DstAlignCanChange = true;
@@ -3650,8 +3666,8 @@ static SDValue getMemmoveLoadsAndStores(SelectionDAG &DAG, DebugLoc dl,
   unsigned Limit = AlwaysInline ? ~0U : TLI.getMaxStoresPerMemmove(OptSize);
 
   if (!FindOptimalMemOpLowering(MemOps, Limit, Size,
-                                (DstAlignCanChange ? 0 : Align),
-                                SrcAlign, true, false, false, DAG, TLI))
+                                (DstAlignCanChange ? 0 : Align), SrcAlign,
+                                false, false, false, false, DAG, TLI))
     return SDValue();
 
   if (DstAlignCanChange) {
@@ -3718,8 +3734,8 @@ static SDValue getMemsetStores(SelectionDAG &DAG, DebugLoc dl,
   bool DstAlignCanChange = false;
   MachineFunction &MF = DAG.getMachineFunction();
   MachineFrameInfo *MFI = MF.getFrameInfo();
-  bool OptSize = MF.getFunction()->getFnAttributes().
-    hasAttribute(Attributes::OptimizeForSize);
+  bool OptSize = MF.getFunction()->getAttributes().
+    hasAttribute(AttributeSet::FunctionIndex, Attribute::OptimizeForSize);
   FrameIndexSDNode *FI = dyn_cast<FrameIndexSDNode>(Dst);
   if (FI && !MFI->isFixedObjectIndex(FI->getIndex()))
     DstAlignCanChange = true;
@@ -3727,7 +3743,7 @@ static SDValue getMemsetStores(SelectionDAG &DAG, DebugLoc dl,
     isa<ConstantSDNode>(Src) && cast<ConstantSDNode>(Src)->isNullValue();
   if (!FindOptimalMemOpLowering(MemOps, TLI.getMaxStoresPerMemset(OptSize),
                                 Size, (DstAlignCanChange ? 0 : Align), 0,
-                                IsZeroVal, false, true, DAG, TLI))
+                                true, IsZeroVal, false, true, DAG, TLI))
     return SDValue();
 
   if (DstAlignCanChange) {