am 31675153: Merge branch \'upstream\' into merge_2

* commit '31675153bd2d7617db8cb6aeb58054934c7b9f73': (542 commits)
  MaximumSpanningTree::EdgeWeightCompare: Make this comparator actually be a strict weak ordering, and don't pass possibly-null pointers to dyn_cast.
  Fix misaligned access in MachO object file reader: despite containing an int64_t, Symbol64TableEntry is actually only stored with 4-byte alignment within the file.
  Fix unaligned memory accesses when performing relocations in X86 JIT. There's no cost to using memcpy here: the fixed code is optimized by LLVM to perfect machine code.
  Don't pass a null pointer to cast<> in its unit tests.
  Don't bind a reference to a dereferenced null pointer (for return value of WeakVH::operator*).
  [ms-inline asm] Do not report a Parser error when matching inline assembly.
  Ignore the documentation-suggested location for compile_commands.json
  The presence of the empty file "foo" unfortunately does not improve LLVM in any way.
  Remove unnecessary cast that was also unnecessarily casting away constness.
  Provide a portability macro for __builtin_trap.
  Fix macros arguments with an underscore, dot or dollar in them. This is based on a patch by Andy/PaX. I added the support for dot and dollar.
  [ms-inline asm] Expose the ErrorInfo from the MatchInstructionImpl.  In general, this is the index of the operand that failed to match.
  Formatting.  No functional change.
  Make the wording in of the "expected identifier" error in the .macro directive consistent with the other "expected identifier" errors. Extracted from the Andy/PaX patch. I added the test.
  Pacify PVS-Studio by changing the type rather than doing a cast, a tweak suggested by David Blaikie.
  Add support for the --param ssp-buffer-size= driver option. PR9673
  Use typedefs. Fix indentation. Extracted from the Andy/PaX patch.
  Remove unused variable. Extracted from the Andy/PaX patch.
  Fix typo. Extracted from the Andy/PaX patch.
  MCJIT: Tidy up the constructor.
  ...

1 files changed, 48 insertions, 34 deletions

diff --git a/lib/Target/X86/X86FastISel.cpp b/lib/Target/X86/X86FastISel.cpp
index 585b7a5..e5952aa 100644
--- a/lib/Target/X86/X86FastISel.cpp
+++ b/lib/Target/X86/X86FastISel.cpp
@@ -57,7 +57,9 @@ class X86FastISel : public FastISel {
   bool X86ScalarSSEf32;
 
 public:
-  explicit X86FastISel(FunctionLoweringInfo &funcInfo) : FastISel(funcInfo) {
+  explicit X86FastISel(FunctionLoweringInfo &funcInfo,
+                       const TargetLibraryInfo *libInfo)
+    : FastISel(funcInfo, libInfo) {
     Subtarget = &TM.getSubtarget<X86Subtarget>();
     StackPtr = Subtarget->is64Bit() ? X86::RSP : X86::ESP;
     X86ScalarSSEf64 = Subtarget->hasSSE2();
@@ -155,9 +157,9 @@ bool X86FastISel::isTypeLegal(Type *Ty, MVT &VT, bool AllowI1) {
   // For now, require SSE/SSE2 for performing floating-point operations,
   // since x87 requires additional work.
   if (VT == MVT::f64 && !X86ScalarSSEf64)
-     return false;
+    return false;
   if (VT == MVT::f32 && !X86ScalarSSEf32)
-     return false;
+    return false;
   // Similarly, no f80 support yet.
   if (VT == MVT::f80)
     return false;
@@ -1516,6 +1518,22 @@ bool X86FastISel::X86SelectCall(const Instruction *I) {
   return DoSelectCall(I, 0);
 }
 
+static unsigned computeBytesPoppedByCallee(const X86Subtarget &Subtarget,
+                                           const ImmutableCallSite &CS) {
+  if (Subtarget.is64Bit())
+    return 0;
+  if (Subtarget.isTargetWindows())
+    return 0;
+  CallingConv::ID CC = CS.getCallingConv();
+  if (CC == CallingConv::Fast || CC == CallingConv::GHC)
+    return 0;
+  if (!CS.paramHasAttr(1, Attribute::StructRet))
+    return 0;
+  if (CS.paramHasAttr(1, Attribute::InReg))
+    return 0;
+  return 4;
+}
+
 // Select either a call, or an llvm.memcpy/memmove/memset intrinsic
 bool X86FastISel::DoSelectCall(const Instruction *I, const char *MemIntName) {
   const CallInst *CI = cast<CallInst>(I);
@@ -1862,12 +1880,7 @@ bool X86FastISel::DoSelectCall(const Instruction *I, const char *MemIntName) {
 
   // Issue CALLSEQ_END
   unsigned AdjStackUp = TII.getCallFrameDestroyOpcode();
-  unsigned NumBytesCallee = 0;
-  if (!Subtarget->is64Bit() && !Subtarget->isTargetWindows() &&
-      !(CS.getCallingConv() == CallingConv::Fast ||
-        CS.getCallingConv() == CallingConv::GHC) &&
-      CS.paramHasAttr(1, Attribute::StructRet))
-    NumBytesCallee = 4;
+  const unsigned NumBytesCallee = computeBytesPoppedByCallee(*Subtarget, CS);
   BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(AdjStackUp))
     .addImm(NumBytes).addImm(NumBytesCallee);
 
@@ -2129,28 +2142,28 @@ unsigned X86FastISel::TargetMaterializeFloatZero(const ConstantFP *CF) {
   unsigned Opc = 0;
   const TargetRegisterClass *RC = NULL;
   switch (VT.SimpleTy) {
-    default: return false;
-    case MVT::f32:
-      if (X86ScalarSSEf32) {
-        Opc = X86::FsFLD0SS;
-        RC  = &X86::FR32RegClass;
-      } else {
-        Opc = X86::LD_Fp032;
-        RC  = &X86::RFP32RegClass;
-      }
-      break;
-    case MVT::f64:
-      if (X86ScalarSSEf64) {
-        Opc = X86::FsFLD0SD;
-        RC  = &X86::FR64RegClass;
-      } else {
-        Opc = X86::LD_Fp064;
-        RC  = &X86::RFP64RegClass;
-      }
-      break;
-    case MVT::f80:
-      // No f80 support yet.
-      return false;
+  default: return false;
+  case MVT::f32:
+    if (X86ScalarSSEf32) {
+      Opc = X86::FsFLD0SS;
+      RC  = &X86::FR32RegClass;
+    } else {
+      Opc = X86::LD_Fp032;
+      RC  = &X86::RFP32RegClass;
+    }
+    break;
+  case MVT::f64:
+    if (X86ScalarSSEf64) {
+      Opc = X86::FsFLD0SD;
+      RC  = &X86::FR64RegClass;
+    } else {
+      Opc = X86::LD_Fp064;
+      RC  = &X86::RFP64RegClass;
+    }
+    break;
+  case MVT::f80:
+    // No f80 support yet.
+    return false;
   }
 
   unsigned ResultReg = createResultReg(RC);
@@ -2169,7 +2182,7 @@ bool X86FastISel::TryToFoldLoad(MachineInstr *MI, unsigned OpNo,
   if (!X86SelectAddress(LI->getOperand(0), AM))
     return false;
 
-  X86InstrInfo &XII = (X86InstrInfo&)TII;
+  const X86InstrInfo &XII = (const X86InstrInfo&)TII;
 
   unsigned Size = TD.getTypeAllocSize(LI->getType());
   unsigned Alignment = LI->getAlignment();
@@ -2188,7 +2201,8 @@ bool X86FastISel::TryToFoldLoad(MachineInstr *MI, unsigned OpNo,
 
 
 namespace llvm {
-  FastISel *X86::createFastISel(FunctionLoweringInfo &funcInfo) {
-    return new X86FastISel(funcInfo);
+  FastISel *X86::createFastISel(FunctionLoweringInfo &funcInfo,
+                                const TargetLibraryInfo *libInfo) {
+    return new X86FastISel(funcInfo, libInfo);
   }
 }