Merge branch 'upstream' into merge_2

Conflicts:
	lib/Target/ARM/ARMCodeEmitter.cpp

Change-Id: I6702d340c733e9721499b5d85b13b96ad9c14eb5

11 files changed, 283 insertions, 146 deletions

diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 747bc44..1c485a0 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -228,6 +228,9 @@ namespace {
     SDValue visitFP_EXTEND(SDNode *N);
     SDValue visitFNEG(SDNode *N);
     SDValue visitFABS(SDNode *N);
+    SDValue visitFCEIL(SDNode *N);
+    SDValue visitFTRUNC(SDNode *N);
+    SDValue visitFFLOOR(SDNode *N);
     SDValue visitBRCOND(SDNode *N);
     SDValue visitBR_CC(SDNode *N);
     SDValue visitLOAD(SDNode *N);
@@ -1140,6 +1143,9 @@ SDValue DAGCombiner::visit(SDNode *N) {
   case ISD::FP_EXTEND:          return visitFP_EXTEND(N);
   case ISD::FNEG:               return visitFNEG(N);
   case ISD::FABS:               return visitFABS(N);
+  case ISD::FFLOOR:             return visitFFLOOR(N);
+  case ISD::FCEIL:              return visitFCEIL(N);
+  case ISD::FTRUNC:             return visitFTRUNC(N);
   case ISD::BRCOND:             return visitBRCOND(N);
   case ISD::BR_CC:              return visitBR_CC(N);
   case ISD::LOAD:               return visitLOAD(N);
@@ -5679,7 +5685,7 @@ SDValue DAGCombiner::visitFADD(SDNode *N) {
   if ((DAG.getTarget().Options.AllowFPOpFusion == FPOpFusion::Fast ||
        DAG.getTarget().Options.UnsafeFPMath) &&
       DAG.getTarget().getTargetLowering()->isFMAFasterThanMulAndAdd(VT) &&
-      TLI.isOperationLegal(ISD::FMA, VT)) {
+      TLI.isOperationLegalOrCustom(ISD::FMA, VT)) {
 
     // fold (fadd (fmul x, y), z) -> (fma x, y, z)
     if (N0.getOpcode() == ISD::FMUL && N0->hasOneUse()) {
@@ -5704,6 +5710,7 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
   ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
   ConstantFPSDNode *N1CFP = dyn_cast<ConstantFPSDNode>(N1);
   EVT VT = N->getValueType(0);
+  DebugLoc dl = N->getDebugLoc();
 
   // fold vector ops
   if (VT.isVector()) {
@@ -5724,11 +5731,11 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
     if (isNegatibleForFree(N1, LegalOperations, TLI, &DAG.getTarget().Options))
       return GetNegatedExpression(N1, DAG, LegalOperations);
     if (!LegalOperations || TLI.isOperationLegal(ISD::FNEG, VT))
-      return DAG.getNode(ISD::FNEG, N->getDebugLoc(), VT, N1);
+      return DAG.getNode(ISD::FNEG, dl, VT, N1);
   }
   // fold (fsub A, (fneg B)) -> (fadd A, B)
   if (isNegatibleForFree(N1, LegalOperations, TLI, &DAG.getTarget().Options))
-    return DAG.getNode(ISD::FADD, N->getDebugLoc(), VT, N0,
+    return DAG.getNode(ISD::FADD, dl, VT, N0,
                        GetNegatedExpression(N1, DAG, LegalOperations));
 
   // If 'unsafe math' is enabled, fold
@@ -5756,23 +5763,34 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
   if ((DAG.getTarget().Options.AllowFPOpFusion == FPOpFusion::Fast ||
        DAG.getTarget().Options.UnsafeFPMath) &&
       DAG.getTarget().getTargetLowering()->isFMAFasterThanMulAndAdd(VT) &&
-      TLI.isOperationLegal(ISD::FMA, VT)) {
+      TLI.isOperationLegalOrCustom(ISD::FMA, VT)) {
 
     // fold (fsub (fmul x, y), z) -> (fma x, y, (fneg z))
     if (N0.getOpcode() == ISD::FMUL && N0->hasOneUse()) {
-      return DAG.getNode(ISD::FMA, N->getDebugLoc(), VT,
+      return DAG.getNode(ISD::FMA, dl, VT,
                          N0.getOperand(0), N0.getOperand(1),
-                         DAG.getNode(ISD::FNEG, N1->getDebugLoc(), VT, N1));
+                         DAG.getNode(ISD::FNEG, dl, VT, N1));
     }
 
     // fold (fsub x, (fmul y, z)) -> (fma (fneg y), z, x)
     // Note: Commutes FSUB operands.
     if (N1.getOpcode() == ISD::FMUL && N1->hasOneUse()) {
-      return DAG.getNode(ISD::FMA, N->getDebugLoc(), VT,
-                         DAG.getNode(ISD::FNEG, N1->getDebugLoc(), VT,
+      return DAG.getNode(ISD::FMA, dl, VT,
+                         DAG.getNode(ISD::FNEG, dl, VT,
                          N1.getOperand(0)),
                          N1.getOperand(1), N0);
     }
+
+    // fold (fsub (-(fmul, x, y)), z) -> (fma (fneg x), y, (fneg z))
+    if (N0.getOpcode() == ISD::FNEG && 
+        N0.getOperand(0).getOpcode() == ISD::FMUL &&
+        N0->hasOneUse() && N0.getOperand(0).hasOneUse()) {
+      SDValue N00 = N0.getOperand(0).getOperand(0);
+      SDValue N01 = N0.getOperand(0).getOperand(1);
+      return DAG.getNode(ISD::FMA, dl, VT,
+                         DAG.getNode(ISD::FNEG, dl, VT, N00), N01,
+                         DAG.getNode(ISD::FNEG, dl, VT, N1));
+    }
   }
 
   return SDValue();
@@ -6231,6 +6249,42 @@ SDValue DAGCombiner::visitFNEG(SDNode *N) {
   return SDValue();
 }
 
+SDValue DAGCombiner::visitFCEIL(SDNode *N) {
+  SDValue N0 = N->getOperand(0);
+  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
+  EVT VT = N->getValueType(0);
+
+  // fold (fceil c1) -> fceil(c1)
+  if (N0CFP && VT != MVT::ppcf128)
+    return DAG.getNode(ISD::FCEIL, N->getDebugLoc(), VT, N0);
+
+  return SDValue();
+}
+
+SDValue DAGCombiner::visitFTRUNC(SDNode *N) {
+  SDValue N0 = N->getOperand(0);
+  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
+  EVT VT = N->getValueType(0);
+
+  // fold (ftrunc c1) -> ftrunc(c1)
+  if (N0CFP && VT != MVT::ppcf128)
+    return DAG.getNode(ISD::FTRUNC, N->getDebugLoc(), VT, N0);
+
+  return SDValue();
+}
+
+SDValue DAGCombiner::visitFFLOOR(SDNode *N) {
+  SDValue N0 = N->getOperand(0);
+  ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
+  EVT VT = N->getValueType(0);
+
+  // fold (ffloor c1) -> ffloor(c1)
+  if (N0CFP && VT != MVT::ppcf128)
+    return DAG.getNode(ISD::FFLOOR, N->getDebugLoc(), VT, N0);
+
+  return SDValue();
+}
+
 SDValue DAGCombiner::visitFABS(SDNode *N) {
   SDValue N0 = N->getOperand(0);
   ConstantFPSDNode *N0CFP = dyn_cast<ConstantFPSDNode>(N0);
@@ -7822,9 +7876,29 @@ SDValue DAGCombiner::visitBUILD_VECTOR(SDNode *N) {
       if (VecIn1.getValueType().getSizeInBits()*2 != VT.getSizeInBits())
         return SDValue();
 
-      // Widen the input vector by adding undef values.
-      VecIn1 = DAG.getNode(ISD::CONCAT_VECTORS, N->getDebugLoc(), VT,
-                           VecIn1, DAG.getUNDEF(VecIn1.getValueType()));
+      // If the element type of the input vector is not the same as
+      // the output element type, make concat_vectors based on input element
+      // type and then bitcast it to the output vector type.
+      //
+      // In another words avoid nodes like this:
+      //  <NODE> v16i8 = concat_vectors v4i16 v4i16
+      // Replace it with this one:
+      //  <NODE0> v8i16 = concat_vectors v4i16 v4i16
+      //  <NODE1> v16i8 = bitcast NODE0
+      EVT ItemType = VecIn1.getValueType().getVectorElementType();
+      if (ItemType != VT.getVectorElementType()) {
+        EVT ConcatVT = EVT::getVectorVT(*DAG.getContext(),
+                                ItemType,
+                                VecIn1.getValueType().getVectorNumElements()*2);
+        // Widen the input vector by adding undef values.
+        VecIn1 = DAG.getNode(ISD::CONCAT_VECTORS, dl, ConcatVT,
+                             VecIn1, DAG.getUNDEF(VecIn1.getValueType()));
+        VecIn1 = DAG.getNode(ISD::BITCAST, dl, VT, VecIn1);
+      } else
+        // Widen the input vector by adding undef values.
+        VecIn1 = DAG.getNode(ISD::CONCAT_VECTORS, dl, VT,
+                             VecIn1, DAG.getUNDEF(VecIn1.getValueType()));
+
     }
 
     // If VecIn2 is unused then change it to undef.
diff --git a/lib/CodeGen/SelectionDAG/FastISel.cpp b/lib/CodeGen/SelectionDAG/FastISel.cpp
index e5ea6e6..683fac6 100644
--- a/lib/CodeGen/SelectionDAG/FastISel.cpp
+++ b/lib/CodeGen/SelectionDAG/FastISel.cpp
@@ -55,6 +55,7 @@
 #include "llvm/Analysis/Loads.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetInstrInfo.h"
+#include "llvm/Target/TargetLibraryInfo.h"
 #include "llvm/Target/TargetLowering.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -789,6 +790,17 @@ FastISel::SelectInstruction(const Instruction *I) {
 
   MachineBasicBlock::iterator SavedInsertPt = FuncInfo.InsertPt;
 
+  // As a special case, don't handle calls to builtin library functions that
+  // may be translated directly to target instructions.
+  if (const CallInst *Call = dyn_cast<CallInst>(I)) {
+    const Function *F = Call->getCalledFunction();
+    LibFunc::Func Func;
+    if (F && !F->hasLocalLinkage() && F->hasName() &&
+        LibInfo->getLibFunc(F->getName(), Func) &&
+        LibInfo->hasOptimizedCodeGen(Func))
+      return false;
+  }
+
   // First, try doing target-independent selection.
   if (SelectOperator(I, I->getOpcode())) {
     ++NumFastIselSuccessIndependent;
@@ -1040,7 +1052,8 @@ FastISel::SelectOperator(const User *I, unsigned Opcode) {
   }
 }
 
-FastISel::FastISel(FunctionLoweringInfo &funcInfo)
+FastISel::FastISel(FunctionLoweringInfo &funcInfo,
+                   const TargetLibraryInfo *libInfo)
   : FuncInfo(funcInfo),
     MRI(FuncInfo.MF->getRegInfo()),
     MFI(*FuncInfo.MF->getFrameInfo()),
@@ -1049,7 +1062,8 @@ FastISel::FastISel(FunctionLoweringInfo &funcInfo)
     TD(*TM.getTargetData()),
     TII(*TM.getInstrInfo()),
     TLI(*TM.getTargetLowering()),
-    TRI(*TM.getRegisterInfo()) {
+    TRI(*TM.getRegisterInfo()),
+    LibInfo(libInfo) {
 }
 
 FastISel::~FastISel() {}
diff --git a/lib/CodeGen/SelectionDAG/InstrEmitter.cpp b/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
index 936c126..4488d27 100644
--- a/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
+++ b/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
@@ -411,6 +411,10 @@ void InstrEmitter::AddOperand(MachineInstr *MI, SDValue Op,
   } else if (BlockAddressSDNode *BA = dyn_cast<BlockAddressSDNode>(Op)) {
     MI->addOperand(MachineOperand::CreateBA(BA->getBlockAddress(),
                                             BA->getTargetFlags()));
+  } else if (TargetIndexSDNode *TI = dyn_cast<TargetIndexSDNode>(Op)) {
+    MI->addOperand(MachineOperand::CreateTargetIndex(TI->getIndex(),
+                                                     TI->getOffset(),
+                                                     TI->getTargetFlags()));
   } else {
     assert(Op.getValueType() != MVT::Other &&
            Op.getValueType() != MVT::Glue &&
diff --git a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index b0776af..908ebb9 100644
--- a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -428,7 +428,7 @@ ExpandUnalignedLoad(LoadSDNode *LD, SelectionDAG &DAG,
   DebugLoc dl = LD->getDebugLoc();
   if (VT.isFloatingPoint() || VT.isVector()) {
     EVT intVT = EVT::getIntegerVT(*DAG.getContext(), LoadedVT.getSizeInBits());
-    if (TLI.isTypeLegal(intVT)) {
+    if (TLI.isTypeLegal(intVT) && TLI.isTypeLegal(LoadedVT)) {
       // Expand to a (misaligned) integer load of the same size,
       // then bitconvert to floating point or vector.
       SDValue newLoad = DAG.getLoad(intVT, dl, Chain, Ptr, LD->getPointerInfo(),
@@ -436,8 +436,9 @@ ExpandUnalignedLoad(LoadSDNode *LD, SelectionDAG &DAG,
                                     LD->isNonTemporal(),
                                     LD->isInvariant(), LD->getAlignment());
       SDValue Result = DAG.getNode(ISD::BITCAST, dl, LoadedVT, newLoad);
-      if (VT.isFloatingPoint() && LoadedVT != VT)
-        Result = DAG.getNode(ISD::FP_EXTEND, dl, VT, Result);
+      if (LoadedVT != VT)
+        Result = DAG.getNode(VT.isFloatingPoint() ? ISD::FP_EXTEND :
+                             ISD::ANY_EXTEND, dl, VT, Result);
 
       ValResult = Result;
       ChainResult = Chain;
diff --git a/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.h b/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.h
index 5384576..84e41fc 100644
--- a/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.h
+++ b/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.h
@@ -61,6 +61,7 @@ namespace llvm {
       if (isa<BasicBlockSDNode>(Node))     return true;
       if (isa<FrameIndexSDNode>(Node))     return true;
       if (isa<ConstantPoolSDNode>(Node))   return true;
+      if (isa<TargetIndexSDNode>(Node))    return true;
       if (isa<JumpTableSDNode>(Node))      return true;
       if (isa<ExternalSymbolSDNode>(Node)) return true;
       if (isa<BlockAddressSDNode>(Node))   return true;
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index b971b69..f4fe892 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -403,6 +403,7 @@ static void AddNodeIDCustom(FoldingSetNodeID &ID, const SDNode *N) {
     ID.AddPointer(GA->getGlobal());
     ID.AddInteger(GA->getOffset());
     ID.AddInteger(GA->getTargetFlags());
+    ID.AddInteger(GA->getAddressSpace());
     break;
   }
   case ISD::BasicBlock:
@@ -438,16 +439,25 @@ static void AddNodeIDCustom(FoldingSetNodeID &ID, const SDNode *N) {
     ID.AddInteger(CP->getTargetFlags());
     break;
   }
+  case ISD::TargetIndex: {
+    const TargetIndexSDNode *TI = cast<TargetIndexSDNode>(N);
+    ID.AddInteger(TI->getIndex());
+    ID.AddInteger(TI->getOffset());
+    ID.AddInteger(TI->getTargetFlags());
+    break;
+  }
   case ISD::LOAD: {
     const LoadSDNode *LD = cast<LoadSDNode>(N);
     ID.AddInteger(LD->getMemoryVT().getRawBits());
     ID.AddInteger(LD->getRawSubclassData());
+    ID.AddInteger(LD->getPointerInfo().getAddrSpace());
     break;
   }
   case ISD::STORE: {
     const StoreSDNode *ST = cast<StoreSDNode>(N);
     ID.AddInteger(ST->getMemoryVT().getRawBits());
     ID.AddInteger(ST->getRawSubclassData());
+    ID.AddInteger(ST->getPointerInfo().getAddrSpace());
     break;
   }
   case ISD::ATOMIC_CMP_SWAP:
@@ -467,6 +477,12 @@ static void AddNodeIDCustom(FoldingSetNodeID &ID, const SDNode *N) {
     const AtomicSDNode *AT = cast<AtomicSDNode>(N);
     ID.AddInteger(AT->getMemoryVT().getRawBits());
     ID.AddInteger(AT->getRawSubclassData());
+    ID.AddInteger(AT->getPointerInfo().getAddrSpace());
+    break;
+  }
+  case ISD::PREFETCH: {
+    const MemSDNode *PF = cast<MemSDNode>(N);
+    ID.AddInteger(PF->getPointerInfo().getAddrSpace());
     break;
   }
   case ISD::VECTOR_SHUFFLE: {
@@ -483,6 +499,10 @@ static void AddNodeIDCustom(FoldingSetNodeID &ID, const SDNode *N) {
     break;
   }
   } // end switch (N->getOpcode())
+
+  // Target specific memory nodes could also have address spaces to check.
+  if (N->isTargetMemoryOpcode())
+    ID.AddInteger(cast<MemSDNode>(N)->getPointerInfo().getAddrSpace());
 }
 
 /// AddNodeIDNode - Generic routine for adding a nodes info to the NodeID
@@ -1100,6 +1120,7 @@ SDValue SelectionDAG::getGlobalAddress(const GlobalValue *GV, DebugLoc DL,
   ID.AddPointer(GV);
   ID.AddInteger(Offset);
   ID.AddInteger(TargetFlags);
+  ID.AddInteger(GV->getType()->getAddressSpace());
   void *IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP))
     return SDValue(E, 0);
@@ -1199,6 +1220,24 @@ SDValue SelectionDAG::getConstantPool(MachineConstantPoolValue *C, EVT VT,
   return SDValue(N, 0);
 }
 
+SDValue SelectionDAG::getTargetIndex(int Index, EVT VT, int64_t Offset,
+                                     unsigned char TargetFlags) {
+  FoldingSetNodeID ID;
+  AddNodeIDNode(ID, ISD::TargetIndex, getVTList(VT), 0, 0);
+  ID.AddInteger(Index);
+  ID.AddInteger(Offset);
+  ID.AddInteger(TargetFlags);
+  void *IP = 0;
+  if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP))
+    return SDValue(E, 0);
+
+  SDNode *N = new (NodeAllocator) TargetIndexSDNode(Index, VT, Offset,
+                                                    TargetFlags);
+  CSEMap.InsertNode(N, IP);
+  AllNodes.push_back(N);
+  return SDValue(N, 0);
+}
+
 SDValue SelectionDAG::getBasicBlock(MachineBasicBlock *MBB) {
   FoldingSetNodeID ID;
   AddNodeIDNode(ID, ISD::BasicBlock, getVTList(MVT::Other), 0, 0);
@@ -2444,6 +2483,24 @@ SDValue SelectionDAG::getNode(unsigned Opcode, DebugLoc DL,
       case ISD::FABS:
         V.clearSign();
         return getConstantFP(V, VT);
+      case ISD::FCEIL: {
+        APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardPositive);
+        if (fs == APFloat::opOK || fs == APFloat::opInexact)
+          return getConstantFP(V, VT);
+        break;
+      }
+      case ISD::FTRUNC: {
+        APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardZero);
+        if (fs == APFloat::opOK || fs == APFloat::opInexact)
+          return getConstantFP(V, VT);
+        break;
+      }
+      case ISD::FFLOOR: {
+        APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardNegative);
+        if (fs == APFloat::opOK || fs == APFloat::opInexact)
+          return getConstantFP(V, VT);
+        break;
+      }
       case ISD::FP_EXTEND: {
         bool ignored;
         // This can return overflow, underflow, or inexact; we don't care.
@@ -3901,6 +3958,7 @@ SDValue SelectionDAG::getAtomic(unsigned Opcode, DebugLoc dl, EVT MemVT,
   ID.AddInteger(MemVT.getRawBits());
   SDValue Ops[] = {Chain, Ptr, Cmp, Swp};
   AddNodeIDNode(ID, Opcode, VTs, Ops, 4);
+  ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
   void* IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP)) {
     cast<AtomicSDNode>(E)->refineAlignment(MMO);
@@ -3973,6 +4031,7 @@ SDValue SelectionDAG::getAtomic(unsigned Opcode, DebugLoc dl, EVT MemVT,
   ID.AddInteger(MemVT.getRawBits());
   SDValue Ops[] = {Chain, Ptr, Val};
   AddNodeIDNode(ID, Opcode, VTs, Ops, 3);
+  ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
   void* IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP)) {
     cast<AtomicSDNode>(E)->refineAlignment(MMO);
@@ -4029,6 +4088,7 @@ SDValue SelectionDAG::getAtomic(unsigned Opcode, DebugLoc dl, EVT MemVT,
   ID.AddInteger(MemVT.getRawBits());
   SDValue Ops[] = {Chain, Ptr};
   AddNodeIDNode(ID, Opcode, VTs, Ops, 2);
+  ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
   void* IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP)) {
     cast<AtomicSDNode>(E)->refineAlignment(MMO);
@@ -4106,6 +4166,7 @@ SelectionDAG::getMemIntrinsicNode(unsigned Opcode, DebugLoc dl, SDVTList VTList,
   if (VTList.VTs[VTList.NumVTs-1] != MVT::Glue) {
     FoldingSetNodeID ID;
     AddNodeIDNode(ID, Opcode, VTList, Ops, NumOps);
+    ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
     void *IP = 0;
     if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP)) {
       cast<MemIntrinsicSDNode>(E)->refineAlignment(MMO);
@@ -4225,6 +4286,7 @@ SelectionDAG::getLoad(ISD::MemIndexedMode AM, ISD::LoadExtType ExtType,
   ID.AddInteger(encodeMemSDNodeFlags(ExtType, AM, MMO->isVolatile(),
                                      MMO->isNonTemporal(), 
                                      MMO->isInvariant()));
+  ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
   void *IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP)) {
     cast<LoadSDNode>(E)->refineAlignment(MMO);
@@ -4314,6 +4376,7 @@ SDValue SelectionDAG::getStore(SDValue Chain, DebugLoc dl, SDValue Val,
   ID.AddInteger(VT.getRawBits());
   ID.AddInteger(encodeMemSDNodeFlags(false, ISD::UNINDEXED, MMO->isVolatile(),
                                      MMO->isNonTemporal(), MMO->isInvariant()));
+  ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
   void *IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP)) {
     cast<StoreSDNode>(E)->refineAlignment(MMO);
@@ -4381,6 +4444,7 @@ SDValue SelectionDAG::getTruncStore(SDValue Chain, DebugLoc dl, SDValue Val,
   ID.AddInteger(SVT.getRawBits());
   ID.AddInteger(encodeMemSDNodeFlags(true, ISD::UNINDEXED, MMO->isVolatile(),
                                      MMO->isNonTemporal(), MMO->isInvariant()));
+  ID.AddInteger(MMO->getPointerInfo().getAddrSpace());
   void *IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP)) {
     cast<StoreSDNode>(E)->refineAlignment(MMO);
@@ -4405,6 +4469,7 @@ SelectionDAG::getIndexedStore(SDValue OrigStore, DebugLoc dl, SDValue Base,
   AddNodeIDNode(ID, ISD::STORE, VTs, Ops, 4);
   ID.AddInteger(ST->getMemoryVT().getRawBits());
   ID.AddInteger(ST->getRawSubclassData());
+  ID.AddInteger(ST->getPointerInfo().getAddrSpace());
   void *IP = 0;
   if (SDNode *E = CSEMap.FindNodeOrInsertPos(ID, IP))
     return SDValue(E, 0);
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index 8cbe818..f3cf758 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -1601,7 +1601,10 @@ void SelectionDAGBuilder::visitSwitchCase(CaseBlock &CB,
 
   // Update successor info
   addSuccessorWithWeight(SwitchBB, CB.TrueBB, CB.TrueWeight);
-  addSuccessorWithWeight(SwitchBB, CB.FalseBB, CB.FalseWeight);
+  // TrueBB and FalseBB are always different unless the incoming IR is
+  // degenerate. This only happens when running llc on weird IR.
+  if (CB.TrueBB != CB.FalseBB)
+    addSuccessorWithWeight(SwitchBB, CB.FalseBB, CB.FalseWeight);
 
   // Set NextBlock to be the MBB immediately after the current one, if any.
   // This is used to avoid emitting unnecessary branches to the next block.
@@ -3460,7 +3463,7 @@ void SelectionDAGBuilder::visitAtomicLoad(const LoadInst &I) {
 
   SDValue InChain = getRoot();
 
-  EVT VT = EVT::getEVT(I.getType());
+  EVT VT = TLI.getValueType(I.getType());
 
   if (I.getAlignment() * 8 < VT.getSizeInBits())
     report_fatal_error("Cannot generate unaligned atomic load");
@@ -3490,7 +3493,7 @@ void SelectionDAGBuilder::visitAtomicStore(const StoreInst &I) {
 
   SDValue InChain = getRoot();
 
-  EVT VT = EVT::getEVT(I.getValueOperand()->getType());
+  EVT VT = TLI.getValueType(I.getValueOperand()->getType());
 
   if (I.getAlignment() * 8 < VT.getSizeInBits())
     report_fatal_error("Cannot generate unaligned atomic store");
@@ -4929,6 +4932,11 @@ SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, unsigned Intrinsic) {
                              getValue(I.getArgOperand(0)).getValueType(),
                              getValue(I.getArgOperand(0))));
     return 0;
+  case Intrinsic::floor:
+    setValue(&I, DAG.getNode(ISD::FFLOOR, dl,
+                             getValue(I.getArgOperand(0)).getValueType(),
+                             getValue(I.getArgOperand(0))));
+    return 0;
   case Intrinsic::fma:
     setValue(&I, DAG.getNode(ISD::FMA, dl,
                              getValue(I.getArgOperand(0)).getValueType(),
@@ -5506,6 +5514,22 @@ bool SelectionDAGBuilder::visitMemCmpCall(const CallInst &I) {
   return false;
 }
 
+/// visitUnaryFloatCall - If a call instruction is a unary floating-point
+/// operation (as expected), translate it to an SDNode with the specified opcode
+/// and return true.
+bool SelectionDAGBuilder::visitUnaryFloatCall(const CallInst &I,
+                                              unsigned Opcode) {
+  // Sanity check that it really is a unary floating-point call.
+  if (I.getNumArgOperands() != 1 ||
+      !I.getArgOperand(0)->getType()->isFloatingPointTy() ||
+      I.getType() != I.getArgOperand(0)->getType() ||
+      !I.onlyReadsMemory())
+    return false;
+
+  SDValue Tmp = getValue(I.getArgOperand(0));
+  setValue(&I, DAG.getNode(Opcode, getCurDebugLoc(), Tmp.getValueType(), Tmp));
+  return true;
+}
 
 void SelectionDAGBuilder::visitCall(const CallInst &I) {
   // Handle inline assembly differently.
@@ -5536,150 +5560,97 @@ void SelectionDAGBuilder::visitCall(const CallInst &I) {
 
     // Check for well-known libc/libm calls.  If the function is internal, it
     // can't be a library call.
-    if (!F->hasLocalLinkage() && F->hasName()) {
-      StringRef Name = F->getName();
-      if ((LibInfo->has(LibFunc::copysign) && Name == "copysign") ||
-          (LibInfo->has(LibFunc::copysignf) && Name == "copysignf") ||
-          (LibInfo->has(LibFunc::copysignl) && Name == "copysignl")) {
+    LibFunc::Func Func;
+    if (!F->hasLocalLinkage() && F->hasName() &&
+        LibInfo->getLibFunc(F->getName(), Func) &&
+        LibInfo->hasOptimizedCodeGen(Func)) {
+      switch (Func) {
+      default: break;
+      case LibFunc::copysign:
+      case LibFunc::copysignf:
+      case LibFunc::copysignl:
         if (I.getNumArgOperands() == 2 &&   // Basic sanity checks.
             I.getArgOperand(0)->getType()->isFloatingPointTy() &&
             I.getType() == I.getArgOperand(0)->getType() &&
-            I.getType() == I.getArgOperand(1)->getType()) {
+            I.getType() == I.getArgOperand(1)->getType() &&
+            I.onlyReadsMemory()) {
           SDValue LHS = getValue(I.getArgOperand(0));
           SDValue RHS = getValue(I.getArgOperand(1));
           setValue(&I, DAG.getNode(ISD::FCOPYSIGN, getCurDebugLoc(),
                                    LHS.getValueType(), LHS, RHS));
           return;
         }
-      } else if ((LibInfo->has(LibFunc::fabs) && Name == "fabs") ||
-                 (LibInfo->has(LibFunc::fabsf) && Name == "fabsf") ||
-                 (LibInfo->has(LibFunc::fabsl) && Name == "fabsl")) {
-        if (I.getNumArgOperands() == 1 &&   // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FABS, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::fabs:
+      case LibFunc::fabsf:
+      case LibFunc::fabsl:
+        if (visitUnaryFloatCall(I, ISD::FABS))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::sin) && Name == "sin") ||
-                 (LibInfo->has(LibFunc::sinf) && Name == "sinf") ||
-                 (LibInfo->has(LibFunc::sinl) && Name == "sinl")) {
-        if (I.getNumArgOperands() == 1 &&   // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType() &&
-            I.onlyReadsMemory()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FSIN, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::sin:
+      case LibFunc::sinf:
+      case LibFunc::sinl:
+        if (visitUnaryFloatCall(I, ISD::FSIN))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::cos) && Name == "cos") ||
-                 (LibInfo->has(LibFunc::cosf) && Name == "cosf") ||
-                 (LibInfo->has(LibFunc::cosl) && Name == "cosl")) {
-        if (I.getNumArgOperands() == 1 &&   // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType() &&
-            I.onlyReadsMemory()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FCOS, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::cos:
+      case LibFunc::cosf:
+      case LibFunc::cosl:
+        if (visitUnaryFloatCall(I, ISD::FCOS))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::sqrt) && Name == "sqrt") ||
-                 (LibInfo->has(LibFunc::sqrtf) && Name == "sqrtf") ||
-                 (LibInfo->has(LibFunc::sqrtl) && Name == "sqrtl")) {
-        if (I.getNumArgOperands() == 1 &&   // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType() &&
-            I.onlyReadsMemory()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FSQRT, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::sqrt:
+      case LibFunc::sqrtf:
+      case LibFunc::sqrtl:
+        if (visitUnaryFloatCall(I, ISD::FSQRT))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::floor) && Name == "floor") ||
-                 (LibInfo->has(LibFunc::floorf) && Name == "floorf") ||
-                 (LibInfo->has(LibFunc::floorl) && Name == "floorl")) {
-        if (I.getNumArgOperands() == 1 && // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FFLOOR, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::floor:
+      case LibFunc::floorf:
+      case LibFunc::floorl:
+        if (visitUnaryFloatCall(I, ISD::FFLOOR))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::nearbyint) && Name == "nearbyint") ||
-                 (LibInfo->has(LibFunc::nearbyintf) && Name == "nearbyintf") ||
-                 (LibInfo->has(LibFunc::nearbyintl) && Name == "nearbyintl")) {
-        if (I.getNumArgOperands() == 1 && // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FNEARBYINT, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::nearbyint:
+      case LibFunc::nearbyintf:
+      case LibFunc::nearbyintl:
+        if (visitUnaryFloatCall(I, ISD::FNEARBYINT))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::ceil) && Name == "ceil") ||
-                 (LibInfo->has(LibFunc::ceilf) && Name == "ceilf") ||
-                 (LibInfo->has(LibFunc::ceill) && Name == "ceill")) {
-        if (I.getNumArgOperands() == 1 && // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FCEIL, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::ceil:
+      case LibFunc::ceilf:
+      case LibFunc::ceill:
+        if (visitUnaryFloatCall(I, ISD::FCEIL))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::rint) && Name == "rint") ||
-                 (LibInfo->has(LibFunc::rintf) && Name == "rintf") ||
-                 (LibInfo->has(LibFunc::rintl) && Name == "rintl")) {
-        if (I.getNumArgOperands() == 1 && // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FRINT, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::rint:
+      case LibFunc::rintf:
+      case LibFunc::rintl:
+        if (visitUnaryFloatCall(I, ISD::FRINT))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::trunc) && Name == "trunc") ||
-                 (LibInfo->has(LibFunc::truncf) && Name == "truncf") ||
-                 (LibInfo->has(LibFunc::truncl) && Name == "truncl")) {
-        if (I.getNumArgOperands() == 1 && // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FTRUNC, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::trunc:
+      case LibFunc::truncf:
+      case LibFunc::truncl:
+        if (visitUnaryFloatCall(I, ISD::FTRUNC))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::log2) && Name == "log2") ||
-                 (LibInfo->has(LibFunc::log2f) && Name == "log2f") ||
-                 (LibInfo->has(LibFunc::log2l) && Name == "log2l")) {
-        if (I.getNumArgOperands() == 1 && // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType() &&
-            I.onlyReadsMemory()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FLOG2, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::log2:
+      case LibFunc::log2f:
+      case LibFunc::log2l:
+        if (visitUnaryFloatCall(I, ISD::FLOG2))
           return;
-        }
-      } else if ((LibInfo->has(LibFunc::exp2) && Name == "exp2") ||
-                 (LibInfo->has(LibFunc::exp2f) && Name == "exp2f") ||
-                 (LibInfo->has(LibFunc::exp2l) && Name == "exp2l")) {
-        if (I.getNumArgOperands() == 1 && // Basic sanity checks.
-            I.getArgOperand(0)->getType()->isFloatingPointTy() &&
-            I.getType() == I.getArgOperand(0)->getType() &&
-            I.onlyReadsMemory()) {
-          SDValue Tmp = getValue(I.getArgOperand(0));
-          setValue(&I, DAG.getNode(ISD::FEXP2, getCurDebugLoc(),
-                                   Tmp.getValueType(), Tmp));
+        break;
+      case LibFunc::exp2:
+      case LibFunc::exp2f:
+      case LibFunc::exp2l:
+        if (visitUnaryFloatCall(I, ISD::FEXP2))
           return;
-        }
-      } else if (Name == "memcmp") {
+        break;
+      case LibFunc::memcmp:
         if (visitMemCmpCall(I))
           return;
+        break;
       }
     }
   }
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
index d0fde6f..4090002 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
+++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.h
@@ -520,6 +520,7 @@ private:
   void visitPHI(const PHINode &I);
   void visitCall(const CallInst &I);
   bool visitMemCmpCall(const CallInst &I);
+  bool visitUnaryFloatCall(const CallInst &I, unsigned Opcode);
   void visitAtomicLoad(const LoadInst &I);
   void visitAtomicStore(const StoreInst &I);
 
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
index 9fc225f..13cd011 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
@@ -100,6 +100,7 @@ std::string SDNode::getOperationName(const SelectionDAG *G) const {
   case ISD::EH_SJLJ_SETJMP:             return "EH_SJLJ_SETJMP";
   case ISD::EH_SJLJ_LONGJMP:            return "EH_SJLJ_LONGJMP";
   case ISD::ConstantPool:               return "ConstantPool";
+  case ISD::TargetIndex:                return "TargetIndex";
   case ISD::ExternalSymbol:             return "ExternalSymbol";
   case ISD::BlockAddress:               return "BlockAddress";
   case ISD::INTRINSIC_WO_CHAIN:
@@ -409,6 +410,10 @@ void SDNode::print_details(raw_ostream &OS, const SelectionDAG *G) const {
       OS << " " << offset;
     if (unsigned int TF = CP->getTargetFlags())
       OS << " [TF=" << TF << ']';
+  } else if (const TargetIndexSDNode *TI = dyn_cast<TargetIndexSDNode>(this)) {
+    OS << "<" << TI->getIndex() << '+' << TI->getOffset() << ">";
+    if (unsigned TF = TI->getTargetFlags())
+      OS << " [TF=" << TF << ']';
   } else if (const BasicBlockSDNode *BBDN = dyn_cast<BasicBlockSDNode>(this)) {
     OS << "<";
     const Value *LBB = (const Value*)BBDN->getBasicBlock()->getBasicBlock();
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
index 287c679..4e5e3ba 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
@@ -979,7 +979,7 @@ void SelectionDAGISel::SelectAllBasicBlocks(const Function &Fn) {
   // Initialize the Fast-ISel state, if needed.
   FastISel *FastIS = 0;
   if (TM.Options.EnableFastISel)
-    FastIS = TLI.createFastISel(*FuncInfo);
+    FastIS = TLI.createFastISel(*FuncInfo, LibInfo);
 
   // Iterate over all basic blocks in the function.
   ReversePostOrderTraversal<const Function*> RPOT(&Fn);
diff --git a/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index dff9b2c..6820175 100644
--- a/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -2303,7 +2303,7 @@ TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
         N0.getOpcode() == ISD::AND)
       if (ConstantSDNode *AndRHS =
                   dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
-        EVT ShiftTy = DCI.isBeforeLegalize() ?
+        EVT ShiftTy = DCI.isBeforeLegalizeOps() ?
           getPointerTy() : getShiftAmountTy(N0.getValueType());
         if (Cond == ISD::SETNE && C1 == 0) {// (X & 8) != 0  -->  (X & 8) >> 3
           // Perform the xform if the AND RHS is a single bit.
@@ -2333,7 +2333,7 @@ TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
           const APInt &AndRHSC = AndRHS->getAPIntValue();
           if ((-AndRHSC).isPowerOf2() && (AndRHSC & C1) == C1) {
             unsigned ShiftBits = AndRHSC.countTrailingZeros();
-            EVT ShiftTy = DCI.isBeforeLegalize() ?
+            EVT ShiftTy = DCI.isBeforeLegalizeOps() ?
               getPointerTy() : getShiftAmountTy(N0.getValueType());
             EVT CmpTy = N0.getValueType();
             SDValue Shift = DAG.getNode(ISD::SRL, dl, CmpTy, N0.getOperand(0),
@@ -2361,7 +2361,7 @@ TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
         }
         NewC = NewC.lshr(ShiftBits);
         if (ShiftBits && isLegalICmpImmediate(NewC.getSExtValue())) {
-          EVT ShiftTy = DCI.isBeforeLegalize() ?
+          EVT ShiftTy = DCI.isBeforeLegalizeOps() ?
             getPointerTy() : getShiftAmountTy(N0.getValueType());
           EVT CmpTy = N0.getValueType();
           SDValue Shift = DAG.getNode(ISD::SRL, dl, CmpTy, N0,
@@ -2464,7 +2464,8 @@ TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
     // Otherwise, we can't fold it.  However, we can simplify it to SETUO/SETO
     // if it is not already.
     ISD::CondCode NewCond = UOF == 0 ? ISD::SETO : ISD::SETUO;
-    if (NewCond != Cond)
+    if (NewCond != Cond && (DCI.isBeforeLegalizeOps() ||
+          getCondCodeAction(NewCond, N0.getValueType()) == Legal))
       return DAG.getSetCC(dl, VT, N0, N1, NewCond);
   }