vendor/llvm/llvm-trunk-r321017

41 files changed, 1822 insertions, 1549 deletions

diff --git a/lib/IR/AsmWriter.cpp b/lib/IR/AsmWriter.cpp
index 170bc544d53f..0fafe82404e4 100644
--- a/lib/IR/AsmWriter.cpp
+++ b/lib/IR/AsmWriter.cpp
@@ -373,7 +373,9 @@ static void PrintCallingConv(unsigned cc, raw_ostream &Out) {
   case CallingConv::HHVM:          Out << "hhvmcc"; break;
   case CallingConv::HHVM_C:        Out << "hhvm_ccc"; break;
   case CallingConv::AMDGPU_VS:     Out << "amdgpu_vs"; break;
+  case CallingConv::AMDGPU_LS:     Out << "amdgpu_ls"; break;
   case CallingConv::AMDGPU_HS:     Out << "amdgpu_hs"; break;
+  case CallingConv::AMDGPU_ES:     Out << "amdgpu_es"; break;
   case CallingConv::AMDGPU_GS:     Out << "amdgpu_gs"; break;
   case CallingConv::AMDGPU_PS:     Out << "amdgpu_ps"; break;
   case CallingConv::AMDGPU_CS:     Out << "amdgpu_cs"; break;
@@ -1046,6 +1048,10 @@ void SlotTracker::CreateFunctionSlot(const Value *V) {
 void SlotTracker::CreateMetadataSlot(const MDNode *N) {
   assert(N && "Can't insert a null Value into SlotTracker!");
 
+  // Don't make slots for DIExpressions. We just print them inline everywhere.
+  if (isa<DIExpression>(N))
+    return;
+
   unsigned DestSlot = mdnNext;
   if (!mdnMap.insert(std::make_pair(N, DestSlot)).second)
     return;
@@ -1102,10 +1108,12 @@ static void writeAtomicRMWOperation(raw_ostream &Out,
 
 static void WriteOptimizationInfo(raw_ostream &Out, const User *U) {
   if (const FPMathOperator *FPO = dyn_cast<const FPMathOperator>(U)) {
-    // Unsafe algebra implies all the others, no need to write them all out
-    if (FPO->hasUnsafeAlgebra())
+    // 'Fast' is an abbreviation for all fast-math-flags.
+    if (FPO->isFast())
       Out << " fast";
     else {
+      if (FPO->hasAllowReassoc())
+        Out << " reassoc";
       if (FPO->hasNoNaNs())
         Out << " nnan";
       if (FPO->hasNoInfs())
@@ -1116,6 +1124,8 @@ static void WriteOptimizationInfo(raw_ostream &Out, const User *U) {
         Out << " arcp";
       if (FPO->hasAllowContract())
         Out << " contract";
+      if (FPO->hasApproxFunc())
+        Out << " afn";
     }
   }
 
@@ -1738,6 +1748,7 @@ static void writeDICompileUnit(raw_ostream &Out, const DICompileUnit *N,
   Printer.printBool("splitDebugInlining", N->getSplitDebugInlining(), true);
   Printer.printBool("debugInfoForProfiling", N->getDebugInfoForProfiling(),
                     false);
+  Printer.printBool("gnuPubnames", N->getGnuPubnames(), false);
   Out << ")";
 }
 
@@ -2073,6 +2084,13 @@ static void WriteAsOperandInternal(raw_ostream &Out, const Metadata *MD,
                                    TypePrinting *TypePrinter,
                                    SlotTracker *Machine, const Module *Context,
                                    bool FromValue) {
+  // Write DIExpressions inline when used as a value. Improves readability of
+  // debug info intrinsics.
+  if (const DIExpression *Expr = dyn_cast<DIExpression>(MD)) {
+    writeDIExpression(Out, Expr, TypePrinter, Machine, Context);
+    return;
+  }
+
   if (const MDNode *N = dyn_cast<MDNode>(MD)) {
     std::unique_ptr<SlotTracker> MachineStorage;
     if (!Machine) {
@@ -2424,7 +2442,16 @@ void AssemblyWriter::printNamedMDNode(const NamedMDNode *NMD) {
   for (unsigned i = 0, e = NMD->getNumOperands(); i != e; ++i) {
     if (i)
       Out << ", ";
-    int Slot = Machine.getMetadataSlot(NMD->getOperand(i));
+
+    // Write DIExpressions inline.
+    // FIXME: Ban DIExpressions in NamedMDNodes, they will serve no purpose.
+    MDNode *Op = NMD->getOperand(i);
+    if (auto *Expr = dyn_cast<DIExpression>(Op)) {
+      writeDIExpression(Out, Expr, nullptr, nullptr, nullptr);
+      continue;
+    }
+
+    int Slot = Machine.getMetadataSlot(Op);
     if (Slot == -1)
       Out << "<badref>";
     else
@@ -2470,6 +2497,11 @@ static void PrintVisibility(GlobalValue::VisibilityTypes Vis,
   }
 }
 
+static void PrintDSOLocation(bool IsDSOLocal, formatted_raw_ostream &Out){
+  if (IsDSOLocal)
+    Out << "dso_local ";
+}
+
 static void PrintDLLStorageClass(GlobalValue::DLLStorageClassTypes SCT,
                                  formatted_raw_ostream &Out) {
   switch (SCT) {
@@ -2540,6 +2572,7 @@ void AssemblyWriter::printGlobal(const GlobalVariable *GV) {
     Out << "external ";
 
   Out << getLinkagePrintName(GV->getLinkage());
+  PrintDSOLocation(GV->isDSOLocal(), Out);
   PrintVisibility(GV->getVisibility(), Out);
   PrintDLLStorageClass(GV->getDLLStorageClass(), Out);
   PrintThreadLocalModel(GV->getThreadLocalMode(), Out);
@@ -2586,6 +2619,7 @@ void AssemblyWriter::printIndirectSymbol(const GlobalIndirectSymbol *GIS) {
   Out << " = ";
 
   Out << getLinkagePrintName(GIS->getLinkage());
+  PrintDSOLocation(GIS->isDSOLocal(), Out);
   PrintVisibility(GIS->getVisibility(), Out);
   PrintDLLStorageClass(GIS->getDLLStorageClass(), Out);
   PrintThreadLocalModel(GIS->getThreadLocalMode(), Out);
@@ -2697,6 +2731,7 @@ void AssemblyWriter::printFunction(const Function *F) {
     Out << "define ";
 
   Out << getLinkagePrintName(F->getLinkage());
+  PrintDSOLocation(F->isDSOLocal(), Out);
   PrintVisibility(F->getVisibility(), Out);
   PrintDLLStorageClass(F->getDLLStorageClass(), Out);
 
@@ -3572,7 +3607,7 @@ static void printMetadataImpl(raw_ostream &ROS, const Metadata &MD,
                          /* FromValue */ true);
 
   auto *N = dyn_cast<MDNode>(&MD);
-  if (OnlyAsOperand || !N)
+  if (OnlyAsOperand || !N || isa<DIExpression>(MD))
     return;
 
   OS << " = ";
diff --git a/lib/IR/Attributes.cpp b/lib/IR/Attributes.cpp
index 8f2e641d64b9..1b19a0474727 100644
--- a/lib/IR/Attributes.cpp
+++ b/lib/IR/Attributes.cpp
@@ -245,6 +245,8 @@ std::string Attribute::getAsString(bool InAttrGrp) const {
 
   if (hasAttribute(Attribute::SanitizeAddress))
     return "sanitize_address";
+  if (hasAttribute(Attribute::SanitizeHWAddress))
+    return "sanitize_hwaddress";
   if (hasAttribute(Attribute::AlwaysInline))
     return "alwaysinline";
   if (hasAttribute(Attribute::ArgMemOnly))
@@ -327,6 +329,8 @@ std::string Attribute::getAsString(bool InAttrGrp) const {
     return "sspstrong";
   if (hasAttribute(Attribute::SafeStack))
     return "safestack";
+  if (hasAttribute(Attribute::StrictFP))
+    return "strictfp";
   if (hasAttribute(Attribute::StructRet))
     return "sret";
   if (hasAttribute(Attribute::SanitizeThread))
@@ -788,14 +792,12 @@ std::string AttributeSetNode::getAsString(bool InAttrGrp) const {
 // AttributeListImpl Definition
 //===----------------------------------------------------------------------===//
 
-/// Map from AttributeList index to the internal array index. Adding one works:
-///   FunctionIndex: ~0U -> 0
-///   ReturnIndex:    0  -> 1
-///   FirstArgIndex: 1.. -> 2..
+/// Map from AttributeList index to the internal array index. Adding one happens
+/// to work, but it relies on unsigned integer wrapping. MSVC warns about
+/// unsigned wrapping in constexpr functions, so write out the conditional. LLVM
+/// folds it to add anyway.
 static constexpr unsigned attrIdxToArrayIdx(unsigned Index) {
-  // MSVC warns about '~0U + 1' wrapping around when this is called on
-  // FunctionIndex, so cast to int first.
-  return static_cast<int>(Index) + 1;
+  return Index == AttributeList::FunctionIndex ? 0 : Index + 1;
 }
 
 AttributeListImpl::AttributeListImpl(LLVMContext &C,
diff --git a/lib/IR/AutoUpgrade.cpp b/lib/IR/AutoUpgrade.cpp
index 80640def955e..c258d1a4e3ad 100644
--- a/lib/IR/AutoUpgrade.cpp
+++ b/lib/IR/AutoUpgrade.cpp
@@ -15,8 +15,6 @@
 
 #include "llvm/IR/AutoUpgrade.h"
 #include "llvm/ADT/StringSwitch.h"
-#include "llvm/IR/CFG.h"
-#include "llvm/IR/CallSite.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DIBuilder.h"
 #include "llvm/IR/DebugInfo.h"
@@ -24,9 +22,9 @@
 #include "llvm/IR/Function.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Instruction.h"
-#include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Module.h"
+#include "llvm/IR/Verifier.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/Regex.h"
 #include <cstring>
@@ -72,12 +70,24 @@ static bool ShouldUpgradeX86Intrinsic(Function *F, StringRef Name) {
   // like to use this information to remove upgrade code for some older
   // intrinsics. It is currently undecided how we will determine that future
   // point.
-  if (Name.startswith("sse2.pcmpeq.") || // Added in 3.1
+  if (Name=="ssse3.pabs.b.128" || // Added in 6.0
+      Name=="ssse3.pabs.w.128" || // Added in 6.0
+      Name=="ssse3.pabs.d.128" || // Added in 6.0
+      Name.startswith("avx512.mask.shuf.i") || // Added in 6.0
+      Name.startswith("avx512.mask.shuf.f") || // Added in 6.0
+      Name.startswith("avx512.kunpck") || //added in 6.0 
+      Name.startswith("avx2.pabs.") || // Added in 6.0
+      Name.startswith("avx512.mask.pabs.") || // Added in 6.0
+      Name.startswith("avx512.broadcastm") || // Added in 6.0
+      Name.startswith("avx512.mask.pbroadcast") || // Added in 6.0
+      Name.startswith("sse2.pcmpeq.") || // Added in 3.1
       Name.startswith("sse2.pcmpgt.") || // Added in 3.1
       Name.startswith("avx2.pcmpeq.") || // Added in 3.1
       Name.startswith("avx2.pcmpgt.") || // Added in 3.1
       Name.startswith("avx512.mask.pcmpeq.") || // Added in 3.9
       Name.startswith("avx512.mask.pcmpgt.") || // Added in 3.9
+      Name.startswith("avx.vperm2f128.") || // Added in 6.0
+      Name == "avx2.vperm2i128" || // Added in 6.0
       Name == "sse.add.ss" || // Added in 4.0
       Name == "sse2.add.sd" || // Added in 4.0
       Name == "sse.sub.ss" || // Added in 4.0
@@ -239,12 +249,19 @@ static bool ShouldUpgradeX86Intrinsic(Function *F, StringRef Name) {
       Name.startswith("avx2.pblendd.") || // Added in 3.7
       Name.startswith("avx.vbroadcastf128") || // Added in 4.0
       Name == "avx2.vbroadcasti128" || // Added in 3.7
+      Name.startswith("avx512.mask.broadcastf") || // Added in 6.0
+      Name.startswith("avx512.mask.broadcasti") || // Added in 6.0
       Name == "xop.vpcmov" || // Added in 3.8
       Name == "xop.vpcmov.256" || // Added in 5.0
       Name.startswith("avx512.mask.move.s") || // Added in 4.0
       Name.startswith("avx512.cvtmask2") || // Added in 5.0
       (Name.startswith("xop.vpcom") && // Added in 3.2
-       F->arg_size() == 2))
+       F->arg_size() == 2) ||
+      Name.startswith("avx512.ptestm") || //Added in 6.0
+      Name.startswith("avx512.ptestnm") || //Added in 6.0
+      Name.startswith("sse2.pavg") || // Added in 6.0
+      Name.startswith("avx2.pavg") || // Added in 6.0
+      Name.startswith("avx512.mask.pavg")) // Added in 6.0
     return true;
 
   return false;
@@ -420,6 +437,14 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
     }
     break;
   }
+  case 'd': {
+    if (Name == "dbg.value" && F->arg_size() == 4) {
+      rename(F);
+      NewFn = Intrinsic::getDeclaration(F->getParent(), Intrinsic::dbg_value);
+      return true;
+    }
+    break;
+  }
   case 'i':
   case 'l': {
     bool IsLifetimeStart = Name.startswith("lifetime.start");
@@ -774,6 +799,20 @@ static Value *UpgradeMaskedLoad(IRBuilder<> &Builder,
   return Builder.CreateMaskedLoad(Ptr, Align, Mask, Passthru);
 }
 
+static Value *upgradeAbs(IRBuilder<> &Builder, CallInst &CI) {
+  Value *Op0 = CI.getArgOperand(0);
+  llvm::Type *Ty = Op0->getType();
+  Value *Zero = llvm::Constant::getNullValue(Ty);
+  Value *Cmp = Builder.CreateICmp(ICmpInst::ICMP_SGT, Op0, Zero);
+  Value *Neg = Builder.CreateNeg(Op0);
+  Value *Res = Builder.CreateSelect(Cmp, Op0, Neg);
+
+  if (CI.getNumArgOperands() == 3)
+    Res = EmitX86Select(Builder,CI.getArgOperand(2), Res, CI.getArgOperand(1));
+
+  return Res;
+}
+
 static Value *upgradeIntMinMax(IRBuilder<> &Builder, CallInst &CI,
                                ICmpInst::Predicate Pred) {
   Value *Op0 = CI.getArgOperand(0);
@@ -787,6 +826,26 @@ static Value *upgradeIntMinMax(IRBuilder<> &Builder, CallInst &CI,
   return Res;
 }
 
+// Applying mask on vector of i1's and make sure result is at least 8 bits wide.
+static Value *ApplyX86MaskOn1BitsVec(IRBuilder<> &Builder,Value *Vec, Value *Mask,
+                                     unsigned NumElts) {
+  const auto *C = dyn_cast<Constant>(Mask);
+  if (!C || !C->isAllOnesValue())
+    Vec = Builder.CreateAnd(Vec, getX86MaskVec(Builder, Mask, NumElts));
+
+  if (NumElts < 8) {
+    uint32_t Indices[8];
+    for (unsigned i = 0; i != NumElts; ++i)
+      Indices[i] = i;
+    for (unsigned i = NumElts; i != 8; ++i)
+      Indices[i] = NumElts + i % NumElts;
+    Vec = Builder.CreateShuffleVector(Vec,
+                                      Constant::getNullValue(Vec->getType()),
+                                      Indices);
+  }
+  return Builder.CreateBitCast(Vec, Builder.getIntNTy(std::max(NumElts, 8U)));
+}
+
 static Value *upgradeMaskedCompare(IRBuilder<> &Builder, CallInst &CI,
                                    unsigned CC, bool Signed) {
   Value *Op0 = CI.getArgOperand(0);
@@ -812,22 +871,8 @@ static Value *upgradeMaskedCompare(IRBuilder<> &Builder, CallInst &CI,
   }
 
   Value *Mask = CI.getArgOperand(CI.getNumArgOperands() - 1);
-  const auto *C = dyn_cast<Constant>(Mask);
-  if (!C || !C->isAllOnesValue())
-    Cmp = Builder.CreateAnd(Cmp, getX86MaskVec(Builder, Mask, NumElts));
 
-  if (NumElts < 8) {
-    uint32_t Indices[8];
-    for (unsigned i = 0; i != NumElts; ++i)
-      Indices[i] = i;
-    for (unsigned i = NumElts; i != 8; ++i)
-      Indices[i] = NumElts + i % NumElts;
-    Cmp = Builder.CreateShuffleVector(Cmp,
-                                      Constant::getNullValue(Cmp->getType()),
-                                      Indices);
-  }
-  return Builder.CreateBitCast(Cmp, IntegerType::get(CI.getContext(),
-                                                     std::max(NumElts, 8U)));
+  return ApplyX86MaskOn1BitsVec(Builder, Cmp, Mask, NumElts);
 }
 
 // Replace a masked intrinsic with an older unmasked intrinsic.
@@ -991,6 +1036,39 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       Rep = Builder.CreateICmp(CmpEq ? ICmpInst::ICMP_EQ : ICmpInst::ICMP_SGT,
                                CI->getArgOperand(0), CI->getArgOperand(1));
       Rep = Builder.CreateSExt(Rep, CI->getType(), "");
+    } else if (IsX86 && (Name.startswith("avx512.broadcastm"))) {
+      Type *ExtTy = Type::getInt32Ty(C);
+      if (CI->getOperand(0)->getType()->isIntegerTy(8))
+        ExtTy = Type::getInt64Ty(C);
+      unsigned NumElts = CI->getType()->getPrimitiveSizeInBits() /
+                         ExtTy->getPrimitiveSizeInBits();
+      Rep = Builder.CreateZExt(CI->getArgOperand(0), ExtTy);
+      Rep = Builder.CreateVectorSplat(NumElts, Rep);
+    } else if (IsX86 && (Name.startswith("avx512.ptestm") ||
+                         Name.startswith("avx512.ptestnm"))) {
+      Value *Op0 = CI->getArgOperand(0);
+      Value *Op1 = CI->getArgOperand(1);
+      Value *Mask = CI->getArgOperand(2);
+      Rep = Builder.CreateAnd(Op0, Op1);
+      llvm::Type *Ty = Op0->getType();
+      Value *Zero = llvm::Constant::getNullValue(Ty);
+      ICmpInst::Predicate Pred =
+        Name.startswith("avx512.ptestm") ? ICmpInst::ICMP_NE : ICmpInst::ICMP_EQ;
+      Rep = Builder.CreateICmp(Pred, Rep, Zero);
+      unsigned NumElts = Op0->getType()->getVectorNumElements();
+      Rep = ApplyX86MaskOn1BitsVec(Builder, Rep, Mask, NumElts);
+    } else if (IsX86 && (Name.startswith("avx512.mask.pbroadcast"))){
+      unsigned NumElts =
+          CI->getArgOperand(1)->getType()->getVectorNumElements();
+      Rep = Builder.CreateVectorSplat(NumElts, CI->getArgOperand(0));
+      Rep = EmitX86Select(Builder, CI->getArgOperand(2), Rep,
+                          CI->getArgOperand(1));
+    } else if (IsX86 && (Name.startswith("avx512.kunpck"))) {
+      uint64_t Shift = CI->getType()->getScalarSizeInBits() / 2;
+      uint64_t And = (1ULL << Shift) - 1; 
+      Value* LowBits =  Builder.CreateAnd(CI->getArgOperand(0), And);
+      Value* HighBits =  Builder.CreateShl(CI->getArgOperand(1), Shift);
+      Rep = Builder.CreateOr(LowBits, HighBits);
     } else if (IsX86 && (Name == "sse.add.ss" || Name == "sse2.add.sd")) {
       Type *I32Ty = Type::getInt32Ty(C);
       Value *Elt0 = Builder.CreateExtractElement(CI->getArgOperand(0),
@@ -1037,6 +1115,12 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
     } else if (IsX86 && Name.startswith("avx512.mask.ucmp")) {
       unsigned Imm = cast<ConstantInt>(CI->getArgOperand(2))->getZExtValue();
       Rep = upgradeMaskedCompare(Builder, *CI, Imm, false);
+    } else if(IsX86 && (Name == "ssse3.pabs.b.128" ||
+                        Name == "ssse3.pabs.w.128" ||
+                        Name == "ssse3.pabs.d.128" ||
+                        Name.startswith("avx2.pabs") ||
+                        Name.startswith("avx512.mask.pabs"))) {
+      Rep = upgradeAbs(Builder, *CI);
     } else if (IsX86 && (Name == "sse41.pmaxsb" ||
                          Name == "sse2.pmaxs.w" ||
                          Name == "sse41.pmaxsd" ||
@@ -1213,6 +1297,43 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       else
         Rep = Builder.CreateShuffleVector(Load, UndefValue::get(Load->getType()),
                                           { 0, 1, 2, 3, 0, 1, 2, 3 });
+    } else if (IsX86 && (Name.startswith("avx512.mask.shuf.i") ||
+                         Name.startswith("avx512.mask.shuf.f"))) {
+      unsigned Imm = cast<ConstantInt>(CI->getArgOperand(2))->getZExtValue();
+      Type *VT = CI->getType();
+      unsigned NumLanes = VT->getPrimitiveSizeInBits() / 128;
+      unsigned NumElementsInLane = 128 / VT->getScalarSizeInBits();
+      unsigned ControlBitsMask = NumLanes - 1;
+      unsigned NumControlBits = NumLanes / 2;
+      SmallVector<uint32_t, 8> ShuffleMask(0);
+
+      for (unsigned l = 0; l != NumLanes; ++l) {
+        unsigned LaneMask = (Imm >> (l * NumControlBits)) & ControlBitsMask;
+        // We actually need the other source.
+        if (l >= NumLanes / 2)
+          LaneMask += NumLanes;
+        for (unsigned i = 0; i != NumElementsInLane; ++i)
+          ShuffleMask.push_back(LaneMask * NumElementsInLane + i);
+      }
+      Rep = Builder.CreateShuffleVector(CI->getArgOperand(0),
+                                        CI->getArgOperand(1), ShuffleMask);
+      Rep = EmitX86Select(Builder, CI->getArgOperand(4), Rep,
+                          CI->getArgOperand(3));
+    }else if (IsX86 && (Name.startswith("avx512.mask.broadcastf") ||
+                         Name.startswith("avx512.mask.broadcasti"))) {
+      unsigned NumSrcElts =
+                        CI->getArgOperand(0)->getType()->getVectorNumElements();
+      unsigned NumDstElts = CI->getType()->getVectorNumElements();
+
+      SmallVector<uint32_t, 8> ShuffleMask(NumDstElts);
+      for (unsigned i = 0; i != NumDstElts; ++i)
+        ShuffleMask[i] = i % NumSrcElts;
+
+      Rep = Builder.CreateShuffleVector(CI->getArgOperand(0),
+                                        CI->getArgOperand(0),
+                                        ShuffleMask);
+      Rep = EmitX86Select(Builder, CI->getArgOperand(2), Rep,
+                          CI->getArgOperand(1));
     } else if (IsX86 && (Name.startswith("avx2.pbroadcast") ||
                          Name.startswith("avx2.vbroadcast") ||
                          Name.startswith("avx512.pbroadcast") ||
@@ -1367,6 +1488,42 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       if (CI->getNumArgOperands() == 4)
         Rep = EmitX86Select(Builder, CI->getArgOperand(3), Rep,
                             CI->getArgOperand(2));
+    } else if (IsX86 && (Name.startswith("avx.vperm2f128.") ||
+                         Name == "avx2.vperm2i128")) {
+      // The immediate permute control byte looks like this:
+      //    [1:0] - select 128 bits from sources for low half of destination
+      //    [2]   - ignore
+      //    [3]   - zero low half of destination
+      //    [5:4] - select 128 bits from sources for high half of destination
+      //    [6]   - ignore
+      //    [7]   - zero high half of destination
+
+      uint8_t Imm = cast<ConstantInt>(CI->getArgOperand(2))->getZExtValue();
+
+      unsigned NumElts = CI->getType()->getVectorNumElements();
+      unsigned HalfSize = NumElts / 2;
+      SmallVector<uint32_t, 8> ShuffleMask(NumElts);
+
+      // Determine which operand(s) are actually in use for this instruction.
+      Value *V0 = (Imm & 0x02) ? CI->getArgOperand(1) : CI->getArgOperand(0);
+      Value *V1 = (Imm & 0x20) ? CI->getArgOperand(1) : CI->getArgOperand(0);
+
+      // If needed, replace operands based on zero mask.
+      V0 = (Imm & 0x08) ? ConstantAggregateZero::get(CI->getType()) : V0;
+      V1 = (Imm & 0x80) ? ConstantAggregateZero::get(CI->getType()) : V1;
+
+      // Permute low half of result.
+      unsigned StartIndex = (Imm & 0x01) ? HalfSize : 0;
+      for (unsigned i = 0; i < HalfSize; ++i)
+        ShuffleMask[i] = StartIndex + i;
+
+      // Permute high half of result.
+      StartIndex = (Imm & 0x10) ? HalfSize : 0;
+      for (unsigned i = 0; i < HalfSize; ++i)
+        ShuffleMask[i + HalfSize] = NumElts + StartIndex + i;
+
+      Rep = Builder.CreateShuffleVector(V0, V1, ShuffleMask);
+
     } else if (IsX86 && (Name.startswith("avx.vpermil.") ||
                          Name == "sse2.pshuf.d" ||
                          Name.startswith("avx512.mask.vpermil.p") ||
@@ -1941,6 +2098,25 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       LoadInst *LI = Builder.CreateAlignedLoad(BC, VTy->getBitWidth() / 8);
       LI->setMetadata(M->getMDKindID("nontemporal"), Node);
       Rep = LI;
+    } else if (IsX86 &&
+               (Name.startswith("sse2.pavg") || Name.startswith("avx2.pavg") ||
+                Name.startswith("avx512.mask.pavg"))) {
+      // llvm.x86.sse2.pavg.b/w, llvm.x86.avx2.pavg.b/w,
+      // llvm.x86.avx512.mask.pavg.b/w
+      Value *A = CI->getArgOperand(0);
+      Value *B = CI->getArgOperand(1);
+      VectorType *ZextType = VectorType::getExtendedElementVectorType(
+          cast<VectorType>(A->getType()));
+      Value *ExtendedA = Builder.CreateZExt(A, ZextType);
+      Value *ExtendedB = Builder.CreateZExt(B, ZextType);
+      Value *Sum = Builder.CreateAdd(ExtendedA, ExtendedB);
+      Value *AddOne = Builder.CreateAdd(Sum, ConstantInt::get(ZextType, 1));
+      Value *ShiftR = Builder.CreateLShr(AddOne, ConstantInt::get(ZextType, 1));
+      Rep = Builder.CreateTrunc(ShiftR, A->getType());
+      if (CI->getNumArgOperands() > 2) {
+        Rep = EmitX86Select(Builder, CI->getArgOperand(3), Rep,
+                            CI->getArgOperand(2));
+      }
     } else if (IsNVVM && (Name == "abs.i" || Name == "abs.ll")) {
       Value *Arg = CI->getArgOperand(0);
       Value *Neg = Builder.CreateNeg(Arg, "neg");
@@ -2055,6 +2231,20 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
     NewCall = Builder.CreateCall(NewFn, {CI->getArgOperand(0)});
     break;
 
+  case Intrinsic::dbg_value:
+    // Upgrade from the old version that had an extra offset argument.
+    assert(CI->getNumArgOperands() == 4);
+    // Drop nonzero offsets instead of attempting to upgrade them.
+    if (auto *Offset = dyn_cast_or_null<Constant>(CI->getArgOperand(1)))
+      if (Offset->isZeroValue()) {
+        NewCall = Builder.CreateCall(
+            NewFn,
+            {CI->getArgOperand(0), CI->getArgOperand(2), CI->getArgOperand(3)});
+        break;
+      }
+    CI->eraseFromParent();
+    return;
+
   case Intrinsic::x86_xop_vfrcz_ss:
   case Intrinsic::x86_xop_vfrcz_sd:
     NewCall = Builder.CreateCall(NewFn, {CI->getArgOperand(1)});
@@ -2227,15 +2417,26 @@ Value *llvm::UpgradeBitCastExpr(unsigned Opc, Constant *C, Type *DestTy) {
 /// info. Return true if module is modified.
 bool llvm::UpgradeDebugInfo(Module &M) {
   unsigned Version = getDebugMetadataVersionFromModule(M);
-  if (Version == DEBUG_METADATA_VERSION)
-    return false;
-
-  bool RetCode = StripDebugInfo(M);
-  if (RetCode) {
+  if (Version == DEBUG_METADATA_VERSION) {
+    bool BrokenDebugInfo = false;
+    if (verifyModule(M, &llvm::errs(), &BrokenDebugInfo))
+      report_fatal_error("Broken module found, compilation aborted!");
+    if (!BrokenDebugInfo)
+      // Everything is ok.
+      return false;
+    else {
+      // Diagnose malformed debug info.
+      DiagnosticInfoIgnoringInvalidDebugMetadata Diag(M);
+      M.getContext().diagnose(Diag);
+    }
+  }
+  bool Modified = StripDebugInfo(M);
+  if (Modified && Version != DEBUG_METADATA_VERSION) {
+    // Diagnose a version mismatch.
     DiagnosticInfoDebugMetadataVersion DiagVersion(M, Version);
     M.getContext().diagnose(DiagVersion);
   }
-  return RetCode;
+  return Modified;
 }
 
 bool llvm::UpgradeModuleFlags(Module &M) {
diff --git a/lib/IR/BasicBlock.cpp b/lib/IR/BasicBlock.cpp
index 2b780adf6c69..22513924a96d 100644
--- a/lib/IR/BasicBlock.cpp
+++ b/lib/IR/BasicBlock.cpp
@@ -447,3 +447,16 @@ bool BasicBlock::isLandingPad() const {
 const LandingPadInst *BasicBlock::getLandingPadInst() const {
   return dyn_cast<LandingPadInst>(getFirstNonPHI());
 }
+
+Optional<uint64_t> BasicBlock::getIrrLoopHeaderWeight() const {
+  const TerminatorInst *TI = getTerminator();
+  if (MDNode *MDIrrLoopHeader =
+      TI->getMetadata(LLVMContext::MD_irr_loop)) {
+    MDString *MDName = cast<MDString>(MDIrrLoopHeader->getOperand(0));
+    if (MDName->getString().equals("loop_header_weight")) {
+      auto *CI = mdconst::extract<ConstantInt>(MDIrrLoopHeader->getOperand(1));
+      return Optional<uint64_t>(CI->getValue().getZExtValue());
+    }
+  }
+  return Optional<uint64_t>();
+}
diff --git a/lib/IR/CMakeLists.txt b/lib/IR/CMakeLists.txt
index 1cc229d68bfc..17822bbbb5cb 100644
--- a/lib/IR/CMakeLists.txt
+++ b/lib/IR/CMakeLists.txt
@@ -17,11 +17,11 @@ add_llvm_library(LLVMCore
   DebugInfo.cpp
   DebugInfoMetadata.cpp
   DebugLoc.cpp
+  DiagnosticHandler.cpp
   DiagnosticInfo.cpp
   DiagnosticPrinter.cpp
   Dominators.cpp
   Function.cpp
-  GCOV.cpp
   GVMaterializer.cpp
   Globals.cpp
   IRBuilder.cpp
diff --git a/lib/IR/ConstantFold.cpp b/lib/IR/ConstantFold.cpp
index 996331e68e83..90b10309b58b 100644
--- a/lib/IR/ConstantFold.cpp
+++ b/lib/IR/ConstantFold.cpp
@@ -629,6 +629,15 @@ Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V,
     if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
       if (CE->getOpcode() == Instruction::GetElementPtr &&
           CE->getOperand(0)->isNullValue()) {
+        // FIXME: Looks like getFoldedSizeOf(), getFoldedOffsetOf() and
+        // getFoldedAlignOf() don't handle the case when DestTy is a vector of
+        // pointers yet. We end up in asserts in CastInst::getCastOpcode (see
+        // test/Analysis/ConstantFolding/cast-vector.ll). I've only seen this
+        // happen in one "real" C-code test case, so it does not seem to be an
+        // important optimization to handle vectors here. For now, simply bail
+        // out.
+        if (DestTy->isVectorTy())
+          return nullptr;
         GEPOperator *GEPO = cast<GEPOperator>(CE);
         Type *Ty = GEPO->getSourceElementType();
         if (CE->getNumOperands() == 2) {
@@ -2062,9 +2071,20 @@ Constant *llvm::ConstantFoldGetElementPtr(Type *PointeeTy, Constant *C,
       Type *Ty = GetElementPtrInst::getIndexedType(PointeeTy, Idxs);
 
       assert(Ty && "Invalid indices for GEP!");
+      Type *OrigGEPTy = PointerType::get(Ty, PtrTy->getAddressSpace());
       Type *GEPTy = PointerType::get(Ty, PtrTy->getAddressSpace());
       if (VectorType *VT = dyn_cast<VectorType>(C->getType()))
-        GEPTy = VectorType::get(GEPTy, VT->getNumElements());
+        GEPTy = VectorType::get(OrigGEPTy, VT->getNumElements());
+
+      // The GEP returns a vector of pointers when one of more of
+      // its arguments is a vector.
+      for (unsigned i = 0, e = Idxs.size(); i != e; ++i) {
+        if (auto *VT = dyn_cast<VectorType>(Idxs[i]->getType())) {
+          GEPTy = VectorType::get(OrigGEPTy, VT->getNumElements());
+          break;
+        }
+      }
+
       return Constant::getNullValue(GEPTy);
     }
   }
@@ -2190,17 +2210,17 @@ Constant *llvm::ConstantFoldGetElementPtr(Type *PointeeTy, Constant *C,
   SmallVector<Constant *, 8> NewIdxs;
   Type *Ty = PointeeTy;
   Type *Prev = C->getType();
-  bool Unknown = !isa<ConstantInt>(Idxs[0]);
+  bool Unknown =
+      !isa<ConstantInt>(Idxs[0]) && !isa<ConstantDataVector>(Idxs[0]);
   for (unsigned i = 1, e = Idxs.size(); i != e;
        Prev = Ty, Ty = cast<CompositeType>(Ty)->getTypeAtIndex(Idxs[i]), ++i) {
-    auto *CI = dyn_cast<ConstantInt>(Idxs[i]);
-    if (!CI) {
+    if (!isa<ConstantInt>(Idxs[i]) && !isa<ConstantDataVector>(Idxs[i])) {
       // We don't know if it's in range or not.
       Unknown = true;
       continue;
     }
-    if (!isa<ConstantInt>(Idxs[i - 1]))
-      // FIXME: add the support of cosntant vector index.
+    if (!isa<ConstantInt>(Idxs[i - 1]) && !isa<ConstantDataVector>(Idxs[i - 1]))
+      // Skip if the type of the previous index is not supported.
       continue;
     if (InRangeIndex && i == *InRangeIndex + 1) {
       // If an index is marked inrange, we cannot apply this canonicalization to
@@ -2218,46 +2238,91 @@ Constant *llvm::ConstantFoldGetElementPtr(Type *PointeeTy, Constant *C,
       Unknown = true;
       continue;
     }
-    if (isIndexInRangeOfArrayType(STy->getNumElements(), CI))
-      // It's in range, skip to the next index.
-      continue;
+    if (ConstantInt *CI = dyn_cast<ConstantInt>(Idxs[i])) {
+      if (isIndexInRangeOfArrayType(STy->getNumElements(), CI))
+        // It's in range, skip to the next index.
+        continue;
+      if (CI->getSExtValue() < 0) {
+        // It's out of range and negative, don't try to factor it.
+        Unknown = true;
+        continue;
+      }
+    } else {
+      auto *CV = cast<ConstantDataVector>(Idxs[i]);
+      bool InRange = true;
+      for (unsigned I = 0, E = CV->getNumElements(); I != E; ++I) {
+        auto *CI = cast<ConstantInt>(CV->getElementAsConstant(I));
+        InRange &= isIndexInRangeOfArrayType(STy->getNumElements(), CI);
+        if (CI->getSExtValue() < 0) {
+          Unknown = true;
+          break;
+        }
+      }
+      if (InRange || Unknown)
+        // It's in range, skip to the next index.
+        // It's out of range and negative, don't try to factor it.
+        continue;
+    }
     if (isa<StructType>(Prev)) {
       // It's out of range, but the prior dimension is a struct
       // so we can't do anything about it.
       Unknown = true;
       continue;
     }
-    if (CI->getSExtValue() < 0) {
-      // It's out of range and negative, don't try to factor it.
-      Unknown = true;
-      continue;
-    }
     // It's out of range, but we can factor it into the prior
     // dimension.
     NewIdxs.resize(Idxs.size());
     // Determine the number of elements in our sequential type.
     uint64_t NumElements = STy->getArrayNumElements();
 
-    ConstantInt *Factor = ConstantInt::get(CI->getType(), NumElements);
-    NewIdxs[i] = ConstantExpr::getSRem(CI, Factor);
+    // Expand the current index or the previous index to a vector from a scalar
+    // if necessary.
+    Constant *CurrIdx = cast<Constant>(Idxs[i]);
+    auto *PrevIdx =
+        NewIdxs[i - 1] ? NewIdxs[i - 1] : cast<Constant>(Idxs[i - 1]);
+    bool IsCurrIdxVector = CurrIdx->getType()->isVectorTy();
+    bool IsPrevIdxVector = PrevIdx->getType()->isVectorTy();
+    bool UseVector = IsCurrIdxVector || IsPrevIdxVector;
+
+    if (!IsCurrIdxVector && IsPrevIdxVector)
+      CurrIdx = ConstantDataVector::getSplat(
+          PrevIdx->getType()->getVectorNumElements(), CurrIdx);
 
-    Constant *PrevIdx = NewIdxs[i-1] ? NewIdxs[i-1] :
-                           cast<Constant>(Idxs[i - 1]);
-    Constant *Div = ConstantExpr::getSDiv(CI, Factor);
+    if (!IsPrevIdxVector && IsCurrIdxVector)
+      PrevIdx = ConstantDataVector::getSplat(
+          CurrIdx->getType()->getVectorNumElements(), PrevIdx);
+
+    Constant *Factor =
+        ConstantInt::get(CurrIdx->getType()->getScalarType(), NumElements);
+    if (UseVector)
+      Factor = ConstantDataVector::getSplat(
+          IsPrevIdxVector ? PrevIdx->getType()->getVectorNumElements()
+                          : CurrIdx->getType()->getVectorNumElements(),
+          Factor);
+
+    NewIdxs[i] = ConstantExpr::getSRem(CurrIdx, Factor);
+
+    Constant *Div = ConstantExpr::getSDiv(CurrIdx, Factor);
 
     unsigned CommonExtendedWidth =
-        std::max(PrevIdx->getType()->getIntegerBitWidth(),
-                 Div->getType()->getIntegerBitWidth());
+        std::max(PrevIdx->getType()->getScalarSizeInBits(),
+                 Div->getType()->getScalarSizeInBits());
     CommonExtendedWidth = std::max(CommonExtendedWidth, 64U);
 
     // Before adding, extend both operands to i64 to avoid
     // overflow trouble.
-    if (!PrevIdx->getType()->isIntegerTy(CommonExtendedWidth))
-      PrevIdx = ConstantExpr::getSExt(
-          PrevIdx, Type::getIntNTy(Div->getContext(), CommonExtendedWidth));
-    if (!Div->getType()->isIntegerTy(CommonExtendedWidth))
-      Div = ConstantExpr::getSExt(
-          Div, Type::getIntNTy(Div->getContext(), CommonExtendedWidth));
+    Type *ExtendedTy = Type::getIntNTy(Div->getContext(), CommonExtendedWidth);
+    if (UseVector)
+      ExtendedTy = VectorType::get(
+          ExtendedTy, IsPrevIdxVector
+                          ? PrevIdx->getType()->getVectorNumElements()
+                          : CurrIdx->getType()->getVectorNumElements());
+
+    if (!PrevIdx->getType()->isIntOrIntVectorTy(CommonExtendedWidth))
+      PrevIdx = ConstantExpr::getSExt(PrevIdx, ExtendedTy);
+
+    if (!Div->getType()->isIntOrIntVectorTy(CommonExtendedWidth))
+      Div = ConstantExpr::getSExt(Div, ExtendedTy);
 
     NewIdxs[i - 1] = ConstantExpr::getAdd(PrevIdx, Div);
   }
diff --git a/lib/IR/ConstantRange.cpp b/lib/IR/ConstantRange.cpp
index 4bd17257016d..48d16f334ba3 100644
--- a/lib/IR/ConstantRange.cpp
+++ b/lib/IR/ConstantRange.cpp
@@ -199,39 +199,63 @@ ConstantRange::makeGuaranteedNoWrapRegion(Instruction::BinaryOps BinOp,
          "NoWrapKind invalid!");
 
   unsigned BitWidth = Other.getBitWidth();
-  if (BinOp != Instruction::Add)
-    // Conservative answer: empty set
-    return ConstantRange(BitWidth, false);
-
-  if (auto *C = Other.getSingleElement())
-    if (C->isNullValue())
-      // Full set: nothing signed / unsigned wraps when added to 0.
-      return ConstantRange(BitWidth);
-
   ConstantRange Result(BitWidth);
 
-  if (NoWrapKind & OBO::NoUnsignedWrap)
-    Result =
-        SubsetIntersect(Result, ConstantRange(APInt::getNullValue(BitWidth),
-                                              -Other.getUnsignedMax()));
-
-  if (NoWrapKind & OBO::NoSignedWrap) {
-    const APInt &SignedMin = Other.getSignedMin();
-    const APInt &SignedMax = Other.getSignedMax();
+  switch (BinOp) {
+  default:
+    // Conservative answer: empty set
+    return ConstantRange(BitWidth, false);
 
-    if (SignedMax.isStrictlyPositive())
-      Result = SubsetIntersect(
-          Result,
-          ConstantRange(APInt::getSignedMinValue(BitWidth),
-                        APInt::getSignedMinValue(BitWidth) - SignedMax));
+  case Instruction::Add:
+    if (auto *C = Other.getSingleElement())
+      if (C->isNullValue())
+        // Full set: nothing signed / unsigned wraps when added to 0.
+        return ConstantRange(BitWidth);
+    if (NoWrapKind & OBO::NoUnsignedWrap)
+      Result =
+          SubsetIntersect(Result, ConstantRange(APInt::getNullValue(BitWidth),
+                                                -Other.getUnsignedMax()));
+    if (NoWrapKind & OBO::NoSignedWrap) {
+      const APInt &SignedMin = Other.getSignedMin();
+      const APInt &SignedMax = Other.getSignedMax();
+      if (SignedMax.isStrictlyPositive())
+        Result = SubsetIntersect(
+            Result,
+            ConstantRange(APInt::getSignedMinValue(BitWidth),
+                          APInt::getSignedMinValue(BitWidth) - SignedMax));
+      if (SignedMin.isNegative())
+        Result = SubsetIntersect(
+            Result,
+            ConstantRange(APInt::getSignedMinValue(BitWidth) - SignedMin,
+                          APInt::getSignedMinValue(BitWidth)));
+    }
+    return Result;
 
-    if (SignedMin.isNegative())
-      Result = SubsetIntersect(
-          Result, ConstantRange(APInt::getSignedMinValue(BitWidth) - SignedMin,
-                                APInt::getSignedMinValue(BitWidth)));
+  case Instruction::Sub:
+    if (auto *C = Other.getSingleElement())
+      if (C->isNullValue())
+        // Full set: nothing signed / unsigned wraps when subtracting 0.
+        return ConstantRange(BitWidth);
+    if (NoWrapKind & OBO::NoUnsignedWrap)
+      Result =
+          SubsetIntersect(Result, ConstantRange(Other.getUnsignedMax(),
+                                                APInt::getMinValue(BitWidth)));
+    if (NoWrapKind & OBO::NoSignedWrap) {
+      const APInt &SignedMin = Other.getSignedMin();
+      const APInt &SignedMax = Other.getSignedMax();
+      if (SignedMax.isStrictlyPositive())
+        Result = SubsetIntersect(
+            Result,
+            ConstantRange(APInt::getSignedMinValue(BitWidth) + SignedMax,
+                          APInt::getSignedMinValue(BitWidth)));
+      if (SignedMin.isNegative())
+        Result = SubsetIntersect(
+            Result,
+            ConstantRange(APInt::getSignedMinValue(BitWidth),
+                          APInt::getSignedMinValue(BitWidth) + SignedMin));
+    }
+    return Result;
   }
-
-  return Result;
 }
 
 bool ConstantRange::isFullSet() const {
@@ -656,6 +680,8 @@ ConstantRange ConstantRange::binaryOp(Instruction::BinaryOps BinOp,
     return shl(Other);
   case Instruction::LShr:
     return lshr(Other);
+  case Instruction::AShr:
+    return ashr(Other);
   case Instruction::And:
     return binaryAnd(Other);
   case Instruction::Or:
@@ -922,6 +948,60 @@ ConstantRange::lshr(const ConstantRange &Other) const {
   return ConstantRange(std::move(min), std::move(max));
 }
 
+ConstantRange
+ConstantRange::ashr(const ConstantRange &Other) const {
+  if (isEmptySet() || Other.isEmptySet())
+    return ConstantRange(getBitWidth(), /*isFullSet=*/false);
+
+  // May straddle zero, so handle both positive and negative cases.
+  // 'PosMax' is the upper bound of the result of the ashr
+  // operation, when Upper of the LHS of ashr is a non-negative.
+  // number. Since ashr of a non-negative number will result in a
+  // smaller number, the Upper value of LHS is shifted right with
+  // the minimum value of 'Other' instead of the maximum value.
+  APInt PosMax = getSignedMax().ashr(Other.getUnsignedMin()) + 1;
+
+  // 'PosMin' is the lower bound of the result of the ashr
+  // operation, when Lower of the LHS is a non-negative number.
+  // Since ashr of a non-negative number will result in a smaller
+  // number, the Lower value of LHS is shifted right with the
+  // maximum value of 'Other'.
+  APInt PosMin = getSignedMin().ashr(Other.getUnsignedMax());
+
+  // 'NegMax' is the upper bound of the result of the ashr
+  // operation, when Upper of the LHS of ashr is a negative number.
+  // Since 'ashr' of a negative number will result in a bigger
+  // number, the Upper value of LHS is shifted right with the
+  // maximum value of 'Other'.
+  APInt NegMax = getSignedMax().ashr(Other.getUnsignedMax()) + 1;
+
+  // 'NegMin' is the lower bound of the result of the ashr
+  // operation, when Lower of the LHS of ashr is a negative number.
+  // Since 'ashr' of a negative number will result in a bigger
+  // number, the Lower value of LHS is shifted right with the
+  // minimum value of 'Other'.
+  APInt NegMin = getSignedMin().ashr(Other.getUnsignedMin());
+
+  APInt max, min;
+  if (getSignedMin().isNonNegative()) {
+    // Upper and Lower of LHS are non-negative.
+    min = PosMin;
+    max = PosMax;
+  } else if (getSignedMax().isNegative()) {
+    // Upper and Lower of LHS are negative.
+    min = NegMin;
+    max = NegMax;
+  } else {
+    // Upper is non-negative and Lower is negative.
+    min = NegMin;
+    max = PosMax;
+  }
+  if (min == max)
+    return ConstantRange(getBitWidth(), /*isFullSet=*/true);
+
+  return ConstantRange(std::move(min), std::move(max));
+}
+
 ConstantRange ConstantRange::inverse() const {
   if (isFullSet())
     return ConstantRange(getBitWidth(), /*isFullSet=*/false);
diff --git a/lib/IR/Constants.cpp b/lib/IR/Constants.cpp
index f56fe7089807..dccba779deb3 100644
--- a/lib/IR/Constants.cpp
+++ b/lib/IR/Constants.cpp
@@ -16,7 +16,6 @@
 #include "LLVMContextImpl.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/StringMap.h"
 #include "llvm/IR/DerivedTypes.h"
 #include "llvm/IR/GetElementPtrTypeIterator.h"
diff --git a/lib/IR/Core.cpp b/lib/IR/Core.cpp
index aba770457e2f..d3c33edec186 100644
--- a/lib/IR/Core.cpp
+++ b/lib/IR/Core.cpp
@@ -85,15 +85,15 @@ LLVMContextRef LLVMGetGlobalContext() { return wrap(&*GlobalContext); }
 void LLVMContextSetDiagnosticHandler(LLVMContextRef C,
                                      LLVMDiagnosticHandler Handler,
                                      void *DiagnosticContext) {
-  unwrap(C)->setDiagnosticHandler(
-      LLVM_EXTENSION reinterpret_cast<LLVMContext::DiagnosticHandlerTy>(
+  unwrap(C)->setDiagnosticHandlerCallBack(
+      LLVM_EXTENSION reinterpret_cast<DiagnosticHandler::DiagnosticHandlerTy>(
           Handler),
       DiagnosticContext);
 }
 
 LLVMDiagnosticHandler LLVMContextGetDiagnosticHandler(LLVMContextRef C) {
   return LLVM_EXTENSION reinterpret_cast<LLVMDiagnosticHandler>(
-      unwrap(C)->getDiagnosticHandler());
+      unwrap(C)->getDiagnosticHandlerCallBack());
 }
 
 void *LLVMContextGetDiagnosticContext(LLVMContextRef C) {
@@ -276,7 +276,8 @@ LLVMBool LLVMPrintModuleToFile(LLVMModuleRef M, const char *Filename,
   dest.close();
 
   if (dest.has_error()) {
-    *ErrorMessage = strdup("Error printing to file");
+    std::string E = "Error printing to file: " + dest.error().message();
+    *ErrorMessage = strdup(E.c_str());
     return true;
   }
 
@@ -451,9 +452,6 @@ LLVMTypeRef LLVMPPCFP128TypeInContext(LLVMContextRef C) {
 LLVMTypeRef LLVMX86MMXTypeInContext(LLVMContextRef C) {
   return (LLVMTypeRef) Type::getX86_MMXTy(*unwrap(C));
 }
-LLVMTypeRef LLVMTokenTypeInContext(LLVMContextRef C) {
-  return (LLVMTypeRef) Type::getTokenTy(*unwrap(C));
-}
 
 LLVMTypeRef LLVMHalfType(void) {
   return LLVMHalfTypeInContext(LLVMGetGlobalContext());
@@ -619,6 +617,12 @@ LLVMTypeRef LLVMVoidTypeInContext(LLVMContextRef C)  {
 LLVMTypeRef LLVMLabelTypeInContext(LLVMContextRef C) {
   return wrap(Type::getLabelTy(*unwrap(C)));
 }
+LLVMTypeRef LLVMTokenTypeInContext(LLVMContextRef C) {
+  return wrap(Type::getTokenTy(*unwrap(C)));
+}
+LLVMTypeRef LLVMMetadataTypeInContext(LLVMContextRef C) {
+  return wrap(Type::getMetadataTy(*unwrap(C)));
+}
 
 LLVMTypeRef LLVMVoidType(void)  {
   return LLVMVoidTypeInContext(LLVMGetGlobalContext());
diff --git a/lib/IR/DIBuilder.cpp b/lib/IR/DIBuilder.cpp
index bce28ba3b950..a00c595d01c5 100644
--- a/lib/IR/DIBuilder.cpp
+++ b/lib/IR/DIBuilder.cpp
@@ -12,6 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/IR/DIBuilder.h"
+#include "llvm/IR/IRBuilder.h"
 #include "LLVMContextImpl.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/BinaryFormat/Dwarf.h"
@@ -24,6 +25,11 @@
 using namespace llvm;
 using namespace llvm::dwarf;
 
+cl::opt<bool>
+    UseDbgAddr("use-dbg-addr",
+               llvm::cl::desc("Use llvm.dbg.addr for all local variables"),
+               cl::init(false), cl::Hidden);
+
 DIBuilder::DIBuilder(Module &m, bool AllowUnresolvedNodes)
   : M(m), VMContext(M.getContext()), CUNode(nullptr),
       DeclareFn(nullptr), ValueFn(nullptr),
@@ -127,7 +133,7 @@ DICompileUnit *DIBuilder::createCompileUnit(
     unsigned Lang, DIFile *File, StringRef Producer, bool isOptimized,
     StringRef Flags, unsigned RunTimeVer, StringRef SplitName,
     DICompileUnit::DebugEmissionKind Kind, uint64_t DWOId,
-    bool SplitDebugInlining, bool DebugInfoForProfiling) {
+    bool SplitDebugInlining, bool DebugInfoForProfiling, bool GnuPubnames) {
 
   assert(((Lang <= dwarf::DW_LANG_Fortran08 && Lang >= dwarf::DW_LANG_C89) ||
           (Lang <= dwarf::DW_LANG_hi_user && Lang >= dwarf::DW_LANG_lo_user)) &&
@@ -137,7 +143,7 @@ DICompileUnit *DIBuilder::createCompileUnit(
   CUNode = DICompileUnit::getDistinct(
       VMContext, Lang, File, Producer, isOptimized, Flags, RunTimeVer,
       SplitName, Kind, nullptr, nullptr, nullptr, nullptr, nullptr, DWOId,
-      SplitDebugInlining, DebugInfoForProfiling);
+      SplitDebugInlining, DebugInfoForProfiling, GnuPubnames);
 
   // Create a named metadata so that it is easier to find cu in a module.
   NamedMDNode *NMD = M.getOrInsertNamedMetadata("llvm.dbg.cu");
@@ -595,6 +601,8 @@ DIGlobalVariableExpression *DIBuilder::createGlobalVariableExpression(
       VMContext, cast_or_null<DIScope>(Context), Name, LinkageName, F,
       LineNumber, Ty, isLocalToUnit, true, cast_or_null<DIDerivedType>(Decl),
       AlignInBits);
+  if (!Expr)
+    Expr = createExpression();
   auto *N = DIGlobalVariableExpression::get(VMContext, GV, Expr);
   AllGVs.push_back(N);
   return N;
@@ -668,12 +676,6 @@ DIExpression *DIBuilder::createExpression(ArrayRef<int64_t> Signed) {
   return createExpression(Addr);
 }
 
-DIExpression *DIBuilder::createFragmentExpression(unsigned OffsetInBytes,
-                                                  unsigned SizeInBytes) {
-  uint64_t Addr[] = {dwarf::DW_OP_LLVM_fragment, OffsetInBytes, SizeInBytes};
-  return DIExpression::get(VMContext, Addr);
-}
-
 template <class... Ts>
 static DISubprogram *getSubprogram(bool IsDistinct, Ts &&... Args) {
   if (IsDistinct)
@@ -770,87 +772,88 @@ DILexicalBlock *DIBuilder::createLexicalBlock(DIScope *Scope, DIFile *File,
                                      File, Line, Col);
 }
 
-static Value *getDbgIntrinsicValueImpl(LLVMContext &VMContext, Value *V) {
-  assert(V && "no value passed to dbg intrinsic");
-  return MetadataAsValue::get(VMContext, ValueAsMetadata::get(V));
-}
-
-static Instruction *withDebugLoc(Instruction *I, const DILocation *DL) {
-  I->setDebugLoc(const_cast<DILocation *>(DL));
-  return I;
-}
-
 Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
                                       DIExpression *Expr, const DILocation *DL,
                                       Instruction *InsertBefore) {
-  assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.declare");
-  assert(DL && "Expected debug loc");
-  assert(DL->getScope()->getSubprogram() ==
-             VarInfo->getScope()->getSubprogram() &&
-         "Expected matching subprograms");
-  if (!DeclareFn)
-    DeclareFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_declare);
-
-  trackIfUnresolved(VarInfo);
-  trackIfUnresolved(Expr);
-  Value *Args[] = {getDbgIntrinsicValueImpl(VMContext, Storage),
-                   MetadataAsValue::get(VMContext, VarInfo),
-                   MetadataAsValue::get(VMContext, Expr)};
-  return withDebugLoc(CallInst::Create(DeclareFn, Args, "", InsertBefore), DL);
+  return insertDeclare(Storage, VarInfo, Expr, DL, InsertBefore->getParent(),
+                       InsertBefore);
 }
 
 Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
                                       DIExpression *Expr, const DILocation *DL,
                                       BasicBlock *InsertAtEnd) {
-  assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.declare");
-  assert(DL && "Expected debug loc");
-  assert(DL->getScope()->getSubprogram() ==
-             VarInfo->getScope()->getSubprogram() &&
-         "Expected matching subprograms");
-  if (!DeclareFn)
-    DeclareFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_declare);
-
-  trackIfUnresolved(VarInfo);
-  trackIfUnresolved(Expr);
-  Value *Args[] = {getDbgIntrinsicValueImpl(VMContext, Storage),
-                   MetadataAsValue::get(VMContext, VarInfo),
-                   MetadataAsValue::get(VMContext, Expr)};
-
-  // If this block already has a terminator then insert this intrinsic
-  // before the terminator.
-  if (TerminatorInst *T = InsertAtEnd->getTerminator())
-    return withDebugLoc(CallInst::Create(DeclareFn, Args, "", T), DL);
-  return withDebugLoc(CallInst::Create(DeclareFn, Args, "", InsertAtEnd), DL);
+  // If this block already has a terminator then insert this intrinsic before
+  // the terminator. Otherwise, put it at the end of the block.
+  Instruction *InsertBefore = InsertAtEnd->getTerminator();
+  return insertDeclare(Storage, VarInfo, Expr, DL, InsertAtEnd, InsertBefore);
 }
 
-Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
+Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V,
                                                 DILocalVariable *VarInfo,
                                                 DIExpression *Expr,
                                                 const DILocation *DL,
                                                 Instruction *InsertBefore) {
-  assert(V && "no value passed to dbg.value");
-  assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.value");
+  return insertDbgValueIntrinsic(
+      V, VarInfo, Expr, DL, InsertBefore ? InsertBefore->getParent() : nullptr,
+      InsertBefore);
+}
+
+Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V,
+                                                DILocalVariable *VarInfo,
+                                                DIExpression *Expr,
+                                                const DILocation *DL,
+                                                BasicBlock *InsertAtEnd) {
+  return insertDbgValueIntrinsic(V, VarInfo, Expr, DL, InsertAtEnd, nullptr);
+}
+
+/// Return an IRBuilder for inserting dbg.declare and dbg.value intrinsics. This
+/// abstracts over the various ways to specify an insert position.
+static IRBuilder<> getIRBForDbgInsertion(const DILocation *DL,
+                                         BasicBlock *InsertBB,
+                                         Instruction *InsertBefore) {
+  IRBuilder<> B(DL->getContext());
+  if (InsertBefore)
+    B.SetInsertPoint(InsertBefore);
+  else if (InsertBB)
+    B.SetInsertPoint(InsertBB);
+  B.SetCurrentDebugLocation(DL);
+  return B;
+}
+
+static Value *getDbgIntrinsicValueImpl(LLVMContext &VMContext, Value *V) {
+  assert(V && "no value passed to dbg intrinsic");
+  return MetadataAsValue::get(VMContext, ValueAsMetadata::get(V));
+}
+
+static Function *getDeclareIntrin(Module &M) {
+  return Intrinsic::getDeclaration(&M, UseDbgAddr ? Intrinsic::dbg_addr
+                                                  : Intrinsic::dbg_declare);
+}
+
+Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
+                                      DIExpression *Expr, const DILocation *DL,
+                                      BasicBlock *InsertBB, Instruction *InsertBefore) {
+  assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.declare");
   assert(DL && "Expected debug loc");
   assert(DL->getScope()->getSubprogram() ==
              VarInfo->getScope()->getSubprogram() &&
          "Expected matching subprograms");
-  if (!ValueFn)
-    ValueFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_value);
+  if (!DeclareFn)
+    DeclareFn = getDeclareIntrin(M);
 
   trackIfUnresolved(VarInfo);
   trackIfUnresolved(Expr);
-  Value *Args[] = {getDbgIntrinsicValueImpl(VMContext, V),
-                   ConstantInt::get(Type::getInt64Ty(VMContext), Offset),
+  Value *Args[] = {getDbgIntrinsicValueImpl(VMContext, Storage),
                    MetadataAsValue::get(VMContext, VarInfo),
                    MetadataAsValue::get(VMContext, Expr)};
-  return withDebugLoc(CallInst::Create(ValueFn, Args, "", InsertBefore), DL);
+
+  IRBuilder<> B = getIRBForDbgInsertion(DL, InsertBB, InsertBefore);
+  return B.CreateCall(DeclareFn, Args);
 }
 
-Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
-                                                DILocalVariable *VarInfo,
-                                                DIExpression *Expr,
-                                                const DILocation *DL,
-                                                BasicBlock *InsertAtEnd) {
+Instruction *DIBuilder::insertDbgValueIntrinsic(
+    Value *V, DILocalVariable *VarInfo, DIExpression *Expr,
+    const DILocation *DL, BasicBlock *InsertBB, Instruction *InsertBefore) {
   assert(V && "no value passed to dbg.value");
   assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.value");
   assert(DL && "Expected debug loc");
@@ -863,15 +866,15 @@ Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
   trackIfUnresolved(VarInfo);
   trackIfUnresolved(Expr);
   Value *Args[] = {getDbgIntrinsicValueImpl(VMContext, V),
-                   ConstantInt::get(Type::getInt64Ty(VMContext), Offset),
                    MetadataAsValue::get(VMContext, VarInfo),
                    MetadataAsValue::get(VMContext, Expr)};
 
-  return withDebugLoc(CallInst::Create(ValueFn, Args, "", InsertAtEnd), DL);
+  IRBuilder<> B = getIRBForDbgInsertion(DL, InsertBB, InsertBefore);
+  return B.CreateCall(ValueFn, Args);
 }
 
 void DIBuilder::replaceVTableHolder(DICompositeType *&T,
-                                    DICompositeType *VTableHolder) {
+                                    DIType *VTableHolder) {
   {
     TypedTrackingMDRef<DICompositeType> N(T);
     N->replaceVTableHolder(VTableHolder);
diff --git a/lib/IR/DataLayout.cpp b/lib/IR/DataLayout.cpp
index 5de281a95237..f4dddeb30d0b 100644
--- a/lib/IR/DataLayout.cpp
+++ b/lib/IR/DataLayout.cpp
@@ -572,6 +572,8 @@ const StructLayout *DataLayout::getStructLayout(StructType *Ty) const {
   int NumElts = Ty->getNumElements();
   StructLayout *L =
     (StructLayout *)malloc(sizeof(StructLayout)+(NumElts-1) * sizeof(uint64_t));
+  if (L == nullptr)
+    report_bad_alloc_error("Allocation of StructLayout elements failed.");
 
   // Set SL before calling StructLayout's ctor.  The ctor could cause other
   // entries to be added to TheMap, invalidating our reference.
diff --git a/lib/IR/DebugInfo.cpp b/lib/IR/DebugInfo.cpp
index 56cec57a4d07..7fff7526b926 100644
--- a/lib/IR/DebugInfo.cpp
+++ b/lib/IR/DebugInfo.cpp
@@ -12,10 +12,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "llvm/IR/DebugInfo.h"
+#include "llvm-c/DebugInfo.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/None.h"
+#include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringRef.h"
@@ -23,6 +24,8 @@
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DebugInfoMetadata.h"
 #include "llvm/IR/DebugLoc.h"
+#include "llvm/IR/DebugInfo.h"
+#include "llvm/IR/DIBuilder.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GVMaterializer.h"
 #include "llvm/IR/Instruction.h"
@@ -290,7 +293,7 @@ static MDNode *stripDebugLocFromLoopID(MDNode *N) {
 
 bool llvm::stripDebugInfo(Function &F) {
   bool Changed = false;
-  if (F.getSubprogram()) {
+  if (F.getMetadata(LLVMContext::MD_dbg)) {
     Changed = true;
     F.setSubprogram(nullptr);
   }
@@ -311,6 +314,9 @@ bool llvm::stripDebugInfo(Function &F) {
     }
 
     auto *TermInst = BB.getTerminator();
+    if (!TermInst)
+      // This is invalid IR, but we may not have run the verifier yet
+      continue;
     if (auto *LoopID = TermInst->getMetadata(LLVMContext::MD_loop)) {
       auto *NewLoopID = LoopIDsMap.lookup(LoopID);
       if (!NewLoopID)
@@ -470,7 +476,7 @@ private:
         CU->getSplitDebugFilename(), DICompileUnit::LineTablesOnly, EnumTypes,
         RetainedTypes, GlobalVariables, ImportedEntities, CU->getMacros(),
         CU->getDWOId(), CU->getSplitDebugInlining(),
-        CU->getDebugInfoForProfiling());
+        CU->getDebugInfoForProfiling(), CU->getGnuPubnames());
   }
 
   DILocation *getReplacementMDLocation(DILocation *MLD) {
@@ -666,3 +672,84 @@ unsigned llvm::getDebugMetadataVersionFromModule(const Module &M) {
     return Val->getZExtValue();
   return 0;
 }
+
+void Instruction::applyMergedLocation(const DILocation *LocA,
+                                      const DILocation *LocB) {
+  setDebugLoc(DILocation::getMergedLocation(LocA, LocB, this));
+}
+
+//===----------------------------------------------------------------------===//
+// LLVM C API implementations.
+//===----------------------------------------------------------------------===//
+
+static unsigned map_from_llvmDWARFsourcelanguage(LLVMDWARFSourceLanguage lang) {
+  switch (lang) {
+#define HANDLE_DW_LANG(ID, NAME, VERSION, VENDOR) \
+case LLVMDWARFSourceLanguage##NAME: return ID;
+#include "llvm/BinaryFormat/Dwarf.def"
+#undef HANDLE_DW_LANG
+  }
+  llvm_unreachable("Unhandled Tag");
+}
+
+unsigned LLVMDebugMetadataVersion() {
+  return DEBUG_METADATA_VERSION;
+}
+
+LLVMDIBuilderRef LLVMCreateDIBuilderDisallowUnresolved(LLVMModuleRef M) {
+  return wrap(new DIBuilder(*unwrap(M), false));
+}
+
+LLVMDIBuilderRef LLVMCreateDIBuilder(LLVMModuleRef M) {
+  return wrap(new DIBuilder(*unwrap(M)));
+}
+
+unsigned LLVMGetModuleDebugMetadataVersion(LLVMModuleRef M) {
+  return getDebugMetadataVersionFromModule(*unwrap(M));
+}
+
+LLVMBool LLVMStripModuleDebugInfo(LLVMModuleRef M) {
+  return StripDebugInfo(*unwrap(M));
+}
+
+void LLVMDisposeDIBuilder(LLVMDIBuilderRef Builder) {
+  delete unwrap(Builder);
+}
+
+void LLVMDIBuilderFinalize(LLVMDIBuilderRef Builder) {
+  unwrap(Builder)->finalize();
+}
+
+LLVMMetadataRef LLVMDIBuilderCreateCompileUnit(
+    LLVMDIBuilderRef Builder, LLVMDWARFSourceLanguage Lang,
+    LLVMMetadataRef FileRef, const char *Producer, size_t ProducerLen,
+    LLVMBool isOptimized, const char *Flags, size_t FlagsLen,
+    unsigned RuntimeVer, const char *SplitName, size_t SplitNameLen,
+    LLVMDWARFEmissionKind Kind, unsigned DWOId, LLVMBool SplitDebugInlining,
+    LLVMBool DebugInfoForProfiling) {
+  auto File = unwrap<DIFile>(FileRef);
+
+  return wrap(unwrap(Builder)->createCompileUnit(
+                 map_from_llvmDWARFsourcelanguage(Lang), File,
+                 StringRef(Producer, ProducerLen), isOptimized,
+                 StringRef(Flags, FlagsLen), RuntimeVer,
+                 StringRef(SplitName, SplitNameLen),
+                 static_cast<DICompileUnit::DebugEmissionKind>(Kind), DWOId,
+                 SplitDebugInlining, DebugInfoForProfiling));
+}
+
+LLVMMetadataRef
+LLVMDIBuilderCreateFile(LLVMDIBuilderRef Builder, const char *Filename,
+                        size_t FilenameLen, const char *Directory,
+                        size_t DirectoryLen) {
+  return wrap(unwrap(Builder)->createFile(StringRef(Filename, FilenameLen),
+                                          StringRef(Directory, DirectoryLen)));
+}
+
+LLVMMetadataRef
+LLVMDIBuilderCreateDebugLocation(LLVMContextRef Ctx, unsigned Line,
+                                 unsigned Column, LLVMMetadataRef Scope,
+                                 LLVMMetadataRef InlinedAt) {
+  return wrap(DILocation::get(*unwrap(Ctx), Line, Column, unwrap(Scope),
+                              unwrap(InlinedAt)));
+}
diff --git a/lib/IR/DebugInfoMetadata.cpp b/lib/IR/DebugInfoMetadata.cpp
index c14940bad45d..75ddd47b2591 100644
--- a/lib/IR/DebugInfoMetadata.cpp
+++ b/lib/IR/DebugInfoMetadata.cpp
@@ -14,9 +14,11 @@
 #include "llvm/IR/DebugInfoMetadata.h"
 #include "LLVMContextImpl.h"
 #include "MetadataImpl.h"
+#include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/StringSwitch.h"
 #include "llvm/IR/DIBuilder.h"
 #include "llvm/IR/Function.h"
+#include "llvm/IR/Instructions.h"
 
 using namespace llvm;
 
@@ -66,6 +68,31 @@ DILocation *DILocation::getImpl(LLVMContext &Context, unsigned Line,
                    Storage, Context.pImpl->DILocations);
 }
 
+const DILocation *
+DILocation::getMergedLocation(const DILocation *LocA, const DILocation *LocB,
+                              const Instruction *ForInst) {
+  if (!LocA || !LocB)
+    return nullptr;
+
+  if (LocA == LocB || !LocA->canDiscriminate(*LocB))
+    return LocA;
+
+  if (!dyn_cast_or_null<CallInst>(ForInst))
+    return nullptr;
+
+  SmallPtrSet<DILocation *, 5> InlinedLocationsA;
+  for (DILocation *L = LocA->getInlinedAt(); L; L = L->getInlinedAt())
+    InlinedLocationsA.insert(L);
+  const DILocation *Result = LocB;
+  for (DILocation *L = LocB->getInlinedAt(); L; L = L->getInlinedAt()) {
+    Result = L;
+    if (InlinedLocationsA.count(L))
+      break;
+  }
+  return DILocation::get(Result->getContext(), 0, 0, Result->getScope(),
+                         Result->getInlinedAt());
+}
+
 DINode::DIFlags DINode::getFlag(StringRef Flag) {
   return StringSwitch<DIFlags>(Flag)
 #define HANDLE_DI_FLAG(ID, NAME) .Case("DIFlag" #NAME, Flag##NAME)
@@ -354,6 +381,8 @@ DISubroutineType *DISubroutineType::getImpl(LLVMContext &Context, DIFlags Flags,
   DEFINE_GETIMPL_STORE(DISubroutineType, (Flags, CC), Ops);
 }
 
+// FIXME: Implement this string-enum correspondence with a .def file and macros,
+// so that the association is explicit rather than implied.
 static const char *ChecksumKindName[DIFile::CSK_Last + 1] = {
   "CSK_None",
   "CSK_MD5",
@@ -391,7 +420,7 @@ DICompileUnit *DICompileUnit::getImpl(
     unsigned EmissionKind, Metadata *EnumTypes, Metadata *RetainedTypes,
     Metadata *GlobalVariables, Metadata *ImportedEntities, Metadata *Macros,
     uint64_t DWOId, bool SplitDebugInlining, bool DebugInfoForProfiling,
-    StorageType Storage, bool ShouldCreate) {
+    bool GnuPubnames, StorageType Storage, bool ShouldCreate) {
   assert(Storage != Uniqued && "Cannot unique DICompileUnit");
   assert(isCanonical(Producer) && "Expected canonical MDString");
   assert(isCanonical(Flags) && "Expected canonical MDString");
@@ -401,11 +430,10 @@ DICompileUnit *DICompileUnit::getImpl(
       File,      Producer,      Flags,           SplitDebugFilename,
       EnumTypes, RetainedTypes, GlobalVariables, ImportedEntities,
       Macros};
-  return storeImpl(new (array_lengthof(Ops))
-                       DICompileUnit(Context, Storage, SourceLanguage,
-                                     IsOptimized, RuntimeVersion, EmissionKind,
-                                     DWOId, SplitDebugInlining,
-                                     DebugInfoForProfiling, Ops),
+  return storeImpl(new (array_lengthof(Ops)) DICompileUnit(
+                       Context, Storage, SourceLanguage, IsOptimized,
+                       RuntimeVersion, EmissionKind, DWOId, SplitDebugInlining,
+                       DebugInfoForProfiling, GnuPubnames, Ops),
                    Storage);
 }
 
@@ -586,6 +614,29 @@ DILocalVariable *DILocalVariable::getImpl(LLVMContext &Context, Metadata *Scope,
   DEFINE_GETIMPL_STORE(DILocalVariable, (Line, Arg, Flags, AlignInBits), Ops);
 }
 
+Optional<uint64_t> DIVariable::getSizeInBits() const {
+  // This is used by the Verifier so be mindful of broken types.
+  const Metadata *RawType = getRawType();
+  while (RawType) {
+    // Try to get the size directly.
+    if (auto *T = dyn_cast<DIType>(RawType))
+      if (uint64_t Size = T->getSizeInBits())
+        return Size;
+
+    if (auto *DT = dyn_cast<DIDerivedType>(RawType)) {
+      // Look at the base type.
+      RawType = DT->getRawBaseType();
+      continue;
+    }
+
+    // Missing type or size.
+    break;
+  }
+
+  // Fail gracefully.
+  return None;
+}
+
 DIExpression *DIExpression::getImpl(LLVMContext &Context,
                                     ArrayRef<uint64_t> Elements,
                                     StorageType Storage, bool ShouldCreate) {
@@ -643,6 +694,7 @@ bool DIExpression::isValid() const {
     case dwarf::DW_OP_plus_uconst:
     case dwarf::DW_OP_plus:
     case dwarf::DW_OP_minus:
+    case dwarf::DW_OP_mul:
     case dwarf::DW_OP_deref:
     case dwarf::DW_OP_xderef:
       break;
@@ -698,12 +750,17 @@ bool DIExpression::extractIfOffset(int64_t &Offset) const {
   return false;
 }
 
-DIExpression *DIExpression::prepend(const DIExpression *Expr, bool Deref,
-                                    int64_t Offset, bool StackValue) {
+DIExpression *DIExpression::prepend(const DIExpression *Expr, bool DerefBefore,
+                                    int64_t Offset, bool DerefAfter,
+                                    bool StackValue) {
   SmallVector<uint64_t, 8> Ops;
+  if (DerefBefore)
+    Ops.push_back(dwarf::DW_OP_deref);
+  
   appendOffset(Ops, Offset);
-  if (Deref)
+  if (DerefAfter)
     Ops.push_back(dwarf::DW_OP_deref);
+
   if (Expr)
     for (auto Op : Expr->expr_ops()) {
       // A DW_OP_stack_value comes at the end, but before a DW_OP_LLVM_fragment.
@@ -724,6 +781,44 @@ DIExpression *DIExpression::prepend(const DIExpression *Expr, bool Deref,
   return DIExpression::get(Expr->getContext(), Ops);
 }
 
+Optional<DIExpression *> DIExpression::createFragmentExpression(
+    const DIExpression *Expr, unsigned OffsetInBits, unsigned SizeInBits) {
+  SmallVector<uint64_t, 8> Ops;
+  // Copy over the expression, but leave off any trailing DW_OP_LLVM_fragment.
+  if (Expr) {
+    for (auto Op : Expr->expr_ops()) {
+      switch (Op.getOp()) {
+      default: break;
+      case dwarf::DW_OP_plus:
+      case dwarf::DW_OP_minus:
+        // We can't safely split arithmetic into multiple fragments because we
+        // can't express carry-over between fragments.
+        //
+        // FIXME: We *could* preserve the lowest fragment of a constant offset
+        // operation if the offset fits into SizeInBits.
+        return None;
+      case dwarf::DW_OP_LLVM_fragment: {
+        // Make the new offset point into the existing fragment.
+        uint64_t FragmentOffsetInBits = Op.getArg(0);
+        // Op.getArg(0) is FragmentOffsetInBits.
+        // Op.getArg(1) is FragmentSizeInBits.
+        assert((OffsetInBits + SizeInBits <= Op.getArg(0) + Op.getArg(1)) &&
+               "new fragment outside of original fragment");
+        OffsetInBits += FragmentOffsetInBits;
+        continue;
+      }
+      }
+      Ops.push_back(Op.getOp());
+      for (unsigned I = 0; I < Op.getNumArgs(); ++I)
+        Ops.push_back(Op.getArg(I));
+    }
+  }
+  Ops.push_back(dwarf::DW_OP_LLVM_fragment);
+  Ops.push_back(OffsetInBits);
+  Ops.push_back(SizeInBits);
+  return DIExpression::get(Expr->getContext(), Ops);
+}
+
 bool DIExpression::isConstant() const {
   // Recognize DW_OP_constu C DW_OP_stack_value (DW_OP_LLVM_fragment Len Ofs)?.
   if (getNumElements() != 3 && getNumElements() != 6)
diff --git a/lib/IR/DebugLoc.cpp b/lib/IR/DebugLoc.cpp
index 6297395b4c00..0a494119c3fe 100644
--- a/lib/IR/DebugLoc.cpp
+++ b/lib/IR/DebugLoc.cpp
@@ -10,7 +10,6 @@
 #include "llvm/IR/DebugLoc.h"
 #include "LLVMContextImpl.h"
 #include "llvm/IR/DebugInfo.h"
-#include "llvm/IR/IntrinsicInst.h"
 using namespace llvm;
 
 //===----------------------------------------------------------------------===//
diff --git a/lib/IR/DiagnosticHandler.cpp b/lib/IR/DiagnosticHandler.cpp
new file mode 100644
index 000000000000..fb1ac438ffbe
--- /dev/null
+++ b/lib/IR/DiagnosticHandler.cpp
@@ -0,0 +1,91 @@
+//===- DiagnosticHandler.h - DiagnosticHandler class for LLVM -------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+//
+//===----------------------------------------------------------------------===//
+#include "llvm/IR/DiagnosticHandler.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Regex.h"
+
+using namespace llvm;
+
+namespace {
+
+/// \brief Regular expression corresponding to the value given in one of the
+/// -pass-remarks* command line flags. Passes whose name matches this regexp
+/// will emit a diagnostic when calling the associated diagnostic function
+/// (emitOptimizationRemark, emitOptimizationRemarkMissed or
+/// emitOptimizationRemarkAnalysis).
+struct PassRemarksOpt {
+  std::shared_ptr<Regex> Pattern;
+
+  void operator=(const std::string &Val) {
+    // Create a regexp object to match pass names for emitOptimizationRemark.
+    if (!Val.empty()) {
+      Pattern = std::make_shared<Regex>(Val);
+      std::string RegexError;
+      if (!Pattern->isValid(RegexError))
+        report_fatal_error("Invalid regular expression '" + Val +
+                               "' in -pass-remarks: " + RegexError,
+                           false);
+    }
+  }
+};
+
+static PassRemarksOpt PassRemarksPassedOptLoc;
+static PassRemarksOpt PassRemarksMissedOptLoc;
+static PassRemarksOpt PassRemarksAnalysisOptLoc;
+
+// -pass-remarks
+//    Command line flag to enable emitOptimizationRemark()
+static cl::opt<PassRemarksOpt, true, cl::parser<std::string>> PassRemarks(
+    "pass-remarks", cl::value_desc("pattern"),
+    cl::desc("Enable optimization remarks from passes whose name match "
+             "the given regular expression"),
+    cl::Hidden, cl::location(PassRemarksPassedOptLoc), cl::ValueRequired,
+    cl::ZeroOrMore);
+
+// -pass-remarks-missed
+//    Command line flag to enable emitOptimizationRemarkMissed()
+static cl::opt<PassRemarksOpt, true, cl::parser<std::string>> PassRemarksMissed(
+    "pass-remarks-missed", cl::value_desc("pattern"),
+    cl::desc("Enable missed optimization remarks from passes whose name match "
+             "the given regular expression"),
+    cl::Hidden, cl::location(PassRemarksMissedOptLoc), cl::ValueRequired,
+    cl::ZeroOrMore);
+
+// -pass-remarks-analysis
+//    Command line flag to enable emitOptimizationRemarkAnalysis()
+static cl::opt<PassRemarksOpt, true, cl::parser<std::string>>
+    PassRemarksAnalysis(
+        "pass-remarks-analysis", cl::value_desc("pattern"),
+        cl::desc(
+            "Enable optimization analysis remarks from passes whose name match "
+            "the given regular expression"),
+        cl::Hidden, cl::location(PassRemarksAnalysisOptLoc), cl::ValueRequired,
+        cl::ZeroOrMore);
+}
+
+bool DiagnosticHandler::isAnalysisRemarkEnabled(StringRef PassName) const {
+  return (PassRemarksAnalysisOptLoc.Pattern &&
+          PassRemarksAnalysisOptLoc.Pattern->match(PassName));
+}
+bool DiagnosticHandler::isMissedOptRemarkEnabled(StringRef PassName) const {
+  return (PassRemarksMissedOptLoc.Pattern &&
+          PassRemarksMissedOptLoc.Pattern->match(PassName));
+}
+bool DiagnosticHandler::isPassedOptRemarkEnabled(StringRef PassName) const {
+  return (PassRemarksPassedOptLoc.Pattern &&
+          PassRemarksPassedOptLoc.Pattern->match(PassName));
+}
+
+bool DiagnosticHandler::isAnyRemarkEnabled() const {
+  return (PassRemarksPassedOptLoc.Pattern || PassRemarksMissedOptLoc.Pattern ||
+          PassRemarksAnalysisOptLoc.Pattern);
+}
diff --git a/lib/IR/DiagnosticInfo.cpp b/lib/IR/DiagnosticInfo.cpp
index 5129d6b9b008..946df1a836ce 100644
--- a/lib/IR/DiagnosticInfo.cpp
+++ b/lib/IR/DiagnosticInfo.cpp
@@ -13,6 +13,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/IR/DiagnosticInfo.h"
+#include "LLVMContextImpl.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/Twine.h"
 #include "llvm/ADT/iterator_range.h"
@@ -41,64 +42,6 @@
 
 using namespace llvm;
 
-namespace {
-
-/// \brief Regular expression corresponding to the value given in one of the
-/// -pass-remarks* command line flags. Passes whose name matches this regexp
-/// will emit a diagnostic when calling the associated diagnostic function
-/// (emitOptimizationRemark, emitOptimizationRemarkMissed or
-/// emitOptimizationRemarkAnalysis).
-struct PassRemarksOpt {
-  std::shared_ptr<Regex> Pattern;
-
-  void operator=(const std::string &Val) {
-    // Create a regexp object to match pass names for emitOptimizationRemark.
-    if (!Val.empty()) {
-      Pattern = std::make_shared<Regex>(Val);
-      std::string RegexError;
-      if (!Pattern->isValid(RegexError))
-        report_fatal_error("Invalid regular expression '" + Val +
-                               "' in -pass-remarks: " + RegexError,
-                           false);
-    }
-  }
-};
-
-} // end anonymous namespace
-
-static PassRemarksOpt PassRemarksOptLoc;
-static PassRemarksOpt PassRemarksMissedOptLoc;
-static PassRemarksOpt PassRemarksAnalysisOptLoc;
-
-// -pass-remarks
-//    Command line flag to enable emitOptimizationRemark()
-static cl::opt<PassRemarksOpt, true, cl::parser<std::string>>
-PassRemarks("pass-remarks", cl::value_desc("pattern"),
-            cl::desc("Enable optimization remarks from passes whose name match "
-                     "the given regular expression"),
-            cl::Hidden, cl::location(PassRemarksOptLoc), cl::ValueRequired,
-            cl::ZeroOrMore);
-
-// -pass-remarks-missed
-//    Command line flag to enable emitOptimizationRemarkMissed()
-static cl::opt<PassRemarksOpt, true, cl::parser<std::string>> PassRemarksMissed(
-    "pass-remarks-missed", cl::value_desc("pattern"),
-    cl::desc("Enable missed optimization remarks from passes whose name match "
-             "the given regular expression"),
-    cl::Hidden, cl::location(PassRemarksMissedOptLoc), cl::ValueRequired,
-    cl::ZeroOrMore);
-
-// -pass-remarks-analysis
-//    Command line flag to enable emitOptimizationRemarkAnalysis()
-static cl::opt<PassRemarksOpt, true, cl::parser<std::string>>
-PassRemarksAnalysis(
-    "pass-remarks-analysis", cl::value_desc("pattern"),
-    cl::desc(
-        "Enable optimization analysis remarks from passes whose name match "
-        "the given regular expression"),
-    cl::Hidden, cl::location(PassRemarksAnalysisOptLoc), cl::ValueRequired,
-    cl::ZeroOrMore);
-
 int llvm::getNextAvailablePluginDiagnosticKind() {
   static std::atomic<int> PluginKindID(DK_FirstPluginKind);
   return ++PluginKindID;
@@ -218,12 +161,39 @@ DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, const Type *T)
   OS << *T;
 }
 
+DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, StringRef S)
+    : Key(Key), Val(S.str()) {}
+
 DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, int N)
     : Key(Key), Val(itostr(N)) {}
 
+DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, long N)
+    : Key(Key), Val(itostr(N)) {}
+
+DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, long long N)
+    : Key(Key), Val(itostr(N)) {}
+
 DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, unsigned N)
     : Key(Key), Val(utostr(N)) {}
 
+DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key,
+                                                   unsigned long N)
+    : Key(Key), Val(utostr(N)) {}
+
+DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key,
+                                                   unsigned long long N)
+    : Key(Key), Val(utostr(N)) {}
+
+DiagnosticInfoOptimizationBase::Argument::Argument(StringRef Key, DebugLoc Loc)
+    : Key(Key), Loc(Loc) {
+  if (Loc) {
+    Val = (Loc->getFilename() + ":" + Twine(Loc.getLine()) + ":" +
+           Twine(Loc.getCol())).str();
+  } else {
+    Val = "<UNKNOWN LOCATION>";
+  }
+}
+
 void DiagnosticInfoOptimizationBase::print(DiagnosticPrinter &DP) const {
   DP << getLocationStr() << ": " << getMsg();
   if (Hotness)
@@ -259,9 +229,10 @@ OptimizationRemark::OptimizationRemark(const char *PassName,
                                    RemarkName, *Func, Func->getSubprogram(),
                                    &getFirstFunctionBlock(Func)) {}
 
-bool OptimizationRemark::isEnabled(StringRef PassName) {
-  return PassRemarksOptLoc.Pattern &&
-         PassRemarksOptLoc.Pattern->match(PassName);
+bool OptimizationRemark::isEnabled() const {
+  const Function &Fn = getFunction();
+  LLVMContext &Ctx = Fn.getContext();
+  return Ctx.getDiagHandlerPtr()->isPassedOptRemarkEnabled(getPassName());
 }
 
 OptimizationRemarkMissed::OptimizationRemarkMissed(
@@ -279,9 +250,10 @@ OptimizationRemarkMissed::OptimizationRemarkMissed(const char *PassName,
                                    *Inst->getParent()->getParent(),
                                    Inst->getDebugLoc(), Inst->getParent()) {}
 
-bool OptimizationRemarkMissed::isEnabled(StringRef PassName) {
-  return PassRemarksMissedOptLoc.Pattern &&
-         PassRemarksMissedOptLoc.Pattern->match(PassName);
+bool OptimizationRemarkMissed::isEnabled() const {
+  const Function &Fn = getFunction();
+  LLVMContext &Ctx = Fn.getContext();
+  return Ctx.getDiagHandlerPtr()->isMissedOptRemarkEnabled(getPassName());
 }
 
 OptimizationRemarkAnalysis::OptimizationRemarkAnalysis(
@@ -306,51 +278,17 @@ OptimizationRemarkAnalysis::OptimizationRemarkAnalysis(
                                    *cast<BasicBlock>(CodeRegion)->getParent(),
                                    Loc, CodeRegion) {}
 
-bool OptimizationRemarkAnalysis::isEnabled(StringRef PassName) {
-  return PassRemarksAnalysisOptLoc.Pattern &&
-         PassRemarksAnalysisOptLoc.Pattern->match(PassName);
+bool OptimizationRemarkAnalysis::isEnabled() const {
+  const Function &Fn = getFunction();
+  LLVMContext &Ctx = Fn.getContext();
+  return Ctx.getDiagHandlerPtr()->isAnalysisRemarkEnabled(getPassName()) ||
+         shouldAlwaysPrint();
 }
 
 void DiagnosticInfoMIRParser::print(DiagnosticPrinter &DP) const {
   DP << Diagnostic;
 }
 
-void llvm::emitOptimizationRemark(LLVMContext &Ctx, const char *PassName,
-                                  const Function &Fn,
-                                  const DiagnosticLocation &Loc,
-                                  const Twine &Msg) {
-  Ctx.diagnose(OptimizationRemark(PassName, Fn, Loc, Msg));
-}
-
-void llvm::emitOptimizationRemarkMissed(LLVMContext &Ctx, const char *PassName,
-                                        const Function &Fn,
-                                        const DiagnosticLocation &Loc,
-                                        const Twine &Msg) {
-  Ctx.diagnose(OptimizationRemarkMissed(PassName, Fn, Loc, Msg));
-}
-
-void llvm::emitOptimizationRemarkAnalysis(LLVMContext &Ctx,
-                                          const char *PassName,
-                                          const Function &Fn,
-                                          const DiagnosticLocation &Loc,
-                                          const Twine &Msg) {
-  Ctx.diagnose(OptimizationRemarkAnalysis(PassName, Fn, Loc, Msg));
-}
-
-void llvm::emitOptimizationRemarkAnalysisFPCommute(
-    LLVMContext &Ctx, const char *PassName, const Function &Fn,
-    const DiagnosticLocation &Loc, const Twine &Msg) {
-  Ctx.diagnose(OptimizationRemarkAnalysisFPCommute(PassName, Fn, Loc, Msg));
-}
-
-void llvm::emitOptimizationRemarkAnalysisAliasing(LLVMContext &Ctx,
-                                                  const char *PassName,
-                                                  const Function &Fn,
-                                                  const DiagnosticLocation &Loc,
-                                                  const Twine &Msg) {
-  Ctx.diagnose(OptimizationRemarkAnalysisAliasing(PassName, Fn, Loc, Msg));
-}
-
 DiagnosticInfoOptimizationFailure::DiagnosticInfoOptimizationFailure(
     const char *PassName, StringRef RemarkName, const DiagnosticLocation &Loc,
     const Value *CodeRegion)
@@ -377,28 +315,20 @@ void DiagnosticInfoISelFallback::print(DiagnosticPrinter &DP) const {
   DP << "Instruction selection used fallback path for " << getFunction();
 }
 
-DiagnosticInfoOptimizationBase &DiagnosticInfoOptimizationBase::
-operator<<(StringRef S) {
+void DiagnosticInfoOptimizationBase::insert(StringRef S) {
   Args.emplace_back(S);
-  return *this;
 }
 
-DiagnosticInfoOptimizationBase &DiagnosticInfoOptimizationBase::
-operator<<(Argument A) {
+void DiagnosticInfoOptimizationBase::insert(Argument A) {
   Args.push_back(std::move(A));
-  return *this;
 }
 
-DiagnosticInfoOptimizationBase &DiagnosticInfoOptimizationBase::
-operator<<(setIsVerbose V) {
+void DiagnosticInfoOptimizationBase::insert(setIsVerbose V) {
   IsVerbose = true;
-  return *this;
 }
 
-DiagnosticInfoOptimizationBase &DiagnosticInfoOptimizationBase::
-operator<<(setExtraArgs EA) {
+void DiagnosticInfoOptimizationBase::insert(setExtraArgs EA) {
   FirstExtraArgIndex = Args.size();
-  return *this;
 }
 
 std::string DiagnosticInfoOptimizationBase::getMsg() const {
@@ -411,3 +341,83 @@ std::string DiagnosticInfoOptimizationBase::getMsg() const {
     OS << Arg.Val;
   return OS.str();
 }
+
+namespace llvm {
+namespace yaml {
+
+void MappingTraits<DiagnosticInfoOptimizationBase *>::mapping(
+    IO &io, DiagnosticInfoOptimizationBase *&OptDiag) {
+  assert(io.outputting() && "input not yet implemented");
+
+  if (io.mapTag("!Passed",
+                (OptDiag->getKind() == DK_OptimizationRemark ||
+                 OptDiag->getKind() == DK_MachineOptimizationRemark)))
+    ;
+  else if (io.mapTag(
+               "!Missed",
+               (OptDiag->getKind() == DK_OptimizationRemarkMissed ||
+                OptDiag->getKind() == DK_MachineOptimizationRemarkMissed)))
+    ;
+  else if (io.mapTag(
+               "!Analysis",
+               (OptDiag->getKind() == DK_OptimizationRemarkAnalysis ||
+                OptDiag->getKind() == DK_MachineOptimizationRemarkAnalysis)))
+    ;
+  else if (io.mapTag("!AnalysisFPCommute",
+                     OptDiag->getKind() ==
+                         DK_OptimizationRemarkAnalysisFPCommute))
+    ;
+  else if (io.mapTag("!AnalysisAliasing",
+                     OptDiag->getKind() ==
+                         DK_OptimizationRemarkAnalysisAliasing))
+    ;
+  else if (io.mapTag("!Failure", OptDiag->getKind() == DK_OptimizationFailure))
+    ;
+  else
+    llvm_unreachable("Unknown remark type");
+
+  // These are read-only for now.
+  DiagnosticLocation DL = OptDiag->getLocation();
+  StringRef FN =
+      GlobalValue::dropLLVMManglingEscape(OptDiag->getFunction().getName());
+
+  StringRef PassName(OptDiag->PassName);
+  io.mapRequired("Pass", PassName);
+  io.mapRequired("Name", OptDiag->RemarkName);
+  if (!io.outputting() || DL.isValid())
+    io.mapOptional("DebugLoc", DL);
+  io.mapRequired("Function", FN);
+  io.mapOptional("Hotness", OptDiag->Hotness);
+  io.mapOptional("Args", OptDiag->Args);
+}
+
+template <> struct MappingTraits<DiagnosticLocation> {
+  static void mapping(IO &io, DiagnosticLocation &DL) {
+    assert(io.outputting() && "input not yet implemented");
+
+    StringRef File = DL.getFilename();
+    unsigned Line = DL.getLine();
+    unsigned Col = DL.getColumn();
+
+    io.mapRequired("File", File);
+    io.mapRequired("Line", Line);
+    io.mapRequired("Column", Col);
+  }
+
+  static const bool flow = true;
+};
+
+// Implement this as a mapping for now to get proper quotation for the value.
+template <> struct MappingTraits<DiagnosticInfoOptimizationBase::Argument> {
+  static void mapping(IO &io, DiagnosticInfoOptimizationBase::Argument &A) {
+    assert(io.outputting() && "input not yet implemented");
+    io.mapRequired(A.Key.data(), A.Val);
+    if (A.Loc.isValid())
+      io.mapOptional("DebugLoc", A.Loc);
+  }
+};
+
+} // end namespace yaml
+} // end namespace llvm
+
+LLVM_YAML_IS_SEQUENCE_VECTOR(DiagnosticInfoOptimizationBase::Argument)
diff --git a/lib/IR/Dominators.cpp b/lib/IR/Dominators.cpp
index 4d7e3040ecd7..ad448a3f240c 100644
--- a/lib/IR/Dominators.cpp
+++ b/lib/IR/Dominators.cpp
@@ -33,9 +33,9 @@ bool llvm::VerifyDomInfo = true;
 #else
 bool llvm::VerifyDomInfo = false;
 #endif
-static cl::opt<bool,true>
-VerifyDomInfoX("verify-dom-info", cl::location(VerifyDomInfo),
-               cl::desc("Verify dominator info (time consuming)"));
+static cl::opt<bool, true>
+    VerifyDomInfoX("verify-dom-info", cl::location(VerifyDomInfo), cl::Hidden,
+                   cl::desc("Verify dominator info (time consuming)"));
 
 bool BasicBlockEdge::isSingleEdge() const {
   const TerminatorInst *TI = Start->getTerminator();
@@ -64,12 +64,12 @@ template class llvm::DomTreeNodeBase<BasicBlock>;
 template class llvm::DominatorTreeBase<BasicBlock, false>; // DomTreeBase
 template class llvm::DominatorTreeBase<BasicBlock, true>; // PostDomTreeBase
 
-template void
-llvm::DomTreeBuilder::Calculate<DomTreeBuilder::BBDomTree, Function>(
-    DomTreeBuilder::BBDomTree &DT, Function &F);
-template void
-llvm::DomTreeBuilder::Calculate<DomTreeBuilder::BBPostDomTree, Function>(
-    DomTreeBuilder::BBPostDomTree &DT, Function &F);
+template struct llvm::DomTreeBuilder::Update<BasicBlock *>;
+
+template void llvm::DomTreeBuilder::Calculate<DomTreeBuilder::BBDomTree>(
+    DomTreeBuilder::BBDomTree &DT);
+template void llvm::DomTreeBuilder::Calculate<DomTreeBuilder::BBPostDomTree>(
+    DomTreeBuilder::BBPostDomTree &DT);
 
 template void llvm::DomTreeBuilder::InsertEdge<DomTreeBuilder::BBDomTree>(
     DomTreeBuilder::BBDomTree &DT, BasicBlock *From, BasicBlock *To);
@@ -81,6 +81,11 @@ template void llvm::DomTreeBuilder::DeleteEdge<DomTreeBuilder::BBDomTree>(
 template void llvm::DomTreeBuilder::DeleteEdge<DomTreeBuilder::BBPostDomTree>(
     DomTreeBuilder::BBPostDomTree &DT, BasicBlock *From, BasicBlock *To);
 
+template void llvm::DomTreeBuilder::ApplyUpdates<DomTreeBuilder::BBDomTree>(
+    DomTreeBuilder::BBDomTree &DT, DomTreeBuilder::BBUpdates);
+template void llvm::DomTreeBuilder::ApplyUpdates<DomTreeBuilder::BBPostDomTree>(
+    DomTreeBuilder::BBPostDomTree &DT, DomTreeBuilder::BBUpdates);
+
 template bool llvm::DomTreeBuilder::Verify<DomTreeBuilder::BBDomTree>(
     const DomTreeBuilder::BBDomTree &DT);
 template bool llvm::DomTreeBuilder::Verify<DomTreeBuilder::BBPostDomTree>(
@@ -310,10 +315,14 @@ void DominatorTree::verifyDomTree() const {
   DominatorTree OtherDT;
   OtherDT.recalculate(F);
   if (compare(OtherDT)) {
-    errs() << "DominatorTree is not up to date!\nComputed:\n";
+    errs() << "DominatorTree for function " << F.getName()
+           << " is not up to date!\nComputed:\n";
     print(errs());
     errs() << "\nActual:\n";
     OtherDT.print(errs());
+    errs() << "\nCFG:\n";
+    F.print(errs());
+    errs().flush();
     abort();
   }
 }
diff --git a/lib/IR/Function.cpp b/lib/IR/Function.cpp
index 85a019856c01..1fff912ecf2f 100644
--- a/lib/IR/Function.cpp
+++ b/lib/IR/Function.cpp
@@ -12,7 +12,6 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/IR/Function.h"
-#include "LLVMContextImpl.h"
 #include "SymbolTableListTraitsImpl.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseSet.h"
@@ -649,7 +648,10 @@ enum IIT_Info {
   IIT_VEC_OF_ANYPTRS_TO_ELT = 34,
   IIT_I128 = 35,
   IIT_V512 = 36,
-  IIT_V1024 = 37
+  IIT_V1024 = 37,
+  IIT_STRUCT6 = 38,
+  IIT_STRUCT7 = 39,
+  IIT_STRUCT8 = 40
 };
 
 static void DecodeIITType(unsigned &NextElt, ArrayRef<unsigned char> Infos,
@@ -798,6 +800,9 @@ static void DecodeIITType(unsigned &NextElt, ArrayRef<unsigned char> Infos,
   case IIT_EMPTYSTRUCT:
     OutputTable.push_back(IITDescriptor::get(IITDescriptor::Struct, 0));
     return;
+  case IIT_STRUCT8: ++StructElts; LLVM_FALLTHROUGH;
+  case IIT_STRUCT7: ++StructElts; LLVM_FALLTHROUGH;
+  case IIT_STRUCT6: ++StructElts; LLVM_FALLTHROUGH;
   case IIT_STRUCT5: ++StructElts; LLVM_FALLTHROUGH;
   case IIT_STRUCT4: ++StructElts; LLVM_FALLTHROUGH;
   case IIT_STRUCT3: ++StructElts; LLVM_FALLTHROUGH;
@@ -874,11 +879,10 @@ static Type *DecodeFixedType(ArrayRef<Intrinsic::IITDescriptor> &Infos,
     return PointerType::get(DecodeFixedType(Infos, Tys, Context),
                             D.Pointer_AddressSpace);
   case IITDescriptor::Struct: {
-    Type *Elts[5];
-    assert(D.Struct_NumElements <= 5 && "Can't handle this yet");
+    SmallVector<Type *, 8> Elts;
     for (unsigned i = 0, e = D.Struct_NumElements; i != e; ++i)
-      Elts[i] = DecodeFixedType(Infos, Tys, Context);
-    return StructType::get(Context, makeArrayRef(Elts,D.Struct_NumElements));
+      Elts.push_back(DecodeFixedType(Infos, Tys, Context));
+    return StructType::get(Context, Elts);
   }
   case IITDescriptor::Argument:
     return Tys[D.getArgumentNumber()];
diff --git a/lib/IR/GCOV.cpp b/lib/IR/GCOV.cpp
deleted file mode 100644
index d4b455228225..000000000000
--- a/lib/IR/GCOV.cpp
+++ /dev/null
@@ -1,821 +0,0 @@
-//===- GCOV.cpp - LLVM coverage tool --------------------------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// GCOV implements the interface to read and write coverage files that use
-// 'gcov' format.
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/Support/GCOV.h"
-#include "llvm/ADT/STLExtras.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Support/FileSystem.h"
-#include "llvm/Support/Format.h"
-#include "llvm/Support/Path.h"
-#include "llvm/Support/raw_ostream.h"
-#include <algorithm>
-#include <system_error>
-
-using namespace llvm;
-
-//===----------------------------------------------------------------------===//
-// GCOVFile implementation.
-
-/// readGCNO - Read GCNO buffer.
-bool GCOVFile::readGCNO(GCOVBuffer &Buffer) {
-  if (!Buffer.readGCNOFormat())
-    return false;
-  if (!Buffer.readGCOVVersion(Version))
-    return false;
-
-  if (!Buffer.readInt(Checksum))
-    return false;
-  while (true) {
-    if (!Buffer.readFunctionTag())
-      break;
-    auto GFun = make_unique<GCOVFunction>(*this);
-    if (!GFun->readGCNO(Buffer, Version))
-      return false;
-    Functions.push_back(std::move(GFun));
-  }
-
-  GCNOInitialized = true;
-  return true;
-}
-
-/// readGCDA - Read GCDA buffer. It is required that readGCDA() can only be
-/// called after readGCNO().
-bool GCOVFile::readGCDA(GCOVBuffer &Buffer) {
-  assert(GCNOInitialized && "readGCDA() can only be called after readGCNO()");
-  if (!Buffer.readGCDAFormat())
-    return false;
-  GCOV::GCOVVersion GCDAVersion;
-  if (!Buffer.readGCOVVersion(GCDAVersion))
-    return false;
-  if (Version != GCDAVersion) {
-    errs() << "GCOV versions do not match.\n";
-    return false;
-  }
-
-  uint32_t GCDAChecksum;
-  if (!Buffer.readInt(GCDAChecksum))
-    return false;
-  if (Checksum != GCDAChecksum) {
-    errs() << "File checksums do not match: " << Checksum
-           << " != " << GCDAChecksum << ".\n";
-    return false;
-  }
-  for (size_t i = 0, e = Functions.size(); i < e; ++i) {
-    if (!Buffer.readFunctionTag()) {
-      errs() << "Unexpected number of functions.\n";
-      return false;
-    }
-    if (!Functions[i]->readGCDA(Buffer, Version))
-      return false;
-  }
-  if (Buffer.readObjectTag()) {
-    uint32_t Length;
-    uint32_t Dummy;
-    if (!Buffer.readInt(Length))
-      return false;
-    if (!Buffer.readInt(Dummy))
-      return false; // checksum
-    if (!Buffer.readInt(Dummy))
-      return false; // num
-    if (!Buffer.readInt(RunCount))
-      return false;
-    Buffer.advanceCursor(Length - 3);
-  }
-  while (Buffer.readProgramTag()) {
-    uint32_t Length;
-    if (!Buffer.readInt(Length))
-      return false;
-    Buffer.advanceCursor(Length);
-    ++ProgramCount;
-  }
-
-  return true;
-}
-
-void GCOVFile::print(raw_ostream &OS) const {
-  for (const auto &FPtr : Functions)
-    FPtr->print(OS);
-}
-
-#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
-/// dump - Dump GCOVFile content to dbgs() for debugging purposes.
-LLVM_DUMP_METHOD void GCOVFile::dump() const {
-  print(dbgs());
-}
-#endif
-
-/// collectLineCounts - Collect line counts. This must be used after
-/// reading .gcno and .gcda files.
-void GCOVFile::collectLineCounts(FileInfo &FI) {
-  for (const auto &FPtr : Functions)
-    FPtr->collectLineCounts(FI);
-  FI.setRunCount(RunCount);
-  FI.setProgramCount(ProgramCount);
-}
-
-//===----------------------------------------------------------------------===//
-// GCOVFunction implementation.
-
-/// readGCNO - Read a function from the GCNO buffer. Return false if an error
-/// occurs.
-bool GCOVFunction::readGCNO(GCOVBuffer &Buff, GCOV::GCOVVersion Version) {
-  uint32_t Dummy;
-  if (!Buff.readInt(Dummy))
-    return false; // Function header length
-  if (!Buff.readInt(Ident))
-    return false;
-  if (!Buff.readInt(Checksum))
-    return false;
-  if (Version != GCOV::V402) {
-    uint32_t CfgChecksum;
-    if (!Buff.readInt(CfgChecksum))
-      return false;
-    if (Parent.getChecksum() != CfgChecksum) {
-      errs() << "File checksums do not match: " << Parent.getChecksum()
-             << " != " << CfgChecksum << " in (" << Name << ").\n";
-      return false;
-    }
-  }
-  if (!Buff.readString(Name))
-    return false;
-  if (!Buff.readString(Filename))
-    return false;
-  if (!Buff.readInt(LineNumber))
-    return false;
-
-  // read blocks.
-  if (!Buff.readBlockTag()) {
-    errs() << "Block tag not found.\n";
-    return false;
-  }
-  uint32_t BlockCount;
-  if (!Buff.readInt(BlockCount))
-    return false;
-  for (uint32_t i = 0, e = BlockCount; i != e; ++i) {
-    if (!Buff.readInt(Dummy))
-      return false; // Block flags;
-    Blocks.push_back(make_unique<GCOVBlock>(*this, i));
-  }
-
-  // read edges.
-  while (Buff.readEdgeTag()) {
-    uint32_t EdgeCount;
-    if (!Buff.readInt(EdgeCount))
-      return false;
-    EdgeCount = (EdgeCount - 1) / 2;
-    uint32_t BlockNo;
-    if (!Buff.readInt(BlockNo))
-      return false;
-    if (BlockNo >= BlockCount) {
-      errs() << "Unexpected block number: " << BlockNo << " (in " << Name
-             << ").\n";
-      return false;
-    }
-    for (uint32_t i = 0, e = EdgeCount; i != e; ++i) {
-      uint32_t Dst;
-      if (!Buff.readInt(Dst))
-        return false;
-      Edges.push_back(make_unique<GCOVEdge>(*Blocks[BlockNo], *Blocks[Dst]));
-      GCOVEdge *Edge = Edges.back().get();
-      Blocks[BlockNo]->addDstEdge(Edge);
-      Blocks[Dst]->addSrcEdge(Edge);
-      if (!Buff.readInt(Dummy))
-        return false; // Edge flag
-    }
-  }
-
-  // read line table.
-  while (Buff.readLineTag()) {
-    uint32_t LineTableLength;
-    // Read the length of this line table.
-    if (!Buff.readInt(LineTableLength))
-      return false;
-    uint32_t EndPos = Buff.getCursor() + LineTableLength * 4;
-    uint32_t BlockNo;
-    // Read the block number this table is associated with.
-    if (!Buff.readInt(BlockNo))
-      return false;
-    if (BlockNo >= BlockCount) {
-      errs() << "Unexpected block number: " << BlockNo << " (in " << Name
-             << ").\n";
-      return false;
-    }
-    GCOVBlock &Block = *Blocks[BlockNo];
-    // Read the word that pads the beginning of the line table. This may be a
-    // flag of some sort, but seems to always be zero.
-    if (!Buff.readInt(Dummy))
-      return false;
-
-    // Line information starts here and continues up until the last word.
-    if (Buff.getCursor() != (EndPos - sizeof(uint32_t))) {
-      StringRef F;
-      // Read the source file name.
-      if (!Buff.readString(F))
-        return false;
-      if (Filename != F) {
-        errs() << "Multiple sources for a single basic block: " << Filename
-               << " != " << F << " (in " << Name << ").\n";
-        return false;
-      }
-      // Read lines up to, but not including, the null terminator.
-      while (Buff.getCursor() < (EndPos - 2 * sizeof(uint32_t))) {
-        uint32_t Line;
-        if (!Buff.readInt(Line))
-          return false;
-        // Line 0 means this instruction was injected by the compiler. Skip it.
-        if (!Line)
-          continue;
-        Block.addLine(Line);
-      }
-      // Read the null terminator.
-      if (!Buff.readInt(Dummy))
-        return false;
-    }
-    // The last word is either a flag or padding, it isn't clear which. Skip
-    // over it.
-    if (!Buff.readInt(Dummy))
-      return false;
-  }
-  return true;
-}
-
-/// readGCDA - Read a function from the GCDA buffer. Return false if an error
-/// occurs.
-bool GCOVFunction::readGCDA(GCOVBuffer &Buff, GCOV::GCOVVersion Version) {
-  uint32_t HeaderLength;
-  if (!Buff.readInt(HeaderLength))
-    return false; // Function header length
-
-  uint64_t EndPos = Buff.getCursor() + HeaderLength * sizeof(uint32_t);
-
-  uint32_t GCDAIdent;
-  if (!Buff.readInt(GCDAIdent))
-    return false;
-  if (Ident != GCDAIdent) {
-    errs() << "Function identifiers do not match: " << Ident
-           << " != " << GCDAIdent << " (in " << Name << ").\n";
-    return false;
-  }
-
-  uint32_t GCDAChecksum;
-  if (!Buff.readInt(GCDAChecksum))
-    return false;
-  if (Checksum != GCDAChecksum) {
-    errs() << "Function checksums do not match: " << Checksum
-           << " != " << GCDAChecksum << " (in " << Name << ").\n";
-    return false;
-  }
-
-  uint32_t CfgChecksum;
-  if (Version != GCOV::V402) {
-    if (!Buff.readInt(CfgChecksum))
-      return false;
-    if (Parent.getChecksum() != CfgChecksum) {
-      errs() << "File checksums do not match: " << Parent.getChecksum()
-             << " != " << CfgChecksum << " (in " << Name << ").\n";
-      return false;
-    }
-  }
-
-  if (Buff.getCursor() < EndPos) {
-    StringRef GCDAName;
-    if (!Buff.readString(GCDAName))
-      return false;
-    if (Name != GCDAName) {
-      errs() << "Function names do not match: " << Name << " != " << GCDAName
-             << ".\n";
-      return false;
-    }
-  }
-
-  if (!Buff.readArcTag()) {
-    errs() << "Arc tag not found (in " << Name << ").\n";
-    return false;
-  }
-
-  uint32_t Count;
-  if (!Buff.readInt(Count))
-    return false;
-  Count /= 2;
-
-  // This for loop adds the counts for each block. A second nested loop is
-  // required to combine the edge counts that are contained in the GCDA file.
-  for (uint32_t BlockNo = 0; Count > 0; ++BlockNo) {
-    // The last block is always reserved for exit block
-    if (BlockNo >= Blocks.size()) {
-      errs() << "Unexpected number of edges (in " << Name << ").\n";
-      return false;
-    }
-    if (BlockNo == Blocks.size() - 1)
-      errs() << "(" << Name << ") has arcs from exit block.\n";
-    GCOVBlock &Block = *Blocks[BlockNo];
-    for (size_t EdgeNo = 0, End = Block.getNumDstEdges(); EdgeNo < End;
-         ++EdgeNo) {
-      if (Count == 0) {
-        errs() << "Unexpected number of edges (in " << Name << ").\n";
-        return false;
-      }
-      uint64_t ArcCount;
-      if (!Buff.readInt64(ArcCount))
-        return false;
-      Block.addCount(EdgeNo, ArcCount);
-      --Count;
-    }
-    Block.sortDstEdges();
-  }
-  return true;
-}
-
-/// getEntryCount - Get the number of times the function was called by
-/// retrieving the entry block's count.
-uint64_t GCOVFunction::getEntryCount() const {
-  return Blocks.front()->getCount();
-}
-
-/// getExitCount - Get the number of times the function returned by retrieving
-/// the exit block's count.
-uint64_t GCOVFunction::getExitCount() const {
-  return Blocks.back()->getCount();
-}
-
-void GCOVFunction::print(raw_ostream &OS) const {
-  OS << "===== " << Name << " (" << Ident << ") @ " << Filename << ":"
-     << LineNumber << "\n";
-  for (const auto &Block : Blocks)
-    Block->print(OS);
-}
-
-#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
-/// dump - Dump GCOVFunction content to dbgs() for debugging purposes.
-LLVM_DUMP_METHOD void GCOVFunction::dump() const {
-  print(dbgs());
-}
-#endif
-
-/// collectLineCounts - Collect line counts. This must be used after
-/// reading .gcno and .gcda files.
-void GCOVFunction::collectLineCounts(FileInfo &FI) {
-  // If the line number is zero, this is a function that doesn't actually appear
-  // in the source file, so there isn't anything we can do with it.
-  if (LineNumber == 0)
-    return;
-
-  for (const auto &Block : Blocks)
-    Block->collectLineCounts(FI);
-  FI.addFunctionLine(Filename, LineNumber, this);
-}
-
-//===----------------------------------------------------------------------===//
-// GCOVBlock implementation.
-
-/// ~GCOVBlock - Delete GCOVBlock and its content.
-GCOVBlock::~GCOVBlock() {
-  SrcEdges.clear();
-  DstEdges.clear();
-  Lines.clear();
-}
-
-/// addCount - Add to block counter while storing the edge count. If the
-/// destination has no outgoing edges, also update that block's count too.
-void GCOVBlock::addCount(size_t DstEdgeNo, uint64_t N) {
-  assert(DstEdgeNo < DstEdges.size()); // up to caller to ensure EdgeNo is valid
-  DstEdges[DstEdgeNo]->Count = N;
-  Counter += N;
-  if (!DstEdges[DstEdgeNo]->Dst.getNumDstEdges())
-    DstEdges[DstEdgeNo]->Dst.Counter += N;
-}
-
-/// sortDstEdges - Sort destination edges by block number, nop if already
-/// sorted. This is required for printing branch info in the correct order.
-void GCOVBlock::sortDstEdges() {
-  if (!DstEdgesAreSorted) {
-    SortDstEdgesFunctor SortEdges;
-    std::stable_sort(DstEdges.begin(), DstEdges.end(), SortEdges);
-  }
-}
-
-/// collectLineCounts - Collect line counts. This must be used after
-/// reading .gcno and .gcda files.
-void GCOVBlock::collectLineCounts(FileInfo &FI) {
-  for (uint32_t N : Lines)
-    FI.addBlockLine(Parent.getFilename(), N, this);
-}
-
-void GCOVBlock::print(raw_ostream &OS) const {
-  OS << "Block : " << Number << " Counter : " << Counter << "\n";
-  if (!SrcEdges.empty()) {
-    OS << "\tSource Edges : ";
-    for (const GCOVEdge *Edge : SrcEdges)
-      OS << Edge->Src.Number << " (" << Edge->Count << "), ";
-    OS << "\n";
-  }
-  if (!DstEdges.empty()) {
-    OS << "\tDestination Edges : ";
-    for (const GCOVEdge *Edge : DstEdges)
-      OS << Edge->Dst.Number << " (" << Edge->Count << "), ";
-    OS << "\n";
-  }
-  if (!Lines.empty()) {
-    OS << "\tLines : ";
-    for (uint32_t N : Lines)
-      OS << (N) << ",";
-    OS << "\n";
-  }
-}
-
-#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
-/// dump - Dump GCOVBlock content to dbgs() for debugging purposes.
-LLVM_DUMP_METHOD void GCOVBlock::dump() const {
-  print(dbgs());
-}
-#endif
-
-//===----------------------------------------------------------------------===//
-// FileInfo implementation.
-
-// Safe integer division, returns 0 if numerator is 0.
-static uint32_t safeDiv(uint64_t Numerator, uint64_t Divisor) {
-  if (!Numerator)
-    return 0;
-  return Numerator / Divisor;
-}
-
-// This custom division function mimics gcov's branch ouputs:
-//   - Round to closest whole number
-//   - Only output 0% or 100% if it's exactly that value
-static uint32_t branchDiv(uint64_t Numerator, uint64_t Divisor) {
-  if (!Numerator)
-    return 0;
-  if (Numerator == Divisor)
-    return 100;
-
-  uint8_t Res = (Numerator * 100 + Divisor / 2) / Divisor;
-  if (Res == 0)
-    return 1;
-  if (Res == 100)
-    return 99;
-  return Res;
-}
-
-namespace {
-struct formatBranchInfo {
-  formatBranchInfo(const GCOV::Options &Options, uint64_t Count, uint64_t Total)
-      : Options(Options), Count(Count), Total(Total) {}
-
-  void print(raw_ostream &OS) const {
-    if (!Total)
-      OS << "never executed";
-    else if (Options.BranchCount)
-      OS << "taken " << Count;
-    else
-      OS << "taken " << branchDiv(Count, Total) << "%";
-  }
-
-  const GCOV::Options &Options;
-  uint64_t Count;
-  uint64_t Total;
-};
-
-static raw_ostream &operator<<(raw_ostream &OS, const formatBranchInfo &FBI) {
-  FBI.print(OS);
-  return OS;
-}
-
-class LineConsumer {
-  std::unique_ptr<MemoryBuffer> Buffer;
-  StringRef Remaining;
-
-public:
-  LineConsumer(StringRef Filename) {
-    ErrorOr<std::unique_ptr<MemoryBuffer>> BufferOrErr =
-        MemoryBuffer::getFileOrSTDIN(Filename);
-    if (std::error_code EC = BufferOrErr.getError()) {
-      errs() << Filename << ": " << EC.message() << "\n";
-      Remaining = "";
-    } else {
-      Buffer = std::move(BufferOrErr.get());
-      Remaining = Buffer->getBuffer();
-    }
-  }
-  bool empty() { return Remaining.empty(); }
-  void printNext(raw_ostream &OS, uint32_t LineNum) {
-    StringRef Line;
-    if (empty())
-      Line = "/*EOF*/";
-    else
-      std::tie(Line, Remaining) = Remaining.split("\n");
-    OS << format("%5u:", LineNum) << Line << "\n";
-  }
-};
-} // end anonymous namespace
-
-/// Convert a path to a gcov filename. If PreservePaths is true, this
-/// translates "/" to "#", ".." to "^", and drops ".", to match gcov.
-static std::string mangleCoveragePath(StringRef Filename, bool PreservePaths) {
-  if (!PreservePaths)
-    return sys::path::filename(Filename).str();
-
-  // This behaviour is defined by gcov in terms of text replacements, so it's
-  // not likely to do anything useful on filesystems with different textual
-  // conventions.
-  llvm::SmallString<256> Result("");
-  StringRef::iterator I, S, E;
-  for (I = S = Filename.begin(), E = Filename.end(); I != E; ++I) {
-    if (*I != '/')
-      continue;
-
-    if (I - S == 1 && *S == '.') {
-      // ".", the current directory, is skipped.
-    } else if (I - S == 2 && *S == '.' && *(S + 1) == '.') {
-      // "..", the parent directory, is replaced with "^".
-      Result.append("^#");
-    } else {
-      if (S < I)
-        // Leave other components intact,
-        Result.append(S, I);
-      // And separate with "#".
-      Result.push_back('#');
-    }
-    S = I + 1;
-  }
-
-  if (S < I)
-    Result.append(S, I);
-  return Result.str();
-}
-
-std::string FileInfo::getCoveragePath(StringRef Filename,
-                                      StringRef MainFilename) {
-  if (Options.NoOutput)
-    // This is probably a bug in gcov, but when -n is specified, paths aren't
-    // mangled at all, and the -l and -p options are ignored. Here, we do the
-    // same.
-    return Filename;
-
-  std::string CoveragePath;
-  if (Options.LongFileNames && !Filename.equals(MainFilename))
-    CoveragePath =
-        mangleCoveragePath(MainFilename, Options.PreservePaths) + "##";
-  CoveragePath += mangleCoveragePath(Filename, Options.PreservePaths) + ".gcov";
-  return CoveragePath;
-}
-
-std::unique_ptr<raw_ostream>
-FileInfo::openCoveragePath(StringRef CoveragePath) {
-  if (Options.NoOutput)
-    return llvm::make_unique<raw_null_ostream>();
-
-  std::error_code EC;
-  auto OS = llvm::make_unique<raw_fd_ostream>(CoveragePath, EC,
-                                              sys::fs::F_Text);
-  if (EC) {
-    errs() << EC.message() << "\n";
-    return llvm::make_unique<raw_null_ostream>();
-  }
-  return std::move(OS);
-}
-
-/// print -  Print source files with collected line count information.
-void FileInfo::print(raw_ostream &InfoOS, StringRef MainFilename,
-                     StringRef GCNOFile, StringRef GCDAFile) {
-  SmallVector<StringRef, 4> Filenames;
-  for (const auto &LI : LineInfo)
-    Filenames.push_back(LI.first());
-  std::sort(Filenames.begin(), Filenames.end());
-
-  for (StringRef Filename : Filenames) {
-    auto AllLines = LineConsumer(Filename);
-
-    std::string CoveragePath = getCoveragePath(Filename, MainFilename);
-    std::unique_ptr<raw_ostream> CovStream = openCoveragePath(CoveragePath);
-    raw_ostream &CovOS = *CovStream;
-
-    CovOS << "        -:    0:Source:" << Filename << "\n";
-    CovOS << "        -:    0:Graph:" << GCNOFile << "\n";
-    CovOS << "        -:    0:Data:" << GCDAFile << "\n";
-    CovOS << "        -:    0:Runs:" << RunCount << "\n";
-    CovOS << "        -:    0:Programs:" << ProgramCount << "\n";
-
-    const LineData &Line = LineInfo[Filename];
-    GCOVCoverage FileCoverage(Filename);
-    for (uint32_t LineIndex = 0; LineIndex < Line.LastLine || !AllLines.empty();
-         ++LineIndex) {
-      if (Options.BranchInfo) {
-        FunctionLines::const_iterator FuncsIt = Line.Functions.find(LineIndex);
-        if (FuncsIt != Line.Functions.end())
-          printFunctionSummary(CovOS, FuncsIt->second);
-      }
-
-      BlockLines::const_iterator BlocksIt = Line.Blocks.find(LineIndex);
-      if (BlocksIt == Line.Blocks.end()) {
-        // No basic blocks are on this line. Not an executable line of code.
-        CovOS << "        -:";
-        AllLines.printNext(CovOS, LineIndex + 1);
-      } else {
-        const BlockVector &Blocks = BlocksIt->second;
-
-        // Add up the block counts to form line counts.
-        DenseMap<const GCOVFunction *, bool> LineExecs;
-        uint64_t LineCount = 0;
-        for (const GCOVBlock *Block : Blocks) {
-          if (Options.AllBlocks) {
-            // Only take the highest block count for that line.
-            uint64_t BlockCount = Block->getCount();
-            LineCount = LineCount > BlockCount ? LineCount : BlockCount;
-          } else {
-            // Sum up all of the block counts.
-            LineCount += Block->getCount();
-          }
-
-          if (Options.FuncCoverage) {
-            // This is a slightly convoluted way to most accurately gather line
-            // statistics for functions. Basically what is happening is that we
-            // don't want to count a single line with multiple blocks more than
-            // once. However, we also don't simply want to give the total line
-            // count to every function that starts on the line. Thus, what is
-            // happening here are two things:
-            // 1) Ensure that the number of logical lines is only incremented
-            //    once per function.
-            // 2) If there are multiple blocks on the same line, ensure that the
-            //    number of lines executed is incremented as long as at least
-            //    one of the blocks are executed.
-            const GCOVFunction *Function = &Block->getParent();
-            if (FuncCoverages.find(Function) == FuncCoverages.end()) {
-              std::pair<const GCOVFunction *, GCOVCoverage> KeyValue(
-                  Function, GCOVCoverage(Function->getName()));
-              FuncCoverages.insert(KeyValue);
-            }
-            GCOVCoverage &FuncCoverage = FuncCoverages.find(Function)->second;
-
-            if (LineExecs.find(Function) == LineExecs.end()) {
-              if (Block->getCount()) {
-                ++FuncCoverage.LinesExec;
-                LineExecs[Function] = true;
-              } else {
-                LineExecs[Function] = false;
-              }
-              ++FuncCoverage.LogicalLines;
-            } else if (!LineExecs[Function] && Block->getCount()) {
-              ++FuncCoverage.LinesExec;
-              LineExecs[Function] = true;
-            }
-          }
-        }
-
-        if (LineCount == 0)
-          CovOS << "    #####:";
-        else {
-          CovOS << format("%9" PRIu64 ":", LineCount);
-          ++FileCoverage.LinesExec;
-        }
-        ++FileCoverage.LogicalLines;
-
-        AllLines.printNext(CovOS, LineIndex + 1);
-
-        uint32_t BlockNo = 0;
-        uint32_t EdgeNo = 0;
-        for (const GCOVBlock *Block : Blocks) {
-          // Only print block and branch information at the end of the block.
-          if (Block->getLastLine() != LineIndex + 1)
-            continue;
-          if (Options.AllBlocks)
-            printBlockInfo(CovOS, *Block, LineIndex, BlockNo);
-          if (Options.BranchInfo) {
-            size_t NumEdges = Block->getNumDstEdges();
-            if (NumEdges > 1)
-              printBranchInfo(CovOS, *Block, FileCoverage, EdgeNo);
-            else if (Options.UncondBranch && NumEdges == 1)
-              printUncondBranchInfo(CovOS, EdgeNo,
-                                    (*Block->dst_begin())->Count);
-          }
-        }
-      }
-    }
-    FileCoverages.push_back(std::make_pair(CoveragePath, FileCoverage));
-  }
-
-  // FIXME: There is no way to detect calls given current instrumentation.
-  if (Options.FuncCoverage)
-    printFuncCoverage(InfoOS);
-  printFileCoverage(InfoOS);
-}
-
-/// printFunctionSummary - Print function and block summary.
-void FileInfo::printFunctionSummary(raw_ostream &OS,
-                                    const FunctionVector &Funcs) const {
-  for (const GCOVFunction *Func : Funcs) {
-    uint64_t EntryCount = Func->getEntryCount();
-    uint32_t BlocksExec = 0;
-    for (const GCOVBlock &Block : Func->blocks())
-      if (Block.getNumDstEdges() && Block.getCount())
-        ++BlocksExec;
-
-    OS << "function " << Func->getName() << " called " << EntryCount
-       << " returned " << safeDiv(Func->getExitCount() * 100, EntryCount)
-       << "% blocks executed "
-       << safeDiv(BlocksExec * 100, Func->getNumBlocks() - 1) << "%\n";
-  }
-}
-
-/// printBlockInfo - Output counts for each block.
-void FileInfo::printBlockInfo(raw_ostream &OS, const GCOVBlock &Block,
-                              uint32_t LineIndex, uint32_t &BlockNo) const {
-  if (Block.getCount() == 0)
-    OS << "    $$$$$:";
-  else
-    OS << format("%9" PRIu64 ":", Block.getCount());
-  OS << format("%5u-block %2u\n", LineIndex + 1, BlockNo++);
-}
-
-/// printBranchInfo - Print conditional branch probabilities.
-void FileInfo::printBranchInfo(raw_ostream &OS, const GCOVBlock &Block,
-                               GCOVCoverage &Coverage, uint32_t &EdgeNo) {
-  SmallVector<uint64_t, 16> BranchCounts;
-  uint64_t TotalCounts = 0;
-  for (const GCOVEdge *Edge : Block.dsts()) {
-    BranchCounts.push_back(Edge->Count);
-    TotalCounts += Edge->Count;
-    if (Block.getCount())
-      ++Coverage.BranchesExec;
-    if (Edge->Count)
-      ++Coverage.BranchesTaken;
-    ++Coverage.Branches;
-
-    if (Options.FuncCoverage) {
-      const GCOVFunction *Function = &Block.getParent();
-      GCOVCoverage &FuncCoverage = FuncCoverages.find(Function)->second;
-      if (Block.getCount())
-        ++FuncCoverage.BranchesExec;
-      if (Edge->Count)
-        ++FuncCoverage.BranchesTaken;
-      ++FuncCoverage.Branches;
-    }
-  }
-
-  for (uint64_t N : BranchCounts)
-    OS << format("branch %2u ", EdgeNo++)
-       << formatBranchInfo(Options, N, TotalCounts) << "\n";
-}
-
-/// printUncondBranchInfo - Print unconditional branch probabilities.
-void FileInfo::printUncondBranchInfo(raw_ostream &OS, uint32_t &EdgeNo,
-                                     uint64_t Count) const {
-  OS << format("unconditional %2u ", EdgeNo++)
-     << formatBranchInfo(Options, Count, Count) << "\n";
-}
-
-// printCoverage - Print generic coverage info used by both printFuncCoverage
-// and printFileCoverage.
-void FileInfo::printCoverage(raw_ostream &OS,
-                             const GCOVCoverage &Coverage) const {
-  OS << format("Lines executed:%.2f%% of %u\n",
-               double(Coverage.LinesExec) * 100 / Coverage.LogicalLines,
-               Coverage.LogicalLines);
-  if (Options.BranchInfo) {
-    if (Coverage.Branches) {
-      OS << format("Branches executed:%.2f%% of %u\n",
-                   double(Coverage.BranchesExec) * 100 / Coverage.Branches,
-                   Coverage.Branches);
-      OS << format("Taken at least once:%.2f%% of %u\n",
-                   double(Coverage.BranchesTaken) * 100 / Coverage.Branches,
-                   Coverage.Branches);
-    } else {
-      OS << "No branches\n";
-    }
-    OS << "No calls\n"; // to be consistent with gcov
-  }
-}
-
-// printFuncCoverage - Print per-function coverage info.
-void FileInfo::printFuncCoverage(raw_ostream &OS) const {
-  for (const auto &FC : FuncCoverages) {
-    const GCOVCoverage &Coverage = FC.second;
-    OS << "Function '" << Coverage.Name << "'\n";
-    printCoverage(OS, Coverage);
-    OS << "\n";
-  }
-}
-
-// printFileCoverage - Print per-file coverage info.
-void FileInfo::printFileCoverage(raw_ostream &OS) const {
-  for (const auto &FC : FileCoverages) {
-    const std::string &Filename = FC.first;
-    const GCOVCoverage &Coverage = FC.second;
-    OS << "File '" << Coverage.Name << "'\n";
-    printCoverage(OS, Coverage);
-    if (!Options.NoOutput)
-      OS << Coverage.Name << ":creating '" << Filename << "'\n";
-    OS << "\n";
-  }
-}
diff --git a/lib/IR/Globals.cpp b/lib/IR/Globals.cpp
index afd4a36270a8..da1b6c5e0c91 100644
--- a/lib/IR/Globals.cpp
+++ b/lib/IR/Globals.cpp
@@ -67,6 +67,7 @@ void GlobalValue::copyAttributesFrom(const GlobalValue *Src) {
   setVisibility(Src->getVisibility());
   setUnnamedAddr(Src->getUnnamedAddr());
   setDLLStorageClass(Src->getDLLStorageClass());
+  setDSOLocal(Src->isDSOLocal());
 }
 
 void GlobalValue::removeFromParent() {
diff --git a/lib/IR/IRBuilder.cpp b/lib/IR/IRBuilder.cpp
index b7fa07c6ffac..027c0255bcec 100644
--- a/lib/IR/IRBuilder.cpp
+++ b/lib/IR/IRBuilder.cpp
@@ -135,8 +135,13 @@ CreateMemCpy(Value *Dst, Value *Src, Value *Size, unsigned Align,
 }
 
 CallInst *IRBuilderBase::CreateElementUnorderedAtomicMemCpy(
-    Value *Dst, Value *Src, Value *Size, uint32_t ElementSize, MDNode *TBAATag,
-    MDNode *TBAAStructTag, MDNode *ScopeTag, MDNode *NoAliasTag) {
+    Value *Dst, unsigned DstAlign, Value *Src, unsigned SrcAlign, Value *Size,
+    uint32_t ElementSize, MDNode *TBAATag, MDNode *TBAAStructTag,
+    MDNode *ScopeTag, MDNode *NoAliasTag) {
+  assert(DstAlign >= ElementSize &&
+         "Pointer alignment must be at least element size");
+  assert(SrcAlign >= ElementSize &&
+         "Pointer alignment must be at least element size");
   Dst = getCastedInt8PtrValue(Dst);
   Src = getCastedInt8PtrValue(Src);
 
@@ -148,6 +153,10 @@ CallInst *IRBuilderBase::CreateElementUnorderedAtomicMemCpy(
 
   CallInst *CI = createCallHelper(TheFn, Ops, this);
 
+  // Set the alignment of the pointer args.
+  CI->addParamAttr(0, Attribute::getWithAlignment(CI->getContext(), DstAlign));
+  CI->addParamAttr(1, Attribute::getWithAlignment(CI->getContext(), SrcAlign));
+
   // Set the TBAA info if present.
   if (TBAATag)
     CI->setMetadata(LLVMContext::MD_tbaa, TBAATag);
@@ -356,6 +365,7 @@ CallInst *IRBuilderBase::CreateMaskedLoad(Value *Ptr, unsigned Align,
   PointerType *PtrTy = cast<PointerType>(Ptr->getType());
   Type *DataTy = PtrTy->getElementType();
   assert(DataTy->isVectorTy() && "Ptr should point to a vector");
+  assert(Mask && "Mask should not be all-ones (null)");
   if (!PassThru)
     PassThru = UndefValue::get(DataTy);
   Type *OverloadedTypes[] = { DataTy, PtrTy };
@@ -375,6 +385,7 @@ CallInst *IRBuilderBase::CreateMaskedStore(Value *Val, Value *Ptr,
   PointerType *PtrTy = cast<PointerType>(Ptr->getType());
   Type *DataTy = PtrTy->getElementType();
   assert(DataTy->isVectorTy() && "Ptr should point to a vector");
+  assert(Mask && "Mask should not be all-ones (null)");
   Type *OverloadedTypes[] = { DataTy, PtrTy };
   Value *Ops[] = { Val, Ptr, getInt32(Align), Mask };
   return CreateMaskedIntrinsic(Intrinsic::masked_store, Ops, OverloadedTypes);
diff --git a/lib/IR/IRPrintingPasses.cpp b/lib/IR/IRPrintingPasses.cpp
index 955fdc749b2b..3b32814bed5c 100644
--- a/lib/IR/IRPrintingPasses.cpp
+++ b/lib/IR/IRPrintingPasses.cpp
@@ -44,8 +44,12 @@ PrintFunctionPass::PrintFunctionPass(raw_ostream &OS, const std::string &Banner)
 
 PreservedAnalyses PrintFunctionPass::run(Function &F,
                                          FunctionAnalysisManager &) {
-  if (isFunctionInPrintList(F.getName()))
-    OS << Banner << static_cast<Value &>(F);
+  if (isFunctionInPrintList(F.getName())) {
+    if (forcePrintModuleIR())
+      OS << Banner << " (function: " << F.getName() << ")\n" << *F.getParent();
+    else
+      OS << Banner << static_cast<Value &>(F);
+  }
   return PreservedAnalyses::all();
 }
 
@@ -71,7 +75,7 @@ public:
     AU.setPreservesAll();
   }
 
-  StringRef getPassName() const override { return "Print Module IR"; }

+  StringRef getPassName() const override { return "Print Module IR"; }
 };
 
 class PrintFunctionPassWrapper : public FunctionPass {
@@ -94,7 +98,7 @@ public:
     AU.setPreservesAll();
   }
 
-  StringRef getPassName() const override { return "Print Function IR"; }

+  StringRef getPassName() const override { return "Print Function IR"; }
 };
 
 class PrintBasicBlockPass : public BasicBlockPass {
diff --git a/lib/IR/InlineAsm.cpp b/lib/IR/InlineAsm.cpp
index ad22efdf0eff..8667d7aab583 100644
--- a/lib/IR/InlineAsm.cpp
+++ b/lib/IR/InlineAsm.cpp
@@ -163,6 +163,7 @@ bool InlineAsm::ConstraintInfo::Parse(StringRef Str,
           return true;
         // Note that operand #n has a matching input.
         scInfo.MatchingInput = ConstraintsSoFar.size();
+        assert(scInfo.MatchingInput >= 0);
       } else {
         if (ConstraintsSoFar[N].hasMatchingInput() &&
             (size_t)ConstraintsSoFar[N].MatchingInput !=
@@ -170,6 +171,7 @@ bool InlineAsm::ConstraintInfo::Parse(StringRef Str,
           return true;
         // Note that operand #n has a matching input.
         ConstraintsSoFar[N].MatchingInput = ConstraintsSoFar.size();
+        assert(ConstraintsSoFar[N].MatchingInput >= 0);
         }
     } else if (*I == '|') {
       multipleAlternativeIndex++;
diff --git a/lib/IR/Instruction.cpp b/lib/IR/Instruction.cpp
index 365cb019aec4..5f2a6146ad81 100644
--- a/lib/IR/Instruction.cpp
+++ b/lib/IR/Instruction.cpp
@@ -13,11 +13,9 @@
 
 #include "llvm/IR/Instruction.h"
 #include "llvm/ADT/DenseSet.h"
-#include "llvm/IR/CallSite.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/MDBuilder.h"
-#include "llvm/IR/Module.h"
 #include "llvm/IR/Operator.h"
 #include "llvm/IR/Type.h"
 using namespace llvm;
@@ -89,6 +87,10 @@ void Instruction::moveBefore(Instruction *MovePos) {
   moveBefore(*MovePos->getParent(), MovePos->getIterator());
 }
 
+void Instruction::moveAfter(Instruction *MovePos) {
+  moveBefore(*MovePos->getParent(), ++MovePos->getIterator());
+}
+
 void Instruction::moveBefore(BasicBlock &BB,
                              SymbolTableList<Instruction>::iterator I) {
   assert(I == BB.end() || I->getParent() == &BB);
@@ -142,9 +144,14 @@ bool Instruction::isExact() const {
   return cast<PossiblyExactOperator>(this)->isExact();
 }
 
-void Instruction::setHasUnsafeAlgebra(bool B) {
+void Instruction::setFast(bool B) {
   assert(isa<FPMathOperator>(this) && "setting fast-math flag on invalid op");
-  cast<FPMathOperator>(this)->setHasUnsafeAlgebra(B);
+  cast<FPMathOperator>(this)->setFast(B);
+}
+
+void Instruction::setHasAllowReassoc(bool B) {
+  assert(isa<FPMathOperator>(this) && "setting fast-math flag on invalid op");
+  cast<FPMathOperator>(this)->setHasAllowReassoc(B);
 }
 
 void Instruction::setHasNoNaNs(bool B) {
@@ -167,6 +174,11 @@ void Instruction::setHasAllowReciprocal(bool B) {
   cast<FPMathOperator>(this)->setHasAllowReciprocal(B);
 }
 
+void Instruction::setHasApproxFunc(bool B) {
+  assert(isa<FPMathOperator>(this) && "setting fast-math flag on invalid op");
+  cast<FPMathOperator>(this)->setHasApproxFunc(B);
+}
+
 void Instruction::setFastMathFlags(FastMathFlags FMF) {
   assert(isa<FPMathOperator>(this) && "setting fast-math flag on invalid op");
   cast<FPMathOperator>(this)->setFastMathFlags(FMF);
@@ -177,9 +189,14 @@ void Instruction::copyFastMathFlags(FastMathFlags FMF) {
   cast<FPMathOperator>(this)->copyFastMathFlags(FMF);
 }
 
-bool Instruction::hasUnsafeAlgebra() const {
+bool Instruction::isFast() const {
+  assert(isa<FPMathOperator>(this) && "getting fast-math flag on invalid op");
+  return cast<FPMathOperator>(this)->isFast();
+}
+
+bool Instruction::hasAllowReassoc() const {
   assert(isa<FPMathOperator>(this) && "getting fast-math flag on invalid op");
-  return cast<FPMathOperator>(this)->hasUnsafeAlgebra();
+  return cast<FPMathOperator>(this)->hasAllowReassoc();
 }
 
 bool Instruction::hasNoNaNs() const {
@@ -207,6 +224,11 @@ bool Instruction::hasAllowContract() const {
   return cast<FPMathOperator>(this)->hasAllowContract();
 }
 
+bool Instruction::hasApproxFunc() const {
+  assert(isa<FPMathOperator>(this) && "getting fast-math flag on invalid op");
+  return cast<FPMathOperator>(this)->hasApproxFunc();
+}
+
 FastMathFlags Instruction::getFastMathFlags() const {
   assert(isa<FPMathOperator>(this) && "getting fast-math flag on invalid op");
   return cast<FPMathOperator>(this)->getFastMathFlags();
@@ -575,7 +597,7 @@ bool Instruction::isAssociative() const {
   switch (Opcode) {
   case FMul:
   case FAdd:
-    return cast<FPMathOperator>(this)->hasUnsafeAlgebra();
+    return cast<FPMathOperator>(this)->isFast();
   default:
     return false;
   }
@@ -621,7 +643,6 @@ void Instruction::copyMetadata(const Instruction &SrcInst,
   }
   if (WL.empty() || WLS.count(LLVMContext::MD_dbg))
     setDebugLoc(SrcInst.getDebugLoc());
-  return;
 }
 
 Instruction *Instruction::clone() const {
diff --git a/lib/IR/Instructions.cpp b/lib/IR/Instructions.cpp
index 2c49564e328b..490fcbce7439 100644
--- a/lib/IR/Instructions.cpp
+++ b/lib/IR/Instructions.cpp
@@ -2299,7 +2299,7 @@ bool CastInst::isLosslessCast() const {
 bool CastInst::isNoopCast(Instruction::CastOps Opcode,
                           Type *SrcTy,
                           Type *DestTy,
-                          Type *IntPtrTy) {
+                          const DataLayout &DL) {
   switch (Opcode) {
     default: llvm_unreachable("Invalid CastOp");
     case Instruction::Trunc:
@@ -2317,30 +2317,16 @@ bool CastInst::isNoopCast(Instruction::CastOps Opcode,
     case Instruction::BitCast:
       return true;  // BitCast never modifies bits.
     case Instruction::PtrToInt:
-      return IntPtrTy->getScalarSizeInBits() ==
+      return DL.getIntPtrType(SrcTy)->getScalarSizeInBits() ==
              DestTy->getScalarSizeInBits();
     case Instruction::IntToPtr:
-      return IntPtrTy->getScalarSizeInBits() ==
+      return DL.getIntPtrType(DestTy)->getScalarSizeInBits() ==
              SrcTy->getScalarSizeInBits();
   }
 }
 
-/// @brief Determine if a cast is a no-op.
-bool CastInst::isNoopCast(Type *IntPtrTy) const {
-  return isNoopCast(getOpcode(), getOperand(0)->getType(), getType(), IntPtrTy);
-}
-
 bool CastInst::isNoopCast(const DataLayout &DL) const {
-  Type *PtrOpTy = nullptr;
-  if (getOpcode() == Instruction::PtrToInt)
-    PtrOpTy = getOperand(0)->getType();
-  else if (getOpcode() == Instruction::IntToPtr)
-    PtrOpTy = getType();
-
-  Type *IntPtrTy =
-      PtrOpTy ? DL.getIntPtrType(PtrOpTy) : DL.getIntPtrType(getContext(), 0);
-
-  return isNoopCast(getOpcode(), getOperand(0)->getType(), getType(), IntPtrTy);
+  return isNoopCast(getOpcode(), getOperand(0)->getType(), getType(), DL);
 }
 
 /// This function determines if a pair of casts can be eliminated and what
@@ -2891,12 +2877,15 @@ bool CastInst::isBitCastable(Type *SrcTy, Type *DestTy) {
 
 bool CastInst::isBitOrNoopPointerCastable(Type *SrcTy, Type *DestTy,
                                           const DataLayout &DL) {
+  // ptrtoint and inttoptr are not allowed on non-integral pointers
   if (auto *PtrTy = dyn_cast<PointerType>(SrcTy))
     if (auto *IntTy = dyn_cast<IntegerType>(DestTy))
-      return IntTy->getBitWidth() == DL.getPointerTypeSizeInBits(PtrTy);
+      return (IntTy->getBitWidth() == DL.getPointerTypeSizeInBits(PtrTy) &&
+              !DL.isNonIntegralPointerType(PtrTy));
   if (auto *PtrTy = dyn_cast<PointerType>(DestTy))
     if (auto *IntTy = dyn_cast<IntegerType>(SrcTy))
-      return IntTy->getBitWidth() == DL.getPointerTypeSizeInBits(PtrTy);
+      return (IntTy->getBitWidth() == DL.getPointerTypeSizeInBits(PtrTy) &&
+              !DL.isNonIntegralPointerType(PtrTy));
 
   return isBitCastable(SrcTy, DestTy);
 }
diff --git a/lib/IR/IntrinsicInst.cpp b/lib/IR/IntrinsicInst.cpp
index 8b12c55937f5..67bd5b69bb0f 100644
--- a/lib/IR/IntrinsicInst.cpp
+++ b/lib/IR/IntrinsicInst.cpp
@@ -14,10 +14,10 @@
 // are all subclasses of the CallInst class.  Note that none of these classes
 // has state or virtual methods, which is an important part of this gross/neat
 // hack working.
-// 
+//
 // In some cases, arguments to intrinsics need to be generic and are defined as
 // type pointer to empty struct { }*.  To access the real item of interest the
-// cast instruction needs to be stripped away. 
+// cast instruction needs to be stripped away.
 //
 //===----------------------------------------------------------------------===//
 
@@ -98,7 +98,7 @@ Value *InstrProfIncrementInst::getStep() const {
 ConstrainedFPIntrinsic::RoundingMode
 ConstrainedFPIntrinsic::getRoundingMode() const {
   unsigned NumOperands = getNumArgOperands();
-  Metadata *MD = 
+  Metadata *MD =
       dyn_cast<MetadataAsValue>(getArgOperand(NumOperands - 2))->getMetadata();
   if (!MD || !isa<MDString>(MD))
     return rmInvalid;
@@ -118,7 +118,7 @@ ConstrainedFPIntrinsic::getRoundingMode() const {
 ConstrainedFPIntrinsic::ExceptionBehavior
 ConstrainedFPIntrinsic::getExceptionBehavior() const {
   unsigned NumOperands = getNumArgOperands();
-  Metadata *MD = 
+  Metadata *MD =
       dyn_cast<MetadataAsValue>(getArgOperand(NumOperands - 1))->getMetadata();
   if (!MD || !isa<MDString>(MD))
     return ebInvalid;
@@ -132,7 +132,7 @@ ConstrainedFPIntrinsic::getExceptionBehavior() const {
 
 bool ConstrainedFPIntrinsic::isUnaryOp() const {
   switch (getIntrinsicID()) {
-    default: 
+    default:
       return false;
     case Intrinsic::experimental_constrained_sqrt:
     case Intrinsic::experimental_constrained_sin:
@@ -147,3 +147,13 @@ bool ConstrainedFPIntrinsic::isUnaryOp() const {
       return true;
   }
 }
+
+bool ConstrainedFPIntrinsic::isTernaryOp() const {
+  switch (getIntrinsicID()) {
+    default:
+      return false;
+    case Intrinsic::experimental_constrained_fma:
+      return true;
+  }
+}
+
diff --git a/lib/IR/LLVMContext.cpp b/lib/IR/LLVMContext.cpp
index c58459d6d5f5..c8b7c10a9a41 100644
--- a/lib/IR/LLVMContext.cpp
+++ b/lib/IR/LLVMContext.cpp
@@ -59,6 +59,8 @@ LLVMContext::LLVMContext() : pImpl(new LLVMContextImpl(*this)) {
     {MD_section_prefix, "section_prefix"},
     {MD_absolute_symbol, "absolute_symbol"},
     {MD_associated, "associated"},
+    {MD_callees, "callees"},
+    {MD_irr_loop, "irr_loop"},
   };
 
   for (auto &MDKind : MDKinds) {
@@ -129,11 +131,17 @@ void *LLVMContext::getInlineAsmDiagnosticContext() const {
   return pImpl->InlineAsmDiagContext;
 }
 
-void LLVMContext::setDiagnosticHandler(DiagnosticHandlerTy DiagnosticHandler,
-                                       void *DiagnosticContext,
-                                       bool RespectFilters) {
-  pImpl->DiagnosticHandler = DiagnosticHandler;
-  pImpl->DiagnosticContext = DiagnosticContext;
+void LLVMContext::setDiagnosticHandlerCallBack(
+    DiagnosticHandler::DiagnosticHandlerTy DiagnosticHandler,
+    void *DiagnosticContext, bool RespectFilters) {
+  pImpl->DiagHandler->DiagHandlerCallback = DiagnosticHandler;
+  pImpl->DiagHandler->DiagnosticContext = DiagnosticContext;
+  pImpl->RespectDiagnosticFilters = RespectFilters;
+}
+
+void LLVMContext::setDiagnosticHandler(std::unique_ptr<DiagnosticHandler> &&DH,
+                                      bool RespectFilters) {
+  pImpl->DiagHandler = std::move(DH);
   pImpl->RespectDiagnosticFilters = RespectFilters;
 }
 
@@ -159,12 +167,13 @@ void LLVMContext::setDiagnosticsOutputFile(std::unique_ptr<yaml::Output> F) {
   pImpl->DiagnosticsOutputFile = std::move(F);
 }
 
-LLVMContext::DiagnosticHandlerTy LLVMContext::getDiagnosticHandler() const {
-  return pImpl->DiagnosticHandler;
+DiagnosticHandler::DiagnosticHandlerTy
+LLVMContext::getDiagnosticHandlerCallBack() const {
+  return pImpl->DiagHandler->DiagHandlerCallback;
 }
 
 void *LLVMContext::getDiagnosticContext() const {
-  return pImpl->DiagnosticContext;
+  return pImpl->DiagHandler->DiagnosticContext;
 }
 
 void LLVMContext::setYieldCallback(YieldCallbackTy Callback, void *OpaqueHandle)
@@ -192,8 +201,12 @@ static bool isDiagnosticEnabled(const DiagnosticInfo &DI) {
   // pattern, passed via one of the -pass-remarks* flags, matches the name of
   // the pass that is emitting the diagnostic. If there is no match, ignore the
   // diagnostic and return.
+  //
+  // Also noisy remarks are only enabled if we have hotness information to sort
+  // them.
   if (auto *Remark = dyn_cast<DiagnosticInfoOptimizationBase>(&DI))
-    return Remark->isEnabled();
+    return Remark->isEnabled() &&
+           (!Remark->isVerbose() || Remark->getHotness());
 
   return true;
 }
@@ -214,12 +227,19 @@ LLVMContext::getDiagnosticMessagePrefix(DiagnosticSeverity Severity) {
 }
 
 void LLVMContext::diagnose(const DiagnosticInfo &DI) {
+  if (auto *OptDiagBase = dyn_cast<DiagnosticInfoOptimizationBase>(&DI)) {
+    yaml::Output *Out = getDiagnosticsOutputFile();
+    if (Out) {
+      // For remarks the << operator takes a reference to a pointer.
+      auto *P = const_cast<DiagnosticInfoOptimizationBase *>(OptDiagBase);
+      *Out << P;
+    }
+  }
   // If there is a report handler, use it.
-  if (pImpl->DiagnosticHandler) {
-    if (!pImpl->RespectDiagnosticFilters || isDiagnosticEnabled(DI))
-      pImpl->DiagnosticHandler(DI, pImpl->DiagnosticContext);
+  if (pImpl->DiagHandler &&
+      (!pImpl->RespectDiagnosticFilters || isDiagnosticEnabled(DI)) &&
+      pImpl->DiagHandler->handleDiagnostics(DI))
     return;
-  }
 
   if (!isDiagnosticEnabled(DI))
     return;
@@ -315,3 +335,11 @@ void LLVMContext::setDiscardValueNames(bool Discard) {
 OptBisect &LLVMContext::getOptBisect() {
   return pImpl->getOptBisect();
 }
+
+const DiagnosticHandler *LLVMContext::getDiagHandlerPtr() const {
+  return pImpl->DiagHandler.get();
+}
+
+std::unique_ptr<DiagnosticHandler> LLVMContext::getDiagnosticHandler() {
+  return std::move(pImpl->DiagHandler);
+}
diff --git a/lib/IR/LLVMContextImpl.cpp b/lib/IR/LLVMContextImpl.cpp
index 57dd08b36fe7..4b44a6b69cad 100644
--- a/lib/IR/LLVMContextImpl.cpp
+++ b/lib/IR/LLVMContextImpl.cpp
@@ -22,7 +22,8 @@
 using namespace llvm;
 
 LLVMContextImpl::LLVMContextImpl(LLVMContext &C)
-  : VoidTy(C, Type::VoidTyID),
+  : DiagHandler(llvm::make_unique<DiagnosticHandler>()),
+    VoidTy(C, Type::VoidTyID),
     LabelTy(C, Type::LabelTyID),
     HalfTy(C, Type::HalfTyID),
     FloatTy(C, Type::FloatTyID),
diff --git a/lib/IR/LLVMContextImpl.h b/lib/IR/LLVMContextImpl.h
index bea2c7ae8ff2..f41acfa8ea9c 100644
--- a/lib/IR/LLVMContextImpl.h
+++ b/lib/IR/LLVMContextImpl.h
@@ -1168,8 +1168,7 @@ public:
   LLVMContext::InlineAsmDiagHandlerTy InlineAsmDiagHandler = nullptr;
   void *InlineAsmDiagContext = nullptr;
 
-  LLVMContext::DiagnosticHandlerTy DiagnosticHandler = nullptr;
-  void *DiagnosticContext = nullptr;
+  std::unique_ptr<DiagnosticHandler> DiagHandler;
   bool RespectDiagnosticFilters = false;
   bool DiagnosticsHotnessRequested = false;
   uint64_t DiagnosticsHotnessThreshold = 0;
diff --git a/lib/IR/LegacyPassManager.cpp b/lib/IR/LegacyPassManager.cpp
index 995e1e570340..8bd9ed6ef0fa 100644
--- a/lib/IR/LegacyPassManager.cpp
+++ b/lib/IR/LegacyPassManager.cpp
@@ -75,21 +75,25 @@ PrintAfter("print-after",
            llvm::cl::desc("Print IR after specified passes"),
            cl::Hidden);
 
+static cl::opt<bool> PrintBeforeAll("print-before-all",
+                                    llvm::cl::desc("Print IR before each pass"),
+                                    cl::init(false), cl::Hidden);
+static cl::opt<bool> PrintAfterAll("print-after-all",
+                                   llvm::cl::desc("Print IR after each pass"),
+                                   cl::init(false), cl::Hidden);
+
 static cl::opt<bool>
-PrintBeforeAll("print-before-all",
-               llvm::cl::desc("Print IR before each pass"),
-               cl::init(false));
-static cl::opt<bool>
-PrintAfterAll("print-after-all",
-              llvm::cl::desc("Print IR after each pass"),
-              cl::init(false));
+    PrintModuleScope("print-module-scope",
+                     cl::desc("When printing IR for print-[before|after]{-all} "
+                              "always print a module IR"),
+                     cl::init(false));
 
 static cl::list<std::string>
     PrintFuncsList("filter-print-funcs", cl::value_desc("function names"),
                    cl::desc("Only print IR for functions whose name "
                             "match this for all print-[before|after][-all] "
                             "options"),
-                   cl::CommaSeparated);
+                   cl::CommaSeparated, cl::Hidden);
 
 /// This is a helper to determine whether to print IR before or
 /// after a pass.
@@ -117,6 +121,8 @@ static bool ShouldPrintAfterPass(const PassInfo *PI) {
   return PrintAfterAll || ShouldPrintBeforeOrAfterPass(PI, PrintAfter);
 }
 
+bool llvm::forcePrintModuleIR() { return PrintModuleScope; }
+
 bool llvm::isFunctionInPrintList(StringRef FunctionName) {
   static std::unordered_set<std::string> PrintFuncNames(PrintFuncsList.begin(),
                                                         PrintFuncsList.end());
@@ -1729,9 +1735,9 @@ bool PassManager::run(Module &M) {
 // TimingInfo implementation
 
 bool llvm::TimePassesIsEnabled = false;
-static cl::opt<bool,true>
-EnableTiming("time-passes", cl::location(TimePassesIsEnabled),
-            cl::desc("Time each pass, printing elapsed time for each on exit"));
+static cl::opt<bool, true> EnableTiming(
+    "time-passes", cl::location(TimePassesIsEnabled), cl::Hidden,
+    cl::desc("Time each pass, printing elapsed time for each on exit"));
 
 // createTheTimeInfo - This method either initializes the TheTimeInfo pointer to
 // a non-null value (if the -time-passes option is enabled) or it leaves it
diff --git a/lib/IR/MDBuilder.cpp b/lib/IR/MDBuilder.cpp
index b9c4f482adf5..6d77a8f2d601 100644
--- a/lib/IR/MDBuilder.cpp
+++ b/lib/IR/MDBuilder.cpp
@@ -14,6 +14,7 @@
 
 #include "llvm/IR/MDBuilder.h"
 #include "llvm/IR/Constants.h"
+#include "llvm/IR/Function.h"
 #include "llvm/IR/Metadata.h"
 using namespace llvm;
 
@@ -62,9 +63,14 @@ MDNode *MDBuilder::createFunctionEntryCount(
   SmallVector<Metadata *, 8> Ops;
   Ops.push_back(createString("function_entry_count"));
   Ops.push_back(createConstant(ConstantInt::get(Int64Ty, Count)));
-  if (Imports)
-    for (auto ID : *Imports)
+  if (Imports) {
+    SmallVector<GlobalValue::GUID, 2> OrderID(Imports->begin(), Imports->end());
+    std::stable_sort(OrderID.begin(), OrderID.end(),
+      [] (GlobalValue::GUID A, GlobalValue::GUID B) {
+        return A < B;});
+    for (auto ID : OrderID)
       Ops.push_back(createConstant(ConstantInt::get(Int64Ty, ID)));
+  }
   return MDNode::get(Context, Ops);
 }
 
@@ -90,6 +96,13 @@ MDNode *MDBuilder::createRange(Constant *Lo, Constant *Hi) {
   return MDNode::get(Context, {createConstant(Lo), createConstant(Hi)});
 }
 
+MDNode *MDBuilder::createCallees(ArrayRef<Function *> Callees) {
+  SmallVector<Metadata *, 4> Ops;
+  for (Function *F : Callees)
+    Ops.push_back(createConstant(F));
+  return MDNode::get(Context, Ops);
+}
+
 MDNode *MDBuilder::createAnonymousAARoot(StringRef Name, MDNode *Extra) {
   // To ensure uniqueness the root node is self-referential.
   auto Dummy = MDNode::getTemporary(Context, None);
@@ -144,7 +157,7 @@ MDNode *MDBuilder::createTBAAStructNode(ArrayRef<TBAAStructField> Fields) {
   for (unsigned i = 0, e = Fields.size(); i != e; ++i) {
     Vals[i * 3 + 0] = createConstant(ConstantInt::get(Int64, Fields[i].Offset));
     Vals[i * 3 + 1] = createConstant(ConstantInt::get(Int64, Fields[i].Size));
-    Vals[i * 3 + 2] = Fields[i].TBAA;
+    Vals[i * 3 + 2] = Fields[i].Type;
   }
   return MDNode::get(Context, Vals);
 }
@@ -184,3 +197,40 @@ MDNode *MDBuilder::createTBAAStructTagNode(MDNode *BaseType, MDNode *AccessType,
   }
   return MDNode::get(Context, {BaseType, AccessType, createConstant(Off)});
 }
+
+MDNode *MDBuilder::createTBAATypeNode(MDNode *Parent, uint64_t Size,
+                                      Metadata *Id,
+                                      ArrayRef<TBAAStructField> Fields) {
+  SmallVector<Metadata *, 4> Ops(3 + Fields.size() * 3);
+  Type *Int64 = Type::getInt64Ty(Context);
+  Ops[0] = Parent;
+  Ops[1] = createConstant(ConstantInt::get(Int64, Size));
+  Ops[2] = Id;
+  for (unsigned I = 0, E = Fields.size(); I != E; ++I) {
+    Ops[I * 3 + 3] = Fields[I].Type;
+    Ops[I * 3 + 4] = createConstant(ConstantInt::get(Int64, Fields[I].Offset));
+    Ops[I * 3 + 5] = createConstant(ConstantInt::get(Int64, Fields[I].Size));
+  }
+  return MDNode::get(Context, Ops);
+}
+
+MDNode *MDBuilder::createTBAAAccessTag(MDNode *BaseType, MDNode *AccessType,
+                                       uint64_t Offset, uint64_t Size,
+                                       bool IsImmutable) {
+  IntegerType *Int64 = Type::getInt64Ty(Context);
+  auto *OffsetNode = createConstant(ConstantInt::get(Int64, Offset));
+  auto *SizeNode = createConstant(ConstantInt::get(Int64, Size));
+  if (IsImmutable) {
+    auto *ImmutabilityFlagNode = createConstant(ConstantInt::get(Int64, 1));
+    return MDNode::get(Context, {BaseType, AccessType, OffsetNode, SizeNode,
+                                 ImmutabilityFlagNode});
+  }
+  return MDNode::get(Context, {BaseType, AccessType, OffsetNode, SizeNode});
+}
+
+MDNode *MDBuilder::createIrrLoopHeaderWeight(uint64_t Weight) {
+  SmallVector<Metadata *, 2> Vals(2);
+  Vals[0] = createString("loop_header_weight");
+  Vals[1] = createConstant(ConstantInt::get(Type::getInt64Ty(Context), Weight));
+  return MDNode::get(Context, Vals);
+}
diff --git a/lib/IR/Metadata.cpp b/lib/IR/Metadata.cpp
index ac02ff76c843..a148ab65fc83 100644
--- a/lib/IR/Metadata.cpp
+++ b/lib/IR/Metadata.cpp
@@ -1431,7 +1431,6 @@ void GlobalObject::setMetadata(StringRef Kind, MDNode *N) {
 MDNode *GlobalObject::getMetadata(unsigned KindID) const {
   SmallVector<MDNode *, 1> MDs;
   getMetadata(KindID, MDs);
-  assert(MDs.size() <= 1 && "Expected at most one metadata attachment");
   if (MDs.empty())
     return nullptr;
   return MDs[0];
diff --git a/lib/IR/OptBisect.cpp b/lib/IR/OptBisect.cpp
index f1c70058fac2..dc7dcd2e4a97 100644
--- a/lib/IR/OptBisect.cpp
+++ b/lib/IR/OptBisect.cpp
@@ -1,4 +1,4 @@
-//===------- llvm/IR/OptBisect/Bisect.cpp - LLVM Bisect support --------===//
+//===- llvm/IR/OptBisect/Bisect.cpp - LLVM Bisect support -----------------===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -6,31 +6,38 @@
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
-///
+//
 /// \file
 /// This file implements support for a bisecting optimizations based on a
 /// command line option.
-///
+//
 //===----------------------------------------------------------------------===//
 
 #include "llvm/IR/OptBisect.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Analysis/CallGraph.h"
 #include "llvm/Analysis/CallGraphSCCPass.h"
-#include "llvm/Analysis/LazyCallGraph.h"
 #include "llvm/Analysis/LoopInfo.h"
 #include "llvm/Analysis/RegionInfo.h"
+#include "llvm/IR/BasicBlock.h"
+#include "llvm/IR/Function.h"
 #include "llvm/IR/Module.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/raw_ostream.h"
+#include <cassert>
+#include <limits>
+#include <string>
 
 using namespace llvm;
 
 static cl::opt<int> OptBisectLimit("opt-bisect-limit", cl::Hidden,
-                                   cl::init(INT_MAX), cl::Optional,
+                                   cl::init(std::numeric_limits<int>::max()),
+                                   cl::Optional,
                                    cl::desc("Maximum optimization to perform"));
 
 OptBisect::OptBisect() {
-  BisectEnabled = OptBisectLimit != INT_MAX;
+  BisectEnabled = OptBisectLimit != std::numeric_limits<int>::max();
 }
 
 static void printPassMessage(const StringRef &Name, int PassNum,
diff --git a/lib/IR/Pass.cpp b/lib/IR/Pass.cpp
index f1b5f2f108dc..5e0b59476c4b 100644
--- a/lib/IR/Pass.cpp
+++ b/lib/IR/Pass.cpp
@@ -14,14 +14,22 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Pass.h"
+#include "llvm/IR/Attributes.h"
+#include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/IRPrintingPasses.h"
+#include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/LegacyPassNameParser.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/OptBisect.h"
+#include "llvm/PassInfo.h"
 #include "llvm/PassRegistry.h"
+#include "llvm/PassSupport.h"
+#include "llvm/Support/Compiler.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
+#include <cassert>
+
 using namespace llvm;
 
 #define DEBUG_TYPE "ir"
@@ -36,11 +44,11 @@ Pass::~Pass() {
 }
 
 // Force out-of-line virtual method.
-ModulePass::~ModulePass() { }
+ModulePass::~ModulePass() = default;
 
-Pass *ModulePass::createPrinterPass(raw_ostream &O,
+Pass *ModulePass::createPrinterPass(raw_ostream &OS,
                                     const std::string &Banner) const {
-  return createPrintModulePass(O, Banner);
+  return createPrintModulePass(OS, Banner);
 }
 
 PassManagerType ModulePass::getPotentialPassManagerType() const {
@@ -63,7 +71,6 @@ void Pass::dumpPassStructure(unsigned Offset) {
 /// getPassName - Return a nice clean name for a pass.  This usually
 /// implemented in terms of the name that is registered by one of the
 /// Registration templates, but can be overloaded directly.
-///
 StringRef Pass::getPassName() const {
   AnalysisID AID =  getPassID();
   const PassInfo *PI = PassRegistry::getPassRegistry()->getPassInfo(AID);
@@ -113,9 +120,8 @@ void Pass::setResolver(AnalysisResolver *AR) {
 // print - Print out the internal state of the pass.  This is called by Analyze
 // to print out the contents of an analysis.  Otherwise it is not necessary to
 // implement this method.
-//
-void Pass::print(raw_ostream &O,const Module*) const {
-  O << "Pass::print not implemented for pass: '" << getPassName() << "'!\n";
+void Pass::print(raw_ostream &OS, const Module *) const {
+  OS << "Pass::print not implemented for pass: '" << getPassName() << "'!\n";
 }
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
@@ -129,7 +135,7 @@ LLVM_DUMP_METHOD void Pass::dump() const {
 // ImmutablePass Implementation
 //
 // Force out-of-line virtual method.
-ImmutablePass::~ImmutablePass() { }
+ImmutablePass::~ImmutablePass() = default;
 
 void ImmutablePass::initializePass() {
   // By default, don't do anything.
@@ -139,9 +145,9 @@ void ImmutablePass::initializePass() {
 // FunctionPass Implementation
 //
 
-Pass *FunctionPass::createPrinterPass(raw_ostream &O,
+Pass *FunctionPass::createPrinterPass(raw_ostream &OS,
                                       const std::string &Banner) const {
-  return createPrintFunctionPass(O, Banner);
+  return createPrintFunctionPass(OS, Banner);
 }
 
 PassManagerType FunctionPass::getPotentialPassManagerType() const {
@@ -164,9 +170,9 @@ bool FunctionPass::skipFunction(const Function &F) const {
 // BasicBlockPass Implementation
 //
 
-Pass *BasicBlockPass::createPrinterPass(raw_ostream &O,
+Pass *BasicBlockPass::createPrinterPass(raw_ostream &OS,
                                         const std::string &Banner) const {
-  return createPrintBasicBlockPass(O, Banner);
+  return createPrintBasicBlockPass(OS, Banner);
 }
 
 bool BasicBlockPass::doInitialization(Function &) {
@@ -219,7 +225,7 @@ Pass *Pass::createPass(AnalysisID ID) {
 //===----------------------------------------------------------------------===//
 
 // RegisterAGBase implementation
-//
+
 RegisterAGBase::RegisterAGBase(StringRef Name, const void *InterfaceID,
                                const void *PassID, bool isDefault)
     : PassInfo(Name, InterfaceID) {
@@ -233,7 +239,6 @@ RegisterAGBase::RegisterAGBase(StringRef Name, const void *InterfaceID,
 
 // enumeratePasses - Iterate over the registered passes, calling the
 // passEnumerate callback on each PassInfo object.
-//
 void PassRegistrationListener::enumeratePasses() {
   PassRegistry::getPassRegistry()->enumerateWith(this);
 }
@@ -243,28 +248,31 @@ PassNameParser::PassNameParser(cl::Option &O)
   PassRegistry::getPassRegistry()->addRegistrationListener(this);
 }
 
-PassNameParser::~PassNameParser() {
-  // This only gets called during static destruction, in which case the
-  // PassRegistry will have already been destroyed by llvm_shutdown().  So
-  // attempting to remove the registration listener is an error.
-}
+// This only gets called during static destruction, in which case the
+// PassRegistry will have already been destroyed by llvm_shutdown().  So
+// attempting to remove the registration listener is an error.
+PassNameParser::~PassNameParser() = default;
 
 //===----------------------------------------------------------------------===//
 //   AnalysisUsage Class Implementation
 //
 
 namespace {
-  struct GetCFGOnlyPasses : public PassRegistrationListener {
-    typedef AnalysisUsage::VectorType VectorType;
-    VectorType &CFGOnlyList;
-    GetCFGOnlyPasses(VectorType &L) : CFGOnlyList(L) {}
 
-    void passEnumerate(const PassInfo *P) override {
-      if (P->isCFGOnlyPass())
-        CFGOnlyList.push_back(P->getTypeInfo());
-    }
-  };
-}
+struct GetCFGOnlyPasses : public PassRegistrationListener {
+  using VectorType = AnalysisUsage::VectorType;
+
+  VectorType &CFGOnlyList;
+
+  GetCFGOnlyPasses(VectorType &L) : CFGOnlyList(L) {}
+
+  void passEnumerate(const PassInfo *P) override {
+    if (P->isCFGOnlyPass())
+      CFGOnlyList.push_back(P->getTypeInfo());
+  }
+};
+
+} // end anonymous namespace
 
 // setPreservesCFG - This function should be called to by the pass, iff they do
 // not:
@@ -274,7 +282,6 @@ namespace {
 //
 // This function annotates the AnalysisUsage info object to say that analyses
 // that only depend on the CFG are preserved by this pass.
-//
 void AnalysisUsage::setPreservesCFG() {
   // Since this transformation doesn't modify the CFG, it preserves all analyses
   // that only depend on the CFG (like dominators, loop info, etc...)
diff --git a/lib/IR/PassRegistry.cpp b/lib/IR/PassRegistry.cpp
index c0f6f07169ff..b0f1a9928725 100644
--- a/lib/IR/PassRegistry.cpp
+++ b/lib/IR/PassRegistry.cpp
@@ -14,8 +14,12 @@
 
 #include "llvm/PassRegistry.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/PassInfo.h"
 #include "llvm/PassSupport.h"
 #include "llvm/Support/ManagedStatic.h"
+#include <cassert>
+#include <memory>
+#include <utility>
 
 using namespace llvm;
 
@@ -33,7 +37,7 @@ PassRegistry *PassRegistry::getPassRegistry() {
 // Accessors
 //
 
-PassRegistry::~PassRegistry() {}
+PassRegistry::~PassRegistry() = default;
 
 const PassInfo *PassRegistry::getPassInfo(const void *TI) const {
   sys::SmartScopedReader<true> Guard(Lock);
@@ -120,6 +124,6 @@ void PassRegistry::addRegistrationListener(PassRegistrationListener *L) {
 void PassRegistry::removeRegistrationListener(PassRegistrationListener *L) {
   sys::SmartScopedWriter<true> Guard(Lock);
 
-  auto I = find(Listeners, L);
+  auto I = llvm::find(Listeners, L);
   Listeners.erase(I);
 }
diff --git a/lib/IR/SafepointIRVerifier.cpp b/lib/IR/SafepointIRVerifier.cpp
index 8b328c221da3..68e0ce39a54e 100644
--- a/lib/IR/SafepointIRVerifier.cpp
+++ b/lib/IR/SafepointIRVerifier.cpp
@@ -32,6 +32,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/PostOrderIterator.h"
 #include "llvm/ADT/SetOperations.h"
 #include "llvm/ADT/SetVector.h"
 #include "llvm/IR/BasicBlock.h"
@@ -60,6 +61,7 @@ static cl::opt<bool> PrintOnly("safepoint-ir-verifier-print-only",
 
 static void Verify(const Function &F, const DominatorTree &DT);
 
+namespace {
 struct SafepointIRVerifier : public FunctionPass {
   static char ID; // Pass identification, replacement for typeid
   DominatorTree DT;
@@ -79,6 +81,7 @@ struct SafepointIRVerifier : public FunctionPass {
 
   StringRef getPassName() const override { return "safepoint verifier"; }
 };
+} // namespace
 
 void llvm::verifySafepointIR(Function &F) {
   SafepointIRVerifier pass;
@@ -134,92 +137,25 @@ static void PrintValueSet(raw_ostream &OS, IteratorTy Begin, IteratorTy End) {
 /// correctly relocated value at that point, and is a subset of the set of
 /// definitions dominating that point.
 
+using AvailableValueSet = DenseSet<const Value *>;
+
 /// State we compute and track per basic block.
 struct BasicBlockState {
   // Set of values available coming in, before the phi nodes
-  DenseSet<const Value *> AvailableIn;
+  AvailableValueSet AvailableIn;
 
   // Set of values available going out
-  DenseSet<const Value *> AvailableOut;
+  AvailableValueSet AvailableOut;
 
   // AvailableOut minus AvailableIn.
   // All elements are Instructions
-  DenseSet<const Value *> Contribution;
+  AvailableValueSet Contribution;
 
   // True if this block contains a safepoint and thus AvailableIn does not
   // contribute to AvailableOut.
   bool Cleared = false;
 };
 
-
-/// Gather all the definitions dominating the start of BB into Result.  This is
-/// simply the Defs introduced by every dominating basic block and the function
-/// arguments.
-static void GatherDominatingDefs(const BasicBlock *BB,
-                                 DenseSet<const Value *> &Result,
-                                 const DominatorTree &DT,
-                    DenseMap<const BasicBlock *, BasicBlockState *> &BlockMap) {
-  DomTreeNode *DTN = DT[const_cast<BasicBlock *>(BB)];
-
-  while (DTN->getIDom()) {
-    DTN = DTN->getIDom();
-    const auto &Defs = BlockMap[DTN->getBlock()]->Contribution;
-    Result.insert(Defs.begin(), Defs.end());
-    // If this block is 'Cleared', then nothing LiveIn to this block can be
-    // available after this block completes.  Note: This turns out to be 
-    // really important for reducing memory consuption of the initial available
-    // sets and thus peak memory usage by this verifier.
-    if (BlockMap[DTN->getBlock()]->Cleared)
-      return;
-  }
-
-  for (const Argument &A : BB->getParent()->args())
-    if (containsGCPtrType(A.getType()))
-      Result.insert(&A);
-}
-
-/// Model the effect of an instruction on the set of available values.
-static void TransferInstruction(const Instruction &I, bool &Cleared,
-                              DenseSet<const Value *> &Available) {
-  if (isStatepoint(I)) {
-    Cleared = true;
-    Available.clear();
-  } else if (containsGCPtrType(I.getType()))
-    Available.insert(&I);
-}
-
-/// Compute the AvailableOut set for BB, based on the
-/// BasicBlockState BBS, which is the BasicBlockState for BB. FirstPass is set
-/// when the verifier runs for the first time computing the AvailableOut set
-/// for BB.
-static void TransferBlock(const BasicBlock *BB,
-                          BasicBlockState &BBS, bool FirstPass) {
-
-  const DenseSet<const Value *> &AvailableIn = BBS.AvailableIn; 
-  DenseSet<const Value *> &AvailableOut  = BBS.AvailableOut;
-
-  if (BBS.Cleared) {
-    // AvailableOut does not change no matter how the input changes, just
-    // leave it be.  We need to force this calculation the first time so that
-    // we have a AvailableOut at all.
-    if (FirstPass) {
-      AvailableOut = BBS.Contribution;
-    }
-  } else {
-    // Otherwise, we need to reduce the AvailableOut set by things which are no
-    // longer in our AvailableIn
-    DenseSet<const Value *> Temp = BBS.Contribution;
-    set_union(Temp, AvailableIn);
-    AvailableOut = std::move(Temp);
-  }
-
-  DEBUG(dbgs() << "Transfered block " << BB->getName() << " from ";
-        PrintValueSet(dbgs(), AvailableIn.begin(), AvailableIn.end());
-        dbgs() << " to ";
-        PrintValueSet(dbgs(), AvailableOut.begin(), AvailableOut.end());
-        dbgs() << "\n";);
-}
-
 /// A given derived pointer can have multiple base pointers through phi/selects.
 /// This type indicates when the base pointer is exclusively constant
 /// (ExclusivelySomeConstant), and if that constant is proven to be exclusively
@@ -291,32 +227,164 @@ static enum BaseType getBaseType(const Value *Val) {
                                       : BaseType::ExclusivelySomeConstant;
 }
 
-static void Verify(const Function &F, const DominatorTree &DT) {
+static bool isNotExclusivelyConstantDerived(const Value *V) {
+  return getBaseType(V) == BaseType::NonConstant;
+}
+
+namespace {
+class InstructionVerifier;
+
+/// Builds BasicBlockState for each BB of the function.
+/// It can traverse function for verification and provides all required
+/// information.
+class GCPtrTracker {
+  const Function &F;
   SpecificBumpPtrAllocator<BasicBlockState> BSAllocator;
   DenseMap<const BasicBlock *, BasicBlockState *> BlockMap;
- 
-  DEBUG(dbgs() << "Verifying gc pointers in function: " << F.getName() << "\n");
-  if (PrintOnly)
-    dbgs() << "Verifying gc pointers in function: " << F.getName() << "\n";
+  // This set contains defs of unrelocated pointers that are proved to be legal
+  // and don't need verification.
+  DenseSet<const Instruction *> ValidUnrelocatedDefs;
+
+public:
+  GCPtrTracker(const Function &F, const DominatorTree &DT);
+
+  BasicBlockState *getBasicBlockState(const BasicBlock *BB);
+  const BasicBlockState *getBasicBlockState(const BasicBlock *BB) const;
+
+  /// Traverse each BB of the function and call
+  /// InstructionVerifier::verifyInstruction for each possibly invalid
+  /// instruction.
+  /// It destructively modifies GCPtrTracker so it's passed via rvalue reference
+  /// in order to prohibit further usages of GCPtrTracker as it'll be in
+  /// inconsistent state.
+  static void verifyFunction(GCPtrTracker &&Tracker,
+                             InstructionVerifier &Verifier);
+
+private:
+  /// Returns true if the instruction may be safely skipped during verification.
+  bool instructionMayBeSkipped(const Instruction *I) const;
+
+  /// Iterates over all BBs from BlockMap and recalculates AvailableIn/Out for
+  /// each of them until it converges.
+  void recalculateBBsStates();
+
+  /// Remove from Contribution all defs that legally produce unrelocated
+  /// pointers and saves them to ValidUnrelocatedDefs.
+  /// Though Contribution should belong to BBS it is passed separately with
+  /// different const-modifier in order to emphasize (and guarantee) that only
+  /// Contribution will be changed.
+  /// Returns true if Contribution was changed otherwise false.
+  bool removeValidUnrelocatedDefs(const BasicBlock *BB,
+                                  const BasicBlockState *BBS,
+                                  AvailableValueSet &Contribution);
+
+  /// Gather all the definitions dominating the start of BB into Result. This is
+  /// simply the defs introduced by every dominating basic block and the
+  /// function arguments.
+  void gatherDominatingDefs(const BasicBlock *BB, AvailableValueSet &Result,
+                            const DominatorTree &DT);
 
+  /// Compute the AvailableOut set for BB, based on the BasicBlockState BBS,
+  /// which is the BasicBlockState for BB.
+  /// ContributionChanged is set when the verifier runs for the first time
+  /// (in this case Contribution was changed from 'empty' to its initial state)
+  /// or when Contribution of this BB was changed since last computation.
+  static void transferBlock(const BasicBlock *BB, BasicBlockState &BBS,
+                            bool ContributionChanged);
 
+  /// Model the effect of an instruction on the set of available values.
+  static void transferInstruction(const Instruction &I, bool &Cleared,
+                                  AvailableValueSet &Available);
+};
+
+/// It is a visitor for GCPtrTracker::verifyFunction. It decides if the
+/// instruction (which uses heap reference) is legal or not, given our safepoint
+/// semantics.
+class InstructionVerifier {
+  bool AnyInvalidUses = false;
+
+public:
+  void verifyInstruction(const GCPtrTracker *Tracker, const Instruction &I,
+                         const AvailableValueSet &AvailableSet);
+
+  bool hasAnyInvalidUses() const { return AnyInvalidUses; }
+
+private:
+  void reportInvalidUse(const Value &V, const Instruction &I);
+};
+} // end anonymous namespace
+
+GCPtrTracker::GCPtrTracker(const Function &F, const DominatorTree &DT) : F(F) {
+  // First, calculate Contribution of each BB.
   for (const BasicBlock &BB : F) {
-    BasicBlockState *BBS = new(BSAllocator.Allocate()) BasicBlockState;
+    BasicBlockState *BBS = new (BSAllocator.Allocate()) BasicBlockState;
     for (const auto &I : BB)
-      TransferInstruction(I, BBS->Cleared, BBS->Contribution);
+      transferInstruction(I, BBS->Cleared, BBS->Contribution);
     BlockMap[&BB] = BBS;
   }
 
+  // Initialize AvailableIn/Out sets of each BB using only information about
+  // dominating BBs.
   for (auto &BBI : BlockMap) {
-    GatherDominatingDefs(BBI.first, BBI.second->AvailableIn, DT, BlockMap);
-    TransferBlock(BBI.first, *BBI.second, true);
+    gatherDominatingDefs(BBI.first, BBI.second->AvailableIn, DT);
+    transferBlock(BBI.first, *BBI.second, true);
   }
 
+  // Simulate the flow of defs through the CFG and recalculate AvailableIn/Out
+  // sets of each BB until it converges. If any def is proved to be an
+  // unrelocated pointer, it will be removed from all BBSs.
+  recalculateBBsStates();
+}
+
+BasicBlockState *GCPtrTracker::getBasicBlockState(const BasicBlock *BB) {
+  auto it = BlockMap.find(BB);
+  assert(it != BlockMap.end() &&
+         "No such BB in BlockMap! Probably BB from another function");
+  return it->second;
+}
+
+const BasicBlockState *GCPtrTracker::getBasicBlockState(
+    const BasicBlock *BB) const {
+  return const_cast<GCPtrTracker *>(this)->getBasicBlockState(BB);
+}
+
+bool GCPtrTracker::instructionMayBeSkipped(const Instruction *I) const {
+  return ValidUnrelocatedDefs.count(I);
+}
+
+void GCPtrTracker::verifyFunction(GCPtrTracker &&Tracker,
+                                  InstructionVerifier &Verifier) {
+  // We need RPO here to a) report always the first error b) report errors in
+  // same order from run to run.
+  ReversePostOrderTraversal<const Function *> RPOT(&Tracker.F);
+  for (const BasicBlock *BB : RPOT) {
+    BasicBlockState *BBS = Tracker.getBasicBlockState(BB);
+    // We destructively modify AvailableIn as we traverse the block instruction
+    // by instruction.
+    AvailableValueSet &AvailableSet = BBS->AvailableIn;
+    for (const Instruction &I : *BB) {
+      if (Tracker.instructionMayBeSkipped(&I))
+        continue; // This instruction shouldn't be added to AvailableSet.
+
+      Verifier.verifyInstruction(&Tracker, I, AvailableSet);
+
+      // Model the effect of current instruction on AvailableSet to keep the set
+      // relevant at each point of BB.
+      bool Cleared = false;
+      transferInstruction(I, Cleared, AvailableSet);
+      (void)Cleared;
+    }
+  }
+}
+
+void GCPtrTracker::recalculateBBsStates() {
   SetVector<const BasicBlock *> Worklist;
+  // TODO: This order is suboptimal, it's better to replace it with priority
+  // queue where priority is RPO number of BB.
   for (auto &BBI : BlockMap)
     Worklist.insert(BBI.first);
 
-  // This loop iterates the AvailableIn and AvailableOut sets to a fixed point.
+  // This loop iterates the AvailableIn/Out sets until it converges.
   // The AvailableIn and AvailableOut sets decrease as we iterate.
   while (!Worklist.empty()) {
     const BasicBlock *BB = Worklist.pop_back_val();
@@ -326,111 +394,205 @@ static void Verify(const Function &F, const DominatorTree &DT) {
     for (const BasicBlock *PBB : predecessors(BB))
       set_intersect(BBS->AvailableIn, BlockMap[PBB]->AvailableOut);
 
-    if (OldInCount == BBS->AvailableIn.size())
-      continue;
+    assert(OldInCount >= BBS->AvailableIn.size() && "invariant!");
 
-    assert(OldInCount > BBS->AvailableIn.size() && "invariant!");
+    bool InputsChanged = OldInCount != BBS->AvailableIn.size();
+    bool ContributionChanged =
+        removeValidUnrelocatedDefs(BB, BBS, BBS->Contribution);
+    if (!InputsChanged && !ContributionChanged)
+      continue;
 
     size_t OldOutCount = BBS->AvailableOut.size();
-    TransferBlock(BB, *BBS, false);
+    transferBlock(BB, *BBS, ContributionChanged);
     if (OldOutCount != BBS->AvailableOut.size()) {
       assert(OldOutCount > BBS->AvailableOut.size() && "invariant!");
       Worklist.insert(succ_begin(BB), succ_end(BB));
     }
   }
+}
 
-  // We now have all the information we need to decide if the use of a heap
-  // reference is legal or not, given our safepoint semantics.
+bool GCPtrTracker::removeValidUnrelocatedDefs(const BasicBlock *BB,
+                                              const BasicBlockState *BBS,
+                                              AvailableValueSet &Contribution) {
+  assert(&BBS->Contribution == &Contribution &&
+         "Passed Contribution should be from the passed BasicBlockState!");
+  AvailableValueSet AvailableSet = BBS->AvailableIn;
+  bool ContributionChanged = false;
+  for (const Instruction &I : *BB) {
+    bool ProducesUnrelocatedPointer = false;
+    if ((isa<GetElementPtrInst>(I) || isa<BitCastInst>(I)) &&
+        containsGCPtrType(I.getType())) {
+      // GEP/bitcast of unrelocated pointer is legal by itself but this
+      // def shouldn't appear in any AvailableSet.
+      for (const Value *V : I.operands())
+        if (containsGCPtrType(V->getType()) &&
+            isNotExclusivelyConstantDerived(V) && !AvailableSet.count(V)) {
+          ProducesUnrelocatedPointer = true;
+          break;
+        }
+    }
+    if (!ProducesUnrelocatedPointer) {
+      bool Cleared = false;
+      transferInstruction(I, Cleared, AvailableSet);
+      (void)Cleared;
+    } else {
+      // Remove def of unrelocated pointer from Contribution of this BB
+      // and trigger update of all its successors.
+      Contribution.erase(&I);
+      ValidUnrelocatedDefs.insert(&I);
+      DEBUG(dbgs() << "Removing " << I << " from Contribution of "
+                   << BB->getName() << "\n");
+      ContributionChanged = true;
+    }
+  }
+  return ContributionChanged;
+}
 
-  bool AnyInvalidUses = false;
+void GCPtrTracker::gatherDominatingDefs(const BasicBlock *BB,
+                                        AvailableValueSet &Result,
+                                        const DominatorTree &DT) {
+  DomTreeNode *DTN = DT[const_cast<BasicBlock *>(BB)];
 
-  auto ReportInvalidUse = [&AnyInvalidUses](const Value &V,
-                                            const Instruction &I) {
-    errs() << "Illegal use of unrelocated value found!\n";
-    errs() << "Def: " << V << "\n";
-    errs() << "Use: " << I << "\n";
-    if (!PrintOnly)
-      abort();
-    AnyInvalidUses = true;
-  };
+  while (DTN->getIDom()) {
+    DTN = DTN->getIDom();
+    const auto &Defs = BlockMap[DTN->getBlock()]->Contribution;
+    Result.insert(Defs.begin(), Defs.end());
+    // If this block is 'Cleared', then nothing LiveIn to this block can be
+    // available after this block completes.  Note: This turns out to be
+    // really important for reducing memory consuption of the initial available
+    // sets and thus peak memory usage by this verifier.
+    if (BlockMap[DTN->getBlock()]->Cleared)
+      return;
+  }
 
-  auto isNotExclusivelyConstantDerived = [](const Value *V) {
-    return getBaseType(V) == BaseType::NonConstant;
-  };
+  for (const Argument &A : BB->getParent()->args())
+    if (containsGCPtrType(A.getType()))
+      Result.insert(&A);
+}
 
-  for (const BasicBlock &BB : F) {
-    // We destructively modify AvailableIn as we traverse the block instruction
-    // by instruction.
-    DenseSet<const Value *> &AvailableSet = BlockMap[&BB]->AvailableIn;
-    for (const Instruction &I : BB) {
-      if (const PHINode *PN = dyn_cast<PHINode>(&I)) {
-        if (containsGCPtrType(PN->getType()))
-          for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
-            const BasicBlock *InBB = PN->getIncomingBlock(i);
-            const Value *InValue = PN->getIncomingValue(i);
+void GCPtrTracker::transferBlock(const BasicBlock *BB, BasicBlockState &BBS,
+                                 bool ContributionChanged) {
+  const AvailableValueSet &AvailableIn = BBS.AvailableIn;
+  AvailableValueSet &AvailableOut = BBS.AvailableOut;
+
+  if (BBS.Cleared) {
+    // AvailableOut will change only when Contribution changed.
+    if (ContributionChanged)
+      AvailableOut = BBS.Contribution;
+  } else {
+    // Otherwise, we need to reduce the AvailableOut set by things which are no
+    // longer in our AvailableIn
+    AvailableValueSet Temp = BBS.Contribution;
+    set_union(Temp, AvailableIn);
+    AvailableOut = std::move(Temp);
+  }
 
-            if (isNotExclusivelyConstantDerived(InValue) &&
-                !BlockMap[InBB]->AvailableOut.count(InValue))
-              ReportInvalidUse(*InValue, *PN);
-          }
-      } else if (isa<CmpInst>(I) &&
-                 containsGCPtrType(I.getOperand(0)->getType())) {
-        Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
-        enum BaseType baseTyLHS = getBaseType(LHS),
-                      baseTyRHS = getBaseType(RHS);
+  DEBUG(dbgs() << "Transfered block " << BB->getName() << " from ";
+        PrintValueSet(dbgs(), AvailableIn.begin(), AvailableIn.end());
+        dbgs() << " to ";
+        PrintValueSet(dbgs(), AvailableOut.begin(), AvailableOut.end());
+        dbgs() << "\n";);
+}
 
-        // Returns true if LHS and RHS are unrelocated pointers and they are
-        // valid unrelocated uses.
-        auto hasValidUnrelocatedUse = [&AvailableSet, baseTyLHS, baseTyRHS, &LHS, &RHS] () {
-            // A cmp instruction has valid unrelocated pointer operands only if
-            // both operands are unrelocated pointers.
-            // In the comparison between two pointers, if one is an unrelocated
-            // use, the other *should be* an unrelocated use, for this
-            // instruction to contain valid unrelocated uses. This unrelocated
-            // use can be a null constant as well, or another unrelocated
-            // pointer.
-            if (AvailableSet.count(LHS) || AvailableSet.count(RHS))
-              return false;
-            // Constant pointers (that are not exclusively null) may have
-            // meaning in different VMs, so we cannot reorder the compare
-            // against constant pointers before the safepoint. In other words,
-            // comparison of an unrelocated use against a non-null constant
-            // maybe invalid.
-            if ((baseTyLHS == BaseType::ExclusivelySomeConstant &&
-                 baseTyRHS == BaseType::NonConstant) ||
-                (baseTyLHS == BaseType::NonConstant &&
-                 baseTyRHS == BaseType::ExclusivelySomeConstant))
-              return false;
-            // All other cases are valid cases enumerated below:
-            // 1. Comparison between an exlusively derived null pointer and a
-            // constant base pointer.
-            // 2. Comparison between an exlusively derived null pointer and a
-            // non-constant unrelocated base pointer.
-            // 3. Comparison between 2 unrelocated pointers.
-            return true;
-        };
-        if (!hasValidUnrelocatedUse()) {
-          // Print out all non-constant derived pointers that are unrelocated
-          // uses, which are invalid.
-          if (baseTyLHS == BaseType::NonConstant && !AvailableSet.count(LHS))
-            ReportInvalidUse(*LHS, I);
-          if (baseTyRHS == BaseType::NonConstant && !AvailableSet.count(RHS))
-            ReportInvalidUse(*RHS, I);
-        }
-      } else {
-        for (const Value *V : I.operands())
-          if (containsGCPtrType(V->getType()) &&
-              isNotExclusivelyConstantDerived(V) && !AvailableSet.count(V))
-            ReportInvalidUse(*V, I);
+void GCPtrTracker::transferInstruction(const Instruction &I, bool &Cleared,
+                                       AvailableValueSet &Available) {
+  if (isStatepoint(I)) {
+    Cleared = true;
+    Available.clear();
+  } else if (containsGCPtrType(I.getType()))
+    Available.insert(&I);
+}
+
+void InstructionVerifier::verifyInstruction(
+    const GCPtrTracker *Tracker, const Instruction &I,
+    const AvailableValueSet &AvailableSet) {
+  if (const PHINode *PN = dyn_cast<PHINode>(&I)) {
+    if (containsGCPtrType(PN->getType()))
+      for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
+        const BasicBlock *InBB = PN->getIncomingBlock(i);
+        const Value *InValue = PN->getIncomingValue(i);
+
+        if (isNotExclusivelyConstantDerived(InValue) &&
+            !Tracker->getBasicBlockState(InBB)->AvailableOut.count(InValue))
+          reportInvalidUse(*InValue, *PN);
       }
+  } else if (isa<CmpInst>(I) &&
+             containsGCPtrType(I.getOperand(0)->getType())) {
+    Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
+    enum BaseType baseTyLHS = getBaseType(LHS),
+                  baseTyRHS = getBaseType(RHS);
 
-      bool Cleared = false;
-      TransferInstruction(I, Cleared, AvailableSet);
-      (void)Cleared;
+    // Returns true if LHS and RHS are unrelocated pointers and they are
+    // valid unrelocated uses.
+    auto hasValidUnrelocatedUse = [&AvailableSet, baseTyLHS, baseTyRHS, &LHS,
+                                   &RHS] () {
+        // A cmp instruction has valid unrelocated pointer operands only if
+        // both operands are unrelocated pointers.
+        // In the comparison between two pointers, if one is an unrelocated
+        // use, the other *should be* an unrelocated use, for this
+        // instruction to contain valid unrelocated uses. This unrelocated
+        // use can be a null constant as well, or another unrelocated
+        // pointer.
+        if (AvailableSet.count(LHS) || AvailableSet.count(RHS))
+          return false;
+        // Constant pointers (that are not exclusively null) may have
+        // meaning in different VMs, so we cannot reorder the compare
+        // against constant pointers before the safepoint. In other words,
+        // comparison of an unrelocated use against a non-null constant
+        // maybe invalid.
+        if ((baseTyLHS == BaseType::ExclusivelySomeConstant &&
+             baseTyRHS == BaseType::NonConstant) ||
+            (baseTyLHS == BaseType::NonConstant &&
+             baseTyRHS == BaseType::ExclusivelySomeConstant))
+          return false;
+        // All other cases are valid cases enumerated below:
+        // 1. Comparison between an exlusively derived null pointer and a
+        // constant base pointer.
+        // 2. Comparison between an exlusively derived null pointer and a
+        // non-constant unrelocated base pointer.
+        // 3. Comparison between 2 unrelocated pointers.
+        return true;
+    };
+    if (!hasValidUnrelocatedUse()) {
+      // Print out all non-constant derived pointers that are unrelocated
+      // uses, which are invalid.
+      if (baseTyLHS == BaseType::NonConstant && !AvailableSet.count(LHS))
+        reportInvalidUse(*LHS, I);
+      if (baseTyRHS == BaseType::NonConstant && !AvailableSet.count(RHS))
+        reportInvalidUse(*RHS, I);
     }
+  } else {
+    for (const Value *V : I.operands())
+      if (containsGCPtrType(V->getType()) &&
+          isNotExclusivelyConstantDerived(V) && !AvailableSet.count(V))
+        reportInvalidUse(*V, I);
   }
+}
+
+void InstructionVerifier::reportInvalidUse(const Value &V,
+                                           const Instruction &I) {
+  errs() << "Illegal use of unrelocated value found!\n";
+  errs() << "Def: " << V << "\n";
+  errs() << "Use: " << I << "\n";
+  if (!PrintOnly)
+    abort();
+  AnyInvalidUses = true;
+}
+
+static void Verify(const Function &F, const DominatorTree &DT) {
+  DEBUG(dbgs() << "Verifying gc pointers in function: " << F.getName() << "\n");
+  if (PrintOnly)
+    dbgs() << "Verifying gc pointers in function: " << F.getName() << "\n";
+
+  GCPtrTracker Tracker(F, DT);
+
+  // We now have all the information we need to decide if the use of a heap
+  // reference is legal or not, given our safepoint semantics.
+
+  InstructionVerifier Verifier;
+  GCPtrTracker::verifyFunction(std::move(Tracker), Verifier);
 
-  if (PrintOnly && !AnyInvalidUses) {
+  if (PrintOnly && !Verifier.hasAnyInvalidUses()) {
     dbgs() << "No illegal uses found by SafepointIRVerifier in: " << F.getName()
            << "\n";
   }
diff --git a/lib/IR/User.cpp b/lib/IR/User.cpp
index d46039107f33..041593f20b57 100644
--- a/lib/IR/User.cpp
+++ b/lib/IR/User.cpp
@@ -10,7 +10,6 @@
 #include "llvm/IR/User.h"
 #include "llvm/IR/Constant.h"
 #include "llvm/IR/GlobalValue.h"
-#include "llvm/IR/Operator.h"
 
 namespace llvm {
 class BasicBlock;
diff --git a/lib/IR/Value.cpp b/lib/IR/Value.cpp
index 51a7d424c1f3..eae697b2e4b9 100644
--- a/lib/IR/Value.cpp
+++ b/lib/IR/Value.cpp
@@ -15,6 +15,7 @@
 #include "LLVMContextImpl.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/SetVector.h"
 #include "llvm/IR/CallSite.h"
 #include "llvm/IR/Constant.h"
 #include "llvm/IR/Constants.h"
@@ -50,6 +51,7 @@ Value::Value(Type *ty, unsigned scid)
     : VTy(checkType(ty)), UseList(nullptr), SubclassID(scid),
       HasValueHandle(0), SubclassOptionalData(0), SubclassData(0),
       NumUserOperands(0), IsUsedByMD(false), HasName(false) {
+  static_assert(ConstantFirstVal == 0, "!(SubclassID < ConstantFirstVal)");
   // FIXME: Why isn't this in the subclass gunk??
   // Note, we cannot call isa<CallInst> before the CallInst has been
   // constructed.
@@ -57,7 +59,7 @@ Value::Value(Type *ty, unsigned scid)
     assert((VTy->isFirstClassType() || VTy->isVoidTy() || VTy->isStructTy()) &&
            "invalid CallInst type!");
   else if (SubclassID != BasicBlockVal &&
-           (SubclassID < ConstantFirstVal || SubclassID > ConstantLastVal))
+           (/*SubclassID < ConstantFirstVal ||*/ SubclassID > ConstantLastVal))
     assert((VTy->isFirstClassType() || VTy->isVoidTy()) &&
            "Cannot create non-first-class values except for constants!");
   static_assert(sizeof(Value) == 2 * sizeof(void *) + 2 * sizeof(unsigned),
@@ -407,7 +409,7 @@ void Value::doRAUW(Value *New, bool NoMetadata) {
   if (!NoMetadata && isUsedByMetadata())
     ValueAsMetadata::handleRAUW(this, New);
 
-  while (!use_empty()) {
+  while (!materialized_use_empty()) {
     Use &U = *UseList;
     // Must handle Constants specially, we cannot call replaceUsesOfWith on a
     // constant because they are uniqued.
@@ -454,6 +456,35 @@ void Value::replaceUsesOutsideBlock(Value *New, BasicBlock *BB) {
   }
 }
 
+void Value::replaceUsesExceptBlockAddr(Value *New) {
+  SmallSetVector<Constant *, 4> Constants;
+  use_iterator UI = use_begin(), E = use_end();
+  for (; UI != E;) {
+    Use &U = *UI;
+    ++UI;
+
+    if (isa<BlockAddress>(U.getUser()))
+      continue;
+
+    // Must handle Constants specially, we cannot call replaceUsesOfWith on a
+    // constant because they are uniqued.
+    if (auto *C = dyn_cast<Constant>(U.getUser())) {
+      if (!isa<GlobalValue>(C)) {
+        // Save unique users to avoid processing operand replacement
+        // more than once.
+        Constants.insert(C);
+        continue;
+      }
+    }
+
+    U.set(New);
+  }
+
+  // Process operand replacement of saved constants.
+  for (auto *C : Constants)
+    C->handleOperandChange(this, New);
+}
+
 namespace {
 // Various metrics for how much to strip off of pointers.
 enum PointerStripKind {
@@ -588,17 +619,17 @@ const Value *Value::stripInBoundsOffsets() const {
   return stripPointerCastsAndOffsets<PSK_InBounds>(this);
 }
 
-unsigned Value::getPointerDereferenceableBytes(const DataLayout &DL,
+uint64_t Value::getPointerDereferenceableBytes(const DataLayout &DL,
                                                bool &CanBeNull) const {
   assert(getType()->isPointerTy() && "must be pointer");
 
-  unsigned DerefBytes = 0;
+  uint64_t DerefBytes = 0;
   CanBeNull = false;
   if (const Argument *A = dyn_cast<Argument>(this)) {
     DerefBytes = A->getDereferenceableBytes();
-    if (DerefBytes == 0 && A->hasByValAttr() && A->getType()->isSized()) {
-      DerefBytes = DL.getTypeStoreSize(A->getType());
-      CanBeNull = false;
+    if (DerefBytes == 0 && A->hasByValAttr()) {
+      Type *PT = cast<PointerType>(A->getType())->getElementType();
+      DerefBytes = DL.getTypeStoreSize(PT);
     }
     if (DerefBytes == 0) {
       DerefBytes = A->getDereferenceableOrNullBytes();
@@ -624,8 +655,10 @@ unsigned Value::getPointerDereferenceableBytes(const DataLayout &DL,
       CanBeNull = true;
     }
   } else if (auto *AI = dyn_cast<AllocaInst>(this)) {
-    if (AI->getAllocatedType()->isSized()) {
-      DerefBytes = DL.getTypeStoreSize(AI->getAllocatedType());
+    const ConstantInt *ArraySize = dyn_cast<ConstantInt>(AI->getArraySize());
+    if (ArraySize && AI->getAllocatedType()->isSized()) {
+      DerefBytes = DL.getTypeStoreSize(AI->getAllocatedType()) *
+        ArraySize->getZExtValue();
       CanBeNull = false;
     }
   } else if (auto *GV = dyn_cast<GlobalVariable>(this)) {
diff --git a/lib/IR/ValueSymbolTable.cpp b/lib/IR/ValueSymbolTable.cpp
index ccdabe0817b4..0da1990c3a3f 100644
--- a/lib/IR/ValueSymbolTable.cpp
+++ b/lib/IR/ValueSymbolTable.cpp
@@ -13,7 +13,9 @@
 
 #include "llvm/IR/ValueSymbolTable.h"
 #include "llvm/ADT/SmallString.h"
+#include "llvm/ADT/Triple.h"
 #include "llvm/IR/GlobalValue.h"
+#include "llvm/IR/Module.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/Value.h"
 #include "llvm/Support/Casting.h"
@@ -45,8 +47,17 @@ ValueName *ValueSymbolTable::makeUniqueName(Value *V,
     // Trim any suffix off and append the next number.
     UniqueName.resize(BaseSize);
     raw_svector_ostream S(UniqueName);
-    if (isa<GlobalValue>(V))
-      S << ".";
+    if (auto *GV = dyn_cast<GlobalValue>(V)) {
+      // A dot is appended to mark it as clone during ABI demangling so that
+      // for example "_Z1fv" and "_Z1fv.1" both demangle to "f()", the second
+      // one being a clone.
+      // On NVPTX we cannot use a dot because PTX only allows [A-Za-z0-9_$] for
+      // identifiers. This breaks ABI demangling but at least ptxas accepts and
+      // compiles the program.
+      const Module *M = GV->getParent();
+      if (!(M && Triple(M->getTargetTriple()).isNVPTX()))
+        S << ".";
+    }
     S << ++LastUnique;
 
     // Try insert the vmap entry with this suffix.
diff --git a/lib/IR/ValueTypes.cpp b/lib/IR/ValueTypes.cpp
index cf6ee063c2d5..22f9fe7a66d7 100644
--- a/lib/IR/ValueTypes.cpp
+++ b/lib/IR/ValueTypes.cpp
@@ -14,7 +14,6 @@
 #include "llvm/CodeGen/ValueTypes.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/IR/DerivedTypes.h"
-#include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Type.h"
 #include "llvm/Support/ErrorHandling.h"
 using namespace llvm;
@@ -149,6 +148,7 @@ std::string EVT::getEVTString() const {
   case MVT::v16i1:   return "v16i1";
   case MVT::v32i1:   return "v32i1";
   case MVT::v64i1:   return "v64i1";
+  case MVT::v128i1:  return "v128i1";
   case MVT::v512i1:  return "v512i1";
   case MVT::v1024i1: return "v1024i1";
   case MVT::v1i8:    return "v1i8";
@@ -228,6 +228,7 @@ Type *EVT::getTypeForEVT(LLVMContext &Context) const {
   case MVT::v16i1:   return VectorType::get(Type::getInt1Ty(Context), 16);
   case MVT::v32i1:   return VectorType::get(Type::getInt1Ty(Context), 32);
   case MVT::v64i1:   return VectorType::get(Type::getInt1Ty(Context), 64);
+  case MVT::v128i1:  return VectorType::get(Type::getInt1Ty(Context), 128);
   case MVT::v512i1:  return VectorType::get(Type::getInt1Ty(Context), 512);
   case MVT::v1024i1: return VectorType::get(Type::getInt1Ty(Context), 1024);
   case MVT::v1i8:    return VectorType::get(Type::getInt8Ty(Context), 1);
diff --git a/lib/IR/Verifier.cpp b/lib/IR/Verifier.cpp
index 454a56a76923..534104686d81 100644
--- a/lib/IR/Verifier.cpp
+++ b/lib/IR/Verifier.cpp
@@ -75,7 +75,6 @@
 #include "llvm/IR/DebugInfoMetadata.h"
 #include "llvm/IR/DebugLoc.h"
 #include "llvm/IR/DerivedTypes.h"
-#include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/Dominators.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalAlias.h"
@@ -115,8 +114,6 @@
 
 using namespace llvm;
 
-static cl::opt<bool> VerifyDebugInfo("verify-debug-info", cl::init(true));
-
 namespace llvm {
 
 struct VerifierSupport {
@@ -468,8 +465,7 @@ private:
   void visitUserOp2(Instruction &I) { visitUserOp1(I); }
   void visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS);
   void visitConstrainedFPIntrinsic(ConstrainedFPIntrinsic &FPI);
-  template <class DbgIntrinsicTy>
-  void visitDbgIntrinsic(StringRef Kind, DbgIntrinsicTy &DII);
+  void visitDbgIntrinsic(StringRef Kind, DbgInfoIntrinsic &DII);
   void visitAtomicCmpXchgInst(AtomicCmpXchgInst &CXI);
   void visitAtomicRMWInst(AtomicRMWInst &RMWI);
   void visitFenceInst(FenceInst &FI);
@@ -507,6 +503,10 @@ private:
   void verifySiblingFuncletUnwinds();
 
   void verifyFragmentExpression(const DbgInfoIntrinsic &I);
+  template <typename ValueOrMetadata>
+  void verifyFragmentExpression(const DIVariable &V,
+                                DIExpression::FragmentInfo Fragment,
+                                ValueOrMetadata *Desc);
   void verifyFnArgs(const DbgInfoIntrinsic &I);
 
   /// Module-level debug info verification...
@@ -565,6 +565,10 @@ void Verifier::visitGlobalValue(const GlobalValue &GV) {
   if (GV.isDeclarationForLinker())
     Assert(!GV.hasComdat(), "Declaration may not be in a Comdat!", &GV);
 
+  if (GV.hasDLLImportStorageClass())
+    Assert(!GV.isDSOLocal(),
+           "GlobalValue with DLLImport Storage is dso_local!", &GV);
+
   forEachUser(&GV, GlobalValueVisited, [&](const Value *V) -> bool {
     if (const Instruction *I = dyn_cast<Instruction>(V)) {
       if (!I->getParent() || !I->getParent()->getParent())
@@ -839,6 +843,8 @@ void Verifier::visitDILocation(const DILocation &N) {
            "location requires a valid scope", &N, N.getRawScope());
   if (auto *IA = N.getRawInlinedAt())
     AssertDI(isa<DILocation>(IA), "inlined-at should be a location", &N, IA);
+  if (auto *SP = dyn_cast<DISubprogram>(N.getRawScope()))
+    AssertDI(SP->isDefinition(), "scope points into the type hierarchy", &N);
 }
 
 void Verifier::visitGenericDINode(const GenericDINode &N) {
@@ -1067,6 +1073,8 @@ void Verifier::visitDILexicalBlockBase(const DILexicalBlockBase &N) {
   AssertDI(N.getTag() == dwarf::DW_TAG_lexical_block, "invalid tag", &N);
   AssertDI(N.getRawScope() && isa<DILocalScope>(N.getRawScope()),
            "invalid local scope", &N, N.getRawScope());
+  if (auto *SP = dyn_cast<DISubprogram>(N.getRawScope()))
+    AssertDI(SP->isDefinition(), "scope points into the type hierarchy", &N);
 }
 
 void Verifier::visitDILexicalBlock(const DILexicalBlock &N) {
@@ -1139,7 +1147,6 @@ void Verifier::visitDITemplateValueParameter(
 void Verifier::visitDIVariable(const DIVariable &N) {
   if (auto *S = N.getRawScope())
     AssertDI(isa<DIScope>(S), "invalid scope", &N, S);
-  AssertDI(isType(N.getRawType()), "invalid type ref", &N, N.getRawType());
   if (auto *F = N.getRawFile())
     AssertDI(isa<DIFile>(F), "invalid file", &N, F);
 }
@@ -1150,6 +1157,8 @@ void Verifier::visitDIGlobalVariable(const DIGlobalVariable &N) {
 
   AssertDI(N.getTag() == dwarf::DW_TAG_variable, "invalid tag", &N);
   AssertDI(!N.getName().empty(), "missing global variable name", &N);
+  AssertDI(isType(N.getRawType()), "invalid type ref", &N, N.getRawType());
+  AssertDI(N.getType(), "missing global variable type", &N);
   if (auto *Member = N.getRawStaticDataMemberDeclaration()) {
     AssertDI(isa<DIDerivedType>(Member),
              "invalid static data member declaration", &N, Member);
@@ -1160,6 +1169,7 @@ void Verifier::visitDILocalVariable(const DILocalVariable &N) {
   // Checks common to all variables.
   visitDIVariable(N);
 
+  AssertDI(isType(N.getRawType()), "invalid type ref", &N, N.getRawType());
   AssertDI(N.getTag() == dwarf::DW_TAG_variable, "invalid tag", &N);
   AssertDI(N.getRawScope() && isa<DILocalScope>(N.getRawScope()),
            "local variable requires a valid scope", &N, N.getRawScope());
@@ -1174,8 +1184,11 @@ void Verifier::visitDIGlobalVariableExpression(
   AssertDI(GVE.getVariable(), "missing variable");
   if (auto *Var = GVE.getVariable())
     visitDIGlobalVariable(*Var);
-  if (auto *Expr = GVE.getExpression())
+  if (auto *Expr = GVE.getExpression()) {
     visitDIExpression(*Expr);
+    if (auto Fragment = Expr->getFragmentInfo())
+      verifyFragmentExpression(*GVE.getVariable(), *Fragment, &GVE);
+  }
 }
 
 void Verifier::visitDIObjCProperty(const DIObjCProperty &N) {
@@ -1361,6 +1374,7 @@ static bool isFuncOnlyAttr(Attribute::AttrKind Kind) {
   case Attribute::NonLazyBind:
   case Attribute::ReturnsTwice:
   case Attribute::SanitizeAddress:
+  case Attribute::SanitizeHWAddress:
   case Attribute::SanitizeThread:
   case Attribute::SanitizeMemory:
   case Attribute::MinSize:
@@ -1377,6 +1391,7 @@ static bool isFuncOnlyAttr(Attribute::AttrKind Kind) {
   case Attribute::InaccessibleMemOrArgMemOnly:
   case Attribute::AllocSize:
   case Attribute::Speculatable:
+  case Attribute::StrictFP:
     return true;
   default:
     break;
@@ -3001,7 +3016,11 @@ void Verifier::visitGetElementPtrInst(GetElementPtrInst &GEP) {
   Assert(isa<PointerType>(TargetTy),
          "GEP base pointer is not a vector or a vector of pointers", &GEP);
   Assert(GEP.getSourceElementType()->isSized(), "GEP into unsized type!", &GEP);
+
   SmallVector<Value*, 16> Idxs(GEP.idx_begin(), GEP.idx_end());
+  Assert(all_of(
+      Idxs, [](Value* V) { return V->getType()->isIntOrIntVectorTy(); }),
+      "GEP indexes must be integers", &GEP);
   Type *ElTy =
       GetElementPtrInst::getIndexedType(GEP.getSourceElementType(), Idxs);
   Assert(ElTy, "Invalid indices for GEP pointer type!", &GEP);
@@ -3969,6 +3988,7 @@ void Verifier::visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS) {
   case Intrinsic::experimental_constrained_fmul:
   case Intrinsic::experimental_constrained_fdiv:
   case Intrinsic::experimental_constrained_frem:
+  case Intrinsic::experimental_constrained_fma:
   case Intrinsic::experimental_constrained_sqrt:
   case Intrinsic::experimental_constrained_pow:
   case Intrinsic::experimental_constrained_powi:
@@ -3987,10 +4007,13 @@ void Verifier::visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS) {
   case Intrinsic::dbg_declare: // llvm.dbg.declare
     Assert(isa<MetadataAsValue>(CS.getArgOperand(0)),
            "invalid llvm.dbg.declare intrinsic call 1", CS);
-    visitDbgIntrinsic("declare", cast<DbgDeclareInst>(*CS.getInstruction()));
+    visitDbgIntrinsic("declare", cast<DbgInfoIntrinsic>(*CS.getInstruction()));
+    break;
+  case Intrinsic::dbg_addr: // llvm.dbg.addr
+    visitDbgIntrinsic("addr", cast<DbgInfoIntrinsic>(*CS.getInstruction()));
     break;
   case Intrinsic::dbg_value: // llvm.dbg.value
-    visitDbgIntrinsic("value", cast<DbgValueInst>(*CS.getInstruction()));
+    visitDbgIntrinsic("value", cast<DbgInfoIntrinsic>(*CS.getInstruction()));
     break;
   case Intrinsic::memcpy:
   case Intrinsic::memmove:
@@ -4008,9 +4031,7 @@ void Verifier::visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS) {
     break;
   }
   case Intrinsic::memcpy_element_unordered_atomic: {
-    const ElementUnorderedAtomicMemCpyInst *MI =
-        cast<ElementUnorderedAtomicMemCpyInst>(CS.getInstruction());
-    ;
+    const AtomicMemCpyInst *MI = cast<AtomicMemCpyInst>(CS.getInstruction());
 
     ConstantInt *ElementSizeCI =
         dyn_cast<ConstantInt>(MI->getRawElementSizeInBytes());
@@ -4045,7 +4066,7 @@ void Verifier::visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS) {
     break;
   }
   case Intrinsic::memmove_element_unordered_atomic: {
-    auto *MI = cast<ElementUnorderedAtomicMemMoveInst>(CS.getInstruction());
+    auto *MI = cast<AtomicMemMoveInst>(CS.getInstruction());
 
     ConstantInt *ElementSizeCI =
         dyn_cast<ConstantInt>(MI->getRawElementSizeInBytes());
@@ -4080,7 +4101,7 @@ void Verifier::visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS) {
     break;
   }
   case Intrinsic::memset_element_unordered_atomic: {
-    auto *MI = cast<ElementUnorderedAtomicMemSetInst>(CS.getInstruction());
+    auto *MI = cast<AtomicMemSetInst>(CS.getInstruction());
 
     ConstantInt *ElementSizeCI =
         dyn_cast<ConstantInt>(MI->getRawElementSizeInBytes());
@@ -4429,8 +4450,9 @@ static DISubprogram *getSubprogram(Metadata *LocalScope) {
 
 void Verifier::visitConstrainedFPIntrinsic(ConstrainedFPIntrinsic &FPI) {
   unsigned NumOperands = FPI.getNumArgOperands();
-  Assert(((NumOperands == 3 && FPI.isUnaryOp()) || (NumOperands == 4)),
-         "invalid arguments for constrained FP intrinsic", &FPI);
+  Assert(((NumOperands == 5 && FPI.isTernaryOp()) ||
+          (NumOperands == 3 && FPI.isUnaryOp()) || (NumOperands == 4)),
+           "invalid arguments for constrained FP intrinsic", &FPI);
   Assert(isa<MetadataAsValue>(FPI.getArgOperand(NumOperands-1)),
          "invalid exception behavior argument", &FPI);
   Assert(isa<MetadataAsValue>(FPI.getArgOperand(NumOperands-2)),
@@ -4441,8 +4463,7 @@ void Verifier::visitConstrainedFPIntrinsic(ConstrainedFPIntrinsic &FPI) {
          "invalid exception behavior argument", &FPI);
 }
 
-template <class DbgIntrinsicTy>
-void Verifier::visitDbgIntrinsic(StringRef Kind, DbgIntrinsicTy &DII) {
+void Verifier::visitDbgIntrinsic(StringRef Kind, DbgInfoIntrinsic &DII) {
   auto *MD = cast<MetadataAsValue>(DII.getArgOperand(0))->getMetadata();
   AssertDI(isa<ValueAsMetadata>(MD) ||
              (isa<MDNode>(MD) && !cast<MDNode>(MD)->getNumOperands()),
@@ -4481,46 +4502,15 @@ void Verifier::visitDbgIntrinsic(StringRef Kind, DbgIntrinsicTy &DII) {
   verifyFnArgs(DII);
 }
 
-static uint64_t getVariableSize(const DILocalVariable &V) {
-  // Be careful of broken types (checked elsewhere).
-  const Metadata *RawType = V.getRawType();
-  while (RawType) {
-    // Try to get the size directly.
-    if (auto *T = dyn_cast<DIType>(RawType))
-      if (uint64_t Size = T->getSizeInBits())
-        return Size;
-
-    if (auto *DT = dyn_cast<DIDerivedType>(RawType)) {
-      // Look at the base type.
-      RawType = DT->getRawBaseType();
-      continue;
-    }
-
-    // Missing type or size.
-    break;
-  }
-
-  // Fail gracefully.
-  return 0;
-}
-
 void Verifier::verifyFragmentExpression(const DbgInfoIntrinsic &I) {
-  DILocalVariable *V;
-  DIExpression *E;
-  if (auto *DVI = dyn_cast<DbgValueInst>(&I)) {
-    V = dyn_cast_or_null<DILocalVariable>(DVI->getRawVariable());
-    E = dyn_cast_or_null<DIExpression>(DVI->getRawExpression());
-  } else {
-    auto *DDI = cast<DbgDeclareInst>(&I);
-    V = dyn_cast_or_null<DILocalVariable>(DDI->getRawVariable());
-    E = dyn_cast_or_null<DIExpression>(DDI->getRawExpression());
-  }
+  DILocalVariable *V = dyn_cast_or_null<DILocalVariable>(I.getRawVariable());
+  DIExpression *E = dyn_cast_or_null<DIExpression>(I.getRawExpression());
 
   // We don't know whether this intrinsic verified correctly.
   if (!V || !E || !E->isValid())
     return;
 
-  // Nothing to do if this isn't a bit piece expression.
+  // Nothing to do if this isn't a DW_OP_LLVM_fragment expression.
   auto Fragment = E->getFragmentInfo();
   if (!Fragment)
     return;
@@ -4534,17 +4524,24 @@ void Verifier::verifyFragmentExpression(const DbgInfoIntrinsic &I) {
   if (V->isArtificial())
     return;
 
+  verifyFragmentExpression(*V, *Fragment, &I);
+}
+
+template <typename ValueOrMetadata>
+void Verifier::verifyFragmentExpression(const DIVariable &V,
+                                        DIExpression::FragmentInfo Fragment,
+                                        ValueOrMetadata *Desc) {
   // If there's no size, the type is broken, but that should be checked
   // elsewhere.
-  uint64_t VarSize = getVariableSize(*V);
+  auto VarSize = V.getSizeInBits();
   if (!VarSize)
     return;
 
-  unsigned FragSize = Fragment->SizeInBits;
-  unsigned FragOffset = Fragment->OffsetInBits;
-  AssertDI(FragSize + FragOffset <= VarSize,
-         "fragment is larger than or outside of variable", &I, V, E);
-  AssertDI(FragSize != VarSize, "fragment covers entire variable", &I, V, E);
+  unsigned FragSize = Fragment.SizeInBits;
+  unsigned FragOffset = Fragment.OffsetInBits;
+  AssertDI(FragSize + FragOffset <= *VarSize,
+         "fragment is larger than or outside of variable", Desc, &V);
+  AssertDI(FragSize != *VarSize, "fragment covers entire variable", Desc, &V);
 }
 
 void Verifier::verifyFnArgs(const DbgInfoIntrinsic &I) {
@@ -4554,18 +4551,11 @@ void Verifier::verifyFnArgs(const DbgInfoIntrinsic &I) {
   if (!HasDebugInfo)
     return;
 
-  DILocalVariable *Var;
-  if (auto *DV = dyn_cast<DbgValueInst>(&I)) {
-    // For performance reasons only check non-inlined ones.
-    if (DV->getDebugLoc()->getInlinedAt())
-      return;
-    Var = DV->getVariable();
-  } else {
-    auto *DD = cast<DbgDeclareInst>(&I);
-    if (DD->getDebugLoc()->getInlinedAt())
-      return;
-    Var = DD->getVariable();
-  }
+  // For performance reasons only check non-inlined ones.
+  if (I.getDebugLoc()->getInlinedAt())
+    return;
+
+  DILocalVariable *Var = I.getVariable();
   AssertDI(Var, "dbg intrinsic without variable");
 
   unsigned ArgNo = Var->getArg();
@@ -4584,6 +4574,11 @@ void Verifier::verifyFnArgs(const DbgInfoIntrinsic &I) {
 }
 
 void Verifier::verifyCompileUnits() {
+  // When more than one Module is imported into the same context, such as during
+  // an LTO build before linking the modules, ODR type uniquing may cause types
+  // to point to a different CU. This check does not make sense in this case.
+  if (M.getContext().isODRUniquingDebugTypes())
+    return;
   auto *CUs = M.getNamedMetadata("llvm.dbg.cu");
   SmallPtrSet<const Metadata *, 2> Listed;
   if (CUs)
@@ -4675,19 +4670,8 @@ struct VerifierLegacyPass : public FunctionPass {
         HasErrors |= !V->verify(F);
 
     HasErrors |= !V->verify();
-    if (FatalErrors) {
-      if (HasErrors)
-        report_fatal_error("Broken module found, compilation aborted!");
-      assert(!V->hasBrokenDebugInfo() && "Module contains invalid debug info");
-    }
-
-    // Strip broken debug info.
-    if (V->hasBrokenDebugInfo()) {
-      DiagnosticInfoIgnoringInvalidDebugMetadata DiagInvalid(M);
-      M.getContext().diagnose(DiagInvalid);
-      if (!StripDebugInfo(M))
-        report_fatal_error("Failed to strip malformed debug info");
-    }
+    if (FatalErrors && (HasErrors || V->hasBrokenDebugInfo()))
+      report_fatal_error("Broken module found, compilation aborted!");
     return false;
   }
 
@@ -4716,7 +4700,8 @@ template <typename... Tys> void TBAAVerifier::CheckFailed(Tys &&... Args) {
 /// TBAA scheme.  This means \p BaseNode is either a scalar node, or a
 /// struct-type node describing an aggregate data structure (like a struct).
 TBAAVerifier::TBAABaseNodeSummary
-TBAAVerifier::verifyTBAABaseNode(Instruction &I, const MDNode *BaseNode) {
+TBAAVerifier::verifyTBAABaseNode(Instruction &I, const MDNode *BaseNode,
+                                 bool IsNewFormat) {
   if (BaseNode->getNumOperands() < 2) {
     CheckFailed("Base nodes must have at least two operands", &I, BaseNode);
     return {true, ~0u};
@@ -4726,7 +4711,7 @@ TBAAVerifier::verifyTBAABaseNode(Instruction &I, const MDNode *BaseNode) {
   if (Itr != TBAABaseNodes.end())
     return Itr->second;
 
-  auto Result = verifyTBAABaseNodeImpl(I, BaseNode);
+  auto Result = verifyTBAABaseNodeImpl(I, BaseNode, IsNewFormat);
   auto InsertResult = TBAABaseNodes.insert({BaseNode, Result});
   (void)InsertResult;
   assert(InsertResult.second && "We just checked!");
@@ -4734,7 +4719,8 @@ TBAAVerifier::verifyTBAABaseNode(Instruction &I, const MDNode *BaseNode) {
 }
 
 TBAAVerifier::TBAABaseNodeSummary
-TBAAVerifier::verifyTBAABaseNodeImpl(Instruction &I, const MDNode *BaseNode) {
+TBAAVerifier::verifyTBAABaseNodeImpl(Instruction &I, const MDNode *BaseNode,
+                                     bool IsNewFormat) {
   const TBAAVerifier::TBAABaseNodeSummary InvalidNode = {true, ~0u};
 
   if (BaseNode->getNumOperands() == 2) {
@@ -4744,13 +4730,32 @@ TBAAVerifier::verifyTBAABaseNodeImpl(Instruction &I, const MDNode *BaseNode) {
                : InvalidNode;
   }
 
-  if (BaseNode->getNumOperands() % 2 != 1) {
-    CheckFailed("Struct tag nodes must have an odd number of operands!",
-                BaseNode);
-    return InvalidNode;
+  if (IsNewFormat) {
+    if (BaseNode->getNumOperands() % 3 != 0) {
+      CheckFailed("Access tag nodes must have the number of operands that is a "
+                  "multiple of 3!", BaseNode);
+      return InvalidNode;
+    }
+  } else {
+    if (BaseNode->getNumOperands() % 2 != 1) {
+      CheckFailed("Struct tag nodes must have an odd number of operands!",
+                  BaseNode);
+      return InvalidNode;
+    }
   }
 
-  if (!isa<MDString>(BaseNode->getOperand(0))) {
+  // Check the type size field.
+  if (IsNewFormat) {
+    auto *TypeSizeNode = mdconst::dyn_extract_or_null<ConstantInt>(
+        BaseNode->getOperand(1));
+    if (!TypeSizeNode) {
+      CheckFailed("Type size nodes must be constants!", &I, BaseNode);
+      return InvalidNode;
+    }
+  }
+
+  // Check the type name field. In the new format it can be anything.
+  if (!IsNewFormat && !isa<MDString>(BaseNode->getOperand(0))) {
     CheckFailed("Struct tag nodes have a string as their first operand",
                 BaseNode);
     return InvalidNode;
@@ -4763,7 +4768,10 @@ TBAAVerifier::verifyTBAABaseNodeImpl(Instruction &I, const MDNode *BaseNode) {
 
   // We've already checked that BaseNode is not a degenerate root node with one
   // operand in \c verifyTBAABaseNode, so this loop should run at least once.
-  for (unsigned Idx = 1; Idx < BaseNode->getNumOperands(); Idx += 2) {
+  unsigned FirstFieldOpNo = IsNewFormat ? 3 : 1;
+  unsigned NumOpsPerField = IsNewFormat ? 3 : 2;
+  for (unsigned Idx = FirstFieldOpNo; Idx < BaseNode->getNumOperands();
+           Idx += NumOpsPerField) {
     const MDOperand &FieldTy = BaseNode->getOperand(Idx);
     const MDOperand &FieldOffset = BaseNode->getOperand(Idx + 1);
     if (!isa<MDNode>(FieldTy)) {
@@ -4805,6 +4813,16 @@ TBAAVerifier::verifyTBAABaseNodeImpl(Instruction &I, const MDNode *BaseNode) {
     }
 
     PrevOffset = OffsetEntryCI->getValue();
+
+    if (IsNewFormat) {
+      auto *MemberSizeNode = mdconst::dyn_extract_or_null<ConstantInt>(
+          BaseNode->getOperand(Idx + 2));
+      if (!MemberSizeNode) {
+        CheckFailed("Member size entries must be constants!", &I, BaseNode);
+        Failed = true;
+        continue;
+      }
+    }
   }
 
   return Failed ? InvalidNode
@@ -4854,7 +4872,8 @@ bool TBAAVerifier::isValidScalarTBAANode(const MDNode *MD) {
 /// We assume we've okayed \p BaseNode via \c verifyTBAABaseNode.
 MDNode *TBAAVerifier::getFieldNodeFromTBAABaseNode(Instruction &I,
                                                    const MDNode *BaseNode,
-                                                   APInt &Offset) {
+                                                   APInt &Offset,
+                                                   bool IsNewFormat) {
   assert(BaseNode->getNumOperands() >= 2 && "Invalid base node!");
 
   // Scalar nodes have only one possible "field" -- their parent in the access
@@ -4863,35 +4882,52 @@ MDNode *TBAAVerifier::getFieldNodeFromTBAABaseNode(Instruction &I,
   if (BaseNode->getNumOperands() == 2)
     return cast<MDNode>(BaseNode->getOperand(1));
 
-  for (unsigned Idx = 1; Idx < BaseNode->getNumOperands(); Idx += 2) {
+  unsigned FirstFieldOpNo = IsNewFormat ? 3 : 1;
+  unsigned NumOpsPerField = IsNewFormat ? 3 : 2;
+  for (unsigned Idx = FirstFieldOpNo; Idx < BaseNode->getNumOperands();
+           Idx += NumOpsPerField) {
     auto *OffsetEntryCI =
         mdconst::extract<ConstantInt>(BaseNode->getOperand(Idx + 1));
     if (OffsetEntryCI->getValue().ugt(Offset)) {
-      if (Idx == 1) {
+      if (Idx == FirstFieldOpNo) {
         CheckFailed("Could not find TBAA parent in struct type node", &I,
                     BaseNode, &Offset);
         return nullptr;
       }
 
+      unsigned PrevIdx = Idx - NumOpsPerField;
       auto *PrevOffsetEntryCI =
-          mdconst::extract<ConstantInt>(BaseNode->getOperand(Idx - 1));
+          mdconst::extract<ConstantInt>(BaseNode->getOperand(PrevIdx + 1));
       Offset -= PrevOffsetEntryCI->getValue();
-      return cast<MDNode>(BaseNode->getOperand(Idx - 2));
+      return cast<MDNode>(BaseNode->getOperand(PrevIdx));
     }
   }
 
+  unsigned LastIdx = BaseNode->getNumOperands() - NumOpsPerField;
   auto *LastOffsetEntryCI = mdconst::extract<ConstantInt>(
-      BaseNode->getOperand(BaseNode->getNumOperands() - 1));
-
+      BaseNode->getOperand(LastIdx + 1));
   Offset -= LastOffsetEntryCI->getValue();
-  return cast<MDNode>(BaseNode->getOperand(BaseNode->getNumOperands() - 2));
+  return cast<MDNode>(BaseNode->getOperand(LastIdx));
+}
+
+static bool isNewFormatTBAATypeNode(llvm::MDNode *Type) {
+  if (!Type || Type->getNumOperands() < 3)
+    return false;
+
+  // In the new format type nodes shall have a reference to the parent type as
+  // its first operand.
+  MDNode *Parent = dyn_cast_or_null<MDNode>(Type->getOperand(0));
+  if (!Parent)
+    return false;
+
+  return true;
 }
 
 bool TBAAVerifier::visitTBAAMetadata(Instruction &I, const MDNode *MD) {
   AssertTBAA(isa<LoadInst>(I) || isa<StoreInst>(I) || isa<CallInst>(I) ||
                  isa<VAArgInst>(I) || isa<AtomicRMWInst>(I) ||
                  isa<AtomicCmpXchgInst>(I),
-             "TBAA is only for loads, stores and calls!", &I);
+             "This instruction shall not have a TBAA access tag!", &I);
 
   bool IsStructPathTBAA =
       isa<MDNode>(MD->getOperand(0)) && MD->getNumOperands() >= 3;
@@ -4900,18 +4936,34 @@ bool TBAAVerifier::visitTBAAMetadata(Instruction &I, const MDNode *MD) {
       IsStructPathTBAA,
       "Old-style TBAA is no longer allowed, use struct-path TBAA instead", &I);
 
-  AssertTBAA(MD->getNumOperands() < 5,
-             "Struct tag metadata must have either 3 or 4 operands", &I, MD);
-
   MDNode *BaseNode = dyn_cast_or_null<MDNode>(MD->getOperand(0));
   MDNode *AccessType = dyn_cast_or_null<MDNode>(MD->getOperand(1));
 
-  if (MD->getNumOperands() == 4) {
-    auto *IsImmutableCI =
-        mdconst::dyn_extract_or_null<ConstantInt>(MD->getOperand(3));
+  bool IsNewFormat = isNewFormatTBAATypeNode(AccessType);
+
+  if (IsNewFormat) {
+    AssertTBAA(MD->getNumOperands() == 4 || MD->getNumOperands() == 5,
+               "Access tag metadata must have either 4 or 5 operands", &I, MD);
+  } else {
+    AssertTBAA(MD->getNumOperands() < 5,
+               "Struct tag metadata must have either 3 or 4 operands", &I, MD);
+  }
+
+  // Check the access size field.
+  if (IsNewFormat) {
+    auto *AccessSizeNode = mdconst::dyn_extract_or_null<ConstantInt>(
+        MD->getOperand(3));
+    AssertTBAA(AccessSizeNode, "Access size field must be a constant", &I, MD);
+  }
+
+  // Check the immutability flag.
+  unsigned ImmutabilityFlagOpNo = IsNewFormat ? 4 : 3;
+  if (MD->getNumOperands() == ImmutabilityFlagOpNo + 1) {
+    auto *IsImmutableCI = mdconst::dyn_extract_or_null<ConstantInt>(
+        MD->getOperand(ImmutabilityFlagOpNo));
     AssertTBAA(IsImmutableCI,
-               "Immutability tag on struct tag metadata must be a constant", &I,
-               MD);
+               "Immutability tag on struct tag metadata must be a constant",
+               &I, MD);
     AssertTBAA(
         IsImmutableCI->isZero() || IsImmutableCI->isOne(),
         "Immutability part of the struct tag metadata must be either 0 or 1",
@@ -4919,13 +4971,15 @@ bool TBAAVerifier::visitTBAAMetadata(Instruction &I, const MDNode *MD) {
   }
 
   AssertTBAA(BaseNode && AccessType,
-             "Malformed struct tag metadata:  base and access-type "
+             "Malformed struct tag metadata: base and access-type "
              "should be non-null and point to Metadata nodes",
              &I, MD, BaseNode, AccessType);
 
-  AssertTBAA(isValidScalarTBAANode(AccessType),
-             "Access type node must be a valid scalar type", &I, MD,
-             AccessType);
+  if (!IsNewFormat) {
+    AssertTBAA(isValidScalarTBAANode(AccessType),
+               "Access type node must be a valid scalar type", &I, MD,
+               AccessType);
+  }
 
   auto *OffsetCI = mdconst::dyn_extract_or_null<ConstantInt>(MD->getOperand(2));
   AssertTBAA(OffsetCI, "Offset must be constant integer", &I, MD);
@@ -4936,7 +4990,8 @@ bool TBAAVerifier::visitTBAAMetadata(Instruction &I, const MDNode *MD) {
   SmallPtrSet<MDNode *, 4> StructPath;
 
   for (/* empty */; BaseNode && !IsRootTBAANode(BaseNode);
-       BaseNode = getFieldNodeFromTBAABaseNode(I, BaseNode, Offset)) {
+       BaseNode = getFieldNodeFromTBAABaseNode(I, BaseNode, Offset,
+                                               IsNewFormat)) {
     if (!StructPath.insert(BaseNode).second) {
       CheckFailed("Cycle detected in struct path", &I, MD);
       return false;
@@ -4944,7 +4999,8 @@ bool TBAAVerifier::visitTBAAMetadata(Instruction &I, const MDNode *MD) {
 
     bool Invalid;
     unsigned BaseNodeBitWidth;
-    std::tie(Invalid, BaseNodeBitWidth) = verifyTBAABaseNode(I, BaseNode);
+    std::tie(Invalid, BaseNodeBitWidth) = verifyTBAABaseNode(I, BaseNode,
+                                                             IsNewFormat);
 
     // If the base node is invalid in itself, then we've already printed all the
     // errors we wanted to print.
@@ -4958,9 +5014,13 @@ bool TBAAVerifier::visitTBAAMetadata(Instruction &I, const MDNode *MD) {
                  &I, MD, &Offset);
 
     AssertTBAA(BaseNodeBitWidth == Offset.getBitWidth() ||
-                   (BaseNodeBitWidth == 0 && Offset == 0),
+                   (BaseNodeBitWidth == 0 && Offset == 0) ||
+                   (IsNewFormat && BaseNodeBitWidth == ~0u),
                "Access bit-width not the same as description bit-width", &I, MD,
                BaseNodeBitWidth, Offset.getBitWidth());
+
+    if (IsNewFormat && SeenAccessTypeInPath)
+      break;
   }
 
   AssertTBAA(SeenAccessTypeInPath, "Did not see access type in access path!",
@@ -4990,19 +5050,9 @@ VerifierAnalysis::Result VerifierAnalysis::run(Function &F,
 
 PreservedAnalyses VerifierPass::run(Module &M, ModuleAnalysisManager &AM) {
   auto Res = AM.getResult<VerifierAnalysis>(M);
-  if (FatalErrors) {
-    if (Res.IRBroken)
-      report_fatal_error("Broken module found, compilation aborted!");
-    assert(!Res.DebugInfoBroken && "Module contains invalid debug info");
-  }
+  if (FatalErrors && (Res.IRBroken || Res.DebugInfoBroken))
+    report_fatal_error("Broken module found, compilation aborted!");
 
-  // Strip broken debug info.
-  if (Res.DebugInfoBroken) {
-    DiagnosticInfoIgnoringInvalidDebugMetadata DiagInvalid(M);
-    M.getContext().diagnose(DiagInvalid);
-    if (!StripDebugInfo(M))
-      report_fatal_error("Failed to strip malformed debug info");
-  }
   return PreservedAnalyses::all();
 }