X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTransforms%2FUtils%2FSimplifyLibCalls.cpp;h=e7eb39d6312fa595b9724412cdfd3434116761fc;hp=95653cc790ff7901a67c8c5ff08c6b8875421ddc;hb=276f2af5364cc935dc486129ab73618e373150ab;hpb=b0ec6c21b7364421f329a5e8274c6232514c401b

diff --git a/lib/Transforms/Utils/SimplifyLibCalls.cpp b/lib/Transforms/Utils/SimplifyLibCalls.cpp
index 95653cc790f..e7eb39d6312 100644
--- a/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -18,6 +18,7 @@
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/StringMap.h"
 #include "llvm/ADT/Triple.h"
+#include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/DiagnosticInfo.h"
@@ -30,8 +31,8 @@
 #include "llvm/IR/PatternMatch.h"
 #include "llvm/Support/Allocator.h"
 #include "llvm/Support/CommandLine.h"
-#include "llvm/Target/TargetLibraryInfo.h"
 #include "llvm/Transforms/Utils/BuildLibCalls.h"
+#include "llvm/Transforms/Utils/Local.h"
 
 using namespace llvm;
 using namespace PatternMatch;
@@ -52,16 +53,8 @@ static cl::opt<bool>
 //===----------------------------------------------------------------------===//
 
 static bool ignoreCallingConv(LibFunc::Func Func) {
-  switch (Func) {
-  case LibFunc::abs:
-  case LibFunc::labs:
-  case LibFunc::llabs:
-  case LibFunc::strlen:
-    return true;
-  default:
-    return false;
-  }
-  llvm_unreachable("All cases should be covered in the switch.");
+  return Func == LibFunc::abs || Func == LibFunc::labs ||
+         Func == LibFunc::llabs || Func == LibFunc::strlen;
 }
 
 /// isOnlyUsedInZeroEqualityComparison - Return true if it only matters that the
@@ -93,16 +86,13 @@ static bool isOnlyUsedInEqualityComparison(Value *V, Value *With) {
 }
 
 static bool callHasFloatingPointArgument(const CallInst *CI) {
-  for (CallInst::const_op_iterator it = CI->op_begin(), e = CI->op_end();
-       it != e; ++it) {
-    if ((*it)->getType()->isFloatingPointTy())
-      return true;
-  }
-  return false;
+  return std::any_of(CI->op_begin(), CI->op_end(), [](const Use &OI) {
+    return OI->getType()->isFloatingPointTy();
+  });
 }
 
 /// \brief Check whether the overloaded unary floating point function
-/// corresponing to \a Ty is available.
+/// corresponding to \a Ty is available.
 static bool hasUnaryFloatFn(const TargetLibraryInfo *TLI, Type *Ty,
                             LibFunc::Func DoubleFn, LibFunc::Func FloatFn,
                             LibFunc::Func LongDoubleFn) {
@@ -116,207 +106,85 @@ static bool hasUnaryFloatFn(const TargetLibraryInfo *TLI, Type *Ty,
   }
 }
 
-//===----------------------------------------------------------------------===//
-// Fortified Library Call Optimizations
-//===----------------------------------------------------------------------===//
+/// \brief Check whether we can use unsafe floating point math for
+/// the function passed as input.
+static bool canUseUnsafeFPMath(Function *F) {
 
-static bool isFortifiedCallFoldable(CallInst *CI, unsigned SizeCIOp, unsigned SizeArgOp,
-                       bool isString) {
-  if (CI->getArgOperand(SizeCIOp) == CI->getArgOperand(SizeArgOp))
-    return true;
-  if (ConstantInt *SizeCI =
-          dyn_cast<ConstantInt>(CI->getArgOperand(SizeCIOp))) {
-    if (SizeCI->isAllOnesValue())
+  // FIXME: For finer-grain optimization, we need intrinsics to have the same
+  // fast-math flag decorations that are applied to FP instructions. For now,
+  // we have to rely on the function-level unsafe-fp-math attribute to do this
+  // optimization because there's no other way to express that the call can be
+  // relaxed.
+  if (F->hasFnAttribute("unsafe-fp-math")) {
+    Attribute Attr = F->getFnAttribute("unsafe-fp-math");
+    if (Attr.getValueAsString() == "true")
       return true;
-    if (isString) {
-      uint64_t Len = GetStringLength(CI->getArgOperand(SizeArgOp));
-      // If the length is 0 we don't know how long it is and so we can't
-      // remove the check.
-      if (Len == 0)
-        return false;
-      return SizeCI->getZExtValue() >= Len;
-    }
-    if (ConstantInt *Arg = dyn_cast<ConstantInt>(CI->getArgOperand(SizeArgOp)))
-      return SizeCI->getZExtValue() >= Arg->getZExtValue();
   }
   return false;
 }
 
-Value *LibCallSimplifier::optimizeMemCpyChk(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  LLVMContext &Context = CI->getContext();
-
-  // Check if this has the right signature.
-  if (FT->getNumParams() != 4 || FT->getReturnType() != FT->getParamType(0) ||
-      !FT->getParamType(0)->isPointerTy() ||
-      !FT->getParamType(1)->isPointerTy() ||
-      FT->getParamType(2) != DL->getIntPtrType(Context) ||
-      FT->getParamType(3) != DL->getIntPtrType(Context))
-    return nullptr;
-
-  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
-    B.CreateMemCpy(CI->getArgOperand(0), CI->getArgOperand(1),
-                   CI->getArgOperand(2), 1);
-    return CI->getArgOperand(0);
-  }
-  return nullptr;
-}
-
-Value *LibCallSimplifier::optimizeMemMoveChk(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  LLVMContext &Context = CI->getContext();
-
-  // Check if this has the right signature.
-  if (FT->getNumParams() != 4 || FT->getReturnType() != FT->getParamType(0) ||
-      !FT->getParamType(0)->isPointerTy() ||
-      !FT->getParamType(1)->isPointerTy() ||
-      FT->getParamType(2) != DL->getIntPtrType(Context) ||
-      FT->getParamType(3) != DL->getIntPtrType(Context))
-    return nullptr;
-
-  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
-    B.CreateMemMove(CI->getArgOperand(0), CI->getArgOperand(1),
-                    CI->getArgOperand(2), 1);
-    return CI->getArgOperand(0);
-  }
-  return nullptr;
-}
-
-Value *LibCallSimplifier::optimizeMemSetChk(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  LLVMContext &Context = CI->getContext();
-
-  // Check if this has the right signature.
-  if (FT->getNumParams() != 4 || FT->getReturnType() != FT->getParamType(0) ||
-      !FT->getParamType(0)->isPointerTy() ||
-      !FT->getParamType(1)->isIntegerTy() ||
-      FT->getParamType(2) != DL->getIntPtrType(Context) ||
-      FT->getParamType(3) != DL->getIntPtrType(Context))
-    return nullptr;
-
-  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
-    Value *Val = B.CreateIntCast(CI->getArgOperand(1), B.getInt8Ty(), false);
-    B.CreateMemSet(CI->getArgOperand(0), Val, CI->getArgOperand(2), 1);
-    return CI->getArgOperand(0);
-  }
-  return nullptr;
-}
-
-Value *LibCallSimplifier::optimizeStrCpyChk(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  StringRef Name = Callee->getName();
-  FunctionType *FT = Callee->getFunctionType();
-  LLVMContext &Context = CI->getContext();
-
-  // Check if this has the right signature.
-  if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
-      FT->getParamType(0) != FT->getParamType(1) ||
-      FT->getParamType(0) != Type::getInt8PtrTy(Context) ||
-      FT->getParamType(2) != DL->getIntPtrType(Context))
-    return nullptr;
-
-  Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
-  if (Dst == Src) // __strcpy_chk(x,x)  -> x
-    return Src;
-
-  // If a) we don't have any length information, or b) we know this will
-  // fit then just lower to a plain strcpy. Otherwise we'll keep our
-  // strcpy_chk call which may fail at runtime if the size is too long.
-  // TODO: It might be nice to get a maximum length out of the possible
-  // string lengths for varying.
-  if (isFortifiedCallFoldable(CI, 2, 1, true)) {
-    Value *Ret = EmitStrCpy(Dst, Src, B, DL, TLI, Name.substr(2, 6));
-    return Ret;
-  } else {
-    // Maybe we can stil fold __strcpy_chk to __memcpy_chk.
-    uint64_t Len = GetStringLength(Src);
-    if (Len == 0)
-      return nullptr;
-
-    // This optimization require DataLayout.
-    if (!DL)
-      return nullptr;
-
-    Value *Ret = EmitMemCpyChk(
-        Dst, Src, ConstantInt::get(DL->getIntPtrType(Context), Len),
-        CI->getArgOperand(2), B, DL, TLI);
-    return Ret;
-  }
-  return nullptr;
-}
-
-Value *LibCallSimplifier::optimizeStpCpyChk(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  StringRef Name = Callee->getName();
-  FunctionType *FT = Callee->getFunctionType();
-  LLVMContext &Context = CI->getContext();
-
-  // Check if this has the right signature.
-  if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
-      FT->getParamType(0) != FT->getParamType(1) ||
-      FT->getParamType(0) != Type::getInt8PtrTy(Context) ||
-      FT->getParamType(2) != DL->getIntPtrType(FT->getParamType(0)))
-    return nullptr;
-
-  Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
-  if (Dst == Src) { // stpcpy(x,x)  -> x+strlen(x)
-    Value *StrLen = EmitStrLen(Src, B, DL, TLI);
-    return StrLen ? B.CreateInBoundsGEP(Dst, StrLen) : nullptr;
-  }
-
-  // If a) we don't have any length information, or b) we know this will
-  // fit then just lower to a plain stpcpy. Otherwise we'll keep our
-  // stpcpy_chk call which may fail at runtime if the size is too long.
-  // TODO: It might be nice to get a maximum length out of the possible
-  // string lengths for varying.
-  if (isFortifiedCallFoldable(CI, 2, 1, true)) {
-    Value *Ret = EmitStrCpy(Dst, Src, B, DL, TLI, Name.substr(2, 6));
-    return Ret;
-  } else {
-    // Maybe we can stil fold __stpcpy_chk to __memcpy_chk.
-    uint64_t Len = GetStringLength(Src);
-    if (Len == 0)
-      return nullptr;
-
-    // This optimization require DataLayout.
-    if (!DL)
-      return nullptr;
-
-    Type *PT = FT->getParamType(0);
-    Value *LenV = ConstantInt::get(DL->getIntPtrType(PT), Len);
-    Value *DstEnd =
-        B.CreateGEP(Dst, ConstantInt::get(DL->getIntPtrType(PT), Len - 1));
-    if (!EmitMemCpyChk(Dst, Src, LenV, CI->getArgOperand(2), B, DL, TLI))
-      return nullptr;
-    return DstEnd;
-  }
-  return nullptr;
-}
-
-Value *LibCallSimplifier::optimizeStrNCpyChk(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  StringRef Name = Callee->getName();
-  FunctionType *FT = Callee->getFunctionType();
-  LLVMContext &Context = CI->getContext();
-
-  // Check if this has the right signature.
-  if (FT->getNumParams() != 4 || FT->getReturnType() != FT->getParamType(0) ||
-      FT->getParamType(0) != FT->getParamType(1) ||
-      FT->getParamType(0) != Type::getInt8PtrTy(Context) ||
-      !FT->getParamType(2)->isIntegerTy() ||
-      FT->getParamType(3) != DL->getIntPtrType(Context))
-    return nullptr;
+/// \brief Returns whether \p F matches the signature expected for the
+/// string/memory copying library function \p Func.
+/// Acceptable functions are st[rp][n]?cpy, memove, memcpy, and memset.
+/// Their fortified (_chk) counterparts are also accepted.
+static bool checkStringCopyLibFuncSignature(Function *F, LibFunc::Func Func) {
+  const DataLayout &DL = F->getParent()->getDataLayout();
+  FunctionType *FT = F->getFunctionType();
+  LLVMContext &Context = F->getContext();
+  Type *PCharTy = Type::getInt8PtrTy(Context);
+  Type *SizeTTy = DL.getIntPtrType(Context);
+  unsigned NumParams = FT->getNumParams();
+
+  // All string libfuncs return the same type as the first parameter.
+  if (FT->getReturnType() != FT->getParamType(0))
+    return false;
 
-  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
-    Value *Ret =
-        EmitStrNCpy(CI->getArgOperand(0), CI->getArgOperand(1),
-                    CI->getArgOperand(2), B, DL, TLI, Name.substr(2, 7));
-    return Ret;
-  }
-  return nullptr;
+  switch (Func) {
+  default:
+    llvm_unreachable("Can't check signature for non-string-copy libfunc.");
+  case LibFunc::stpncpy_chk:
+  case LibFunc::strncpy_chk:
+    --NumParams; // fallthrough
+  case LibFunc::stpncpy:
+  case LibFunc::strncpy: {
+    if (NumParams != 3 || FT->getParamType(0) != FT->getParamType(1) ||
+        FT->getParamType(0) != PCharTy || !FT->getParamType(2)->isIntegerTy())
+      return false;
+    break;
+  }
+  case LibFunc::strcpy_chk:
+  case LibFunc::stpcpy_chk:
+    --NumParams; // fallthrough
+  case LibFunc::stpcpy:
+  case LibFunc::strcpy: {
+    if (NumParams != 2 || FT->getParamType(0) != FT->getParamType(1) ||
+        FT->getParamType(0) != PCharTy)
+      return false;
+    break;
+  }
+  case LibFunc::memmove_chk:
+  case LibFunc::memcpy_chk:
+    --NumParams; // fallthrough
+  case LibFunc::memmove:
+  case LibFunc::memcpy: {
+    if (NumParams != 3 || !FT->getParamType(0)->isPointerTy() ||
+        !FT->getParamType(1)->isPointerTy() || FT->getParamType(2) != SizeTTy)
+      return false;
+    break;
+  }
+  case LibFunc::memset_chk:
+    --NumParams; // fallthrough
+  case LibFunc::memset: {
+    if (NumParams != 3 || !FT->getParamType(0)->isPointerTy() ||
+        !FT->getParamType(1)->isIntegerTy() || FT->getParamType(2) != SizeTTy)
+      return false;
+    break;
+  }
+  }
+  // If this is a fortified libcall, the last parameter is a size_t.
+  if (NumParams == FT->getNumParams() - 1)
+    return FT->getParamType(FT->getNumParams() - 1) == SizeTTy;
+  return true;
 }
 
 //===----------------------------------------------------------------------===//
@@ -347,10 +215,6 @@ Value *LibCallSimplifier::optimizeStrCat(CallInst *CI, IRBuilder<> &B) {
   if (Len == 0)
     return Dst;
 
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
-
   return emitStrLenMemCpy(Src, Dst, Len, B);
 }
 
@@ -365,13 +229,13 @@ Value *LibCallSimplifier::emitStrLenMemCpy(Value *Src, Value *Dst, uint64_t Len,
   // Now that we have the destination's length, we must index into the
   // destination's pointer to get the actual memcpy destination (end of
   // the string .. we're concatenating).
-  Value *CpyDst = B.CreateGEP(Dst, DstLen, "endptr");
+  Value *CpyDst = B.CreateGEP(B.getInt8Ty(), Dst, DstLen, "endptr");
 
   // We have enough information to now generate the memcpy call to do the
   // concatenation for us.  Make a memcpy to copy the nul byte with align = 1.
-  B.CreateMemCpy(
-      CpyDst, Src,
-      ConstantInt::get(DL->getIntPtrType(Src->getContext()), Len + 1), 1);
+  B.CreateMemCpy(CpyDst, Src,
+                 ConstantInt::get(DL.getIntPtrType(Src->getContext()), Len + 1),
+                 1);
   return Dst;
 }
 
@@ -408,10 +272,6 @@ Value *LibCallSimplifier::optimizeStrNCat(CallInst *CI, IRBuilder<> &B) {
   if (SrcLen == 0 || Len == 0)
     return Dst;
 
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
-
   // We don't optimize this case
   if (Len < SrcLen)
     return nullptr;
@@ -436,25 +296,21 @@ Value *LibCallSimplifier::optimizeStrChr(CallInst *CI, IRBuilder<> &B) {
   // of the input string and turn this into memchr.
   ConstantInt *CharC = dyn_cast<ConstantInt>(CI->getArgOperand(1));
   if (!CharC) {
-    // These optimizations require DataLayout.
-    if (!DL)
-      return nullptr;
-
     uint64_t Len = GetStringLength(SrcStr);
     if (Len == 0 || !FT->getParamType(1)->isIntegerTy(32)) // memchr needs i32.
       return nullptr;
 
-    return EmitMemChr(
-        SrcStr, CI->getArgOperand(1), // include nul.
-        ConstantInt::get(DL->getIntPtrType(CI->getContext()), Len), B, DL, TLI);
+    return EmitMemChr(SrcStr, CI->getArgOperand(1), // include nul.
+                      ConstantInt::get(DL.getIntPtrType(CI->getContext()), Len),
+                      B, DL, TLI);
   }
 
   // Otherwise, the character is a constant, see if the first argument is
   // a string literal.  If so, we can constant fold.
   StringRef Str;
   if (!getConstantStringInfo(SrcStr, Str)) {
-    if (DL && CharC->isZero()) // strchr(p, 0) -> p + strlen(p)
-      return B.CreateGEP(SrcStr, EmitStrLen(SrcStr, B, DL, TLI), "strchr");
+    if (CharC->isZero()) // strchr(p, 0) -> p + strlen(p)
+      return B.CreateGEP(B.getInt8Ty(), SrcStr, EmitStrLen(SrcStr, B, DL, TLI), "strchr");
     return nullptr;
   }
 
@@ -467,7 +323,7 @@ Value *LibCallSimplifier::optimizeStrChr(CallInst *CI, IRBuilder<> &B) {
     return Constant::getNullValue(CI->getType());
 
   // strchr(s+n,c)  -> gep(s+n+i,c)
-  return B.CreateGEP(SrcStr, B.getInt64(I), "strchr");
+  return B.CreateGEP(B.getInt8Ty(), SrcStr, B.getInt64(I), "strchr");
 }
 
 Value *LibCallSimplifier::optimizeStrRChr(CallInst *CI, IRBuilder<> &B) {
@@ -489,8 +345,8 @@ Value *LibCallSimplifier::optimizeStrRChr(CallInst *CI, IRBuilder<> &B) {
   StringRef Str;
   if (!getConstantStringInfo(SrcStr, Str)) {
     // strrchr(s, 0) -> strchr(s, 0)
-    if (DL && CharC->isZero())
-      return EmitStrChr(SrcStr, '\0', B, DL, TLI);
+    if (CharC->isZero())
+      return EmitStrChr(SrcStr, '\0', B, TLI);
     return nullptr;
   }
 
@@ -502,7 +358,7 @@ Value *LibCallSimplifier::optimizeStrRChr(CallInst *CI, IRBuilder<> &B) {
     return Constant::getNullValue(CI->getType());
 
   // strrchr(s+n,c) -> gep(s+n+i,c)
-  return B.CreateGEP(SrcStr, B.getInt64(I), "strrchr");
+  return B.CreateGEP(B.getInt8Ty(), SrcStr, B.getInt64(I), "strrchr");
 }
 
 Value *LibCallSimplifier::optimizeStrCmp(CallInst *CI, IRBuilder<> &B) {
@@ -537,12 +393,8 @@ Value *LibCallSimplifier::optimizeStrCmp(CallInst *CI, IRBuilder<> &B) {
   uint64_t Len1 = GetStringLength(Str1P);
   uint64_t Len2 = GetStringLength(Str2P);
   if (Len1 && Len2) {
-    // These optimizations require DataLayout.
-    if (!DL)
-      return nullptr;
-
     return EmitMemCmp(Str1P, Str2P,
-                      ConstantInt::get(DL->getIntPtrType(CI->getContext()),
+                      ConstantInt::get(DL.getIntPtrType(CI->getContext()),
                                        std::min(Len1, Len2)),
                       B, DL, TLI);
   }
@@ -574,7 +426,7 @@ Value *LibCallSimplifier::optimizeStrNCmp(CallInst *CI, IRBuilder<> &B) {
   if (Length == 0) // strncmp(x,y,0)   -> 0
     return ConstantInt::get(CI->getType(), 0);
 
-  if (DL && Length == 1) // strncmp(x,y,1) -> memcmp(x,y,1)
+  if (Length == 1) // strncmp(x,y,1) -> memcmp(x,y,1)
     return EmitMemCmp(Str1P, Str2P, CI->getArgOperand(2), B, DL, TLI);
 
   StringRef Str1, Str2;
@@ -600,21 +452,14 @@ Value *LibCallSimplifier::optimizeStrNCmp(CallInst *CI, IRBuilder<> &B) {
 
 Value *LibCallSimplifier::optimizeStrCpy(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  // Verify the "strcpy" function prototype.
-  FunctionType *FT = Callee->getFunctionType();
-  if (FT->getNumParams() != 2 || FT->getReturnType() != FT->getParamType(0) ||
-      FT->getParamType(0) != FT->getParamType(1) ||
-      FT->getParamType(0) != B.getInt8PtrTy())
+
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::strcpy))
     return nullptr;
 
   Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
   if (Dst == Src) // strcpy(x,x)  -> x
     return Src;
 
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
-
   // See if we can get the length of the input string.
   uint64_t Len = GetStringLength(Src);
   if (Len == 0)
@@ -623,27 +468,19 @@ Value *LibCallSimplifier::optimizeStrCpy(CallInst *CI, IRBuilder<> &B) {
   // We have enough information to now generate the memcpy call to do the
   // copy for us.  Make a memcpy to copy the nul byte with align = 1.
   B.CreateMemCpy(Dst, Src,
-                 ConstantInt::get(DL->getIntPtrType(CI->getContext()), Len), 1);
+                 ConstantInt::get(DL.getIntPtrType(CI->getContext()), Len), 1);
   return Dst;
 }
 
 Value *LibCallSimplifier::optimizeStpCpy(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  // Verify the "stpcpy" function prototype.
-  FunctionType *FT = Callee->getFunctionType();
-  if (FT->getNumParams() != 2 || FT->getReturnType() != FT->getParamType(0) ||
-      FT->getParamType(0) != FT->getParamType(1) ||
-      FT->getParamType(0) != B.getInt8PtrTy())
-    return nullptr;
-
-  // These optimizations require DataLayout.
-  if (!DL)
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::stpcpy))
     return nullptr;
 
   Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
   if (Dst == Src) { // stpcpy(x,x)  -> x+strlen(x)
     Value *StrLen = EmitStrLen(Src, B, DL, TLI);
-    return StrLen ? B.CreateInBoundsGEP(Dst, StrLen) : nullptr;
+    return StrLen ? B.CreateInBoundsGEP(B.getInt8Ty(), Dst, StrLen) : nullptr;
   }
 
   // See if we can get the length of the input string.
@@ -651,10 +488,10 @@ Value *LibCallSimplifier::optimizeStpCpy(CallInst *CI, IRBuilder<> &B) {
   if (Len == 0)
     return nullptr;
 
-  Type *PT = FT->getParamType(0);
-  Value *LenV = ConstantInt::get(DL->getIntPtrType(PT), Len);
+  Type *PT = Callee->getFunctionType()->getParamType(0);
+  Value *LenV = ConstantInt::get(DL.getIntPtrType(PT), Len);
   Value *DstEnd =
-      B.CreateGEP(Dst, ConstantInt::get(DL->getIntPtrType(PT), Len - 1));
+      B.CreateGEP(B.getInt8Ty(), Dst, ConstantInt::get(DL.getIntPtrType(PT), Len - 1));
 
   // We have enough information to now generate the memcpy call to do the
   // copy for us.  Make a memcpy to copy the nul byte with align = 1.
@@ -664,11 +501,7 @@ Value *LibCallSimplifier::optimizeStpCpy(CallInst *CI, IRBuilder<> &B) {
 
 Value *LibCallSimplifier::optimizeStrNCpy(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
-      FT->getParamType(0) != FT->getParamType(1) ||
-      FT->getParamType(0) != B.getInt8PtrTy() ||
-      !FT->getParamType(2)->isIntegerTy())
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::strncpy))
     return nullptr;
 
   Value *Dst = CI->getArgOperand(0);
@@ -696,17 +529,13 @@ Value *LibCallSimplifier::optimizeStrNCpy(CallInst *CI, IRBuilder<> &B) {
   if (Len == 0)
     return Dst; // strncpy(x, y, 0) -> x
 
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
-
   // Let strncpy handle the zero padding
   if (Len > SrcLen + 1)
     return nullptr;
 
-  Type *PT = FT->getParamType(0);
+  Type *PT = Callee->getFunctionType()->getParamType(0);
   // strncpy(x, s, c) -> memcpy(x, s, c, 1) [s and c are constant]
-  B.CreateMemCpy(Dst, Src, ConstantInt::get(DL->getIntPtrType(PT), Len), 1);
+  B.CreateMemCpy(Dst, Src, ConstantInt::get(DL.getIntPtrType(PT), Len), 1);
 
   return Dst;
 }
@@ -770,12 +599,12 @@ Value *LibCallSimplifier::optimizeStrPBrk(CallInst *CI, IRBuilder<> &B) {
     if (I == StringRef::npos) // No match.
       return Constant::getNullValue(CI->getType());
 
-    return B.CreateGEP(CI->getArgOperand(0), B.getInt64(I), "strpbrk");
+    return B.CreateGEP(B.getInt8Ty(), CI->getArgOperand(0), B.getInt64(I), "strpbrk");
   }
 
   // strpbrk(s, "a") -> strchr(s, 'a')
-  if (DL && HasS2 && S2.size() == 1)
-    return EmitStrChr(CI->getArgOperand(0), S2[0], B, DL, TLI);
+  if (HasS2 && S2.size() == 1)
+    return EmitStrChr(CI->getArgOperand(0), S2[0], B, TLI);
 
   return nullptr;
 }
@@ -851,7 +680,7 @@ Value *LibCallSimplifier::optimizeStrCSpn(CallInst *CI, IRBuilder<> &B) {
   }
 
   // strcspn(s, "") -> strlen(s)
-  if (DL && HasS2 && S2.empty())
+  if (HasS2 && S2.empty())
     return EmitStrLen(CI->getArgOperand(0), B, DL, TLI);
 
   return nullptr;
@@ -870,7 +699,7 @@ Value *LibCallSimplifier::optimizeStrStr(CallInst *CI, IRBuilder<> &B) {
     return B.CreateBitCast(CI->getArgOperand(0), CI->getType());
 
   // fold strstr(a, b) == a -> strncmp(a, b, strlen(b)) == 0
-  if (DL && isOnlyUsedInEqualityComparison(CI, CI->getArgOperand(0))) {
+  if (isOnlyUsedInEqualityComparison(CI, CI->getArgOperand(0))) {
     Value *StrLen = EmitStrLen(CI->getArgOperand(1), B, DL, TLI);
     if (!StrLen)
       return nullptr;
@@ -912,12 +741,98 @@ Value *LibCallSimplifier::optimizeStrStr(CallInst *CI, IRBuilder<> &B) {
 
   // fold strstr(x, "y") -> strchr(x, 'y').
   if (HasStr2 && ToFindStr.size() == 1) {
-    Value *StrChr = EmitStrChr(CI->getArgOperand(0), ToFindStr[0], B, DL, TLI);
+    Value *StrChr = EmitStrChr(CI->getArgOperand(0), ToFindStr[0], B, TLI);
     return StrChr ? B.CreateBitCast(StrChr, CI->getType()) : nullptr;
   }
   return nullptr;
 }
 
+Value *LibCallSimplifier::optimizeMemChr(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+  FunctionType *FT = Callee->getFunctionType();
+  if (FT->getNumParams() != 3 || !FT->getParamType(0)->isPointerTy() ||
+      !FT->getParamType(1)->isIntegerTy(32) ||
+      !FT->getParamType(2)->isIntegerTy() ||
+      !FT->getReturnType()->isPointerTy())
+    return nullptr;
+
+  Value *SrcStr = CI->getArgOperand(0);
+  ConstantInt *CharC = dyn_cast<ConstantInt>(CI->getArgOperand(1));
+  ConstantInt *LenC = dyn_cast<ConstantInt>(CI->getArgOperand(2));
+
+  // memchr(x, y, 0) -> null
+  if (LenC && LenC->isNullValue())
+    return Constant::getNullValue(CI->getType());
+
+  // From now on we need at least constant length and string.
+  StringRef Str;
+  if (!LenC || !getConstantStringInfo(SrcStr, Str, 0, /*TrimAtNul=*/false))
+    return nullptr;
+
+  // Truncate the string to LenC. If Str is smaller than LenC we will still only
+  // scan the string, as reading past the end of it is undefined and we can just
+  // return null if we don't find the char.
+  Str = Str.substr(0, LenC->getZExtValue());
+
+  // If the char is variable but the input str and length are not we can turn
+  // this memchr call into a simple bit field test. Of course this only works
+  // when the return value is only checked against null.
+  //
+  // It would be really nice to reuse switch lowering here but we can't change
+  // the CFG at this point.
+  //
+  // memchr("\r\n", C, 2) != nullptr -> (C & ((1 << '\r') | (1 << '\n'))) != 0
+  //   after bounds check.
+  if (!CharC && !Str.empty() && isOnlyUsedInZeroEqualityComparison(CI)) {
+    unsigned char Max =
+        *std::max_element(reinterpret_cast<const unsigned char *>(Str.begin()),
+                          reinterpret_cast<const unsigned char *>(Str.end()));
+
+    // Make sure the bit field we're about to create fits in a register on the
+    // target.
+    // FIXME: On a 64 bit architecture this prevents us from using the
+    // interesting range of alpha ascii chars. We could do better by emitting
+    // two bitfields or shifting the range by 64 if no lower chars are used.
+    if (!DL.fitsInLegalInteger(Max + 1))
+      return nullptr;
+
+    // For the bit field use a power-of-2 type with at least 8 bits to avoid
+    // creating unnecessary illegal types.
+    unsigned char Width = NextPowerOf2(std::max((unsigned char)7, Max));
+
+    // Now build the bit field.
+    APInt Bitfield(Width, 0);
+    for (char C : Str)
+      Bitfield.setBit((unsigned char)C);
+    Value *BitfieldC = B.getInt(Bitfield);
+
+    // First check that the bit field access is within bounds.
+    Value *C = B.CreateZExtOrTrunc(CI->getArgOperand(1), BitfieldC->getType());
+    Value *Bounds = B.CreateICmp(ICmpInst::ICMP_ULT, C, B.getIntN(Width, Width),
+                                 "memchr.bounds");
+
+    // Create code that checks if the given bit is set in the field.
+    Value *Shl = B.CreateShl(B.getIntN(Width, 1ULL), C);
+    Value *Bits = B.CreateIsNotNull(B.CreateAnd(Shl, BitfieldC), "memchr.bits");
+
+    // Finally merge both checks and cast to pointer type. The inttoptr
+    // implicitly zexts the i1 to intptr type.
+    return B.CreateIntToPtr(B.CreateAnd(Bounds, Bits, "memchr"), CI->getType());
+  }
+
+  // Check if all arguments are constants.  If so, we can constant fold.
+  if (!CharC)
+    return nullptr;
+
+  // Compute the offset.
+  size_t I = Str.find(CharC->getSExtValue() & 0xFF);
+  if (I == StringRef::npos) // Didn't find the char.  memchr returns null.
+    return Constant::getNullValue(CI->getType());
+
+  // memchr(s+n,c,l) -> gep(s+n+i,c)
+  return B.CreateGEP(B.getInt8Ty(), SrcStr, B.getInt64(I), "memchr");
+}
+
 Value *LibCallSimplifier::optimizeMemCmp(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
   FunctionType *FT = Callee->getFunctionType();
@@ -949,6 +864,27 @@ Value *LibCallSimplifier::optimizeMemCmp(CallInst *CI, IRBuilder<> &B) {
     return B.CreateSub(LHSV, RHSV, "chardiff");
   }
 
+  // memcmp(S1,S2,N/8)==0 -> (*(intN_t*)S1 != *(intN_t*)S2)==0
+  if (DL.isLegalInteger(Len * 8) && isOnlyUsedInZeroEqualityComparison(CI)) {
+
+    IntegerType *IntType = IntegerType::get(CI->getContext(), Len * 8);
+    unsigned PrefAlignment = DL.getPrefTypeAlignment(IntType);
+
+    if (getKnownAlignment(LHS, DL, CI) >= PrefAlignment &&
+        getKnownAlignment(RHS, DL, CI) >= PrefAlignment) {
+
+      Type *LHSPtrTy =
+          IntType->getPointerTo(LHS->getType()->getPointerAddressSpace());
+      Type *RHSPtrTy =
+          IntType->getPointerTo(RHS->getType()->getPointerAddressSpace());
+
+      Value *LHSV = B.CreateLoad(B.CreateBitCast(LHS, LHSPtrTy, "lhsc"), "lhsv");
+      Value *RHSV = B.CreateLoad(B.CreateBitCast(RHS, RHSPtrTy, "rhsc"), "rhsv");
+
+      return B.CreateZExt(B.CreateICmpNE(LHSV, RHSV), CI->getType(), "memcmp");
+    }
+  }
+
   // Constant folding: memcmp(x, y, l) -> cnst (all arguments are constant)
   StringRef LHSStr, RHSStr;
   if (getConstantStringInfo(LHS, LHSStr) &&
@@ -972,15 +908,8 @@ Value *LibCallSimplifier::optimizeMemCmp(CallInst *CI, IRBuilder<> &B) {
 
 Value *LibCallSimplifier::optimizeMemCpy(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
 
-  FunctionType *FT = Callee->getFunctionType();
-  if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
-      !FT->getParamType(0)->isPointerTy() ||
-      !FT->getParamType(1)->isPointerTy() ||
-      FT->getParamType(2) != DL->getIntPtrType(CI->getContext()))
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memcpy))
     return nullptr;
 
   // memcpy(x, y, n) -> llvm.memcpy(x, y, n, 1)
@@ -991,15 +920,8 @@ Value *LibCallSimplifier::optimizeMemCpy(CallInst *CI, IRBuilder<> &B) {
 
 Value *LibCallSimplifier::optimizeMemMove(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
 
-  FunctionType *FT = Callee->getFunctionType();
-  if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
-      !FT->getParamType(0)->isPointerTy() ||
-      !FT->getParamType(1)->isPointerTy() ||
-      FT->getParamType(2) != DL->getIntPtrType(CI->getContext()))
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memmove))
     return nullptr;
 
   // memmove(x, y, n) -> llvm.memmove(x, y, n, 1)
@@ -1010,15 +932,8 @@ Value *LibCallSimplifier::optimizeMemMove(CallInst *CI, IRBuilder<> &B) {
 
 Value *LibCallSimplifier::optimizeMemSet(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
 
-  FunctionType *FT = Callee->getFunctionType();
-  if (FT->getNumParams() != 3 || FT->getReturnType() != FT->getParamType(0) ||
-      !FT->getParamType(0)->isPointerTy() ||
-      !FT->getParamType(1)->isIntegerTy() ||
-      FT->getParamType(2) != DL->getIntPtrType(FT->getParamType(0)))
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memset))
     return nullptr;
 
   // memset(p, v, n) -> llvm.memset(p, v, n, 1)
@@ -1044,10 +959,10 @@ static Value *valueHasFloatPrecision(Value *Val) {
   }
   if (ConstantFP *Const = dyn_cast<ConstantFP>(Val)) {
     APFloat F = Const->getValueAPF();
-    bool loosesInfo;
+    bool losesInfo;
     (void)F.convert(APFloat::IEEEsingle, APFloat::rmNearestTiesToEven,
-                    &loosesInfo);
-    if (!loosesInfo)
+                    &losesInfo);
+    if (!losesInfo)
       return ConstantFP::get(Const->getContext(), F);
   }
   return nullptr;
@@ -1081,7 +996,7 @@ Value *LibCallSimplifier::optimizeUnaryDoubleFP(CallInst *CI, IRBuilder<> &B,
   // floor((double)floatval) -> (double)floorf(floatval)
   if (Callee->isIntrinsic()) {
     Module *M = CI->getParent()->getParent()->getParent();
-    Intrinsic::ID IID = (Intrinsic::ID) Callee->getIntrinsicID();
+    Intrinsic::ID IID = Callee->getIntrinsicID();
     Function *F = Intrinsic::getDeclaration(M, IID, B.getFloatTy());
     V = B.CreateCall(F, V);
   } else {
@@ -1123,9 +1038,9 @@ Value *LibCallSimplifier::optimizeBinaryDoubleFP(CallInst *CI, IRBuilder<> &B) {
 Value *LibCallSimplifier::optimizeCos(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
   Value *Ret = nullptr;
-  if (UnsafeFPShrink && Callee->getName() == "cos" && TLI->has(LibFunc::cosf)) {
+  StringRef Name = Callee->getName();
+  if (UnsafeFPShrink && Name == "cos" && hasFloatVersion(Name))
     Ret = optimizeUnaryDoubleFP(CI, B, true);
-  }
 
   FunctionType *FT = Callee->getFunctionType();
   // Just make sure this has 1 argument of FP type, which matches the
@@ -1145,11 +1060,10 @@ Value *LibCallSimplifier::optimizeCos(CallInst *CI, IRBuilder<> &B) {
 
 Value *LibCallSimplifier::optimizePow(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-
   Value *Ret = nullptr;
-  if (UnsafeFPShrink && Callee->getName() == "pow" && TLI->has(LibFunc::powf)) {
+  StringRef Name = Callee->getName();
+  if (UnsafeFPShrink && Name == "pow" && hasFloatVersion(Name))
     Ret = optimizeUnaryDoubleFP(CI, B, true);
-  }
 
   FunctionType *FT = Callee->getFunctionType();
   // Just make sure this has 2 arguments of the same FP type, which match the
@@ -1168,7 +1082,8 @@ Value *LibCallSimplifier::optimizePow(CallInst *CI, IRBuilder<> &B) {
     if (Op1C->isExactlyValue(2.0) &&
         hasUnaryFloatFn(TLI, Op1->getType(), LibFunc::exp2, LibFunc::exp2f,
                         LibFunc::exp2l))
-      return EmitUnaryFloatFnCall(Op2, "exp2", B, Callee->getAttributes());
+      return EmitUnaryFloatFnCall(Op2, TLI->getName(LibFunc::exp2), B,
+                                  Callee->getAttributes());
     // pow(10.0, x) -> exp10(x)
     if (Op1C->isExactlyValue(10.0) &&
         hasUnaryFloatFn(TLI, Op1->getType(), LibFunc::exp10, LibFunc::exp10f,
@@ -1177,6 +1092,32 @@ Value *LibCallSimplifier::optimizePow(CallInst *CI, IRBuilder<> &B) {
                                   Callee->getAttributes());
   }
 
+  bool unsafeFPMath = canUseUnsafeFPMath(CI->getParent()->getParent());
+
+  // pow(exp(x), y) -> exp(x*y)
+  // pow(exp2(x), y) -> exp2(x * y)
+  // We enable these only under fast-math. Besides rounding
+  // differences the transformation changes overflow and
+  // underflow behavior quite dramatically.
+  // Example: x = 1000, y = 0.001.
+  // pow(exp(x), y) = pow(inf, 0.001) = inf, whereas exp(x*y) = exp(1).
+  if (unsafeFPMath) {
+    if (auto *OpC = dyn_cast<CallInst>(Op1)) {
+      IRBuilder<>::FastMathFlagGuard Guard(B);
+      FastMathFlags FMF;
+      FMF.setUnsafeAlgebra();
+      B.SetFastMathFlags(FMF);
+
+      LibFunc::Func Func;
+      Function *OpCCallee = OpC->getCalledFunction();
+      if (OpCCallee && TLI->getLibFunc(OpCCallee->getName(), Func) &&
+          TLI->has(Func) && (Func == LibFunc::exp || Func == LibFunc::exp2))
+        return EmitUnaryFloatFnCall(
+            B.CreateFMul(OpC->getArgOperand(0), Op2, "mul"),
+            OpCCallee->getName(), B, OpCCallee->getAttributes());
+    }
+  }
+
   ConstantFP *Op2C = dyn_cast<ConstantFP>(Op2);
   if (!Op2C)
     return Ret;
@@ -1189,10 +1130,15 @@ Value *LibCallSimplifier::optimizePow(CallInst *CI, IRBuilder<> &B) {
                       LibFunc::sqrtl) &&
       hasUnaryFloatFn(TLI, Op2->getType(), LibFunc::fabs, LibFunc::fabsf,
                       LibFunc::fabsl)) {
+
+    // In -ffast-math, pow(x, 0.5) -> sqrt(x).
+    if (unsafeFPMath)
+      return EmitUnaryFloatFnCall(Op1, TLI->getName(LibFunc::sqrt), B,
+                                  Callee->getAttributes());
+
     // Expand pow(x, 0.5) to (x == -infinity ? +infinity : fabs(sqrt(x))).
     // This is faster than calling pow, and still handles negative zero
     // and negative infinity correctly.
-    // TODO: In fast-math mode, this could be just sqrt(x).
     // TODO: In finite-only mode, this could be just fabs(sqrt(x)).
     Value *Inf = ConstantFP::getInfinity(CI->getType());
     Value *NegInf = ConstantFP::getInfinity(CI->getType(), true);
@@ -1216,12 +1162,10 @@ Value *LibCallSimplifier::optimizePow(CallInst *CI, IRBuilder<> &B) {
 Value *LibCallSimplifier::optimizeExp2(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
   Function *Caller = CI->getParent()->getParent();
-
   Value *Ret = nullptr;
-  if (UnsafeFPShrink && Callee->getName() == "exp2" &&
-      TLI->has(LibFunc::exp2f)) {
+  StringRef Name = Callee->getName();
+  if (UnsafeFPShrink && Name == "exp2" && hasFloatVersion(Name))
     Ret = optimizeUnaryDoubleFP(CI, B, true);
-  }
 
   FunctionType *FT = Callee->getFunctionType();
   // Just make sure this has 1 argument of FP type, which matches the
@@ -1258,7 +1202,7 @@ Value *LibCallSimplifier::optimizeExp2(CallInst *CI, IRBuilder<> &B) {
       Value *Callee =
           M->getOrInsertFunction(TLI->getName(LdExp), Op->getType(),
                                  Op->getType(), B.getInt32Ty(), nullptr);
-      CallInst *CI = B.CreateCall2(Callee, One, LdExpArg);
+      CallInst *CI = B.CreateCall(Callee, {One, LdExpArg});
       if (const Function *F = dyn_cast<Function>(Callee->stripPointerCasts()))
         CI->setCallingConv(F->getCallingConv());
 
@@ -1270,11 +1214,10 @@ Value *LibCallSimplifier::optimizeExp2(CallInst *CI, IRBuilder<> &B) {
 
 Value *LibCallSimplifier::optimizeFabs(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-
   Value *Ret = nullptr;
-  if (Callee->getName() == "fabs" && TLI->has(LibFunc::fabsf)) {
+  StringRef Name = Callee->getName();
+  if (Name == "fabs" && hasFloatVersion(Name))
     Ret = optimizeUnaryDoubleFP(CI, B, false);
-  }
 
   FunctionType *FT = Callee->getFunctionType();
   // Make sure this has 1 argument of FP type which matches the result type.
@@ -1292,29 +1235,110 @@ Value *LibCallSimplifier::optimizeFabs(CallInst *CI, IRBuilder<> &B) {
   return Ret;
 }
 
-Value *LibCallSimplifier::optimizeSqrt(CallInst *CI, IRBuilder<> &B) {
+Value *LibCallSimplifier::optimizeFMinFMax(CallInst *CI, IRBuilder<> &B) {
+  // If we can shrink the call to a float function rather than a double
+  // function, do that first.
   Function *Callee = CI->getCalledFunction();
-  
-  Value *Ret = nullptr;
-  if (TLI->has(LibFunc::sqrtf) && (Callee->getName() == "sqrt" ||
-                                   Callee->getIntrinsicID() == Intrinsic::sqrt))
-    Ret = optimizeUnaryDoubleFP(CI, B, true);
+  StringRef Name = Callee->getName();
+  if ((Name == "fmin" && hasFloatVersion(Name)) ||
+      (Name == "fmax" && hasFloatVersion(Name))) {
+    Value *Ret = optimizeBinaryDoubleFP(CI, B);
+    if (Ret)
+      return Ret;
+  }
 
-  // FIXME: For finer-grain optimization, we need intrinsics to have the same
-  // fast-math flag decorations that are applied to FP instructions. For now,
-  // we have to rely on the function-level unsafe-fp-math attribute to do this
-  // optimization because there's no other way to express that the sqrt can be
-  // reassociated.
+  // Make sure this has 2 arguments of FP type which match the result type.
+  FunctionType *FT = Callee->getFunctionType();
+  if (FT->getNumParams() != 2 || FT->getReturnType() != FT->getParamType(0) ||
+      FT->getParamType(0) != FT->getParamType(1) ||
+      !FT->getParamType(0)->isFloatingPointTy())
+    return nullptr;
+
+  IRBuilder<>::FastMathFlagGuard Guard(B);
+  FastMathFlags FMF;
   Function *F = CI->getParent()->getParent();
-  if (F->hasFnAttribute("unsafe-fp-math")) {
-    // Check for unsafe-fp-math = true.
-    Attribute Attr = F->getFnAttribute("unsafe-fp-math");
+  if (canUseUnsafeFPMath(F)) {
+    // Unsafe algebra sets all fast-math-flags to true.
+    FMF.setUnsafeAlgebra();
+  } else {
+    // At a minimum, no-nans-fp-math must be true.
+    Attribute Attr = F->getFnAttribute("no-nans-fp-math");
     if (Attr.getValueAsString() != "true")
-      return Ret;
-  }
-  Value *Op = CI->getArgOperand(0);
-  if (Instruction *I = dyn_cast<Instruction>(Op)) {
-    if (I->getOpcode() == Instruction::FMul && I->hasUnsafeAlgebra()) {
+      return nullptr;
+    // No-signed-zeros is implied by the definitions of fmax/fmin themselves:
+    // "Ideally, fmax would be sensitive to the sign of zero, for example
+    // fmax(-0. 0, +0. 0) would return +0; however, implementation in software
+    // might be impractical."
+    FMF.setNoSignedZeros();
+    FMF.setNoNaNs();
+  }
+  B.SetFastMathFlags(FMF);
+
+  // We have a relaxed floating-point environment. We can ignore NaN-handling
+  // and transform to a compare and select. We do not have to consider errno or
+  // exceptions, because fmin/fmax do not have those.
+  Value *Op0 = CI->getArgOperand(0);
+  Value *Op1 = CI->getArgOperand(1);
+  Value *Cmp = Callee->getName().startswith("fmin") ?
+    B.CreateFCmpOLT(Op0, Op1) : B.CreateFCmpOGT(Op0, Op1);
+  return B.CreateSelect(Cmp, Op0, Op1);
+}
+
+Value *LibCallSimplifier::optimizeLog(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+  Value *Ret = nullptr;
+  StringRef Name = Callee->getName();
+  if (UnsafeFPShrink && hasFloatVersion(Name))
+    Ret = optimizeUnaryDoubleFP(CI, B, true);
+  FunctionType *FT = Callee->getFunctionType();
+
+  // Just make sure this has 1 argument of FP type, which matches the
+  // result type.
+  if (FT->getNumParams() != 1 || FT->getReturnType() != FT->getParamType(0) ||
+      !FT->getParamType(0)->isFloatingPointTy())
+    return Ret;
+
+  if (!canUseUnsafeFPMath(CI->getParent()->getParent()))
+    return Ret;
+  Value *Op1 = CI->getArgOperand(0);
+  auto *OpC = dyn_cast<CallInst>(Op1);
+  if (!OpC)
+    return Ret;
+
+  // log(pow(x,y)) -> y*log(x)
+  // This is only applicable to log, log2, log10.
+  if (Name != "log" && Name != "log2" && Name != "log10")
+    return Ret;
+
+  IRBuilder<>::FastMathFlagGuard Guard(B);
+  FastMathFlags FMF;
+  FMF.setUnsafeAlgebra();
+  B.SetFastMathFlags(FMF);
+
+  LibFunc::Func Func;
+  Function *F = OpC->getCalledFunction();
+  StringRef FuncName = F->getName();
+  if ((TLI->getLibFunc(FuncName, Func) && TLI->has(Func) &&
+      Func == LibFunc::pow) || F->getIntrinsicID() == Intrinsic::pow)
+    return B.CreateFMul(OpC->getArgOperand(1),
+      EmitUnaryFloatFnCall(OpC->getOperand(0), Callee->getName(), B,
+                           Callee->getAttributes()), "mul");
+  return Ret;
+}
+
+Value *LibCallSimplifier::optimizeSqrt(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+  
+  Value *Ret = nullptr;
+  if (TLI->has(LibFunc::sqrtf) && (Callee->getName() == "sqrt" ||
+                                   Callee->getIntrinsicID() == Intrinsic::sqrt))
+    Ret = optimizeUnaryDoubleFP(CI, B, true);
+  if (!canUseUnsafeFPMath(CI->getParent()->getParent()))
+    return Ret;
+
+  Value *Op = CI->getArgOperand(0);
+  if (Instruction *I = dyn_cast<Instruction>(Op)) {
+    if (I->getOpcode() == Instruction::FMul && I->hasUnsafeAlgebra()) {
       // We're looking for a repeated factor in a multiplication tree,
       // so we can do this fold: sqrt(x * x) -> fabs(x);
       // or this fold: sqrt(x * x * y) -> fabs(x) * sqrt(y).
@@ -1346,8 +1370,7 @@ Value *LibCallSimplifier::optimizeSqrt(CallInst *CI, IRBuilder<> &B) {
         // and multiply.
         // FIXME: We're not checking the sqrt because it doesn't have
         // fast-math-flags (see earlier comment).
-        IRBuilder<true, ConstantFolder,
-          IRBuilderDefaultInserter<true> >::FastMathFlagGuard Guard(B);
+        IRBuilder<>::FastMathFlagGuard Guard(B);
         B.SetFastMathFlags(I->getFastMathFlags());
         // If we found a repeated factor, hoist it out of the square root and
         // replace it with the fabs of that factor.
@@ -1370,6 +1393,40 @@ Value *LibCallSimplifier::optimizeSqrt(CallInst *CI, IRBuilder<> &B) {
   return Ret;
 }
 
+Value *LibCallSimplifier::optimizeTan(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+  Value *Ret = nullptr;
+  StringRef Name = Callee->getName();
+  if (UnsafeFPShrink && Name == "tan" && hasFloatVersion(Name))
+    Ret = optimizeUnaryDoubleFP(CI, B, true);
+  FunctionType *FT = Callee->getFunctionType();
+
+  // Just make sure this has 1 argument of FP type, which matches the
+  // result type.
+  if (FT->getNumParams() != 1 || FT->getReturnType() != FT->getParamType(0) ||
+      !FT->getParamType(0)->isFloatingPointTy())
+    return Ret;
+
+  if (!canUseUnsafeFPMath(CI->getParent()->getParent()))
+    return Ret;
+  Value *Op1 = CI->getArgOperand(0);
+  auto *OpC = dyn_cast<CallInst>(Op1);
+  if (!OpC)
+    return Ret;
+
+  // tan(atan(x)) -> x
+  // tanf(atanf(x)) -> x
+  // tanl(atanl(x)) -> x
+  LibFunc::Func Func;
+  Function *F = OpC->getCalledFunction();
+  if (F && TLI->getLibFunc(F->getName(), Func) && TLI->has(Func) &&
+      ((Func == LibFunc::atan && Callee->getName() == "tan") ||
+       (Func == LibFunc::atanf && Callee->getName() == "tanf") ||
+       (Func == LibFunc::atanl && Callee->getName() == "tanl")))
+    Ret = OpC->getArgOperand(0);
+  return Ret;
+}
+
 static bool isTrigLibCall(CallInst *CI);
 static void insertSinCosCall(IRBuilder<> &B, Function *OrigCallee, Value *Arg,
                              bool UseFloat, Value *&Sin, Value *&Cos,
@@ -1437,9 +1494,9 @@ LibCallSimplifier::classifyArgUse(Value *Val, BasicBlock *BB, bool IsFloat,
     return;
 
   Function *Callee = CI->getCalledFunction();
-  StringRef FuncName = Callee->getName();
   LibFunc::Func Func;
-  if (!TLI->getLibFunc(FuncName, Func) || !TLI->has(Func) || !isTrigLibCall(CI))
+  if (!Callee || !TLI->getLibFunc(Callee->getName(), Func) || !TLI->has(Func) ||
+      !isTrigLibCall(CI))
     return;
 
   if (IsFloat) {
@@ -1461,10 +1518,8 @@ LibCallSimplifier::classifyArgUse(Value *Val, BasicBlock *BB, bool IsFloat,
 
 void LibCallSimplifier::replaceTrigInsts(SmallVectorImpl<CallInst *> &Calls,
                                          Value *Res) {
-  for (SmallVectorImpl<CallInst *>::iterator I = Calls.begin(), E = Calls.end();
-       I != E; ++I) {
-    replaceAllUsesWith(*I, Res);
-  }
+  for (CallInst *C : Calls)
+    replaceAllUsesWith(C, Res);
 }
 
 void insertSinCosCall(IRBuilder<> &B, Function *OrigCallee, Value *Arg,
@@ -1495,8 +1550,7 @@ void insertSinCosCall(IRBuilder<> &B, Function *OrigCallee, Value *Arg,
   if (Instruction *ArgInst = dyn_cast<Instruction>(Arg)) {
     // If the argument is an instruction, it must dominate all uses so put our
     // sincos call there.
-    BasicBlock::iterator Loc = ArgInst;
-    B.SetInsertPoint(ArgInst->getParent(), ++Loc);
+    B.SetInsertPoint(ArgInst->getParent(), ++ArgInst->getIterator());
   } else {
     // Otherwise (e.g. for a constant) the beginning of the function is as
     // good a place as any.
@@ -1521,15 +1575,16 @@ void insertSinCosCall(IRBuilder<> &B, Function *OrigCallee, Value *Arg,
 // Integer Library Call Optimizations
 //===----------------------------------------------------------------------===//
 
+static bool checkIntUnaryReturnAndParam(Function *Callee) {
+  FunctionType *FT = Callee->getFunctionType();
+  return FT->getNumParams() == 1 && FT->getReturnType()->isIntegerTy(32) &&
+    FT->getParamType(0)->isIntegerTy();
+}
+
 Value *LibCallSimplifier::optimizeFFS(CallInst *CI, IRBuilder<> &B) {
   Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  // Just make sure this has 2 arguments of the same FP type, which match the
-  // result type.
-  if (FT->getNumParams() != 1 || !FT->getReturnType()->isIntegerTy(32) ||
-      !FT->getParamType(0)->isIntegerTy())
+  if (!checkIntUnaryReturnAndParam(Callee))
     return nullptr;
-
   Value *Op = CI->getArgOperand(0);
 
   // Constant fold.
@@ -1544,7 +1599,7 @@ Value *LibCallSimplifier::optimizeFFS(CallInst *CI, IRBuilder<> &B) {
   Type *ArgType = Op->getType();
   Value *F =
       Intrinsic::getDeclaration(Callee->getParent(), Intrinsic::cttz, ArgType);
-  Value *V = B.CreateCall2(F, Op, B.getFalse(), "cttz");
+  Value *V = B.CreateCall(F, {Op, B.getTrue()}, "cttz");
   V = B.CreateAdd(V, ConstantInt::get(V->getType(), 1));
   V = B.CreateIntCast(V, B.getInt32Ty(), false);
 
@@ -1569,11 +1624,7 @@ Value *LibCallSimplifier::optimizeAbs(CallInst *CI, IRBuilder<> &B) {
 }
 
 Value *LibCallSimplifier::optimizeIsDigit(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  // We require integer(i32)
-  if (FT->getNumParams() != 1 || !FT->getReturnType()->isIntegerTy() ||
-      !FT->getParamType(0)->isIntegerTy(32))
+  if (!checkIntUnaryReturnAndParam(CI->getCalledFunction()))
     return nullptr;
 
   // isdigit(c) -> (c-'0') <u 10
@@ -1584,11 +1635,7 @@ Value *LibCallSimplifier::optimizeIsDigit(CallInst *CI, IRBuilder<> &B) {
 }
 
 Value *LibCallSimplifier::optimizeIsAscii(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  // We require integer(i32)
-  if (FT->getNumParams() != 1 || !FT->getReturnType()->isIntegerTy() ||
-      !FT->getParamType(0)->isIntegerTy(32))
+  if (!checkIntUnaryReturnAndParam(CI->getCalledFunction()))
     return nullptr;
 
   // isascii(c) -> c <u 128
@@ -1598,11 +1645,7 @@ Value *LibCallSimplifier::optimizeIsAscii(CallInst *CI, IRBuilder<> &B) {
 }
 
 Value *LibCallSimplifier::optimizeToAscii(CallInst *CI, IRBuilder<> &B) {
-  Function *Callee = CI->getCalledFunction();
-  FunctionType *FT = Callee->getFunctionType();
-  // We require i32(i32)
-  if (FT->getNumParams() != 1 || FT->getReturnType() != FT->getParamType(0) ||
-      !FT->getParamType(0)->isIntegerTy(32))
+  if (!checkIntUnaryReturnAndParam(CI->getCalledFunction()))
     return nullptr;
 
   // toascii(c) -> c & 0x7f
@@ -1637,10 +1680,7 @@ Value *LibCallSimplifier::optimizeErrorReporting(CallInst *CI, IRBuilder<> &B,
 }
 
 static bool isReportingError(Function *Callee, CallInst *CI, int StreamArg) {
-  if (!ColdErrorCalls)
-    return false;
-
-  if (!Callee || !Callee->isDeclaration())
+  if (!ColdErrorCalls || !Callee || !Callee->isDeclaration())
     return false;
 
   if (StreamArg < 0)
@@ -1678,7 +1718,7 @@ Value *LibCallSimplifier::optimizePrintFString(CallInst *CI, IRBuilder<> &B) {
 
   // printf("x") -> putchar('x'), even for '%'.
   if (FormatStr.size() == 1) {
-    Value *Res = EmitPutChar(B.getInt32(FormatStr[0]), B, DL, TLI);
+    Value *Res = EmitPutChar(B.getInt32(FormatStr[0]), B, TLI);
     if (CI->use_empty() || !Res)
       return Res;
     return B.CreateIntCast(Res, CI->getType(), true);
@@ -1691,7 +1731,7 @@ Value *LibCallSimplifier::optimizePrintFString(CallInst *CI, IRBuilder<> &B) {
     // pass to be run after this pass, to merge duplicate strings.
     FormatStr = FormatStr.drop_back();
     Value *GV = B.CreateGlobalString(FormatStr, "str");
-    Value *NewCI = EmitPutS(GV, B, DL, TLI);
+    Value *NewCI = EmitPutS(GV, B, TLI);
     return (CI->use_empty() || !NewCI)
                ? NewCI
                : ConstantInt::get(CI->getType(), FormatStr.size() + 1);
@@ -1701,7 +1741,7 @@ Value *LibCallSimplifier::optimizePrintFString(CallInst *CI, IRBuilder<> &B) {
   // printf("%c", chr) --> putchar(chr)
   if (FormatStr == "%c" && CI->getNumArgOperands() > 1 &&
       CI->getArgOperand(1)->getType()->isIntegerTy()) {
-    Value *Res = EmitPutChar(CI->getArgOperand(1), B, DL, TLI);
+    Value *Res = EmitPutChar(CI->getArgOperand(1), B, TLI);
 
     if (CI->use_empty() || !Res)
       return Res;
@@ -1711,7 +1751,7 @@ Value *LibCallSimplifier::optimizePrintFString(CallInst *CI, IRBuilder<> &B) {
   // printf("%s\n", str) --> puts(str)
   if (FormatStr == "%s\n" && CI->getNumArgOperands() > 1 &&
       CI->getArgOperand(1)->getType()->isPointerTy()) {
-    return EmitPutS(CI->getArgOperand(1), B, DL, TLI);
+    return EmitPutS(CI->getArgOperand(1), B, TLI);
   }
   return nullptr;
 }
@@ -1757,16 +1797,11 @@ Value *LibCallSimplifier::optimizeSPrintFString(CallInst *CI, IRBuilder<> &B) {
       if (FormatStr[i] == '%')
         return nullptr; // we found a format specifier, bail out.
 
-    // These optimizations require DataLayout.
-    if (!DL)
-      return nullptr;
-
     // sprintf(str, fmt) -> llvm.memcpy(str, fmt, strlen(fmt)+1, 1)
-    B.CreateMemCpy(
-        CI->getArgOperand(0), CI->getArgOperand(1),
-        ConstantInt::get(DL->getIntPtrType(CI->getContext()),
-                         FormatStr.size() + 1),
-        1); // Copy the null byte.
+    B.CreateMemCpy(CI->getArgOperand(0), CI->getArgOperand(1),
+                   ConstantInt::get(DL.getIntPtrType(CI->getContext()),
+                                    FormatStr.size() + 1),
+                   1); // Copy the null byte.
     return ConstantInt::get(CI->getType(), FormatStr.size());
   }
 
@@ -1784,17 +1819,13 @@ Value *LibCallSimplifier::optimizeSPrintFString(CallInst *CI, IRBuilder<> &B) {
     Value *V = B.CreateTrunc(CI->getArgOperand(2), B.getInt8Ty(), "char");
     Value *Ptr = CastToCStr(CI->getArgOperand(0), B);
     B.CreateStore(V, Ptr);
-    Ptr = B.CreateGEP(Ptr, B.getInt32(1), "nul");
+    Ptr = B.CreateGEP(B.getInt8Ty(), Ptr, B.getInt32(1), "nul");
     B.CreateStore(B.getInt8(0), Ptr);
 
     return ConstantInt::get(CI->getType(), 1);
   }
 
   if (FormatStr[1] == 's') {
-    // These optimizations require DataLayout.
-    if (!DL)
-      return nullptr;
-
     // sprintf(dest, "%s", str) -> llvm.memcpy(dest, str, strlen(str)+1, 1)
     if (!CI->getArgOperand(2)->getType()->isPointerTy())
       return nullptr;
@@ -1859,13 +1890,9 @@ Value *LibCallSimplifier::optimizeFPrintFString(CallInst *CI, IRBuilder<> &B) {
       if (FormatStr[i] == '%') // Could handle %% -> % if we cared.
         return nullptr;        // We found a format specifier.
 
-    // These optimizations require DataLayout.
-    if (!DL)
-      return nullptr;
-
     return EmitFWrite(
         CI->getArgOperand(1),
-        ConstantInt::get(DL->getIntPtrType(CI->getContext()), FormatStr.size()),
+        ConstantInt::get(DL.getIntPtrType(CI->getContext()), FormatStr.size()),
         CI->getArgOperand(0), B, DL, TLI);
   }
 
@@ -1880,14 +1907,14 @@ Value *LibCallSimplifier::optimizeFPrintFString(CallInst *CI, IRBuilder<> &B) {
     // fprintf(F, "%c", chr) --> fputc(chr, F)
     if (!CI->getArgOperand(2)->getType()->isIntegerTy())
       return nullptr;
-    return EmitFPutC(CI->getArgOperand(2), CI->getArgOperand(0), B, DL, TLI);
+    return EmitFPutC(CI->getArgOperand(2), CI->getArgOperand(0), B, TLI);
   }
 
   if (FormatStr[1] == 's') {
     // fprintf(F, "%s", str) --> fputs(str, F)
     if (!CI->getArgOperand(2)->getType()->isPointerTy())
       return nullptr;
-    return EmitFPutS(CI->getArgOperand(2), CI->getArgOperand(0), B, DL, TLI);
+    return EmitFPutS(CI->getArgOperand(2), CI->getArgOperand(0), B, TLI);
   }
   return nullptr;
 }
@@ -1947,7 +1974,7 @@ Value *LibCallSimplifier::optimizeFWrite(CallInst *CI, IRBuilder<> &B) {
   // This optimisation is only valid, if the return value is unused.
   if (Bytes == 1 && CI->use_empty()) { // fwrite(S,1,1,F) -> fputc(S[0],F)
     Value *Char = B.CreateLoad(CastToCStr(CI->getArgOperand(0), B), "char");
-    Value *NewCI = EmitFPutC(Char, CI->getArgOperand(3), B, DL, TLI);
+    Value *NewCI = EmitFPutC(Char, CI->getArgOperand(3), B, TLI);
     return NewCI ? ConstantInt::get(CI->getType(), 1) : nullptr;
   }
 
@@ -1959,10 +1986,6 @@ Value *LibCallSimplifier::optimizeFPuts(CallInst *CI, IRBuilder<> &B) {
 
   Function *Callee = CI->getCalledFunction();
 
-  // These optimizations require DataLayout.
-  if (!DL)
-    return nullptr;
-
   // Require two pointers.  Also, we can't optimize if return value is used.
   FunctionType *FT = Callee->getFunctionType();
   if (FT->getNumParams() != 2 || !FT->getParamType(0)->isPointerTy() ||
@@ -1977,7 +2000,7 @@ Value *LibCallSimplifier::optimizeFPuts(CallInst *CI, IRBuilder<> &B) {
   // Known to have no uses (see above).
   return EmitFWrite(
       CI->getArgOperand(0),
-      ConstantInt::get(DL->getIntPtrType(CI->getContext()), Len - 1),
+      ConstantInt::get(DL.getIntPtrType(CI->getContext()), Len - 1),
       CI->getArgOperand(1), B, DL, TLI);
 }
 
@@ -1996,7 +2019,7 @@ Value *LibCallSimplifier::optimizePuts(CallInst *CI, IRBuilder<> &B) {
 
   if (Str.empty() && CI->use_empty()) {
     // puts("") -> putchar('\n')
-    Value *Res = EmitPutChar(B.getInt32('\n'), B, DL, TLI);
+    Value *Res = EmitPutChar(B.getInt32('\n'), B, TLI);
     if (CI->use_empty() || !Res)
       return Res;
     return B.CreateIntCast(Res, CI->getType(), true);
@@ -2014,53 +2037,18 @@ bool LibCallSimplifier::hasFloatVersion(StringRef FuncName) {
   return false;
 }
 
-Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
-  if (CI->isNoBuiltin())
-    return nullptr;
-
+Value *LibCallSimplifier::optimizeStringMemoryLibCall(CallInst *CI,
+                                                      IRBuilder<> &Builder) {
   LibFunc::Func Func;
   Function *Callee = CI->getCalledFunction();
   StringRef FuncName = Callee->getName();
-  IRBuilder<> Builder(CI);
-  bool isCallingConvC = CI->getCallingConv() == llvm::CallingConv::C;
-
-  // Command-line parameter overrides function attribute.
-  if (EnableUnsafeFPShrink.getNumOccurrences() > 0)
-    UnsafeFPShrink = EnableUnsafeFPShrink;
-  else if (Callee->hasFnAttribute("unsafe-fp-math")) {
-    // FIXME: This is the same problem as described in optimizeSqrt().
-    // If calls gain access to IR-level FMF, then use that instead of a
-    // function attribute.
-
-    // Check for unsafe-fp-math = true.
-    Attribute Attr = Callee->getFnAttribute("unsafe-fp-math");
-    if (Attr.getValueAsString() == "true")
-      UnsafeFPShrink = true;
-  }
-
-  // First, check for intrinsics.
-  if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(CI)) {
-    if (!isCallingConvC)
-      return nullptr;
-    switch (II->getIntrinsicID()) {
-    case Intrinsic::pow:
-      return optimizePow(CI, Builder);
-    case Intrinsic::exp2:
-      return optimizeExp2(CI, Builder);
-    case Intrinsic::fabs:
-      return optimizeFabs(CI, Builder);
-    case Intrinsic::sqrt:
-      return optimizeSqrt(CI, Builder);
-    default:
-      return nullptr;
-    }
-  }
 
-  // Then check for known library functions.
+  // Check for string/memory library functions.
   if (TLI->getLibFunc(FuncName, Func) && TLI->has(Func)) {
-    // We never change the calling convention.
-    if (!ignoreCallingConv(Func) && !isCallingConvC)
-      return nullptr;
+    // Make sure we never change the calling convention.
+    assert((ignoreCallingConv(Func) ||
+            CI->getCallingConv() == llvm::CallingConv::C) &&
+      "Optimizing string/memory libcall would change the calling convention");
     switch (Func) {
     case LibFunc::strcat:
       return optimizeStrCat(CI, Builder);
@@ -2098,6 +2086,8 @@ Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
       return optimizeStrCSpn(CI, Builder);
     case LibFunc::strstr:
       return optimizeStrStr(CI, Builder);
+    case LibFunc::memchr:
+      return optimizeMemChr(CI, Builder);
     case LibFunc::memcmp:
       return optimizeMemCmp(CI, Builder);
     case LibFunc::memcpy:
@@ -2106,6 +2096,75 @@ Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
       return optimizeMemMove(CI, Builder);
     case LibFunc::memset:
       return optimizeMemSet(CI, Builder);
+    default:
+      break;
+    }
+  }
+  return nullptr;
+}
+
+Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
+  if (CI->isNoBuiltin())
+    return nullptr;
+
+  LibFunc::Func Func;
+  Function *Callee = CI->getCalledFunction();
+  StringRef FuncName = Callee->getName();
+  IRBuilder<> Builder(CI);
+  bool isCallingConvC = CI->getCallingConv() == llvm::CallingConv::C;
+
+  // Command-line parameter overrides function attribute.
+  if (EnableUnsafeFPShrink.getNumOccurrences() > 0)
+    UnsafeFPShrink = EnableUnsafeFPShrink;
+  else if (canUseUnsafeFPMath(Callee))
+    UnsafeFPShrink = true;
+
+  // First, check for intrinsics.
+  if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(CI)) {
+    if (!isCallingConvC)
+      return nullptr;
+    switch (II->getIntrinsicID()) {
+    case Intrinsic::pow:
+      return optimizePow(CI, Builder);
+    case Intrinsic::exp2:
+      return optimizeExp2(CI, Builder);
+    case Intrinsic::fabs:
+      return optimizeFabs(CI, Builder);
+    case Intrinsic::log:
+      return optimizeLog(CI, Builder);
+    case Intrinsic::sqrt:
+      return optimizeSqrt(CI, Builder);
+    default:
+      return nullptr;
+    }
+  }
+
+  // Also try to simplify calls to fortified library functions.
+  if (Value *SimplifiedFortifiedCI = FortifiedSimplifier.optimizeCall(CI)) {
+    // Try to further simplify the result.
+    CallInst *SimplifiedCI = dyn_cast<CallInst>(SimplifiedFortifiedCI);
+    if (SimplifiedCI && SimplifiedCI->getCalledFunction()) {
+      // Use an IR Builder from SimplifiedCI if available instead of CI
+      // to guarantee we reach all uses we might replace later on.
+      IRBuilder<> TmpBuilder(SimplifiedCI);
+      if (Value *V = optimizeStringMemoryLibCall(SimplifiedCI, TmpBuilder)) {
+        // If we were able to further simplify, remove the now redundant call.
+        SimplifiedCI->replaceAllUsesWith(V);
+        SimplifiedCI->eraseFromParent();
+        return V;
+      }
+    }
+    return SimplifiedFortifiedCI;
+  }
+
+  // Then check for known library functions.
+  if (TLI->getLibFunc(FuncName, Func) && TLI->has(Func)) {
+    // We never change the calling convention.
+    if (!ignoreCallingConv(Func) && !isCallingConvC)
+      return nullptr;
+    if (Value *V = optimizeStringMemoryLibCall(CI, Builder))
+      return V;
+    switch (Func) {
     case LibFunc::cosf:
     case LibFunc::cos:
     case LibFunc::cosl:
@@ -2155,8 +2214,18 @@ Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
       return optimizeFWrite(CI, Builder);
     case LibFunc::fputs:
       return optimizeFPuts(CI, Builder);
+    case LibFunc::log:
+    case LibFunc::log10:
+    case LibFunc::log1p:
+    case LibFunc::log2:
+    case LibFunc::logb:
+      return optimizeLog(CI, Builder);
     case LibFunc::puts:
       return optimizePuts(CI, Builder);
+    case LibFunc::tan:
+    case LibFunc::tanf:
+    case LibFunc::tanl:
+      return optimizeTan(CI, Builder);
     case LibFunc::perror:
       return optimizeErrorReporting(CI, Builder);
     case LibFunc::vfprintf:
@@ -2184,55 +2253,39 @@ Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
     case LibFunc::exp:
     case LibFunc::exp10:
     case LibFunc::expm1:
-    case LibFunc::log:
-    case LibFunc::log10:
-    case LibFunc::log1p:
-    case LibFunc::log2:
-    case LibFunc::logb:
     case LibFunc::sin:
     case LibFunc::sinh:
-    case LibFunc::tan:
     case LibFunc::tanh:
       if (UnsafeFPShrink && hasFloatVersion(FuncName))
         return optimizeUnaryDoubleFP(CI, Builder, true);
       return nullptr;
     case LibFunc::copysign:
-    case LibFunc::fmin:
-    case LibFunc::fmax:
       if (hasFloatVersion(FuncName))
         return optimizeBinaryDoubleFP(CI, Builder);
       return nullptr;
-    case LibFunc::memcpy_chk:
-      return optimizeMemCpyChk(CI, Builder);
-    case LibFunc::memmove_chk:
-      return optimizeMemMoveChk(CI, Builder);
-    case LibFunc::memset_chk:
-      return optimizeMemSetChk(CI, Builder);
-    case LibFunc::strcpy_chk:
-      return optimizeStrCpyChk(CI, Builder);
-    case LibFunc::stpcpy_chk:
-      return optimizeStpCpyChk(CI, Builder);
-    case LibFunc::stpncpy_chk:
-    case LibFunc::strncpy_chk:
-      return optimizeStrNCpyChk(CI, Builder);
+    case LibFunc::fminf:
+    case LibFunc::fmin:
+    case LibFunc::fminl:
+    case LibFunc::fmaxf:
+    case LibFunc::fmax:
+    case LibFunc::fmaxl:
+      return optimizeFMinFMax(CI, Builder);
     default:
       return nullptr;
     }
   }
-
   return nullptr;
 }
 
-LibCallSimplifier::LibCallSimplifier(const DataLayout *DL,
-                                     const TargetLibraryInfo *TLI) :
-                                     DL(DL),
-                                     TLI(TLI),
-                                     UnsafeFPShrink(false) {
-}
+LibCallSimplifier::LibCallSimplifier(
+    const DataLayout &DL, const TargetLibraryInfo *TLI,
+    function_ref<void(Instruction *, Value *)> Replacer)
+    : FortifiedSimplifier(TLI), DL(DL), TLI(TLI), UnsafeFPShrink(false),
+      Replacer(Replacer) {}
 
-void LibCallSimplifier::replaceAllUsesWith(Instruction *I, Value *With) const {
-  I->replaceAllUsesWith(With);
-  I->eraseFromParent();
+void LibCallSimplifier::replaceAllUsesWith(Instruction *I, Value *With) {
+  // Indirect through the replacer used in this instance.
+  Replacer(I, With);
 }
 
 // TODO:
@@ -2241,25 +2294,22 @@ void LibCallSimplifier::replaceAllUsesWith(Instruction *I, Value *With) const {
 // cbrt:
 //   * cbrt(expN(X))  -> expN(x/3)
 //   * cbrt(sqrt(x))  -> pow(x,1/6)
-//   * cbrt(sqrt(x))  -> pow(x,1/9)
+//   * cbrt(cbrt(x))  -> pow(x,1/9)
 //
 // exp, expf, expl:
 //   * exp(log(x))  -> x
 //
 // log, logf, logl:
 //   * log(exp(x))   -> x
-//   * log(x**y)     -> y*log(x)
 //   * log(exp(y))   -> y*log(e)
 //   * log(exp2(y))  -> y*log(2)
 //   * log(exp10(y)) -> y*log(10)
 //   * log(sqrt(x))  -> 0.5*log(x)
-//   * log(pow(x,y)) -> y*log(x)
 //
 // lround, lroundf, lroundl:
 //   * lround(cnst) -> cnst'
 //
 // pow, powf, powl:
-//   * pow(exp(x),y)  -> exp(x*y)
 //   * pow(sqrt(x),y) -> pow(x,y*0.5)
 //   * pow(pow(x,y),z)-> pow(x,y*z)
 //
@@ -2275,10 +2325,192 @@ void LibCallSimplifier::replaceAllUsesWith(Instruction *I, Value *With) const {
 //   * sqrt(Nroot(x)) -> pow(x,1/(2*N))
 //   * sqrt(pow(x,y)) -> pow(|x|,y*0.5)
 //
-// tan, tanf, tanl:
-//   * tan(atan(x)) -> x
-//
 // trunc, truncf, truncl:
 //   * trunc(cnst) -> cnst'
 //
 //
+
+//===----------------------------------------------------------------------===//
+// Fortified Library Call Optimizations
+//===----------------------------------------------------------------------===//
+
+bool FortifiedLibCallSimplifier::isFortifiedCallFoldable(CallInst *CI,
+                                                         unsigned ObjSizeOp,
+                                                         unsigned SizeOp,
+                                                         bool isString) {
+  if (CI->getArgOperand(ObjSizeOp) == CI->getArgOperand(SizeOp))
+    return true;
+  if (ConstantInt *ObjSizeCI =
+          dyn_cast<ConstantInt>(CI->getArgOperand(ObjSizeOp))) {
+    if (ObjSizeCI->isAllOnesValue())
+      return true;
+    // If the object size wasn't -1 (unknown), bail out if we were asked to.
+    if (OnlyLowerUnknownSize)
+      return false;
+    if (isString) {
+      uint64_t Len = GetStringLength(CI->getArgOperand(SizeOp));
+      // If the length is 0 we don't know how long it is and so we can't
+      // remove the check.
+      if (Len == 0)
+        return false;
+      return ObjSizeCI->getZExtValue() >= Len;
+    }
+    if (ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getArgOperand(SizeOp)))
+      return ObjSizeCI->getZExtValue() >= SizeCI->getZExtValue();
+  }
+  return false;
+}
+
+Value *FortifiedLibCallSimplifier::optimizeMemCpyChk(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memcpy_chk))
+    return nullptr;
+
+  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
+    B.CreateMemCpy(CI->getArgOperand(0), CI->getArgOperand(1),
+                   CI->getArgOperand(2), 1);
+    return CI->getArgOperand(0);
+  }
+  return nullptr;
+}
+
+Value *FortifiedLibCallSimplifier::optimizeMemMoveChk(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memmove_chk))
+    return nullptr;
+
+  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
+    B.CreateMemMove(CI->getArgOperand(0), CI->getArgOperand(1),
+                    CI->getArgOperand(2), 1);
+    return CI->getArgOperand(0);
+  }
+  return nullptr;
+}
+
+Value *FortifiedLibCallSimplifier::optimizeMemSetChk(CallInst *CI, IRBuilder<> &B) {
+  Function *Callee = CI->getCalledFunction();
+
+  if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memset_chk))
+    return nullptr;
+
+  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
+    Value *Val = B.CreateIntCast(CI->getArgOperand(1), B.getInt8Ty(), false);
+    B.CreateMemSet(CI->getArgOperand(0), Val, CI->getArgOperand(2), 1);
+    return CI->getArgOperand(0);
+  }
+  return nullptr;
+}
+
+Value *FortifiedLibCallSimplifier::optimizeStrpCpyChk(CallInst *CI,
+                                                      IRBuilder<> &B,
+                                                      LibFunc::Func Func) {
+  Function *Callee = CI->getCalledFunction();
+  StringRef Name = Callee->getName();
+  const DataLayout &DL = CI->getModule()->getDataLayout();
+
+  if (!checkStringCopyLibFuncSignature(Callee, Func))
+    return nullptr;
+
+  Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1),
+        *ObjSize = CI->getArgOperand(2);
+
+  // __stpcpy_chk(x,x,...)  -> x+strlen(x)
+  if (Func == LibFunc::stpcpy_chk && !OnlyLowerUnknownSize && Dst == Src) {
+    Value *StrLen = EmitStrLen(Src, B, DL, TLI);
+    return StrLen ? B.CreateInBoundsGEP(B.getInt8Ty(), Dst, StrLen) : nullptr;
+  }
+
+  // If a) we don't have any length information, or b) we know this will
+  // fit then just lower to a plain st[rp]cpy. Otherwise we'll keep our
+  // st[rp]cpy_chk call which may fail at runtime if the size is too long.
+  // TODO: It might be nice to get a maximum length out of the possible
+  // string lengths for varying.
+  if (isFortifiedCallFoldable(CI, 2, 1, true))
+    return EmitStrCpy(Dst, Src, B, TLI, Name.substr(2, 6));
+
+  if (OnlyLowerUnknownSize)
+    return nullptr;
+
+  // Maybe we can stil fold __st[rp]cpy_chk to __memcpy_chk.
+  uint64_t Len = GetStringLength(Src);
+  if (Len == 0)
+    return nullptr;
+
+  Type *SizeTTy = DL.getIntPtrType(CI->getContext());
+  Value *LenV = ConstantInt::get(SizeTTy, Len);
+  Value *Ret = EmitMemCpyChk(Dst, Src, LenV, ObjSize, B, DL, TLI);
+  // If the function was an __stpcpy_chk, and we were able to fold it into
+  // a __memcpy_chk, we still need to return the correct end pointer.
+  if (Ret && Func == LibFunc::stpcpy_chk)
+    return B.CreateGEP(B.getInt8Ty(), Dst, ConstantInt::get(SizeTTy, Len - 1));
+  return Ret;
+}
+
+Value *FortifiedLibCallSimplifier::optimizeStrpNCpyChk(CallInst *CI,
+                                                       IRBuilder<> &B,
+                                                       LibFunc::Func Func) {
+  Function *Callee = CI->getCalledFunction();
+  StringRef Name = Callee->getName();
+
+  if (!checkStringCopyLibFuncSignature(Callee, Func))
+    return nullptr;
+  if (isFortifiedCallFoldable(CI, 3, 2, false)) {
+    Value *Ret = EmitStrNCpy(CI->getArgOperand(0), CI->getArgOperand(1),
+                             CI->getArgOperand(2), B, TLI, Name.substr(2, 7));
+    return Ret;
+  }
+  return nullptr;
+}
+
+Value *FortifiedLibCallSimplifier::optimizeCall(CallInst *CI) {
+  // FIXME: We shouldn't be changing "nobuiltin" or TLI unavailable calls here.
+  // Some clang users checked for _chk libcall availability using:
+  //   __has_builtin(__builtin___memcpy_chk)
+  // When compiling with -fno-builtin, this is always true.
+  // When passing -ffreestanding/-mkernel, which both imply -fno-builtin, we
+  // end up with fortified libcalls, which isn't acceptable in a freestanding
+  // environment which only provides their non-fortified counterparts.
+  //
+  // Until we change clang and/or teach external users to check for availability
+  // differently, disregard the "nobuiltin" attribute and TLI::has.
+  //
+  // PR23093.
+
+  LibFunc::Func Func;
+  Function *Callee = CI->getCalledFunction();
+  StringRef FuncName = Callee->getName();
+  IRBuilder<> Builder(CI);
+  bool isCallingConvC = CI->getCallingConv() == llvm::CallingConv::C;
+
+  // First, check that this is a known library functions.
+  if (!TLI->getLibFunc(FuncName, Func))
+    return nullptr;
+
+  // We never change the calling convention.
+  if (!ignoreCallingConv(Func) && !isCallingConvC)
+    return nullptr;
+
+  switch (Func) {
+  case LibFunc::memcpy_chk:
+    return optimizeMemCpyChk(CI, Builder);
+  case LibFunc::memmove_chk:
+    return optimizeMemMoveChk(CI, Builder);
+  case LibFunc::memset_chk:
+    return optimizeMemSetChk(CI, Builder);
+  case LibFunc::stpcpy_chk:
+  case LibFunc::strcpy_chk:
+    return optimizeStrpCpyChk(CI, Builder, Func);
+  case LibFunc::stpncpy_chk:
+  case LibFunc::strncpy_chk:
+    return optimizeStrpNCpyChk(CI, Builder, Func);
+  default:
+    break;
+  }
+  return nullptr;
+}
+
+FortifiedLibCallSimplifier::FortifiedLibCallSimplifier(
+    const TargetLibraryInfo *TLI, bool OnlyLowerUnknownSize)
+    : TLI(TLI), OnlyLowerUnknownSize(OnlyLowerUnknownSize) {}