Add strictfp attribute to prevent unwanted optimizations of libm calls

[andrew.w.kaylor]

Summary

In preparation for supporting the STDC FENV_ACCESS pragma (and related other pragmas), we need a way to prevent optimizations to libm function calls. To accomplish this, I am adding a new function attribute, strictfp, which tells the optimizer to handle these calls specially. This attribute should behave similarly to the nobuiltin attribute, but it isn't quite as restrictive. In particular, it allows optimizations that would be blocked by nobuiltin as long as the optimization is not related to FP semantics.

Since we don't expect (or want) the front end to maintain knowledge about which functions the optimizer recognizes as LibFuncs, the expectation is that the front end will attach the strictfp attribute to all function calls within any scope where FENV_ACCESS is enabled. This should have no effect for functions that are not FP-related.

[craig.topper]

include/llvm/Bitcode/LLVMBitCodes.h

552 ↗

(On Diff #102387)

Maybe just leave a trailnig comma on the end here so the next guy doesn't have to add one. Looks like the 2 enums below have a trailing comma.

lib/AsmParser/LLParser.cpp

1124 ↗

(On Diff #102387)

Should this be after sspstrong to match the order in LLLexer.cpp which seems to be mostly alphabetical except for the sanitize and safestack?

lib/AsmParser/LLToken.h

212 ↗

(On Diff #102387)

Wow how many different not alphabetical orders do we have on these keywords?

lib/Transforms/Utils/SimplifyLibCalls.cpp

2055 ↗

(On Diff #102387)

Doesnt' the caller already check getLibFunc and TLI->has?

[andrew.w.kaylor]

lib/AsmParser/LLToken.h

212 ↗

(On Diff #102387)

Yeah, if you don't object I'm going to leave this patch in the "approximately alphabetical" state it's in and maybe clean them all up afterward in a separate patch.

lib/Transforms/Utils/SimplifyLibCalls.cpp

2055 ↗

(On Diff #102387)

Yes, but I need to call getLibFunc() here to get the Func (or I could pass it as an argument). I chose to do it this way to mimic the behavior of optimizeStringMemoryLibCall(), but I suppose there is some benefit to avoiding the churn that getLibFunc goes through with the StringRef.

[andrew.w.kaylor]

-Addressed review feedback
-Rebased

[craig.topper]

LGTM.

Can you also add strictfp to utils/vim/syntax/llvm.vim where the other keywords are listed?

llvm/trunk/docs/LangRef.rst

``sspstrong``
     #. Variables that have had their address taken are 3rd closest to the
        protector.
 
     This overrides the ``ssp`` function attribute.
 
     If a function that has an ``sspstrong`` attribute is inlined into a
     function that doesn't have an ``sspstrong`` attribute, then the
     resulting function will have an ``sspstrong`` attribute.
+``strictfp``
+    This attribute indicates that the function was called from a scope that
+    requires strict floating point semantics.  LLVM will not attempt any
+    optimizations that require assumptions about the floating point rounding
+    mode or that might alter the state of floating point status flags that
+    might otherwise be set or cleared by calling this function.
 ``"thunk"``
     This attribute indicates that the function will delegate to some other
     function with a tail call. The prototype of a thunk should not be used for
     optimization purposes. The caller is expected to cast the thunk prototype to
     match the thunk target prototype.
 ``uwtable``
     This attribute indicates that the ABI being targeted requires that
     an unwind table entry be produced for this function even if we can

llvm/trunk/include/llvm/Bitcode/LLVMBitCodes.h

enum AttributeKindCodes {
   ATTR_KIND_ARGMEMONLY = 45,
   ATTR_KIND_SWIFT_SELF = 46,
   ATTR_KIND_SWIFT_ERROR = 47,
   ATTR_KIND_NO_RECURSE = 48,
   ATTR_KIND_INACCESSIBLEMEM_ONLY = 49,
   ATTR_KIND_INACCESSIBLEMEM_OR_ARGMEMONLY = 50,
   ATTR_KIND_ALLOC_SIZE = 51,
   ATTR_KIND_WRITEONLY = 52,
-  ATTR_KIND_SPECULATABLE = 53
+  ATTR_KIND_SPECULATABLE = 53,
+  ATTR_KIND_STRICT_FP = 54,
 };
 
 enum ComdatSelectionKindCodes {
   COMDAT_SELECTION_KIND_ANY = 1,
   COMDAT_SELECTION_KIND_EXACT_MATCH = 2,
   COMDAT_SELECTION_KIND_LARGEST = 3,
   COMDAT_SELECTION_KIND_NO_DUPLICATES = 4,
   COMDAT_SELECTION_KIND_SAME_SIZE = 5,

llvm/trunk/include/llvm/IR/Attributes.td

 def StackProtect : EnumAttr<"ssp">;
 
 /// Stack protection required.
 def StackProtectReq : EnumAttr<"sspreq">;
 
 /// Strong Stack protection.
 def StackProtectStrong : EnumAttr<"sspstrong">;
 
+/// Function was called in a scope requiring strict floating point semantics.
+def StrictFP : EnumAttr<"strictfp">;
+
 /// Hidden pointer to structure to return.
 def StructRet : EnumAttr<"sret">;
 
 /// AddressSanitizer is on.
 def SanitizeAddress : EnumAttr<"sanitize_address">;
 
 /// ThreadSanitizer is on.
 def SanitizeThread : EnumAttr<"sanitize_thread">;

llvm/trunk/include/llvm/IR/CallSite.h

bool returnDoesNotAlias() const {
     CALLSITE_DELEGATE_GETTER(returnDoesNotAlias());
   }
 
   /// Return true if the call should not be treated as a call to a builtin.
   bool isNoBuiltin() const {
     CALLSITE_DELEGATE_GETTER(isNoBuiltin());
   }
 
+  /// Return true if the call requires strict floating point semantics.
+  bool isStrictFP() const {
+    CALLSITE_DELEGATE_GETTER(isStrictFP());
+  }
+
   /// Return true if the call should not be inlined.
   bool isNoInline() const {
     CALLSITE_DELEGATE_GETTER(isNoInline());
   }
   void setIsNoInline(bool Value = true) {
     CALLSITE_DELEGATE_SETTER(setIsNoInline(Value));
   }
 

llvm/trunk/include/llvm/IR/Instructions.h

public:
 
   /// Return true if the call should not be treated as a call to a
   /// builtin.
   bool isNoBuiltin() const {
     return hasFnAttrImpl(Attribute::NoBuiltin) &&
       !hasFnAttrImpl(Attribute::Builtin);
   }
 
+  /// Determine if the call requires strict floating point semantics.
+  bool isStrictFP() const { return hasFnAttr(Attribute::StrictFP); }
+
   /// Return true if the call should not be inlined.
   bool isNoInline() const { return hasFnAttr(Attribute::NoInline); }
   void setIsNoInline() {
     addAttribute(AttributeList::FunctionIndex, Attribute::NoInline);
   }
 
   /// Return true if the call can return twice
   bool canReturnTwice() const {
public:
   /// builtin.
   bool isNoBuiltin() const {
     // We assert in hasFnAttr if one passes in Attribute::NoBuiltin, so we have
     // to check it by hand.
     return hasFnAttrImpl(Attribute::NoBuiltin) &&
       !hasFnAttrImpl(Attribute::Builtin);
   }
 
+  /// Determine if the call requires strict floating point semantics.
+  bool isStrictFP() const { return hasFnAttr(Attribute::StrictFP); }
+
   /// Return true if the call should not be inlined.
   bool isNoInline() const { return hasFnAttr(Attribute::NoInline); }
   void setIsNoInline() {
     addAttribute(AttributeList::FunctionIndex, Attribute::NoInline);
   }
 
   /// Determine if the call does not access memory.
   bool doesNotAccessMemory() const {

llvm/trunk/include/llvm/Transforms/Utils/SimplifyLibCalls.h

private:
   Value *optimizeCos(CallInst *CI, IRBuilder<> &B);
   Value *optimizePow(CallInst *CI, IRBuilder<> &B);
   Value *optimizeExp2(CallInst *CI, IRBuilder<> &B);
   Value *optimizeFMinFMax(CallInst *CI, IRBuilder<> &B);
   Value *optimizeLog(CallInst *CI, IRBuilder<> &B);
   Value *optimizeSqrt(CallInst *CI, IRBuilder<> &B);
   Value *optimizeSinCosPi(CallInst *CI, IRBuilder<> &B);
   Value *optimizeTan(CallInst *CI, IRBuilder<> &B);
+  // Wrapper for all floating point library call optimizations
+  Value *optimizeFloatingPointLibCall(CallInst *CI, LibFunc Func,
+                                      IRBuilder<> &B);
 
   // Integer Library Call Optimizations
   Value *optimizeFFS(CallInst *CI, IRBuilder<> &B);
   Value *optimizeFls(CallInst *CI, IRBuilder<> &B);
   Value *optimizeAbs(CallInst *CI, IRBuilder<> &B);
   Value *optimizeIsDigit(CallInst *CI, IRBuilder<> &B);
   Value *optimizeIsAscii(CallInst *CI, IRBuilder<> &B);
   Value *optimizeToAscii(CallInst *CI, IRBuilder<> &B);

llvm/trunk/lib/Analysis/ConstantFolding.cpp

llvm::ConstantFoldLoadThroughGEPIndices(Constant *C,
   return C;
 }
 
 //===----------------------------------------------------------------------===//
 //  Constant Folding for Calls
 //
 
 bool llvm::canConstantFoldCallTo(ImmutableCallSite CS, const Function *F) {
-  if (CS.isNoBuiltin())
+  if (CS.isNoBuiltin() || CS.isStrictFP())
     return false;
   switch (F->getIntrinsicID()) {
   case Intrinsic::fabs:
   case Intrinsic::minnum:
   case Intrinsic::maxnum:
   case Intrinsic::log:
   case Intrinsic::log2:
   case Intrinsic::log10:

 }
 
 } // end anonymous namespace
 
 Constant *
 llvm::ConstantFoldCall(ImmutableCallSite CS, Function *F,
                        ArrayRef<Constant *> Operands,
                        const TargetLibraryInfo *TLI) {
-  if (CS.isNoBuiltin())
+  if (CS.isNoBuiltin() || CS.isStrictFP())
     return nullptr;
   if (!F->hasName())
     return nullptr;
   StringRef Name = F->getName();
 
   Type *Ty = F->getReturnType();
 
   if (auto *VTy = dyn_cast<VectorType>(Ty))
     return ConstantFoldVectorCall(Name, F->getIntrinsicID(), VTy, Operands,
                                   F->getParent()->getDataLayout(), TLI);
 
   return ConstantFoldScalarCall(Name, F->getIntrinsicID(), Ty, Operands, TLI);
 }
 
 bool llvm::isMathLibCallNoop(CallSite CS, const TargetLibraryInfo *TLI) {
   // FIXME: Refactor this code; this duplicates logic in LibCallsShrinkWrap
   // (and to some extent ConstantFoldScalarCall).
-  if (CS.isNoBuiltin())
+  if (CS.isNoBuiltin() || CS.isStrictFP())
     return false;
   Function *F = CS.getCalledFunction();
   if (!F)
     return false;
 
   LibFunc Func;
   if (!TLI || !TLI->getLibFunc(*F, Func))
     return false;

llvm/trunk/lib/AsmParser/LLLexer.cpp

#define KEYWORD(STR)                                                           \
   KEYWORD(returned);
   KEYWORD(returns_twice);
   KEYWORD(signext);
   KEYWORD(speculatable);
   KEYWORD(sret);
   KEYWORD(ssp);
   KEYWORD(sspreq);
   KEYWORD(sspstrong);
+  KEYWORD(strictfp);
   KEYWORD(safestack);
   KEYWORD(sanitize_address);
   KEYWORD(sanitize_thread);
   KEYWORD(sanitize_memory);
   KEYWORD(swifterror);
   KEYWORD(swiftself);
   KEYWORD(uwtable);
   KEYWORD(writeonly);

llvm/trunk/lib/AsmParser/LLParser.cpp

case lltok::kw_sspstrong:
       B.addAttribute(Attribute::StackProtectStrong); break;
     case lltok::kw_safestack: B.addAttribute(Attribute::SafeStack); break;
     case lltok::kw_sanitize_address:
       B.addAttribute(Attribute::SanitizeAddress); break;
     case lltok::kw_sanitize_thread:
       B.addAttribute(Attribute::SanitizeThread); break;
     case lltok::kw_sanitize_memory:
       B.addAttribute(Attribute::SanitizeMemory); break;
+    case lltok::kw_strictfp: B.addAttribute(Attribute::StrictFP); break;
     case lltok::kw_uwtable: B.addAttribute(Attribute::UWTable); break;
     case lltok::kw_writeonly: B.addAttribute(Attribute::WriteOnly); break;
 
     // Error handling.
     case lltok::kw_inreg:
     case lltok::kw_signext:
     case lltok::kw_zeroext:
       HaveError |=
while (true) {
     case lltok::kw_returns_twice:
     case lltok::kw_sanitize_address:
     case lltok::kw_sanitize_memory:
     case lltok::kw_sanitize_thread:
     case lltok::kw_ssp:
     case lltok::kw_sspreq:
     case lltok::kw_sspstrong:
     case lltok::kw_safestack:
+    case lltok::kw_strictfp:
     case lltok::kw_uwtable:
       HaveError |= Error(Lex.getLoc(), "invalid use of function-only attribute");
       break;
     }
 
     Lex.Lex();
   }
 }
while (true) {
     case lltok::kw_returns_twice:
     case lltok::kw_sanitize_address:
     case lltok::kw_sanitize_memory:
     case lltok::kw_sanitize_thread:
     case lltok::kw_ssp:
     case lltok::kw_sspreq:
     case lltok::kw_sspstrong:
     case lltok::kw_safestack:
+    case lltok::kw_strictfp:
     case lltok::kw_uwtable:
       HaveError |= Error(Lex.getLoc(), "invalid use of function-only attribute");
       break;
 
     case lltok::kw_readnone:
     case lltok::kw_readonly:
       HaveError |= Error(Lex.getLoc(), "invalid use of attribute on return type");
     }

llvm/trunk/lib/AsmParser/LLToken.h

enum Kind {
   kw_speculatable,
   kw_ssp,
   kw_sspreq,
   kw_sspstrong,
   kw_safestack,
   kw_sret,
   kw_sanitize_thread,
   kw_sanitize_memory,
+  kw_strictfp,
   kw_swifterror,
   kw_swiftself,
   kw_uwtable,
   kw_writeonly,
   kw_zeroext,
 
   kw_type,
   kw_opaque,

llvm/trunk/lib/Bitcode/Reader/BitcodeReader.cpp

static uint64_t getRawAttributeMask(Attribute::AttrKind Val) {
   case Attribute::SafeStack:       return 1ULL << 47;
   case Attribute::NoRecurse:       return 1ULL << 48;
   case Attribute::InaccessibleMemOnly:         return 1ULL << 49;
   case Attribute::InaccessibleMemOrArgMemOnly: return 1ULL << 50;
   case Attribute::SwiftSelf:       return 1ULL << 51;
   case Attribute::SwiftError:      return 1ULL << 52;
   case Attribute::WriteOnly:       return 1ULL << 53;
   case Attribute::Speculatable:    return 1ULL << 54;
+  case Attribute::StrictFP:        return 1ULL << 55;
   case Attribute::Dereferenceable:
     llvm_unreachable("dereferenceable attribute not supported in raw format");
     break;
   case Attribute::DereferenceableOrNull:
     llvm_unreachable("dereferenceable_or_null attribute not supported in raw "
                      "format");
     break;
   case Attribute::ArgMemOnly:
static Attribute::AttrKind getAttrFromCode(uint64_t Code) {
   case bitc::ATTR_KIND_STACK_PROTECT:
     return Attribute::StackProtect;
   case bitc::ATTR_KIND_STACK_PROTECT_REQ:
     return Attribute::StackProtectReq;
   case bitc::ATTR_KIND_STACK_PROTECT_STRONG:
     return Attribute::StackProtectStrong;
   case bitc::ATTR_KIND_SAFESTACK:
     return Attribute::SafeStack;
+  case bitc::ATTR_KIND_STRICT_FP:
+    return Attribute::StrictFP;
   case bitc::ATTR_KIND_STRUCT_RET:
     return Attribute::StructRet;
   case bitc::ATTR_KIND_SANITIZE_ADDRESS:
     return Attribute::SanitizeAddress;
   case bitc::ATTR_KIND_SANITIZE_THREAD:
     return Attribute::SanitizeThread;
   case bitc::ATTR_KIND_SANITIZE_MEMORY:
     return Attribute::SanitizeMemory;

llvm/trunk/lib/Bitcode/Writer/BitcodeWriter.cpp

static uint64_t getAttrKindEncoding(Attribute::AttrKind Kind) {
   case Attribute::StackProtect:
     return bitc::ATTR_KIND_STACK_PROTECT;
   case Attribute::StackProtectReq:
     return bitc::ATTR_KIND_STACK_PROTECT_REQ;
   case Attribute::StackProtectStrong:
     return bitc::ATTR_KIND_STACK_PROTECT_STRONG;
   case Attribute::SafeStack:
     return bitc::ATTR_KIND_SAFESTACK;
+  case Attribute::StrictFP:
+    return bitc::ATTR_KIND_STRICT_FP;
   case Attribute::StructRet:
     return bitc::ATTR_KIND_STRUCT_RET;
   case Attribute::SanitizeAddress:
     return bitc::ATTR_KIND_SANITIZE_ADDRESS;
   case Attribute::SanitizeThread:
     return bitc::ATTR_KIND_SANITIZE_THREAD;
   case Attribute::SanitizeMemory:
     return bitc::ATTR_KIND_SANITIZE_MEMORY;

llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

if (F->isDeclaration()) {
         RenameFn = visitIntrinsicCall(I, IID);
         if (!RenameFn)
           return;
       }
     }
 
     // Check for well-known libc/libm calls.  If the function is internal, it
     // can't be a library call.  Don't do the check if marked as nobuiltin for
-    // some reason.
+    // some reason or the call site requires strict floating point semantics.
     LibFunc Func;
-    if (!I.isNoBuiltin() && !F->hasLocalLinkage() && F->hasName() &&
-        LibInfo->getLibFunc(*F, Func) &&
+    if (!I.isNoBuiltin() && !I.isStrictFP() && !F->hasLocalLinkage() &&
+        F->hasName() && LibInfo->getLibFunc(*F, Func) &&
         LibInfo->hasOptimizedCodeGen(Func)) {
       switch (Func) {
       default: break;
       case LibFunc_copysign:
       case LibFunc_copysignf:
       case LibFunc_copysignl:
         // We already checked this call's prototype; verify it doesn't modify
         // errno.

llvm/trunk/lib/IR/Attributes.cpp

std::string Attribute::getAsString(bool InAttrGrp) const {
   if (hasAttribute(Attribute::StackProtect))
     return "ssp";
   if (hasAttribute(Attribute::StackProtectReq))
     return "sspreq";
   if (hasAttribute(Attribute::StackProtectStrong))
     return "sspstrong";
   if (hasAttribute(Attribute::SafeStack))
     return "safestack";
+  if (hasAttribute(Attribute::StrictFP))
+    return "strictfp";
   if (hasAttribute(Attribute::StructRet))
     return "sret";
   if (hasAttribute(Attribute::SanitizeThread))
     return "sanitize_thread";
   if (hasAttribute(Attribute::SanitizeMemory))
     return "sanitize_memory";
   if (hasAttribute(Attribute::UWTable))
     return "uwtable";

llvm/trunk/lib/IR/Verifier.cpp

static bool isFuncOnlyAttr(Attribute::AttrKind Kind) {
   case Attribute::JumpTable:
   case Attribute::Convergent:
   case Attribute::ArgMemOnly:
   case Attribute::NoRecurse:
   case Attribute::InaccessibleMemOnly:
   case Attribute::InaccessibleMemOrArgMemOnly:
   case Attribute::AllocSize:
   case Attribute::Speculatable:
+  case Attribute::StrictFP:
     return true;
   default:
     break;
   }
   return false;
 }
 
 /// Return true if this is a function attribute that can also appear on

llvm/trunk/lib/Transforms/IPO/ForceFunctionAttrs.cpp

return StringSwitch<Attribute::AttrKind>(Kind)
       .Case("returns_twice", Attribute::ReturnsTwice)
       .Case("safestack", Attribute::SafeStack)
       .Case("sanitize_address", Attribute::SanitizeAddress)
       .Case("sanitize_memory", Attribute::SanitizeMemory)
       .Case("sanitize_thread", Attribute::SanitizeThread)
       .Case("ssp", Attribute::StackProtect)
       .Case("sspreq", Attribute::StackProtectReq)
       .Case("sspstrong", Attribute::StackProtectStrong)
+      .Case("strictfp", Attribute::StrictFP)
       .Case("uwtable", Attribute::UWTable)
       .Default(Attribute::None);
 }
 
 /// If F has any forced attributes given on the command line, add them.
 static void addForcedAttributes(Function &F) {
   for (auto &S : ForceAttributes) {
     auto KV = StringRef(S).split(':');

llvm/trunk/lib/Transforms/Utils/SimplifyLibCalls.cpp

case LibFunc_wcslen:
       return optimizeWcslen(CI, Builder);
     default:
       break;
     }
   }
   return nullptr;
 }
 
+Value *LibCallSimplifier::optimizeFloatingPointLibCall(CallInst *CI,
+                                                       LibFunc Func,
+                                                       IRBuilder<> &Builder) {
+  // Don't optimize calls that require strict floating point semantics.
+  if (CI->isStrictFP())
+    return nullptr;
+
+  switch (Func) {
+  case LibFunc_cosf:
+  case LibFunc_cos:
+  case LibFunc_cosl:
+    return optimizeCos(CI, Builder);
+  case LibFunc_sinpif:
+  case LibFunc_sinpi:
+  case LibFunc_cospif:
+  case LibFunc_cospi:
+    return optimizeSinCosPi(CI, Builder);
+  case LibFunc_powf:
+  case LibFunc_pow:
+  case LibFunc_powl:
+    return optimizePow(CI, Builder);
+  case LibFunc_exp2l:
+  case LibFunc_exp2:
+  case LibFunc_exp2f:
+    return optimizeExp2(CI, Builder);
+  case LibFunc_fabsf:
+  case LibFunc_fabs:
+  case LibFunc_fabsl:
+    return replaceUnaryCall(CI, Builder, Intrinsic::fabs);
+  case LibFunc_sqrtf:
+  case LibFunc_sqrt:
+  case LibFunc_sqrtl:
+    return optimizeSqrt(CI, Builder);
+  case LibFunc_log:
+  case LibFunc_log10:
+  case LibFunc_log1p:
+  case LibFunc_log2:
+  case LibFunc_logb:
+    return optimizeLog(CI, Builder);
+  case LibFunc_tan:
+  case LibFunc_tanf:
+  case LibFunc_tanl:
+    return optimizeTan(CI, Builder);
+  case LibFunc_ceil:
+    return replaceUnaryCall(CI, Builder, Intrinsic::ceil);
+  case LibFunc_floor:
+    return replaceUnaryCall(CI, Builder, Intrinsic::floor);
+  case LibFunc_round:
+    return replaceUnaryCall(CI, Builder, Intrinsic::round);
+  case LibFunc_nearbyint:
+    return replaceUnaryCall(CI, Builder, Intrinsic::nearbyint);
+  case LibFunc_rint:
+    return replaceUnaryCall(CI, Builder, Intrinsic::rint);
+  case LibFunc_trunc:
+    return replaceUnaryCall(CI, Builder, Intrinsic::trunc);
+  case LibFunc_acos:
+  case LibFunc_acosh:
+  case LibFunc_asin:
+  case LibFunc_asinh:
+  case LibFunc_atan:
+  case LibFunc_atanh:
+  case LibFunc_cbrt:
+  case LibFunc_cosh:
+  case LibFunc_exp:
+  case LibFunc_exp10:
+  case LibFunc_expm1:
+  case LibFunc_sin:
+  case LibFunc_sinh:
+  case LibFunc_tanh:
+    if (UnsafeFPShrink && hasFloatVersion(CI->getCalledFunction()->getName()))
+      return optimizeUnaryDoubleFP(CI, Builder, true);
+    return nullptr;
+  case LibFunc_copysign:
+    if (hasFloatVersion(CI->getCalledFunction()->getName()))
+      return optimizeBinaryDoubleFP(CI, Builder);
+    return nullptr;
+  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;
+  }
+}
+
 Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
+  // TODO: Split out the code below that operates on FP calls so that
+  //       we can all non-FP calls with the StrictFP attribute to be
+  //       optimized.
   if (CI->isNoBuiltin())
     return nullptr;
 
   LibFunc Func;
   Function *Callee = CI->getCalledFunction();
-  StringRef FuncName = Callee->getName();
 
   SmallVector<OperandBundleDef, 2> OpBundles;
   CI->getOperandBundlesAsDefs(OpBundles);
   IRBuilder<> Builder(CI, /*FPMathTag=*/nullptr, OpBundles);
   bool isCallingConvC = isCallingConvCCompatible(CI);
 
   // Command-line parameter overrides instruction attribute.
+  // This can't be moved to optimizeFloatingPointLibCall() because it may be
+  // used by the intrinsic optimizations. 
   if (EnableUnsafeFPShrink.getNumOccurrences() > 0)
     UnsafeFPShrink = EnableUnsafeFPShrink;
   else if (isa<FPMathOperator>(CI) && CI->hasUnsafeAlgebra())
     UnsafeFPShrink = true;
 
   // First, check for intrinsics.
   if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(CI)) {
     if (!isCallingConvC)
       return nullptr;
+    // The FP intrinsics have corresponding constrained versions so we don't
+    // need to check for the StrictFP attribute here.
     switch (II->getIntrinsicID()) {
     case Intrinsic::pow:
       return optimizePow(CI, Builder);
     case Intrinsic::exp2:
       return optimizeExp2(CI, Builder);
     case Intrinsic::log:
       return optimizeLog(CI, Builder);
     case Intrinsic::sqrt:
Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
 
   // Then check for known library functions.
   if (TLI->getLibFunc(*Callee, Func) && TLI->has(Func)) {
     // We never change the calling convention.
     if (!ignoreCallingConv(Func) && !isCallingConvC)
       return nullptr;
     if (Value *V = optimizeStringMemoryLibCall(CI, Builder))
       return V;
+    if (Value *V = optimizeFloatingPointLibCall(CI, Func, Builder))
+      return V;
     switch (Func) {
-    case LibFunc_cosf:
-    case LibFunc_cos:
-    case LibFunc_cosl:
-      return optimizeCos(CI, Builder);
-    case LibFunc_sinpif:
-    case LibFunc_sinpi:
-    case LibFunc_cospif:
-    case LibFunc_cospi:
-      return optimizeSinCosPi(CI, Builder);
-    case LibFunc_powf:
-    case LibFunc_pow:
-    case LibFunc_powl:
-      return optimizePow(CI, Builder);
-    case LibFunc_exp2l:
-    case LibFunc_exp2:
-    case LibFunc_exp2f:
-      return optimizeExp2(CI, Builder);
-    case LibFunc_fabsf:
-    case LibFunc_fabs:
-    case LibFunc_fabsl:
-      return replaceUnaryCall(CI, Builder, Intrinsic::fabs);
-    case LibFunc_sqrtf:
-    case LibFunc_sqrt:
-    case LibFunc_sqrtl:
-      return optimizeSqrt(CI, Builder);
     case LibFunc_ffs:
     case LibFunc_ffsl:
     case LibFunc_ffsll:
       return optimizeFFS(CI, Builder);
     case LibFunc_fls:
     case LibFunc_flsl:
     case LibFunc_flsll:
       return optimizeFls(CI, Builder);
if (TLI->getLibFunc(*Callee, Func) && TLI->has(Func)) {
     case LibFunc_sprintf:
       return optimizeSPrintF(CI, Builder);
     case LibFunc_fprintf:
       return optimizeFPrintF(CI, Builder);
     case LibFunc_fwrite:
       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:
     case LibFunc_fiprintf:
       return optimizeErrorReporting(CI, Builder, 0);
     case LibFunc_fputc:
       return optimizeErrorReporting(CI, Builder, 1);
-    case LibFunc_ceil:
-      return replaceUnaryCall(CI, Builder, Intrinsic::ceil);
-    case LibFunc_floor:
-      return replaceUnaryCall(CI, Builder, Intrinsic::floor);
-    case LibFunc_round:
-      return replaceUnaryCall(CI, Builder, Intrinsic::round);
-    case LibFunc_nearbyint:
-      return replaceUnaryCall(CI, Builder, Intrinsic::nearbyint);
-    case LibFunc_rint:
-      return replaceUnaryCall(CI, Builder, Intrinsic::rint);
-    case LibFunc_trunc:
-      return replaceUnaryCall(CI, Builder, Intrinsic::trunc);
-    case LibFunc_acos:
-    case LibFunc_acosh:
-    case LibFunc_asin:
-    case LibFunc_asinh:
-    case LibFunc_atan:
-    case LibFunc_atanh:
-    case LibFunc_cbrt:
-    case LibFunc_cosh:
-    case LibFunc_exp:
-    case LibFunc_exp10:
-    case LibFunc_expm1:
-    case LibFunc_sin:
-    case LibFunc_sinh:
-    case LibFunc_tanh:
-      if (UnsafeFPShrink && hasFloatVersion(FuncName))
-        return optimizeUnaryDoubleFP(CI, Builder, true);
-      return nullptr;
-    case LibFunc_copysign:
-      if (hasFloatVersion(FuncName))
-        return optimizeBinaryDoubleFP(CI, Builder);
-      return nullptr;
-    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(

llvm/trunk/test/Bitcode/compatibility.ll

 declare void @f.kvpair() "cpu"="cortex-a8"
 ; CHECK:declare void @f.kvpair() #31
 declare void @f.norecurse() norecurse
 ; CHECK: declare void @f.norecurse() #32
 declare void @f.inaccessiblememonly() inaccessiblememonly
 ; CHECK: declare void @f.inaccessiblememonly() #33
 declare void @f.inaccessiblemem_or_argmemonly() inaccessiblemem_or_argmemonly
 ; CHECK: declare void @f.inaccessiblemem_or_argmemonly() #34
+declare void @f.strictfp() #35
 
 ; Functions -- section
 declare void @f.section() section "80"
 ; CHECK: declare void @f.section() section "80"
 
 ; Functions -- comdat
 define void @f.comdat_any() comdat($comdat.any) {
 ; CHECK: define void @f.comdat_any() comdat($comdat.any)
exit:
   select i1 true, i32 0, i32 1
   ; CHECK: select i1 true, i32 0, i32 1
   select <2 x i1> <i1 true, i1 false>, <2 x i8> <i8 2, i8 3>, <2 x i8> <i8 3, i8 2>
   ; CHECK: select <2 x i1> <i1 true, i1 false>, <2 x i8> <i8 2, i8 3>, <2 x i8> <i8 3, i8 2>
 
   call void @f.nobuiltin() builtin
   ; CHECK: call void @f.nobuiltin() #42
 
+  call void @f.strictfp() strictfp
+  ; CHECK: call void @f.strictfp() #43
+
   call fastcc noalias i32* @f.noalias() noinline
   ; CHECK: call fastcc noalias i32* @f.noalias() #12
   tail call ghccc nonnull i32* @f.nonnull() minsize
   ; CHECK: tail call ghccc nonnull i32* @f.nonnull() #7
 
   ret void
 }
 

 ; CHECK: attributes #35 = { nounwind readnone }
 ; CHECK: attributes #36 = { argmemonly nounwind readonly }
 ; CHECK: attributes #37 = { argmemonly nounwind }
 ; CHECK: attributes #38 = { nounwind readonly }
 ; CHECK: attributes #39 = { inaccessiblemem_or_argmemonly nounwind }
 ; CHECK: attributes #40 = { writeonly }
 ; CHECK: attributes #41 = { speculatable }
 ; CHECK: attributes #42 = { builtin }
+; CHECK: attributes #43 = { strictfp }
 
 ;; Metadata
 
 ; Metadata -- Module flags
 !llvm.module.flags = !{!0, !1, !2, !4, !5, !6}
 ; CHECK: !llvm.module.flags = !{!0, !1, !2, !4, !5, !6}
 
 !0 = !{i32 1, !"mod1", i32 0}

llvm/trunk/test/Transforms/DCE/calls-errno.ll

 ; cos(inf) is a domain error
 ; CHECK-NEXT: %cos2 = call double @cos(double 0x7FF0000000000000)
   %cos2 = call double @cos(double 0x7FF0000000000000)
 
 ; cos(0) nobuiltin may have side effects 
 ; CHECK-NEXT: %cos3 = call double @cos(double 0.000000e+00)
   %cos3 = call double @cos(double 0.000000e+00) nobuiltin
 
+; cos(1) strictfp sets FP status flags
+; CHECK-NEXT: %cos4 = call double @cos(double 1.000000e+00)
+  %cos4 = call double @cos(double 1.000000e+00) strictfp
+
 ; pow(0, 1) is 0
   %pow1 = call double @pow(double 0x7FF0000000000000, double 1.000000e+00)
 
 ; pow(0, -1) is a pole error
 ; FIXME: It fails on mingw host. Suppress checking.
 ; %pow2 = call double @pow(double 0.000000e+00, double -1.000000e+00)
 
 ; fmod(inf, nan) is nan

llvm/trunk/test/Transforms/InstCombine/constant-fold-libfunc.ll

 ; RUN: opt < %s -instcombine -S | FileCheck %s
 
 declare double @acos(double)
 
 ; Check that functions without any function attributes are simplified.
 
 define double @test_simplify_acos() {
 ; CHECK-LABEL: @test_simplify_acos
   %pi = call double @acos(double -1.000000e+00)
 ; CHECK-NOT: call double @acos
 ; CHECK: ret double 0x400921FB54442D18
   ret double %pi
 }
 
+; Check that we don't constant fold builtin functions.
+
 define double @test_acos_nobuiltin() {
 ; CHECK-LABEL: @test_acos_nobuiltin
   %pi = call double @acos(double -1.000000e+00) nobuiltin 
 ; CHECK: call double @acos(double -1.000000e+00)
   ret double %pi
 }
+
+; Check that we don't constant fold strictfp results that require rounding.
+
+define double @test_acos_strictfp() {
+; CHECK-LABEL: @test_acos_strictfp
+  %pi = call double @acos(double -1.000000e+00) strictfp 
+; CHECK: call double @acos(double -1.000000e+00)
+  ret double %pi
+}

llvm/trunk/test/Transforms/InstCombine/memcpy-1.ll

 ; CHECK-LABEL: @test_simplify1(
 ; CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* %mem1, i8* %mem2, i32 %size, i32 1, i1 false)
 ; CHECK-NEXT:    ret i8* %mem1
 ;
   %ret = call i8* @memcpy(i8* %mem1, i8* %mem2, i32 %size)
   ret i8* %ret
 }
 
+; Verify that the strictfp attr doesn't block this optimization.
+
+define i8* @test_simplify2(i8* %mem1, i8* %mem2, i32 %size) {
+; CHECK-LABEL: @test_simplify2(
+  %ret = call i8* @memcpy(i8* %mem1, i8* %mem2, i32 %size) strictfp
+; CHECK: call void @llvm.memcpy
+  ret i8* %ret
+; CHECK: ret i8* %mem1
+}

llvm/trunk/utils/vim/syntax/llvm.vim

syn keyword llvmKeyword
       \ source_filename
       \ speculatable
       \ spir_func
       \ spir_kernel
       \ sret
       \ ssp
       \ sspreq
       \ sspstrong
+      \ strictfp
       \ swiftcc
       \ tail
       \ target
       \ thread_local
       \ to
       \ triple
       \ unnamed_addr
       \ unordered