#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"
#include "llvm/IR/PatternMatch.h"
#include "llvm/Support/Allocator.h"
#include "llvm/Support/CommandLine.h"
-#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Transforms/Utils/BuildLibCalls.h"
#include "llvm/Transforms/Utils/Local.h"
//===----------------------------------------------------------------------===//
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
}
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
// 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.
+ // 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")
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
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
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,
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
// 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 (canUseUnsafeFPMath(CI->getParent()->getParent())) {
+ if (unsafeFPMath) {
if (auto *OpC = dyn_cast<CallInst>(Op1)) {
IRBuilder<>::FastMathFlagGuard Guard(B);
FastMathFlags FMF;
B.SetFastMathFlags(FMF);
LibFunc::Func Func;
- Function *Callee = OpC->getCalledFunction();
- StringRef FuncName = Callee->getName();
-
- if (TLI->getLibFunc(FuncName, Func) && TLI->has(Func) &&
- (Func == LibFunc::exp || Func == LibFunc::exp2))
+ 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"), FuncName, B,
- Callee->getAttributes());
+ B.CreateFMul(OpC->getArgOperand(0), Op2, "mul"),
+ OpCCallee->getName(), B, OpCCallee->getAttributes());
}
}
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);
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
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.
// If we can shrink the call to a float function rather than a double
// function, do that first.
Function *Callee = CI->getCalledFunction();
- if ((Callee->getName() == "fmin" && TLI->has(LibFunc::fminf)) ||
- (Callee->getName() == "fmax" && TLI->has(LibFunc::fmaxf))) {
+ StringRef Name = Callee->getName();
+ if ((Name == "fmin" && hasFloatVersion(Name)) ||
+ (Name == "fmax" && hasFloatVersion(Name))) {
Value *Ret = optimizeBinaryDoubleFP(CI, B);
if (Ret)
return Ret;
Value *LibCallSimplifier::optimizeTan(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
Value *Ret = nullptr;
- if (UnsafeFPShrink && Callee->getName() == "tan" && TLI->has(LibFunc::tanf))
+ StringRef Name = Callee->getName();
+ if (UnsafeFPShrink && Name == "tan" && hasFloatVersion(Name))
Ret = optimizeUnaryDoubleFP(CI, B, true);
FunctionType *FT = Callee->getFunctionType();
// tanl(atanl(x)) -> x
LibFunc::Func Func;
Function *F = OpC->getCalledFunction();
- StringRef FuncName = F->getName();
- if (TLI->getLibFunc(FuncName, Func) && TLI->has(Func) &&
+ 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")))
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) {
// * 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)
//
// * 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'
//