Skip to content

AMDGPU: Replace sqrt OpenCL libcalls with llvm.sqrt #74197

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jan 9, 2024
Merged

AMDGPU: Replace sqrt OpenCL libcalls with llvm.sqrt #74197

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 3 additions & 29 deletions llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -87,9 +87,6 @@ class AMDGPULibCalls {
Constant *copr0, Constant *copr1);
bool evaluateCall(CallInst *aCI, const FuncInfo &FInfo);

// sqrt
bool fold_sqrt(FPMathOperator *FPOp, IRBuilder<> &B, const FuncInfo &FInfo);

/// Insert a value to sincos function \p Fsincos. Returns (value of sin, value
/// of cos, sincos call).
std::tuple<Value *, Value *, Value *> insertSinCos(Value *Arg,
Expand Down Expand Up @@ -673,8 +670,6 @@ bool AMDGPULibCalls::fold(CallInst *CI) {

// Specialized optimizations for each function call.
//
// TODO: Handle other simple intrinsic wrappers. Sqrt.
//
// TODO: Handle native functions
switch (FInfo.getId()) {
case AMDGPULibFunc::EI_EXP:
Expand Down Expand Up @@ -795,7 +790,9 @@ bool AMDGPULibCalls::fold(CallInst *CI) {
case AMDGPULibFunc::EI_ROOTN:
return fold_rootn(FPOp, B, FInfo);
case AMDGPULibFunc::EI_SQRT:
return fold_sqrt(FPOp, B, FInfo);
// TODO: Allow with strictfp + constrained intrinsic
return tryReplaceLibcallWithSimpleIntrinsic(
B, CI, Intrinsic::sqrt, true, true, /*AllowStrictFP=*/false);
case AMDGPULibFunc::EI_COS:
case AMDGPULibFunc::EI_SIN:
return fold_sincos(FPOp, B, FInfo);
Expand Down Expand Up @@ -1275,29 +1272,6 @@ bool AMDGPULibCalls::tryReplaceLibcallWithSimpleIntrinsic(
return true;
}

// fold sqrt -> native_sqrt (x)
bool AMDGPULibCalls::fold_sqrt(FPMathOperator *FPOp, IRBuilder<> &B,
const FuncInfo &FInfo) {
if (!isUnsafeMath(FPOp))
return false;

if (getArgType(FInfo) == AMDGPULibFunc::F32 && (getVecSize(FInfo) == 1) &&
(FInfo.getPrefix() != AMDGPULibFunc::NATIVE)) {
Module *M = B.GetInsertBlock()->getModule();

if (FunctionCallee FPExpr = getNativeFunction(
M, AMDGPULibFunc(AMDGPULibFunc::EI_SQRT, FInfo))) {
Value *opr0 = FPOp->getOperand(0);
LLVM_DEBUG(errs() << "AMDIC: " << *FPOp << " ---> "
<< "sqrt(" << *opr0 << ")\n");
Value *nval = CreateCallEx(B,FPExpr, opr0, "__sqrt");
replaceCall(FPOp, nval);
return true;
}
}
return false;
}

std::tuple<Value *, Value *, Value *>
AMDGPULibCalls::insertSinCos(Value *Arg, FastMathFlags FMF, IRBuilder<> &B,
FunctionCallee Fsincos) {
Expand Down
Loading