Skip to content

[AMDGPU] Remove NoNaNsFPMath uses #180469

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
paperchalice merged 1 commit into from
Feb 9, 2026
Merged

[AMDGPU] Remove NoNaNsFPMath uses #180469

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
4 changes: 1 addition & 3 deletions llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2776,7 +2776,6 @@ SDValue AMDGPUTargetLowering::LowerFLOGCommon(SDValue Op,
const bool IsLog10 = Op.getOpcode() == ISD::FLOG10;
assert(IsLog10 || Op.getOpcode() == ISD::FLOG);

const auto &Options = getTargetMachine().Options;
if (VT == MVT::f16 || Flags.hasApproximateFuncs()) {

if (VT == MVT::f16 && !isTypeLegal(MVT::f16)) {
Expand Down Expand Up @@ -2845,8 +2844,7 @@ SDValue AMDGPUTargetLowering::LowerFLOGCommon(SDValue Op,
R = getMad(DAG, DL, VT, YH, CH, Mad1);
}

const bool IsFiniteOnly =
(Flags.hasNoNaNs() || Options.NoNaNsFPMath) && Flags.hasNoInfs();
const bool IsFiniteOnly = Flags.hasNoNaNs() && Flags.hasNoInfs();

// TODO: Check if known finite from source value.
if (!IsFiniteOnly) {
Expand Down
7 changes: 1 addition & 6 deletions llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3554,14 +3554,10 @@ bool AMDGPULegalizerInfo::legalizeFlogCommon(MachineInstr &MI,
Register X = MI.getOperand(1).getReg();
unsigned Flags = MI.getFlags();
const LLT Ty = MRI.getType(X);
MachineFunction &MF = B.getMF();

const LLT F32 = LLT::scalar(32);
const LLT F16 = LLT::scalar(16);

const AMDGPUTargetMachine &TM =
static_cast<const AMDGPUTargetMachine &>(MF.getTarget());

if (Ty == F16 || MI.getFlag(MachineInstr::FmAfn)) {
if (Ty == F16 && !ST.has16BitInsts()) {
Register LogVal = MRI.createGenericVirtualRegister(F32);
Expand Down Expand Up @@ -3630,8 +3626,7 @@ bool AMDGPULegalizerInfo::legalizeFlogCommon(MachineInstr &MI,
}

const bool IsFiniteOnly =
(MI.getFlag(MachineInstr::FmNoNans) || TM.Options.NoNaNsFPMath) &&
MI.getFlag(MachineInstr::FmNoInfs);
MI.getFlag(MachineInstr::FmNoNans) && MI.getFlag(MachineInstr::FmNoInfs);

if (!IsFiniteOnly) {
// Expand isfinite(x) => fabs(x) < inf
Expand Down