Skip to content

[InstCombine] Fix flag propagation in foldSelectIntoOp #162003

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 5 commits into from
Oct 11, 2025
+172 −30
Merged

[InstCombine] Fix flag propagation in foldSelectIntoOp #162003

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
9716818
[InstCombine] Add pre-commit tests. NFC.
dtcxzyw Oct 5, 2025
b1fbcdf
[InstCombine] Fix flag propagation in `foldSelectIntoOp`
dtcxzyw Oct 5, 2025
3b28b6f
[InstCombine] Address review comments.
dtcxzyw Oct 10, 2025
dfd4bdf
[InstCombine] format
dtcxzyw Oct 10, 2025
f4af43a
Fix typo
dtcxzyw Oct 10, 2025
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
5 changes: 4 additions & 1 deletion llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -544,8 +544,11 @@ Instruction *InstCombinerImpl::foldSelectIntoOp(SelectInst &SI, Value *TrueVal,

Value *NewSel = Builder.CreateSelect(SI.getCondition(), Swapped ? C : OOp,
Swapped ? OOp : C, "", &SI);
if (isa<FPMathOperator>(&SI))
if (isa<FPMathOperator>(&SI)) {
cast<Instruction>(NewSel)->setFastMathFlags(FMF);
if (!TVI->hasNoInfs() && !FMF.noNaNs())
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you add a TODO that you can preserve this for operators which cannot introduce infs

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This fold works for fadd, fsub, and fmul. Unfortunately, there are no exceptions. All of these patterns have counterexamples to block the flag propagation. I have updated comments and proofs.

cast<Instruction>(NewSel)->setHasNoInfs(false);
}
NewSel->takeName(TVI);
BinaryOperator *BO =
BinaryOperator::Create(TVI->getOpcode(), FalseVal, NewSel);
Expand Down
44 changes: 44 additions & 0 deletions llvm/test/Transforms/InstCombine/select-binop-foldable-floating-point.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,50 @@ define float @select_fpclass_fadd(i1 %cond, float nofpclass(nan) %A, float %B) {
ret float %D
}

define float @select_fpclass_fadd_ninf1(i1 %cond, float nofpclass(nan) %A, float %B) {
; CHECK-LABEL: @select_fpclass_fadd_ninf1(
; CHECK-NEXT: [[C:%.*]] = select i1 [[COND:%.*]], float [[B:%.*]], float -0.000000e+00
; CHECK-NEXT: [[D:%.*]] = fadd float [[A:%.*]], [[C]]
; CHECK-NEXT: ret float [[D]]
;
%C = fadd ninf float %A, %B
%D = select i1 %cond, float %C, float %A
ret float %D
}

define float @select_fpclass_fadd_ninf2(i1 %cond, float nofpclass(nan) %A, float %B) {
; CHECK-LABEL: @select_fpclass_fadd_ninf2(
; CHECK-NEXT: [[C:%.*]] = select i1 [[COND:%.*]], float [[B:%.*]], float -0.000000e+00
; CHECK-NEXT: [[D:%.*]] = fadd float [[A:%.*]], [[C]]
; CHECK-NEXT: ret float [[D]]
;
%C = fadd float %A, %B
%D = select ninf i1 %cond, float %C, float %A
ret float %D
}

define float @select_fpclass_fadd_ninf3(i1 %cond, float nofpclass(nan) %A, float %B) {
; CHECK-LABEL: @select_fpclass_fadd_ninf3(
; CHECK-NEXT: [[C:%.*]] = select ninf i1 [[COND:%.*]], float [[B:%.*]], float -0.000000e+00
; CHECK-NEXT: [[D:%.*]] = fadd ninf float [[A:%.*]], [[C]]
; CHECK-NEXT: ret float [[D]]
;
%C = fadd ninf float %A, %B
%D = select ninf i1 %cond, float %C, float %A
ret float %D
}

define float @select_fpclass_fadd_nnan_ninf(i1 %cond, float nofpclass(nan) %A, float %B) {
; CHECK-LABEL: @select_fpclass_fadd_nnan_ninf(
; CHECK-NEXT: [[C:%.*]] = select nnan ninf i1 [[COND:%.*]], float [[B:%.*]], float -0.000000e+00
; CHECK-NEXT: [[D:%.*]] = fadd float [[A:%.*]], [[C]]
; CHECK-NEXT: ret float [[D]]
;
%C = fadd float %A, %B
%D = select nnan ninf i1 %cond, float %C, float %A
ret float %D
}

define float @select_nnan_fadd(i1 %cond, float %A, float %B) {
; CHECK-LABEL: @select_nnan_fadd(
; CHECK-NEXT: [[C:%.*]] = select nnan i1 [[COND:%.*]], float [[B:%.*]], float -0.000000e+00
Expand Down