Skip to content

[ValueTracking] Support non-constant idx for computeKnownFPClass of insertelement #87708

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

Closed
wants to merge 2 commits into from
Closed

[ValueTracking] Support non-constant idx for computeKnownFPClass of insertelement #87708

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4fcbb91
[ValueTracking] Add tests for non-constant idx for fpclass of `insert…
goldsteinn Apr 4, 2024
190d61f
[ValueTracking] Support non-constant idx for `computeKnownFPClass` of…
goldsteinn Apr 4, 2024
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
19 changes: 10 additions & 9 deletions llvm/lib/Analysis/ValueTracking.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5355,14 +5355,17 @@ void computeKnownFPClass(const Value *V, const APInt &DemandedElts,
const Value *Vec = Op->getOperand(0);
const Value *Elt = Op->getOperand(1);
auto *CIdx = dyn_cast<ConstantInt>(Op->getOperand(2));
// Early out if the index is non-constant or out-of-range.
unsigned NumElts = DemandedElts.getBitWidth();
if (!CIdx || CIdx->getValue().uge(NumElts))
return;
APInt DemandedVecElts = DemandedElts;
bool NeedsElt = true;
// If we know the index we are inserting too, clear it from Vec check.
Copy link
Contributor

Choose a reason for hiding this comment

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

Suggested change
// If we know the index we are inserting too, clear it from Vec check.
// If we know the index we are inserting to, clear it from Vec check.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Will update before pushing.

if (CIdx && CIdx->getValue().ult(NumElts)) {
DemandedVecElts.clearBit(CIdx->getZExtValue());
NeedsElt = DemandedElts[CIdx->getZExtValue()];
}

unsigned EltIdx = CIdx->getZExtValue();
// Do we demand the inserted element?
if (DemandedElts[EltIdx]) {
if (NeedsElt) {
computeKnownFPClass(Elt, Known, InterestedClasses, Depth + 1, Q);
// If we don't know any bits, early out.
if (Known.isUnknown())
Expand All @@ -5371,10 +5374,8 @@ void computeKnownFPClass(const Value *V, const APInt &DemandedElts,
Known.KnownFPClasses = fcNone;
}

// We don't need the base vector element that has been inserted.
APInt DemandedVecElts = DemandedElts;
DemandedVecElts.clearBit(EltIdx);
if (!!DemandedVecElts) {
// Do we need anymore elements from Vec?
if (!DemandedVecElts.isZero()) {
KnownFPClass Known2;
computeKnownFPClass(Vec, DemandedVecElts, InterestedClasses, Known2,
Depth + 1, Q);
Expand Down
21 changes: 20 additions & 1 deletion llvm/test/Transforms/Attributor/nofpclass.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1513,6 +1513,25 @@ define <4 x float> @insertelement_constant_chain() {
ret <4 x float> %ins.3
}

define <4 x float> @insertelement_non_constant_chain(i32 %idx) {
; CHECK: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
; CHECK-LABEL: define nofpclass(nan inf nzero sub) <4 x float> @insertelement_non_constant_chain
; CHECK-SAME: (i32 [[IDX:%.*]]) #[[ATTR3]] {
; CHECK-NEXT: [[INS_0:%.*]] = insertelement <4 x float> poison, float 1.000000e+00, i32 0
; CHECK-NEXT: [[INS_1:%.*]] = insertelement <4 x float> [[INS_0]], float 0.000000e+00, i32 1
; CHECK-NEXT: [[INS_2:%.*]] = insertelement <4 x float> [[INS_1]], float -9.000000e+00, i32 2
; CHECK-NEXT: [[INS_4:%.*]] = insertelement <4 x float> [[INS_2]], float 3.000000e+00, i32 3
; CHECK-NEXT: [[INS_3:%.*]] = insertelement <4 x float> [[INS_2]], float 4.000000e+00, i32 [[IDX]]
; CHECK-NEXT: ret <4 x float> [[INS_3]]
;
%ins.0 = insertelement <4 x float> poison, float 1.0, i32 0
%ins.1 = insertelement <4 x float> %ins.0, float 0.0, i32 1
%ins.2 = insertelement <4 x float> %ins.1, float -9.0, i32 2
%ins.3 = insertelement <4 x float> %ins.2, float 3.0, i32 3
%ins.4 = insertelement <4 x float> %ins.2, float 4.0, i32 %idx
ret <4 x float> %ins.4
}

define <vscale x 4 x float> @insertelement_scalable_constant_chain() {
; CHECK: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
; CHECK-LABEL: define <vscale x 4 x float> @insertelement_scalable_constant_chain
Expand Down Expand Up @@ -1582,7 +1601,7 @@ define float @insertelement_extractelement_unknown(<4 x float> nofpclass(zero) %

define <4 x float> @insertelement_index_oob_chain() {
; CHECK: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none)
; CHECK-LABEL: define <4 x float> @insertelement_index_oob_chain
; CHECK-LABEL: define nofpclass(nan ninf nzero sub norm) <4 x float> @insertelement_index_oob_chain
; CHECK-SAME: () #[[ATTR3]] {
; CHECK-NEXT: [[INSERT:%.*]] = insertelement <4 x float> zeroinitializer, float 0x7FF0000000000000, i32 4
; CHECK-NEXT: ret <4 x float> [[INSERT]]
Expand Down