Skip to content

[BasicAA] Handle scalable type sizes with constant offsets #80445

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.

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
Feb 5, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
46 changes: 28 additions & 18 deletions llvm/lib/Analysis/BasicAliasAnalysis.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -1113,10 +1113,6 @@ AliasResult BasicAAResult::aliasGEP(
return BaseAlias;
}

// Bail on analysing scalable LocationSize
if (V1Size.isScalable() || V2Size.isScalable())
return AliasResult::MayAlias;

// If there is a constant difference between the pointers, but the difference
// is less than the size of the associated memory object, then we know
// that the objects are partially overlapping. If the difference is
Expand Down Expand Up @@ -1146,24 +1142,38 @@ AliasResult BasicAAResult::aliasGEP(
if (!VLeftSize.hasValue())
return AliasResult::MayAlias;

const uint64_t LSize = VLeftSize.getValue();
if (Off.ult(LSize)) {
// Conservatively drop processing if a phi was visited and/or offset is
// too big.
AliasResult AR = AliasResult::PartialAlias;
if (VRightSize.hasValue() && Off.ule(INT32_MAX) &&
(Off + VRightSize.getValue()).ule(LSize)) {
// Memory referenced by right pointer is nested. Save the offset in
// cache. Note that originally offset estimated as GEP1-V2, but
// AliasResult contains the shift that represents GEP1+Offset=V2.
AR.setOffset(-Off.getSExtValue());
AR.swap(Swapped);
const TypeSize LSize = VLeftSize.getValue();
if (!LSize.isScalable()) {
if (Off.ult(LSize)) {
// Conservatively drop processing if a phi was visited and/or offset is
// too big.
AliasResult AR = AliasResult::PartialAlias;
if (VRightSize.hasValue() && !VRightSize.isScalable() &&
Off.ule(INT32_MAX) && (Off + VRightSize.getValue()).ule(LSize)) {
// Memory referenced by right pointer is nested. Save the offset in
// cache. Note that originally offset estimated as GEP1-V2, but
// AliasResult contains the shift that represents GEP1+Offset=V2.
AR.setOffset(-Off.getSExtValue());
AR.swap(Swapped);
}
return AR;
}
return AR;
return AliasResult::NoAlias;
} else {
// We can use the getVScaleRange to prove that Off >= (CR.upper * LSize).
ConstantRange CR = getVScaleRange(&F, Off.getBitWidth());
bool Overflow;
APInt UpperRange = CR.getUnsignedMax().umul_ov(
APInt(Off.getBitWidth(), LSize.getKnownMinValue()), Overflow);
if (!Overflow && Off.uge(UpperRange))
return AliasResult::NoAlias;
}
return AliasResult::NoAlias;
}

// Bail on analysing scalable LocationSize
if (V1Size.isScalable() || V2Size.isScalable())
return AliasResult::MayAlias;

// We need to know both acess sizes for all the following heuristics.
if (!V1Size.hasValue() || !V2Size.hasValue())
return AliasResult::MayAlias;
Expand Down
3 changes: 2 additions & 1 deletion llvm/test/Analysis/AliasSet/memloc-vscale.ll
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,8 @@ define void @ss2(ptr %p) {
ret void
}
; CHECK-LABEL: Alias sets for function 'son':
; CHECK: AliasSet[{{.*}}, 2] may alias, Mod Memory locations: (ptr %g, LocationSize::precise(vscale x 16)), (ptr %p, LocationSize::precise(8))
; CHECK: AliasSet[{{.*}}, 1] must alias, Mod Memory locations: (ptr %g, LocationSize::precise(vscale x 16))
; CHECK: AliasSet[{{.*}}, 1] must alias, Mod Memory locations: (ptr %p, LocationSize::precise(8))
define void @son(ptr %p) {
%g = getelementptr i8, ptr %p, i64 8
store <vscale x 2 x i64> zeroinitializer, ptr %g, align 2
Expand Down
14 changes: 7 additions & 7 deletions llvm/test/Analysis/BasicAA/vscale.ll
Original file line number Diff line number Diff line change
Expand Up @@ -292,7 +292,7 @@ define void @gep_pos_scalable(ptr %p) vscale_range(1,16) {
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %m16, <4 x i32>* %p
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %m16, <vscale x 4 x i32>* %vm16
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %m16, <4 x i32>* %vm16
; CHECK-DAG: MayAlias: <4 x i32>* %m16, <vscale x 4 x i32>* %p
; CHECK-DAG: NoAlias: <4 x i32>* %m16, <vscale x 4 x i32>* %p
; CHECK-DAG: NoAlias: <4 x i32>* %m16, <4 x i32>* %p
; CHECK-DAG: MayAlias: <4 x i32>* %m16, <vscale x 4 x i32>* %vm16
; CHECK-DAG: MayAlias: <4 x i32>* %m16, <4 x i32>* %vm16
Expand Down Expand Up @@ -453,14 +453,14 @@ define void @gep_2048(ptr %p) {
; CHECK-LABEL: gep_2048_vscalerange
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %off255, <vscale x 4 x i32>* %p
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %noff255, <vscale x 4 x i32>* %p
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %noff255, <vscale x 4 x i32>* %off255
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %off256, <vscale x 4 x i32>* %p
; CHECK-DAG: NoAlias: <vscale x 4 x i32>* %noff255, <vscale x 4 x i32>* %off255
; CHECK-DAG: NoAlias: <vscale x 4 x i32>* %off256, <vscale x 4 x i32>* %p
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %off255, <vscale x 4 x i32>* %off256
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %noff255, <vscale x 4 x i32>* %off256
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %noff256, <vscale x 4 x i32>* %p
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %noff256, <vscale x 4 x i32>* %off255
; CHECK-DAG: NoAlias: <vscale x 4 x i32>* %noff255, <vscale x 4 x i32>* %off256
; CHECK-DAG: NoAlias: <vscale x 4 x i32>* %noff256, <vscale x 4 x i32>* %p
; CHECK-DAG: NoAlias: <vscale x 4 x i32>* %noff256, <vscale x 4 x i32>* %off255
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %noff255, <vscale x 4 x i32>* %noff256
; CHECK-DAG: MayAlias: <vscale x 4 x i32>* %noff256, <vscale x 4 x i32>* %off256
; CHECK-DAG: NoAlias: <vscale x 4 x i32>* %noff256, <vscale x 4 x i32>* %off256
define void @gep_2048_vscalerange(ptr %p) vscale_range(1,16) {
%off255 = getelementptr i8, ptr %p, i64 255
%noff255 = getelementptr i8, ptr %p, i64 -255
Expand Down