Skip to content

[Matrix] Add tests identifying GVN and DSE opportunities for matrix store / load intrinsics #163573

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 2 commits into from
Oct 22, 2025
Merged

[Matrix] Add tests identifying GVN and DSE opportunities for matrix store / load intrinsics #163573

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
7de09f4
[Matrix] Add tests identifying GVN and DSE opportunities for matrix s…
cofibrant Oct 13, 2025
31e47b6
Update tests
cofibrant Oct 22, 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
30 changes: 30 additions & 0 deletions llvm/test/Analysis/BasicAA/matrix-intrinsics.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
; RUN: opt %s -aa-pipeline=basic-aa -passes=aa-eval -print-all-alias-modref-info -disable-output 2>&1 | FileCheck %s

; BasicAA should prove that loads from sufficiently large static offsets
; don't overlap with matrix loads with a statically known size.

define <8 x double> @non_overlapping_strided_load(ptr %src) {
; CHECK-LABEL: Function: non_overlapping_strided_load:
; Just Ref: %l = call <8 x double> @llvm.matrix.column.major.load.v8f64.i32(ptr %src.offset, i32 8, i1 false, i32 4, i32 2) <-> call void @llvm.matrix.column.major.store.v8f64.i32(<8 x double> %l, ptr %src, i32 8, i1 false, i32 4, i32 2)
; Just Mod: call void @llvm.matrix.column.major.store.v8f64.i32(<8 x double> %l, ptr %src, i32 8, i1 false, i32 4, i32 2) <-> %l = call <8 x double> @llvm.matrix.column.major.load.v8f64.i32(ptr %src.offset, i32 8, i1 false, i32 4, i32 2)
entry:
%src.offset = getelementptr inbounds double, ptr %src, i32 12
%l = call <8 x double> @llvm.matrix.column.major.load.v8f64.i32(ptr %src.offset, i32 8, i1 false, i32 4, i32 2)
call void @llvm.matrix.column.major.store(<8 x double> %l, ptr %src, i32 8, i1 false, i32 4, i32 2)
ret <8 x double> %l
}

Copy link
Contributor

Choose a reason for hiding this comment

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

for DSE, we probably also want tests that cover the interaction between matrix intrinsics and normal loads/stores:

  • normal store is dead because the last read was from a matrix intrinsic that happened before the store
  • normal store is not dead because the last read by a matrix intrinsic happens after the store
  • matrix intrinsic store is dead because the last read was from a normal store that happened before the store
  • matrix intrinsic store is not dead because the last read from a normal store happens after the store

define <8 x double> @overlapping_strided_load(ptr %src) {
; CHECK-LABEL: Function: overlapping_strided_load:
; CHECK: Just Ref: %l = call <8 x double> @llvm.matrix.column.major.load.v8f64.i32(ptr %src.offset, i32 8, i1 false, i32 4, i32 2) <-> call void @llvm.matrix.column.major.store.v8f64.i32(<8 x double> %l, ptr %src, i32 8, i1 false, i32 4, i32 2)
; CHECK: Just Mod: call void @llvm.matrix.column.major.store.v8f64.i32(<8 x double> %l, ptr %src, i32 8, i1 false, i32 4, i32 2) <-> %l = call <8 x double> @llvm.matrix.column.major.load.v8f64.i32(ptr %src.offset, i32 8, i1 false, i32 4, i32 2)
;
entry:
%src.offset = getelementptr inbounds double, ptr %src, i32 11
%l = call <8 x double> @llvm.matrix.column.major.load.v8f64.i32(ptr %src.offset, i32 8, i1 false, i32 4, i32 2)
call void @llvm.matrix.column.major.store(<8 x double> %l, ptr %src, i32 8, i1 false, i32 4, i32 2)
ret <8 x double> %l
}

declare <8 x double> @llvm.matrix.column.major.load.v8f64.i32(ptr, i32, i1, i32, i32)
declare void @llvm.matrix.column.major.store.v8f64.i32(<8 x double>, ptr, i32, i1, i32, i32)
Loading