Use convergent attribute to funcs for GPU targets#153995
Merged
rust-bors[bot] merged 1 commit intorust-lang:mainfrom Apr 8, 2026
Merged
Use convergent attribute to funcs for GPU targets#153995rust-bors[bot] merged 1 commit intorust-lang:mainfrom
rust-bors[bot] merged 1 commit intorust-lang:mainfrom
Conversation
On targets with convergent operations, we need to add the convergent attribute to all functions that run convergent operations. Following clang, we can conservatively apply the attribute to all functions when compiling for such a target and rely on LLVM optimizing away the attribute in cases where it is not necessary. This affects the amdgpu and nvptx targets.
rustbot
added
A-LLVM
nnethercote
approved these changes
Apr 8, 2026
Contributor
|
@bors r+ |
Contributor
rust-bors
Bot
added
S-waiting-on-bors
JonathanBrouwer
added a commit
to JonathanBrouwer/rust
that referenced
this pull request
Apr 8, 2026
…hercote Use convergent attribute to funcs for GPU targets On targets with convergent operations, we need to add the convergent attribute to all functions that run convergent operations. Following clang, we can conservatively apply the attribute to all functions when compiling for such a target and rely on LLVM optimizing away the attribute in cases where it is not necessary. This affects the amdgpu and nvptx targets. cc @kjetilkjeka, @kulst for nvptx cc @ZuseZ4 r? @nnethercote, as you already reviewed this in the other PR Split out from rust-lang#149637, the part here should be uncontroversial.
This was referenced Apr 8, 2026
rust-bors Bot
pushed a commit
that referenced
this pull request
Apr 8, 2026
…uwer Rollup of 15 pull requests Successful merges: - #153995 (Use convergent attribute to funcs for GPU targets) - #154184 (stabilize s390x vector registers) - #151898 (constify DoubleEndedIterator) - #154235 (remove unnecessary variables and delimiter check) - #154473 (move borrow checker tests) - #154745 (Replace span_look_ahead with span_followed_by) - #154778 (make field representing types invariant over the base type) - #154867 (Fix private fields diagnostics and improve error messages) - #154879 (Don't store `pattern_ty` in `TestableCase`) - #154910 (Suppress `unreachable_code` lint in `derive(PartialEq, Clone)`) - #154923 (Fix ICE in next-solver dyn-compatibility check) - #154934 (Add getters for `rustc_pattern_analysis::constructor::Slice` fields) - #154938 (match exhaustiveness: Show the guard exhaustivity note only when it's the guards alone that cause non-exhaustiveness) - #154961 (Use derived impl for `GappedRange` subdiagnostic) - #154980 (rustc-dev-guide subtree update)
rust-timer
added a commit
that referenced
this pull request
Apr 8, 2026
Rollup merge of #153995 - Flakebi:gpu-use-convergent, r=nnethercote Use convergent attribute to funcs for GPU targets On targets with convergent operations, we need to add the convergent attribute to all functions that run convergent operations. Following clang, we can conservatively apply the attribute to all functions when compiling for such a target and rely on LLVM optimizing away the attribute in cases where it is not necessary. This affects the amdgpu and nvptx targets. cc @kjetilkjeka, @kulst for nvptx cc @ZuseZ4 r? @nnethercote, as you already reviewed this in the other PR Split out from #149637, the part here should be uncontroversial.
26 tasks
github-actions Bot
pushed a commit
to rust-lang/rustc-dev-guide
that referenced
this pull request
Apr 13, 2026
…uwer Rollup of 15 pull requests Successful merges: - rust-lang/rust#153995 (Use convergent attribute to funcs for GPU targets) - rust-lang/rust#154184 (stabilize s390x vector registers) - rust-lang/rust#151898 (constify DoubleEndedIterator) - rust-lang/rust#154235 (remove unnecessary variables and delimiter check) - rust-lang/rust#154473 (move borrow checker tests) - rust-lang/rust#154745 (Replace span_look_ahead with span_followed_by) - rust-lang/rust#154778 (make field representing types invariant over the base type) - rust-lang/rust#154867 (Fix private fields diagnostics and improve error messages) - rust-lang/rust#154879 (Don't store `pattern_ty` in `TestableCase`) - rust-lang/rust#154910 (Suppress `unreachable_code` lint in `derive(PartialEq, Clone)`) - rust-lang/rust#154923 (Fix ICE in next-solver dyn-compatibility check) - rust-lang/rust#154934 (Add getters for `rustc_pattern_analysis::constructor::Slice` fields) - rust-lang/rust#154938 (match exhaustiveness: Show the guard exhaustivity note only when it's the guards alone that cause non-exhaustiveness) - rust-lang/rust#154961 (Use derived impl for `GappedRange` subdiagnostic) - rust-lang/rust#154980 (rustc-dev-guide subtree update)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
On targets with convergent operations, we need to add the convergent attribute to all functions that run convergent operations. Following clang, we can conservatively apply the attribute to all functions when compiling for such a target and rely on LLVM optimizing away the attribute in cases where it is not necessary.
This affects the amdgpu and nvptx targets.
cc @kjetilkjeka, @kulst for nvptx
cc @ZuseZ4
r? @nnethercote, as you already reviewed this in the other PR
Split out from #149637, the part here should be uncontroversial.