Guard shared Metal helper source across custom kernels - #65
Open
audreyt wants to merge 1 commit into
Open
Conversation
Core AI can compile multiple custom kernels and scalar specializations into one Metal library. Re-emitting identical helper_src functions then fails with a symbol redefinition unless every caller manually adds matching preprocessor guards. Wrap each rendered helper block in a deterministic content-hash guard. Hash after dtype substitution so distinct templated variants remain independent while identical rendered helpers deduplicate. Add coverage for shared helpers, dtype variants, and kernels without helpers.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
helper_srcblock in a deterministic content-hash include guardProblem
Core AI can compile several generated custom-kernel sources into one Metal library. Two
TorchMetalKernelinstances that reuse an unguarded static helper then emit the same function twice and fail at runtime with a Metal redefinition error. Callers currently have to coordinate matching#ifndefguards manually.A minimal two-kernel runtime repro on upstream
main(98eecc0) aborts with:With this change, the same asset loads and executes successfully (
x + 2).The hash covers the fully rendered helper text. Identical helpers therefore share one guard, while helpers specialized to
half,float, or other template dtypes receive distinct guards. Kernels without helpers are unchanged.Verification
uv run pytest tests/dsl/test_kernel_collisions.py -q: 10 passeduv run ruff check ...: PASSuv run ruff format --check ...: PASS