Repository navigation
ci: disable sccache preprocessor cache mode - #141
Merged
atmnp merged 3 commits intoJun 24, 2026
Merged
Conversation
The preprocessor cache mode produces minimal/zero hits across CI runs because pip's per-run build-isolation overlay path leaks into compiler flags. See NVIDIA/cuda-python#2220 for the full investigation. LLVM 7 / ToT builds are unaffected — they use the S3 backend configured via ci/setup-sccache.sh and don't go through cibuildwheel.
Contributor
Member
Author
|
/ok to test 1708603 |
This comment has been minimized.
This comment has been minimized.
leofang
marked this pull request as ready for review
June 16, 2026 05:08
atmnp
approved these changes
Jun 23, 2026
Collaborator
|
/ok to test e7372c8 |
atmnp
enabled auto-merge (squash)
June 23, 2026 23:47
|
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Mirrors NVIDIA/cuda-python#2221 — see NVIDIA/cuda-python#2220 for the full investigation.
Same root cause applies here: pip's per-run
/tmp/pip-build-env-<RANDOM>/...overlay path leaks into-Iflags during the cibuildwheel build, so sccache's preprocessor cache key is randomized every run. In cuSIMT the symptom is milder than in cuda-python (CMake doesn't propagate every overlay path through to gcc — last main run had 9 hits / 25 misses) but it's still net overhead.LLVM 7 / modern LLVM builds (
.github/workflows/build-llvm.yml+ci/setup-sccache.sh) are unaffected: they use the S3 backend with[cache.s3.preprocessor_cache_mode]and don't go through cibuildwheel.