fix: make Linux OpenBLAS an explicit backend choice - #99
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Linux builds previously selected system BLAS based on whatever pkg-config found on the host, including incompatible generic BLAS / ILP64 alternatives. Default native builds now use Rust kernels;
system-openblasexplicitly enables the existing LP64 OpenBLAS branches and fails clearly when prerequisites are missing. Darwin implementation and scoreboard floors are unchanged.Add artifact linkage and numerical checks for both choices, install OpenBLAS in all-features CI lanes, and keep release assets and scheduled Linux quality evidence on the Rust default. Document the feature, prerequisites and backend reporting; remove the unused default Python pkg-config dependency.
Validation: Linux artifact/missing-prerequisite checks; 58/59 kernel tests; 943 CLI/FFI/OpenBLAS library tests and 875 local-only tests; clippy for default and all-features; dependency and standalone-lockfile gates; formatting and documentation links. Local semver-checks completed (the existing 0.21 → 0.22 major-change policy skips compatibility lints). CI exposed an additional OpenBLAS prerequisite in its semver-checks job, now installed explicitly.
Same-host evidence is in benchmarks/results/linux-blas-selection-2026-09-24/. Rust default matches the baseline per-file quality on Vox-3, fixed Vox-10 and AMI-1. OpenBLAS changes clustering on the larger subsets. Unrelated CPU load makes throughput unsuitable for performance certification; RSS observations and repeated-run ranges are preserved without claiming memory improvements. Full-corpus performance qualification was not completed. Darwin scoreboard CI passed; no Darwin implementation or locked limits changed.