shrink GPU wheel under PyPI 320 MiB limit#262
Merged
jameslehoux merged 1 commit intomasterfrom Apr 24, 2026
Merged
Conversation
Previous build succeeded and produced a valid wheel but PyPI rejected the upload with HTTP 400 "File too large. Limit for project 'openimpala-cuda' is 320 MB." Two levers bring it under the cap: 1) Trim CUDA architectures from 7 to 5, keep SASS-only Before: 60;70;75;80;86;89;90 (each arch compiles SASS + PTX ≈ 7×2 fat-binary slices per kernel). After: 70-real;75-real;80-real;89-real;90-real;90-virtual (5 SASS slices + one PTX for forward-compat / JIT on any new arch). Drops Pascal P100 (60 — obsolete, 2016) and Ampere consumer RTX 30 (86 — JITs fine from 80 PTX via 90-virtual). Keeps the actually-used cloud/HPC GPUs: V100 (70), T4 (75), A100 (80), L4 (89), H100 (90). The -real suffix suppresses PTX emission per arch, and the single 90-virtual entry gives future-arch JIT without embedding 6 redundant PTX slices. 2) auditwheel repair --strip Strips ELF debug symbols from every bundled .so in the wheel. No effect on runtime behaviour; typical 20-40% size reduction on C++ scientific libraries. Cache key bumped to reflect the arch change so the next build can't restore the oversized deps.tar.gz. https://claude.ai/code/session_011dJ5Bwq4Tnr8wxH597XJFf
Code Coverage ReportGenerated by CI — coverage data from gcovr |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Previous build succeeded and produced a valid wheel but PyPI rejected the upload with HTTP 400 "File too large. Limit for project 'openimpala-cuda' is 320 MB." Two levers bring it under the cap:
Trim CUDA architectures from 7 to 5, keep SASS-only
Before: 60;70;75;80;86;89;90 (each arch compiles SASS + PTX ≈ 7×2
fat-binary slices per kernel).
After: 70-real;75-real;80-real;89-real;90-real;90-virtual (5 SASS
slices + one PTX for forward-compat / JIT on any new arch).
Drops Pascal P100 (60 — obsolete, 2016) and Ampere consumer RTX 30
(86 — JITs fine from 80 PTX via 90-virtual). Keeps the actually-used
cloud/HPC GPUs: V100 (70), T4 (75), A100 (80), L4 (89), H100 (90).
The -real suffix suppresses PTX emission per arch, and the single
90-virtual entry gives future-arch JIT without embedding 6 redundant
PTX slices.
auditwheel repair --strip
Strips ELF debug symbols from every bundled .so in the wheel. No
effect on runtime behaviour; typical 20-40% size reduction on C++
scientific libraries.
Cache key bumped to reflect the arch change so the next build can't restore the oversized deps.tar.gz.