Apply LLVM customizations for Python wheel build #2504
Merged
Conversation
This file contains 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
Signed-off-by: Ben Howe <bhowe@nvidia.com>
bmhowe23
force-pushed
the
pr-apply-llvm-patches-for-manylinux
branch
from
675e685 to
51e7147
Compare
Signed-off-by: Ben Howe <bhowe@nvidia.com>
|
CUDA Quantum Docs Bot: A preview of the documentation can be found here. |
bettinaheim
approved these changes
Jan 13, 2025
This was referenced Jan 14, 2025
bmhowe23
added a commit
to NVIDIA/cudaqx
that referenced
this pull request
Jan 21, 2025
## Primary Changes * Update build_wheels.yaml to test CUDA-QX wheels after building them. * Update build_wheels.yaml to build CUDA-Q wheels rather than using released PyPi CUDA-Q wheels. This was necessary because the current 0.9.1 CUDA-Q wheels have an ARM bug that were impacting our ability to do testing. This was fixed in NVIDIA/cuda-quantum#2504, which has not been released yet. * Fix RPATH of decoder plugin example (thanks for the suggestion, @melody-ren) to fix a wheel packaging problem where `auditwheel` as duplicating (aka grafting) the `libcudaq-qec.so` file, causing `extension_point` registration anomalies and decoder plugin lookup failures. ## Secondary Changes * In .github/actions/get-cudaq-version/action.yaml, skip `jq` install if it is already installed * Add get-cudaq-wheels action; currently only invoked on manually commanded build_wheels.yaml workflow but we may decide to invoke this on a per-PR basis in the future. * Change retention period of all uploaded artifacts to 14 days * Update `getCUDAQXLibraryPath` to work for both QEC and Solvers * Change decoder-plugins path from a build-time constant (which will not work for wheels) to use `getCUDAQXLibraryPath` instead. * Change use of shared_ptr to unique_ptr in decoder plugin handling * Move decoder plugin helper functions into namespaces rather than global namespace * Change cudaqx wheels to be depending on cuda-quantum >= 0.9 rather than ~= 0.9.0 for better futureproofing of our released wheels Sample run result w/ custom CUDA-Q wheels: https://github.com/NVIDIA/cudaqx/actions/runs/12855925162 Sample run result w/ released PyPI CUDA-Q wheels: https://github.com/NVIDIA/cudaqx/actions/runs/12856394702 --------- Signed-off-by: Ben Howe <bhowe@nvidia.com>
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.
This PR fixes an LLVM discrepancy in our Docker images vs our Python wheels. This should fix #1421 and #1799 for our Python wheels. (The Docker images were already correct.)
Just to complicate things even more than they already are, our 0.9.*, the Python wheels were already showing correct behavior for the exact test case in #1799. The exact reason is spelled out in the details of #1799. However, the above table describes the theoretical situation that we could've been vulnerable to the region simplification bug in other conditions.