tests: enumerate CUDA devices for GPU-only system checks - #2916
Merged
Merged
Conversation
github-actions
Bot
added
the
cuda.core
isVoid
added
test
Contributor
Author
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Contributor
|
1 of 2 tasks
Andy-Jost
added a commit
that referenced
this pull request
Sep 23, 2026
* cuda.core: host-only memory resources need no CUDA context (#2773) Since 1.2.0, Buffer.from_handle binds a default-stream deallocation token to the current context for every owning memory resource. Memory the device cannot access has no stream ordering to preserve, so skip the binding when mr.is_device_accessible is False. Such buffers record no deallocation stream and never call the driver. Fixes #2769 Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com> (cherry picked from commit 3fef50b) * Add Windows-on-ARM support (cherry picked from commit 9d762dd) * Fix Windows on ARM build (cherry picked from commit 7cfdad7) * Exclude Python 3.10 on Windows-on-ARM (cherry picked from commit a6f6adc) * make platform tag explicit (#2832) (cherry picked from commit c280d47) * tests: enumerate CUDA devices for GPU-only system checks (#2916) (cherry picked from commit f59229e) * Prepare cuda.core v1.2.1 with Windows on ARM support Release branch off the cuda-core-v1.2.0 tag. Windows on ARM wheels need the CUDA 13.4 toolkit, so the build moves to CUDA 13.4.2 for every platform (ci/versions.yml and the pixi cuda-version pins). The WoA build jobs themselves are cherry-picked from main in the preceding commits. From the CUDA 13.4 integration on main (#2437, #2788, merged via #2789): - Map the ClocksEventReasons BOARD_LIMIT and RELIABILITY values that CUDA 13.4 added. Device.current_clock_event_reasons and Device.supported_clock_event_reasons raised ValueError for an unknown reason bit on devices that report them. - Map the NVIDIA DLA, vGameDev and NPU brand types instead of "Unknown". - Test updates: accept the "DLA-" UUID prefix, exclude the new 13.4 memory-location enum value from the coverage check, and cover Device.get_all_devices. The 1.2.1 release notes and the install page describe the new wheels: CUDA 13 only, Python 3.11 and newer, cuda-bindings 13.4.1 or newer. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * chore: refresh pixi.lock for cuda_core (#2830) Scheduled pixi update --no-install for this workspace only. Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> (cherry picked from commit 04d4753) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * chore: refresh pixi.lock for cuda_bindings (#2828) Scheduled pixi update --no-install for this workspace only. Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> (cherry picked from commit 7aa0648) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * ci: drop the Python 3.15 test jobs on the 1.2.1 release branch This branch builds with CUDA 13.4.2 but tests on 13.3.0, so the test jobs install published cuda-bindings 13.3.x to cover the real-world combination. No 13.3.x wheel exists for Python 3.15, so the three 3.15 and 3.15t test jobs fail at pip install before any test runs. 1.2.1 ships no 3.15 wheels, so nothing is lost; the 3.15 build jobs remain. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> --------- Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com> Co-authored-by: Michael Droettboom <mdroettboom@nvidia.com> Co-authored-by: Leo Fang <leof@nvidia.com> Co-authored-by: Michael Wang <13521008+isVoid@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
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
2 participants
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.
Description
cuda 1.2.0 contains some tests that queries using CUDA API, which would fail on N1X systems due to the non-CUDA devices. Previously we had fixes for some of the test to enumerate only on CUDADevice.get_all_devices(), but they don't cover all tests impacted. This PR fixes that.