Fix cuda_bindings conjugate_gradient_multi_block_cg.py example#1922
Merged
leofang merged 2 commits intoNVIDIA:mainfrom Apr 16, 2026
Merged
Fix cuda_bindings conjugate_gradient_multi_block_cg.py example#1922leofang merged 2 commits intoNVIDIA:mainfrom
leofang merged 2 commits intoNVIDIA:mainfrom
Conversation
Remove the unconditional NVRTC waiver from the renamed example so CI can exercise its real execution path again. While re-enabling it, replace the standalone pytest.skip() checks with requirement_not_met() and simplify the platform gating. The waived code path had been hiding several Python-side runtime bugs that required these fixes: - replaced the invalid C-style %d f-string formatting - fixed gen_tridiag() variable shadowing so the CSR row-offset array is actually populated - passed the computed dynamic shared-memory size into cuLaunchCooperativeKernel() and made that size integer-valued - stopped overwriting managed-memory pointer variables with loop indices before kernel launch and cleanup - cached the residual before freeing dot_result, which removed the teardown segfault Made-with: Cursor
QNX is an operating system rather than a machine architecture, so checking platform.machine() can miss the requirement_not_met() path on QNX hosts. Use platform.system() so the example is waived consistently on that platform. Made-with: Cursor
rwgk
added
bug
Contributor
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
Contributor
Author
|
/ok to test |
This comment has been minimized.
This comment has been minimized.
Contributor
Author
|
Cursor-generated analysis of CI logs for commit be43e41:
I unpacked and inspected the combined CI logs locally under For completeness, these are the log archives the combined directory came from: I then scanned all Summary
Skip reasonAll 16 skips appear to be consistent with the same grouped pytest summary reason: I did not see any skips for this example caused by its own runtime gates At a high level, this looks as expected:
Passed in
Skipped in
|
mdboom
approved these changes
Apr 16, 2026
Contributor
mdboom
left a comment
mdboom
left a comment
There was a problem hiding this comment.
LGTM.
The pointer arithmetic in this example should probably be removed someday, but that was pre-existing.
|
mdboom
pushed a commit
to mdboom/cuda-python
that referenced
this pull request
Apr 20, 2026
…A#1922) * fix: re-enable and fix the conjugate gradient multi-block example Remove the unconditional NVRTC waiver from the renamed example so CI can exercise its real execution path again. While re-enabling it, replace the standalone pytest.skip() checks with requirement_not_met() and simplify the platform gating. The waived code path had been hiding several Python-side runtime bugs that required these fixes: - replaced the invalid C-style %d f-string formatting - fixed gen_tridiag() variable shadowing so the CSR row-offset array is actually populated - passed the computed dynamic shared-memory size into cuLaunchCooperativeKernel() and made that size integer-valued - stopped overwriting managed-memory pointer variables with loop indices before kernel launch and cleanup - cached the residual before freeing dot_result, which removed the teardown segfault Made-with: Cursor * fix: check the QNX example gate via platform.system() QNX is an operating system rather than a machine architecture, so checking platform.machine() can miss the requirement_not_met() path on QNX hosts. Use platform.system() so the example is waived consistently on that platform. Made-with: Cursor
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
3 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.
This PR started out as a humble attempt to eliminate the only remaining
pytest.skip()in our examples (similar to #1861), but it turned into a bug-fix pass once the example actually began running.cuda_bindings/examples/4_CUDA_Libraries/conjugate_gradient_multi_block_cg.pyby removing its unconditional NVRTC waiver (which dates all the way back to the initial cuda-python commit 4a36f83) and replacing standalonepytest.skip()calls withrequirement_not_met()%dformatting inside an f-stringgen_tridiag()variable shadowing that broke CSR constructiondot_result, which caused a teardown segfaultcuda_bindings/examples/3_CUDA_Features/global_to_shmem_async_copy.pyby checkingplatform.system() == "QNX"instead ofplatform.machine() == "qnx"— QNX is an operating system rather than a machine architecture, sochecking platform.machine() == "qnx"does not detect QNX hosts.