[isVoid]

Description

Closes #3167.

Use CU_DEVICE_ATTRIBUTE_INTEGRATED instead of the Tegra model-name helper to determine the minimum expected BAR size in test_get_bar_size_in_kb.

GB10 is an integrated GPU and successfully reports zero BAR size, but get_tegra_kind() does not recognize it. The existing test therefore incorrectly requires a positive size.

The test still runs on GB10 and asserts an integer, nonnegative result. Discrete GPUs still require a positive BAR size. The device-attribute query must return CUDA_SUCCESS; existing cuFile error handling is unchanged. No production binding or CI workflow changes.

This is separate from #2299, which concerned cuFile error-status handling on GH200 rather than rejecting a successfully returned zero.

Validation

• pytest -v -rs cuda_bindings/tests/test_cufile.py::test_get_bar_size_in_kb: 1 passed on an NVIDIA A30 (discrete GPU), driver 595.84.
• Eight mocked cases checked integrated/discrete GPUs with zero, positive, negative, or non-integer BAR values and a failing CUDA attribute query. Zero passes only for integrated GPUs; invalid results still fail.
• Ruff check/format, applicable pre-commit hooks, and git diff --check passed.
• Local Python 3.13 used the existing cuda-bindings 13.4.2.dev243+gf9de5797e wheel environment. No new source build or GB10 rerun for this public branch; integrated behavior was checked with mocks locally.
• Five existing unknown-thread_unsafe-marker warnings were emitted by the local environment, which lacks pytest-run-parallel.

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes. (Test-only change; no public API changes.)

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Configuration used: Repository: NVIDIA/cuda-python/.coderabbit.yaml
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 37f3055d-6cec-41c8-80ac-dc417456011b

📥 Commits

Reviewing files that changed from the base of the PR and between bdb6910 and 3ff8bda.

📒 Files selected for processing: 1

• cuda_bindings_13/tests/test_cufile.py

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 10 remain after this review.

📝 Summary

Summary by CodeRabbit

• Bug Fixes
  • Updated BAR size validation to correctly account for integrated and discrete GPUs.
  • Removed the previous platform-specific exception in favor of CUDA’s GPU classification.

The cuFile BAR-size test now queries CUDA’s integrated-device attribute. It permits a minimum BAR size of zero for integrated GPUs and retains a minimum of one KB for discrete GPUs.

Changes: cuFile BAR-size test

Layer / File(s)	Summary
Select BAR-size minimum cuda_bindings_13/tests/test_cufile.py	The test uses the integrated-device attribute to select a minimum BAR size of zero for integrated GPUs or one KB for discrete GPUs. It asserts that the attribute query succeeds.

Assessment against linked issues

Objective	Addressed	Explanation
[3167] Use CU_DEVICE_ATTRIBUTE_INTEGRATED to allow a zero BAR size for integrated GPUs, retain the positive-size requirement for discrete GPUs, and assert that the attribute query succeeds.	✅

Suggested reviewers: rwgk

Priority: ⬇️ Low

Change: Bug fix · Severity of issue fixed: Low

Merge Risk: ⚪ Minimal · up to 3ff8b

The BAR-size test change is consistent with the stated integrated-GPU behavior and has no unresolved merge-blocking risk.

✨ Finishing Touches: 💡 1

🛠️ Fix failing CI checks: 💡

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Doc Preview CI
Preview removed because the pull request was closed or merged.