[martindurant]

Description

Solves parts of #2910

In VirtualMemoryResourceOptions, •

• Reject allocation_type="managed" with ValueError.
• Default Windows handles to None, preserving Linux’s POSIX descriptor default.

These are points 3 and 6 of the linked issue.

Checklist

• I have read the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Summary

Summary by CodeRabbit

• Bug Fixes
  • Virtual memory allocations now reject the unsupported MANAGED allocation type with a ValueError; use the managed memory resource for managed allocations.
  • The default handle type is platform-specific: None on Windows and POSIX_FD on other platforms. Host allocations require None.
• Documentation
  • Clarified supported allocation types, platform-specific handle defaults, accepted handle type values, and the requirement to use None for host allocations.

Virtual memory resource options now use platform-specific handle defaults and accept None. VirtualMemoryResource supports pinned allocations and validates allocation types during option initialization and configuration checks. Documentation, release notes, and tests reflect these changes.

Changes: Virtual memory resource options

Layer / File(s)	Summary
Platform-specific handle defaults cuda_core/cuda/core/_memory/_virtual_memory_resource.pyi, cuda_core/cuda/core/_memory/_virtual_memory_resource.pyx, cuda_core/tests/test_memory.py, cuda_core/docs/source/release/1.3.0-notes.rst	handle_type accepts strings and None. Its default is None on Windows and POSIX_FD elsewhere. Host-location checks and tests reflect the handle requirements.
Pinned-only allocation validation cuda_core/cuda/core/_memory/_virtual_memory_resource.pyx, cuda_core/cuda/core/_memory/_virtual_memory_resource.pyi, cuda_core/cuda/core/typing.py, cuda_core/tests/test_memory.py, cuda_core/docs/source/release/1.3.0-notes.rst	The allocation map contains only PINNED. Option initialization and configuration checks validate allocation types. Documentation identifies MANAGED as unsupported, and tests check managed allocation rejection.

Possibly related issues

• NVIDIA/cuda-python issue 2910: The change rejects allocation_type="managed" and sets the Windows handle default to None, while retaining the POSIX descriptor default elsewhere.

Suggested reviewers: andy-jost

Change: Bug fix

Merge Risk: Low · up to 55c15

Windows users could follow the documentation into an unsupported VMM configuration; correct the platform guidance before merging.

- [ ] Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Tip

🎯 Customize your risk assessments

Guide how CodeRabbit assesses merge risk and security architecture risk. Highlight sensitive changes and explain your security boundaries using Merge Risk Instructions and Security Risk Instructions in your review settings. Security architecture risk assessments require the Advanced plan or above.

✅ Automatic approvals, with your rules

Let CodeRabbit approve pull requests once required threads are resolved, pre-merge checks pass, and your risk limits are met. Enable Automatic approvals in review settings, with overrides for teams or file paths.

[Andy-Jost]

The code changes look good. I agree with the coderabbit items. Will approve once those are complete.

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to GitHub limitations.

⚠️ Outside diff range comments (1)

Minor · 🎯 Functional Correctness · _virtual_memory_resource.pyx:109

cuda_core/cuda/core/_memory/_virtual_memory_resource.pyx:109
Minor · Functional Correctness · Quick win

important: Restore the Linux qualifier for posix_fd import/export guidance.

cuda.core transports POSIX descriptors for VMM IPC only on Linux. The current wording can direct Windows users to an unsupported configuration. Apply the same correction in the .pyi documentation.

-        :attr:`VirtualMemoryResource.is_ipc_enabled`). Use ``"posix_fd"`` on
-        if you plan to import/export the allocation. Use `None` if you
+        :attr:`VirtualMemoryResource.is_ipc_enabled`). Use ``"posix_fd"`` on
+        Linux if you plan to import/export the allocation. Use `None` if you

Review info

Run configuration

• Source mode: Auto
• Repository: NVIDIA/cuda-python
• Branch: main
• File: .coderabbit.yaml
• Commit: f707a1f
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 735d1998-8076-4425-916b-96fc3578b775

Commits

Reviewing files that changed from the base of the PR and between a519baa and 55c156e.

Files selected for processing: 3

• cuda_core/cuda/core/_memory/_virtual_memory_resource.pyi
• cuda_core/cuda/core/_memory/_virtual_memory_resource.pyx
• cuda_core/tests/test_memory.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.

[Andy-Jost]

/ok to test 55c156e

[github-actions]

Doc Preview CI
🚀 View preview at https://nvidia.github.io/cuda-python/pr-preview/pr-3343/
https://nvidia.github.io/cuda-python/pr-preview/pr-3343/cuda-core/
https://nvidia.github.io/cuda-python/pr-preview/pr-3343/cuda-bindings/
https://nvidia.github.io/cuda-python/pr-preview/pr-3343/cuda-pathfinder/
Preview will be ready when the GitHub Pages deployment is complete.