fix: align is_sm120f_supported with SM12x family semantics

[leonardHONG]

📌 Description

is_sm120f_supported requires minor == 0, so it returns False on sm_121 even though the 120f gencode target covers both sm_120 and sm_121. Drop the minor == 0 check.

No callers in flashinfer/ or tests/. Only CLAUDE.md mentions it. Could delete the function instead if that's cleaner.

🔍 Related Issues

Refs #3170 (Action Item 5 / A1).

🚀 Pull Request Checklist

✅ Pre-commit Checks

• I have installed pre-commit by running pip install pre-commit (or
  used your preferred method).
• I have installed the hooks with pre-commit install.
• I have run the hooks manually with pre-commit run --all-files and
  fixed any reported issues.

Ran pre-commit run --files flashinfer/utils.py; passed.

🧪 Tests

• Tests have been added or updated as needed.
• All tests are passing (unittest, etc.).

No new tests; the function is currently uncalled.

Reviewer Notes

Summary by CodeRabbit

• Bug Fixes
  • Expanded SM120f GPU support compatibility to include additional GPU variants within the compute capability 12 family while maintaining CUDA version requirements.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: e1694fca-f63d-4888-9bd6-7b73278e0d85

📥 Commits

Reviewing files that changed from the base of the PR and between 24c4aee and 1a876d3.

📒 Files selected for processing (1)

• flashinfer/utils.py

---

📝 Walkthrough

Walkthrough

The is_sm120f_supported function in flashinfer/utils.py was modified to check only the GPU's compute capability major version (major == 12) instead of requiring a specific minor version, while maintaining the CUDA version requirement.

Changes

Cohort / File(s)	Summary
GPU Compute Capability Check flashinfer/utils.py	Modified is_sm120f_supported to ignore compute capability minor version, now checking only major == 12 instead of major == 12 and minor == 0.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Possibly related PRs

• fix: allow fmha_v2_prefill_deepseek on SM121 (DGX Spark) #2559 — Both PRs relax SM12x support checks by accepting a broader range of SM12 variants instead of a single specific version.
• feat: add is_sm12x_supported() helper for SM12x family detection #2574 — Both PRs modify SM12x GPU detection logic in flashinfer/utils.py by adjusting how compute-capability minor versions are handled.

Suggested labels

run-ci

Suggested reviewers

• yzh119
• bkryu
• cyx-6
• nvmbreughe

Poem

🐰 A minor version checked, now let it go,
Any SM12 will gracefully flow,
No more strict bounds on our GPU's might,
Flexibility blooms—compatibility takes flight! ✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately reflects the main change: removing the minor version check to align SM120f support with SM12x family semantics.
Description check	✅ Passed	The description covers the core issue (minor == 0 check prevents sm_121 support), rationale, related issue reference, and pre-commit status. However, test section is incomplete.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[gemini-code-assist]

Code Review

This pull request updates the is_sm120f_supported function in flashinfer/utils.py to simplify the compute capability check by only verifying the major version (12) and removing the specific minor version check. I have no feedback to provide.

[sricketts]

/bot run

[flashinfer-bot]

GitLab MR !614 has been created, and the CI pipeline #49717071 is currently running. I'll report back once the pipeline job completes.