Skip to content

remove-occupancy-nvrtc-dependency - #494

Merged
Anerudhan merged 1 commit into
NVIDIA:developfrom
yanqinz2:fix-occupancy-blackwell-ci
Aug 7, 2026
Merged

remove-occupancy-nvrtc-dependency#494
Anerudhan merged 1 commit into
NVIDIA:developfrom
yanqinz2:fix-occupancy-blackwell-ci

Conversation

@yanqinz2

@yanqinz2 yanqinz2 commented Aug 5, 2026

Copy link
Copy Markdown
Contributor

Before submitting

  • I agree to license this contribution under the terms of LICENSE.txt.
  • I ran pre-commit run and committed any formatting changes.
  • I added GitHub labels: one cat-*, one or more mod-*, and one orig-* (see label list).

Affected area

Summary

Remove nvrtc dependency when calculating the occupancy map, to fix the ci issue

Related issues

API and compatibility impact

Testing

Summary by CodeRabbit

  • Performance

    • Improved GPU occupancy detection for newer SM90-class devices by using precompiled resources, reducing runtime compilation overhead.
    • Added cached detection of each device’s multiprocessor count for faster repeated queries.
  • Compatibility

    • Added reliable fallback occupancy behavior for older GPU architectures, including support for single-cluster workloads.

@coderabbitai

coderabbitai Bot commented Aug 5, 2026

Copy link
Copy Markdown
Contributor

Review Change Stack

📝 Walkthrough

Walkthrough

Changes

Device-Aware Occupancy Probing

Layer / File(s) Summary
Multiprocessor count query
python/cudnn/frost/device.py
Adds a cached CUDA driver query for a device’s multiprocessor count.
Occupancy map capability handling
python/cudnn/frost/occupancy.py
Uses precompiled SM90 PTX for occupancy probing and returns fallback values for devices below SM90.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: anerudhan

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name Status Explanation Resolution
Description check ⚠️ Warning The description includes a summary but omits the affected area, rationale, API impact, and testing details. Select the affected area and complete the Why, API and compatibility impact, and Testing sections.
✅ Passed checks (4 passed)
Check name Status Explanation
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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.
Title check ✅ Passed The title clearly and concisely states that the NVRTC dependency is removed from occupancy processing.
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests

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

@Anerudhan Anerudhan added this to the Frontend 1.28.0 milestone Aug 6, 2026
@Anerudhan
Anerudhan self-requested a review August 6, 2026 19:04
@Anerudhan Anerudhan added cat-bug Reports of incorrect behavior, crashes, regressions, or unexpected results. orig-nv-eng Reported or requested by NVIDIA engineering. mod-frost labels Aug 6, 2026
@Anerudhan

Copy link
Copy Markdown
Collaborator

@cudnn-ci-bot run frost

@cudnn-ci-bot

Copy link
Copy Markdown

🚀 Running mirror pipeline

Branch: cudnn-gh/pr-494-ef71629
Pipeline: 61448138
Targets: frost

@Anerudhan
Anerudhan merged commit b232b44 into NVIDIA:develop Aug 7, 2026
1 check passed
@yanqinz2
yanqinz2 deleted the fix-occupancy-blackwell-ci branch August 7, 2026 18:05
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

cat-bug Reports of incorrect behavior, crashes, regressions, or unexpected results. mod-frost orig-nv-eng Reported or requested by NVIDIA engineering.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants