Attempt to stabilize flaky benchmarks

[nvtw]

Description

Stabilize SDF benchmarks by limiting gating coverage to resolutions with repeatable timings, and strengthen hydroelastic contact-surface repeatability coverage.

Checklist

• New or existing tests cover these changes
• The documentation is up to date with these changes
• CHANGELOG.md has been updated (if user-facing change)

Test plan

uv run --extra dev -m newton.tests -k test_no_degenerate_triangles_deep_penetration

Summary by CodeRabbit

• Bug Fixes

  • Improved stability and repeatability of hydroelastic contact-surface generation during deep penetration.
  • Ensured generated contact surfaces avoid degenerate triangles.

• Benchmarks

  • Updated SDF performance benchmarks to use stable grid resolutions, excluding resolution 128 and higher.

• Documentation

  • Updated the unreleased changelog entry with the related issue references.

[coderabbitai]

📝 Walkthrough

Walkthrough

The hydroelastic deep-penetration test now checks repeatability across SDF rebuilds. The SDF benchmark excludes resolution 128 because of bimodal timing behavior, and the changelog references both related fixes.

Changes

Hydroelastic SDF stability

Layer / File(s)	Summary
Deep-penetration repeatability checks newton/tests/test_hydroelastic.py, CHANGELOG.md	The test repeats selected deep-penetration builds, checks contact-surface triangle degeneracy, and asserts stable face counts across rebuilds. The changelog adds the related fix reference.
SDF benchmark resolution coverage asv/benchmarks/setup/bench_sdf.py	SDF build measurements now cover resolutions 32 and 64, excluding resolution 128 from batching and benchmark parameters.

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

Possibly related PRs

• newton-physics/newton#2424: Introduced the hydroelastic marching-cubes degeneracy fix exercised by the updated test.
• newton-physics/newton#2747: Updated the same deep-penetration hydroelastic test setup and verification logic.
• newton-physics/newton#3417: Adjusted the same SDF benchmark resolution coverage.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 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.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title matches the main goal of stabilizing flaky benchmark-related tests and benchmark measurements.

✨ 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.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!