[infra] Limit PR CUDA builds to runner architectures

[vikuznetsov-nvidia]

Build x86 targets for the detected runner GPU and pin Jetson jobs to their platform architectures to avoid compiling unused CUDA variants.

Summary by CodeRabbit

• Chores
  • Updated GPU build and test configurations to target the appropriate CUDA architectures across supported hardware.
  • Improved consistency for rerun-enabled builds by applying the correct CUDA architecture settings.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Enterprise

Run ID: a322b59f-e9b6-48a3-94f5-e1efd1323ead

📥 Commits

Reviewing files that changed from the base of the PR and between b8c4992 and e57f3a9.

📒 Files selected for processing (1)

• .github/workflows/pr-verify.yml

---

📝 Walkthrough

Walkthrough

The pull request updates GPU build and test workflow jobs to pass explicit CUDA architectures through EXTRA_CMAKE_ARGS, including the x86 rerun build while preserving USE_RERUN=ON.

Changes

CUDA build configuration

Layer / File(s)	Summary
Configure CUDA architectures for GPU jobs .github/workflows/pr-verify.yml	The x86, Orin, and Thor jobs set CUDA architectures to native, 87, and 110; the x86 rerun build combines native with USE_RERUN=ON.

Estimated code review effort: 1 (Trivial) | ~5 minutes

Suggested reviewers: hrabeti-nvidia, slepichev, hguillen

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main change: restricting PR CUDA builds to runner and platform architectures.
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.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch vikuznetsov/pin-pr-cuda-architectures

---

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

[github-actions]

Test Results

Status	Platform	Language	Total	Passed	Failed	Errors	Skipped
✅	Orin	C++	16	16	0	0	0
✅	Orin	Python	68	67	0	0	1
✅	Thor	C++	16	16	0	0	0
✅	Thor	Python	68	67	0	0	1
✅	x86_64	C++	16	16	0	0	0
✅	x86_64	Python	68	67	0	0	1

cuVSLAM Evaluation KPIs

Config	Dataset	ATE,%	ARE,º/m	Kabsch,	Losts,	diff ATE,%	diff ARE,º/m	diff Kabsch,	diff Losts,	FPS,Hz
x86_64-cuda12.6.3-ubuntu24.04	KITTI-STEREO_ODOM	0.8468	0.0024	2.909	0	NA	NA	NA	NA	297.4
x86_64-cuda12.6.3-ubuntu24.04	KITTI-STEREO_SLAM	0.7625	0.002	2.0586	0	NA	NA	NA	NA	176.0

Artifacts

• eval-html-report
• eval-kpis-pr
• test-results-x86_64
• docs-html
• test-results-orin
• test-results-thor