Omniverse USD Performance Tuning for Codex: Compare Profiles
Use NVIDIA Omniverse's USD tuning workflow to define a baseline and an after profile for the same scene and separate measurable changes from diagnosis.
中文Setup for Codex (NVIDIA plugin)
Codex with two synthetic profile records. Full USD profiling depends on the chosen runtime and may use Omniverse Kit, USD Python, Asset Validator or other tools.
- Open the pinned GitHub source for this version. Copy the entire `skills/omniverse-usd-performance-tuning/` folder with its references into your scratch project's `.agents/skills/omniverse-usd-performance-tuning/`.
- Start Codex from that project and load the skill. Read the entry skill's profile and assessment milestones. Keep the scene identifier, settings and units the same in the before and after fixtures.
Example task
Create a local baseline for scene demo-stage: loadMs 2000, peakMemoryMb 600. The after record uses the same scene and settings with loadMs 1500, peakMemoryMb 660. Ask for the two percentage changes and a trade-off note.Expected result
Load time decreases by 25 percent and peak memory increases by 10 percent. The report keeps the two metrics separate and notes the added memory cost.
Upstream documentation
Original SKILL.md on GitHub · Apache-2.0 license · Source reviewed
Frequently asked question
What makes an Omniverse USD performance comparison credible?
Profile the same workload before and after with comparable settings, metrics and units. Record both improvement and regression where they occur.