What problem does it solve?
Measuring and improving Isaac Sim performance requires coordinating benchmark runs, GPU frame-time recording, Tracy profile capture, and CSV comparison across iterations, which is error-prone when done manually.
Core Features & Use Cases
- Benchmark Execution with Tracy Capture: Run any of the built-in Isaac Sim benchmarks (camera, RTX lidar, robots, SDG, scene loading) through the tracy_capture.py wrapper to record profiles, export CSVs, and save full logs in one command.
- Zone-Level Comparison: Compare two Tracy CSV exports with the bundled compare_tracy_csvs.py script to surface top zones by total time, new hot paths, and disappeared zones.
- Iterative Optimization Workflow: Track frame time, GPU time, and FPS across multiple optimization iterations with side-by-side comparison tables.
- Extension Test Profiling: Discover the child kit process command from extension unit tests and rerun it directly with Tracy settings to profile test workloads.
- Use Case: A robotics engineer notices slow rendering in the nova_carter benchmark, captures a baseline Tracy profile, modifies physics or rendering code, and re-runs the identical benchmark to verify a measurable frame-time improvement.
Quick Start
Profile the robots_nova_carter benchmark with GPU frame timing for 600 frames and compare the resulting Tracy CSV against my previous baseline run.
