What problem does it solve?
Metal GPU debugging and profiling on macOS requires coordinating multiple tools (gpu_trace, xctrace, xcrun metal) to capture, parse, and interpret GPU activity and shader behavior. This skill provides an end-to-end workflow to automate environment validation, trace collection, data extraction, and reporting for Metal workloads on Apple Silicon.
Core Features & Use Cases
- Automated session lifecycle: doctor/checks, record, export, parse, analyze, and report traces.
- GPU trace tooling integration: gpu_trace, xctrace, and Metal validation layers for comprehensive debugging.
- Performance insights: per-kernel timing, hardware counters, and HUD metrics to identify bottlenecks.
Quick Start
Record a Metal System Trace for your app with gpu_trace run, then analyze the trace with gpu_trace analyze and review the generated output.