system-profile

To profile a Python script, you specify the target file like 'train.py' and the system runs profiling to generate structured performance metrics across CPU and memory. It produces actionable recommendations and saves artifacts to a dedicated profile_output directory.

Yes, you can profile a running process by providing its PID. The system attaches to the target process to measure memory usage and CPU hotspots, generating structured reports with actionable recommendations without requiring application restarts.

The best way to profile GPU workloads and interconnect bandwidth is to specify 'gpu' as your target. The system analyzes the GPU workload and interconnect topology, producing structured tables and charts that highlight bandwidth bottlenecks.

You need minimal instrumentation to measure hotspots and bottlenecks accurately. The system provides specific instrumentation guidelines designed to capture performance metrics without significantly perturbing the application's runtime behavior.

Performance profiling works effectively in development, testing, and production environments. The system supports profiling Python scripts, running processes, and GPU workloads across these distinct contexts to identify bottlenecks.

Profiling reports are saved as structured artifacts containing tables and charts in a dedicated profile_output directory. These reports include CPU overhead metrics, memory footprints, and actionable recommendations for performance optimization.