profile-hotpath

What problem does it solve?

This Skill helps you measure where a backend feature spends time so you can focus optimization work on the true hot path instead of guessing.

Core Features & Use Cases

• Backend-aware profiling: Uses the right profiler for CUDA, SYCL, HIP, Metal, or CPU fallback paths.
• Symbolicated performance analysis: Builds with debug symbols when needed, then produces flamegraphs and hot-function rankings.
• Actionable optimization guidance: Turns benchmark results into concrete suggestions for reducing latency and improving throughput.
• Use case: Profile a GPU rendering feature, compare bottlenecks across backends, and identify the functions that dominate frame time.

Quick Start

Ask the skill to profile the specified backend and feature with your preferred resolution and frame count so it can run the benchmark, generate a flamegraph, and summarize the hottest functions.