GPU performance and validation skills from AMD Research.
AMD Research skills for kernel isolation, profiling, line analysis, execution inspection, and correctness validation.
Arranged in the author's recommended order. Walk through them in sequence, or open any one on its own.
Profile GPU kernels when performance analysis or optimization is required.
Validate GPU kernel correctness by comparing reference and optimized outputs.
Extract GPU kernel assembly and HIP source from HSA packet traces.
Profile GPU kernels at source-line granularity with cycle-level timing and stall analysis.
Test local kerncap changes end-to-end by profiling an application, extracting a kernel, and validating the reproducer.