Skill Claude CodeCodex
Profile inside a CuTe DSL kernel with IKET (In-Kernel Event Tracing) and the run-iket profiler. Use when developing a CuTe DSL (cutlass.cute) kernel and tuning its performance — to get per-warp timelines of kernel phases (setup/mainloop/epilogue), see producer/consumer (TMA/MMA) overlap, and find where warps stall on…
5 29d ago A 170 tokens