{"slug":"nvidia-tilegym-improve-cutile-kernel-perf","source_name":"nvidia/tilegym-improve-cutile-kernel-perf","name":"Nvidia/Tilegym Improve Cutile Kernel Perf","description":"Iteratively optimize cuTile kernel performance through systematic profiling, bottleneck analysis, IR comparison, and targeted tuning. Covers tile sizes, occupancy, autotune configs, TMA, latency hints, persistent scheduling, num_ctas, flush_to_zero, and IR-level debugging. Use when asked to \"optimize cutile kernel\", \"improve kernel perf\", \"tune cutile performance\", \"make kernel faster\", or iteratively benchmark and refine a cuTile GPU kernel in the TileGym project.","version":1,"lift":{"pass_rate_delta_pts":43.48,"pass_rate_pct":78.3,"total_cases":23,"passed_cases":18,"tokens_delta_pct":56.7,"turns_delta_pct":0,"verdict":"mixed","benchmark_model":"gemini-3.6-flash","grading_method":"judged","completed_at":"2026-08-12T23:54:12.370045+00:00"},"skill_score":null,"benchmark_models":[{"model":"gemini-3.6-flash","headline":true,"delta_pts":43.48,"with_pass_pct":78.3,"without_pass_pct":34.8,"tokens_delta_pct":56.7,"turns_delta_pct":0,"total_cases":23,"cases_aggregated":18,"verdict":"mixed","never_hurt":false,"completed_at":"2026-08-12T23:54:12.370045+00:00","run_id":"d75aef74-f9f3-47fa-8dcf-2d2c0f00b932","version_number":1,"is_latest_version":true,"gate":null}],"trust":{"skill_safety":"passed","safety_status":"clean","intent_verdict":"safe","content_status":"clean","indexable":true},"license":"CC-BY-4.0 AND Apache-2.0","install_count":0,"manifest_hash":"372fa218dd1799955057d8751096cd8051d77b2ee1bee6ab4083e9e3c8e45bf1","raw_url":"https://app.decimal.ai/s/nvidia-tilegym-improve-cutile-kernel-perf/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/nvidia-tilegym-improve-cutile-kernel-perf"}