How to benchmark compiler opts: inspect emitted code, split cold vs steady
cfregly · x · 2026-10-22
Takeaway from the series: inspect emitted compiler fragments and CUDA replay activity, and measure cold first use separately from repeated execution — otherwise optimizations like nested graphs (big steady-state win, modest cold-start cost) are easy to misread.
Related event: Nested PyTorch graphs speed steady-state 4.5x at cold-start cost(6 posts)→
More from Infra
- RL environment task prices have gone bananas, says insider on the data market — jia_seed · 2026-10-22
- Cloudflare ramps builds back up, promises full post mortem tomorrow — edgarpavlovsky · 2026-10-22
- SpaceX's Bastrop site balloons into industrial city with 1.1M sq ft solar factory — XFreeze · 2026-10-22
- Nested CUDA graphs cut replay launches from 16 to 2 per call — cfregly · 2026-10-22
- Reddit asks: which 50-150B MoE models are worth running locally besides Qwen Flash Next — FlyingFishMakeAWish · 2026-10-22
- Open-source tool condenses millions of logs into an LLM-sized snapshot — CommanderAlPachino · 2026-10-22