Rank #99 open-sources 5 H200 kernels from Open Kernel Challenge, up to 2.66x over SGLang
abhijithneil · x · 2026-11-10
abhijithneil open-sourced his Open Kernel Challenge submissions (rank #99, double digits), covering 5 NVIDIA H200 kernels with learning notes. Speedups over the SGLang reference implementation:
- Dense projections: 1.59x
- Attention decode: 2.66x
- Routed experts: 2.43x
- Decoder layer: 2.30x
- Prefill layer: 1.93x
The repo includes a recommended reading order (start with dense projections) and concise explanations of terms like token, projection, expert, and pooler — a practical entry point for developers wanting to learn GPU kernel optimization.
More from Infra
- Musk reveals Grok 4.8: entire RL and inference stack rewritten from scratch in C/C++ — elonmusk · 2026-11-10
- Running open-weight models locally: Kimi K3.1 needs 594GB, ~$38k for 15 tok/s — ravikantagrawal · 2026-11-10
- InfraRSI Applies Recursive Self-Improvement to Training Infra: 2.75x Speedup, MFU Lifted from 17.4% to 30% — TokenRhythm · 2026-11-10
- ClickBench Playground Lets You Run SQL Against 110+ Databases Side by Side — mgill25 · 2026-11-10
- Musk: new Grok training stack derived from Starlink's in-house engineering — elonmusk · 2026-11-10
- 4x RTX 3090 vs one RTX 5090: Malaysian builder weighs VRAM against newer architecture — whity2773 · 2026-11-10