BBBK kernels fuse quantization into GEMM epilogue, saturating NVLink and Tensor Cores
snowclipsed · x · 2026-11-14
A technical discussion of the BBBK kernel engineering approach: to eliminate stalls, BBBK keeps as many hardware pipelines saturated as possible — fusing RHT and S2F6 quantization directly into the w2 GEMM epilogue, combined with aggressive epilogue fusion, ALU vectorization, instruction scheduling, register planning/packing, and speculative megakernel scheduling. In training it simultaneously saturates NVLink, ALUs, and Tensor Cores; in RL sampling it hits the latency floor set by NVLink and HBM. The thread also notes k=96 sampling, likening it to a Cursor-like recipe.
Related event: MoE Training Communication Slashed via 8-bit Quantization and Fused Kernels(4 posts)→
More from Infra
- Atreides sends letter to Nvidia board outlining minimum governance standards — GavinSBaker · 2026-11-14
- OSINT deep-dive: FAA nationwide grounding caused by long-haul ATC comms, radar and GPS fine — eigenrobot · 2026-11-14
- Airtel chairman: Starlink V3 delivers 1 Tbps per launch, matching all of OneWeb — elonmusk · 2026-11-14
- Dev goes all-in on Cloudflare, citing its excellent MCP as the deciding factor — ritakozlov · 2026-11-14
- Brave Leo adds verifiable encrypted LLM inference via TEEs, calls ZDR clauses a pinky promise — gnukeith · 2026-11-14
- 97% of coding-agent tokens are cache reads, but cache writes eat 31% of the cost — Nitjsefnie · 2026-11-14