Default Branch

master fc4747668e · paper(§5): add roofline, capacity planning, parallelism selection subsections · Updated 2026-07-30 22:56:41 +00:00

ZIP  TAR.GZ

Branches

feat/gqa-composite-commands a4a2683aad · paper(gqa): single-rank decode-streaming study — composite's masked bandwidth win · Updated 2026-06-24 02:39:45 +00:00    ywkang	77 1	Compare	ZIP  TAR.GZ
sccl-distributed-allreduce fd56b6cacd · adr: add ADR-0043/0044 (eval harnesses); reconcile ADR-0024/0032 for SIP w/h · Updated 2026-05-21 17:26:25 +00:00    mukesh	186 0	Included	ZIP  TAR.GZ
kernbench_Apr16 1d8b9401e5 · Intercube allreduce: pe0 cube-mesh reduce + multi-SIP ring/torus/mesh · Updated 2026-04-17 00:33:42 +00:00    mukesh	238 0	Included	ZIP  TAR.GZ