Skip to content

perf(conv): vectorized staging, per-shape block tiles and fp32 forward reads in the WMMA kernels - #14

Merged
AlexanderSoftCode merged 5 commits into
mainfrom
perf/conv-wmma-staging
Oct 30, 2026
Merged

AlexanderSoftCode merged 5 commits into
mainfrom
perf/conv-wmma-staging

Commits

  1. Commits on Oct 30, 2026