Skip to content

[Perf] Add async GMEM->LDS infra; switch KV tile load to inverse-swizzle addressing#212

Open
diptorupd wants to merge 3 commits into
ROCm:amd-integrationfrom
diptorupd:perf/async-pipeline
Open

[Perf] Add async GMEM->LDS infra; switch KV tile load to inverse-swizzle addressing#212
diptorupd wants to merge 3 commits into
ROCm:amd-integrationfrom
diptorupd:perf/async-pipeline

Commits

Commits on Apr 16, 2026

Commits on Apr 18, 2026