Triton Ascend Case Reduction Amax Large

非reduce轴很小、reduce轴很大的归约优化:将reduce轴映射到多核(而非常规的非reduce轴),使用原子操作跨线程块归约,通过二次切分避免超UB,适用于极端shape比例(M<<N如16×262144)的归约场景

wenyi-li Updated 10 repo stars

File contents

wenyi-li/awesome-agent-kernel-skills commit c9dd200c98

Frequently asked questions

npx get-skill add wenyi-li/triton-ascend-case-reduction-amax-large