Skip to content

triton-ascend-case-reduction-amin-medium

大规模2D归约(amin)reduce轴很大优化:在优先占满UB前提下为reduce轴分配较大切分尺寸(BLOCK_SIZE_N=16384最优),减少循环次数但需权衡单次迭代负载,适用于非reduce轴中等、reduce轴很大(50万级元素)的场景

Repository Source folder

Details

Path
akg_agents/python/akg_agents/op/resources/skills/triton-ascend/cases/triton-ascend-case-reduction-amin-medium/SKILL.md
Dependencies
3

FAQ