composable_kernel/include/utility/amd_inline_asm.hpp · 3bf52e60c5374c9a63256dff5e3442a4046c81dc · yangql / composable_kernel-1 · GitLab

Find file Blame History Permalink

Hybrid direct + implicit GEMM forward convolution NCHWc v5r1 (#25) · 792a20fa
zjing14 authored Apr 07, 2021
```
* Hybrid direct + implicit GEMM forward convolution NCHWc v5r1. Input tensor bypass LDS. Support fp32/fp16/int8
```
792a20fa

amd_inline_asm.hpp 13.5 KB