MISA
MISA copied to clipboard
ROCm
Machine Intelligence Shader Autogen. AMDGPU ML shader code generator. (previously iGEMMgen)
why we need nxb since wave_tile_n exists? and why gemm_n split into nxb ?
@qianfengz We are going to enforce two factor authentication in (https://github.com/ROCmSoftwarePlatform/) organization on 29th April, 2022 . Since we identified you as outside collaborator for ROCmSoftwarePlatform organization, you need to...
req: - [x] input/output: nchw->nchw-vecc nchw-vecc->nchw - [ ] weight: nchw->chwn-vecc - [ ] padding transpose: for cases c%vecc!=0, padding 0 at vecc's tail
As we continue optimizing the performance and stability for igemmGen and this tool can generate more efficient kernels for igemm or direct conv, we may think about how to merge...
There are three problems： 1、In the generated code, there is no instruction to output the result 2、without the the first problem solved ，the verification is still successful。 the problem can...
need to generalize code generation logic for different direction, precision, arch * global load/store: - [ ] support different precision, fp32/fp16(short)/ubyte - [ ] support 2d/3d load, and have exec...
- [x] unify fwd/bwd/wrw direction branch code in conv_driver.cpp - [ ] unify fp32/fp16/int8 logic - [x] unify fwd/bwd/wrw driver code