ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 05:01:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add fp16/fp8 support into Grouped gemm FixedNK (#874)

Browse Source 
* move all arguments into device

* add b2c_tile_map

* add examples

* add SetDeviceKernelArgs

* dedicated fixed_nk solution

* init client api

* add grouped_gemm_bias example

* add a instance

* add instances

* formatting

* fixed cmake

* Update EnableCompilerWarnings.cmake

* Update cmake-ck-dev.sh

* clean; fixed comments

* fixed comment

* add instances for fp32 output

* add instances for fp32 output

* add fp32 out client example

* fixed CI

* init commit for kbatch

* add splitk gridwise

* format

* fixed

* clean deviceop

* clean code

* finish splitk

* fixed instances

* change m_loops to tile_loops

* add setkbatch

* clean code

* add splitK+bias

* add instances

* opt mk_nk instances

* clean examples

* fixed CI

* remove zero

* finished non-zero

* clean

* clean code

* optimized global_barrier

* fixed ci

* fixed CI

* instance and client

* removed AddBias

* format

* fixed CI

* fixed CI

* move 20_grouped_gemm to 21_grouped_gemm

* clean

* formatting

* clean

* clean

* fixed computeType

---------

Co-authored-by: Jing Zhang <jizha@amd.com>
This commit is contained in:
zjing14
2023-09-14 21:04:10 -05:00
committed by GitHub
parent 0d8efaa13d
commit f9d0eddb90
17 changed files with 1749 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
include/ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp
View File

@@ -75,12 +75,24 @@ struct PassThrough
y = type_convert<bhalf_t>(x);
}
template <>
__host__ __device__ void operator()<float, half_t>(float& y, const half_t& x) const
{
y = type_convert<float>(x);
}
template <>
__host__ __device__ void operator()<int8_t, int8_t>(int8_t& y, const int8_t& x) const
{
y = x;
}
template <>
__host__ __device__ void operator()<half_t, int8_t>(half_t& y, const int8_t& x) const
{
y = type_convert<half_t>(x);
}
template <>
__host__ __device__ void operator()<int8_t, int32_t>(int8_t& y, const int32_t& x) const
{