ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-12 01:10:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Improve 4k gemm perf (#1047)

Browse Source 
* improve 4k gemm perf

* add f8 instances

* format

---------

Co-authored-by: Jing Zhang <jizha@amd.com>
This commit is contained in:
zjing14
2023-11-17 07:06:24 -06:00
committed by GitHub
parent e1fa00917c
commit e8cddfdc3b
7 changed files with 198 additions and 109 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
library/include/ck/library/tensor_operation_instance/gpu/gemm.hpp
View File

@@ -377,7 +377,7 @@ struct DeviceOperationInstanceFactory<
if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
#endif
@@ -386,7 +386,7 @@ struct DeviceOperationInstanceFactory<
else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
#endif
@@ -395,7 +395,7 @@ struct DeviceOperationInstanceFactory<
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
#endif
@@ -404,7 +404,7 @@ struct DeviceOperationInstanceFactory<
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
#endif
@@ -418,7 +418,7 @@ struct DeviceOperationInstanceFactory<
if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_mk_kn_mn_irregular_instances(op_ptrs);
@@ -430,7 +430,7 @@ struct DeviceOperationInstanceFactory<
else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_mk_nk_mn_irregular_instances(op_ptrs);
@@ -443,7 +443,7 @@ struct DeviceOperationInstanceFactory<
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_km_kn_mn_irregular_instances(op_ptrs);
@@ -455,7 +455,7 @@ struct DeviceOperationInstanceFactory<
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
/// add_device_gemm_xdl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
#ifdef DL_KERNELS
add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
add_device_gemm_dl_f16_f16_f16_km_nk_mn_irregular_instances(op_ptrs);