ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-20 21:09:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add fp8 gemm instances (#920)

Browse Source 
* Add fp8 gemm instances

* Update instance naming

[ROCm/composable_kernel commit: 94bfa50256]
This commit is contained in:
Rostyslav Geyyer
2023-09-26 14:59:33 -05:00
committed by GitHub
parent 37f4626e3e
commit b74d4f5fc6
8 changed files with 350 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
library/include/ck/library/tensor_operation_instance/gpu/gemm.hpp
View File

@@ -312,6 +312,23 @@ void add_device_gemm_xdl_f64_f64_f64_mk_nk_mn_instances(
DeviceGemm<Row, Col, Row, F64, F64, F64, PassThrough, PassThrough, PassThrough>>>&
instances);
#endif
#ifdef CK_ENABLE_FP8
void add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_kn_mn_instances(
std::vector<std::unique_ptr<
DeviceGemm<Col, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
void add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_nk_mn_instances(
std::vector<std::unique_ptr<
DeviceGemm<Col, Col, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances(
std::vector<std::unique_ptr<
DeviceGemm<Row, Row, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
void add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances(
std::vector<std::unique_ptr<
DeviceGemm<Row, Col, Row, F8, F8, F8, PassThrough, PassThrough, PassThrough>>>& instances);
#endif
template <typename ALayout,
typename BLayout,
typename CLayout,
@@ -505,6 +522,32 @@ struct DeviceOperationInstanceFactory<
#endif
}
}
#endif
#ifdef CK_ENABLE_FP8
else if constexpr(is_same_v<ADataType, ck::f8_t> && is_same_v<BDataType, ck::f8_t> &&
is_same_v<CDataType, ck::f8_t>)
{
if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_f8_f8_f8_mk_nk_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Row> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_kn_mn_instances(op_ptrs);
}
else if constexpr(is_same_v<ALayout, Col> && is_same_v<BLayout, Col> &&
is_same_v<CLayout, Row>)
{
add_device_gemm_xdl_c_shuffle_f8_f8_f8_km_nk_mn_instances(op_ptrs);
}
}
#endif
return op_ptrs;
}