ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-13 02:27:33 +00:00
Code Issues Packages Projects Releases Wiki Activity

update instances

Browse Source
This commit is contained in:
aska-0096
2024-12-13 03:29:15 +00:00
parent 2e9901bfa6
commit e8ca3daf4e
11 changed files with 298 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

61
library/include/ck/library/tensor_operation_instance/gpu/gemm_multiply_multiply.hpp
View File

@@ -69,6 +69,58 @@ void add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_kpadding_ins
PassThrough,
MultiplyMultiply>>>& instances);
void add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_default_instances_part1(
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitK<Row,
Col,
Tuple<Row, Col>,
Row,
F8,
F8,
Tuple<F32, F32>,
BF16,
PassThrough,
PassThrough,
MultiplyMultiply>>>& instances);
void add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_kpadding_instances_part1(
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitK<Row,
Col,
Tuple<Row, Col>,
Row,
F8,
F8,
Tuple<F32, F32>,
BF16,
PassThrough,
PassThrough,
MultiplyMultiply>>>& instances);
void add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_default_instances_part2(
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitK<Row,
Col,
Tuple<Row, Col>,
Row,
F8,
F8,
Tuple<F32, F32>,
BF16,
PassThrough,
PassThrough,
MultiplyMultiply>>>& instances);
void add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_kpadding_instances_part2(
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitK<Row,
Col,
Tuple<Row, Col>,
Row,
F8,
F8,
Tuple<F32, F32>,
BF16,
PassThrough,
PassThrough,
MultiplyMultiply>>>& instances);
void add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_mem_v1_default_instances(
std::vector<std::unique_ptr<DeviceGemmMultipleDSplitK<Row,
Col,
@@ -256,6 +308,15 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGemmMu
add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_kpadding_instances_part2(
op_ptrs);
add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_default_instances_part1(
op_ptrs);
add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_kpadding_instances_part1(
op_ptrs);
add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_default_instances_part2(
op_ptrs);
add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_comp_mfma16x16_kpadding_instances_part2(
op_ptrs);
add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_mem_v1_default_instances(
op_ptrs);
add_device_gemm_multiply_multiply_xdl_f8_f8_bf16_mk_nk_mn_mem_v1_kpadding_instances(