Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
composable_kernel
Commits
0197b61e
Commit
0197b61e
authored
May 10, 2023
by
Po-Yen, Chen
Browse files
Remove more unrelated instances
parent
e10d1043
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
7 additions
and
113 deletions
+7
-113
library/include/ck/library/tensor_operation_instance/gpu/gemm.hpp
...include/ck/library/tensor_operation_instance/gpu/gemm.hpp
+1
-107
library/src/tensor_operation_instance/gpu/gemm/CMakeLists.txt
...ary/src/tensor_operation_instance/gpu/gemm/CMakeLists.txt
+6
-6
No files found.
library/include/ck/library/tensor_operation_instance/gpu/gemm.hpp
View file @
0197b61e
...
...
@@ -257,124 +257,18 @@ struct DeviceOperationInstanceFactory<
{
std
::
vector
<
std
::
unique_ptr
<
DeviceOp
>>
op_ptrs
;
// if constexpr(is_same_v<ADataType, float> && is_same_v<BDataType, float> &&
// is_same_v<CDataType, float>)
// {
// 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_dl_f32_f32_f32_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_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_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_f32_f32_f32_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_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_f32_f32_f32_km_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_f32_f32_f32_km_kn_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_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_f32_f32_f32_km_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_f32_f32_f32_km_nk_mn_instances(op_ptrs);
// add_device_gemm_xdl_c_shuffle_f32_f32_f32_km_nk_mn_instances(op_ptrs);
// }
// }
if
constexpr
(
is_same_v
<
ADataType
,
half_t
>
&&
is_same_v
<
BDataType
,
half_t
>
&&
is_same_v
<
CDataType
,
half_t
>
)
{
if
constexpr
(
is_same_v
<
ALayout
,
Row
>
&&
is_same_v
<
BLayout
,
Row
>
&&
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_kn_mn_instances
(
op_ptrs
);
// add_device_gemm_dl_f16_f16_f16_mk_kn_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_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_f16_f16_f16_mk_nk_mn_instances
(
op_ptrs
);
// add_device_gemm_dl_f16_f16_f16_mk_nk_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instances
(
op_ptrs
);
add_device_gemm_xdl_c_shuffle_2_stage_f16_f16_f16_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_f16_f16_f16_km_kn_mn_instances
(
op_ptrs
);
// add_device_gemm_dl_f16_f16_f16_km_kn_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_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_f16_f16_f16_km_nk_mn_instances
(
op_ptrs
);
// add_device_gemm_dl_f16_f16_f16_km_nk_mn_instances(op_ptrs);
add_device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instances
(
op_ptrs
);
}
}
// else if constexpr(is_same_v<ADataType, ck::bhalf_t> && is_same_v<BDataType, ck::bhalf_t>
// &&
// is_same_v<CDataType, ck::bhalf_t>)
// {
// if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_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_bf16_bf16_bf16_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_bf16_bf16_bf16_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_bf16_bf16_bf16_km_nk_mn_instances(op_ptrs);
// }
// }
// else if constexpr(is_same_v<ADataType, int8_t> && is_same_v<BDataType, int8_t> &&
// is_same_v<CDataType, int8_t>)
// {
// if constexpr(is_same_v<ALayout, Row> && is_same_v<BLayout, Row> &&
// is_same_v<CLayout, Row>)
// {
// add_device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_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_i8_i8_i8_mk_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_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_i8_i8_i8_km_kn_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_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_i8_i8_i8_km_nk_mn_instances(op_ptrs);
// add_device_gemm_dl_i8_i8_i8_km_nk_mn_instances(op_ptrs);
// }
// }
return
op_ptrs
;
}
...
...
library/src/tensor_operation_instance/gpu/gemm/CMakeLists.txt
View file @
0197b61e
...
...
@@ -7,10 +7,10 @@ add_instance_library(device_gemm_instance
# device_gemm_xdl_f32_f32_f32_mk_nk_mn_instance.cpp
# device_gemm_xdl_f32_f32_f32_km_kn_mn_instance.cpp
# device_gemm_xdl_f32_f32_f32_km_nk_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_mk_kn_mn_instance.cpp
#
device_gemm_xdl_f16_f16_f16_mk_kn_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_mk_nk_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_km_kn_mn_instance.cpp
device_gemm_xdl_f16_f16_f16_km_nk_mn_instance.cpp
#
device_gemm_xdl_f16_f16_f16_km_kn_mn_instance.cpp
#
device_gemm_xdl_f16_f16_f16_km_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_i8_i8_i8_mk_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_i8_i8_i8_mk_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_i8_i8_i8_km_kn_mn_instance.cpp
...
...
@@ -19,10 +19,10 @@ add_instance_library(device_gemm_instance
# device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_bf16_bf16_bf16_km_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
#
device_gemm_xdl_c_shuffle_f16_f16_f16_mk_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_mk_nk_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
#
device_gemm_xdl_c_shuffle_f16_f16_f16_km_kn_mn_instance.cpp
#
device_gemm_xdl_c_shuffle_f16_f16_f16_km_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f32_f32_f32_mk_kn_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f32_f32_f32_mk_nk_mn_instance.cpp
# device_gemm_xdl_c_shuffle_f32_f32_f32_km_kn_mn_instance.cpp
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment