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
65c56e56
Commit
65c56e56
authored
Jul 25, 2022
by
Chao Liu
Browse files
update Tensor
parent
028171e9
Changes
60
Hide whitespace changes
Inline
Side-by-side
Showing
20 changed files
with
76 additions
and
73 deletions
+76
-73
example/01_gemm/gemm_dl_fp16.cpp
example/01_gemm/gemm_dl_fp16.cpp
+3
-3
example/01_gemm/gemm_dl_fp32.cpp
example/01_gemm/gemm_dl_fp32.cpp
+3
-3
example/01_gemm/gemm_dl_int8.cpp
example/01_gemm/gemm_dl_int8.cpp
+3
-3
example/01_gemm/gemm_xdl_bf16.cpp
example/01_gemm/gemm_xdl_bf16.cpp
+3
-3
example/01_gemm/gemm_xdl_fp16.cpp
example/01_gemm/gemm_xdl_fp16.cpp
+3
-3
example/01_gemm/gemm_xdl_fp64.cpp
example/01_gemm/gemm_xdl_fp64.cpp
+3
-3
example/01_gemm/gemm_xdl_int8.cpp
example/01_gemm/gemm_xdl_int8.cpp
+3
-3
example/02_gemm_bilinear/gemm_bilinear_xdl_fp16.cpp
example/02_gemm_bilinear/gemm_bilinear_xdl_fp16.cpp
+4
-4
example/03_gemm_bias_relu/gemm_bias_relu_xdl_fp16.cpp
example/03_gemm_bias_relu/gemm_bias_relu_xdl_fp16.cpp
+4
-4
example/04_gemm_add_add_fastgelu/gemm_add_add_fastgelu_xdl_fp16.cpp
..._gemm_add_add_fastgelu/gemm_add_add_fastgelu_xdl_fp16.cpp
+5
-5
example/06_conv2d_fwd_bias_relu/conv2d_fwd_bias_relu_xdl_fp16.cpp
...06_conv2d_fwd_bias_relu/conv2d_fwd_bias_relu_xdl_fp16.cpp
+5
-4
example/06_conv2d_fwd_bias_relu/convnd_fwd_bias_common.hpp
example/06_conv2d_fwd_bias_relu/convnd_fwd_bias_common.hpp
+4
-4
example/07_conv2d_fwd_bias_relu_add/conv2d_fwd_bias_relu_add_xdl_fp16.cpp
...d_fwd_bias_relu_add/conv2d_fwd_bias_relu_add_xdl_fp16.cpp
+6
-5
example/09_convnd_fwd/convnd_fwd_common.hpp
example/09_convnd_fwd/convnd_fwd_common.hpp
+4
-4
example/12_reduce/reduce_blockwise.cpp
example/12_reduce/reduce_blockwise.cpp
+3
-3
example/12_reduce/reduce_blockwise_two_call.cpp
example/12_reduce/reduce_blockwise_two_call.cpp
+4
-4
example/13_pool2d_fwd/pool2d_fwd_common.hpp
example/13_pool2d_fwd/pool2d_fwd_common.hpp
+4
-3
example/14_gemm_xdl_requant_relu_requant/gemm_xdl_requant_relu_requant_int8.cpp
...quant_relu_requant/gemm_xdl_requant_relu_requant_int8.cpp
+3
-3
example/15_grouped_gemm/grouped_gemm_xdl_fp16.cpp
example/15_grouped_gemm/grouped_gemm_xdl_fp16.cpp
+5
-5
example/16_gemm_reduce/gemm_reduce_xdl_max_fp16.cpp
example/16_gemm_reduce/gemm_reduce_xdl_max_fp16.cpp
+4
-4
No files found.
example/01_gemm/gemm_dl_fp16.cpp
View file @
65c56e56
...
...
@@ -142,9 +142,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_Sequential
<
1
>
{});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/01_gemm/gemm_dl_fp32.cpp
View file @
65c56e56
...
...
@@ -141,9 +141,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_Sequential
<
1
>
{});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/01_gemm/gemm_dl_int8.cpp
View file @
65c56e56
...
...
@@ -139,9 +139,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_Sequential
<
1
>
{});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/01_gemm/gemm_xdl_bf16.cpp
View file @
65c56e56
...
...
@@ -170,9 +170,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_3
<
BDataType
>
{
-
0.5
,
0.5
});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/01_gemm/gemm_xdl_fp16.cpp
View file @
65c56e56
...
...
@@ -155,9 +155,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_Sequential
<
1
>
{});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/01_gemm/gemm_xdl_fp64.cpp
View file @
65c56e56
...
...
@@ -165,9 +165,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_1
<
BDataType
>
{
1
});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/01_gemm/gemm_xdl_int8.cpp
View file @
65c56e56
...
...
@@ -167,9 +167,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_3
<
BDataType
>
{
-
0.5
,
0.5
});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/02_gemm_bilinear/gemm_bilinear_xdl_fp16.cpp
View file @
65c56e56
...
...
@@ -213,10 +213,10 @@ int main(int argc, char* argv[])
d_m_n
.
GenerateTensorValue
(
GeneratorTensor_3
<
DDataType
>
{
-
0.5
,
0.5
});
}
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
d_device_buf
(
sizeof
(
DDataType
)
*
d_m_n
.
mDesc
.
GetElementSpace
());
DeviceMem
e_device_buf
(
sizeof
(
EDataType
)
*
e_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
d_device_buf
(
sizeof
(
DDataType
)
*
d_m_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
e_device_buf
(
sizeof
(
EDataType
)
*
e_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/03_gemm_bias_relu/gemm_bias_relu_xdl_fp16.cpp
View file @
65c56e56
...
...
@@ -191,10 +191,10 @@ int main(int argc, char* argv[])
d_m_n
.
GenerateTensorValue
(
GeneratorTensor_3
<
DDataType
>
{
0.0
,
1.0
});
}
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
d_device_buf
(
sizeof
(
DDataType
)
*
d_m_n
.
mDesc
.
GetElementSpace
());
DeviceMem
e_device_buf
(
sizeof
(
EDataType
)
*
e_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
d_device_buf
(
sizeof
(
DDataType
)
*
d_m_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
e_device_buf
(
sizeof
(
EDataType
)
*
e_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/04_gemm_add_add_fastgelu/gemm_add_add_fastgelu_xdl_fp16.cpp
View file @
65c56e56
...
...
@@ -156,11 +156,11 @@ int main(int argc, char* argv[])
d1_m_n
.
GenerateTensorValue
(
GeneratorTensor_3
<
D1DataType
>
{
0.0
,
1.0
});
}
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
d0_device_buf
(
sizeof
(
D0DataType
)
*
d0_m_n
.
mDesc
.
GetElementSpace
());
DeviceMem
d1_device_buf
(
sizeof
(
D1DataType
)
*
d1_m_n
.
mDesc
.
GetElementSpace
());
DeviceMem
e_device_buf
(
sizeof
(
EDataType
)
*
e_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
d0_device_buf
(
sizeof
(
D0DataType
)
*
d0_m_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
d1_device_buf
(
sizeof
(
D1DataType
)
*
d1_m_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
e_device_buf
(
sizeof
(
EDataType
)
*
e_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/06_conv2d_fwd_bias_relu/conv2d_fwd_bias_relu_xdl_fp16.cpp
View file @
65c56e56
...
...
@@ -235,10 +235,11 @@ int main(int argc, char* argv[])
bias_k
.
GenerateTensorValue
(
GeneratorTensor_3
<
OutDataType
>
{
0.0
,
1.0
});
}
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in_n_hi_wi_c
.
mDesc
.
GetElementSpace
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei_k_y_x_c
.
mDesc
.
GetElementSpace
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias_k
.
mDesc
.
GetElementSpace
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_n_ho_wo_k_device
.
mDesc
.
GetElementSpace
());
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in_n_hi_wi_c
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei_k_y_x_c
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias_k
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_n_ho_wo_k_device
.
mDesc
.
GetElementSpaceSize
());
in_device_buf
.
ToDevice
(
in_n_hi_wi_c
.
mData
.
data
());
wei_device_buf
.
ToDevice
(
wei_k_y_x_c
.
mData
.
data
());
...
...
example/06_conv2d_fwd_bias_relu/convnd_fwd_bias_common.hpp
View file @
65c56e56
...
...
@@ -138,10 +138,10 @@ int run_conv_fwd_bias(bool do_verification,
bias
.
GenerateTensorValue
(
GeneratorTensor_3
<
OutDataType
>
{
-
0.5
,
0.5
});
}
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in
.
mDesc
.
GetElementSpace
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei
.
mDesc
.
GetElementSpace
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias
.
mDesc
.
GetElementSpace
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_device
.
mDesc
.
GetElementSpace
());
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_device
.
mDesc
.
GetElementSpace
Size
());
in_device_buf
.
ToDevice
(
in
.
mData
.
data
());
wei_device_buf
.
ToDevice
(
wei
.
mData
.
data
());
...
...
example/07_conv2d_fwd_bias_relu_add/conv2d_fwd_bias_relu_add_xdl_fp16.cpp
View file @
65c56e56
...
...
@@ -238,11 +238,12 @@ int main(int argc, char* argv[])
resi_n_ho_wo_k
.
GenerateTensorValue
(
GeneratorTensor_3
<
OutDataType
>
{
0.0
,
1.0
});
}
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in_n_hi_wi_c
.
mDesc
.
GetElementSpace
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei_k_y_x_c
.
mDesc
.
GetElementSpace
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias_k
.
mDesc
.
GetElementSpace
());
DeviceMem
resi_device_buf
(
sizeof
(
OutDataType
)
*
resi_n_ho_wo_k
.
mDesc
.
GetElementSpace
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_n_ho_wo_k_device
.
mDesc
.
GetElementSpace
());
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in_n_hi_wi_c
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei_k_y_x_c
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias_k
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
resi_device_buf
(
sizeof
(
OutDataType
)
*
resi_n_ho_wo_k
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_n_ho_wo_k_device
.
mDesc
.
GetElementSpaceSize
());
in_device_buf
.
ToDevice
(
in_n_hi_wi_c
.
mData
.
data
());
wei_device_buf
.
ToDevice
(
wei_k_y_x_c
.
mData
.
data
());
...
...
example/09_convnd_fwd/convnd_fwd_common.hpp
View file @
65c56e56
...
...
@@ -146,10 +146,10 @@ int run_conv_fwd(bool do_verification,
bias
.
GenerateTensorValue
(
GeneratorTensor_3
<
OutDataType
>
{
-
0.5
,
0.5
});
}
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in
.
mDesc
.
GetElementSpace
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei
.
mDesc
.
GetElementSpace
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias
.
mDesc
.
GetElementSpace
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_device
.
mDesc
.
GetElementSpace
());
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
wei_device_buf
(
sizeof
(
WeiDataType
)
*
wei
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
bias_device_buf
(
sizeof
(
OutDataType
)
*
bias
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_device
.
mDesc
.
GetElementSpace
Size
());
in_device_buf
.
ToDevice
(
in
.
mData
.
data
());
wei_device_buf
.
ToDevice
(
wei
.
mData
.
data
());
...
...
example/12_reduce/reduce_blockwise.cpp
View file @
65c56e56
...
...
@@ -230,13 +230,13 @@ int main(int argc, char* argv[])
}
if
(
beta
!=
0.0
f
)
for
(
size_t
i
=
0
;
i
<
out_ref
.
mDesc
.
GetElementSpace
();
i
++
)
for
(
size_t
i
=
0
;
i
<
out_ref
.
mDesc
.
GetElementSpace
Size
();
i
++
)
out
.
mData
[
i
]
=
out_ref
.
mData
[
i
];
};
// these buffers are usually provided by the user application
DeviceMem
in_dev
(
sizeof
(
InDataType
)
*
in
.
mDesc
.
GetElementSpace
());
DeviceMem
out_dev
(
sizeof
(
OutDataType
)
*
out
.
mDesc
.
GetElementSpace
());
DeviceMem
in_dev
(
sizeof
(
InDataType
)
*
in
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
out_dev
(
sizeof
(
OutDataType
)
*
out
.
mDesc
.
GetElementSpace
Size
());
in_dev
.
ToDevice
(
in
.
mData
.
data
());
...
...
example/12_reduce/reduce_blockwise_two_call.cpp
View file @
65c56e56
...
...
@@ -171,13 +171,13 @@ int main(int argc, char* argv[])
}
if
(
beta
!=
0.0
f
)
for
(
size_t
i
=
0
;
i
<
out_ref
.
mDesc
.
GetElementSpace
();
i
++
)
for
(
size_t
i
=
0
;
i
<
out_ref
.
mDesc
.
GetElementSpace
Size
();
i
++
)
out
.
mData
[
i
]
=
out_ref
.
mData
[
i
];
};
DeviceMem
in_1_dev
(
sizeof
(
InOutDataType
)
*
in_1
.
mDesc
.
GetElementSpace
());
DeviceMem
in_2_dev
(
sizeof
(
InOutDataType
)
*
in_2
.
mDesc
.
GetElementSpace
());
DeviceMem
out_dev
(
sizeof
(
InOutDataType
)
*
out
.
mDesc
.
GetElementSpace
());
DeviceMem
in_1_dev
(
sizeof
(
InOutDataType
)
*
in_1
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
in_2_dev
(
sizeof
(
InOutDataType
)
*
in_2
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
out_dev
(
sizeof
(
InOutDataType
)
*
out
.
mDesc
.
GetElementSpace
Size
());
in_1_dev
.
ToDevice
(
in_1
.
mData
.
data
());
...
...
example/13_pool2d_fwd/pool2d_fwd_common.hpp
View file @
65c56e56
...
...
@@ -204,10 +204,11 @@ bool pool_test(bool do_verification,
default:
in_n_c_hi_wi
.
GenerateTensorValue
(
GeneratorTensor_3
<
InDataType
>
{
-
5.0
,
5.0
});
}
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in_n_c_hi_wi
.
mDesc
.
GetElementSpace
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_n_c_ho_wo_device
.
mDesc
.
GetElementSpace
());
DeviceMem
in_device_buf
(
sizeof
(
InDataType
)
*
in_n_c_hi_wi
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
out_device_buf
(
sizeof
(
OutDataType
)
*
out_n_c_ho_wo_device
.
mDesc
.
GetElementSpaceSize
());
DeviceMem
out_indices_device_buf
(
sizeof
(
IndexDataType
)
*
out_indices_n_c_ho_wo_device
.
mDesc
.
GetElementSpace
());
out_indices_n_c_ho_wo_device
.
mDesc
.
GetElementSpace
Size
());
in_device_buf
.
ToDevice
(
in_n_c_hi_wi
.
mData
.
data
());
...
...
example/14_gemm_xdl_requant_relu_requant/gemm_xdl_requant_relu_requant_int8.cpp
View file @
65c56e56
...
...
@@ -190,9 +190,9 @@ int main(int argc, char* argv[])
b_k_n
.
GenerateTensorValue
(
GeneratorTensor_3
<
BDataType
>
{
-
0.5
,
0.5
});
}
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_m_k_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_k_n_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_m_n_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
a_m_k_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_k_n_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
example/15_grouped_gemm/grouped_gemm_xdl_fp16.cpp
View file @
65c56e56
...
...
@@ -170,12 +170,12 @@ int main(int argc, char* argv[])
for
(
std
::
size_t
i
=
0
;
i
<
gemm_shapes
.
size
();
i
++
)
{
a_tensors_device
.
emplace_back
(
std
::
make_unique
<
DeviceMem
>
(
sizeof
(
ADataType
)
*
a_tensors
[
i
].
mDesc
.
GetElementSpace
()));
b_tensors_device
.
emplace_back
(
std
::
make_unique
<
DeviceMem
>
(
sizeof
(
BDataType
)
*
b_tensors
[
i
].
mDesc
.
GetElementSpace
()));
a_tensors_device
.
emplace_back
(
std
::
make_unique
<
DeviceMem
>
(
sizeof
(
ADataType
)
*
a_tensors
[
i
].
mDesc
.
GetElementSpace
Size
()));
b_tensors_device
.
emplace_back
(
std
::
make_unique
<
DeviceMem
>
(
sizeof
(
BDataType
)
*
b_tensors
[
i
].
mDesc
.
GetElementSpace
Size
()));
c_tensors_device
.
emplace_back
(
std
::
make_unique
<
DeviceMem
>
(
sizeof
(
CDataType
)
*
c_device_tensors
[
i
].
mDesc
.
GetElementSpace
()));
sizeof
(
CDataType
)
*
c_device_tensors
[
i
].
mDesc
.
GetElementSpace
Size
()));
a_tensors_device
[
i
]
->
ToDevice
(
a_tensors
[
i
].
mData
.
data
());
b_tensors_device
[
i
]
->
ToDevice
(
b_tensors
[
i
].
mData
.
data
());
...
...
example/16_gemm_reduce/gemm_reduce_xdl_max_fp16.cpp
View file @
65c56e56
...
...
@@ -173,11 +173,11 @@ int main(int argc, char* argv[])
break
;
}
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
());
DeviceMem
c_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
());
DeviceMem
a_device_buf
(
sizeof
(
ADataType
)
*
a_m_k
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
b_device_buf
(
sizeof
(
BDataType
)
*
b_k_n
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
c_device_buf
(
sizeof
(
CDataType
)
*
c_m_n_device_result
.
mDesc
.
GetElementSpace
Size
());
DeviceMem
reduce_device_buf
(
sizeof
(
ReduceDataType
)
*
reduce_m_device_result
.
mDesc
.
GetElementSpace
());
reduce_m_device_result
.
mDesc
.
GetElementSpace
Size
());
a_device_buf
.
ToDevice
(
a_m_k
.
mData
.
data
());
b_device_buf
.
ToDevice
(
b_k_n
.
mData
.
data
());
...
...
Prev
1
2
3
Next
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