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
8159be33
Commit
8159be33
authored
Dec 06, 2021
by
Chao Liu
Browse files
adding conv+bias+relu
parent
29c6b47c
Changes
4
Expand all
Show whitespace changes
Inline
Side-by-side
Showing
4 changed files
with
1187 additions
and
5 deletions
+1187
-5
composable_kernel/include/tensor_operation/gridwise_gemm_xdlops_v2r6.hpp
...el/include/tensor_operation/gridwise_gemm_xdlops_v2r6.hpp
+655
-0
composable_kernel/include/tensor_operation/threadwise_tensor_slice_transfer_v1r5.hpp
...ensor_operation/threadwise_tensor_slice_transfer_v1r5.hpp
+522
-0
composable_kernel/include/utility/config.hpp
composable_kernel/include/utility/config.hpp
+5
-0
device_operation/include/device_conv2d_fwd_xdl_bias_activation_nhwc_kyxc_nhwk.hpp
.../device_conv2d_fwd_xdl_bias_activation_nhwc_kyxc_nhwk.hpp
+5
-5
No files found.
composable_kernel/include/tensor_operation/gridwise_gemm_xdlops_v2r6.hpp
0 → 100644
View file @
8159be33
This diff is collapsed.
Click to expand it.
composable_kernel/include/tensor_operation/threadwise_tensor_slice_transfer_v1r5.hpp
0 → 100644
View file @
8159be33
This diff is collapsed.
Click to expand it.
composable_kernel/include/utility/config.hpp
View file @
8159be33
...
@@ -141,6 +141,11 @@
...
@@ -141,6 +141,11 @@
#define CK_WORKAROUND_SWDEV_XXXXXX_THREAD_WISE_COPY_V1R4_TYPE_CONVERT_ISSUE 1
#define CK_WORKAROUND_SWDEV_XXXXXX_THREAD_WISE_COPY_V1R4_TYPE_CONVERT_ISSUE 1
#endif
#endif
#ifndef CK_WORKAROUND_SWDEV_XXXXXX_THREAD_WISE_COPY_V1R5_TYPE_CONVERT_ISSUE
#define CK_WORKAROUND_SWDEV_XXXXXX_THREAD_WISE_COPY_V1R5_TYPE_CONVERT_ISSUE 1
#endif
namespace
ck
{
namespace
ck
{
enum
InMemoryDataOperationEnum_t
enum
InMemoryDataOperationEnum_t
...
...
device_operation/include/device_conv2d_fwd_xdl_bias_activation_nhwc_kyxc_nhwk.hpp
View file @
8159be33
...
@@ -10,7 +10,7 @@
...
@@ -10,7 +10,7 @@
#include "tensor_layout.hpp"
#include "tensor_layout.hpp"
#include "tensor_descriptor.hpp"
#include "tensor_descriptor.hpp"
#include "tensor_descriptor_helper.hpp"
#include "tensor_descriptor_helper.hpp"
#include "gridwise_gemm_xdlops_v2r
5
.hpp"
#include "gridwise_gemm_xdlops_v2r
6
.hpp"
namespace
ck
{
namespace
ck
{
namespace
tensor_operation
{
namespace
tensor_operation
{
...
@@ -269,7 +269,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
...
@@ -269,7 +269,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
static
constexpr
auto
b_k0_n_k1_grid_move_slice_window_step_hacks
=
Sequence
<
0
,
0
,
0
,
0
,
0
>
{};
static
constexpr
auto
b_k0_n_k1_grid_move_slice_window_step_hacks
=
Sequence
<
0
,
0
,
0
,
0
,
0
>
{};
// GridwiseGemm
// GridwiseGemm
using
GridwiseGemm
=
GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r
5
<
using
GridwiseGemm
=
GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r
6
<
BlockSize
,
BlockSize
,
ABDataType
,
// TODO: distinguish A/B datatype
ABDataType
,
// TODO: distinguish A/B datatype
AccDataType
,
AccDataType
,
...
@@ -462,7 +462,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
...
@@ -462,7 +462,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
arg
.
N01_
))
arg
.
N01_
))
{
{
throw
std
::
runtime_error
(
throw
std
::
runtime_error
(
"wrong! GridwiseGemm_km_kn_m0m1n0n1_xdlops_v2r
5
has invalid setting"
);
"wrong! GridwiseGemm_km_kn_m0m1n0n1_xdlops_v2r
6
has invalid setting"
);
}
}
const
index_t
grid_size
=
GridwiseGemm
::
CalculateGridSize
(
arg
.
c_grid_desc_m_n_
);
const
index_t
grid_size
=
GridwiseGemm
::
CalculateGridSize
(
arg
.
c_grid_desc_m_n_
);
...
@@ -475,7 +475,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
...
@@ -475,7 +475,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
if
(
has_main_k0_block_loop
)
if
(
has_main_k0_block_loop
)
{
{
const
auto
kernel
=
kernel_gemm_xdlops_v2r
5
<
const
auto
kernel
=
kernel_gemm_xdlops_v2r
6
<
GridwiseGemm
,
GridwiseGemm
,
ADataType
,
// TODO: distiguish A/B datatype
ADataType
,
// TODO: distiguish A/B datatype
CDataType
,
CDataType
,
...
@@ -512,7 +512,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
...
@@ -512,7 +512,7 @@ struct DeviceConv2dFwdXdl_Bias_Activation_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
}
}
else
else
{
{
const
auto
kernel
=
kernel_gemm_xdlops_v2r
5
<
const
auto
kernel
=
kernel_gemm_xdlops_v2r
6
<
GridwiseGemm
,
GridwiseGemm
,
ADataType
,
// TODO: distiguish A/B datatype
ADataType
,
// TODO: distiguish A/B datatype
CDataType
,
CDataType
,
...
...
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