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
16f02f76
Commit
16f02f76
authored
Nov 18, 2022
by
Astha Rai
Browse files
Merge branch 'gridwise_2d' of github.com:ROCmSoftwarePlatform/composable_kernel into gridwise_2d
parents
7d653017
9b3365e1
Changes
187
Hide whitespace changes
Inline
Side-by-side
Showing
20 changed files
with
216 additions
and
1067 deletions
+216
-1067
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp
...or_operation_instance/gpu/grouped_convolution_forward.hpp
+2
-2
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_dl.hpp
...operation_instance/gpu/grouped_convolution_forward_dl.hpp
+1
-3
library/include/ck/library/utility/algorithm.hpp
library/include/ck/library/utility/algorithm.hpp
+43
-0
library/include/ck/library/utility/check_err.hpp
library/include/ck/library/utility/check_err.hpp
+46
-43
library/include/ck/library/utility/fill.hpp
library/include/ck/library/utility/fill.hpp
+14
-3
library/include/ck/library/utility/host_tensor.hpp
library/include/ck/library/utility/host_tensor.hpp
+28
-26
library/include/ck/library/utility/iterator.hpp
library/include/ck/library/utility/iterator.hpp
+22
-0
library/include/ck/library/utility/ranges.hpp
library/include/ck/library/utility/ranges.hpp
+60
-0
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/CMakeLists.txt
...r_operation_instance/gpu/conv1d_bwd_weight/CMakeLists.txt
+0
-5
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_bf16_instance.cpp
...evice_conv1d_bwd_weight_xdl_nwc_kxc_nwk_bf16_instance.cpp
+0
-102
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f16_instance.cpp
...device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f16_instance.cpp
+0
-102
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f32_instance.cpp
...device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f32_instance.cpp
+0
-101
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/CMakeLists.txt
...r_operation_instance/gpu/conv2d_bwd_weight/CMakeLists.txt
+0
-6
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_bf16_instance.cpp
...ce_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_bf16_instance.cpp
+0
-102
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instance.cpp
...ice_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instance.cpp
+0
-130
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instance.cpp
...ice_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instance.cpp
+0
-128
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/CMakeLists.txt
...r_operation_instance/gpu/conv3d_bwd_weight/CMakeLists.txt
+0
-5
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_bf16_instance.cpp
...conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_bf16_instance.cpp
+0
-104
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f16_instance.cpp
..._conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f16_instance.cpp
+0
-103
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f32_instance.cpp
..._conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f32_instance.cpp
+0
-102
No files found.
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp
View file @
16f02f76
...
...
@@ -3,11 +3,11 @@
#pragma once
#include <
cstdlib
>
#include <
vector
>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_fwd.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
...
...
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_dl.hpp
View file @
16f02f76
...
...
@@ -3,11 +3,9 @@
#pragma once
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_fwd.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
...
...
library/include/ck/library/utility/algorithm.hpp
0 → 100644
View file @
16f02f76
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <algorithm>
#include <iterator>
#include <type_traits>
#include <utility>
namespace
ck
{
namespace
ranges
{
template
<
typename
InputRange
,
typename
OutputIterator
>
auto
copy
(
InputRange
&&
range
,
OutputIterator
iter
)
->
decltype
(
std
::
copy
(
std
::
begin
(
std
::
forward
<
InputRange
>
(
range
)),
std
::
end
(
std
::
forward
<
InputRange
>
(
range
)),
iter
))
{
return
std
::
copy
(
std
::
begin
(
std
::
forward
<
InputRange
>
(
range
)),
std
::
end
(
std
::
forward
<
InputRange
>
(
range
)),
iter
);
}
template
<
typename
T
,
typename
OutputRange
>
auto
fill
(
OutputRange
&&
range
,
const
T
&
init
)
->
std
::
void_t
<
decltype
(
std
::
fill
(
std
::
begin
(
std
::
forward
<
OutputRange
>
(
range
)),
std
::
end
(
std
::
forward
<
OutputRange
>
(
range
)),
init
))
>
{
std
::
fill
(
std
::
begin
(
std
::
forward
<
OutputRange
>
(
range
)),
std
::
end
(
std
::
forward
<
OutputRange
>
(
range
)),
init
);
}
template
<
typename
InputRange
,
typename
OutputIterator
,
typename
UnaryOperation
>
auto
transform
(
InputRange
&&
range
,
OutputIterator
iter
,
UnaryOperation
unary_op
)
->
decltype
(
std
::
transform
(
std
::
begin
(
range
),
std
::
end
(
range
),
iter
,
unary_op
))
{
return
std
::
transform
(
std
::
begin
(
range
),
std
::
end
(
range
),
iter
,
unary_op
);
}
}
// namespace ranges
}
// namespace ck
library/include/ck/library/utility/check_err.hpp
View file @
16f02f76
...
...
@@ -15,18 +15,22 @@
#include "ck/ck.hpp"
#include "ck/utility/data_type.hpp"
#include "ck/utility/span.hpp"
#include "ck/utility/type.hpp"
#include "ck/host_utility/io.hpp"
#include "ck/library/utility/ranges.hpp"
namespace
ck
{
namespace
utils
{
template
<
typename
T
>
typename
std
::
enable_if
<
std
::
is_floating_point
<
T
>::
value
&&
!
std
::
is_same
<
T
,
half_t
>::
value
,
bool
>::
type
check_err
(
const
std
::
vector
<
T
>&
out
,
const
std
::
vector
<
T
>&
ref
,
template
<
typename
Range
,
typename
RefRange
>
typename
std
::
enable_if
<
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
ranges
::
range_value_t
<
RefRange
>>
&&
std
::
is_floating_point_v
<
ranges
::
range_value_t
<
Range
>>
&&
!
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
half_t
>
,
bool
>::
type
check_err
(
const
Range
&
out
,
const
RefRange
&
ref
,
const
std
::
string
&
msg
=
"Error: Incorrect results!"
,
double
rtol
=
1e-5
,
double
atol
=
3e-6
)
...
...
@@ -44,15 +48,17 @@ check_err(const std::vector<T>& out,
double
max_err
=
std
::
numeric_limits
<
double
>::
min
();
for
(
std
::
size_t
i
=
0
;
i
<
ref
.
size
();
++
i
)
{
err
=
std
::
abs
(
out
[
i
]
-
ref
[
i
]);
if
(
err
>
atol
+
rtol
*
std
::
abs
(
ref
[
i
])
||
!
std
::
isfinite
(
out
[
i
])
||
!
std
::
isfinite
(
ref
[
i
]))
const
double
o
=
*
std
::
next
(
std
::
begin
(
out
),
i
);
const
double
r
=
*
std
::
next
(
std
::
begin
(
ref
),
i
);
err
=
std
::
abs
(
o
-
r
);
if
(
err
>
atol
+
rtol
*
std
::
abs
(
r
)
||
!
std
::
isfinite
(
o
)
||
!
std
::
isfinite
(
r
))
{
max_err
=
err
>
max_err
?
err
:
max_err
;
err_count
++
;
if
(
err_count
<
5
)
{
std
::
cerr
<<
msg
<<
std
::
setw
(
12
)
<<
std
::
setprecision
(
7
)
<<
" out["
<<
i
<<
"] != ref["
<<
i
<<
"]: "
<<
o
ut
[
i
]
<<
" != "
<<
r
ef
[
i
]
<<
std
::
endl
;
<<
"] != ref["
<<
i
<<
"]: "
<<
o
<<
" != "
<<
r
<<
std
::
endl
;
}
res
=
false
;
}
...
...
@@ -64,10 +70,13 @@ check_err(const std::vector<T>& out,
return
res
;
}
template
<
typename
T
>
typename
std
::
enable_if
<
std
::
is_same
<
T
,
bhalf_t
>::
value
,
bool
>::
type
check_err
(
const
std
::
vector
<
T
>&
out
,
const
std
::
vector
<
T
>&
ref
,
template
<
typename
Range
,
typename
RefRange
>
typename
std
::
enable_if
<
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
ranges
::
range_value_t
<
RefRange
>>
&&
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
bhalf_t
>
,
bool
>::
type
check_err
(
const
Range
&
out
,
const
RefRange
&
ref
,
const
std
::
string
&
msg
=
"Error: Incorrect results!"
,
double
rtol
=
1e-3
,
double
atol
=
1e-3
)
...
...
@@ -86,9 +95,9 @@ check_err(const std::vector<T>& out,
double
max_err
=
std
::
numeric_limits
<
float
>::
min
();
for
(
std
::
size_t
i
=
0
;
i
<
ref
.
size
();
++
i
)
{
double
o
=
type_convert
<
float
>
(
out
[
i
]
);
double
r
=
type_convert
<
float
>
(
ref
[
i
]
);
err
=
std
::
abs
(
o
-
r
);
const
double
o
=
type_convert
<
float
>
(
*
std
::
next
(
std
::
begin
(
out
),
i
)
);
const
double
r
=
type_convert
<
float
>
(
*
std
::
next
(
std
::
begin
(
ref
),
i
)
);
err
=
std
::
abs
(
o
-
r
);
if
(
err
>
atol
+
rtol
*
std
::
abs
(
r
)
||
!
std
::
isfinite
(
o
)
||
!
std
::
isfinite
(
r
))
{
max_err
=
err
>
max_err
?
err
:
max_err
;
...
...
@@ -108,10 +117,13 @@ check_err(const std::vector<T>& out,
return
res
;
}
template
<
typename
T
>
typename
std
::
enable_if
<
std
::
is_same_v
<
T
,
half_t
>
,
bool
>::
type
check_err
(
span
<
const
T
>
out
,
span
<
const
T
>
ref
,
template
<
typename
Range
,
typename
RefRange
>
typename
std
::
enable_if
<
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
ranges
::
range_value_t
<
RefRange
>>
&&
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
half_t
>
,
bool
>::
type
check_err
(
const
Range
&
out
,
const
RefRange
&
ref
,
const
std
::
string
&
msg
=
"Error: Incorrect results!"
,
double
rtol
=
1e-3
,
double
atol
=
1e-3
)
...
...
@@ -126,12 +138,12 @@ check_err(span<const T> out,
bool
res
{
true
};
int
err_count
=
0
;
double
err
=
0
;
double
max_err
=
std
::
numeric_limits
<
T
>::
min
();
double
max_err
=
std
::
numeric_limits
<
ranges
::
range_value_t
<
Range
>
>::
min
();
for
(
std
::
size_t
i
=
0
;
i
<
ref
.
size
();
++
i
)
{
double
o
=
type_convert
<
float
>
(
out
[
i
]
);
double
r
=
type_convert
<
float
>
(
ref
[
i
]
);
err
=
std
::
abs
(
o
-
r
);
const
double
o
=
type_convert
<
float
>
(
*
std
::
next
(
std
::
begin
(
out
),
i
)
);
const
double
r
=
type_convert
<
float
>
(
*
std
::
next
(
std
::
begin
(
ref
),
i
)
);
err
=
std
::
abs
(
o
-
r
);
if
(
err
>
atol
+
rtol
*
std
::
abs
(
r
)
||
!
std
::
isfinite
(
o
)
||
!
std
::
isfinite
(
r
))
{
max_err
=
err
>
max_err
?
err
:
max_err
;
...
...
@@ -151,26 +163,17 @@ check_err(span<const T> out,
return
res
;
}
template
<
typename
T
>
typename
std
::
enable_if
<
std
::
is_same
<
T
,
half_t
>::
value
,
bool
>::
type
check_err
(
const
std
::
vector
<
T
>&
out
,
const
std
::
vector
<
T
>&
ref
,
const
std
::
string
&
msg
=
"Error: Incorrect results!"
,
double
rtol
=
1e-3
,
double
atol
=
1e-3
)
{
return
check_err
(
span
<
const
T
>
{
out
},
span
<
const
T
>
{
ref
},
msg
,
rtol
,
atol
);
}
template
<
typename
T
>
std
::
enable_if_t
<
(
std
::
is_integral_v
<
T
>
&&
!
std
::
is_same_v
<
T
,
bhalf_t
>
)
template
<
typename
Range
,
typename
RefRange
>
std
::
enable_if_t
<
(
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
ranges
::
range_value_t
<
RefRange
>>
&&
std
::
is_integral_v
<
ranges
::
range_value_t
<
Range
>>
&&
!
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
bhalf_t
>
)
#ifdef CK_EXPERIMENTAL_BIT_INT_EXTENSION_INT4
||
std
::
is_same_v
<
T
,
int4_t
>
||
std
::
is_same_v
<
ranges
::
range_value_t
<
Range
>
,
int4_t
>
#endif
,
bool
>
check_err
(
const
std
::
vector
<
T
>
&
out
,
const
std
::
vector
<
T
>
&
ref
,
check_err
(
const
Range
&
out
,
const
RefRange
&
ref
,
const
std
::
string
&
msg
=
"Error: Incorrect results!"
,
double
=
0
,
double
atol
=
0
)
...
...
@@ -188,9 +191,9 @@ check_err(const std::vector<T>& out,
int64_t
max_err
=
std
::
numeric_limits
<
int64_t
>::
min
();
for
(
std
::
size_t
i
=
0
;
i
<
ref
.
size
();
++
i
)
{
int64_t
o
=
out
[
i
]
;
int64_t
r
=
ref
[
i
]
;
err
=
std
::
abs
(
o
-
r
);
const
int64_t
o
=
*
std
::
next
(
std
::
begin
(
out
),
i
)
;
const
int64_t
r
=
*
std
::
next
(
std
::
begin
(
ref
),
i
)
;
err
=
std
::
abs
(
o
-
r
);
if
(
err
>
atol
)
{
...
...
library/include/ck/library/utility/fill.hpp
View file @
16f02f76
...
...
@@ -30,9 +30,10 @@ struct FillUniformDistribution
}
template
<
typename
ForwardRange
>
auto
operator
()(
ForwardRange
&&
range
)
->
std
::
void_t
<
decltype
(
std
::
declval
<
FillUniformDistribution
>
()(
std
::
begin
(
std
::
forward
<
ForwardRange
>
(
range
)),
std
::
end
(
std
::
forward
<
ForwardRange
>
(
range
))))
>
auto
operator
()(
ForwardRange
&&
range
)
const
->
std
::
void_t
<
decltype
(
std
::
declval
<
const
FillUniformDistribution
&>
()(
std
::
begin
(
std
::
forward
<
ForwardRange
>
(
range
)),
std
::
end
(
std
::
forward
<
ForwardRange
>
(
range
))))
>
{
(
*
this
)(
std
::
begin
(
std
::
forward
<
ForwardRange
>
(
range
)),
std
::
end
(
std
::
forward
<
ForwardRange
>
(
range
)));
...
...
@@ -72,6 +73,16 @@ struct FillUniformDistributionIntegerValue
std
::
generate
(
first
,
last
,
[
&
dis
,
&
gen
]()
{
return
ck
::
type_convert
<
T
>
(
std
::
round
(
dis
(
gen
)));
});
}
template
<
typename
ForwardRange
>
auto
operator
()(
ForwardRange
&&
range
)
const
->
std
::
void_t
<
decltype
(
std
::
declval
<
const
FillUniformDistributionIntegerValue
&>
()(
std
::
begin
(
std
::
forward
<
ForwardRange
>
(
range
)),
std
::
end
(
std
::
forward
<
ForwardRange
>
(
range
))))
>
{
(
*
this
)(
std
::
begin
(
std
::
forward
<
ForwardRange
>
(
range
)),
std
::
end
(
std
::
forward
<
ForwardRange
>
(
range
)));
}
};
template
<
typename
T
>
...
...
library/include/ck/library/utility/host_tensor.hpp
View file @
16f02f76
...
...
@@ -14,6 +14,9 @@
#include "ck/utility/data_type.hpp"
#include "ck/utility/span.hpp"
#include "ck/library/utility/algorithm.hpp"
#include "ck/library/utility/ranges.hpp"
template
<
typename
Range
>
std
::
ostream
&
LogRange
(
std
::
ostream
&
os
,
Range
&&
range
,
std
::
string
delim
)
{
...
...
@@ -84,10 +87,10 @@ struct HostTensorDescriptor
this
->
CalculateStrides
();
}
template
<
typename
Range
,
template
<
typename
Lengths
,
typename
=
std
::
enable_if_t
<
std
::
is_convertible_v
<
decltype
(
*
std
::
begin
(
std
::
declval
<
Range
>()))
,
std
::
size_t
>>>
HostTensorDescriptor
(
const
Range
&
lens
)
:
mLens
(
lens
.
begin
(),
lens
.
end
())
std
::
is_convertible_v
<
ck
::
ranges
::
range_value_t
<
Lengths
>
,
std
::
size_t
>>>
HostTensorDescriptor
(
const
Lengths
&
lens
)
:
mLens
(
lens
.
begin
(),
lens
.
end
())
{
this
->
CalculateStrides
();
}
...
...
@@ -102,13 +105,12 @@ struct HostTensorDescriptor
{
}
template
<
typename
Range1
,
typename
Range2
,
typename
=
std
::
enable_if_t
<
std
::
is_convertible_v
<
decltype
(
*
std
::
begin
(
std
::
declval
<
Range1
>())),
std
::
size_t
>
&&
std
::
is_convertible_v
<
decltype
(
*
std
::
begin
(
std
::
declval
<
Range2
>
())),
std
::
size_t
>>>
HostTensorDescriptor
(
const
Range1
&
lens
,
const
Range2
&
strides
)
template
<
typename
Lengths
,
typename
Strides
,
typename
=
std
::
enable_if_t
<
std
::
is_convertible_v
<
ck
::
ranges
::
range_value_t
<
Lengths
>,
std
::
size_t
>
&&
std
::
is_convertible_v
<
ck
::
ranges
::
range_value_t
<
Strides
>
,
std
::
size_t
>>>
HostTensorDescriptor
(
const
Lengths
&
lens
,
const
Strides
&
strides
)
:
mLens
(
lens
.
begin
(),
lens
.
end
()),
mStrides
(
strides
.
begin
(),
strides
.
end
())
{
}
...
...
@@ -244,14 +246,20 @@ struct Tensor
{
}
template
<
typename
X
>
Tensor
(
std
::
vector
<
X
>
lens
)
:
mDesc
(
lens
),
mData
(
mDesc
.
GetElementSpaceSize
())
template
<
typename
X
,
typename
Y
>
Tensor
(
std
::
initializer_list
<
X
>
lens
,
std
::
initializer_list
<
Y
>
strides
)
:
mDesc
(
lens
,
strides
),
mData
(
mDesc
.
GetElementSpaceSize
())
{
}
template
<
typename
X
,
typename
Y
>
Tensor
(
std
::
vector
<
X
>
lens
,
std
::
vector
<
Y
>
strides
)
:
mDesc
(
lens
,
strides
),
mData
(
mDesc
.
GetElementSpaceSize
())
template
<
typename
Lengths
>
Tensor
(
const
Lengths
&
lens
)
:
mDesc
(
lens
),
mData
(
mDesc
.
GetElementSpaceSize
())
{
}
template
<
typename
Lengths
,
typename
Strides
>
Tensor
(
const
Lengths
&
lens
,
const
Strides
&
strides
)
:
mDesc
(
lens
,
strides
),
mData
(
GetElementSpaceSize
())
{
}
...
...
@@ -261,10 +269,10 @@ struct Tensor
Tensor
<
OutT
>
CopyAsType
()
const
{
Tensor
<
OutT
>
ret
(
mDesc
);
for
(
size_t
i
=
0
;
i
<
mData
.
size
();
i
++
)
{
ret
.
mData
[
i
]
=
ck
::
type_convert
<
OutT
>
(
mData
[
i
]
);
}
ck
::
ranges
::
transform
(
mData
,
ret
.
mData
.
begin
(),
[](
auto
value
)
{
return
ck
::
type_convert
<
OutT
>
(
value
);
}
);
return
ret
;
}
...
...
@@ -294,13 +302,7 @@ struct Tensor
std
::
size_t
GetElementSpaceSizeInBytes
()
const
{
return
sizeof
(
T
)
*
GetElementSpaceSize
();
}
void
SetZero
()
{
for
(
auto
&
v
:
mData
)
{
v
=
T
{
0
};
}
}
void
SetZero
()
{
ck
::
ranges
::
fill
<
T
>
(
mData
,
0
);
}
template
<
typename
F
>
void
ForEach_impl
(
F
&&
f
,
std
::
vector
<
size_t
>&
idx
,
size_t
rank
)
...
...
library/include/ck/library/utility/iterator.hpp
0 → 100644
View file @
16f02f76
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iterator>
#include <utility>
#include "ck/utility/type.hpp"
namespace
ck
{
template
<
typename
T
>
using
iter_value_t
=
typename
std
::
iterator_traits
<
remove_cvref_t
<
T
>>::
value_type
;
template
<
typename
T
>
using
iter_reference_t
=
decltype
(
*
std
::
declval
<
T
&>
());
template
<
typename
T
>
using
iter_difference_t
=
typename
std
::
iterator_traits
<
remove_cvref_t
<
T
>>::
difference_type
;
}
// namespace ck
library/include/ck/library/utility/ranges.hpp
0 → 100644
View file @
16f02f76
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iterator>
#include <type_traits>
#include <utility>
#include "ck/library/utility/iterator.hpp"
namespace
ck
{
namespace
ranges
{
template
<
typename
R
>
using
iterator_t
=
decltype
(
std
::
begin
(
std
::
declval
<
R
&>
()));
template
<
typename
R
>
using
sentinel_t
=
decltype
(
std
::
end
(
std
::
declval
<
R
&>
()));
template
<
typename
R
>
using
range_size_t
=
decltype
(
std
::
size
(
std
::
declval
<
R
&>
()));
template
<
typename
R
>
using
range_difference_t
=
ck
::
iter_difference_t
<
ranges
::
iterator_t
<
R
>>
;
template
<
typename
R
>
using
range_value_t
=
iter_value_t
<
ranges
::
iterator_t
<
R
>>
;
template
<
typename
R
>
using
range_reference_t
=
iter_reference_t
<
ranges
::
iterator_t
<
R
>>
;
template
<
typename
T
,
typename
=
void
>
struct
is_range
:
std
::
false_type
{
};
template
<
typename
T
>
struct
is_range
<
T
,
std
::
void_t
<
decltype
(
std
::
begin
(
std
::
declval
<
T
&>
())),
decltype
(
std
::
end
(
std
::
declval
<
T
&>
()))
>>
:
std
::
true_type
{
};
template
<
typename
T
>
inline
constexpr
bool
is_range_v
=
is_range
<
T
>::
value
;
template
<
typename
T
,
typename
=
void
>
struct
is_sized_range
:
std
::
false_type
{
};
template
<
typename
T
>
struct
is_sized_range
<
T
,
std
::
void_t
<
decltype
(
std
::
size
(
std
::
declval
<
T
&>
()))
>>
:
std
::
bool_constant
<
is_range_v
<
T
>>
{
};
}
// namespace ranges
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/CMakeLists.txt
deleted
100644 → 0
View file @
7d653017
add_instance_library
(
device_conv1d_bwd_weight_instance
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f16_instance.cpp
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f32_instance.cpp
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_bf16_instance.cpp
)
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_bf16_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
BF16
=
bhalf_t
;
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NWC
=
ck
::
tensor_layout
::
convolution
::
NWC
;
using
KXC
=
ck
::
tensor_layout
::
convolution
::
KXC
;
using
NWK
=
ck
::
tensor_layout
::
convolution
::
NWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, wi, c] * wei[k, x, c] = out[n, wo, k]
using
device_conv1d_bwd_weight_xdl_c_shuffle_nwc_kxc_nwk_bf16_f32_bf16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
using
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_1x1_s1_p0_bf16_f32_bf16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
void
add_device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_bf16_f32_bf16_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
1
,
NWC
,
KXC
,
NWK
,
BF16
,
F32
,
BF16
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv1d_bwd_weight_xdl_c_shuffle_nwc_kxc_nwk_bf16_f32_bf16_instances
{});
add_device_operation_instances
(
instances
,
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_1x1_s1_p0_bf16_f32_bf16_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f16_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
F16
=
ck
::
half_t
;
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NWC
=
ck
::
tensor_layout
::
convolution
::
NWC
;
using
KXC
=
ck
::
tensor_layout
::
convolution
::
KXC
;
using
NWK
=
ck
::
tensor_layout
::
convolution
::
NWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, wi, c] * wei[k, x, c] = out[n, wo, k]
using
device_conv1d_bwd_weight_xdl_c_shuffle_nwc_kxc_nwk_f16_default_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
// clang-format on
>
;
using
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_1x1_s1_p0_f16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
// clang-format on
>
;
void
add_device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f16_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
1
,
NWC
,
KXC
,
NWK
,
F16
,
F16
,
F16
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv1d_bwd_weight_xdl_c_shuffle_nwc_kxc_nwk_f16_default_instances
{});
add_device_operation_instances
(
instances
,
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_1x1_s1_p0_f16_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv1d_bwd_weight/device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f32_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NWC
=
ck
::
tensor_layout
::
convolution
::
NWC
;
using
KXC
=
ck
::
tensor_layout
::
convolution
::
KXC
;
using
NWK
=
ck
::
tensor_layout
::
convolution
::
NWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, wi, c] * wei[k, x, c] = out[n, wo, k]
using
device_conv1d_bwd_weight_xdl_c_shuffle_nwc_kxc_nwk_f32_default_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
using
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_1x1_s1_p0_f32_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
1
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
void
add_device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_f32_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
1
,
NWC
,
KXC
,
NWK
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv1d_bwd_weight_xdl_c_shuffle_nwc_kxc_nwk_f32_default_instances
{});
add_device_operation_instances
(
instances
,
device_conv1d_bwd_weight_xdl_nwc_kxc_nwk_1x1_s1_p0_f32_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/CMakeLists.txt
deleted
100644 → 0
View file @
7d653017
add_instance_library
(
device_conv2d_bwd_weight_instance
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instance.cpp
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instance.cpp
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_bf16_instance.cpp
)
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_bf16_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
BF16
=
bhalf_t
;
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NHWC
=
ck
::
tensor_layout
::
convolution
::
NHWC
;
using
KYXC
=
ck
::
tensor_layout
::
convolution
::
KYXC
;
using
NHWK
=
ck
::
tensor_layout
::
convolution
::
NHWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
using
device_conv2d_bwd_weight_xdl_c_shuffle_nhwc_kyxc_nhwk_bf16_f32_bf16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
using
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_f32_bf16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
void
add_device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_bf16_f32_bf16_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
2
,
NHWC
,
KYXC
,
NHWK
,
BF16
,
F32
,
BF16
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv2d_bwd_weight_xdl_c_shuffle_nhwc_kyxc_nhwk_bf16_f32_bf16_instances
{});
add_device_operation_instances
(
instances
,
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_bf16_f32_bf16_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
// TODO: retire dedicated 2d version
#include "ck/tensor_operation/gpu/device/impl/device_conv2d_backward_weight_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
F16
=
ck
::
half_t
;
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NHWC
=
ck
::
tensor_layout
::
convolution
::
NHWC
;
using
KYXC
=
ck
::
tensor_layout
::
convolution
::
KYXC
;
using
NHWK
=
ck
::
tensor_layout
::
convolution
::
NHWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
using
device_conv2d_bwd_weight_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_default_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
// clang-format on
>
;
using
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
// clang-format on
>
;
// TODO: retire dedicated 2d version
// Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
using
device_conv_dedicated_2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instances
=
std
::
tuple
<
// clang-format off
//#################################################################################| InData| WeiData| OutData| AccData| In| Wei| Out| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#################################################################################| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#################################################################################| | | | | Operation| Operation| Operation| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#################################################################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
// clang-format on
>
;
void
add_device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
2
,
NHWC
,
KYXC
,
NHWK
,
F16
,
F16
,
F16
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv2d_bwd_weight_xdl_c_shuffle_nhwc_kyxc_nhwk_f16_default_instances
{});
add_device_operation_instances
(
instances
,
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f16_instances
{});
add_device_operation_instances
(
instances
,
device_conv_dedicated_2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f16_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv2d_bwd_weight/device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
// TODO: retire dedicated 2d version
#include "ck/tensor_operation/gpu/device/impl/device_conv2d_backward_weight_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NHWC
=
ck
::
tensor_layout
::
convolution
::
NHWC
;
using
KYXC
=
ck
::
tensor_layout
::
convolution
::
KYXC
;
using
NHWK
=
ck
::
tensor_layout
::
convolution
::
NHWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
using
device_conv2d_bwd_weight_xdl_c_shuffle_nhwc_kyxc_nhwk_f32_default_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
using
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f32_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
2
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
// Compilation parameters for in[n, hi, wi, c] * wei[k, y, x, c] = out[n, ho, wo, k]
using
device_conv_dedicated_2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instances
=
std
::
tuple
<
// clang-format off
//#################################################################################| InData| WeiData| OutData| AccData| In| Wei| Out| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#################################################################################| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#################################################################################| | | | | Operation| Operation| Operation| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#################################################################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
256
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
128
,
256
,
4
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
128
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
128
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
64
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
64
,
64
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
256
,
64
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
128
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
128
,
32
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
64
,
64
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConv2dBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
<
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
64
,
32
,
64
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
void
add_device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
2
,
NHWC
,
KYXC
,
NHWK
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv2d_bwd_weight_xdl_c_shuffle_nhwc_kyxc_nhwk_f32_default_instances
{});
add_device_operation_instances
(
instances
,
device_conv2d_bwd_weight_xdl_nhwc_kyxc_nhwk_1x1_s1_p0_f32_instances
{});
add_device_operation_instances
(
instances
,
device_conv_dedicated_2d_bwd_weight_xdl_nhwc_kyxc_nhwk_f32_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/CMakeLists.txt
deleted
100644 → 0
View file @
7d653017
add_instance_library
(
device_conv3d_bwd_weight_instance
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f16_instance.cpp
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f32_instance.cpp
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_bf16_instance.cpp
)
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_bf16_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
BF16
=
bhalf_t
;
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NDHWC
=
ck
::
tensor_layout
::
convolution
::
NDHWC
;
using
KZYXC
=
ck
::
tensor_layout
::
convolution
::
KZYXC
;
using
NDHWK
=
ck
::
tensor_layout
::
convolution
::
NDHWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, di, hi, wi, c] * wei[k, z, y, x, c] = out[n, do, ho, wo, k]
using
device_conv3d_bwd_weight_xdl_c_shuffle_ndhwc_kzyxc_ndhwk_bf16_f32_bf16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
using
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_bf16_f32_bf16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
BF16
,
F32
,
BF16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
void
add_device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_bf16_f32_bf16_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
3
,
NDHWC
,
KZYXC
,
NDHWK
,
BF16
,
F32
,
BF16
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv3d_bwd_weight_xdl_c_shuffle_ndhwc_kzyxc_ndhwk_bf16_f32_bf16_instances
{});
add_device_operation_instances
(
instances
,
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_bf16_f32_bf16_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f16_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
F16
=
ck
::
half_t
;
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NDHWC
=
ck
::
tensor_layout
::
convolution
::
NDHWC
;
using
KZYXC
=
ck
::
tensor_layout
::
convolution
::
KZYXC
;
using
NDHWK
=
ck
::
tensor_layout
::
convolution
::
NDHWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, di, hi, wi, c] * wei[k, z, y, x, c] = out[n, do, ho, wo, k]
using
device_conv3d_bwd_weight_xdl_c_shuffle_ndhwc_kzyxc_ndhwk_f16_default_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
// clang-format on
>
;
using
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_f16_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
8
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
8
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
8
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
8
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
1
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
8
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F16
,
F16
,
F16
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
8
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
4
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
2
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
8
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
8
>
// clang-format on
>
;
void
add_device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f16_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
3
,
NDHWC
,
KZYXC
,
NDHWK
,
F16
,
F16
,
F16
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv3d_bwd_weight_xdl_c_shuffle_ndhwc_kzyxc_ndhwk_f16_default_instances
{});
add_device_operation_instances
(
instances
,
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_f16_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
library/src/tensor_operation_instance/gpu/conv3d_bwd_weight/device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f32_instance.cpp
deleted
100644 → 0
View file @
7d653017
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_convnd_bwd_weight_nwc_kxc_nwk_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace
ck
{
namespace
tensor_operation
{
namespace
device
{
namespace
instance
{
using
F32
=
float
;
template
<
ck
::
index_t
...
Is
>
using
S
=
ck
::
Sequence
<
Is
...
>
;
using
NDHWC
=
ck
::
tensor_layout
::
convolution
::
NDHWC
;
using
KZYXC
=
ck
::
tensor_layout
::
convolution
::
KZYXC
;
using
NDHWK
=
ck
::
tensor_layout
::
convolution
::
NDHWK
;
using
PassThrough
=
ck
::
tensor_operation
::
element_wise
::
PassThrough
;
static
constexpr
auto
ConvBwdWeightDefault
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Default
;
static
constexpr
auto
ConvBwdWeightFilter1x1Stride1Pad0
=
ck
::
tensor_operation
::
device
::
ConvolutionBackwardWeightSpecialization
::
Filter1x1Stride1Pad0
;
// Compilation parameters for in[n, di, hi, wi, c] * wei[k, z, y, x, c] = out[n, do, ho, wo, k]
using
device_conv3d_bwd_weight_xdl_c_shuffle_ndhwc_kzyxc_ndhwk_f32_default_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
256
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
256
,
4
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
64
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
256
,
64
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
128
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
128
,
32
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
64
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightDefault
,
64
,
32
,
64
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
using
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_f32_instances
=
std
::
tuple
<
// clang-format off
//#########################################| Num| InData| WeiData| OutData| AccData| In| Wei| Out| ConvBackward| Block| MPer| NPer| K0Per| K1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransfer| CBlockTransfer|
//#########################################| Dim| Type| Type| Type| Type| Elementwise| Elementwise| Elementwise| Weight| Size| Block| Block| Block| | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| ClusterLengths| ScalarPerVector|
//#########################################| Spatial| | | | | Operation| Operation| Operation| Specialization| | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| MBlock_MPerBlock| NWaveNPerXdl|
//#########################################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | NBlock_NPerBlock| |
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
256
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
256
,
4
,
4
,
32
,
32
,
2
,
4
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
64
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
8
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
128
,
4
,
4
,
32
,
32
,
4
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
64
,
128
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
64
,
4
,
4
,
32
,
32
,
2
,
2
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
128
,
64
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
256
,
64
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
16
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
128
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
128
,
32
,
128
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
4
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
1
,
true
,
S
<
1
,
4
,
32
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
32
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
64
,
32
,
4
,
4
,
32
,
32
,
2
,
1
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
,
DeviceConvNdBwdWeightNwcKxcNwk_Xdl_CShuffle
<
3
,
F32
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
,
ConvBwdWeightFilter1x1Stride1Pad0
,
64
,
32
,
64
,
4
,
4
,
32
,
32
,
1
,
2
,
S
<
1
,
4
,
8
,
2
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
2
,
true
,
S
<
1
,
4
,
16
,
1
>
,
S
<
0
,
3
,
1
,
2
>
,
S
<
0
,
2
,
1
,
3
>
,
2
,
4
,
4
,
true
,
1
,
1
,
S
<
1
,
16
,
1
,
4
>
,
4
>
// clang-format on
>
;
void
add_device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_f32_instances
(
std
::
vector
<
std
::
unique_ptr
<
DeviceConvBwdWeight
<
3
,
NDHWC
,
KZYXC
,
NDHWK
,
F32
,
F32
,
F32
,
PassThrough
,
PassThrough
,
PassThrough
>>>&
instances
)
{
add_device_operation_instances
(
instances
,
device_conv3d_bwd_weight_xdl_c_shuffle_ndhwc_kzyxc_ndhwk_f32_default_instances
{});
add_device_operation_instances
(
instances
,
device_conv3d_bwd_weight_xdl_ndhwc_kzyxc_ndhwk_1x1_s1_p0_f32_instances
{});
}
}
// namespace instance
}
// namespace device
}
// namespace tensor_operation
}
// namespace ck
Prev
1
…
3
4
5
6
7
8
9
10
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