Skip to content
项目
群组
代码片段
帮助
当前项目
正在载入...
登录 / 注册
切换导航面板
P
pytensor
项目
项目
详情
活动
周期分析
仓库
仓库
文件
提交
分支
标签
贡献者
图表
比较
统计图
议题
0
议题
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
CI / CD
CI / CD
流水线
作业
日程
统计图
Wiki
Wiki
代码片段
代码片段
成员
成员
折叠边栏
关闭边栏
活动
图像
聊天
创建新问题
作业
提交
问题看板
Open sidebar
testgroup
pytensor
Commits
f38a3f3d
提交
f38a3f3d
authored
1月 07, 2014
作者:
Arnaud Bergeron
提交者：
Frederic
1月 07, 2014
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Clean up types in printfs to make sure the printouts will work on 32 and 64 bits machines.
上级
d19a4777
隐藏空白字符变更
内嵌
并排
正在显示
1 个修改的文件
包含
100 行增加
和
79 行删除
+100
-79
conv.cu
theano/sandbox/gpuarray/conv.cu
+100
-79
没有找到文件。
theano/sandbox/gpuarray/conv.cu
浏览文件 @
f38a3f3d
...
...
@@ -132,9 +132,9 @@ PyGpuArray_conv_valid(const PyGpuArrayObject *img,
const
int
kern_size_byte
=
kern_size
*
sizeof
(
float
);
const
int
out_size_byte
=
out_size
*
sizeof
(
float
);
if
(
!
((
THEANO_KERN_WID
==
PyGpuArray_DIMS
(
kern
)[
3
])
||
(
THEANO_KERN_WID
==
0
))){
PyErr_Format
(
PyExc_ValueError
,
"ERROR: This GpuConv code was compiled for"
" %d kernel columns, but the kernel we received had %ud columns!"
,
THEANO_KERN_WID
,
PyGpuArray_DIMS
(
kern
)[
3
]);
PyErr_Format
(
PyExc_ValueError
,
"ERROR: This GpuConv code was compiled for"
" %d kernel columns, but the kernel we received had %
ll
ud columns!"
,
THEANO_KERN_WID
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
3
]);
return
-
1
;
}
...
...
@@ -319,19 +319,21 @@ PyGpuArray_conv_valid(const PyGpuArrayObject *img,
" img_c_contiguous_2d=%i,"
" kern_c_contiguous_2d=%i, nb_split=%i,"
" preload_full_kernel=%i,"
" subsample_rows=%
i, subsample_cols=%i
\n
"
,
" subsample_rows=%
llu, subsample_cols=%llu
\n
"
,
threads
.
x
,
threads
.
y
,
grid
.
x
,
grid
.
y
,
shared_size
,
threads
.
x
*
threads
.
y
,
THEANO_KERN_WID
,
img_contiguous_2d
,
kern_contiguous_2d
,
nb_split
,
preload_full_kernel
,
subsample_rows
,
subsample_cols
);
(
unsigned
long
long
)
subsample_rows
,
(
unsigned
long
long
)
subsample_cols
);
if
(
verbose
)
fprintf
(
stderr
,
"INFO: used 'conv_patch_stack' version with nb_split=%i"
" and preload_full_kernel=%i,"
" subsample_rows=%
i, subsample_cols=%i
\n
"
,
" subsample_rows=%
llu, subsample_cols=%llu
\n
"
,
nb_split
,
preload_full_kernel
,
subsample_rows
,
subsample_cols
);
(
unsigned
long
long
)
subsample_rows
,
(
unsigned
long
long
)
subsample_cols
);
work_complete
=
true
;
}
else
...
...
@@ -344,12 +346,13 @@ PyGpuArray_conv_valid(const PyGpuArrayObject *img,
" kern_width=%i, img_c_contiguous_2d=%i,"
" kern_c_contiguous_2d=%i, nb_split=%i,"
" preload_full_kernel=%i,"
" subsample_rows=%
i, subsample_cols=%i
\n
"
,
" subsample_rows=%
llu, subsample_cols=%llu
\n
"
,
threads
.
x
,
threads
.
y
,
grid
.
x
,
grid
.
y
,
shared_size
,
threads
.
x
*
threads
.
y
,
THEANO_KERN_WID
,
img_contiguous_2d
,
kern_contiguous_2d
,
nb_split
,
preload_full_kernel
,
subsample_rows
,
subsample_cols
);
(
unsigned
long
long
)
subsample_rows
,
(
unsigned
long
long
)
subsample_cols
);
if
(
verbose
)
fprintf
(
stderr
,
"INFO: impl 'conv_patch_stack' failed (%s),"
...
...
@@ -818,30 +821,35 @@ PyGpuArray_conv_valid(const PyGpuArrayObject *img,
if
(
verbose
)
fprintf
(
stderr
,
"INFO: launching conv_reference_valid
\n
"
);
if
(
verbose
>
1
)
fprintf
(
stderr
,
" img : %i %i %i %i %p %i %i %i %i
\n
"
,
nbatch
,
PyGpuArray_DIMS
(
img
)[
1
],
img_len
,
img_wid
,
fprintf
(
stderr
,
" img : %i %llu %i %i %p "
"%lld %lld %lld %lld
\n
"
,
nbatch
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
1
],
img_len
,
img_wid
,
cuda_get_ptr
(
img
),
PyGpuArray_STRIDES
(
img
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
3
]
/
4
);
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
3
]
/
4
);
if
(
verbose
>
1
)
fprintf
(
stderr
,
" kern: %i %i %i %i %p %i %i %i %i
\n
"
,
fprintf
(
stderr
,
" kern: %i %i %i %i %p "
"%lld %lld %lld %lld
\n
"
,
nkern
,
nstack
,
kern_len
,
kern_wid
,
cuda_get_ptr
(
kern
),
PyGpuArray_STRIDES
(
kern
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
3
]
/
4
);
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
3
]
/
4
);
if
(
verbose
>
1
)
fprintf
(
stderr
,
" out : %i %i %i %i %p %i %i %i %i
\n
"
,
PyGpuArray_DIMS
(
out
)[
0
],
PyGpuArray_DIMS
(
out
)[
1
],
out_len
,
out_wid
,
fprintf
(
stderr
,
" out : %llu %llu %i %i %p "
"%lld %lld %lld %lld
\n
"
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
0
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
1
],
out_len
,
out_wid
,
cuda_get_ptr
(
out
),
PyGpuArray_STRIDES
(
out
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
3
]
/
4
);
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
3
]
/
4
);
if
(
verbose
>
1
)
fprintf
(
stderr
,
" launch params: %i %i %i
\n
"
,
outsize
,
n_blocks
,
n_threads
);
...
...
@@ -968,8 +976,8 @@ PyGpuArray_conv_full(const PyGpuArrayObject *img, const PyGpuArrayObject * kern,
PyErr_Format
(
PyExc_ValueError
,
"ERROR: This GpuConv code was compiled for"
" %d kernel columns, but the kernel we received"
" had %ud columns!"
,
THEANO_KERN_WID
,
PyGpuArray_DIMS
(
kern
)[
3
]);
" had %
ll
ud columns!"
,
THEANO_KERN_WID
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
3
]);
return
-
1
;
}
bool
subsample
=
subsample_rows
!=
1
||
subsample_cols
!=
1
;
...
...
@@ -1007,27 +1015,36 @@ PyGpuArray_conv_full(const PyGpuArrayObject *img, const PyGpuArrayObject * kern,
{
printf
(
"INFO: Running conv_full version=%d,"
" MACRO kern_width=%d with inputs:
\n
"
,
version
,
THEANO_KERN_WID
);
printf
(
"INFO: img dim: %i %i %i %i img stride: %i %i %i %i
\n
"
,
PyGpuArray_DIMS
(
img
)[
0
],
PyGpuArray_DIMS
(
img
)[
1
],
PyGpuArray_DIMS
(
img
)[
2
],
PyGpuArray_DIMS
(
img
)[
3
],
PyGpuArray_STRIDES
(
img
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
3
]
/
4
);
printf
(
"INFO: kern dim: %i %i %i %i kern stride: %i %i %i %i
\n
"
,
PyGpuArray_DIMS
(
kern
)[
0
],
PyGpuArray_DIMS
(
kern
)[
1
],
PyGpuArray_DIMS
(
kern
)[
2
],
PyGpuArray_DIMS
(
kern
)[
3
],
PyGpuArray_STRIDES
(
kern
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
3
]
/
4
);
printf
(
"INFO: out dim: %i %i %i %i out stride: %i %i %i %i
\n
"
,
PyGpuArray_DIMS
(
out
)[
0
],
PyGpuArray_DIMS
(
out
)[
1
],
PyGpuArray_DIMS
(
out
)[
2
],
PyGpuArray_DIMS
(
out
)[
3
],
PyGpuArray_STRIDES
(
out
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
3
]
/
4
);
printf
(
"INFO: img dim: %llu %llu %llu %llu "
"img stride: %lld %lld %lld %lld
\n
"
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
0
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
1
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
2
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
3
],
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
3
]
/
4
);
printf
(
"INFO: kern dim: %llu %llu %llu %llu "
"kern stride: %lld %lld %lld %lld
\n
"
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
0
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
1
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
2
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
3
],
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
3
]
/
4
);
printf
(
"INFO: out dim: %llu %llu %llu %llu "
"out stride: %lld %lld %lld %lld
\n
"
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
0
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
1
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
2
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
3
],
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
3
]
/
4
);
}
if
(
!
subsample
&&
...
...
@@ -1313,45 +1330,49 @@ PyGpuArray_conv_full(const PyGpuArrayObject *img, const PyGpuArrayObject * kern,
if
(
verbose
)
fprintf
(
stderr
,
"INFO: launching conv_reference_valid
\n
"
);
if
(
verbose
)
fprintf
(
stderr
,
" img : %i %i %i %i %p %i %i %i %i
\n
"
,
PyGpuArray_DIMS
(
img
)[
0
],
PyGpuArray_DIMS
(
img
)[
1
],
PyGpuArray_DIMS
(
img
)[
2
],
PyGpuArray_DIMS
(
img
)[
3
],
fprintf
(
stderr
,
" img : %llu %llu %llu %llu %p "
"%lld %lld %lld %lld
\n
"
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
0
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
1
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
2
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
img
)[
3
],
cuda_get_ptr
(
img
),
PyGpuArray_STRIDES
(
img
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
img
)[
3
]
/
4
);
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
img
)[
3
]
/
4
);
if
(
verbose
)
fprintf
(
stderr
,
" kern: %i %i %i %i %p %i %i %i %i
\n
"
,
PyGpuArray_DIMS
(
kern
)[
0
],
PyGpuArray_DIMS
(
kern
)[
1
],
PyGpuArray_DIMS
(
kern
)[
2
],
PyGpuArray_DIMS
(
kern
)[
3
],
fprintf
(
stderr
,
" kern: %llu %llu %llu %llu %p "
"%lld %lld %lld %lld
\n
"
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
0
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
1
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
2
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
kern
)[
3
],
cuda_get_ptr
(
kern
),
PyGpuArray_STRIDES
(
kern
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
kern
)[
3
]
/
4
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
kern
)[
3
]
/
4
);
if
(
verbose
)
fprintf
(
stderr
,
" out : %i %i %i %i %p %i %i %i %i
\n
"
,
PyGpuArray_DIMS
(
out
)[
0
],
PyGpuArray_DIMS
(
out
)[
1
],
PyGpuArray_DIMS
(
out
)[
2
],
PyGpuArray_DIMS
(
out
)[
3
],
fprintf
(
stderr
,
" out : %llu %llu %llu %llu %p "
"%lld %lld %lld %lld
\n
"
,
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
0
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
1
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
2
],
(
unsigned
long
long
)
PyGpuArray_DIMS
(
out
)[
3
],
cuda_get_ptr
(
out
),
PyGpuArray_STRIDES
(
out
)[
0
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
1
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
2
]
/
4
,
PyGpuArray_STRIDES
(
out
)[
3
]
/
4
);
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
0
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
1
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
2
]
/
4
,
(
long
long
)
PyGpuArray_STRIDES
(
out
)[
3
]
/
4
);
if
(
verbose
)
fprintf
(
stderr
,
" launch params: %i %i %i
\n
"
,
outsize
,
n_blocks
,
n_threads
);
if
(
verbose
)
fprintf
(
stderr
,
" subsample params: %i %i
\n
"
,
subsample_rows
,
subsample_cols
);
fprintf
(
stderr
,
" subsample params: %llu %llu
\n
"
,
(
unsigned
long
long
)
subsample_rows
,
(
unsigned
long
long
)
subsample_cols
);
}
conv_reference_full
<<<
n_blocks
,
n_threads
>>>
(
PyGpuArray_DIMS
(
img
)[
0
],
PyGpuArray_DIMS
(
kern
)[
0
],
...
...
编写
预览
Markdown
格式
0%
重试
或
添加新文件
添加附件
取消
您添加了
0
人
到此讨论。请谨慎行事。
请先完成此评论的编辑！
取消
请
注册
或者
登录
后发表评论
