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
218094ba
提交
218094ba
authored
10月 01, 2015
作者:
Frédéric Bastien
浏览文件
操作
浏览文件
下载
差异文件
Merge pull request #3167 from abergeron/fix_amdlibm
Rework amdlibm exclusion on GPU.
上级
49b55484
c95d77b1
显示空白字符变更
内嵌
并排
正在显示
7 个修改的文件
包含
113 行增加
和
46 行删除
+113
-46
cop.txt
doc/extending/cop.txt
+46
-10
ctype.txt
doc/extending/ctype.txt
+20
-6
cc.py
theano/gof/cc.py
+27
-17
cmodule.py
theano/gof/cmodule.py
+6
-1
nvcc_compiler.py
theano/sandbox/cuda/nvcc_compiler.py
+2
-0
basic.py
theano/scalar/basic.py
+6
-6
type.py
theano/tensor/type.py
+6
-6
没有找到文件。
doc/extending/cop.txt
浏览文件 @
218094ba
...
@@ -57,39 +57,63 @@ There are less methods to define for an Op than for a Type:
...
@@ -57,39 +57,63 @@ There are less methods to define for an Op than for a Type:
*Default:* The default behavior is to do nothing.
*Default:* The default behavior is to do nothing.
.. method:: c_headers()
.. method:: c_headers(
[c_compiler]
)
Returns a list of headers to include in the file. 'Python.h' is
Returns a list of headers to include in the file. 'Python.h' is
included by default so you don't need to specify it. Also all
included by default so you don't need to specify it. Also all
of the header required by the Types involved (inputs and
of the header
s
required by the Types involved (inputs and
outputs) will also be included.
outputs) will also be included.
.. method:: c_header_dirs()
The `c_compiler` [#2v]_ parameter is the C compiler that will
be used to compile the code for the node. You may get multiple
calls with different C compilers.
.. method:: c_header_dirs([c_compiler])
Returns a list of directories to search for headers (arguments
Returns a list of directories to search for headers (arguments
to -I).
to -I).
.. method:: c_libraries()
The `c_compiler` [#2v]_ parameter is the C compiler that will
be used to compile the code for the node. You may get multiple
calls with different C compilers.
.. method:: c_libraries([c_compiler])
Returns a list of library names that your op needs to link to.
Returns a list of library names that your op needs to link to.
All ops are automatically linked with 'python' and the
All ops are automatically linked with 'python' and the
libraries their types require. (arguments to -l)
libraries their types require. (arguments to -l)
.. method:: c_lib_dirs()
The `c_compiler` [#2v]_ parameter is the C compiler that will
be used to compile the code for the node. You may get multiple
calls with different C compilers.
.. method:: c_lib_dirs([c_compiler])
Returns a list of directory to search for libraries (arguments
Returns a list of directory to search for libraries (arguments
to -L).
to -L).
.. method:: c_compile_args()
The `c_compiler` [#2v]_ parameter is the C compiler that will
be used to compile the code for the node. You may get multiple
calls with different C compilers.
.. method:: c_compile_args([c_compiler])
Allows to specify additional arbitrary arguments to the C
compiler. This is not usually required.
Allows to specify additional arbitrary arguments to g++. This
The `c_compiler` [#2v]_ parameter is the C compiler that will
is not usually required.
be used to compile the code for the node. You may get multiple
calls with different C compilers.
.. method:: c_no_compile_args()
.. method:: c_no_compile_args(
[c_compiler]
)
Returns a list of
g++
arguments that are forbidden when
Returns a list of
C compiler
arguments that are forbidden when
compiling this Op.
compiling this Op.
The `c_compiler` [#2v]_ parameter is the C compiler that will
be used to compile the code for the node. You may get multiple
calls with different C compilers.
.. method:: c_init_code()
.. method:: c_init_code()
Allows you to specify code that will be executed once when the
Allows you to specify code that will be executed once when the
...
@@ -245,6 +269,18 @@ In a nutshell, ``input_names`` and ``output_names`` parameterize the
...
@@ -245,6 +269,18 @@ In a nutshell, ``input_names`` and ``output_names`` parameterize the
names of the inputs your operation needs to use and the outputs it
names of the inputs your operation needs to use and the outputs it
needs to put variables into. But this will be clear with the examples.
needs to put variables into. But this will be clear with the examples.
.. rubric:: Footnotes
.. [#2v] There are actually two versions of this method one with a
`c_compiler` parameter and one without. The calling code will
try the version with c_compiler and try the version without
if it does not work. Defining both versions is pointless
since the one without `c_compiler` will never get called.
Note that these methods are not specific to a single apply
node so they may get called more than once on the same object
with different values for c_compiler.
Defining the methods
Defining the methods
====================
====================
...
...
doc/extending/ctype.txt
浏览文件 @
218094ba
...
@@ -78,18 +78,32 @@ the most important ones:
...
@@ -78,18 +78,32 @@ the most important ones:
When we are done using the data, clean up whatever we allocated and
When we are done using the data, clean up whatever we allocated and
decrease the appropriate reference counts.
decrease the appropriate reference counts.
.. method:: c_headers()
.. method:: c_headers(
[c_compiler]
)
c_libraries()
c_libraries(
[c_compiler]
)
c_header_dirs()
c_header_dirs(
[c_compiler]
)
c_lib_dirs()
c_lib_dirs(
[c_compiler]
)
Allows you to specify headers, libraries and associated directories.
Allows you to specify headers, libraries and associated directories.
.. method:: c_compile_args()
These methods have two versions, one with a `c_compiler`
c_no_compile_args()
argument and one without. The version with c_compiler is tried
first and if it doesn't work, the one without is.
The `c_compiler` argument is the C compiler that will be used
to compile the C code for the node that uses this type.
.. method:: c_compile_args([c_compiler])
c_no_compile_args([c_compiler])
Allows to specify special compiler arguments to add/exclude.
Allows to specify special compiler arguments to add/exclude.
These methods have two versions, one with a `c_compiler`
argument and one without. The version with c_compiler is tried
first and if it doesn't work, the one without is.
The `c_compiler` argument is the C compiler that will be used
to compile the C code for the node that uses this type.
.. method:: c_init_code()
.. method:: c_init_code()
Allows you to specify code that will be executed once when the
Allows you to specify code that will be executed once when the
...
...
theano/gof/cc.py
浏览文件 @
218094ba
...
@@ -930,15 +930,19 @@ class CLinker(link.Linker):
...
@@ -930,15 +930,19 @@ class CLinker(link.Linker):
"-Wno-unused-variable"
,
# idem as the precedent
"-Wno-unused-variable"
,
# idem as the precedent
"-Wno-write-strings"
,
# generated by our code generator...
"-Wno-write-strings"
,
# generated by our code generator...
]
]
c_compiler
=
self
.
c_compiler
()
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
y
.
op
for
y
in
self
.
node_order
]:
y
.
op
for
y
in
self
.
node_order
]:
try
:
try
:
try
:
ret
+=
x
.
c_compile_args
(
c_compiler
)
except
TypeError
:
ret
+=
x
.
c_compile_args
()
ret
+=
x
.
c_compile_args
()
except
utils
.
MethodNotDefined
:
except
utils
.
MethodNotDefined
:
pass
pass
c_compiler
=
self
.
c_compiler
()
ret
=
utils
.
uniq
(
ret
)
# to remove duplicate
ret
=
utils
.
uniq
(
ret
)
# to remove duplicate
# The args set by the compiler include the user flags. We do not want
# The args set by the compiler include the user flags. We do not want
# to reorder them
# to reorder them
...
@@ -946,7 +950,11 @@ class CLinker(link.Linker):
...
@@ -946,7 +950,11 @@ class CLinker(link.Linker):
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
y
.
op
for
y
in
self
.
node_order
]:
y
.
op
for
y
in
self
.
node_order
]:
try
:
try
:
for
i
in
x
.
c_no_compile_args
():
try
:
no_comp
=
x
.
c_no_compile_args
(
c_compiler
)
except
TypeError
:
no_comp
=
x
.
c_no_compile_args
()
for
i
in
no_comp
:
try
:
try
:
ret
.
remove
(
i
)
ret
.
remove
(
i
)
except
ValueError
:
except
ValueError
:
...
@@ -966,9 +974,13 @@ class CLinker(link.Linker):
...
@@ -966,9 +974,13 @@ class CLinker(link.Linker):
"""
"""
ret
=
[]
ret
=
[]
c_compiler
=
self
.
c_compiler
()
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
y
.
op
for
y
in
self
.
node_order
]:
y
.
op
for
y
in
self
.
node_order
]:
try
:
try
:
try
:
ret
+=
x
.
c_headers
(
c_compiler
)
except
TypeError
:
ret
+=
x
.
c_headers
()
ret
+=
x
.
c_headers
()
except
utils
.
MethodNotDefined
:
except
utils
.
MethodNotDefined
:
pass
pass
...
@@ -1023,9 +1035,13 @@ class CLinker(link.Linker):
...
@@ -1023,9 +1035,13 @@ class CLinker(link.Linker):
"""
"""
ret
=
[]
ret
=
[]
c_compiler
=
self
.
c_compiler
()
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
y
.
op
for
y
in
self
.
node_order
]:
y
.
op
for
y
in
self
.
node_order
]:
try
:
try
:
try
:
ret
+=
x
.
c_header_dirs
(
c_compiler
)
except
TypeError
:
ret
+=
x
.
c_header_dirs
()
ret
+=
x
.
c_header_dirs
()
except
utils
.
MethodNotDefined
:
except
utils
.
MethodNotDefined
:
pass
pass
...
@@ -1042,9 +1058,13 @@ class CLinker(link.Linker):
...
@@ -1042,9 +1058,13 @@ class CLinker(link.Linker):
"""
"""
ret
=
[]
ret
=
[]
c_compiler
=
self
.
c_compiler
()
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
y
.
op
for
y
in
self
.
node_order
]:
y
.
op
for
y
in
self
.
node_order
]:
try
:
try
:
try
:
ret
+=
x
.
c_libraries
(
c_compiler
)
except
TypeError
:
ret
+=
x
.
c_libraries
()
ret
+=
x
.
c_libraries
()
except
utils
.
MethodNotDefined
:
except
utils
.
MethodNotDefined
:
pass
pass
...
@@ -1061,9 +1081,13 @@ class CLinker(link.Linker):
...
@@ -1061,9 +1081,13 @@ class CLinker(link.Linker):
"""
"""
ret
=
[]
ret
=
[]
c_compiler
=
self
.
c_compiler
()
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
for
x
in
[
y
.
type
for
y
in
self
.
variables
]
+
[
y
.
op
for
y
in
self
.
node_order
]:
y
.
op
for
y
in
self
.
node_order
]:
try
:
try
:
try
:
ret
+=
x
.
c_lib_dirs
(
c_compiler
)
except
TypeError
:
ret
+=
x
.
c_lib_dirs
()
ret
+=
x
.
c_lib_dirs
()
except
utils
.
MethodNotDefined
:
except
utils
.
MethodNotDefined
:
pass
pass
...
@@ -1431,20 +1455,6 @@ class CLinker(link.Linker):
...
@@ -1431,20 +1455,6 @@ class CLinker(link.Linker):
c_compiler
=
self
.
c_compiler
()
c_compiler
=
self
.
c_compiler
()
libs
=
self
.
libraries
()
libs
=
self
.
libraries
()
preargs
=
self
.
compile_args
()
preargs
=
self
.
compile_args
()
compiler_name
=
c_compiler
.
__name__
if
compiler_name
==
'NVCC_compiler'
and
config
.
lib
.
amdlibm
:
# This lib does not work correctly with nvcc in device code.
# and newer version of g++ as 4.5.1.
# example of errors: "/usr/lib/gcc/x86_64-redhat-linux/4.5.1/
# include/mmintrin.h(49): error: identifier
# "__builtin_ia32_emms" is undefined"
if
'<amdlibm.h>'
in
mod
.
includes
:
mod
.
includes
.
remove
(
'<amdlibm.h>'
)
if
'-DREPLACE_WITH_AMDLIBM'
in
preargs
:
preargs
.
remove
(
'-DREPLACE_WITH_AMDLIBM'
)
if
'amdlibm'
in
libs
:
libs
.
remove
(
'amdlibm'
)
# We want to compute the code without the lock
# We want to compute the code without the lock
src_code
=
mod
.
code
()
src_code
=
mod
.
code
()
get_lock
()
get_lock
()
...
...
theano/gof/cmodule.py
浏览文件 @
218094ba
...
@@ -253,7 +253,10 @@ static struct PyModuleDef moduledef = {{
...
@@ -253,7 +253,10 @@ static struct PyModuleDef moduledef = {{
self
.
print_init
(
sio
)
self
.
print_init
(
sio
)
rval
=
sio
.
getvalue
()
rval
=
sio
.
getvalue
()
self
.
code_hash
=
hash_from_code
(
rval
)
# Make sure the hash of the code hasn't changed
h
=
hash_from_code
(
rval
)
assert
self
.
code_hash
is
None
or
self
.
code_hash
==
h
self
.
code_hash
=
h
rval
=
re
.
sub
(
self
.
hash_placeholder
,
self
.
code_hash
,
rval
)
rval
=
re
.
sub
(
self
.
hash_placeholder
,
self
.
code_hash
,
rval
)
# Finalize the Module, so no support code or function
# Finalize the Module, so no support code or function
# can be added
# can be added
...
@@ -1767,6 +1770,8 @@ class GCC_compiler(Compiler):
...
@@ -1767,6 +1770,8 @@ class GCC_compiler(Compiler):
# The equivalent flags of --march=native used by g++.
# The equivalent flags of --march=native used by g++.
march_flags
=
None
march_flags
=
None
supports_amdlibm
=
True
@staticmethod
@staticmethod
def
version_str
():
def
version_str
():
return
theano
.
config
.
cxx
+
" "
+
gcc_version_str
return
theano
.
config
.
cxx
+
" "
+
gcc_version_str
...
...
theano/sandbox/cuda/nvcc_compiler.py
浏览文件 @
218094ba
...
@@ -131,6 +131,8 @@ def add_standard_rpath(rpath):
...
@@ -131,6 +131,8 @@ def add_standard_rpath(rpath):
class
NVCC_compiler
(
Compiler
):
class
NVCC_compiler
(
Compiler
):
supports_amdlibm
=
False
@staticmethod
@staticmethod
def
try_compile_tmp
(
src_code
,
tmp_prefix
=
''
,
flags
=
(),
def
try_compile_tmp
(
src_code
,
tmp_prefix
=
''
,
flags
=
(),
try_run
=
False
,
output
=
False
):
try_run
=
False
,
output
=
False
):
...
...
theano/scalar/basic.py
浏览文件 @
218094ba
...
@@ -201,24 +201,24 @@ class Scalar(Type):
...
@@ -201,24 +201,24 @@ class Scalar(Type):
def
values_eq_approx
(
self
,
a
,
b
,
tolerance
=
1e-4
):
def
values_eq_approx
(
self
,
a
,
b
,
tolerance
=
1e-4
):
return
abs
(
a
-
b
)
<=
((
abs
(
a
)
+
abs
(
b
))
*
tolerance
)
return
abs
(
a
-
b
)
<=
((
abs
(
a
)
+
abs
(
b
))
*
tolerance
)
def
c_headers
(
self
):
def
c_headers
(
self
,
c_compiler
):
l
=
[
'<math.h>'
]
l
=
[
'<math.h>'
]
# These includes are needed by Scalar and TensorType,
# These includes are needed by Scalar and TensorType,
# we declare them here and they will be re-used by TensorType
# we declare them here and they will be re-used by TensorType
l
.
append
(
'<numpy/arrayobject.h>'
)
l
.
append
(
'<numpy/arrayobject.h>'
)
l
.
append
(
'<numpy/arrayscalars.h>'
)
l
.
append
(
'<numpy/arrayscalars.h>'
)
if
config
.
lib
.
amdlibm
:
if
config
.
lib
.
amdlibm
and
c_compiler
.
supports_amdlibm
:
l
+=
[
'<amdlibm.h>'
]
l
+=
[
'<amdlibm.h>'
]
return
l
return
l
def
c_libraries
(
self
):
def
c_libraries
(
self
,
c_compiler
):
l
=
[]
l
=
[]
if
config
.
lib
.
amdlibm
:
if
config
.
lib
.
amdlibm
and
c_compiler
.
supports_amdlibm
:
l
+=
[
'amdlibm'
]
l
+=
[
'amdlibm'
]
return
l
return
l
def
c_compile_args
(
self
):
def
c_compile_args
(
self
,
c_compiler
):
if
config
.
lib
.
amdlibm
:
if
config
.
lib
.
amdlibm
and
c_compiler
.
supports_amdlibm
:
return
[
'-DREPLACE_WITH_AMDLIBM'
]
return
[
'-DREPLACE_WITH_AMDLIBM'
]
else
:
else
:
return
[]
return
[]
...
...
theano/tensor/type.py
浏览文件 @
218094ba
...
@@ -596,18 +596,18 @@ class TensorType(Type):
...
@@ -596,18 +596,18 @@ class TensorType(Type):
}
}
"""
%
locals
()
"""
%
locals
()
def
c_headers
(
self
):
def
c_headers
(
self
,
c_compiler
):
"""
"""
Override `CLinkerObject.c_headers`.
Override `CLinkerObject.c_headers`.
"""
"""
return
scal
.
get_scalar_type
(
self
.
dtype
)
.
c_headers
()
return
scal
.
get_scalar_type
(
self
.
dtype
)
.
c_headers
(
c_compiler
)
def
c_libraries
(
self
):
def
c_libraries
(
self
,
c_compiler
):
return
scal
.
get_scalar_type
(
self
.
dtype
)
.
c_libraries
()
return
scal
.
get_scalar_type
(
self
.
dtype
)
.
c_libraries
(
c_compiler
)
def
c_compile_args
(
self
):
def
c_compile_args
(
self
,
c_compiler
):
return
scal
.
get_scalar_type
(
self
.
dtype
)
.
c_compile_args
()
return
scal
.
get_scalar_type
(
self
.
dtype
)
.
c_compile_args
(
c_compiler
)
def
c_support_code
(
self
):
def
c_support_code
(
self
):
"""
"""
...
...
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