Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
afeixing77
Paddle
提交
524f6e9b
P
Paddle
项目概览
afeixing77
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
0
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
0
列表
看板
标记
里程碑
合并请求
0
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
0
Issue
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
提交
524f6e9b
编写于
9月 29, 2018
作者:
Y
Yu Yang
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Refine code
上级
5cf395be
变更
7
隐藏空白更改
内联
并排
Showing
7 changed file
with
79 addition
and
26 deletion
+79
-26
paddle/fluid/memory/allocation/CMakeLists.txt
paddle/fluid/memory/allocation/CMakeLists.txt
+3
-2
paddle/fluid/memory/allocation/allocator_facade.cc
paddle/fluid/memory/allocation/allocator_facade.cc
+3
-1
paddle/fluid/memory/allocation/cuda_allocator.cc
paddle/fluid/memory/allocation/cuda_allocator.cc
+3
-22
paddle/fluid/operators/math/selected_rows_functor_test.cu.cc
paddle/fluid/operators/math/selected_rows_functor_test.cu.cc
+2
-1
paddle/fluid/platform/CMakeLists.txt
paddle/fluid/platform/CMakeLists.txt
+1
-0
paddle/fluid/platform/cuda_device_guard.cc
paddle/fluid/platform/cuda_device_guard.cc
+22
-0
paddle/fluid/platform/cuda_device_guard.h
paddle/fluid/platform/cuda_device_guard.h
+45
-0
未找到文件。
paddle/fluid/memory/allocation/CMakeLists.txt
浏览文件 @
524f6e9b
...
@@ -2,7 +2,7 @@ cc_library(allocator SRCS allocator.cc DEPS place)
...
@@ -2,7 +2,7 @@ cc_library(allocator SRCS allocator.cc DEPS place)
cc_library
(
cpu_allocator SRCS cpu_allocator.cc DEPS allocator
)
cc_library
(
cpu_allocator SRCS cpu_allocator.cc DEPS allocator
)
cc_library
(
best_fit_allocator SRCS best_fit_allocator.cc DEPS allocator
)
cc_library
(
best_fit_allocator SRCS best_fit_allocator.cc DEPS allocator
)
cc_library
(
locked_allocator SRCS locked_allocator.cc DEPS allocator
)
cc_library
(
locked_allocator SRCS locked_allocator.cc DEPS allocator
)
nv_library
(
cuda_allocator SRCS cuda_allocator.cc DEPS allocator
gpu_info
)
nv_library
(
cuda_allocator SRCS cuda_allocator.cc DEPS allocator
cuda_device_guard
)
if
(
WITH_GPU
)
if
(
WITH_GPU
)
nv_test
(
best_fit_allocator_test
nv_test
(
best_fit_allocator_test
...
@@ -40,4 +40,5 @@ cc_library(allocator_facade SRCS allocator_facade.cc DEPS
...
@@ -40,4 +40,5 @@ cc_library(allocator_facade SRCS allocator_facade.cc DEPS
locked_allocator
locked_allocator
best_fit_allocator
best_fit_allocator
naive_managed_allocator
naive_managed_allocator
aligned_allocator
)
aligned_allocator
cuda_device_guard
)
paddle/fluid/memory/allocation/allocator_facade.cc
浏览文件 @
524f6e9b
...
@@ -21,6 +21,7 @@
...
@@ -21,6 +21,7 @@
#include "paddle/fluid/memory/allocation/cpu_allocator.h"
#include "paddle/fluid/memory/allocation/cpu_allocator.h"
#include "paddle/fluid/memory/allocation/locked_allocator.h"
#include "paddle/fluid/memory/allocation/locked_allocator.h"
#include "paddle/fluid/memory/allocation/naive_managed_allocator.h"
#include "paddle/fluid/memory/allocation/naive_managed_allocator.h"
#include "paddle/fluid/platform/cuda_device_guard.h"
#include "paddle/fluid/platform/gpu_info.h"
#include "paddle/fluid/platform/gpu_info.h"
#include "paddle/fluid/platform/place.h"
#include "paddle/fluid/platform/place.h"
#ifdef PADDLE_WITH_CUDA
#ifdef PADDLE_WITH_CUDA
...
@@ -45,6 +46,7 @@ class AllocatorFacadePrivate {
...
@@ -45,6 +46,7 @@ class AllocatorFacadePrivate {
}
}
AllocatorFacadePrivate
()
{
AllocatorFacadePrivate
()
{
std
::
cout
<<
"Init Allocator Facade"
<<
std
::
endl
;
InitCPUAllocator
();
InitCPUAllocator
();
InitCUDAAllocator
();
InitCUDAAllocator
();
}
}
...
@@ -60,10 +62,10 @@ class AllocatorFacadePrivate {
...
@@ -60,10 +62,10 @@ class AllocatorFacadePrivate {
void
InitCUDAAllocator
()
{
void
InitCUDAAllocator
()
{
#ifdef PADDLE_WITH_CUDA
#ifdef PADDLE_WITH_CUDA
for
(
int
dev_id
=
0
;
dev_id
<
platform
::
GetCUDADeviceCount
();
++
dev_id
)
{
for
(
int
dev_id
=
0
;
dev_id
<
platform
::
GetCUDADeviceCount
();
++
dev_id
)
{
platform
::
CUDADeviceGuard
guard
(
dev_id
);
auto
cuda_allocator
=
auto
cuda_allocator
=
NaiveManagedAllocator
::
Create
(
std
::
unique_ptr
<
Allocator
>
(
NaiveManagedAllocator
::
Create
(
std
::
unique_ptr
<
Allocator
>
(
new
CUDAAllocator
(
platform
::
CUDAPlace
(
dev_id
))));
new
CUDAAllocator
(
platform
::
CUDAPlace
(
dev_id
))));
auto
allocation
=
cuda_allocator
->
Allocate
(
platform
::
GpuMaxChunkSize
());
auto
allocation
=
cuda_allocator
->
Allocate
(
platform
::
GpuMaxChunkSize
());
auto
allocator
=
NaiveManagedAllocator
::
Create
(
std
::
unique_ptr
<
Allocator
>
(
auto
allocator
=
NaiveManagedAllocator
::
Create
(
std
::
unique_ptr
<
Allocator
>
(
new
LockedAllocator
(
std
::
unique_ptr
<
Allocator
>
(
new
LockedAllocator
(
std
::
unique_ptr
<
Allocator
>
(
...
...
paddle/fluid/memory/allocation/cuda_allocator.cc
浏览文件 @
524f6e9b
...
@@ -16,34 +16,14 @@
...
@@ -16,34 +16,14 @@
#include <cuda.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <cuda_runtime.h>
#include <string>
#include <string>
#include "paddle/fluid/platform/cuda_device_guard.h"
#include "paddle/fluid/platform/gpu_info.h"
#include "paddle/fluid/platform/gpu_info.h"
namespace
paddle
{
namespace
paddle
{
namespace
memory
{
namespace
memory
{
namespace
allocation
{
namespace
allocation
{
class
CUDADeviceGuard
{
public:
explicit
CUDADeviceGuard
(
int
dev_id
)
{
int
prev_id
=
platform
::
GetCurrentDeviceId
();
if
(
prev_id
!=
dev_id
)
{
prev_id_
=
prev_id
;
platform
::
SetDeviceId
(
dev_id
);
}
}
~
CUDADeviceGuard
()
{
if
(
prev_id_
!=
-
1
)
{
platform
::
SetDeviceId
(
prev_id_
);
}
}
private:
int
prev_id_
{
-
1
};
};
std
::
unique_ptr
<
Allocation
>
CUDAAllocator
::
Allocate
(
size_t
size
,
Attr
attr
)
{
std
::
unique_ptr
<
Allocation
>
CUDAAllocator
::
Allocate
(
size_t
size
,
Attr
attr
)
{
CUDADeviceGuard
guard
(
place_
.
device
);
platform
::
CUDADeviceGuard
guard
(
place_
.
device
);
void
*
ptr
;
void
*
ptr
;
auto
status
=
cudaMalloc
(
&
ptr
,
size
);
auto
status
=
cudaMalloc
(
&
ptr
,
size
);
if
(
UNLIKELY
(
status
!=
cudaSuccess
))
{
if
(
UNLIKELY
(
status
!=
cudaSuccess
))
{
...
@@ -57,6 +37,7 @@ std::unique_ptr<Allocation> CUDAAllocator::Allocate(size_t size, Attr attr) {
...
@@ -57,6 +37,7 @@ std::unique_ptr<Allocation> CUDAAllocator::Allocate(size_t size, Attr attr) {
}
}
void
CUDAAllocator
::
Free
(
Allocation
*
allocation
)
{
void
CUDAAllocator
::
Free
(
Allocation
*
allocation
)
{
platform
::
CUDADeviceGuard
guard
(
place_
.
device
);
auto
*
cuda_allocation
=
dynamic_cast
<
CUDAAllocation
*>
(
allocation
);
auto
*
cuda_allocation
=
dynamic_cast
<
CUDAAllocation
*>
(
allocation
);
PADDLE_ENFORCE_NOT_NULL
(
cuda_allocation
);
PADDLE_ENFORCE_NOT_NULL
(
cuda_allocation
);
PADDLE_ENFORCE_EQ
(
boost
::
get
<
platform
::
CUDAPlace
>
(
cuda_allocation
->
place
()),
PADDLE_ENFORCE_EQ
(
boost
::
get
<
platform
::
CUDAPlace
>
(
cuda_allocation
->
place
()),
...
...
paddle/fluid/operators/math/selected_rows_functor_test.cu
→
paddle/fluid/operators/math/selected_rows_functor_test.cu
.cc
浏览文件 @
524f6e9b
...
@@ -12,10 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
...
@@ -12,10 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
See the License for the specific language governing permissions and
limitations under the License. */
limitations under the License. */
#include "paddle/fluid/operators/math/selected_rows_functor.h"
#include <vector>
#include <vector>
#include "gtest/gtest.h"
#include "gtest/gtest.h"
#include "paddle/fluid/operators/math/math_function.h"
#include "paddle/fluid/operators/math/math_function.h"
#include "paddle/fluid/operators/math/selected_rows_functor.h"
TEST
(
selected_rows_functor
,
gpu_add
)
{
TEST
(
selected_rows_functor
,
gpu_add
)
{
paddle
::
platform
::
CUDAPlace
gpu_place
(
0
);
paddle
::
platform
::
CUDAPlace
gpu_place
(
0
);
...
@@ -38,6 +38,7 @@ TEST(selected_rows_functor, gpu_add) {
...
@@ -38,6 +38,7 @@ TEST(selected_rows_functor, gpu_add) {
{
static_cast
<
int64_t
>
(
rows1
.
size
()),
row_numel
}),
{
static_cast
<
int64_t
>
(
rows1
.
size
()),
row_numel
}),
gpu_place
);
gpu_place
);
functor
(
ctx
,
in1_value
,
1.0
);
functor
(
ctx
,
in1_value
,
1.0
);
PADDLE_ENFORCE
(
cudaDeviceSynchronize
());
std
::
vector
<
int64_t
>
rows2
{
0
,
5
,
7
,
9
};
std
::
vector
<
int64_t
>
rows2
{
0
,
5
,
7
,
9
};
std
::
unique_ptr
<
paddle
::
framework
::
SelectedRows
>
selected_rows2
{
std
::
unique_ptr
<
paddle
::
framework
::
SelectedRows
>
selected_rows2
{
...
...
paddle/fluid/platform/CMakeLists.txt
浏览文件 @
524f6e9b
...
@@ -73,3 +73,4 @@ cc_test(float16_test SRCS float16_test.cc DEPS lod_tensor)
...
@@ -73,3 +73,4 @@ cc_test(float16_test SRCS float16_test.cc DEPS lod_tensor)
IF
(
WITH_GPU
)
IF
(
WITH_GPU
)
nv_test
(
cuda_helper_test SRCS cuda_helper_test.cu
)
nv_test
(
cuda_helper_test SRCS cuda_helper_test.cu
)
ENDIF
()
ENDIF
()
nv_library
(
cuda_device_guard SRCS cuda_device_guard.cc DEPS gpu_info
)
paddle/fluid/platform/cuda_device_guard.cc
0 → 100644
浏览文件 @
524f6e9b
// Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "paddle/fluid/platform/cuda_device_guard.h"
namespace
paddle
{
namespace
platform
{
// Even this source file does not contains any code, it is better to keep this
// source file for cmake dependency.
}
// namespace platform
}
// namespace paddle
paddle/fluid/platform/cuda_device_guard.h
0 → 100644
浏览文件 @
524f6e9b
// Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#pragma once
#include "paddle/fluid/platform/gpu_info.h"
namespace
paddle
{
namespace
platform
{
class
CUDADeviceGuard
{
public:
explicit
inline
CUDADeviceGuard
(
int
dev_id
)
{
int
prev_id
=
platform
::
GetCurrentDeviceId
();
if
(
prev_id
!=
dev_id
)
{
prev_id_
=
prev_id
;
platform
::
SetDeviceId
(
dev_id
);
}
}
inline
~
CUDADeviceGuard
()
{
if
(
prev_id_
!=
-
1
)
{
platform
::
SetDeviceId
(
prev_id_
);
}
}
CUDADeviceGuard
(
const
CUDADeviceGuard
&
o
)
=
delete
;
CUDADeviceGuard
&
operator
=
(
const
CUDADeviceGuard
&
o
)
=
delete
;
private:
int
prev_id_
{
-
1
};
};
}
// namespace platform
}
// namespace paddle
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录