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
PaddlePaddle
Paddle
提交
1432e3d2
P
Paddle
项目概览
PaddlePaddle
/
Paddle
大约 1 年 前同步成功
通知
2299
Star
20931
Fork
5422
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
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
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
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
未验证
提交
1432e3d2
编写于
12月 06, 2021
作者:
R
ronnywang
提交者:
GitHub
12月 06, 2021
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Fix bug (#37868)
上级
6a1e4de2
变更
4
显示空白变更内容
内联
并排
Showing
4 changed file
with
5 addition
and
5 deletion
+5
-5
paddle/fluid/imperative/nccl_context.cc
paddle/fluid/imperative/nccl_context.cc
+2
-2
paddle/fluid/imperative/tests/CMakeLists.txt
paddle/fluid/imperative/tests/CMakeLists.txt
+1
-1
paddle/fluid/imperative/tests/heter_ccl_context_test.cc
paddle/fluid/imperative/tests/heter_ccl_context_test.cc
+1
-1
paddle/fluid/imperative/tests/nccl_context_test.cc
paddle/fluid/imperative/tests/nccl_context_test.cc
+1
-1
未找到文件。
paddle/fluid/imperative/nccl_context.cc
浏览文件 @
1432e3d2
...
@@ -27,8 +27,8 @@
...
@@ -27,8 +27,8 @@
#include "paddle/fluid/framework/scope.h"
#include "paddle/fluid/framework/scope.h"
#include "paddle/fluid/framework/variable.h"
#include "paddle/fluid/framework/variable.h"
#include "paddle/fluid/platform/device/gpu/nccl_helper.h"
#include "paddle/fluid/platform/device_context.h"
#include "paddle/fluid/platform/device_context.h"
#include "paddle/fluid/platform/nccl_helper.h"
#include "paddle/fluid/platform/place.h"
#include "paddle/fluid/platform/place.h"
namespace
paddle
{
namespace
paddle
{
...
@@ -145,7 +145,7 @@ void NCCLParallelContext::Broadcast(framework::Variable *src, int ring_id) {
...
@@ -145,7 +145,7 @@ void NCCLParallelContext::Broadcast(framework::Variable *src, int ring_id) {
void
*
src_ptr
=
src_tensor
->
data
<
void
>
();
void
*
src_ptr
=
src_tensor
->
data
<
void
>
();
auto
nccl_dtype
=
platform
::
ToNCCLDataType
(
src_tensor
->
type
());
auto
nccl_dtype
=
platform
::
ToNCCLDataType
(
src_tensor
->
type
());
PADDLE_ENFORCE_
CUDA
_SUCCESS
(
platform
::
dynload
::
ncclBcast
(
PADDLE_ENFORCE_
GPU
_SUCCESS
(
platform
::
dynload
::
ncclBcast
(
src_ptr
,
src_tensor
->
numel
(),
nccl_dtype
,
0
,
comm
->
comm
(),
stream
));
src_ptr
,
src_tensor
->
numel
(),
nccl_dtype
,
0
,
comm
->
comm
(),
stream
));
}
}
...
...
paddle/fluid/imperative/tests/CMakeLists.txt
浏览文件 @
1432e3d2
if
(
WIN32
)
if
(
WIN32
)
cc_test
(
nccl_context_test SRCS nccl_context_test.cc DEPS device_context
)
cc_test
(
nccl_context_test SRCS nccl_context_test.cc DEPS device_context
)
else
()
else
()
if
(
WITH_
NCCL OR WITH_RCCL
)
if
(
WITH_
GLOO
AND
(
WITH_NCCL OR WITH_RCCL
)
)
cc_test
(
nccl_context_test SRCS nccl_context_test.cc DEPS nccl_context
)
cc_test
(
nccl_context_test SRCS nccl_context_test.cc DEPS nccl_context
)
cc_test
(
heter_ccl_context_test SRCS heter_ccl_context_test.cc DEPS heter_ccl_context nccl_context imperative_gloo_context gloo_context gloo_wrapper gloo fs shell
)
cc_test
(
heter_ccl_context_test SRCS heter_ccl_context_test.cc DEPS heter_ccl_context nccl_context imperative_gloo_context gloo_context gloo_wrapper gloo fs shell
)
#set_tests_properties(heter_ccl_context_test PROPERTIES LABELS "RUN_TYPE=DIST")
#set_tests_properties(heter_ccl_context_test PROPERTIES LABELS "RUN_TYPE=DIST")
...
...
paddle/fluid/imperative/tests/heter_ccl_context_test.cc
浏览文件 @
1432e3d2
...
@@ -79,7 +79,7 @@ void AllReduceByStream(int local_rank, int device_id) {
...
@@ -79,7 +79,7 @@ void AllReduceByStream(int local_rank, int device_id) {
}
}
TEST
(
AllReduceByStream
,
Run
)
{
TEST
(
AllReduceByStream
,
Run
)
{
if
(
platform
::
Get
CUDA
DeviceCount
()
>=
2
)
{
if
(
platform
::
Get
GPU
DeviceCount
()
>=
2
)
{
std
::
thread
t0
(
AllReduceByStream
,
0
,
0
);
std
::
thread
t0
(
AllReduceByStream
,
0
,
0
);
std
::
thread
t1
(
AllReduceByStream
,
1
,
1
);
std
::
thread
t1
(
AllReduceByStream
,
1
,
1
);
t0
.
join
();
t0
.
join
();
...
...
paddle/fluid/imperative/tests/nccl_context_test.cc
浏览文件 @
1432e3d2
...
@@ -111,7 +111,7 @@ void Broadcast(int local_rank, int device_id) {
...
@@ -111,7 +111,7 @@ void Broadcast(int local_rank, int device_id) {
}
}
TEST
(
Broadcast
,
Run
)
{
TEST
(
Broadcast
,
Run
)
{
if
(
platform
::
Get
CUDA
DeviceCount
()
>=
2
)
{
if
(
platform
::
Get
GPU
DeviceCount
()
>=
2
)
{
std
::
thread
t0
(
Broadcast
,
0
,
0
);
std
::
thread
t0
(
Broadcast
,
0
,
0
);
std
::
thread
t1
(
Broadcast
,
1
,
1
);
std
::
thread
t1
(
Broadcast
,
1
,
1
);
t0
.
join
();
t0
.
join
();
...
...
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