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
csdn_franckjun
Paddle
提交
51a9fca3
P
Paddle
项目概览
csdn_franckjun
/
Paddle
与 Fork 源项目一致
Fork自
PaddlePaddle / Paddle
通知
1
Star
1
Fork
0
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1
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
1
Issue
1
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 >>
未验证
提交
51a9fca3
编写于
12月 24, 2018
作者:
Q
qingqing01
提交者:
GitHub
12月 24, 2018
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Async memory copy (#15013)
上级
93870574
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
13 addition
and
6 deletion
+13
-6
paddle/fluid/inference/api/analysis_predictor.cc
paddle/fluid/inference/api/analysis_predictor.cc
+5
-2
paddle/fluid/inference/api/api_impl.cc
paddle/fluid/inference/api/api_impl.cc
+5
-2
paddle/fluid/operators/detection/density_prior_box_op.cu
paddle/fluid/operators/detection/density_prior_box_op.cu
+3
-2
未找到文件。
paddle/fluid/inference/api/analysis_predictor.cc
浏览文件 @
51a9fca3
...
@@ -231,11 +231,14 @@ bool AnalysisPredictor::SetFeed(const std::vector<PaddleTensor> &inputs,
...
@@ -231,11 +231,14 @@ bool AnalysisPredictor::SetFeed(const std::vector<PaddleTensor> &inputs,
inputs
[
i
].
data
.
length
());
inputs
[
i
].
data
.
length
());
}
else
{
}
else
{
#ifdef PADDLE_WITH_CUDA
#ifdef PADDLE_WITH_CUDA
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
auto
*
dev_ctx
=
static_cast
<
const
platform
::
CUDADeviceContext
*>
(
pool
.
Get
(
place_
));
auto
dst_gpu_place
=
boost
::
get
<
platform
::
CUDAPlace
>
(
place_
);
auto
dst_gpu_place
=
boost
::
get
<
platform
::
CUDAPlace
>
(
place_
);
memory
::
Copy
(
dst_gpu_place
,
static_cast
<
void
*>
(
input_ptr
),
memory
::
Copy
(
dst_gpu_place
,
static_cast
<
void
*>
(
input_ptr
),
platform
::
CPUPlace
(),
inputs
[
i
].
data
.
data
(),
platform
::
CPUPlace
(),
inputs
[
i
].
data
.
data
(),
inputs
[
i
].
data
.
length
(),
inputs
[
i
].
data
.
length
(),
dev_ctx
->
stream
());
0
);
// stream 0 for sync copy
#else
#else
PADDLE_THROW
(
"Not compile with CUDA, should not reach here."
);
PADDLE_THROW
(
"Not compile with CUDA, should not reach here."
);
#endif
#endif
...
...
paddle/fluid/inference/api/api_impl.cc
浏览文件 @
51a9fca3
...
@@ -208,11 +208,14 @@ bool NativePaddlePredictor::SetFeed(const std::vector<PaddleTensor> &inputs,
...
@@ -208,11 +208,14 @@ bool NativePaddlePredictor::SetFeed(const std::vector<PaddleTensor> &inputs,
inputs
[
i
].
data
.
length
());
inputs
[
i
].
data
.
length
());
}
else
{
}
else
{
#ifdef PADDLE_WITH_CUDA
#ifdef PADDLE_WITH_CUDA
platform
::
DeviceContextPool
&
pool
=
platform
::
DeviceContextPool
::
Instance
();
auto
*
dev_ctx
=
static_cast
<
const
platform
::
CUDADeviceContext
*>
(
pool
.
Get
(
place_
));
auto
dst_gpu_place
=
boost
::
get
<
platform
::
CUDAPlace
>
(
place_
);
auto
dst_gpu_place
=
boost
::
get
<
platform
::
CUDAPlace
>
(
place_
);
memory
::
Copy
(
dst_gpu_place
,
static_cast
<
void
*>
(
input_ptr
),
memory
::
Copy
(
dst_gpu_place
,
static_cast
<
void
*>
(
input_ptr
),
platform
::
CPUPlace
(),
inputs
[
i
].
data
.
data
(),
platform
::
CPUPlace
(),
inputs
[
i
].
data
.
data
(),
inputs
[
i
].
data
.
length
(),
inputs
[
i
].
data
.
length
(),
dev_ctx
->
stream
());
0
);
// stream 0 for sync copy
#else
#else
PADDLE_THROW
(
"Not compile with CUDA, should not reach here."
);
PADDLE_THROW
(
"Not compile with CUDA, should not reach here."
);
#endif
#endif
...
...
paddle/fluid/operators/detection/density_prior_box_op.cu
浏览文件 @
51a9fca3
...
@@ -142,12 +142,13 @@ class DensityPriorBoxOpCUDAKernel : public framework::OpKernel<T> {
...
@@ -142,12 +142,13 @@ class DensityPriorBoxOpCUDAKernel : public framework::OpKernel<T> {
vars
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
vars
->
mutable_data
<
T
>
(
ctx
.
GetPlace
());
framework
::
Tensor
d_temp
;
framework
::
Tensor
d_temp
;
framework
::
TensorCopy
Sync
(
h_temp
,
ctx
.
GetPlace
(),
&
d_temp
);
framework
::
TensorCopy
(
h_temp
,
ctx
.
GetPlace
(),
&
d_temp
);
// At least use 32 threads, at most 512 threads.
// At least use 32 threads, at most 512 threads.
// blockx is multiple of 32.
// blockx is multiple of 32.
int
blockx
=
std
::
min
(
int
blockx
=
std
::
min
(
static_cast
<
long
>
(((
feature_width
*
num_priors
+
31
)
>>
5
)
<<
5
),
512L
);
static_cast
<
int64_t
>
(((
feature_width
*
num_priors
+
31
)
>>
5
)
<<
5
),
512L
);
int
gridx
=
(
feature_width
*
num_priors
+
blockx
-
1
)
/
blockx
;
int
gridx
=
(
feature_width
*
num_priors
+
blockx
-
1
)
/
blockx
;
dim3
threads
(
blockx
,
1
);
dim3
threads
(
blockx
,
1
);
dim3
grids
(
gridx
,
feature_height
);
dim3
grids
(
gridx
,
feature_height
);
...
...
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