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
毕竟曾有刹那
Mace
提交
fe4cefd1
Mace
项目概览
毕竟曾有刹那
/
Mace
与 Fork 源项目一致
Fork自
Xiaomi / Mace
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
DevOps
流水线
流水线任务
计划
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
Mace
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
DevOps
DevOps
流水线
流水线任务
计划
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
流水线任务
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
提交
fe4cefd1
编写于
4月 18, 2018
作者:
L
liuqi
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Use small input for out of range test for mali gpu.
上级
9e305d67
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
70 addition
and
72 deletion
+70
-72
mace/core/runtime/opencl/opencl_runtime.cc
mace/core/runtime/opencl/opencl_runtime.cc
+45
-44
mace/core/runtime/opencl/opencl_runtime.h
mace/core/runtime/opencl/opencl_runtime.h
+1
-1
mace/kernels/opencl/out_of_range_check_test.cc
mace/kernels/opencl/out_of_range_check_test.cc
+24
-27
未找到文件。
mace/core/runtime/opencl/opencl_runtime.cc
浏览文件 @
fe4cefd1
...
...
@@ -161,7 +161,47 @@ void OpenCLPrintfCallback(const char *buffer,
void
*
user_data
)
{
fwrite
(
buffer
,
1
,
length
,
stdout
);
}
void
GetAdrenoContextProperties
(
std
::
vector
<
cl_context_properties
>
*
properties
,
GPUPerfHint
gpu_perf_hint
,
GPUPriorityHint
gpu_priority_hint
)
{
MACE_CHECK_NOTNULL
(
properties
);
switch
(
gpu_perf_hint
)
{
case
GPUPerfHint
::
PERF_LOW
:
properties
->
push_back
(
CL_CONTEXT_PERF_HINT_QCOM
);
properties
->
push_back
(
CL_PERF_HINT_LOW_QCOM
);
break
;
case
GPUPerfHint
::
PERF_NORMAL
:
properties
->
push_back
(
CL_CONTEXT_PERF_HINT_QCOM
);
properties
->
push_back
(
CL_PERF_HINT_NORMAL_QCOM
);
break
;
case
GPUPerfHint
::
PERF_HIGH
:
properties
->
push_back
(
CL_CONTEXT_PERF_HINT_QCOM
);
properties
->
push_back
(
CL_PERF_HINT_HIGH_QCOM
);
break
;
default:
break
;
}
switch
(
gpu_priority_hint
)
{
case
GPUPriorityHint
::
PRIORITY_LOW
:
properties
->
push_back
(
CL_CONTEXT_PRIORITY_HINT_QCOM
);
properties
->
push_back
(
CL_PRIORITY_HINT_LOW_QCOM
);
break
;
case
GPUPriorityHint
::
PRIORITY_NORMAL
:
properties
->
push_back
(
CL_CONTEXT_PRIORITY_HINT_QCOM
);
properties
->
push_back
(
CL_PRIORITY_HINT_NORMAL_QCOM
);
break
;
case
GPUPriorityHint
::
PRIORITY_HIGH
:
properties
->
push_back
(
CL_CONTEXT_PRIORITY_HINT_QCOM
);
properties
->
push_back
(
CL_PRIORITY_HINT_HIGH_QCOM
);
break
;
default:
break
;
}
// The properties list should be terminated with 0
properties
->
push_back
(
0
);
}
}
// namespace
void
OpenCLProfilingTimer
::
StartTiming
()
{}
...
...
@@ -193,7 +233,7 @@ GPUPriorityHint OpenCLRuntime::kGPUPriorityHint =
GPUPriorityHint
::
PRIORITY_DEFAULT
;
OpenCLRuntime
*
OpenCLRuntime
::
Global
()
{
static
OpenCLRuntime
runtime
(
kGPUPerfHint
,
kGPUPriorityHint
)
;
static
OpenCLRuntime
runtime
;
return
&
runtime
;
}
...
...
@@ -203,48 +243,8 @@ void OpenCLRuntime::Configure(GPUPerfHint gpu_perf_hint,
OpenCLRuntime
::
kGPUPriorityHint
=
gpu_priority_hint
;
}
void
GetAdrenoContextProperties
(
std
::
vector
<
cl_context_properties
>
*
properties
,
GPUPerfHint
gpu_perf_hint
,
GPUPriorityHint
gpu_priority_hint
)
{
MACE_CHECK_NOTNULL
(
properties
);
switch
(
gpu_perf_hint
)
{
case
GPUPerfHint
::
PERF_LOW
:
properties
->
push_back
(
CL_CONTEXT_PERF_HINT_QCOM
);
properties
->
push_back
(
CL_PERF_HINT_LOW_QCOM
);
break
;
case
GPUPerfHint
::
PERF_NORMAL
:
properties
->
push_back
(
CL_CONTEXT_PERF_HINT_QCOM
);
properties
->
push_back
(
CL_PERF_HINT_NORMAL_QCOM
);
break
;
case
GPUPerfHint
::
PERF_HIGH
:
properties
->
push_back
(
CL_CONTEXT_PERF_HINT_QCOM
);
properties
->
push_back
(
CL_PERF_HINT_HIGH_QCOM
);
break
;
default:
break
;
}
switch
(
gpu_priority_hint
)
{
case
GPUPriorityHint
::
PRIORITY_LOW
:
properties
->
push_back
(
CL_CONTEXT_PRIORITY_HINT_QCOM
);
properties
->
push_back
(
CL_PRIORITY_HINT_LOW_QCOM
);
break
;
case
GPUPriorityHint
::
PRIORITY_NORMAL
:
properties
->
push_back
(
CL_CONTEXT_PRIORITY_HINT_QCOM
);
properties
->
push_back
(
CL_PRIORITY_HINT_NORMAL_QCOM
);
break
;
case
GPUPriorityHint
::
PRIORITY_HIGH
:
properties
->
push_back
(
CL_CONTEXT_PRIORITY_HINT_QCOM
);
properties
->
push_back
(
CL_PRIORITY_HINT_HIGH_QCOM
);
break
;
default:
break
;
}
// The properties list should be terminated with 0
properties
->
push_back
(
0
);
}
OpenCLRuntime
::
OpenCLRuntime
(
GPUPerfHint
gpu_perf_hint
,
GPUPriorityHint
gpu_priority_hint
)
:
OpenCLRuntime
::
OpenCLRuntime
()
:
storage_
(
nullptr
),
is_profiling_enabled_
(
false
)
{
LoadOpenCLLibrary
();
...
...
@@ -302,8 +302,9 @@ OpenCLRuntime::OpenCLRuntime(GPUPerfHint gpu_perf_hint,
if
(
gpu_type_
==
GPUType
::
QUALCOMM_ADRENO
)
{
std
::
vector
<
cl_context_properties
>
context_properties
;
context_properties
.
reserve
(
5
);
GetAdrenoContextProperties
(
&
context_properties
,
gpu_perf_hint
,
gpu_priority_hint
);
GetAdrenoContextProperties
(
&
context_properties
,
OpenCLRuntime
::
kGPUPerfHint
,
OpenCLRuntime
::
kGPUPriorityHint
);
context_
=
std
::
shared_ptr
<
cl
::
Context
>
(
new
cl
::
Context
({
*
device_
},
context_properties
.
data
(),
nullptr
,
nullptr
,
&
err
));
...
...
mace/core/runtime/opencl/opencl_runtime.h
浏览文件 @
fe4cefd1
...
...
@@ -80,7 +80,7 @@ class OpenCLRuntime {
const
bool
is_profiling_enabled
()
const
;
private:
OpenCLRuntime
(
GPUPerfHint
,
GPUPriorityHint
);
OpenCLRuntime
();
~
OpenCLRuntime
();
OpenCLRuntime
(
const
OpenCLRuntime
&
)
=
delete
;
OpenCLRuntime
&
operator
=
(
const
OpenCLRuntime
&
)
=
delete
;
...
...
mace/kernels/opencl/out_of_range_check_test.cc
浏览文件 @
fe4cefd1
...
...
@@ -118,34 +118,31 @@ class OutOfRangeCheckTest : public ::testing::Test {
};
TEST
(
OutOfRangeCheckTest
,
RandomTest
)
{
static
unsigned
int
seed
=
time
(
NULL
);
for
(
int
round
=
0
;
round
<
10
;
++
round
)
{
index_t
batch
=
11
+
rand_r
(
&
seed
)
%
10
;
index_t
height
=
12
+
rand_r
(
&
seed
)
%
100
;
index_t
width
=
13
+
rand_r
(
&
seed
)
%
100
;
index_t
channels
=
14
+
rand_r
(
&
seed
)
%
50
;
std
::
vector
<
index_t
>
buffer_shape
=
{
batch
,
height
,
width
,
channels
};
Workspace
ws
;
Tensor
*
buffer
=
ws
.
CreateTensor
(
"Buffer"
,
GetDeviceAllocator
(
DeviceType
::
OPENCL
),
DataTypeToEnum
<
float
>::
v
());
buffer
->
Resize
(
buffer_shape
);
std
::
vector
<
size_t
>
image_shape
;
Tensor
*
image
=
ws
.
CreateTensor
(
"Image"
,
GetDeviceAllocator
(
DeviceType
::
OPENCL
),
DataTypeToEnum
<
float
>::
v
());
CalImage2DShape
(
buffer
->
shape
(),
IN_OUT_CHANNEL
,
&
image_shape
);
image
->
ResizeImage
(
buffer
->
shape
(),
image_shape
);
ASSERT_FALSE
(
BufferToImageOpImpl
(
buffer
,
image
,
image_shape
));
std
::
vector
<
size_t
>
overflow_image_shape
=
image_shape
;
for
(
int
i
=
0
;
i
<
overflow_image_shape
.
size
();
++
i
)
{
overflow_image_shape
[
i
]
+=
1
;
}
ASSERT_TRUE
(
BufferToImageOpImpl
(
buffer
,
image
,
overflow_image_shape
));
index_t
batch
=
3
;
index_t
height
=
5
;
index_t
width
=
7
;
index_t
channels
=
11
;
std
::
vector
<
index_t
>
buffer_shape
=
{
batch
,
height
,
width
,
channels
};
Workspace
ws
;
Tensor
*
buffer
=
ws
.
CreateTensor
(
"Buffer"
,
GetDeviceAllocator
(
DeviceType
::
OPENCL
),
DataTypeToEnum
<
float
>::
v
());
buffer
->
Resize
(
buffer_shape
);
std
::
vector
<
size_t
>
image_shape
;
Tensor
*
image
=
ws
.
CreateTensor
(
"Image"
,
GetDeviceAllocator
(
DeviceType
::
OPENCL
),
DataTypeToEnum
<
float
>::
v
());
CalImage2DShape
(
buffer
->
shape
(),
IN_OUT_CHANNEL
,
&
image_shape
);
image
->
ResizeImage
(
buffer
->
shape
(),
image_shape
);
ASSERT_FALSE
(
BufferToImageOpImpl
(
buffer
,
image
,
image_shape
));
std
::
vector
<
size_t
>
overflow_image_shape
=
image_shape
;
for
(
int
i
=
0
;
i
<
overflow_image_shape
.
size
();
++
i
)
{
overflow_image_shape
[
i
]
+=
1
;
}
ASSERT_TRUE
(
BufferToImageOpImpl
(
buffer
,
image
,
overflow_image_shape
));
}
}
// namespace kernels
...
...
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