1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
|
/* Copyright 2015 Google Inc. 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 "tensorflow/stream_executor/cuda/cuda_timer.h"
#include "tensorflow/stream_executor/cuda/cuda_driver.h"
#include "tensorflow/stream_executor/cuda/cuda_gpu_executor.h"
#include "tensorflow/stream_executor/cuda/cuda_stream.h"
#include "tensorflow/stream_executor/lib/status.h"
namespace perftools {
namespace gputools {
namespace cuda {
bool CUDATimer::Init() {
CHECK(start_event_ == nullptr && stop_event_ == nullptr);
CUcontext context = parent_->cuda_context();
if (!CUDADriver::CreateEvent(context, &start_event_,
CUDADriver::EventFlags::kDefault)
.ok()) {
return false;
}
if (!CUDADriver::CreateEvent(context, &stop_event_,
CUDADriver::EventFlags::kDefault)
.ok()) {
port::Status status = CUDADriver::DestroyEvent(context, &start_event_);
if (!status.ok()) {
LOG(ERROR) << status;
}
return false;
}
CHECK(start_event_ != nullptr && stop_event_ != nullptr);
return true;
}
void CUDATimer::Destroy() {
CUcontext context = parent_->cuda_context();
port::Status status = CUDADriver::DestroyEvent(context, &start_event_);
if (!status.ok()) {
LOG(ERROR) << status;
}
status = CUDADriver::DestroyEvent(context, &stop_event_);
if (!status.ok()) {
LOG(ERROR) << status;
}
}
float CUDATimer::GetElapsedMilliseconds() const {
CHECK(start_event_ != nullptr && stop_event_ != nullptr);
// TODO(leary) provide a way to query timer resolution?
// CUDA docs say a resolution of about 0.5us
float elapsed_milliseconds = NAN;
(void)CUDADriver::GetEventElapsedTime(parent_->cuda_context(),
&elapsed_milliseconds, start_event_,
stop_event_);
return elapsed_milliseconds;
}
bool CUDATimer::Start(CUDAStream *stream) {
return CUDADriver::RecordEvent(parent_->cuda_context(), start_event_,
stream->cuda_stream())
.ok();
}
bool CUDATimer::Stop(CUDAStream *stream) {
return CUDADriver::RecordEvent(parent_->cuda_context(), stop_event_,
stream->cuda_stream())
.ok();
}
} // namespace cuda
} // namespace gputools
} // namespace perftools
|
