FazBrowse GitHub Viewer
|
Trending
|
URL:
|
Home
Tools:
[Download Repo ZIP]
[View Raw Code]
[Original HTTPS Page]
opencv/modules/core/src/parallel_impl.cpp at master · pythonmjs/opencv · GitHub
pythonmjs
/
opencv
Public
forked from
opencv/opencv
Notifications
You must be signed in to change notification settings
Fork
0
Star
0
Code
Pull requests
0
Actions
Projects
Wiki
Security and quality
0
Insights
Additional navigation options
Code
Pull requests
Actions
Projects
Wiki
Security and quality
Insights
Expand file tree
Breadcrumbs
opencv
/
modules
/
core
/
src
/
parallel_impl.cpp
Copy path
More file actions
More file actions
Latest commit
History
History
History
723 lines (648 loc) · 25.2 KB
Breadcrumbs
opencv
/
modules
/
core
/
src
/
parallel_impl.cpp
Copy path
File metadata and controls
723 lines (648 loc) · 25.2 KB
Raw
Copy raw file
Download raw file
Open symbols panel
Edit and raw actions
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
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
//
This file is part of OpenCV project.
//
It is subject to the license terms in the LICENSE file found in the top-level directory
//
of this distribution and at http://opencv.org/license.html.
#
include
"
precomp.hpp
"
#
include
"
parallel_impl.hpp
"
#
ifdef
HAVE_PTHREADS_PF
#
include
<
pthread.h
>
#
include
<
opencv2/core/utils/configuration.private.hpp
>
#
include
<
opencv2/core/utils/logger.defines.hpp
>
//
#undef CV_LOG_STRIP_LEVEL
//
#define CV_LOG_STRIP_LEVEL CV_LOG_LEVEL_VERBOSE + 1
#
include
<
opencv2/core/utils/logger.hpp
>
//
#define CV_PROFILE_THREADS 64
//
#define getTickCount getCPUTickCount // use this if getTickCount() calls are expensive (and getCPUTickCount() is accurate)
//
#define CV_USE_GLOBAL_WORKERS_COND_VAR // not effective on many-core systems (10+)
#
include
<
atomic
>
//
Spin lock's OS-level yield
#
ifdef
DECLARE_CV_YIELD
DECLARE_CV_YIELD
#
endif
#
ifndef
CV_YIELD
#
include
<
thread
>
#
define
CV_YIELD
() std::this_thread::yield()
#
endif
//
CV_YIELD
//
Spin lock's CPU-level yield (required for Hyper-Threading)
#
ifdef
DECLARE_CV_PAUSE
DECLARE_CV_PAUSE
#
endif
#
ifndef
CV_PAUSE
#
if
defined __GNUC__ && (defined __i386__ || defined __x86_64__)
#
if
!defined(__SSE__)
static
inline
void
cv_non_sse_mm_pause
() {
__asm__
__volatile__
(
"
rep; nop
"
); }
#
define
_mm_pause
cv_non_sse_mm_pause
#
endif
#
define
CV_PAUSE
(
v
)
do
{
for
(
int
__delay = (v); __delay >
0
; --__delay) {
_mm_pause
(); } }
while
(
0
)
#
elif
defined __GNUC__ && defined __aarch64__
#
define
CV_PAUSE
(
v
)
do
{
for
(
int
__delay = (v); __delay >
0
; --__delay) {
asm
volatile
(
"
yield
"
:::
"
memory
"
); } }
while
(
0
)
#
elif
defined __GNUC__ && defined __arm__
#
define
CV_PAUSE
(
v
)
do
{
for
(
int
__delay = (v); __delay >
0
; --__delay) {
asm
volatile
(
"
"
:::
"
memory
"
); } }
while
(
0
)
#
elif
defined __GNUC__ && defined __PPC64__
#
define
CV_PAUSE
(
v
)
do
{
for
(
int
__delay = (v); __delay >
0
; --__delay) {
asm
volatile
(
"
or 27,27,27
"
:::
"
memory
"
); } }
while
(
0
)
#
else
#
warning
"Can't detect 'pause' (CPU-yield) instruction on the target platform. Specify CV_PAUSE() definition via compiler flags."
#
define
CV_PAUSE
(...)
do
{
/*
no-op: works, but not effective
*/
}
while
(
0
)
#
endif
#
endif
//
CV_PAUSE
namespace
cv
{
static
int
CV_ACTIVE_WAIT_PAUSE_LIMIT
= (
int
)utils::getConfigurationParameterSizeT(
"
OPENCV_THREAD_POOL_ACTIVE_WAIT_PAUSE_LIMIT
"
,
16
);
//
iterations
static
int
CV_WORKER_ACTIVE_WAIT
= (
int
)utils::getConfigurationParameterSizeT(
"
OPENCV_THREAD_POOL_ACTIVE_WAIT_WORKER
"
,
2000
);
//
iterations
static
int
CV_MAIN_THREAD_ACTIVE_WAIT
= (
int
)utils::getConfigurationParameterSizeT(
"
OPENCV_THREAD_POOL_ACTIVE_WAIT_MAIN
"
,
10000
);
//
iterations
static
int
CV_WORKER_ACTIVE_WAIT_THREADS_LIMIT
= (
int
)utils::getConfigurationParameterSizeT(
"
OPENCV_THREAD_POOL_ACTIVE_WAIT_THREADS_LIMIT
"
,
0
);
//
number of real cores
class
WorkerThread
;
class
ParallelJob
;
class
ThreadPool
{
public:
static
ThreadPool&
instance
()
{
CV_SINGLETON_LAZY_INIT_REF
(ThreadPool,
new
ThreadPool
())
}
static
void
stop
()
{
ThreadPool& manager =
instance
();
manager.
reconfigure
(
0
);
}
void
reconfigure
(
unsigned
new_threads_count)
{
if
(new_threads_count == threads.
size
())
return
;
pthread_mutex_lock
(&mutex);
reconfigure_
(new_threads_count);
pthread_mutex_unlock
(&mutex);
}
bool
reconfigure_
(
unsigned
new_threads_count);
//
internal implementation
void
run
(
const
Range& range,
const
ParallelLoopBody& body,
double
nstripes);
size_t
getNumOfThreads
();
void
setNumOfThreads
(
unsigned
n);
ThreadPool
();
~ThreadPool
();
unsigned
num_threads;
pthread_mutex_t
mutex;
//
guards fields (job/threads) from non-worker threads (concurrent parallel_for calls)
#
if
defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_cond_t
cond_thread_wake;
#
endif
pthread_mutex_t
mutex_notify;
pthread_cond_t
cond_thread_task_complete;
std::vector< Ptr<WorkerThread> > threads;
Ptr<ParallelJob> job;
#
ifdef
CV_PROFILE_THREADS
double
tickFreq;
int64 jobSubmitTime;
struct
ThreadStatistics
{
ThreadStatistics
() : threadWait(
0
)
{
reset
();
}
void
reset
()
{
threadWake =
0
;
threadExecuteStart =
0
;
threadExecuteStop =
0
;
executedTasks =
0
;
keepActive =
false
;
threadPing =
getTickCount
();
}
int64 threadWait;
//
don't reset by default
int64 threadPing;
//
don't reset by default
int64 threadWake;
int64 threadExecuteStart;
int64 threadExecuteStop;
int64 threadFree;
unsigned
executedTasks;
bool
keepActive;
int64 dummy_[
8
];
//
separate cache lines
void
dump
(
int
id, int64 baseTime,
double
tickFreq)
{
if
(id <
0
)
std::cout <<
"
Main:
"
;
else
printf
(
"
T%03d:
"
, id +
2
);
printf
(
"
wait=% 10.1f ping=% 6.1f
"
,
threadWait >
0
? (threadWait - baseTime) / tickFreq *
1e6
: -
0.0
,
threadPing >
0
? (threadPing - baseTime) / tickFreq *
1e6
: -
0.0
);
if
(threadWake >
0
)
printf
(
"
wake=% 6.1f
"
,
(threadWake >
0
? (threadWake - baseTime) / tickFreq *
1e6
: -
0.0
));
if
(threadExecuteStart >
0
)
{
printf
(
"
exec=% 6.1f - % 6.1f tasksDone=%5u free=% 6.1f
"
,
(threadExecuteStart >
0
? (threadExecuteStart - baseTime) / tickFreq *
1e6
: -
0.0
),
(threadExecuteStop >
0
? (threadExecuteStop - baseTime) / tickFreq *
1e6
: -
0.0
),
executedTasks,
(threadFree >
0
? (threadFree - baseTime) / tickFreq *
1e6
: -
0.0
));
if
(id >=
0
)
printf
(
"
active=%s
\n
"
, keepActive ?
"
true
"
:
"
false
"
);
else
printf
(
"
\n
"
);
}
else
printf
(
"
------------------------------------------------------------------------------
\n
"
);
}
};
ThreadStatistics threads_stat[
CV_PROFILE_THREADS
];
//
0 - main thread, 1..N - worker threads
#
endif
};
class
WorkerThread
{
public:
ThreadPool& thread_pool;
const
unsigned
id;
pthread_t
posix_thread;
bool
is_created;
volatile
bool
stop_thread;
volatile
bool
has_wake_signal;
Ptr<ParallelJob> job;
pthread_mutex_t
mutex;
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
volatile
bool
isActive;
pthread_cond_t
cond_thread_wake;
#
endif
WorkerThread
(ThreadPool& thread_pool_,
unsigned
id_) :
thread_pool
(thread_pool_),
id
(id_),
posix_thread
(
0
),
is_created
(
false
),
stop_thread
(
false
),
has_wake_signal
(
false
)
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
, isActive(
true
)
#
endif
{
CV_LOG_VERBOSE
(
NULL
,
1
,
"
MainThread: initializing new worker:
"
<< id);
int
res =
pthread_mutex_init
(&mutex,
NULL
);
if
(res !=
0
)
{
CV_LOG_ERROR
(
NULL
, id <<
"
: Can't create thread mutex: res =
"
<< res);
return
;
}
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
res =
pthread_cond_init
(&cond_thread_wake,
NULL
);
if
(res !=
0
)
{
CV_LOG_ERROR
(
NULL
, id <<
"
: Can't create thread condition variable: res =
"
<< res);
return
;
}
#
endif
res =
pthread_create
(&posix_thread,
NULL
, thread_loop_wrapper, (
void
*)
this
);
if
(res !=
0
)
{
CV_LOG_ERROR
(
NULL
, id <<
"
: Can't spawn new thread: res =
"
<< res);
}
else
{
is_created =
true
;
}
}
~WorkerThread
()
{
CV_LOG_VERBOSE
(
NULL
,
1
,
"
MainThread: destroy worker thread:
"
<< id);
if
(is_created)
{
if
(!stop_thread)
{
pthread_mutex_lock
(&mutex);
//
to avoid signal miss due pre-check
stop_thread =
true
;
pthread_mutex_unlock
(&mutex);
#
if
defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_cond_broadcast
(&thread_pool.
cond_thread_wake
);
#
else
pthread_cond_signal
(&cond_thread_wake);
#
endif
}
pthread_join
(posix_thread,
NULL
);
}
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_cond_destroy
(&cond_thread_wake);
#
endif
pthread_mutex_destroy
(&mutex);
}
void
thread_body
();
static
void
*
thread_loop_wrapper
(
void
* thread_object)
{
((WorkerThread*)thread_object)->
thread_body
();
return
0
;
}
};
class
ParallelJob
{
public:
ParallelJob
(
const
ThreadPool& thread_pool_,
const
Range& range_,
const
ParallelLoopBody& body_,
int
nstripes_) :
thread_pool
(thread_pool_),
body
(body_),
range
(range_),
nstripes
((
unsigned
)nstripes_),
is_completed
(
false
)
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
ParallelJob::ParallelJob(
"
<< (
void
*)
this
<<
"
)
"
);
current_task.
store
(
0
, std::memory_order_relaxed);
active_thread_count.
store
(
0
, std::memory_order_relaxed);
completed_thread_count.
store
(
0
, std::memory_order_relaxed);
dummy0_[
0
] =
0
, dummy1_[
0
] =
0
, dummy2_[
0
] =
0
;
//
compiler warning
}
~ParallelJob
()
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
ParallelJob::~ParallelJob(
"
<< (
void
*)
this
<<
"
)
"
);
}
unsigned
execute
(
bool
is_worker_thread)
{
unsigned
executed_tasks =
0
;
const
int
task_count = range.
size
();
const
int
remaining_multiplier =
std::min
(nstripes,
std::max
(
std::min
(
100u
, thread_pool.
num_threads
*
4
),
thread_pool.
num_threads
*
2
));
//
experimental value
for
(;;)
{
int
chunk_size =
std::max
(
1
, (task_count - current_task) / remaining_multiplier);
int
id = current_task.
fetch_add
(chunk_size, std::memory_order_seq_cst);
if
(id >= task_count)
break
;
//
no more free tasks
executed_tasks += chunk_size;
int
start_id = id;
int
end_id =
std::min
(task_count, id + chunk_size);
CV_LOG_VERBOSE
(
NULL
,
9
,
"
Thread: job
"
<< start_id <<
"
-
"
<< end_id);
//
TODO: if (not pending exception)
{
body.
operator
()(
Range
(range.
start
+ start_id, range.
start
+ end_id));
}
if
(is_worker_thread && is_completed)
{
CV_LOG_ERROR
(
NULL
,
"
\t\t\t\t
BUG! Job:
"
<< (
void
*)
this
<<
"
"
<< id <<
"
"
<< active_thread_count <<
"
"
<< completed_thread_count);
CV_Assert
(!is_completed);
//
TODO Dbg this
}
}
return
executed_tasks;
}
const
ThreadPool& thread_pool;
const
ParallelLoopBody& body;
const
Range range;
const
unsigned
nstripes;
std::atomic<
int
> current_task;
//
next free part of job
int64 dummy0_[
8
];
//
avoid cache-line reusing for the same atomics
std::atomic<
int
> active_thread_count;
//
number of threads worked on this job
int64 dummy1_[
8
];
//
avoid cache-line reusing for the same atomics
std::atomic<
int
> completed_thread_count;
//
number of threads completed any activities on this job
int64 dummy2_[
8
];
//
avoid cache-line reusing for the same atomics
volatile
bool
is_completed;
//
std::atomic_flag ?
//
TODO exception handling
};
void
WorkerThread::thread_body
()
{
(
void
)
cv::utils::getThreadID
();
//
notify OpenCV about new thread
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: new thread:
"
<< id);
bool
allow_active_wait =
true
;
#
ifdef
CV_PROFILE_THREADS
ThreadPool::ThreadStatistics& stat = thread_pool.
threads_stat
[id +
1
];
#
endif
while
(!stop_thread)
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: ... loop iteration: allow_active_wait=
"
<< allow_active_wait <<
"
has_wake_signal=
"
<< has_wake_signal);
if
(allow_active_wait &&
CV_WORKER_ACTIVE_WAIT
>
0
)
{
allow_active_wait =
false
;
for
(
int
i =
0
; i <
CV_WORKER_ACTIVE_WAIT
; i++)
{
if
(has_wake_signal)
break
;
if
(
CV_ACTIVE_WAIT_PAUSE_LIMIT
>
0
&& (i <
CV_ACTIVE_WAIT_PAUSE_LIMIT
|| (i &
1
)))
CV_PAUSE
(
16
);
else
CV_YIELD
();
}
}
pthread_mutex_lock
(&mutex);
#
ifdef
CV_PROFILE_THREADS
stat.
threadWait
=
getTickCount
();
#
endif
while
(!has_wake_signal)
//
to handle spurious wakeups
{
//
CV_LOG_VERBOSE(NULL, 5, "Thread: wait (sleep) ...");
#
if
defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_cond_wait
(&thread_pool.
cond_thread_wake
, &mutex);
#
else
isActive =
false
;
pthread_cond_wait
(&cond_thread_wake, &mutex);
isActive =
true
;
#
endif
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: wake ... (has_wake_signal=
"
<< has_wake_signal <<
"
stop_thread=
"
<< stop_thread <<
"
)
"
)
}
#
ifdef
CV_PROFILE_THREADS
stat.
threadWake
=
getTickCount
();
#
endif
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: checking for new job
"
);
if
(
CV_WORKER_ACTIVE_WAIT_THREADS_LIMIT
==
0
)
allow_active_wait =
true
;
Ptr<ParallelJob> j_ptr;
swap
(j_ptr, job);
has_wake_signal =
false
;
//
TODO .store(false, std::memory_order_release)
pthread_mutex_unlock
(&mutex);
if
(!stop_thread)
{
ParallelJob* j = j_ptr;
if
(j)
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: job size=
"
<< j->
range
.
size
() <<
"
done=
"
<< j->
current_task
);
if
(j->
current_task
< j->
range
.
size
())
{
int
other = j->
active_thread_count
.
fetch_add
(
1
, std::memory_order_seq_cst);
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: processing new job (with
"
<< other <<
"
other threads)
"
);
CV_UNUSED
(other);
#
ifdef
CV_PROFILE_THREADS
stat.
threadExecuteStart
=
getTickCount
();
stat.
executedTasks
= j->
execute
(
true
);
stat.
threadExecuteStop
=
getTickCount
();
#
else
j->
execute
(
true
);
#
endif
int
completed = j->
completed_thread_count
.
fetch_add
(
1
, std::memory_order_seq_cst) +
1
;
int
active = j->
active_thread_count
.
load
(std::memory_order_acquire);
if
(
CV_WORKER_ACTIVE_WAIT_THREADS_LIMIT
>
0
)
{
allow_active_wait =
true
;
if
(active >=
CV_WORKER_ACTIVE_WAIT_THREADS_LIMIT
&& (id &
1
) ==
0
)
//
turn off a half of threads
allow_active_wait =
false
;
}
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: completed job processing:
"
<< active <<
"
"
<< completed);
if
(active == completed)
{
bool
need_signal = !j->
is_completed
;
j->
is_completed
=
true
;
j =
NULL
; j_ptr.
release
();
if
(need_signal)
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: job finished => notifying the main thread
"
);
pthread_mutex_lock
(&thread_pool.
mutex_notify
);
//
to avoid signal miss due pre-check condition
//
empty
pthread_mutex_unlock
(&thread_pool.
mutex_notify
);
pthread_cond_broadcast
/*
pthread_cond_signal
*/
(&thread_pool.
cond_thread_task_complete
);
}
}
}
else
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
Thread: no free job tasks
"
);
}
}
}
#
ifdef
CV_PROFILE_THREADS
stat.
threadFree
=
getTickCount
();
stat.
keepActive
= allow_active_wait;
#
endif
}
}
ThreadPool::ThreadPool
()
{
#
ifdef
CV_PROFILE_THREADS
tickFreq =
getTickFrequency
();
#
endif
int
res =
0
;
res |=
pthread_mutex_init
(&mutex,
NULL
);
res |=
pthread_mutex_init
(&mutex_notify,
NULL
);
#
if
defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
res |=
pthread_cond_init
(&cond_thread_wake,
NULL
);
#
endif
res |=
pthread_cond_init
(&cond_thread_task_complete,
NULL
);
if
(
0
!= res)
{
CV_LOG_FATAL
(
NULL
,
"
Failed to initialize ThreadPool (pthreads)
"
);
}
num_threads =
defaultNumberOfThreads
();
}
bool
ThreadPool::reconfigure_
(
unsigned
new_threads_count)
{
if
(new_threads_count == threads.
size
())
return
false
;
if
(new_threads_count < threads.
size
())
{
CV_LOG_VERBOSE
(
NULL
,
1
,
"
MainThread: reduce worker pool:
"
<< threads.
size
() <<
"
=>
"
<< new_threads_count);
std::vector< Ptr<WorkerThread> >
release_threads
(threads.
size
() - new_threads_count);
for
(
size_t
i = new_threads_count; i < threads.
size
(); ++i)
{
pthread_mutex_lock
(&threads[i]->
mutex
);
//
to avoid signal miss due pre-check
threads[i]->
stop_thread
=
true
;
threads[i]->
has_wake_signal
=
true
;
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_mutex_unlock
(&threads[i]->
mutex
);
pthread_cond_broadcast
/*
pthread_cond_signal
*/
(&threads[i]->
cond_thread_wake
);
//
wake thread
#
else
pthread_mutex_unlock
(&threads[i]->
mutex
);
#
endif
std::swap
(threads[i], release_threads[i - new_threads_count]);
}
#
if
defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
CV_LOG_VERBOSE
(
NULL
,
1
,
"
MainThread: notify worker threads about termination...
"
);
pthread_cond_broadcast
(&cond_thread_wake);
//
wake all threads
#
endif
threads.
resize
(new_threads_count);
release_threads.
clear
();
//
calls thread_join which want to lock mutex
return
false
;
}
else
{
CV_LOG_VERBOSE
(
NULL
,
1
,
"
MainThread: upgrade worker pool:
"
<< threads.
size
() <<
"
=>
"
<< new_threads_count);
for
(
size_t
i = threads.
size
(); i < new_threads_count; ++i)
{
threads.
push_back
(Ptr<WorkerThread>(
new
WorkerThread
(*
this
, (
unsigned
)i)));
//
spawn more threads
}
}
return
false
;
}
ThreadPool::~ThreadPool
()
{
reconfigure
(
0
);
pthread_cond_destroy
(&cond_thread_task_complete);
#
if
defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_cond_destroy
(&cond_thread_wake);
#
endif
pthread_mutex_destroy
(&mutex);
pthread_mutex_destroy
(&mutex_notify);
}
void
ThreadPool::run
(
const
Range& range,
const
ParallelLoopBody& body,
double
nstripes)
{
CV_LOG_VERBOSE
(
NULL
,
1
,
"
MainThread: new parallel job: num_threads=
"
<< num_threads <<
"
range=
"
<< range.
size
() <<
"
nstripes=
"
<< nstripes <<
"
job=
"
<< (
void
*)job);
#
ifdef
CV_PROFILE_THREADS
jobSubmitTime =
getTickCount
();
threads_stat[
0
].
reset
();
threads_stat[
0
].
threadWait
= jobSubmitTime;
threads_stat[
0
].
threadWake
= jobSubmitTime;
#
endif
if
(
getNumOfThreads
() >
1
&&
job ==
NULL
&&
(range.
size
() * nstripes >=
2
|| (range.
size
() >
1
&& nstripes <=
0
))
)
{
pthread_mutex_lock
(&mutex);
if
(job !=
NULL
)
{
pthread_mutex_unlock
(&mutex);
body
(range);
return
;
}
reconfigure_
(num_threads -
1
);
{
CV_LOG_VERBOSE
(
NULL
,
1
,
"
MainThread: initialize parallel job:
"
<< range.
size
());
job = Ptr<ParallelJob>(
new
ParallelJob
(*
this
, range, body, nstripes));
pthread_mutex_unlock
(&mutex);
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: wake worker threads...
"
);
for
(
size_t
i =
0
; i < threads.
size
(); ++i)
{
WorkerThread& thread = *(threads[i].
get
());
if
(
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
thread.
isActive
||
#
endif
thread.
has_wake_signal
|| !thread.
job
.
empty
()
//
#10881
)
{
pthread_mutex_lock
(&thread.
mutex
);
thread.
job
= job;
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
bool
isActive = thread.
isActive
;
#
endif
thread.
has_wake_signal
=
true
;
#
ifdef
CV_PROFILE_THREADS
threads_stat[i +
1
].
reset
();
#
endif
pthread_mutex_unlock
(&thread.
mutex
);
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
if
(!isActive)
{
pthread_cond_broadcast
/*
pthread_cond_signal
*/
(&thread.
cond_thread_wake
);
//
wake thread
}
#
endif
}
else
{
CV_Assert
(thread.
job
.
empty
());
thread.
job
= job;
thread.
has_wake_signal
=
true
;
#
ifdef
CV_PROFILE_THREADS
threads_stat[i +
1
].
reset
();
#
endif
#
if
!defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_cond_broadcast
/*
pthread_cond_signal
*/
(&thread.
cond_thread_wake
);
//
wake thread
#
endif
}
}
#
ifdef
CV_PROFILE_THREADS
threads_stat[
0
].
threadPing
=
getTickCount
();
#
endif
#
if
defined(CV_USE_GLOBAL_WORKERS_COND_VAR)
pthread_cond_broadcast
(&cond_thread_wake);
//
wake all threads
#
endif
#
ifdef
CV_PROFILE_THREADS
threads_stat[
0
].
threadWake
=
getTickCount
();
#
endif
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: wake worker threads... (done)
"
);
{
ParallelJob& j = *(
this
->
job
);
#
ifdef
CV_PROFILE_THREADS
threads_stat[
0
].
threadExecuteStart
=
getTickCount
();
threads_stat[
0
].
executedTasks
= j.
execute
(
false
);
threads_stat[
0
].
threadExecuteStop
=
getTickCount
();
#
else
j.
execute
(
false
);
#
endif
CV_Assert
(j.
current_task
>= j.
range
.
size
());
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: complete self-tasks:
"
<< j.
active_thread_count
<<
"
"
<< j.
completed_thread_count
);
if
(job->
is_completed
|| j.
active_thread_count
==
0
)
{
job->
is_completed
=
true
;
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: no WIP worker threads
"
);
}
else
{
if
(
CV_MAIN_THREAD_ACTIVE_WAIT
>
0
)
{
for
(
int
i =
0
; i <
CV_MAIN_THREAD_ACTIVE_WAIT
; i++)
//
don't spin too much in any case (inaccurate getTickCount())
{
if
(job->
is_completed
)
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: job finalize (active wait)
"
<< j.
active_thread_count
<<
"
"
<< j.
completed_thread_count
);
break
;
}
if
(
CV_ACTIVE_WAIT_PAUSE_LIMIT
>
0
&& (i <
CV_ACTIVE_WAIT_PAUSE_LIMIT
|| (i &
1
)))
CV_PAUSE
(
16
);
else
CV_YIELD
();
}
}
if
(!job->
is_completed
)
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: prepare wait
"
<< j.
active_thread_count
<<
"
"
<< j.
completed_thread_count
);
pthread_mutex_lock
(&mutex_notify);
for
(;;)
{
if
(job->
is_completed
)
{
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: job finalize (wait)
"
<< j.
active_thread_count
<<
"
"
<< j.
completed_thread_count
);
break
;
}
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: wait completion (sleep) ...
"
);
pthread_cond_wait
(&cond_thread_task_complete, &mutex_notify);
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: wake
"
);
}
pthread_mutex_unlock
(&mutex_notify);
}
}
}
#
ifdef
CV_PROFILE_THREADS
threads_stat[
0
].
threadFree
=
getTickCount
();
std::cout <<
"
Job: sz=
"
<< range.
size
() <<
"
nstripes=
"
<< nstripes <<
"
Time:
"
<< (threads_stat[
0
].
threadFree
- jobSubmitTime) / tickFreq *
1e6
<<
"
usec
"
<< std::endl;
for
(
int
i =
0
; i < (
int
)threads.
size
() +
1
; i++)
{
threads_stat[i].
dump
(i -
1
, jobSubmitTime, tickFreq);
}
#
endif
if
(job)
{
pthread_mutex_lock
(&mutex);
CV_LOG_VERBOSE
(
NULL
,
5
,
"
MainThread: job release
"
);
CV_Assert
(job->
is_completed
);
job.
release
();
pthread_mutex_unlock
(&mutex);
}
}
}
else
{
body
(range);
}
}
size_t
ThreadPool::getNumOfThreads
()
{
return
num_threads;
}
void
ThreadPool::setNumOfThreads
(
unsigned
n)
{
if
(n != num_threads)
{
num_threads = n;
if
(n ==
1
)
if
(job ==
NULL
)
reconfigure
(
0
);
//
stop worker threads immediately
}
}
size_t
parallel_pthreads_get_threads_num
()
{
return
ThreadPool::instance
().
getNumOfThreads
();
}
void
parallel_pthreads_set_threads_num
(
int
num)
{
if
(num <
0
)
{
ThreadPool::instance
().
setNumOfThreads
(
0
);
}
else
{
ThreadPool::instance
().
setNumOfThreads
(
unsigned
(num));
}
}
void
parallel_for_pthreads
(
const
Range& range,
const
ParallelLoopBody& body,
double
nstripes)
{
ThreadPool::instance
().
run
(range, body, nstripes);
}
}
#
endif
Back
|
FazBrowse Home
|
New Git URL