Coverage Report

Created: 2026-08-31 06:22

next uncovered line (L), next uncovered region (R), next uncovered branch (B)
/work/vvenc/source/Lib/Utilities/NoMallocThreadPool.cpp
Line
Count
Source
1
/* -----------------------------------------------------------------------------
2
The copyright in this software is being made available under the Clear BSD
3
License, included below. No patent rights, trademark rights and/or 
4
other Intellectual Property Rights other than the copyrights concerning 
5
the Software are granted under this license.
6
7
The Clear BSD License
8
9
Copyright (c) 2019-2026, Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. & The VVenC Authors.
10
All rights reserved.
11
12
Redistribution and use in source and binary forms, with or without modification,
13
are permitted (subject to the limitations in the disclaimer below) provided that
14
the following conditions are met:
15
16
     * Redistributions of source code must retain the above copyright notice,
17
     this list of conditions and the following disclaimer.
18
19
     * Redistributions in binary form must reproduce the above copyright
20
     notice, this list of conditions and the following disclaimer in the
21
     documentation and/or other materials provided with the distribution.
22
23
     * Neither the name of the copyright holder nor the names of its
24
     contributors may be used to endorse or promote products derived from this
25
     software without specific prior written permission.
26
27
NO EXPRESS OR IMPLIED LICENSES TO ANY PARTY'S PATENT RIGHTS ARE GRANTED BY
28
THIS LICENSE. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
29
CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
31
PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
32
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
33
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
34
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
35
BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
36
IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
37
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
38
POSSIBILITY OF SUCH DAMAGE.
39
40
41
------------------------------------------------------------------------------------------- */
42
43
44
/** \file     NoMallocThreadPool.cpp
45
    \brief    thread pool
46
*/
47
48
#include "NoMallocThreadPool.h"
49
50
#ifdef HAVE_PTHREADS
51
#  include <pthread.h>
52
4.82k
#  define THREAD_MIN_STACK_SIZE 1024 * 1024
53
#endif
54
55
56
//! \ingroup Utilities
57
//! \{
58
59
namespace vvenc {
60
61
#if ENABLE_TIME_PROFILING_MT_MODE
62
thread_local std::unique_ptr<TProfiler> ptls;
63
#endif
64
65
NoMallocThreadPool::NoMallocThreadPool( int numThreads, const char * threadPoolName, const VVEncCfg* encCfg )
66
1.20k
  : m_poolName( threadPoolName )
67
1.20k
{
68
1.20k
  if( numThreads < 0 )
69
0
  {
70
0
    numThreads = std::thread::hardware_concurrency();
71
0
  }
72
73
6.02k
  for( int i = 0; i < numThreads; ++i )
74
4.82k
  {
75
4.82k
    m_threads.emplace_back( &NoMallocThreadPool::threadProc, this, i, *encCfg );
76
4.82k
  }
77
1.20k
}
78
79
NoMallocThreadPool::~NoMallocThreadPool()
80
1.20k
{
81
1.20k
  m_exitThreads = true;
82
83
1.20k
  waitForThreads();
84
1.20k
}
85
86
bool NoMallocThreadPool::processTasksOnMainThread()
87
0
{
88
0
  CHECK( m_threads.size() != 0, "should not be used with multiple threads" );
89
90
0
  bool         progress      = false;
91
0
  TaskIterator firstFailedIt = m_tasks.end();
92
0
  for( auto taskIt = findNextTask( 0, m_tasks.begin() ); taskIt.isValid(); taskIt = findNextTask( 0, taskIt ) )
93
0
  {
94
0
    const bool success = processTask( 0, *taskIt );
95
0
    progress |= success;
96
97
0
    if( taskIt == firstFailedIt )
98
0
    {
99
0
      if( success )
100
0
      {
101
        // first failed was successful -> reset
102
0
        firstFailedIt = m_tasks.end();
103
0
      }
104
0
      else if( progress )
105
0
      {
106
        // reset progress, try another round
107
0
        progress = false;
108
0
      }
109
0
      else
110
0
      {
111
        // no progress -> exit
112
0
        break;
113
0
      }
114
0
    }
115
0
    else if( !success && !firstFailedIt.isValid() )
116
0
    {
117
0
      firstFailedIt = taskIt;
118
0
    }
119
0
  }
120
121
  // return true if all done (-> false if some tasks blocked due to barriers)
122
0
  return std::all_of( m_tasks.begin(), m_tasks.end(), []( Slot& t ) { return t.state == FREE; } );
123
0
}
124
125
void NoMallocThreadPool::shutdown( bool block )
126
1.20k
{
127
1.20k
  m_exitThreads = true;
128
1.20k
  if( block )
129
1.20k
  {
130
1.20k
    waitForThreads();
131
1.20k
  }
132
1.20k
}
133
134
void NoMallocThreadPool::waitForThreads()
135
2.41k
{
136
2.41k
  for( auto& t: m_threads )
137
9.64k
  {
138
9.64k
    if( t.joinable() )
139
4.82k
      t.join();
140
9.64k
  }
141
2.41k
}
142
143
void NoMallocThreadPool::threadProc( int threadId, const VVEncCfg& encCfg )
144
4.81k
{
145
4.81k
#if __linux
146
4.81k
  if( !m_poolName.empty() )
147
4.82k
  {
148
4.82k
    std::string threadName( m_poolName + std::to_string( threadId ) );
149
4.82k
    pthread_setname_np( pthread_self(), threadName.c_str() );
150
4.82k
  }
151
4.81k
#endif
152
#if ENABLE_TIME_PROFILING_MT_MODE
153
  ptls.reset( timeProfilerCreate( encCfg ) );
154
  {
155
    std::unique_lock< std::mutex > lock( m_nextFillSlotMutex );
156
    TProfiler *tp = ptls.get();
157
    profilers.push_back( tp );
158
  }
159
#endif
160
161
4.81k
  auto nextTaskIt = m_tasks.begin();
162
31.3k
  while( !m_exitThreads )
163
31.3k
  {
164
31.3k
    auto taskIt = findNextTask( threadId, nextTaskIt );
165
31.3k
    if( !taskIt.isValid() )
166
17.2k
    {
167
17.2k
      std::unique_lock<std::mutex> l( m_idleMutex, std::defer_lock );
168
169
17.2k
      ITT_TASKSTART( itt_domain_thrd, itt_handle_TPspinWait );
170
17.2k
      m_waitingThreads.fetch_add( 1, std::memory_order_relaxed );
171
17.2k
      const auto startWait = std::chrono::steady_clock::now();
172
86.7M
      while( !m_exitThreads )
173
86.7M
      {
174
86.7M
        taskIt = findNextTask( threadId, nextTaskIt );
175
86.7M
        if( taskIt.isValid() || m_exitThreads )
176
15.6k
        {
177
15.6k
          break;
178
15.6k
        }
179
180
86.7M
        if( !l.owns_lock()
181
784k
            && m_waitingThreads.load( std::memory_order_relaxed ) > 1
182
702k
            && ( BUSY_WAIT_TIME.count() == 0 || std::chrono::steady_clock::now() - startWait > BUSY_WAIT_TIME )
183
10.2k
            && !m_exitThreads )
184
10.2k
        {
185
10.2k
          ITT_TASKSTART(itt_domain_thrd, itt_handle_TPblocked);
186
10.2k
          l.lock();
187
10.2k
          ITT_TASKEND(itt_domain_thrd, itt_handle_TPblocked);
188
10.2k
        }
189
86.7M
        else
190
86.7M
        {
191
86.7M
          std::this_thread::yield();
192
86.7M
        }
193
86.7M
      }
194
17.2k
      m_waitingThreads.fetch_sub( 1, std::memory_order_relaxed );
195
17.2k
      ITT_TASKEND( itt_domain_thrd, itt_handle_TPspinWait );
196
17.2k
    }
197
31.3k
    if( m_exitThreads )
198
4.80k
    {
199
4.80k
      return;
200
4.80k
    }
201
202
26.5k
    processTask( threadId, *taskIt );
203
204
26.5k
    nextTaskIt = taskIt;
205
26.5k
    nextTaskIt.incWrap();
206
26.5k
  }
207
4.81k
}
208
209
NoMallocThreadPool::TaskIterator NoMallocThreadPool::findNextTask( int threadId, TaskIterator startSearch )
210
86.7M
{
211
86.7M
  if( !startSearch.isValid() )
212
0
  {
213
0
    startSearch = m_tasks.begin();
214
0
  }
215
86.7M
  bool first = true;
216
11.1G
  for( auto it = startSearch; it != startSearch || first; it.incWrap() )
217
11.0G
  {
218
#if ENABLE_VALGRIND_CODE
219
    MutexLock lock( m_extraMutex );
220
#endif
221
222
11.0G
    first = false;
223
224
11.0G
    Slot& t = *it;
225
11.0G
    auto expected = WAITING;
226
11.0G
    if( t.state.load( std::memory_order_relaxed ) == WAITING && t.state.compare_exchange_strong( expected, RUNNING ) )
227
128M
    {
228
128M
      if( !t.barriers.empty() )
229
38.1M
      {
230
38.1M
        if( std::any_of( t.barriers.cbegin(), t.barriers.cend(), []( const Barrier* b ) { return b && b->isBlocked(); } ) )
231
38.1M
        {
232
          // reschedule
233
38.1M
          t.state.store( WAITING );
234
38.1M
          continue;
235
38.1M
        }
236
1.20k
        t.barriers.clear();   // clear barriers, so we don't need to check them on the next try (we assume they won't get locked again)
237
1.20k
      }
238
90.3M
      if( t.readyCheck && t.readyCheck( threadId, t.param ) == false )
239
90.3M
      {
240
        // reschedule
241
90.3M
        t.state.store( WAITING );
242
90.3M
        continue;
243
90.3M
      }
244
245
38.6k
      return it;
246
90.3M
    }
247
11.0G
  }
248
86.7M
  return {};
249
86.7M
}
250
251
bool NoMallocThreadPool::processTask( int threadId, NoMallocThreadPool::Slot& task )
252
26.5k
{
253
26.5k
  const bool success = task.func( threadId, task.param );
254
#if ENABLE_VALGRIND_CODE
255
  MutexLock lock( m_extraMutex );
256
#endif
257
26.5k
  if( !success )
258
21.5k
  {
259
21.5k
    task.state = WAITING;
260
21.5k
    return false;
261
21.5k
  }
262
263
4.94k
  if( task.done != nullptr )
264
0
  {
265
0
    task.done->unlock();
266
0
  }
267
4.94k
  if( task.counter != nullptr )
268
3.76k
  {
269
3.76k
    --(*task.counter);
270
3.76k
  }
271
272
4.94k
  task.state = FREE;
273
274
4.94k
  return true;
275
26.5k
}
276
277
#ifdef HAVE_PTHREADS
278
279
template<class TFunc, class... TArgs>
280
NoMallocThreadPool::PThread::PThread( TFunc&& func, TArgs&&... args )
281
4.82k
{
282
4.82k
  using WrappedCall     = std::function<void()>;
283
4.82k
  std::unique_ptr<WrappedCall> call = std::make_unique<WrappedCall>( std::bind( func, args... ) );
284
285
4.82k
  using PThreadsStartFn = void* (*) ( void* );
286
4.82k
  PThreadsStartFn threadFn = []( void* p ) -> void*
287
4.82k
  {
288
4.81k
    std::unique_ptr<WrappedCall> call( static_cast<WrappedCall*>( p ) );
289
290
4.81k
    ( *call )();
291
292
4.81k
    return nullptr;
293
4.81k
  };
294
295
4.82k
  pthread_attr_t attr;
296
4.82k
  int ret = pthread_attr_init( &attr );
297
4.82k
  CHECK( ret != 0, "pthread_attr_init() failed" );
298
299
4.82k
  try
300
4.82k
  {
301
4.82k
    size_t currStackSize = 0;
302
4.82k
    ret = pthread_attr_getstacksize( &attr, &currStackSize );
303
4.82k
    CHECK( ret != 0, "pthread_attr_getstacksize() failed" );
304
305
4.82k
    if( currStackSize < THREAD_MIN_STACK_SIZE )
306
0
    {
307
0
      ret = pthread_attr_setstacksize( &attr, THREAD_MIN_STACK_SIZE );
308
0
      CHECK( ret != 0, "pthread_attr_setstacksize() failed" );
309
310
0
#  if defined( _DEBUG ) && !defined( __MINGW32__ ) && !defined( __MINGW64__ )
311
0
      ret = pthread_attr_setguardsize( &attr, 1024 * 1024 );   // set stack guard size to 1MB to more reliably deteck stack overflows
312
0
      CHECK( ret != 0, "pthread_attr_setguardsize() failed" );
313
0
#  endif
314
0
    }
315
4.82k
    m_joinable = 0 == pthread_create( &m_id, &attr, threadFn, call.get() );
316
4.82k
    CHECK( !m_joinable, "pthread_create() faild" );
317
318
4.82k
    call.release();   // will now be freed by the thread
319
320
4.82k
    pthread_attr_destroy( &attr );
321
4.82k
  }
322
4.82k
  catch( ... )
323
4.82k
  {
324
0
    pthread_attr_destroy( &attr );
325
0
    throw;
326
0
  }
327
4.82k
}
328
329
NoMallocThreadPool::PThread& NoMallocThreadPool::PThread::operator=( PThread&& other )
330
3.61k
{
331
3.61k
  m_id             = other.m_id;
332
3.61k
  m_joinable       = other.m_joinable;
333
3.61k
  other.m_id       = 0;
334
3.61k
  other.m_joinable = false;
335
3.61k
  return *this;
336
3.61k
}
337
338
void NoMallocThreadPool::PThread::join()
339
4.82k
{
340
4.82k
  if( m_joinable )
341
4.82k
  {
342
4.82k
    m_joinable = false;
343
4.82k
    pthread_join( m_id, nullptr );
344
4.82k
  }
345
4.82k
}
346
347
#endif   // HAVE_PTHREADS
348
349
} // namespace vvenc
350
351
//! \}
352