| 1 | /* |
| 2 | Copyright (c) 2005-2019 Intel Corporation |
| 3 | |
| 4 | Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | you may not use this file except in compliance with the License. |
| 6 | You may obtain a copy of the License at |
| 7 | |
| 8 | http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | |
| 10 | Unless required by applicable law or agreed to in writing, software |
| 11 | distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | See the License for the specific language governing permissions and |
| 14 | limitations under the License. |
| 15 | */ |
| 16 | |
| 17 | #define HARNESS_DEFAULT_MIN_THREADS 6 |
| 18 | #define HARNESS_DEFAULT_MAX_THREADS 8 |
| 19 | |
| 20 | #include "tbb/concurrent_monitor.h" |
| 21 | #include "tbb/atomic.h" |
| 22 | #include "tbb/task_scheduler_init.h" |
| 23 | #include "tbb/parallel_for.h" |
| 24 | #include "tbb/blocked_range.h" |
| 25 | #include "harness.h" |
| 26 | #if _WIN32||_WIN64 |
| 27 | #include "tbb/dynamic_link.cpp" |
| 28 | #endif |
| 29 | |
| 30 | #include "tbb/semaphore.cpp" |
| 31 | #include "tbb/concurrent_monitor.cpp" |
| 32 | |
| 33 | #if _MSC_VER && !defined(__INTEL_COMPILER) |
| 34 | // Workaround for overzealous compiler warnings |
| 35 | // Suppress compiler warning about constant conditional expression |
| 36 | #pragma warning (disable: 4127) |
| 37 | #endif |
| 38 | |
| 39 | using namespace tbb; |
| 40 | |
| 41 | //! Queuing lock with concurrent_monitor; to test concurrent_monitor::notify( Predicate p ) |
| 42 | class QueuingMutex { |
| 43 | public: |
| 44 | //! Construct unacquired mutex. |
| 45 | QueuingMutex() { q_tail = NULL; } |
| 46 | |
| 47 | //! The scoped locking pattern |
| 48 | class ScopedLock: internal::no_copy { |
| 49 | void Initialize() { mutex = NULL; } |
| 50 | public: |
| 51 | ScopedLock() {Initialize();} |
| 52 | ScopedLock( QueuingMutex& m, size_t test_mode ) { Initialize(); Acquire(m,test_mode); } |
| 53 | ~ScopedLock() { if( mutex ) Release(); } |
| 54 | void Acquire( QueuingMutex& m, size_t test_mode ); |
| 55 | void Release(); |
| 56 | void SleepPerhaps(); |
| 57 | |
| 58 | private: |
| 59 | QueuingMutex* mutex; |
| 60 | ScopedLock* next; |
| 61 | uintptr_t going; |
| 62 | internal::concurrent_monitor::thread_context thr_ctx; |
| 63 | }; |
| 64 | |
| 65 | friend class ScopedLock; |
| 66 | private: |
| 67 | //! The last competitor requesting the lock |
| 68 | atomic<ScopedLock*> q_tail; |
| 69 | internal::concurrent_monitor waitq; |
| 70 | }; |
| 71 | |
| 72 | struct PredicateEq { |
| 73 | uintptr_t p; |
| 74 | PredicateEq( uintptr_t p_ ) : p(p_) {} |
| 75 | bool operator() ( uintptr_t v ) const {return p==v;} |
| 76 | }; |
| 77 | |
| 78 | struct QueuingMutex_Context { |
| 79 | const QueuingMutex::ScopedLock* lck; |
| 80 | QueuingMutex_Context( QueuingMutex::ScopedLock* l_ ) : lck(l_) {} |
| 81 | uintptr_t operator()() { return uintptr_t(lck); } |
| 82 | }; |
| 83 | |
| 84 | struct QueuingMutex_Until : NoAssign { |
| 85 | uintptr_t& flag; |
| 86 | QueuingMutex_Until( uintptr_t& f_ ) : flag(f_) {} |
| 87 | bool operator()() { return flag!=0ul; } |
| 88 | }; |
| 89 | |
| 90 | //! A method to acquire QueuingMutex lock |
| 91 | void QueuingMutex::ScopedLock::Acquire( QueuingMutex& m, size_t test_mode ) |
| 92 | { |
| 93 | // Must set all fields before the fetch_and_store, because once the |
| 94 | // fetch_and_store executes, *this becomes accessible to other threads. |
| 95 | mutex = &m; |
| 96 | next = NULL; |
| 97 | going = 0; |
| 98 | |
| 99 | // The fetch_and_store must have release semantics, because we are |
| 100 | // "sending" the fields initialized above to other processors. |
| 101 | ScopedLock* pred = m.q_tail.fetch_and_store<tbb::release>(this); |
| 102 | if( pred ) { |
| 103 | #if TBB_USE_ASSERT |
| 104 | __TBB_control_consistency_helper(); // on "m.q_tail" |
| 105 | ASSERT( !pred->next, "the predecessor has another successor!" ); |
| 106 | #endif |
| 107 | pred->next = this; |
| 108 | for( int i=0; i<16; ++i ) { |
| 109 | if( going!=0ul ) break; |
| 110 | __TBB_Yield(); |
| 111 | } |
| 112 | int x = int( test_mode%3 ); |
| 113 | switch( x ) { |
| 114 | case 0: |
| 115 | mutex->waitq.wait( QueuingMutex_Until(going), QueuingMutex_Context(this) ); |
| 116 | break; |
| 117 | #if __TBB_CPP11_LAMBDAS_PRESENT |
| 118 | case 1: |
| 119 | mutex->waitq.wait( [&](){ return going!=0ul; }, [=]() { return (uintptr_t)this; } ); |
| 120 | break; |
| 121 | #endif |
| 122 | default: |
| 123 | SleepPerhaps(); |
| 124 | break; |
| 125 | } |
| 126 | } |
| 127 | |
| 128 | // Acquire critical section indirectly from previous owner or directly from predecessor. |
| 129 | __TBB_control_consistency_helper(); // on either "m.q_tail" or "going" |
| 130 | } |
| 131 | |
| 132 | //! A method to release QueuingMutex lock |
| 133 | void QueuingMutex::ScopedLock::Release( ) |
| 134 | { |
| 135 | if( !next ) { |
| 136 | if( this == mutex->q_tail.compare_and_swap<tbb::release>(NULL, this) ) { |
| 137 | // this was the only item in the queue, and the queue is now empty. |
| 138 | goto done; |
| 139 | } |
| 140 | // Someone in the queue |
| 141 | spin_wait_while_eq( next, (ScopedLock*)0 ); |
| 142 | } |
| 143 | __TBB_store_with_release(next->going, 1); |
| 144 | mutex->waitq.notify( PredicateEq(uintptr_t(next)) ); |
| 145 | done: |
| 146 | Initialize(); |
| 147 | } |
| 148 | |
| 149 | //! Yield and block; go to sleep |
| 150 | void QueuingMutex::ScopedLock::SleepPerhaps() |
| 151 | { |
| 152 | bool slept = false; |
| 153 | internal::concurrent_monitor& mq = mutex->waitq; |
| 154 | mq.prepare_wait( thr_ctx, uintptr_t(this) ); |
| 155 | while( going==0ul ) { |
| 156 | if( (slept=mq.commit_wait( thr_ctx ))==true && going!=0ul ) |
| 157 | break; |
| 158 | slept = false; |
| 159 | mq.prepare_wait( thr_ctx, uintptr_t(this) ); |
| 160 | } |
| 161 | if( !slept ) |
| 162 | mq.cancel_wait( thr_ctx ); |
| 163 | } |
| 164 | |
| 165 | // Spin lock with concurrent_monitor; to test concurrent_monitor::notify_all() and concurrent_monitor::notify() |
| 166 | class SpinMutex { |
| 167 | public: |
| 168 | //! Construct unacquired mutex. |
| 169 | SpinMutex() : toggle(false) { flag = 0; } |
| 170 | |
| 171 | //! The scoped locking pattern |
| 172 | class ScopedLock: internal::no_copy { |
| 173 | void Initialize() { mutex = NULL; } |
| 174 | public: |
| 175 | ScopedLock() {Initialize();} |
| 176 | ScopedLock( SpinMutex& m, size_t test_mode ) { Initialize(); Acquire(m,test_mode); } |
| 177 | ~ScopedLock() { if( mutex ) Release(); } |
| 178 | void Acquire( SpinMutex& m, size_t test_mode ); |
| 179 | void Release(); |
| 180 | void SleepPerhaps(); |
| 181 | |
| 182 | private: |
| 183 | SpinMutex* mutex; |
| 184 | internal::concurrent_monitor::thread_context thr_ctx; |
| 185 | }; |
| 186 | |
| 187 | friend class ScopedLock; |
| 188 | friend struct SpinMutex_Until; |
| 189 | private: |
| 190 | tbb::atomic<unsigned> flag; |
| 191 | bool toggle; |
| 192 | internal::concurrent_monitor waitq; |
| 193 | }; |
| 194 | |
| 195 | struct SpinMutex_Context { |
| 196 | const SpinMutex::ScopedLock* lck; |
| 197 | SpinMutex_Context( SpinMutex::ScopedLock* l_ ) : lck(l_) {} |
| 198 | uintptr_t operator()() { return uintptr_t(lck); } |
| 199 | }; |
| 200 | |
| 201 | struct SpinMutex_Until { |
| 202 | const SpinMutex* mtx; |
| 203 | SpinMutex_Until( SpinMutex* m_ ) : mtx(m_) {} |
| 204 | bool operator()() { return mtx->flag==0; } |
| 205 | }; |
| 206 | |
| 207 | //! A method to acquire SpinMutex lock |
| 208 | void SpinMutex::ScopedLock::Acquire( SpinMutex& m, size_t test_mode ) |
| 209 | { |
| 210 | mutex = &m; |
| 211 | retry: |
| 212 | if( m.flag.compare_and_swap( 1, 0 )!=0 ) { |
| 213 | int x = int( test_mode%3 ); |
| 214 | switch( x ) { |
| 215 | case 0: |
| 216 | mutex->waitq.wait( SpinMutex_Until(mutex), SpinMutex_Context(this) ); |
| 217 | break; |
| 218 | #if __TBB_CPP11_LAMBDAS_PRESENT |
| 219 | case 1: |
| 220 | mutex->waitq.wait( [&](){ return mutex->flag==0; }, [=]() { return (uintptr_t)this; } ); |
| 221 | break; |
| 222 | #endif |
| 223 | default: |
| 224 | SleepPerhaps(); |
| 225 | break; |
| 226 | } |
| 227 | goto retry; |
| 228 | } |
| 229 | } |
| 230 | |
| 231 | //! A method to release SpinMutex lock |
| 232 | void SpinMutex::ScopedLock::Release() |
| 233 | { |
| 234 | bool old_toggle = mutex->toggle; |
| 235 | mutex->toggle = !mutex->toggle; |
| 236 | mutex->flag = 0; |
| 237 | if( old_toggle ) |
| 238 | mutex->waitq.notify_one(); |
| 239 | else |
| 240 | mutex->waitq.notify_all(); |
| 241 | } |
| 242 | |
| 243 | //! Yield and block; go to sleep |
| 244 | void SpinMutex::ScopedLock::SleepPerhaps() |
| 245 | { |
| 246 | bool slept = false; |
| 247 | internal::concurrent_monitor& mq = mutex->waitq; |
| 248 | mq.prepare_wait( thr_ctx, uintptr_t(this) ); |
| 249 | while( mutex->flag ) { |
| 250 | if( (slept=mq.commit_wait( thr_ctx ))==true ) |
| 251 | break; |
| 252 | mq.prepare_wait( thr_ctx, uintptr_t(this) ); |
| 253 | } |
| 254 | if( !slept ) |
| 255 | mq.cancel_wait( thr_ctx ); |
| 256 | } |
| 257 | |
| 258 | //! A value protected by a mutex. |
| 259 | template<typename M> |
| 260 | struct Counter { |
| 261 | typedef M mutex_type; |
| 262 | M mutex; |
| 263 | long value; |
| 264 | }; |
| 265 | |
| 266 | //! Function object for use with parallel_for.h. |
| 267 | template<typename C, int D> |
| 268 | struct AddOne: NoAssign { |
| 269 | C& counter; |
| 270 | /** Increments counter once for each iteration in the iteration space. */ |
| 271 | void operator()( tbb::blocked_range<size_t>& range ) const { |
| 272 | for( size_t i=range.begin(); i!=range.end(); ++i ) { |
| 273 | typename C::mutex_type::ScopedLock lock(counter.mutex, i); |
| 274 | counter.value = counter.value+1; |
| 275 | if( D>0 ) |
| 276 | for( int j=0; j<D; ++j ) __TBB_Yield(); |
| 277 | } |
| 278 | } |
| 279 | AddOne( C& counter_ ) : counter(counter_) {} |
| 280 | }; |
| 281 | |
| 282 | //! Generic test with TBB mutex type M, max range R, and delay D. |
| 283 | template<typename M,int R, int D> |
| 284 | void Test( int p ) { |
| 285 | Counter<M> counter; |
| 286 | counter.value = 0; |
| 287 | const int n = R; |
| 288 | tbb::task_scheduler_init init(p); |
| 289 | tbb::parallel_for(tbb::blocked_range<size_t>(0,n,n/10),AddOne<Counter<M>,D>(counter)); |
| 290 | if( counter.value!=n ) |
| 291 | REPORT("ERROR : counter.value=%ld (instead of %ld)\n" ,counter.value,n); |
| 292 | } |
| 293 | |
| 294 | #if TBB_USE_EXCEPTIONS |
| 295 | #define NTHRS_USED_IN_DESTRUCTOR_TEST 8 |
| 296 | |
| 297 | atomic<size_t> n_sleepers; |
| 298 | |
| 299 | #if defined(_MSC_VER) && defined(_Wp64) |
| 300 | // Workaround for overzealous compiler warnings in /Wp64 mode |
| 301 | #pragma warning (disable: 4244 4267) |
| 302 | #endif |
| 303 | |
| 304 | struct AllButOneSleep : NoAssign { |
| 305 | internal::concurrent_monitor*& mon; |
| 306 | static const size_t VLN = 1024*1024; |
| 307 | void operator()( int i ) const { |
| 308 | internal::concurrent_monitor::thread_context thr_ctx; |
| 309 | |
| 310 | if( i==0 ) { |
| 311 | size_t n_expected_sleepers = NTHRS_USED_IN_DESTRUCTOR_TEST-1; |
| 312 | while( n_sleepers<n_expected_sleepers ) |
| 313 | __TBB_Yield(); |
| 314 | while( n_sleepers.compare_and_swap( VLN+NTHRS_USED_IN_DESTRUCTOR_TEST, n_expected_sleepers )!=n_expected_sleepers ) |
| 315 | __TBB_Yield(); |
| 316 | |
| 317 | for( int j=0; j<100; ++j ) |
| 318 | Harness::Sleep( 1 ); |
| 319 | delete mon; |
| 320 | mon = NULL; |
| 321 | } else { |
| 322 | mon->prepare_wait( thr_ctx, uintptr_t(this) ); |
| 323 | while( n_sleepers<VLN ) { |
| 324 | try { |
| 325 | ++n_sleepers; |
| 326 | mon->commit_wait( thr_ctx ); |
| 327 | if( --n_sleepers>VLN ) |
| 328 | break; |
| 329 | } catch( tbb::user_abort& ) { |
| 330 | // can no longer access 'mon' |
| 331 | break; |
| 332 | } |
| 333 | mon->prepare_wait( thr_ctx, uintptr_t(this) ); |
| 334 | } |
| 335 | } |
| 336 | } |
| 337 | AllButOneSleep( internal::concurrent_monitor*& m_ ) : mon(m_) {} |
| 338 | }; |
| 339 | #endif /* TBB_USE_EXCEPTIONS */ |
| 340 | |
| 341 | void TestDestructor() { |
| 342 | #if TBB_USE_EXCEPTIONS |
| 343 | tbb::task_scheduler_init init(NTHRS_USED_IN_DESTRUCTOR_TEST); |
| 344 | internal::concurrent_monitor* my_mon = new internal::concurrent_monitor; |
| 345 | REMARK( "testing the destructor\n" ); |
| 346 | n_sleepers = 0; |
| 347 | NativeParallelFor(NTHRS_USED_IN_DESTRUCTOR_TEST,AllButOneSleep(my_mon)); |
| 348 | ASSERT( my_mon==NULL, "" ); |
| 349 | #endif /* TBB_USE_EXCEPTIONS */ |
| 350 | } |
| 351 | |
| 352 | int TestMain () { |
| 353 | for( int p=MinThread; p<=MaxThread; ++p ) { |
| 354 | REMARK( "testing with %d workers\n" , static_cast<int>(p) ); |
| 355 | // test the predicated notify |
| 356 | Test<QueuingMutex,100000,0>( p ); |
| 357 | Test<QueuingMutex,1000,10000>( p ); |
| 358 | // test the notify_all method |
| 359 | Test<SpinMutex,100000,0>( p ); |
| 360 | Test<SpinMutex,1000,10000>( p ); |
| 361 | REMARK( "calling destructor for task_scheduler_init\n" ); |
| 362 | } |
| 363 | TestDestructor(); |
| 364 | return Harness::Done; |
| 365 | } |
| 366 | |