| 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 2 |
| 18 | #define HARNESS_DEFAULT_MAX_THREADS 4 |
| 19 | |
| 20 | #include "harness.h" |
| 21 | |
| 22 | #if __TBB_TASK_GROUP_CONTEXT |
| 23 | |
| 24 | #include <limits.h> // for INT_MAX |
| 25 | #include "tbb/task_scheduler_init.h" |
| 26 | #include "tbb/tbb_exception.h" |
| 27 | #include "tbb/task.h" |
| 28 | #include "tbb/atomic.h" |
| 29 | #include "tbb/parallel_for.h" |
| 30 | #include "tbb/parallel_reduce.h" |
| 31 | #include "tbb/parallel_do.h" |
| 32 | #include "tbb/pipeline.h" |
| 33 | #include "tbb/parallel_scan.h" |
| 34 | #include "tbb/blocked_range.h" |
| 35 | #include "harness_assert.h" |
| 36 | |
| 37 | #define FLAT_RANGE 100000 |
| 38 | #define FLAT_GRAIN 100 |
| 39 | #define OUTER_RANGE 100 |
| 40 | #define OUTER_GRAIN 10 |
| 41 | #define INNER_RANGE (FLAT_RANGE / OUTER_RANGE) |
| 42 | #define INNER_GRAIN (FLAT_GRAIN / OUTER_GRAIN) |
| 43 | |
| 44 | tbb::atomic<intptr_t> g_FedTasksCount; // number of tasks added by parallel_do feeder |
| 45 | tbb::atomic<intptr_t> g_OuterParCalls; // number of actual invocations of the outer construct executed. |
| 46 | tbb::atomic<intptr_t> g_TGCCancelled; // Number of times a task sees its group cancelled at start |
| 47 | |
| 48 | inline intptr_t Existed () { return INT_MAX; } |
| 49 | |
| 50 | #include "harness_eh.h" |
| 51 | /******************************** |
| 52 | Variables in test |
| 53 | |
| 54 | __ Test control variables |
| 55 | g_ExceptionInMaster -- only the master thread is allowed to throw. If false, the master cannot throw |
| 56 | g_SolitaryException -- only one throw may be executed. |
| 57 | |
| 58 | -- controls for ThrowTestException for pipeline tests |
| 59 | g_NestedPipelines -- are inner pipelines being run? |
| 60 | g_PipelinesStarted -- how many pipelines have run their first filter at least once. |
| 61 | |
| 62 | -- Information variables |
| 63 | |
| 64 | g_Master -- Thread ID of the "master" thread |
| 65 | In pipelines sometimes the master thread does not participate, so the tests have to be resilient to this. |
| 66 | |
| 67 | -- Measurement variables |
| 68 | |
| 69 | g_OuterParCalls -- how many outer parallel ranges or filters started |
| 70 | g_TGCCancelled -- how many inner parallel ranges or filters saw task::self().is_cancelled() |
| 71 | g_ExceptionsThrown -- number of throws executed (counted in ThrowTestException) |
| 72 | g_MasterExecutedThrow -- number of times master thread actually executed a throw |
| 73 | g_NonMasterExecutedThrow -- number of times non-master thread actually executed a throw |
| 74 | g_ExceptionCaught -- one of PropagatedException or unknown exception was caught. (Other exceptions cause assertions.) |
| 75 | |
| 76 | -- Tallies for the task bodies which have executed (counted in each inner body, sampled in ThrowTestException) |
| 77 | g_CurExecuted -- total number of inner ranges or filters which executed |
| 78 | g_ExecutedAtLastCatch -- value of g_CurExecuted when last catch was made, 0 if none. |
| 79 | g_ExecutedAtFirstCatch -- value of g_CurExecuted when first catch is made, 0 if none. |
| 80 | *********************************/ |
| 81 | |
| 82 | inline void ResetGlobals ( bool throwException = true, bool flog = false ) { |
| 83 | ResetEhGlobals( throwException, flog ); |
| 84 | g_FedTasksCount = 0; |
| 85 | g_OuterParCalls = 0; |
| 86 | g_NestedPipelines = false; |
| 87 | g_TGCCancelled = 0; |
| 88 | } |
| 89 | |
| 90 | //////////////////////////////////////////////////////////////////////////////// |
| 91 | // Tests for tbb::parallel_for and tbb::parallel_reduce |
| 92 | |
| 93 | typedef size_t count_type; |
| 94 | typedef tbb::blocked_range<count_type> range_type; |
| 95 | |
| 96 | inline intptr_t CountSubranges(range_type r) { |
| 97 | if(!r.is_divisible()) return intptr_t(1); |
| 98 | range_type r2(r,tbb::split()); |
| 99 | return CountSubranges(r) + CountSubranges(r2); |
| 100 | } |
| 101 | |
| 102 | inline intptr_t NumSubranges ( intptr_t length, intptr_t grain ) { |
| 103 | return CountSubranges(range_type(0,length,grain)); |
| 104 | } |
| 105 | |
| 106 | template<class Body> |
| 107 | intptr_t TestNumSubrangesCalculation ( intptr_t length, intptr_t grain, intptr_t inner_length, intptr_t inner_grain ) { |
| 108 | ResetGlobals(); |
| 109 | g_ThrowException = false; |
| 110 | intptr_t outerCalls = NumSubranges(length, grain), |
| 111 | innerCalls = NumSubranges(inner_length, inner_grain), |
| 112 | maxExecuted = outerCalls * (innerCalls + 1); |
| 113 | tbb::parallel_for( range_type(0, length, grain), Body() ); |
| 114 | ASSERT (g_CurExecuted == maxExecuted, "Wrong estimation of bodies invocation count" ); |
| 115 | return maxExecuted; |
| 116 | } |
| 117 | |
| 118 | class NoThrowParForBody { |
| 119 | public: |
| 120 | void operator()( const range_type& r ) const { |
| 121 | volatile count_type x = 0; |
| 122 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 123 | else g_NonMasterExecuted = true; |
| 124 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 125 | count_type end = r.end(); |
| 126 | for( count_type i=r.begin(); i<end; ++i ) |
| 127 | x += i; |
| 128 | } |
| 129 | }; |
| 130 | |
| 131 | #if TBB_USE_EXCEPTIONS |
| 132 | |
| 133 | void Test0 () { |
| 134 | ResetGlobals(); |
| 135 | tbb::simple_partitioner p; |
| 136 | for( size_t i=0; i<10; ++i ) { |
| 137 | tbb::parallel_for( range_type(0, 0, 1), NoThrowParForBody() ); |
| 138 | tbb::parallel_for( range_type(0, 0, 1), NoThrowParForBody(), p ); |
| 139 | tbb::parallel_for( range_type(0, 128, 8), NoThrowParForBody() ); |
| 140 | tbb::parallel_for( range_type(0, 128, 8), NoThrowParForBody(), p ); |
| 141 | } |
| 142 | } // void Test0 () |
| 143 | |
| 144 | //! Template that creates a functor suitable for parallel_reduce from a functor for parallel_for. |
| 145 | template<typename ParForBody> |
| 146 | class SimpleParReduceBody: NoAssign { |
| 147 | ParForBody m_Body; |
| 148 | public: |
| 149 | void operator()( const range_type& r ) const { m_Body(r); } |
| 150 | SimpleParReduceBody() {} |
| 151 | SimpleParReduceBody( SimpleParReduceBody& left, tbb::split ) : m_Body(left.m_Body) {} |
| 152 | void join( SimpleParReduceBody& /*right*/ ) {} |
| 153 | }; // SimpleParReduceBody |
| 154 | |
| 155 | //! Test parallel_for and parallel_reduce for a given partitioner. |
| 156 | /** The Body need only be suitable for a parallel_for. */ |
| 157 | template<typename ParForBody, typename Partitioner> |
| 158 | void TestParallelLoopAux() { |
| 159 | Partitioner partitioner; |
| 160 | for( int i=0; i<2; ++i ) { |
| 161 | ResetGlobals(); |
| 162 | TRY(); |
| 163 | if( i==0 ) |
| 164 | tbb::parallel_for( range_type(0, FLAT_RANGE, FLAT_GRAIN), ParForBody(), partitioner ); |
| 165 | else { |
| 166 | SimpleParReduceBody<ParForBody> rb; |
| 167 | tbb::parallel_reduce( range_type(0, FLAT_RANGE, FLAT_GRAIN), rb, partitioner ); |
| 168 | } |
| 169 | CATCH_AND_ASSERT(); |
| 170 | // two cases: g_SolitaryException and !g_SolitaryException |
| 171 | // 1) g_SolitaryException: only one thread actually threw. There is only one context, so the exception |
| 172 | // (when caught) will cause that context to be cancelled. After this event, there may be one or |
| 173 | // more threads which are "in-flight", up to g_NumThreads, but no more will be started. The threads, |
| 174 | // when they start, if they see they are cancelled, TGCCancelled is incremented. |
| 175 | // 2) !g_SolitaryException: more than one thread can throw. The number of threads that actually |
| 176 | // threw is g_MasterExecutedThrow if only the master is allowed, else g_NonMasterExecutedThrow. |
| 177 | // Only one context, so TGCCancelled should be <= g_NumThreads. |
| 178 | // |
| 179 | // the reasoning is similar for nested algorithms in a single context (Test2). |
| 180 | // |
| 181 | // If a thread throws in a context, more than one subsequent task body may see the |
| 182 | // cancelled state (if they are scheduled before the state is propagated.) this is |
| 183 | // infrequent, but it occurs. So what was to be an assertion must be a remark. |
| 184 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks ran after exception thrown" ); |
| 185 | if( g_TGCCancelled > g_NumThreads) REMARK( "Too many tasks ran after exception thrown (%d vs. %d)\n" , |
| 186 | (int)g_TGCCancelled, (int)g_NumThreads); |
| 187 | ASSERT(g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 188 | if ( g_SolitaryException ) { |
| 189 | ASSERT(g_NumExceptionsCaught == 1, "No try_blocks in any body expected in this test" ); |
| 190 | ASSERT(g_NumExceptionsCaught == (g_ExceptionInMaster ? g_MasterExecutedThrow : g_NonMasterExecutedThrow), |
| 191 | "Not all throws were caught" ); |
| 192 | ASSERT(g_ExecutedAtFirstCatch == g_ExecutedAtLastCatch, "Too many exceptions occurred" ); |
| 193 | } |
| 194 | else { |
| 195 | ASSERT(g_NumExceptionsCaught >= 1, "No try blocks in any body expected in this test" ); |
| 196 | } |
| 197 | } |
| 198 | } // TestParallelLoopAux |
| 199 | |
| 200 | //! Test with parallel_for and parallel_reduce, over all three kinds of partitioners. |
| 201 | /** The Body only needs to be suitable for tbb::parallel_for. */ |
| 202 | template<typename Body> |
| 203 | void TestParallelLoop() { |
| 204 | // The simple and auto partitioners should be const, but not the affinity partitioner. |
| 205 | TestParallelLoopAux<Body, const tbb::simple_partitioner >(); |
| 206 | TestParallelLoopAux<Body, const tbb::auto_partitioner >(); |
| 207 | #define __TBB_TEMPORARILY_DISABLED 1 |
| 208 | #if !__TBB_TEMPORARILY_DISABLED |
| 209 | // TODO: Improve the test so that it tolerates delayed start of tasks with affinity_partitioner |
| 210 | TestParallelLoopAux<Body, /***/ tbb::affinity_partitioner>(); |
| 211 | #endif |
| 212 | #undef __TBB_TEMPORARILY_DISABLED |
| 213 | } |
| 214 | |
| 215 | class SimpleParForBody: NoAssign { |
| 216 | public: |
| 217 | void operator()( const range_type& r ) const { |
| 218 | Harness::ConcurrencyTracker ct; |
| 219 | volatile long x = 0; |
| 220 | ++g_CurExecuted; |
| 221 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 222 | else g_NonMasterExecuted = true; |
| 223 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 224 | for( count_type i = r.begin(); i != r.end(); ++i ) |
| 225 | x += 0; |
| 226 | WaitUntilConcurrencyPeaks(); |
| 227 | ThrowTestException(1); |
| 228 | } |
| 229 | }; |
| 230 | |
| 231 | void Test1() { |
| 232 | // non-nested parallel_for/reduce with throwing body, one context |
| 233 | TestParallelLoop<SimpleParForBody>(); |
| 234 | } // void Test1 () |
| 235 | |
| 236 | class OuterParForBody: NoAssign { |
| 237 | public: |
| 238 | void operator()( const range_type& ) const { |
| 239 | Harness::ConcurrencyTracker ct; |
| 240 | ++g_OuterParCalls; |
| 241 | tbb::parallel_for( tbb::blocked_range<size_t>(0, INNER_RANGE, INNER_GRAIN), SimpleParForBody() ); |
| 242 | } |
| 243 | }; |
| 244 | |
| 245 | //! Uses parallel_for body containing an inner parallel_for with the default context not wrapped by a try-block. |
| 246 | /** Inner algorithms are spawned inside the new bound context by default. Since |
| 247 | exceptions thrown from the inner parallel_for are not handled by the caller |
| 248 | (outer parallel_for body) in this test, they will cancel all the sibling inner |
| 249 | algorithms. **/ |
| 250 | void Test2 () { |
| 251 | TestParallelLoop<OuterParForBody>(); |
| 252 | } // void Test2 () |
| 253 | |
| 254 | class OuterParForBodyWithIsolatedCtx { |
| 255 | public: |
| 256 | void operator()( const range_type& ) const { |
| 257 | tbb::task_group_context ctx(tbb::task_group_context::isolated); |
| 258 | ++g_OuterParCalls; |
| 259 | tbb::parallel_for( tbb::blocked_range<size_t>(0, INNER_RANGE, INNER_GRAIN), SimpleParForBody(), tbb::simple_partitioner(), ctx ); |
| 260 | } |
| 261 | }; |
| 262 | |
| 263 | //! Uses parallel_for body invoking an inner parallel_for with an isolated context without a try-block. |
| 264 | /** Even though exceptions thrown from the inner parallel_for are not handled |
| 265 | by the caller in this test, they will not affect sibling inner algorithms |
| 266 | already running because of the isolated contexts. However because the first |
| 267 | exception cancels the root parallel_for only the first g_NumThreads subranges |
| 268 | will be processed (which launch inner parallel_fors) **/ |
| 269 | void Test3 () { |
| 270 | ResetGlobals(); |
| 271 | typedef OuterParForBodyWithIsolatedCtx body_type; |
| 272 | intptr_t innerCalls = NumSubranges(INNER_RANGE, INNER_GRAIN), |
| 273 | // we expect one thread to throw without counting, the rest to run to completion |
| 274 | // this formula assumes g_numThreads outer pfor ranges will be started, but that is not the |
| 275 | // case; the SimpleParFor subranges are started up as part of the outer ones, and when |
| 276 | // the amount of concurrency reaches g_NumThreads no more outer Pfor ranges are started. |
| 277 | // so we have to count the number of outer Pfors actually started. |
| 278 | minExecuted = (g_NumThreads - 1) * innerCalls; |
| 279 | TRY(); |
| 280 | tbb::parallel_for( range_type(0, OUTER_RANGE, OUTER_GRAIN), body_type() ); |
| 281 | CATCH_AND_ASSERT(); |
| 282 | minExecuted = (g_OuterParCalls - 1) * innerCalls; // see above |
| 283 | |
| 284 | // The first formula above assumes all ranges of the outer parallel for are executed, and one |
| 285 | // cancels. In the event, we have a smaller number of ranges that start before the exception |
| 286 | // is caught. |
| 287 | // |
| 288 | // g_SolitaryException:One inner range throws. Outer parallel_For is cancelled, but sibling |
| 289 | // parallel_fors continue to completion (unless the threads that execute |
| 290 | // are not allowed to throw, in which case we will not see any exceptions). |
| 291 | // !g_SolitaryException:multiple inner ranges may throw. Any which throws will stop, and the |
| 292 | // corresponding range of the outer pfor will stop also. |
| 293 | // |
| 294 | // In either case, once the outer pfor gets the exception it will stop executing further ranges. |
| 295 | |
| 296 | // if the only threads executing were not allowed to throw, then not seeing an exception is okay. |
| 297 | bool okayNoExceptionsCaught = (g_ExceptionInMaster && !g_MasterExecuted) || (!g_ExceptionInMaster && !g_NonMasterExecuted); |
| 298 | if ( g_SolitaryException ) { |
| 299 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived exception" ); |
| 300 | ASSERT (g_CurExecuted > minExecuted, "Too few tasks survived exception" ); |
| 301 | ASSERT (g_CurExecuted <= minExecuted + (g_ExecutedAtLastCatch + g_NumThreads), "Too many tasks survived exception" ); |
| 302 | ASSERT (g_NumExceptionsCaught == 1 || okayNoExceptionsCaught, "No try_blocks in any body expected in this test" ); |
| 303 | } |
| 304 | else { |
| 305 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 306 | ASSERT (g_NumExceptionsCaught >= 1 || okayNoExceptionsCaught, "No try_blocks in any body expected in this test" ); |
| 307 | } |
| 308 | } // void Test3 () |
| 309 | |
| 310 | class OuterParForExceptionSafeBody { |
| 311 | public: |
| 312 | void operator()( const range_type& ) const { |
| 313 | tbb::task_group_context ctx(tbb::task_group_context::isolated); |
| 314 | ++g_OuterParCalls; |
| 315 | TRY(); |
| 316 | tbb::parallel_for( tbb::blocked_range<size_t>(0, INNER_RANGE, INNER_GRAIN), SimpleParForBody(), tbb::simple_partitioner(), ctx ); |
| 317 | CATCH(); // this macro sets g_ExceptionCaught |
| 318 | } |
| 319 | }; |
| 320 | |
| 321 | //! Uses parallel_for body invoking an inner parallel_for (with isolated context) inside a try-block. |
| 322 | /** Since exception(s) thrown from the inner parallel_for are handled by the caller |
| 323 | in this test, they do not affect neither other tasks of the the root parallel_for |
| 324 | nor sibling inner algorithms. **/ |
| 325 | void Test4 () { |
| 326 | ResetGlobals( true, true ); |
| 327 | intptr_t innerCalls = NumSubranges(INNER_RANGE, INNER_GRAIN), |
| 328 | outerCalls = NumSubranges(OUTER_RANGE, OUTER_GRAIN); |
| 329 | TRY(); |
| 330 | tbb::parallel_for( range_type(0, OUTER_RANGE, OUTER_GRAIN), OuterParForExceptionSafeBody() ); |
| 331 | CATCH(); |
| 332 | // g_SolitaryException : one inner pfor will throw, the rest will execute to completion. |
| 333 | // so the count should be (outerCalls -1) * innerCalls, if a throw happened. |
| 334 | // !g_SolitaryException : possible multiple inner pfor throws. Should be approximately |
| 335 | // (outerCalls - g_NumExceptionsCaught) * innerCalls, give or take a few |
| 336 | intptr_t minExecuted = (outerCalls - g_NumExceptionsCaught) * innerCalls; |
| 337 | bool okayNoExceptionsCaught = (g_ExceptionInMaster && !g_MasterExecuted) || (!g_ExceptionInMaster && !g_NonMasterExecuted); |
| 338 | if ( g_SolitaryException ) { |
| 339 | // only one task had exception thrown. That task had at least one execution (the one that threw). |
| 340 | // There may be an arbitrary number of ranges executed after the throw but before the exception |
| 341 | // is caught in the scheduler and cancellation is signaled. (seen 9, 11 and 62 (!) for 8 threads) |
| 342 | ASSERT (g_NumExceptionsCaught == 1 || okayNoExceptionsCaught, "No exception registered" ); |
| 343 | ASSERT (g_CurExecuted >= minExecuted, "Too few tasks executed" ); |
| 344 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived exception" ); |
| 345 | // a small number of threads can execute in a throwing sub-pfor, if the task which is |
| 346 | // to do the solitary throw swaps out after registering its intent to throw but before it |
| 347 | // actually does so. (Or is this caused by the extra threads participating? No, the |
| 348 | // number of extra tasks is sometimes far greater than the number of extra threads.) |
| 349 | ASSERT (g_CurExecuted <= minExecuted + g_NumThreads, "Too many tasks survived exception" ); |
| 350 | if(g_CurExecuted > minExecuted + g_NumThreads) REMARK("Unusual number of tasks executed after signal (%d vs. %d)\n" , |
| 351 | (int)g_CurExecuted, minExecuted + g_NumThreads); |
| 352 | } |
| 353 | else { |
| 354 | ASSERT ((g_NumExceptionsCaught >= 1 && g_NumExceptionsCaught <= outerCalls) || okayNoExceptionsCaught, "Unexpected actual number of exceptions" ); |
| 355 | ASSERT (g_CurExecuted >= minExecuted, "Too few executed tasks reported" ); |
| 356 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived multiple exceptions" ); |
| 357 | if(g_CurExecuted > g_ExecutedAtLastCatch + g_NumThreads) REMARK("Unusual number of tasks executed after signal (%d vs. %d)\n" , |
| 358 | (int)g_CurExecuted, g_ExecutedAtLastCatch + g_NumThreads); |
| 359 | ASSERT (g_CurExecuted <= outerCalls * (1 + g_NumThreads), "Too many tasks survived exception" ); |
| 360 | } |
| 361 | } // void Test4 () |
| 362 | |
| 363 | #endif /* TBB_USE_EXCEPTIONS */ |
| 364 | |
| 365 | class ParForBodyToCancel { |
| 366 | public: |
| 367 | void operator()( const range_type& ) const { |
| 368 | ++g_CurExecuted; |
| 369 | CancellatorTask::WaitUntilReady(); |
| 370 | } |
| 371 | }; |
| 372 | |
| 373 | template<class B> |
| 374 | class ParForLauncherTask : public tbb::task { |
| 375 | tbb::task_group_context &my_ctx; |
| 376 | |
| 377 | tbb::task* execute () __TBB_override { |
| 378 | tbb::parallel_for( range_type(0, FLAT_RANGE, FLAT_GRAIN), B(), tbb::simple_partitioner(), my_ctx ); |
| 379 | return NULL; |
| 380 | } |
| 381 | public: |
| 382 | ParForLauncherTask ( tbb::task_group_context& ctx ) : my_ctx(ctx) {} |
| 383 | }; |
| 384 | |
| 385 | //! Test for cancelling an algorithm from outside (from a task running in parallel with the algorithm). |
| 386 | void TestCancelation1 () { |
| 387 | ResetGlobals( false ); |
| 388 | RunCancellationTest<ParForLauncherTask<ParForBodyToCancel>, CancellatorTask>( NumSubranges(FLAT_RANGE, FLAT_GRAIN) / 4 ); |
| 389 | } |
| 390 | |
| 391 | class CancellatorTask2 : public tbb::task { |
| 392 | tbb::task_group_context &m_GroupToCancel; |
| 393 | |
| 394 | tbb::task* execute () __TBB_override { |
| 395 | Harness::ConcurrencyTracker ct; |
| 396 | WaitUntilConcurrencyPeaks(); |
| 397 | m_GroupToCancel.cancel_group_execution(); |
| 398 | g_ExecutedAtLastCatch = g_CurExecuted; |
| 399 | return NULL; |
| 400 | } |
| 401 | public: |
| 402 | CancellatorTask2 ( tbb::task_group_context& ctx, intptr_t ) : m_GroupToCancel(ctx) {} |
| 403 | }; |
| 404 | |
| 405 | class ParForBodyToCancel2 { |
| 406 | public: |
| 407 | void operator()( const range_type& ) const { |
| 408 | ++g_CurExecuted; |
| 409 | Harness::ConcurrencyTracker ct; |
| 410 | // The test will hang (and be timed out by the test system) if is_cancelled() is broken |
| 411 | while( !tbb::task::self().is_cancelled() ) |
| 412 | __TBB_Yield(); |
| 413 | } |
| 414 | }; |
| 415 | |
| 416 | //! Test for cancelling an algorithm from outside (from a task running in parallel with the algorithm). |
| 417 | /** This version also tests task::is_cancelled() method. **/ |
| 418 | void TestCancelation2 () { |
| 419 | ResetGlobals(); |
| 420 | RunCancellationTest<ParForLauncherTask<ParForBodyToCancel2>, CancellatorTask2>(); |
| 421 | ASSERT (g_ExecutedAtLastCatch < g_NumThreads, "Somehow worker tasks started their execution before the cancellator task" ); |
| 422 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived cancellation" ); |
| 423 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Some tasks were executed after cancellation" ); |
| 424 | } |
| 425 | |
| 426 | //////////////////////////////////////////////////////////////////////////////// |
| 427 | // Regression test based on the contribution by the author of the following forum post: |
| 428 | // http://softwarecommunity.intel.com/isn/Community/en-US/forums/thread/30254959.aspx |
| 429 | |
| 430 | class Worker { |
| 431 | static const int max_nesting = 3; |
| 432 | static const int reduce_range = 1024; |
| 433 | static const int reduce_grain = 256; |
| 434 | public: |
| 435 | int DoWork (int level); |
| 436 | int Validate (int start_level) { |
| 437 | int expected = 1; // identity for multiplication |
| 438 | for(int i=start_level+1; i<max_nesting; ++i) |
| 439 | expected *= reduce_range; |
| 440 | return expected; |
| 441 | } |
| 442 | }; |
| 443 | |
| 444 | class RecursiveParReduceBodyWithSharedWorker { |
| 445 | Worker * m_SharedWorker; |
| 446 | int m_NestingLevel; |
| 447 | int m_Result; |
| 448 | public: |
| 449 | RecursiveParReduceBodyWithSharedWorker ( RecursiveParReduceBodyWithSharedWorker& src, tbb::split ) |
| 450 | : m_SharedWorker(src.m_SharedWorker) |
| 451 | , m_NestingLevel(src.m_NestingLevel) |
| 452 | , m_Result(0) |
| 453 | {} |
| 454 | RecursiveParReduceBodyWithSharedWorker ( Worker *w, int outer ) |
| 455 | : m_SharedWorker(w) |
| 456 | , m_NestingLevel(outer) |
| 457 | , m_Result(0) |
| 458 | {} |
| 459 | |
| 460 | void operator() ( const tbb::blocked_range<size_t>& r ) { |
| 461 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 462 | else g_NonMasterExecuted = true; |
| 463 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 464 | for (size_t i = r.begin (); i != r.end (); ++i) { |
| 465 | m_Result += m_SharedWorker->DoWork (m_NestingLevel); |
| 466 | } |
| 467 | } |
| 468 | void join (const RecursiveParReduceBodyWithSharedWorker & x) { |
| 469 | m_Result += x.m_Result; |
| 470 | } |
| 471 | int result () { return m_Result; } |
| 472 | }; |
| 473 | |
| 474 | int Worker::DoWork ( int level ) { |
| 475 | ++level; |
| 476 | if ( level < max_nesting ) { |
| 477 | RecursiveParReduceBodyWithSharedWorker rt (this, level); |
| 478 | tbb::parallel_reduce (tbb::blocked_range<size_t>(0, reduce_range, reduce_grain), rt); |
| 479 | return rt.result(); |
| 480 | } |
| 481 | else |
| 482 | return 1; |
| 483 | } |
| 484 | |
| 485 | //! Regression test for hanging that occurred with the first version of cancellation propagation |
| 486 | void TestCancelation3 () { |
| 487 | Worker w; |
| 488 | int result = w.DoWork (0); |
| 489 | int expected = w.Validate(0); |
| 490 | ASSERT ( result == expected, "Wrong calculation result" ); |
| 491 | } |
| 492 | |
| 493 | struct StatsCounters { |
| 494 | tbb::atomic<size_t> my_total_created; |
| 495 | tbb::atomic<size_t> my_total_deleted; |
| 496 | StatsCounters() { |
| 497 | my_total_created = 0; |
| 498 | my_total_deleted = 0; |
| 499 | } |
| 500 | }; |
| 501 | |
| 502 | class ParReduceBody { |
| 503 | StatsCounters* my_stats; |
| 504 | size_t my_id; |
| 505 | bool my_exception; |
| 506 | |
| 507 | public: |
| 508 | ParReduceBody( StatsCounters& s_, bool e_ ) : my_stats(&s_), my_exception(e_) { |
| 509 | my_id = my_stats->my_total_created++; |
| 510 | } |
| 511 | |
| 512 | ParReduceBody( const ParReduceBody& lhs ) { |
| 513 | my_stats = lhs.my_stats; |
| 514 | my_id = my_stats->my_total_created++; |
| 515 | } |
| 516 | |
| 517 | ParReduceBody( ParReduceBody& lhs, tbb::split ) { |
| 518 | my_stats = lhs.my_stats; |
| 519 | my_id = my_stats->my_total_created++; |
| 520 | } |
| 521 | |
| 522 | ~ParReduceBody(){ ++my_stats->my_total_deleted; } |
| 523 | |
| 524 | void operator()( const tbb::blocked_range<std::size_t>& /*range*/ ) const { |
| 525 | //Do nothing, except for one task (chosen arbitrarily) |
| 526 | if( my_id >= 12 ) { |
| 527 | if( my_exception ) |
| 528 | ThrowTestException(1); |
| 529 | else |
| 530 | tbb::task::self().cancel_group_execution(); |
| 531 | } |
| 532 | } |
| 533 | |
| 534 | void join( ParReduceBody& /*rhs*/ ) {} |
| 535 | }; |
| 536 | |
| 537 | void TestCancelation4() { |
| 538 | StatsCounters statsObj; |
| 539 | __TBB_TRY { |
| 540 | tbb::task_group_context tgc1, tgc2; |
| 541 | ParReduceBody body_for_cancellation(statsObj, false), body_for_exception(statsObj, true); |
| 542 | tbb::parallel_reduce( tbb::blocked_range<std::size_t>(0,100000000,100), body_for_cancellation, tbb::simple_partitioner(), tgc1 ); |
| 543 | tbb::parallel_reduce( tbb::blocked_range<std::size_t>(0,100000000,100), body_for_exception, tbb::simple_partitioner(), tgc2 ); |
| 544 | } __TBB_CATCH(...) {} |
| 545 | ASSERT ( statsObj.my_total_created==statsObj.my_total_deleted, "Not all parallel_reduce body objects created were reclaimed" ); |
| 546 | } |
| 547 | |
| 548 | void RunParForAndReduceTests () { |
| 549 | REMARK( "parallel for and reduce tests\n" ); |
| 550 | tbb::task_scheduler_init init (g_NumThreads); |
| 551 | g_Master = Harness::CurrentTid(); |
| 552 | |
| 553 | #if TBB_USE_EXCEPTIONS && !__TBB_THROW_ACROSS_MODULE_BOUNDARY_BROKEN |
| 554 | Test0(); |
| 555 | Test1(); |
| 556 | Test2(); |
| 557 | Test3(); |
| 558 | Test4(); |
| 559 | #endif /* TBB_USE_EXCEPTIONS && !__TBB_THROW_ACROSS_MODULE_BOUNDARY_BROKEN */ |
| 560 | TestCancelation1(); |
| 561 | TestCancelation2(); |
| 562 | TestCancelation3(); |
| 563 | TestCancelation4(); |
| 564 | } |
| 565 | |
| 566 | //////////////////////////////////////////////////////////////////////////////// |
| 567 | // Tests for tbb::parallel_do |
| 568 | |
| 569 | #define ITER_RANGE 1000 |
| 570 | #define ITEMS_TO_FEED 50 |
| 571 | #define INNER_ITER_RANGE 100 |
| 572 | #define OUTER_ITER_RANGE 50 |
| 573 | |
| 574 | #define PREPARE_RANGE(Iterator, rangeSize) \ |
| 575 | size_t test_vector[rangeSize + 1]; \ |
| 576 | for (int i =0; i < rangeSize; i++) \ |
| 577 | test_vector[i] = i; \ |
| 578 | Iterator begin(&test_vector[0]); \ |
| 579 | Iterator end(&test_vector[rangeSize]) |
| 580 | |
| 581 | void Feed ( tbb::parallel_do_feeder<size_t> &feeder, size_t val ) { |
| 582 | if (g_FedTasksCount < ITEMS_TO_FEED) { |
| 583 | ++g_FedTasksCount; |
| 584 | feeder.add(val); |
| 585 | } |
| 586 | } |
| 587 | |
| 588 | #include "harness_iterator.h" |
| 589 | |
| 590 | #if TBB_USE_EXCEPTIONS |
| 591 | |
| 592 | // Simple functor object with exception |
| 593 | class SimpleParDoBody { |
| 594 | public: |
| 595 | void operator() ( size_t &value ) const { |
| 596 | ++g_CurExecuted; |
| 597 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 598 | else g_NonMasterExecuted = true; |
| 599 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 600 | Harness::ConcurrencyTracker ct; |
| 601 | value += 1000; |
| 602 | WaitUntilConcurrencyPeaks(); |
| 603 | ThrowTestException(1); |
| 604 | } |
| 605 | }; |
| 606 | |
| 607 | // Simple functor object with exception and feeder |
| 608 | class SimpleParDoBodyWithFeeder : SimpleParDoBody { |
| 609 | public: |
| 610 | void operator() ( size_t &value, tbb::parallel_do_feeder<size_t> &feeder ) const { |
| 611 | Feed(feeder, 0); |
| 612 | SimpleParDoBody::operator()(value); |
| 613 | } |
| 614 | }; |
| 615 | |
| 616 | // Tests exceptions without nesting |
| 617 | template <class Iterator, class simple_body> |
| 618 | void Test1_parallel_do () { |
| 619 | ResetGlobals(); |
| 620 | PREPARE_RANGE(Iterator, ITER_RANGE); |
| 621 | TRY(); |
| 622 | tbb::parallel_do<Iterator, simple_body>(begin, end, simple_body() ); |
| 623 | CATCH_AND_ASSERT(); |
| 624 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 625 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived cancellation" ); |
| 626 | ASSERT (g_NumExceptionsCaught == 1, "No try_blocks in any body expected in this test" ); |
| 627 | if ( !g_SolitaryException ) |
| 628 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 629 | |
| 630 | } // void Test1_parallel_do () |
| 631 | |
| 632 | template <class Iterator> |
| 633 | class OuterParDoBody { |
| 634 | public: |
| 635 | void operator()( size_t& /*value*/ ) const { |
| 636 | ++g_OuterParCalls; |
| 637 | PREPARE_RANGE(Iterator, INNER_ITER_RANGE); |
| 638 | tbb::parallel_do<Iterator, SimpleParDoBody>(begin, end, SimpleParDoBody()); |
| 639 | } |
| 640 | }; |
| 641 | |
| 642 | template <class Iterator> |
| 643 | class OuterParDoBodyWithFeeder : OuterParDoBody<Iterator> { |
| 644 | public: |
| 645 | void operator()( size_t& value, tbb::parallel_do_feeder<size_t>& feeder ) const { |
| 646 | Feed(feeder, 0); |
| 647 | OuterParDoBody<Iterator>::operator()(value); |
| 648 | } |
| 649 | }; |
| 650 | |
| 651 | //! Uses parallel_do body containing an inner parallel_do with the default context not wrapped by a try-block. |
| 652 | /** Inner algorithms are spawned inside the new bound context by default. Since |
| 653 | exceptions thrown from the inner parallel_do are not handled by the caller |
| 654 | (outer parallel_do body) in this test, they will cancel all the sibling inner |
| 655 | algorithms. **/ |
| 656 | template <class Iterator, class outer_body> |
| 657 | void Test2_parallel_do () { |
| 658 | ResetGlobals(); |
| 659 | PREPARE_RANGE(Iterator, ITER_RANGE); |
| 660 | TRY(); |
| 661 | tbb::parallel_do<Iterator, outer_body >(begin, end, outer_body() ); |
| 662 | CATCH_AND_ASSERT(); |
| 663 | //if ( g_SolitaryException ) |
| 664 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 665 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived cancellation" ); |
| 666 | ASSERT (g_NumExceptionsCaught == 1, "No try_blocks in any body expected in this test" ); |
| 667 | if ( !g_SolitaryException ) |
| 668 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 669 | } // void Test2_parallel_do () |
| 670 | |
| 671 | template <class Iterator> |
| 672 | class OuterParDoBodyWithIsolatedCtx { |
| 673 | public: |
| 674 | void operator()( size_t& /*value*/ ) const { |
| 675 | tbb::task_group_context ctx(tbb::task_group_context::isolated); |
| 676 | ++g_OuterParCalls; |
| 677 | PREPARE_RANGE(Iterator, INNER_ITER_RANGE); |
| 678 | tbb::parallel_do<Iterator, SimpleParDoBody>(begin, end, SimpleParDoBody(), ctx); |
| 679 | } |
| 680 | }; |
| 681 | |
| 682 | template <class Iterator> |
| 683 | class OuterParDoBodyWithIsolatedCtxWithFeeder : OuterParDoBodyWithIsolatedCtx<Iterator> { |
| 684 | public: |
| 685 | void operator()( size_t& value, tbb::parallel_do_feeder<size_t> &feeder ) const { |
| 686 | Feed(feeder, 0); |
| 687 | OuterParDoBodyWithIsolatedCtx<Iterator>::operator()(value); |
| 688 | } |
| 689 | }; |
| 690 | |
| 691 | //! Uses parallel_do body invoking an inner parallel_do with an isolated context without a try-block. |
| 692 | /** Even though exceptions thrown from the inner parallel_do are not handled |
| 693 | by the caller in this test, they will not affect sibling inner algorithms |
| 694 | already running because of the isolated contexts. However because the first |
| 695 | exception cancels the root parallel_do, at most the first g_NumThreads subranges |
| 696 | will be processed (which launch inner parallel_dos) **/ |
| 697 | template <class Iterator, class outer_body> |
| 698 | void Test3_parallel_do () { |
| 699 | ResetGlobals(); |
| 700 | PREPARE_RANGE(Iterator, OUTER_ITER_RANGE); |
| 701 | intptr_t innerCalls = INNER_ITER_RANGE, |
| 702 | // The assumption here is the same as in outer parallel fors. |
| 703 | minExecuted = (g_NumThreads - 1) * innerCalls; |
| 704 | g_Master = Harness::CurrentTid(); |
| 705 | TRY(); |
| 706 | tbb::parallel_do<Iterator, outer_body >(begin, end, outer_body()); |
| 707 | CATCH_AND_ASSERT(); |
| 708 | // figure actual number of expected executions given the number of outer PDos started. |
| 709 | minExecuted = (g_OuterParCalls - 1) * innerCalls; |
| 710 | // one extra thread may run a task that sees cancellation. Infrequent but possible |
| 711 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived exception" ); |
| 712 | if(g_TGCCancelled > g_NumThreads) REMARK("Extra thread(s) executed after cancel (%d vs. %d)\n" , |
| 713 | (int)g_TGCCancelled, (int)g_NumThreads); |
| 714 | if ( g_SolitaryException ) { |
| 715 | ASSERT (g_CurExecuted > minExecuted, "Too few tasks survived exception" ); |
| 716 | ASSERT (g_CurExecuted <= minExecuted + (g_ExecutedAtLastCatch + g_NumThreads), "Too many tasks survived exception" ); |
| 717 | } |
| 718 | ASSERT (g_NumExceptionsCaught == 1, "No try_blocks in any body expected in this test" ); |
| 719 | if ( !g_SolitaryException ) |
| 720 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 721 | } // void Test3_parallel_do () |
| 722 | |
| 723 | template <class Iterator> |
| 724 | class OuterParDoWithEhBody { |
| 725 | public: |
| 726 | void operator()( size_t& /*value*/ ) const { |
| 727 | tbb::task_group_context ctx(tbb::task_group_context::isolated); |
| 728 | ++g_OuterParCalls; |
| 729 | PREPARE_RANGE(Iterator, INNER_ITER_RANGE); |
| 730 | TRY(); |
| 731 | tbb::parallel_do<Iterator, SimpleParDoBody>(begin, end, SimpleParDoBody(), ctx); |
| 732 | CATCH(); |
| 733 | } |
| 734 | }; |
| 735 | |
| 736 | template <class Iterator> |
| 737 | class OuterParDoWithEhBodyWithFeeder : NoAssign, OuterParDoWithEhBody<Iterator> { |
| 738 | public: |
| 739 | void operator()( size_t &value, tbb::parallel_do_feeder<size_t> &feeder ) const { |
| 740 | Feed(feeder, 0); |
| 741 | OuterParDoWithEhBody<Iterator>::operator()(value); |
| 742 | } |
| 743 | }; |
| 744 | |
| 745 | //! Uses parallel_for body invoking an inner parallel_for (with default bound context) inside a try-block. |
| 746 | /** Since exception(s) thrown from the inner parallel_for are handled by the caller |
| 747 | in this test, they do not affect neither other tasks of the the root parallel_for |
| 748 | nor sibling inner algorithms. **/ |
| 749 | template <class Iterator, class outer_body_with_eh> |
| 750 | void Test4_parallel_do () { |
| 751 | ResetGlobals( true, true ); |
| 752 | PREPARE_RANGE(Iterator, OUTER_ITER_RANGE); |
| 753 | g_Master = Harness::CurrentTid(); |
| 754 | TRY(); |
| 755 | tbb::parallel_do<Iterator, outer_body_with_eh>(begin, end, outer_body_with_eh()); |
| 756 | CATCH(); |
| 757 | ASSERT (!l_ExceptionCaughtAtCurrentLevel, "All exceptions must have been handled in the parallel_do body" ); |
| 758 | intptr_t innerCalls = INNER_ITER_RANGE, |
| 759 | outerCalls = OUTER_ITER_RANGE + g_FedTasksCount, |
| 760 | maxExecuted = outerCalls * innerCalls, |
| 761 | minExecuted = 0; |
| 762 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived exception" ); |
| 763 | if ( g_SolitaryException ) { |
| 764 | minExecuted = maxExecuted - innerCalls; |
| 765 | ASSERT (g_NumExceptionsCaught == 1, "No exception registered" ); |
| 766 | ASSERT (g_CurExecuted >= minExecuted, "Too few tasks executed" ); |
| 767 | // This test has the same property as Test4 (parallel_for); the exception can be |
| 768 | // thrown, but some number of tasks from the outer Pdo can execute after the throw but |
| 769 | // before the cancellation is signaled (have seen 36). |
| 770 | ASSERT_WARNING(g_CurExecuted < maxExecuted || g_TGCCancelled, "All tasks survived exception. Oversubscription?" ); |
| 771 | } |
| 772 | else { |
| 773 | minExecuted = g_NumExceptionsCaught; |
| 774 | ASSERT (g_NumExceptionsCaught > 1 && g_NumExceptionsCaught <= outerCalls, "Unexpected actual number of exceptions" ); |
| 775 | ASSERT (g_CurExecuted >= minExecuted, "Too many executed tasks reported" ); |
| 776 | ASSERT (g_CurExecuted < g_ExecutedAtLastCatch + g_NumThreads + outerCalls, "Too many tasks survived multiple exceptions" ); |
| 777 | ASSERT (g_CurExecuted <= outerCalls * (1 + g_NumThreads), "Too many tasks survived exception" ); |
| 778 | } |
| 779 | } // void Test4_parallel_do () |
| 780 | |
| 781 | // This body throws an exception only if the task was added by feeder |
| 782 | class ParDoBodyWithThrowingFeederTasks { |
| 783 | public: |
| 784 | //! This form of the function call operator can be used when the body needs to add more work during the processing |
| 785 | void operator() ( size_t &value, tbb::parallel_do_feeder<size_t> &feeder ) const { |
| 786 | ++g_CurExecuted; |
| 787 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 788 | else g_NonMasterExecuted = true; |
| 789 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 790 | Feed(feeder, 1); |
| 791 | if (value == 1) |
| 792 | ThrowTestException(1); |
| 793 | } |
| 794 | }; // class ParDoBodyWithThrowingFeederTasks |
| 795 | |
| 796 | // Test exception in task, which was added by feeder. |
| 797 | template <class Iterator> |
| 798 | void Test5_parallel_do () { |
| 799 | ResetGlobals(); |
| 800 | PREPARE_RANGE(Iterator, ITER_RANGE); |
| 801 | g_Master = Harness::CurrentTid(); |
| 802 | TRY(); |
| 803 | tbb::parallel_do<Iterator, ParDoBodyWithThrowingFeederTasks>(begin, end, ParDoBodyWithThrowingFeederTasks()); |
| 804 | CATCH(); |
| 805 | if (g_SolitaryException) { |
| 806 | // Failure occurs when g_ExceptionInMaster is false, but all the 1 values in the range |
| 807 | // are handled by the master thread. In this case no throw occurs. |
| 808 | ASSERT (l_ExceptionCaughtAtCurrentLevel // we saw an exception |
| 809 | || (!g_ExceptionInMaster && !g_NonMasterExecutedThrow) // non-master throws but none tried |
| 810 | || (g_ExceptionInMaster && !g_MasterExecutedThrow) // master throws but master didn't try |
| 811 | , "At least one exception should occur" ); |
| 812 | if(!g_ExceptionCaught) { |
| 813 | if(g_ExceptionInMaster) |
| 814 | REMARK("PDo exception not thrown; non-masters handled all throwing values.\n" ); |
| 815 | else |
| 816 | REMARK("PDo exception not thrown; master handled all throwing values.\n" ); |
| 817 | } |
| 818 | } |
| 819 | } // void Test5_parallel_do () |
| 820 | |
| 821 | #endif /* TBB_USE_EXCEPTIONS */ |
| 822 | |
| 823 | class ParDoBodyToCancel { |
| 824 | public: |
| 825 | void operator()( size_t& /*value*/ ) const { |
| 826 | ++g_CurExecuted; |
| 827 | CancellatorTask::WaitUntilReady(); |
| 828 | } |
| 829 | }; |
| 830 | |
| 831 | class ParDoBodyToCancelWithFeeder : ParDoBodyToCancel { |
| 832 | public: |
| 833 | void operator()( size_t& value, tbb::parallel_do_feeder<size_t> &feeder ) const { |
| 834 | Feed(feeder, 0); |
| 835 | ParDoBodyToCancel::operator()(value); |
| 836 | } |
| 837 | }; |
| 838 | |
| 839 | template<class B, class Iterator> |
| 840 | class ParDoWorkerTask : public tbb::task { |
| 841 | tbb::task_group_context &my_ctx; |
| 842 | |
| 843 | tbb::task* execute () __TBB_override { |
| 844 | PREPARE_RANGE(Iterator, INNER_ITER_RANGE); |
| 845 | tbb::parallel_do<Iterator, B>( begin, end, B(), my_ctx ); |
| 846 | return NULL; |
| 847 | } |
| 848 | public: |
| 849 | ParDoWorkerTask ( tbb::task_group_context& ctx ) : my_ctx(ctx) {} |
| 850 | }; |
| 851 | |
| 852 | //! Test for cancelling an algorithm from outside (from a task running in parallel with the algorithm). |
| 853 | template <class Iterator, class body_to_cancel> |
| 854 | void TestCancelation1_parallel_do () { |
| 855 | ResetGlobals( false ); |
| 856 | intptr_t threshold = 10; |
| 857 | tbb::task_group_context ctx; |
| 858 | ctx.reset(); |
| 859 | tbb::empty_task &r = *new( tbb::task::allocate_root() ) tbb::empty_task; |
| 860 | r.set_ref_count(3); |
| 861 | r.spawn( *new( r.allocate_child() ) CancellatorTask(ctx, threshold) ); |
| 862 | __TBB_Yield(); |
| 863 | r.spawn( *new( r.allocate_child() ) ParDoWorkerTask<body_to_cancel, Iterator>(ctx) ); |
| 864 | TRY(); |
| 865 | r.wait_for_all(); |
| 866 | CATCH_AND_FAIL(); |
| 867 | ASSERT (g_CurExecuted < g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks were executed after cancellation" ); |
| 868 | r.destroy(r); |
| 869 | } |
| 870 | |
| 871 | class ParDoBodyToCancel2 { |
| 872 | public: |
| 873 | void operator()( size_t& /*value*/ ) const { |
| 874 | ++g_CurExecuted; |
| 875 | Harness::ConcurrencyTracker ct; |
| 876 | // The test will hang (and be timed out by the test system) if is_cancelled() is broken |
| 877 | while( !tbb::task::self().is_cancelled() ) |
| 878 | __TBB_Yield(); |
| 879 | } |
| 880 | }; |
| 881 | |
| 882 | class ParDoBodyToCancel2WithFeeder : ParDoBodyToCancel2 { |
| 883 | public: |
| 884 | void operator()( size_t& value, tbb::parallel_do_feeder<size_t> &feeder ) const { |
| 885 | Feed(feeder, 0); |
| 886 | ParDoBodyToCancel2::operator()(value); |
| 887 | } |
| 888 | }; |
| 889 | |
| 890 | //! Test for cancelling an algorithm from outside (from a task running in parallel with the algorithm). |
| 891 | /** This version also tests task::is_cancelled() method. **/ |
| 892 | template <class Iterator, class body_to_cancel> |
| 893 | void TestCancelation2_parallel_do () { |
| 894 | ResetGlobals(); |
| 895 | RunCancellationTest<ParDoWorkerTask<body_to_cancel, Iterator>, CancellatorTask2>(); |
| 896 | } |
| 897 | |
| 898 | #define RunWithSimpleBody(func, body) \ |
| 899 | func<Harness::RandomIterator<size_t>, body>(); \ |
| 900 | func<Harness::RandomIterator<size_t>, body##WithFeeder>(); \ |
| 901 | func<Harness::ForwardIterator<size_t>, body>(); \ |
| 902 | func<Harness::ForwardIterator<size_t>, body##WithFeeder>() |
| 903 | |
| 904 | #define RunWithTemplatedBody(func, body) \ |
| 905 | func<Harness::RandomIterator<size_t>, body<Harness::RandomIterator<size_t> > >(); \ |
| 906 | func<Harness::RandomIterator<size_t>, body##WithFeeder<Harness::RandomIterator<size_t> > >(); \ |
| 907 | func<Harness::ForwardIterator<size_t>, body<Harness::ForwardIterator<size_t> > >(); \ |
| 908 | func<Harness::ForwardIterator<size_t>, body##WithFeeder<Harness::ForwardIterator<size_t> > >() |
| 909 | |
| 910 | void RunParDoTests() { |
| 911 | REMARK( "parallel do tests\n" ); |
| 912 | tbb::task_scheduler_init init (g_NumThreads); |
| 913 | g_Master = Harness::CurrentTid(); |
| 914 | #if TBB_USE_EXCEPTIONS && !__TBB_THROW_ACROSS_MODULE_BOUNDARY_BROKEN |
| 915 | RunWithSimpleBody(Test1_parallel_do, SimpleParDoBody); |
| 916 | RunWithTemplatedBody(Test2_parallel_do, OuterParDoBody); |
| 917 | RunWithTemplatedBody(Test3_parallel_do, OuterParDoBodyWithIsolatedCtx); |
| 918 | RunWithTemplatedBody(Test4_parallel_do, OuterParDoWithEhBody); |
| 919 | Test5_parallel_do<Harness::ForwardIterator<size_t> >(); |
| 920 | Test5_parallel_do<Harness::RandomIterator<size_t> >(); |
| 921 | #endif /* TBB_USE_EXCEPTIONS && !__TBB_THROW_ACROSS_MODULE_BOUNDARY_BROKEN */ |
| 922 | RunWithSimpleBody(TestCancelation1_parallel_do, ParDoBodyToCancel); |
| 923 | RunWithSimpleBody(TestCancelation2_parallel_do, ParDoBodyToCancel2); |
| 924 | } |
| 925 | |
| 926 | //////////////////////////////////////////////////////////////////////////////// |
| 927 | // Tests for tbb::pipeline |
| 928 | |
| 929 | #define NUM_ITEMS 100 |
| 930 | |
| 931 | const size_t c_DataEndTag = size_t(~0); |
| 932 | |
| 933 | int g_NumTokens = 0; |
| 934 | |
| 935 | // Simple input filter class, it assigns 1 to all array members |
| 936 | // It stops when it receives item equal to -1 |
| 937 | class InputFilter: public tbb::filter { |
| 938 | tbb::atomic<size_t> m_Item; |
| 939 | size_t m_Buffer[NUM_ITEMS + 1]; |
| 940 | public: |
| 941 | InputFilter() : tbb::filter(parallel) { |
| 942 | m_Item = 0; |
| 943 | for (size_t i = 0; i < NUM_ITEMS; ++i ) |
| 944 | m_Buffer[i] = 1; |
| 945 | m_Buffer[NUM_ITEMS] = c_DataEndTag; |
| 946 | } |
| 947 | |
| 948 | void* operator()( void* ) __TBB_override { |
| 949 | size_t item = m_Item.fetch_and_increment(); |
| 950 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 951 | else g_NonMasterExecuted = true; |
| 952 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 953 | if(item == 1) { |
| 954 | ++g_PipelinesStarted; // count on emitting the first item. |
| 955 | } |
| 956 | if ( item >= NUM_ITEMS ) |
| 957 | return NULL; |
| 958 | m_Buffer[item] = 1; |
| 959 | return &m_Buffer[item]; |
| 960 | } |
| 961 | |
| 962 | size_t* buffer() { return m_Buffer; } |
| 963 | }; // class InputFilter |
| 964 | |
| 965 | // Pipeline filter, without exceptions throwing |
| 966 | class NoThrowFilter : public tbb::filter { |
| 967 | size_t m_Value; |
| 968 | public: |
| 969 | enum operation { |
| 970 | addition, |
| 971 | subtraction, |
| 972 | multiplication |
| 973 | } m_Operation; |
| 974 | |
| 975 | NoThrowFilter(operation _operation, size_t value, bool is_parallel) |
| 976 | : filter(is_parallel? tbb::filter::parallel : tbb::filter::serial_in_order), |
| 977 | m_Value(value), m_Operation(_operation) |
| 978 | {} |
| 979 | void* operator()(void* item) __TBB_override { |
| 980 | size_t &value = *(size_t*)item; |
| 981 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 982 | else g_NonMasterExecuted = true; |
| 983 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 984 | ASSERT(value != c_DataEndTag, "terminator element is being processed" ); |
| 985 | switch (m_Operation){ |
| 986 | case addition: |
| 987 | value += m_Value; |
| 988 | break; |
| 989 | case subtraction: |
| 990 | value -= m_Value; |
| 991 | break; |
| 992 | case multiplication: |
| 993 | value *= m_Value; |
| 994 | break; |
| 995 | default: |
| 996 | ASSERT(0, "Wrong operation parameter passed to NoThrowFilter" ); |
| 997 | } // switch (m_Operation) |
| 998 | return item; |
| 999 | } |
| 1000 | }; |
| 1001 | |
| 1002 | // Test pipeline without exceptions throwing |
| 1003 | void Test0_pipeline () { |
| 1004 | ResetGlobals(); |
| 1005 | // Run test when serial filter is the first non-input filter |
| 1006 | InputFilter inputFilter; //Emits NUM_ITEMS items |
| 1007 | NoThrowFilter filter1(NoThrowFilter::addition, 99, false); |
| 1008 | NoThrowFilter filter2(NoThrowFilter::subtraction, 90, true); |
| 1009 | NoThrowFilter filter3(NoThrowFilter::multiplication, 5, false); |
| 1010 | // Result should be 50 for all items except the last |
| 1011 | tbb::pipeline p; |
| 1012 | p.add_filter(inputFilter); |
| 1013 | p.add_filter(filter1); |
| 1014 | p.add_filter(filter2); |
| 1015 | p.add_filter(filter3); |
| 1016 | p.run(8); |
| 1017 | for (size_t i = 0; i < NUM_ITEMS; ++i) |
| 1018 | ASSERT(inputFilter.buffer()[i] == 50, "pipeline didn't process items properly" ); |
| 1019 | } // void Test0_pipeline () |
| 1020 | |
| 1021 | #if TBB_USE_EXCEPTIONS |
| 1022 | |
| 1023 | // Simple filter with exception throwing. If parallel, will wait until |
| 1024 | // as many parallel filters start as there are threads. |
| 1025 | class SimpleFilter : public tbb::filter { |
| 1026 | bool m_canThrow; |
| 1027 | public: |
| 1028 | SimpleFilter (tbb::filter::mode _mode, bool canThrow ) : filter (_mode), m_canThrow(canThrow) {} |
| 1029 | void* operator()(void* item) __TBB_override { |
| 1030 | ++g_CurExecuted; |
| 1031 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 1032 | else g_NonMasterExecuted = true; |
| 1033 | if( tbb::task::self().is_cancelled() ) ++g_TGCCancelled; |
| 1034 | if ( m_canThrow ) { |
| 1035 | if ( !is_serial() ) { |
| 1036 | Harness::ConcurrencyTracker ct; |
| 1037 | WaitUntilConcurrencyPeaks( min(g_NumTokens, g_NumThreads) ); |
| 1038 | } |
| 1039 | ThrowTestException(1); |
| 1040 | } |
| 1041 | return item; |
| 1042 | } |
| 1043 | }; // class SimpleFilter |
| 1044 | |
| 1045 | // This enumeration represents filters order in pipeline |
| 1046 | struct FilterSet { |
| 1047 | tbb::filter::mode mode1, |
| 1048 | mode2; |
| 1049 | bool throw1, |
| 1050 | throw2; |
| 1051 | |
| 1052 | FilterSet( tbb::filter::mode m1, tbb::filter::mode m2, bool t1, bool t2 ) |
| 1053 | : mode1(m1), mode2(m2), throw1(t1), throw2(t2) |
| 1054 | {} |
| 1055 | }; // struct FilterSet |
| 1056 | |
| 1057 | FilterSet serial_parallel( tbb::filter::serial, tbb::filter::parallel, /*throw1*/false, /*throw2*/true ); |
| 1058 | |
| 1059 | template<typename InFilter, typename Filter> |
| 1060 | class CustomPipeline : protected tbb::pipeline { |
| 1061 | InFilter inputFilter; |
| 1062 | Filter filter1; |
| 1063 | Filter filter2; |
| 1064 | public: |
| 1065 | CustomPipeline( const FilterSet& filters ) |
| 1066 | : filter1(filters.mode1, filters.throw1), filter2(filters.mode2, filters.throw2) |
| 1067 | { |
| 1068 | add_filter(inputFilter); |
| 1069 | add_filter(filter1); |
| 1070 | add_filter(filter2); |
| 1071 | } |
| 1072 | void run () { tbb::pipeline::run(g_NumTokens); } |
| 1073 | void run ( tbb::task_group_context& ctx ) { tbb::pipeline::run(g_NumTokens, ctx); } |
| 1074 | |
| 1075 | using tbb::pipeline::add_filter; |
| 1076 | }; |
| 1077 | |
| 1078 | typedef CustomPipeline<InputFilter, SimpleFilter> SimplePipeline; |
| 1079 | |
| 1080 | // Tests exceptions without nesting |
| 1081 | void Test1_pipeline ( const FilterSet& filters ) { |
| 1082 | ResetGlobals(); |
| 1083 | SimplePipeline testPipeline(filters); |
| 1084 | TRY(); |
| 1085 | testPipeline.run(); |
| 1086 | if ( g_CurExecuted == 2 * NUM_ITEMS ) { |
| 1087 | // all the items were processed, though an exception was supposed to occur. |
| 1088 | if(!g_ExceptionInMaster && g_NonMasterExecutedThrow > 0) { |
| 1089 | // if !g_ExceptionInMaster, the master thread is not allowed to throw. |
| 1090 | // if g_nonMasterExcutedThrow > 0 then a thread besides the master tried to throw. |
| 1091 | ASSERT(filters.mode1 != tbb::filter::parallel && filters.mode2 != tbb::filter::parallel, "Unusual count" ); |
| 1092 | } |
| 1093 | else { |
| 1094 | REMARK("test1_Pipeline with %d threads: Only the master thread tried to throw, and it is not allowed to.\n" , (int)g_NumThreads); |
| 1095 | } |
| 1096 | // In case of all serial filters they might be all executed in the thread(s) |
| 1097 | // where exceptions are not allowed by the common test logic. So we just quit. |
| 1098 | return; |
| 1099 | } |
| 1100 | CATCH_AND_ASSERT(); |
| 1101 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived exception" ); |
| 1102 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 1103 | ASSERT (g_NumExceptionsCaught == 1, "No try_blocks in any body expected in this test" ); |
| 1104 | if ( !g_SolitaryException ) |
| 1105 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 1106 | |
| 1107 | } // void Test1_pipeline () |
| 1108 | |
| 1109 | // Filter with nesting |
| 1110 | class OuterFilter : public tbb::filter { |
| 1111 | public: |
| 1112 | OuterFilter (tbb::filter::mode _mode, bool ) : filter (_mode) {} |
| 1113 | |
| 1114 | void* operator()(void* item) __TBB_override { |
| 1115 | ++g_OuterParCalls; |
| 1116 | SimplePipeline testPipeline(serial_parallel); |
| 1117 | testPipeline.run(); |
| 1118 | return item; |
| 1119 | } |
| 1120 | }; // class OuterFilter |
| 1121 | |
| 1122 | //! Uses pipeline containing an inner pipeline with the default context not wrapped by a try-block. |
| 1123 | /** Inner algorithms are spawned inside the new bound context by default. Since |
| 1124 | exceptions thrown from the inner pipeline are not handled by the caller |
| 1125 | (outer pipeline body) in this test, they will cancel all the sibling inner |
| 1126 | algorithms. **/ |
| 1127 | void Test2_pipeline ( const FilterSet& filters ) { |
| 1128 | ResetGlobals(); |
| 1129 | g_NestedPipelines = true; |
| 1130 | CustomPipeline<InputFilter, OuterFilter> testPipeline(filters); |
| 1131 | TRY(); |
| 1132 | testPipeline.run(); |
| 1133 | CATCH_AND_ASSERT(); |
| 1134 | bool okayNoExceptionCaught = (g_ExceptionInMaster && !g_MasterExecutedThrow) || (!g_ExceptionInMaster && !g_NonMasterExecutedThrow); |
| 1135 | ASSERT (g_NumExceptionsCaught == 1 || okayNoExceptionCaught, "No try_blocks in any body expected in this test" ); |
| 1136 | if ( g_SolitaryException ) { |
| 1137 | if( g_TGCCancelled > g_NumThreads) REMARK( "Extra tasks ran after exception thrown (%d vs. %d)\n" , |
| 1138 | (int)g_TGCCancelled, (int)g_NumThreads); |
| 1139 | } |
| 1140 | else { |
| 1141 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived exception" ); |
| 1142 | } |
| 1143 | } // void Test2_pipeline () |
| 1144 | |
| 1145 | //! creates isolated inner pipeline and runs it. |
| 1146 | class OuterFilterWithIsolatedCtx : public tbb::filter { |
| 1147 | public: |
| 1148 | OuterFilterWithIsolatedCtx(tbb::filter::mode m, bool ) : filter(m) {} |
| 1149 | |
| 1150 | void* operator()(void* item) __TBB_override { |
| 1151 | ++g_OuterParCalls; |
| 1152 | tbb::task_group_context ctx(tbb::task_group_context::isolated); |
| 1153 | // create inner pipeline with serial input, parallel output filter, second filter throws |
| 1154 | SimplePipeline testPipeline(serial_parallel); |
| 1155 | testPipeline.run(ctx); |
| 1156 | return item; |
| 1157 | } |
| 1158 | }; // class OuterFilterWithIsolatedCtx |
| 1159 | |
| 1160 | //! Uses pipeline invoking an inner pipeline with an isolated context without a try-block. |
| 1161 | /** Even though exceptions thrown from the inner pipeline are not handled |
| 1162 | by the caller in this test, they will not affect sibling inner algorithms |
| 1163 | already running because of the isolated contexts. However because the first |
| 1164 | exception cancels the root parallel_do only the first g_NumThreads subranges |
| 1165 | will be processed (which launch inner pipelines) **/ |
| 1166 | void Test3_pipeline ( const FilterSet& filters ) { |
| 1167 | for( int nTries = 1; nTries <= 4; ++nTries) { |
| 1168 | ResetGlobals(); |
| 1169 | g_NestedPipelines = true; |
| 1170 | g_Master = Harness::CurrentTid(); |
| 1171 | intptr_t innerCalls = NUM_ITEMS, |
| 1172 | minExecuted = (g_NumThreads - 1) * innerCalls; |
| 1173 | CustomPipeline<InputFilter, OuterFilterWithIsolatedCtx> testPipeline(filters); |
| 1174 | TRY(); |
| 1175 | testPipeline.run(); |
| 1176 | CATCH_AND_ASSERT(); |
| 1177 | |
| 1178 | bool okayNoExceptionCaught = (g_ExceptionInMaster && !g_MasterExecuted) || |
| 1179 | (!g_ExceptionInMaster && !g_NonMasterExecuted); |
| 1180 | // only test assertions if the test threw an exception (or we don't care) |
| 1181 | bool testSucceeded = okayNoExceptionCaught || g_NumExceptionsCaught > 0; |
| 1182 | if(testSucceeded) { |
| 1183 | if (g_SolitaryException) { |
| 1184 | |
| 1185 | // The test is one outer pipeline with two NestedFilters that each start an inner pipeline. |
| 1186 | // Each time the input filter of a pipeline delivers its first item, it increments |
| 1187 | // g_PipelinesStarted. When g_SolitaryException, the throw will not occur until |
| 1188 | // g_PipelinesStarted >= 3. (This is so at least a second pipeline in its own isolated |
| 1189 | // context will start; that is what we're testing.) |
| 1190 | // |
| 1191 | // There are two pipelines which will NOT run to completion when a solitary throw |
| 1192 | // happens in an isolated inner context: the outer pipeline and the pipeline which |
| 1193 | // throws. All the other pipelines which start should run to completion. But only |
| 1194 | // inner body invocations are counted. |
| 1195 | // |
| 1196 | // So g_CurExecuted should be about |
| 1197 | // |
| 1198 | // (2*NUM_ITEMS) * (g_PipelinesStarted - 2) + 1 |
| 1199 | // ^ executions for each completed pipeline |
| 1200 | // ^ completing pipelines (remembering two will not complete) |
| 1201 | // ^ one for the inner throwing pipeline |
| 1202 | |
| 1203 | minExecuted = (2*NUM_ITEMS) * (g_PipelinesStarted - 2) + 1; |
| 1204 | // each failing pipeline must execute at least two tasks |
| 1205 | ASSERT(g_CurExecuted >= minExecuted, "Too few tasks survived exception" ); |
| 1206 | // no more than g_NumThreads tasks will be executed in a cancelled context. Otherwise |
| 1207 | // tasks not executing at throw were scheduled. |
| 1208 | ASSERT( g_TGCCancelled <= g_NumThreads, "Tasks not in-flight were executed" ); |
| 1209 | ASSERT(g_NumExceptionsCaught == 1, "Should have only one exception" ); |
| 1210 | // if we're only throwing from the master thread, and that thread didn't |
| 1211 | // participate in the pipelines, then no throw occurred. |
| 1212 | if(g_ExceptionInMaster && !g_MasterExecuted) { |
| 1213 | REMARK_ONCE("Master expected to throw, but didn't participate.\n" ); |
| 1214 | } |
| 1215 | else if(!g_ExceptionInMaster && !g_NonMasterExecuted) { |
| 1216 | REMARK_ONCE("Non-master expected to throw, but didn't participate.\n" ); |
| 1217 | } |
| 1218 | } |
| 1219 | ASSERT (g_NumExceptionsCaught == 1 || okayNoExceptionCaught, "No try_blocks in any body expected in this test" ); |
| 1220 | ASSERT ((g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads) || okayNoExceptionCaught, "Too many tasks survived exception" ); |
| 1221 | if(nTries > 1) REMARK("Test3_pipeline succeeeded on try %d\n" , nTries); |
| 1222 | return; |
| 1223 | } |
| 1224 | } |
| 1225 | REMARK_ONCE("Test3_pipeline failed for g_NumThreads==%d, g_ExceptionInMaster==%s , g_SolitaryException==%s\n" , |
| 1226 | g_NumThreads, g_ExceptionInMaster?"T" :"F" , g_SolitaryException?"T" :"F" ); |
| 1227 | } // void Test3_pipeline () |
| 1228 | |
| 1229 | class OuterFilterWithEhBody : public tbb::filter { |
| 1230 | public: |
| 1231 | OuterFilterWithEhBody(tbb::filter::mode m, bool ) : filter(m) {} |
| 1232 | |
| 1233 | void* operator()(void* item) __TBB_override { |
| 1234 | tbb::task_group_context ctx(tbb::task_group_context::isolated); |
| 1235 | ++g_OuterParCalls; |
| 1236 | SimplePipeline testPipeline(serial_parallel); |
| 1237 | TRY(); |
| 1238 | testPipeline.run(ctx); |
| 1239 | CATCH(); |
| 1240 | return item; |
| 1241 | } |
| 1242 | }; // class OuterFilterWithEhBody |
| 1243 | |
| 1244 | //! Uses pipeline body invoking an inner pipeline (with isolated context) inside a try-block. |
| 1245 | /** Since exception(s) thrown from the inner pipeline are handled by the caller |
| 1246 | in this test, they do not affect other tasks of the the root pipeline |
| 1247 | nor sibling inner algorithms. **/ |
| 1248 | void Test4_pipeline ( const FilterSet& filters ) { |
| 1249 | #if __GNUC__ && !__INTEL_COMPILER |
| 1250 | if ( strncmp(__VERSION__, "4.1.0" , 5) == 0 ) { |
| 1251 | REMARK_ONCE("Known issue: one of exception handling tests is skipped.\n" ); |
| 1252 | return; |
| 1253 | } |
| 1254 | #endif |
| 1255 | ResetGlobals( true, true ); |
| 1256 | // each outer pipeline stage will start NUM_ITEMS inner pipelines. |
| 1257 | // each inner pipeline that doesn't throw will process NUM_ITEMS items. |
| 1258 | // for solitary exception there will be one pipeline that only processes one stage, one item. |
| 1259 | // innerCalls should be 2*NUM_ITEMS |
| 1260 | intptr_t innerCalls = 2*NUM_ITEMS, |
| 1261 | outerCalls = 2 * NUM_ITEMS, |
| 1262 | maxExecuted = outerCalls * innerCalls; // the number of invocations of the inner pipelines |
| 1263 | CustomPipeline<InputFilter, OuterFilterWithEhBody> testPipeline(filters); |
| 1264 | TRY(); |
| 1265 | testPipeline.run(); |
| 1266 | CATCH_AND_ASSERT(); |
| 1267 | intptr_t minExecuted = 0; |
| 1268 | bool okayNoExceptionCaught = (g_ExceptionInMaster && !g_MasterExecuted) || |
| 1269 | (!g_ExceptionInMaster && !g_NonMasterExecuted); |
| 1270 | if ( g_SolitaryException ) { |
| 1271 | minExecuted = maxExecuted - innerCalls; // one throwing inner pipeline |
| 1272 | ASSERT (g_NumExceptionsCaught == 1 || okayNoExceptionCaught, "No exception registered" ); |
| 1273 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived exception" ); // probably will assert. |
| 1274 | } |
| 1275 | else { |
| 1276 | // we assume throwing pipelines will not count |
| 1277 | minExecuted = (outerCalls - g_NumExceptionsCaught) * innerCalls; |
| 1278 | ASSERT((g_NumExceptionsCaught >= 1 && g_NumExceptionsCaught <= outerCalls)||okayNoExceptionCaught, "Unexpected actual number of exceptions" ); |
| 1279 | ASSERT (g_CurExecuted >= minExecuted, "Too many executed tasks reported" ); |
| 1280 | // too many already-scheduled tasks are started after the first exception is |
| 1281 | // thrown. And g_ExecutedAtLastCatch is updated every time an exception is caught. |
| 1282 | // So with multiple exceptions there are a variable number of tasks that have been |
| 1283 | // discarded because of the signals. |
| 1284 | // each throw is caught, so we will see many cancelled tasks. g_ExecutedAtLastCatch is |
| 1285 | // updated with each throw, so the value will be the number of tasks executed at the last |
| 1286 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks survived multiple exceptions" ); |
| 1287 | } |
| 1288 | } // void Test4_pipeline () |
| 1289 | |
| 1290 | //! Testing filter::finalize method |
| 1291 | #define BUFFER_SIZE 32 |
| 1292 | #define NUM_BUFFERS 1024 |
| 1293 | |
| 1294 | tbb::atomic<size_t> g_AllocatedCount; // Number of currently allocated buffers |
| 1295 | tbb::atomic<size_t> g_TotalCount; // Total number of allocated buffers |
| 1296 | |
| 1297 | //! Base class for all filters involved in finalize method testing |
| 1298 | class FinalizationBaseFilter : public tbb::filter { |
| 1299 | public: |
| 1300 | FinalizationBaseFilter ( tbb::filter::mode m ) : filter(m) {} |
| 1301 | |
| 1302 | // Deletes buffers if exception occurred |
| 1303 | virtual void finalize( void* item ) __TBB_override { |
| 1304 | size_t* m_Item = (size_t*)item; |
| 1305 | delete[] m_Item; |
| 1306 | --g_AllocatedCount; |
| 1307 | } |
| 1308 | }; |
| 1309 | |
| 1310 | //! Input filter to test finalize method |
| 1311 | class InputFilterWithFinalization: public FinalizationBaseFilter { |
| 1312 | public: |
| 1313 | InputFilterWithFinalization() : FinalizationBaseFilter(tbb::filter::serial) { |
| 1314 | g_TotalCount = 0; |
| 1315 | } |
| 1316 | void* operator()( void* ) __TBB_override { |
| 1317 | if (g_TotalCount == NUM_BUFFERS) |
| 1318 | return NULL; |
| 1319 | size_t* item = new size_t[BUFFER_SIZE]; |
| 1320 | for (int i = 0; i < BUFFER_SIZE; i++) |
| 1321 | item[i] = 1; |
| 1322 | ++g_TotalCount; |
| 1323 | ++g_AllocatedCount; |
| 1324 | return item; |
| 1325 | } |
| 1326 | }; |
| 1327 | |
| 1328 | // The filter multiplies each buffer item by 10. |
| 1329 | class ProcessingFilterWithFinalization : public FinalizationBaseFilter { |
| 1330 | public: |
| 1331 | ProcessingFilterWithFinalization (tbb::filter::mode _mode, bool) : FinalizationBaseFilter (_mode) {} |
| 1332 | |
| 1333 | void* operator()( void* item) __TBB_override { |
| 1334 | if(g_Master == Harness::CurrentTid()) g_MasterExecuted = true; |
| 1335 | else g_NonMasterExecuted = true; |
| 1336 | if( tbb::task::self().is_cancelled()) ++g_TGCCancelled; |
| 1337 | if (g_TotalCount > NUM_BUFFERS / 2) |
| 1338 | ThrowTestException(1); |
| 1339 | size_t* m_Item = (size_t*)item; |
| 1340 | for (int i = 0; i < BUFFER_SIZE; i++) |
| 1341 | m_Item[i] *= 10; |
| 1342 | return item; |
| 1343 | } |
| 1344 | }; |
| 1345 | |
| 1346 | // Output filter deletes previously allocated buffer |
| 1347 | class OutputFilterWithFinalization : public FinalizationBaseFilter { |
| 1348 | public: |
| 1349 | OutputFilterWithFinalization (tbb::filter::mode m) : FinalizationBaseFilter (m) {} |
| 1350 | |
| 1351 | void* operator()( void* item) __TBB_override { |
| 1352 | size_t* m_Item = (size_t*)item; |
| 1353 | delete[] m_Item; |
| 1354 | --g_AllocatedCount; |
| 1355 | return NULL; |
| 1356 | } |
| 1357 | }; |
| 1358 | |
| 1359 | //! Tests filter::finalize method |
| 1360 | void Test5_pipeline ( const FilterSet& filters ) { |
| 1361 | ResetGlobals(); |
| 1362 | g_AllocatedCount = 0; |
| 1363 | CustomPipeline<InputFilterWithFinalization, ProcessingFilterWithFinalization> testPipeline(filters); |
| 1364 | OutputFilterWithFinalization my_output_filter(tbb::filter::parallel); |
| 1365 | |
| 1366 | testPipeline.add_filter(my_output_filter); |
| 1367 | TRY(); |
| 1368 | testPipeline.run(); |
| 1369 | CATCH(); |
| 1370 | ASSERT (g_AllocatedCount == 0, "Memory leak: Some my_object weren't destroyed" ); |
| 1371 | } // void Test5_pipeline () |
| 1372 | |
| 1373 | //! Tests pipeline function passed with different combination of filters |
| 1374 | template<void testFunc(const FilterSet&)> |
| 1375 | void TestWithDifferentFilters() { |
| 1376 | const int NumFilterTypes = 3; |
| 1377 | const tbb::filter::mode modes[NumFilterTypes] = { |
| 1378 | tbb::filter::parallel, |
| 1379 | tbb::filter::serial, |
| 1380 | tbb::filter::serial_out_of_order |
| 1381 | }; |
| 1382 | for ( int i = 0; i < NumFilterTypes; ++i ) { |
| 1383 | for ( int j = 0; j < NumFilterTypes; ++j ) { |
| 1384 | for ( int k = 0; k < 2; ++k ) |
| 1385 | testFunc( FilterSet(modes[i], modes[j], k == 0, k != 0) ); |
| 1386 | } |
| 1387 | } |
| 1388 | } |
| 1389 | |
| 1390 | #endif /* TBB_USE_EXCEPTIONS */ |
| 1391 | |
| 1392 | class FilterToCancel : public tbb::filter { |
| 1393 | public: |
| 1394 | FilterToCancel(bool is_parallel) |
| 1395 | : filter( is_parallel ? tbb::filter::parallel : tbb::filter::serial_in_order ) |
| 1396 | {} |
| 1397 | void* operator()(void* item) __TBB_override { |
| 1398 | ++g_CurExecuted; |
| 1399 | CancellatorTask::WaitUntilReady(); |
| 1400 | return item; |
| 1401 | } |
| 1402 | }; // class FilterToCancel |
| 1403 | |
| 1404 | template <class Filter_to_cancel> |
| 1405 | class PipelineLauncherTask : public tbb::task { |
| 1406 | tbb::task_group_context &my_ctx; |
| 1407 | public: |
| 1408 | PipelineLauncherTask ( tbb::task_group_context& ctx ) : my_ctx(ctx) {} |
| 1409 | |
| 1410 | tbb::task* execute () __TBB_override { |
| 1411 | // Run test when serial filter is the first non-input filter |
| 1412 | InputFilter inputFilter; |
| 1413 | Filter_to_cancel filterToCancel(true); |
| 1414 | tbb::pipeline p; |
| 1415 | p.add_filter(inputFilter); |
| 1416 | p.add_filter(filterToCancel); |
| 1417 | p.run(g_NumTokens, my_ctx); |
| 1418 | return NULL; |
| 1419 | } |
| 1420 | }; |
| 1421 | |
| 1422 | //! Test for cancelling an algorithm from outside (from a task running in parallel with the algorithm). |
| 1423 | void TestCancelation1_pipeline () { |
| 1424 | ResetGlobals(); |
| 1425 | g_ThrowException = false; |
| 1426 | intptr_t threshold = 10; |
| 1427 | tbb::task_group_context ctx; |
| 1428 | ctx.reset(); |
| 1429 | tbb::empty_task &r = *new( tbb::task::allocate_root() ) tbb::empty_task; |
| 1430 | r.set_ref_count(3); |
| 1431 | r.spawn( *new( r.allocate_child() ) CancellatorTask(ctx, threshold) ); |
| 1432 | __TBB_Yield(); |
| 1433 | r.spawn( *new( r.allocate_child() ) PipelineLauncherTask<FilterToCancel>(ctx) ); |
| 1434 | TRY(); |
| 1435 | r.wait_for_all(); |
| 1436 | CATCH_AND_FAIL(); |
| 1437 | r.destroy(r); |
| 1438 | ASSERT( g_TGCCancelled <= g_NumThreads, "Too many tasks survived cancellation" ); |
| 1439 | ASSERT (g_CurExecuted < g_ExecutedAtLastCatch + g_NumThreads, "Too many tasks were executed after cancellation" ); |
| 1440 | } |
| 1441 | |
| 1442 | class FilterToCancel2 : public tbb::filter { |
| 1443 | public: |
| 1444 | FilterToCancel2(bool is_parallel) |
| 1445 | : filter ( is_parallel ? tbb::filter::parallel : tbb::filter::serial_in_order) |
| 1446 | {} |
| 1447 | |
| 1448 | void* operator()(void* item) __TBB_override { |
| 1449 | ++g_CurExecuted; |
| 1450 | Harness::ConcurrencyTracker ct; |
| 1451 | // The test will hang (and be timed out by the test system) if is_cancelled() is broken |
| 1452 | while( !tbb::task::self().is_cancelled() ) |
| 1453 | __TBB_Yield(); |
| 1454 | return item; |
| 1455 | } |
| 1456 | }; |
| 1457 | |
| 1458 | //! Test for cancelling an algorithm from outside (from a task running in parallel with the algorithm). |
| 1459 | /** This version also tests task::is_cancelled() method. **/ |
| 1460 | void TestCancelation2_pipeline () { |
| 1461 | ResetGlobals(); |
| 1462 | RunCancellationTest<PipelineLauncherTask<FilterToCancel2>, CancellatorTask2>(); |
| 1463 | // g_CurExecuted is always >= g_ExecutedAtLastCatch, because the latter is always a snapshot of the |
| 1464 | // former, and g_CurExecuted is monotonic increasing. so the comparison should be at least ==. |
| 1465 | // If another filter is started after cancel but before cancellation is propagated, then the |
| 1466 | // number will be larger. |
| 1467 | ASSERT (g_CurExecuted <= g_ExecutedAtLastCatch, "Some tasks were executed after cancellation" ); |
| 1468 | } |
| 1469 | |
| 1470 | void RunPipelineTests() { |
| 1471 | REMARK( "pipeline tests\n" ); |
| 1472 | tbb::task_scheduler_init init (g_NumThreads); |
| 1473 | g_Master = Harness::CurrentTid(); |
| 1474 | g_NumTokens = 2 * g_NumThreads; |
| 1475 | |
| 1476 | Test0_pipeline(); |
| 1477 | #if TBB_USE_EXCEPTIONS && !__TBB_THROW_ACROSS_MODULE_BOUNDARY_BROKEN |
| 1478 | TestWithDifferentFilters<Test1_pipeline>(); |
| 1479 | TestWithDifferentFilters<Test2_pipeline>(); |
| 1480 | TestWithDifferentFilters<Test3_pipeline>(); |
| 1481 | TestWithDifferentFilters<Test4_pipeline>(); |
| 1482 | TestWithDifferentFilters<Test5_pipeline>(); |
| 1483 | #endif /* TBB_USE_EXCEPTIONS && !__TBB_THROW_ACROSS_MODULE_BOUNDARY_BROKEN */ |
| 1484 | TestCancelation1_pipeline(); |
| 1485 | TestCancelation2_pipeline(); |
| 1486 | } |
| 1487 | |
| 1488 | |
| 1489 | #if TBB_USE_EXCEPTIONS |
| 1490 | |
| 1491 | class MyCapturedException : public tbb::captured_exception { |
| 1492 | public: |
| 1493 | static int m_refCount; |
| 1494 | |
| 1495 | MyCapturedException () : tbb::captured_exception("MyCapturedException" , "test" ) { ++m_refCount; } |
| 1496 | ~MyCapturedException () throw() { --m_refCount; } |
| 1497 | |
| 1498 | MyCapturedException* move () throw() __TBB_override { |
| 1499 | MyCapturedException* movee = (MyCapturedException*)malloc(sizeof(MyCapturedException)); |
| 1500 | return ::new (movee) MyCapturedException; |
| 1501 | } |
| 1502 | void destroy () throw() __TBB_override { |
| 1503 | this->~MyCapturedException(); |
| 1504 | free(this); |
| 1505 | } |
| 1506 | void operator delete ( void* p ) { free(p); } |
| 1507 | }; |
| 1508 | |
| 1509 | int MyCapturedException::m_refCount = 0; |
| 1510 | |
| 1511 | void DeleteTbbException ( volatile tbb::tbb_exception* pe ) { |
| 1512 | delete pe; |
| 1513 | } |
| 1514 | |
| 1515 | void TestTbbExceptionAPI () { |
| 1516 | const char *name = "Test captured exception" , |
| 1517 | *reason = "Unit testing" ; |
| 1518 | tbb::captured_exception e(name, reason); |
| 1519 | ASSERT (strcmp(e.name(), name) == 0, "Setting captured exception name failed" ); |
| 1520 | ASSERT (strcmp(e.what(), reason) == 0, "Setting captured exception reason failed" ); |
| 1521 | tbb::captured_exception c(e); |
| 1522 | ASSERT (strcmp(c.name(), e.name()) == 0, "Copying captured exception name failed" ); |
| 1523 | ASSERT (strcmp(c.what(), e.what()) == 0, "Copying captured exception reason failed" ); |
| 1524 | tbb::captured_exception *m = e.move(); |
| 1525 | ASSERT (strcmp(m->name(), name) == 0, "Moving captured exception name failed" ); |
| 1526 | ASSERT (strcmp(m->what(), reason) == 0, "Moving captured exception reason failed" ); |
| 1527 | ASSERT (!e.name() && !e.what(), "Moving semantics broken" ); |
| 1528 | m->destroy(); |
| 1529 | |
| 1530 | MyCapturedException mce; |
| 1531 | MyCapturedException *mmce = mce.move(); |
| 1532 | ASSERT( MyCapturedException::m_refCount == 2, NULL ); |
| 1533 | DeleteTbbException(mmce); |
| 1534 | ASSERT( MyCapturedException::m_refCount == 1, NULL ); |
| 1535 | } |
| 1536 | |
| 1537 | #endif /* TBB_USE_EXCEPTIONS */ |
| 1538 | |
| 1539 | /** If min and max thread numbers specified on the command line are different, |
| 1540 | the test is run only for 2 sizes of the thread pool (MinThread and MaxThread) |
| 1541 | to be able to test the high and low contention modes while keeping the test reasonably fast **/ |
| 1542 | int TestMain () { |
| 1543 | if(tbb::task_scheduler_init::default_num_threads() == 1) { |
| 1544 | REPORT("Known issue: tests require multiple hardware threads\n" ); |
| 1545 | return Harness::Skipped; |
| 1546 | } |
| 1547 | REMARK ("Using %s\n" , TBB_USE_CAPTURED_EXCEPTION ? "tbb:captured_exception" : "exact exception propagation" ); |
| 1548 | MinThread = min(tbb::task_scheduler_init::default_num_threads(), max(2, MinThread)); |
| 1549 | MaxThread = max(MinThread, min(tbb::task_scheduler_init::default_num_threads(), MaxThread)); |
| 1550 | ASSERT (FLAT_RANGE >= FLAT_GRAIN * MaxThread, "Fix defines" ); |
| 1551 | int step = max((MaxThread - MinThread + 1)/2, 1); |
| 1552 | for ( g_NumThreads = MinThread; g_NumThreads <= MaxThread; g_NumThreads += step ) { |
| 1553 | REMARK ("Number of threads %d\n" , g_NumThreads); |
| 1554 | // Execute in all the possible modes |
| 1555 | for ( size_t j = 0; j < 4; ++j ) { |
| 1556 | g_ExceptionInMaster = (j & 1) != 0; |
| 1557 | g_SolitaryException = (j & 2) != 0; |
| 1558 | REMARK("g_ExceptionInMaster==%s, g_SolitaryException==%s\n" , g_ExceptionInMaster?"T" :"F" , g_SolitaryException?"T" :"F" ); |
| 1559 | RunParForAndReduceTests(); |
| 1560 | RunParDoTests(); |
| 1561 | RunPipelineTests(); |
| 1562 | } |
| 1563 | } |
| 1564 | #if TBB_USE_EXCEPTIONS |
| 1565 | TestTbbExceptionAPI(); |
| 1566 | #endif |
| 1567 | #if __TBB_THROW_ACROSS_MODULE_BOUNDARY_BROKEN |
| 1568 | REPORT("Known issue: exception handling tests are skipped.\n" ); |
| 1569 | #endif |
| 1570 | return Harness::Done; |
| 1571 | } |
| 1572 | |
| 1573 | #else /* !__TBB_TASK_GROUP_CONTEXT */ |
| 1574 | |
| 1575 | int TestMain () { |
| 1576 | return Harness::Skipped; |
| 1577 | } |
| 1578 | |
| 1579 | #endif /* !__TBB_TASK_GROUP_CONTEXT */ |
| 1580 | |