]>
Commit | Line | Data |
---|---|---|
4710dd51 | 1 | /* record_replay.h -*-C++-*- |
2 | * | |
3 | ************************************************************************* | |
4 | * | |
0657c20f | 5 | * Copyright (C) 2012-2016, Intel Corporation |
4710dd51 | 6 | * All rights reserved. |
7 | * | |
4710dd51 | 8 | * Redistribution and use in source and binary forms, with or without |
9 | * modification, are permitted provided that the following conditions | |
10 | * are met: | |
11 | * | |
12 | * * Redistributions of source code must retain the above copyright | |
13 | * notice, this list of conditions and the following disclaimer. | |
14 | * * Redistributions in binary form must reproduce the above copyright | |
15 | * notice, this list of conditions and the following disclaimer in | |
16 | * the documentation and/or other materials provided with the | |
17 | * distribution. | |
18 | * * Neither the name of Intel Corporation nor the names of its | |
19 | * contributors may be used to endorse or promote products derived | |
20 | * from this software without specific prior written permission. | |
21 | * | |
4710dd51 | 22 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
23 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
24 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
25 | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
26 | * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, | |
27 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, | |
28 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS | |
29 | * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED | |
30 | * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
31 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY | |
32 | * WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | |
33 | * POSSIBILITY OF SUCH DAMAGE. | |
0657c20f | 34 | * |
35 | * ********************************************************************* | |
36 | * | |
37 | * PLEASE NOTE: This file is a downstream copy of a file mainitained in | |
38 | * a repository at cilkplus.org. Changes made to this file that are not | |
39 | * submitted through the contribution process detailed at | |
40 | * http://www.cilkplus.org/submit-cilk-contribution will be lost the next | |
41 | * time that a new version is released. Changes only submitted to the | |
42 | * GNU compiler collection or posted to the git repository at | |
43 | * https://bitbucket.org/intelcilkruntime/intel-cilk-runtime.git are | |
44 | * not tracked. | |
45 | * | |
46 | * We welcome your contributions to this open source project. Thank you | |
47 | * for your assistance in helping us improve Cilk Plus. | |
4710dd51 | 48 | * |
49 | **************************************************************************/ | |
50 | ||
51 | /** | |
52 | * @file record-replay.h | |
53 | * | |
54 | * @brief record-replay.h and .cpp encapsulate most of the functionality to | |
55 | * record and play back a Cilk Plus application. | |
56 | * | |
57 | * Recording is directed by the setting of the CILK_RECORD_LOG environment | |
58 | * variable. If it's defined, the value specifies the root we'll use to | |
59 | * generate files for each worker using the following format string: | |
60 | * "%s%d.cilklog", where the integer is the value of w->self. | |
61 | * | |
62 | * Replay is directed by the setting of the CILK_REPLAY_LOG environment | |
63 | * variable, interpreted the same way as CILK_RECORD_LOG. If both | |
64 | * CILK_RECORD_LOG and CILK_REPLAY_LOG are defined, a warning will be given | |
65 | * and the attempt to record a log will be ignored. | |
66 | * | |
67 | * Recording is relatively straightforward. We write all information about a | |
68 | * worker to a per-worker file. | |
69 | * | |
70 | * Each pedigree record consists of the following fields. All fields must be | |
71 | * present in every record to make parsing easy. | |
72 | * - Type - A string identifying the pedigree record. See the PED_TYPE_STR_ | |
73 | * macros for the currently defined values. | |
74 | * - Pedigree - A string of pedigree values, with underscores between | |
75 | * adjacent values. | |
76 | * - i1 - Record type-specific value. -1 if not used. | |
77 | * - i2 - Record type-specific value. -1 if not used. | |
78 | * | |
79 | * WORKERS record - only written to the file for worker 0. Note that this is | |
80 | * the first worker in the workers array. Worker 0 is the first system worker, | |
81 | * *NOT* a user worker. | |
82 | * - Type: "Workers" | |
83 | * - Pedigree: Always "0" - ignored | |
84 | * - i1: Number of workers (g->P) when we recorded the log. A mismatch when | |
85 | * we attempt to replay the log will result in aborting the execution. | |
86 | * - i2: Log version number - Specified by PED_VERSION in record-replay.cpp | |
87 | * | |
88 | * STEAL record - written after a successful steal. | |
89 | * - Type: "Steal" | |
90 | * - Pedigree: Pedigree of stolen frame | |
91 | * - i1: Worker the frame was stolen from | |
92 | * - i2: -1 | |
93 | * | |
94 | * SYNC record - written after a worker continues from a sync. | |
95 | * - Type: "Sync" | |
96 | * - Pedigree: Pedigree of sync. Note that this is the pedigree *before* | |
97 | * the pedigree in incremented in setup_for_execution_pedigree(). | |
98 | * - i1: -1 | |
99 | * - i2: -1 | |
100 | * | |
101 | * ORPHANED record - saved on a return to a stolen parent. | |
102 | * - Type: "Orphaned" | |
103 | * - Pedigree: Pedigree of the parent frame *before* the pedigree is | |
104 | * incremented by the return | |
105 | * - i1: -1 | |
106 | * - i2: -1 | |
107 | * | |
108 | * On replay, the data is loaded into a per-worker array, and the data is | |
109 | * consumed in order as needed. | |
110 | */ | |
111 | ||
112 | #ifndef INCLUDED_RECORD_REPLAY_DOT_H | |
113 | #define INCLUDED_RECORD_REPLAY_DOT_H | |
114 | ||
115 | #include "cilk/common.h" | |
116 | #include "global_state.h" | |
117 | ||
118 | /** | |
119 | * Define CILK_RECORD_REPLAY to enable record/replay functionality. If | |
120 | * CILK_RECORD_REPLAY is not defined, all of the record/replay functions in | |
121 | * record-replay.h will be stubbed out. Since they're declared as inline, | |
122 | * functions, the resulting build should have no performance impact due to | |
123 | * the implementation or record/replay. | |
124 | */ | |
125 | #define CILK_RECORD_REPLAY 1 | |
126 | ||
127 | /** | |
128 | * Define RECORD_ON_REPLAY=1 to write logs when we're replaying a log. This | |
129 | * should only be needed when debugging the replay functionality. This should | |
130 | * always be defined as 0 when record-replay.h is checked in. | |
131 | */ | |
132 | #define RECORD_ON_REPLAY 0 | |
133 | ||
134 | __CILKRTS_BEGIN_EXTERN_C | |
135 | ||
136 | #ifdef CILK_RECORD_REPLAY | |
137 | // Declarations of internal record/replay functions. The inlined versions | |
138 | // further down do some preliminary testing (like if we're not recording or | |
139 | // replaying) and will stub out the functionality if we've compiled out the | |
140 | // record/replay feature | |
141 | int replay_match_sync_pedigree_internal(__cilkrts_worker *w); | |
142 | void replay_wait_for_steal_if_parent_was_stolen_internal(__cilkrts_worker *w); | |
143 | void replay_record_steal_internal(__cilkrts_worker *w, int32_t victim_id); | |
144 | void replay_record_sync_internal(__cilkrts_worker *w); | |
145 | void replay_record_orphaned_internal(__cilkrts_worker *w); | |
146 | int replay_match_victim_pedigree_internal(__cilkrts_worker *w, __cilkrts_worker *victim); | |
147 | void replay_advance_from_sync_internal (__cilkrts_worker *w); | |
148 | int replay_get_next_recorded_victim_internal(__cilkrts_worker *w); | |
149 | #endif // CILK_RECORD_REPLAY | |
150 | ||
151 | // Publically defined record/replay API | |
152 | ||
153 | /** | |
154 | * If we're replaying a log, wait for our parent to be stolen if it was when | |
155 | * the log was recorded. If record/replay is compiled out, this is a noop. | |
156 | * | |
157 | * @param w The __cilkrts_worker we're executing on. The worker's replay | |
158 | * list will be checked for a ORPHANED record with a matching pedigree. If | |
159 | * there is a match, the ORPHANED record will be consumed. | |
160 | */ | |
161 | #ifdef CILK_RECORD_REPLAY | |
162 | __CILKRTS_INLINE | |
163 | void replay_wait_for_steal_if_parent_was_stolen(__cilkrts_worker *w) | |
164 | { | |
165 | // Only check if we're replaying a log | |
166 | if (REPLAY_LOG == w->g->record_or_replay) | |
167 | replay_wait_for_steal_if_parent_was_stolen_internal(w); | |
168 | } | |
169 | #else | |
170 | __CILKRTS_INLINE | |
171 | void replay_wait_for_steal_if_parent_was_stolen(__cilkrts_worker *w) | |
172 | { | |
173 | // If record/replay is disabled, we never wait | |
174 | } | |
175 | #endif // CILK_RECORD_REPLAY | |
176 | ||
177 | /** | |
178 | * Called from random_steal() to override the ID of the randomly chosen victim | |
179 | * worker which this worker will attempt to steal from. Returns the worker id | |
180 | * of the next victim this worker was recorded stealing from, or -1 if the | |
181 | * next record in the log is not a STEAL. | |
182 | * | |
183 | * @note This call does NOT attempt to match the pedigree. That will be done | |
184 | * by replay_match_victim_pedigree() after random_steal() has locked the victim | |
185 | * worker. | |
186 | * | |
187 | * @param w The __cilkrts_worker we're executing on. The worker's replay log | |
188 | * is checked for a STEAL record. If we've got one, the stolen worker ID is | |
189 | * returned. | |
190 | * @param id The randomly chosen victim worker ID. If we're not replaying a | |
191 | * log, or if record/replay has been compiled out, this is the value that | |
192 | * will be returned. | |
193 | * | |
194 | * @return id if we're not replaying a log | |
195 | * @return -1 if the next record is not a STEAL | |
196 | * @return recorded stolen worker ID if we've got a matching STEAL record | |
197 | */ | |
198 | #ifdef CILK_RECORD_REPLAY | |
199 | __CILKRTS_INLINE | |
200 | int replay_get_next_recorded_victim(__cilkrts_worker *w, int id) | |
201 | { | |
202 | // Only check if we're replaying a log | |
203 | if (REPLAY_LOG == w->g->record_or_replay) | |
204 | return replay_get_next_recorded_victim_internal(w); | |
205 | else | |
206 | return id; | |
207 | } | |
208 | #else | |
209 | __CILKRTS_INLINE | |
210 | int replay_get_next_recorded_victim(__cilkrts_worker *w, int id) | |
211 | { | |
212 | // Record/replay is disabled. Always return the original worker id | |
213 | return id; | |
214 | } | |
215 | #endif // CILK_RECORD_REPLAY | |
216 | ||
217 | /** | |
218 | * Initialize per-worker data for record/replay. A noop if record/replay | |
219 | * is disabled, or if we're not recording or replaying anything. | |
220 | * | |
221 | * If we're recording a log, this will ready us to create the per-worker | |
222 | * logs. | |
223 | * | |
224 | * If we're replaying a log, this will read the logs into the per-worker | |
225 | * structures. | |
226 | * | |
227 | * @param g Cilk runtime global state | |
228 | */ | |
229 | void replay_init_workers(global_state_t *g); | |
230 | ||
231 | /** | |
232 | * Record a record on a successful steal. A noop if record/replay is | |
233 | * diabled, or if we're not recording anything | |
234 | * | |
235 | * @param w The __cilkrts_worker we're executing on. The pedigree of | |
236 | * the stolen frame will be walked to generate the STEAL record. | |
237 | * | |
238 | * @param victim_id The worker ID of the worker w stole from. | |
239 | */ | |
240 | #ifdef CILK_RECORD_REPLAY | |
241 | __CILKRTS_INLINE | |
242 | void replay_record_steal(__cilkrts_worker *w, int32_t victim_id) | |
243 | { | |
244 | #if RECORD_ON_REPLAY | |
245 | // If we're recording on replay, write the record if we're recording or | |
246 | // replaying | |
247 | if (RECORD_REPLAY_NONE == w->g->record_or_replay) | |
248 | return; | |
249 | #else | |
250 | // Only write the record if we're recording | |
251 | if (RECORD_LOG != w->g->record_or_replay) | |
252 | return; | |
253 | #endif | |
254 | ||
255 | replay_record_steal_internal(w, victim_id); | |
256 | } | |
257 | #else | |
258 | __CILKRTS_INLINE | |
259 | void replay_record_steal(__cilkrts_worker *w, int32_t victim_id) | |
260 | { | |
261 | } | |
262 | #endif // CILK_RECORD_REPLAY | |
263 | ||
264 | /** | |
265 | * Record a record when continuing after a sync. A noop if record/replay is | |
266 | * diabled, or if we're not recording anything, or if the sync was abandoned, | |
267 | * meaning this isn't the worker that continues from the sync. | |
268 | * | |
269 | * @param w The __cilkrts_worker for we're executing on. The pedigree of | |
270 | * the sync-ing frame will be walked to generate the SYNC record. | |
271 | * | |
272 | * @param continuing True if this worker will be continuing from the | |
273 | * cilk_sync. A SYNC record will only be generated if continuing is true. | |
274 | */ | |
275 | #ifdef CILK_RECORD_REPLAY | |
276 | __CILKRTS_INLINE | |
277 | void replay_record_sync(__cilkrts_worker *w, int continuing) | |
278 | { | |
279 | // If this was not the last worker to the syn, return | |
280 | if (! continuing) | |
281 | return; | |
282 | ||
283 | #if RECORD_ON_REPLAY | |
284 | // If we're recording on replay, write the record if we're recording or | |
285 | // replaying | |
286 | if (RECORD_REPLAY_NONE == w->g->record_or_replay) | |
287 | return; | |
288 | #else | |
289 | // Only write the record if we're recording | |
290 | if (RECORD_LOG != w->g->record_or_replay) | |
291 | return; | |
292 | #endif | |
293 | ||
294 | replay_record_sync_internal(w); | |
295 | } | |
296 | #else | |
297 | __CILKRTS_INLINE | |
298 | void replay_record_sync(__cilkrts_worker *w, int abandoned) | |
299 | { | |
300 | } | |
301 | #endif // CILK_RECORD_REPLAY | |
302 | ||
303 | /** | |
304 | * Record a record on a return to a stolen parent. A noop if record/replay is | |
305 | * diabled, or if we're not recording anything. | |
306 | * | |
307 | * @param w The __cilkrts_worker for we're executing on. The pedigree of the | |
308 | * frame that has discovered that its parent has been stolken will be walked | |
309 | * to generate the ORPHANED record. | |
310 | */ | |
311 | #ifdef CILK_RECORD_REPLAY | |
312 | __CILKRTS_INLINE | |
313 | void replay_record_orphaned(__cilkrts_worker *w) | |
314 | { | |
315 | #if RECORD_ON_REPLAY | |
316 | // If we're recording on replay, write the record if we're recording or | |
317 | // replaying | |
318 | if (RECORD_REPLAY_NONE == w->g->record_or_replay) | |
319 | return; | |
320 | #else | |
321 | // Only write the record if we're recording | |
322 | if (RECORD_LOG != w->g->record_or_replay) | |
323 | return; | |
324 | #endif | |
325 | ||
326 | replay_record_orphaned_internal(w); | |
327 | } | |
328 | #else | |
329 | __CILKRTS_INLINE | |
330 | void replay_record_orphaned(__cilkrts_worker *w) | |
331 | { | |
332 | } | |
333 | #endif // CILK_RECORD_REPLAY | |
334 | ||
335 | /** | |
336 | * Test whether the frame at the head of the victim matches the pedigree of | |
337 | * the frame that was recorded being stolen. Called in random steal to verify | |
338 | * that we're about to steal the correct frame. | |
339 | * | |
340 | * @param w The __cilkrts_worker for we're executing on. The current worker | |
341 | * is needed to find the replay entry to be checked. | |
342 | * | |
343 | * @param victim The __cilkrts_worker for we're proposing to steal a frame | |
344 | * from. The victim's head entry is | |
345 | * is needed to find the replay entry to be checked. | |
346 | * | |
347 | * @return 0 if we're replaying a log and the victim's pedigree does NOT match | |
348 | * the next frame the worker is expected to steal. | |
349 | * | |
350 | * @return 1 in all other cases to indicate that the steal attempt should | |
351 | * continue | |
352 | */ | |
353 | #ifdef CILK_RECORD_REPLAY | |
354 | __CILKRTS_INLINE | |
355 | int replay_match_victim_pedigree(__cilkrts_worker *w, __cilkrts_worker *victim) | |
356 | { | |
357 | // We're not replaying a log. The victim is always acceptable | |
358 | if (REPLAY_LOG != w->g->record_or_replay) | |
359 | return 1; | |
360 | ||
361 | // Return 1 if the victim's pedigree matches the frame the worker stole | |
362 | // when we recorded the log | |
363 | return replay_match_victim_pedigree_internal(w, victim); | |
364 | } | |
365 | #else | |
366 | __CILKRTS_INLINE | |
367 | int replay_match_victim_pedigree(__cilkrts_worker *w, __cilkrts_worker *victim) | |
368 | { | |
369 | // Record/replay is disabled. The victim is always acceptable | |
370 | return 1; | |
371 | } | |
372 | #endif // CILK_RECORD_REPLAY | |
373 | ||
374 | /** | |
375 | * Test whether the current replay entry is a sync record matching the | |
376 | * worker's pedigree. | |
377 | * | |
378 | * @param w The __cilkrts_worker for we're executing on. | |
379 | * | |
380 | * @return 1 if the current replay entry matches the current pedigree. | |
381 | * @return 0 if there's no match, or if we're not replaying a log. | |
382 | */ | |
383 | #ifdef CILK_RECORD_REPLAY | |
384 | __CILKRTS_INLINE | |
385 | int replay_match_sync_pedigree(__cilkrts_worker *w) | |
386 | { | |
387 | // If we're not replaying, assume no match | |
388 | if (REPLAY_LOG != w->g->record_or_replay) | |
389 | return 0; | |
390 | ||
391 | return replay_match_sync_pedigree_internal(w); | |
392 | } | |
393 | #else | |
394 | __CILKRTS_INLINE | |
395 | int replay_match_sync_pedigree(__cilkrts_worker *w) | |
396 | { | |
397 | // Record/replay is disabled. Assume no match | |
398 | return 0; | |
399 | } | |
400 | #endif | |
401 | ||
402 | /** | |
403 | * Marks a sync record seen, advancing to the next record in the replay list. | |
404 | * | |
405 | * This function will only advance to the next record if: | |
406 | * - Record/replay hasn't been compiled out AND | |
407 | * - We're replaying a log AND | |
408 | * - A match was found AND | |
409 | * - The sync is not being abandoned | |
410 | * | |
411 | * @param w The __cilkrts_worker for we're executing on. | |
412 | * @param match_found The value returned by replay_match_sync_pedigree(). If | |
413 | * match_found is false, nothing is done. | |
414 | * @param continuing Flag indicating whether this worker will continue from | |
415 | * the sync (it's the last worker to the sync) or if it will abandon the work | |
416 | * and go to the scheduling loop to look for more work it can steal. | |
417 | */ | |
418 | #ifdef CILK_RECORD_REPLAY | |
419 | __CILKRTS_INLINE | |
420 | void replay_advance_from_sync(__cilkrts_worker *w, int match_found, int continuing) | |
421 | { | |
422 | // If we're replaying a log, and the current sync wasn't abandoned, and we | |
423 | // found a match in the log, mark the sync record seen. | |
424 | if ((REPLAY_LOG == w->g->record_or_replay) && match_found && continuing) | |
425 | replay_advance_from_sync_internal(w); | |
426 | } | |
427 | #else | |
428 | __CILKRTS_INLINE | |
429 | void replay_advance_from_sync(__cilkrts_worker *w, int match_found, int continuing) | |
430 | { | |
431 | } | |
432 | #endif | |
433 | ||
434 | /** | |
435 | * Release any resources used to read or write a replay log. | |
436 | * | |
437 | * @param g Cilk runtime global state | |
438 | */ | |
439 | void replay_term(global_state_t *g); | |
440 | ||
441 | __CILKRTS_END_EXTERN_C | |
442 | ||
443 | #endif // ! defined(INCLUDED_RECORD_REPLAY_DOT_H) |