engine_marktasks.c 18.6 KB
Newer Older
Pedro Gonnet's avatar
Pedro Gonnet committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
/*******************************************************************************
 * This file is part of SWIFT.
 * Copyright (c) 2012 Pedro Gonnet (pedro.gonnet@durham.ac.uk)
 *                    Matthieu Schaller (matthieu.schaller@durham.ac.uk)
 *               2015 Peter W. Draper (p.w.draper@durham.ac.uk)
 *                    Angus Lepper (angus.lepper@ed.ac.uk)
 *               2016 John A. Regan (john.a.regan@durham.ac.uk)
 *                    Tom Theuns (tom.theuns@durham.ac.uk)
 *
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published
 * by the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 *
 ******************************************************************************/

/* Config parameters. */
#include "../config.h"

/* Some standard headers. */
#include <stdlib.h>
#include <unistd.h>

/* MPI headers. */
#ifdef WITH_MPI
#include <mpi.h>
#endif

/* Load the profiler header, if needed. */
#ifdef WITH_PROFILER
#include <gperftools/profiler.h>
#endif

/* This object's header. */
#include "engine.h"

/* Local headers. */
#include "active.h"
#include "atomic.h"
#include "cell.h"
#include "clocks.h"
#include "cycle.h"
#include "debug.h"
#include "error.h"
#include "proxy.h"
#include "timers.h"

/**
 * @brief Mark tasks to be un-skipped and set the sort flags accordingly.
 *        Threadpool mapper function.
 *
 * @param map_data pointer to the tasks
 * @param num_elements number of tasks
 * @param extra_data pointer to int that will define if a rebuild is needed.
 */
void engine_marktasks_mapper(void *map_data, int num_elements,
                             void *extra_data) {
  /* Unpack the arguments. */
  struct task *tasks = (struct task *)map_data;
  size_t *rebuild_space = &((size_t *)extra_data)[1];
  struct scheduler *s = (struct scheduler *)(((size_t *)extra_data)[2]);
  struct engine *e = (struct engine *)((size_t *)extra_data)[0];
  const int nodeID = e->nodeID;

  for (int ind = 0; ind < num_elements; ind++) {

    /* Get basic task information */
    struct task *t = &tasks[ind];
    const enum task_types t_type = t->type;
    const enum task_subtypes t_subtype = t->subtype;

    /* Single-cell task? */
    if (t_type == task_type_self || t_type == task_type_sub_self) {

      /* Local pointer. */
      struct cell *ci = t->ci;

      if (ci->nodeID != engine_rank) error("Non-local self task found");

      /* Activate the hydro drift */
      if (t_type == task_type_self && t_subtype == task_subtype_density) {
        if (cell_is_active_hydro(ci, e)) {
          scheduler_activate(s, t);
          cell_activate_drift_part(ci, s);
        }
      }

      /* Store current values of dx_max and h_max. */
      else if (t_type == task_type_sub_self &&
               t_subtype == task_subtype_density) {
        if (cell_is_active_hydro(ci, e)) {
          scheduler_activate(s, t);
          cell_activate_subcell_hydro_tasks(ci, NULL, s);
        }
      }

      else if (t_type == task_type_self && t_subtype == task_subtype_force) {
        if (cell_is_active_hydro(ci, e)) scheduler_activate(s, t);
      }

      else if (t_type == task_type_sub_self &&
               t_subtype == task_subtype_force) {
        if (cell_is_active_hydro(ci, e)) scheduler_activate(s, t);
      }

#ifdef EXTRA_HYDRO_LOOP
      else if (t_type == task_type_self && t_subtype == task_subtype_gradient) {
        if (cell_is_active_hydro(ci, e)) scheduler_activate(s, t);
      }

      else if (t_type == task_type_sub_self &&
               t_subtype == task_subtype_gradient) {
        if (cell_is_active_hydro(ci, e)) scheduler_activate(s, t);
      }
#endif

      /* Activate the star density */
      else if (t_type == task_type_self &&
               t_subtype == task_subtype_stars_density) {
        if (cell_is_active_stars(ci, e)) {
          scheduler_activate(s, t);
          cell_activate_drift_part(ci, s);
          cell_activate_drift_gpart(ci, s);
        }
      }

      /* Store current values of dx_max and h_max. */
      else if (t_type == task_type_sub_self &&
               t_subtype == task_subtype_stars_density) {
        if (cell_is_active_stars(ci, e)) {
          scheduler_activate(s, t);
          cell_activate_subcell_stars_tasks(ci, NULL, s);
        }
      }

      /* Activate the gravity drift */
      else if (t_type == task_type_self && t_subtype == task_subtype_grav) {
        if (cell_is_active_gravity(ci, e)) {
          scheduler_activate(s, t);
          cell_activate_subcell_grav_tasks(t->ci, NULL, s);
        }
      }

      /* Activate the gravity drift */
      else if (t_type == task_type_self &&
               t_subtype == task_subtype_external_grav) {
        if (cell_is_active_gravity(ci, e)) {
          scheduler_activate(s, t);
          cell_activate_drift_gpart(t->ci, s);
        }
      }

#ifdef SWIFT_DEBUG_CHECKS
      else {
        error("Invalid task type / sub-type encountered");
      }
#endif
    }

    /* Pair? */
    else if (t_type == task_type_pair || t_type == task_type_sub_pair) {

      /* Local pointers. */
      struct cell *ci = t->ci;
      struct cell *cj = t->cj;
#ifdef WITH_MPI
      const int ci_nodeID = ci->nodeID;
      const int cj_nodeID = cj->nodeID;
#else
      const int ci_nodeID = nodeID;
      const int cj_nodeID = nodeID;
#endif
      const int ci_active_hydro = cell_is_active_hydro(ci, e);
      const int cj_active_hydro = cell_is_active_hydro(cj, e);
      const int ci_active_gravity = cell_is_active_gravity(ci, e);
      const int cj_active_gravity = cell_is_active_gravity(cj, e);
      const int ci_active_stars = cell_is_active_stars(ci, e);
      const int cj_active_stars = cell_is_active_stars(cj, e);

      /* Only activate tasks that involve a local active cell. */
      if ((t_subtype == task_subtype_density ||
           t_subtype == task_subtype_gradient ||
           t_subtype == task_subtype_force) &&
          ((ci_active_hydro && ci_nodeID == nodeID) ||
           (cj_active_hydro && cj_nodeID == nodeID))) {

        scheduler_activate(s, t);

        /* Set the correct sorting flags */
        if (t_type == task_type_pair &&
            (t_subtype == task_subtype_density ||
             t_subtype == task_subtype_stars_density)) {

          /* Store some values. */
          atomic_or(&ci->hydro.requires_sorts, 1 << t->flags);
          atomic_or(&cj->hydro.requires_sorts, 1 << t->flags);
          ci->hydro.dx_max_sort_old = ci->hydro.dx_max_sort;
          cj->hydro.dx_max_sort_old = cj->hydro.dx_max_sort;

          /* Activate the hydro drift tasks. */
          if (ci_nodeID == nodeID) cell_activate_drift_part(ci, s);
          if (cj_nodeID == nodeID) cell_activate_drift_part(cj, s);

          /* Check the sorts and activate them if needed. */
          cell_activate_sorts(ci, t->flags, s);
          cell_activate_sorts(cj, t->flags, s);

        }

        /* Store current values of dx_max and h_max. */
        else if (t_type == task_type_sub_pair &&
                 (t_subtype == task_subtype_density ||
                  t_subtype == task_subtype_stars_density)) {
          cell_activate_subcell_hydro_tasks(t->ci, t->cj, s);
        }
      }

      /* Stars */
      if (t_subtype == task_subtype_stars_density &&
          ((ci_active_stars && ci->nodeID == engine_rank) ||
           (cj_active_stars && cj->nodeID == engine_rank))) {

        scheduler_activate(s, t);

        /* Set the correct sorting flags */
        if (t_type == task_type_pair) {

          /* Store some values. */
          atomic_or(&ci->hydro.requires_sorts, 1 << t->flags);
          atomic_or(&cj->hydro.requires_sorts, 1 << t->flags);
          ci->hydro.dx_max_sort_old = ci->hydro.dx_max_sort;
          cj->hydro.dx_max_sort_old = cj->hydro.dx_max_sort;

          /* Activate the hydro drift tasks. */
          if (ci_nodeID == nodeID) {
            cell_activate_drift_part(ci, s);
            cell_activate_drift_gpart(ci, s);
          }
          if (cj_nodeID == nodeID) {
            cell_activate_drift_part(cj, s);
            cell_activate_drift_gpart(cj, s);
          }

          /* Check the sorts and activate them if needed. */
          cell_activate_sorts(ci, t->flags, s);
          cell_activate_sorts(cj, t->flags, s);

        }

        /* Store current values of dx_max and h_max. */
        else if (t_type == task_type_sub_pair) {
          cell_activate_subcell_stars_tasks(t->ci, t->cj, s);
        }
      }

      if ((t_subtype == task_subtype_grav) &&
          ((ci_active_gravity && ci_nodeID == nodeID) ||
           (cj_active_gravity && cj_nodeID == nodeID))) {

        scheduler_activate(s, t);

        if (t_type == task_type_pair && t_subtype == task_subtype_grav) {
          /* Activate the gravity drift */
          cell_activate_subcell_grav_tasks(t->ci, t->cj, s);
        }

#ifdef SWIFT_DEBUG_CHECKS
        else if (t_type == task_type_sub_pair &&
                 t_subtype == task_subtype_grav) {
          error("Invalid task sub-type encountered");
        }
#endif
      }

      /* Only interested in density tasks as of here. */
      if (t_subtype == task_subtype_density) {

        /* Too much particle movement? */
        if (cell_need_rebuild_for_pair(ci, cj)) *rebuild_space = 1;

#ifdef WITH_MPI
        /* Activate the send/recv tasks. */
        if (ci_nodeID != nodeID) {

          /* If the local cell is active, receive data from the foreign cell. */
          if (cj_active_hydro) {
            scheduler_activate(s, ci->mpi.hydro.recv_xv);
            if (ci_active_hydro) {
              scheduler_activate(s, ci->mpi.hydro.recv_rho);
#ifdef EXTRA_HYDRO_LOOP
              scheduler_activate(s, ci->mpi.hydro.recv_gradient);
#endif
            }
          }

          /* If the foreign cell is active, we want its ti_end values. */
          if (ci_active_hydro) scheduler_activate(s, ci->mpi.recv_ti);

          /* Is the foreign cell active and will need stuff from us? */
          if (ci_active_hydro) {

            struct link *l =
                scheduler_activate_send(s, cj->mpi.hydro.send_xv, ci_nodeID);

            /* Drift the cell which will be sent at the level at which it is
               sent, i.e. drift the cell specified in the send task (l->t)
               itself. */
            cell_activate_drift_part(l->t->ci, s);

            /* If the local cell is also active, more stuff will be needed. */
            if (cj_active_hydro) {
              scheduler_activate_send(s, cj->mpi.hydro.send_rho, ci_nodeID);

#ifdef EXTRA_HYDRO_LOOP
              scheduler_activate_send(s, cj->mpi.hydro.send_gradient,
                                      ci_nodeID);
#endif
            }
          }

          /* If the local cell is active, send its ti_end values. */
          if (cj_active_hydro)
            scheduler_activate_send(s, cj->mpi.send_ti, ci_nodeID);

        } else if (cj_nodeID != nodeID) {

          /* If the local cell is active, receive data from the foreign cell. */
          if (ci_active_hydro) {
            scheduler_activate(s, cj->mpi.hydro.recv_xv);
            if (cj_active_hydro) {
              scheduler_activate(s, cj->mpi.hydro.recv_rho);
#ifdef EXTRA_HYDRO_LOOP
              scheduler_activate(s, cj->mpi.hydro.recv_gradient);
#endif
            }
          }

          /* If the foreign cell is active, we want its ti_end values. */
          if (cj_active_hydro) scheduler_activate(s, cj->mpi.recv_ti);

          /* Is the foreign cell active and will need stuff from us? */
          if (cj_active_hydro) {

            struct link *l =
                scheduler_activate_send(s, ci->mpi.hydro.send_xv, cj_nodeID);

            /* Drift the cell which will be sent at the level at which it is
               sent, i.e. drift the cell specified in the send task (l->t)
               itself. */
            cell_activate_drift_part(l->t->ci, s);

            /* If the local cell is also active, more stuff will be needed. */
            if (ci_active_hydro) {

              scheduler_activate_send(s, ci->mpi.hydro.send_rho, cj_nodeID);

#ifdef EXTRA_HYDRO_LOOP
              scheduler_activate_send(s, ci->mpi.hydro.send_gradient,
                                      cj_nodeID);
#endif
            }
          }

          /* If the local cell is active, send its ti_end values. */
          if (ci_active_hydro)
            scheduler_activate_send(s, ci->mpi.send_ti, cj_nodeID);
        }
#endif
      }

      /* Only interested in stars_density tasks as of here. */
      if (t->subtype == task_subtype_stars_density) {

        /* Too much particle movement? */
        if (cell_need_rebuild_for_pair(ci, cj)) *rebuild_space = 1;

        // LOIC: Need implementing MPI case
      }

      /* Only interested in gravity tasks as of here. */
      if (t_subtype == task_subtype_grav) {

#ifdef WITH_MPI
        /* Activate the send/recv tasks. */
        if (ci_nodeID != nodeID) {

          /* If the local cell is active, receive data from the foreign cell. */
          if (cj_active_gravity) scheduler_activate(s, ci->mpi.grav.recv);

          /* If the foreign cell is active, we want its ti_end values. */
          if (ci_active_gravity) scheduler_activate(s, ci->mpi.recv_ti);

          /* Is the foreign cell active and will need stuff from us? */
          if (ci_active_gravity) {

            struct link *l =
                scheduler_activate_send(s, cj->mpi.grav.send, ci_nodeID);

            /* Drift the cell which will be sent at the level at which it is
               sent, i.e. drift the cell specified in the send task (l->t)
               itself. */
            cell_activate_drift_gpart(l->t->ci, s);
          }

          /* If the local cell is active, send its ti_end values. */
          if (cj_active_gravity)
            scheduler_activate_send(s, cj->mpi.send_ti, ci_nodeID);

        } else if (cj_nodeID != nodeID) {

          /* If the local cell is active, receive data from the foreign cell. */
          if (ci_active_gravity) scheduler_activate(s, cj->mpi.grav.recv);

          /* If the foreign cell is active, we want its ti_end values. */
          if (cj_active_gravity) scheduler_activate(s, cj->mpi.recv_ti);

          /* Is the foreign cell active and will need stuff from us? */
          if (cj_active_gravity) {

            struct link *l =
                scheduler_activate_send(s, ci->mpi.grav.send, cj_nodeID);

            /* Drift the cell which will be sent at the level at which it is
               sent, i.e. drift the cell specified in the send task (l->t)
               itself. */
            cell_activate_drift_gpart(l->t->ci, s);
          }

          /* If the local cell is active, send its ti_end values. */
          if (ci_active_gravity)
            scheduler_activate_send(s, ci->mpi.send_ti, cj_nodeID);
        }
#endif
      }
    }

    /* End force ? */
446
447
    else if (t_type == task_type_end_force ||
             t_type == task_type_end_force_in) {
Pedro Gonnet's avatar
Pedro Gonnet committed
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553

      if (cell_is_active_hydro(t->ci, e) || cell_is_active_gravity(t->ci, e))
        scheduler_activate(s, t);
    }

    /* Kick ? */
    else if (t_type == task_type_kick1 || t_type == task_type_kick2) {

      if (cell_is_active_hydro(t->ci, e) || cell_is_active_gravity(t->ci, e))
        scheduler_activate(s, t);
    }

    /* Hydro ghost tasks ? */
    else if (t_type == task_type_ghost || t_type == task_type_extra_ghost ||
             t_type == task_type_ghost_in || t_type == task_type_ghost_out) {
      if (cell_is_active_hydro(t->ci, e)) scheduler_activate(s, t);
    }

    /* logger tasks ? */
    else if (t->type == task_type_logger) {
      if (cell_is_active_hydro(t->ci, e) || cell_is_active_gravity(t->ci, e) ||
          cell_is_active_stars(t->ci, e))
        scheduler_activate(s, t);
    }

    /* Gravity stuff ? */
    else if (t_type == task_type_grav_down || t_type == task_type_grav_mesh ||
             t_type == task_type_grav_long_range ||
             t_type == task_type_init_grav ||
             t_type == task_type_init_grav_out ||
             t_type == task_type_grav_down_in) {
      if (cell_is_active_gravity(t->ci, e)) scheduler_activate(s, t);
    }

    /* Multipole - Multipole interaction task */
    else if (t_type == task_type_grav_mm) {

      /* Local pointers. */
      const struct cell *ci = t->ci;
      const struct cell *cj = t->cj;
#ifdef WITH_MPI
      const int ci_nodeID = ci->nodeID;
      const int cj_nodeID = (cj != NULL) ? cj->nodeID : -1;
#else
      const int ci_nodeID = nodeID;
      const int cj_nodeID = nodeID;
#endif
      const int ci_active_gravity = cell_is_active_gravity_mm(ci, e);
      const int cj_active_gravity = cell_is_active_gravity_mm(cj, e);

      if ((ci_active_gravity && ci_nodeID == nodeID) ||
          (cj_active_gravity && cj_nodeID == nodeID))
        scheduler_activate(s, t);
    }

    /* Star ghost tasks ? */
    else if (t_type == task_type_stars_ghost ||
             t_type == task_type_stars_ghost_in ||
             t_type == task_type_stars_ghost_out) {
      if (cell_is_active_stars(t->ci, e)) scheduler_activate(s, t);
    }

    /* Time-step? */
    else if (t_type == task_type_timestep) {
      t->ci->hydro.updated = 0;
      t->ci->grav.updated = 0;
      t->ci->stars.updated = 0;
      if (cell_is_active_hydro(t->ci, e) || cell_is_active_gravity(t->ci, e))
        scheduler_activate(s, t);
    }

    /* Subgrid tasks */
    else if (t_type == task_type_cooling) {
      if (cell_is_active_hydro(t->ci, e) || cell_is_active_gravity(t->ci, e))
        scheduler_activate(s, t);
    } else if (t_type == task_type_star_formation) {
      if (cell_is_active_hydro(t->ci, e) || cell_is_active_gravity(t->ci, e))
        scheduler_activate(s, t);
    }
  }
}

/**
 * @brief Mark tasks to be un-skipped and set the sort flags accordingly.
 *
 * @return 1 if the space has to be rebuilt, 0 otherwise.
 */
int engine_marktasks(struct engine *e) {

  struct scheduler *s = &e->sched;
  const ticks tic = getticks();
  int rebuild_space = 0;

  /* Run through the tasks and mark as skip or not. */
  size_t extra_data[3] = {(size_t)e, (size_t)rebuild_space, (size_t)&e->sched};
  threadpool_map(&e->threadpool, engine_marktasks_mapper, s->tasks, s->nr_tasks,
                 sizeof(struct task), 0, extra_data);
  rebuild_space = extra_data[1];

  if (e->verbose)
    message("took %.3f %s.", clocks_from_ticks(getticks() - tic),
            clocks_getunit());

  /* All is well... */
  return rebuild_space;
}