engine.c 167 KB
Newer Older
Pedro Gonnet's avatar
Pedro Gonnet committed
1
/*******************************************************************************
2
 * This file is part of SWIFT.
3
 * Copyright (c) 2012 Pedro Gonnet (pedro.gonnet@durham.ac.uk)
4
 *                    Matthieu Schaller (matthieu.schaller@durham.ac.uk)
Peter W. Draper's avatar
Peter W. Draper committed
5
 *               2015 Peter W. Draper (p.w.draper@durham.ac.uk)
6
7
8
 *                    Angus Lepper (angus.lepper@ed.ac.uk)
 *               2016 John A. Regan (john.a.regan@durham.ac.uk)
 *                    Tom Theuns (tom.theuns@durham.ac.uk)
9
 *
Pedro Gonnet's avatar
Pedro Gonnet committed
10
11
12
13
 * 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.
14
 *
Pedro Gonnet's avatar
Pedro Gonnet committed
15
16
17
18
 * 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.
19
 *
Pedro Gonnet's avatar
Pedro Gonnet committed
20
21
 * 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/>.
22
 *
Pedro Gonnet's avatar
Pedro Gonnet committed
23
24
25
26
27
28
29
30
31
 ******************************************************************************/

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

/* Some standard headers. */
#include <float.h>
#include <limits.h>
#include <sched.h>
32
#include <stdbool.h>
33
34
35
36
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
Pedro Gonnet's avatar
Pedro Gonnet committed
37

38
39
/* MPI headers. */
#ifdef WITH_MPI
40
#include <mpi.h>
41
42
#endif

Angus Lepper's avatar
Angus Lepper committed
43
44
#ifdef HAVE_LIBNUMA
#include <numa.h>
45
46
#endif

47
/* This object's header. */
Pedro Gonnet's avatar
Pedro Gonnet committed
48
#include "engine.h"
49
50

/* Local headers. */
51
#include "active.h"
52
#include "atomic.h"
53
#include "cell.h"
54
#include "clocks.h"
55
56
#include "cycle.h"
#include "debug.h"
57
#include "error.h"
58
#include "gravity.h"
59
#include "hydro.h"
60
#include "map.h"
61
#include "minmax.h"
62
#include "parallel_io.h"
63
#include "part.h"
64
#include "partition.h"
James Willis's avatar
James Willis committed
65
#include "profiler.h"
66
67
#include "proxy.h"
#include "runner.h"
68
69
#include "serial_io.h"
#include "single_io.h"
70
#include "sort_part.h"
71
#include "statistics.h"
72
#include "timers.h"
73
#include "tools.h"
74
#include "units.h"
75
#include "version.h"
Pedro Gonnet's avatar
Pedro Gonnet committed
76

77
78
79
/* Particle cache size. */
#define CACHE_SIZE 512

80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
const char *engine_policy_names[] = {"none",
                                     "rand",
                                     "steal",
                                     "keep",
                                     "block",
                                     "cpu_tight",
                                     "mpi",
                                     "numa_affinity",
                                     "hydro",
                                     "self_gravity",
                                     "external_gravity",
                                     "cosmology_integration",
                                     "drift_all",
                                     "reconstruct_mpoles",
                                     "cooling",
                                     "sourceterms",
                                     "stars"};
Pedro Gonnet's avatar
Pedro Gonnet committed
97

98
99
100
/** The rank of the engine as a global variable (for messages). */
int engine_rank;

101
102
103
104
/**
 * @brief Link a density/force task to a cell.
 *
 * @param e The #engine.
105
 * @param l A pointer to the #link, will be modified atomically.
106
107
108
109
 * @param t The #task.
 *
 * @return The new #link pointer.
 */
110
void engine_addlink(struct engine *e, struct link **l, struct task *t) {
111

112
  /* Get the next free link. */
113
114
115
116
117
  const int ind = atomic_inc(&e->nr_links);
  if (ind >= e->size_links) {
    error("Link table overflow.");
  }
  struct link *res = &e->links[ind];
118

119
  /* Set it atomically. */
120
  res->t = t;
121
  res->next = atomic_swap(l, res);
122
}
123

124
125
126
127
128
/**
 * @brief Recursively add non-implicit ghost tasks to a cell hierarchy.
 */
void engine_add_ghosts(struct engine *e, struct cell *c, struct task *ghost_in,
                       struct task *ghost_out) {
129
130
131
132
133

  /* Break the recursion if we have no part to play with */
  if (c->count == 0) return;

  /* If we have reached the leaf OR have to few particles to play with*/
134
  if (!c->split || c->count < engine_max_parts_per_ghost) {
135
136

    /* Add the ghost task and its dependencies */
137
138
139
140
141
142
    struct scheduler *s = &e->sched;
    c->ghost =
        scheduler_addtask(s, task_type_ghost, task_subtype_none, 0, 0, c, NULL);
    scheduler_addunlock(s, ghost_in, c->ghost);
    scheduler_addunlock(s, c->ghost, ghost_out);
  } else {
143
    /* Keep recursing */
144
145
146
147
148
149
    for (int k = 0; k < 8; k++)
      if (c->progeny[k] != NULL)
        engine_add_ghosts(e, c->progeny[k], ghost_in, ghost_out);
  }
}

150
151
152
153
154
155
/**
 * @brief Generate the hydro hierarchical tasks for a hierarchy of cells -
 * i.e. all the O(Npart) tasks.
 *
 * Tasks are only created here. The dependencies will be added later on.
 *
156
157
 * Note that there is no need to recurse below the super-cell. Note also
 * that we only add tasks if the relevant particles are present in the cell.
158
 *
159
160
161
 * @param e The #engine.
 * @param c The #cell.
 */
162
void engine_make_hierarchical_tasks(struct engine *e, struct cell *c) {
163
164

  struct scheduler *s = &e->sched;
165
  const int periodic = e->s->periodic;
166
  const int is_with_hydro = (e->policy & engine_policy_hydro);
167
  const int is_self_gravity = (e->policy & engine_policy_self_gravity);
168
  const int is_external_gravity = (e->policy & engine_policy_external_gravity);
169
  const int is_with_cooling = (e->policy & engine_policy_cooling);
Matthieu Schaller's avatar
Matthieu Schaller committed
170
  const int is_with_sourceterms = (e->policy & engine_policy_sourceterms);
171
172
  const int has_part = (c->count > 0);
  const int has_gpart = (c->gcount > 0);
173

174
  /* Are we in a super-cell ? */
175
  if (c->super == c) {
176

177
    /* Add the sort task. */
178
    if (is_with_hydro && has_part) {
179
180
      c->sorts = scheduler_addtask(s, task_type_sort, task_subtype_none, 0, 0,
                                   c, NULL);
181
    }
182

183
184
185
    /* Local tasks only... */
    if (c->nodeID == e->nodeID) {

186
      /* Add the drift tasks corresponding to the policy. */
187
      if (is_with_hydro && has_part) {
188
189
        c->drift_part = scheduler_addtask(s, task_type_drift_part,
                                          task_subtype_none, 0, 0, c, NULL);
190
      }
191
      if ((is_self_gravity || is_external_gravity) && has_gpart) {
192
193
194
        c->drift_gpart = scheduler_addtask(s, task_type_drift_gpart,
                                           task_subtype_none, 0, 0, c, NULL);
      }
195

196
197
      /* Add the two half kicks */
      c->kick1 = scheduler_addtask(s, task_type_kick1, task_subtype_none, 0, 0,
198
                                   c, NULL);
199
200

      c->kick2 = scheduler_addtask(s, task_type_kick2, task_subtype_none, 0, 0,
201
                                   c, NULL);
Tom Theuns's avatar
Tom Theuns committed
202

203
204
      /* Add the time-step calculation task and its dependency */
      c->timestep = scheduler_addtask(s, task_type_timestep, task_subtype_none,
205
                                      0, 0, c, NULL);
206
207

      scheduler_addunlock(s, c->kick2, c->timestep);
208
      scheduler_addunlock(s, c->timestep, c->kick1);
209

210
      /* Add the self-gravity tasks */
211
      if (is_self_gravity && has_gpart) {
212

213
214
        /* Initialisation of the multipoles */
        c->init_grav = scheduler_addtask(s, task_type_init_grav,
215
                                         task_subtype_none, 0, 0, c, NULL);
216

217
218
        /* Gravity non-neighbouring pm calculations */
        c->grav_long_range = scheduler_addtask(
219
            s, task_type_grav_long_range, task_subtype_none, 0, 0, c, NULL);
220

221
222
        /* Gravity recursive down-pass */
        c->grav_down = scheduler_addtask(s, task_type_grav_down,
223
                                         task_subtype_none, 0, 0, c, NULL);
224

225
	if(periodic) scheduler_addunlock(s, c->init_grav, c->grav_ghost[0]);
226
        scheduler_addunlock(s, c->init_grav, c->grav_long_range);
227
228
        scheduler_addunlock(s, c->grav_long_range, c->grav_down);
        scheduler_addunlock(s, c->grav_down, c->kick2);
229
230
      }

231
      /* Add the hydrodynamics tasks */
232
      if (is_with_hydro && has_part) {
233
234

        /* Generate the ghost tasks. */
235
236
237
238
239
240
241
        c->ghost_in =
            scheduler_addtask(s, task_type_ghost, task_subtype_none, 0,
                              /* implicit = */ 1, c, NULL);
        c->ghost_out =
            scheduler_addtask(s, task_type_ghost, task_subtype_none, 0,
                              /* implicit = */ 1, c, NULL);
        engine_add_ghosts(e, c, c->ghost_in, c->ghost_out);
242
243

#ifdef EXTRA_HYDRO_LOOP
244
        /* Generate the extra ghost task. */
Matthieu Schaller's avatar
Matthieu Schaller committed
245
        c->extra_ghost = scheduler_addtask(s, task_type_extra_ghost,
246
                                           task_subtype_none, 0, 0, c, NULL);
247
#endif
248
      }
249

250
      /* Cooling task */
251
      if (is_with_cooling && has_part) {
Matthieu Schaller's avatar
Matthieu Schaller committed
252
        c->cooling = scheduler_addtask(s, task_type_cooling, task_subtype_none,
253
                                       0, 0, c, NULL);
254
255
256
257

        scheduler_addunlock(s, c->cooling, c->kick2);
      }

258
      /* add source terms */
259
      if (is_with_sourceterms && has_part) {
260
        c->sourceterms = scheduler_addtask(s, task_type_sourceterms,
261
                                           task_subtype_none, 0, 0, c, NULL);
262
      }
263
264
    }

265
  } else { /* We are above the super-cell so need to go deeper */
266

267
268
269
#ifdef SWIFT_DEBUG_CHECKS
    if (c->super != NULL) error("Incorrectly set super pointer");
#endif
270

271
272
273
274
    /* Recurse. */
    if (c->split)
      for (int k = 0; k < 8; k++)
        if (c->progeny[k] != NULL)
275
          engine_make_hierarchical_tasks(e, c->progeny[k]);
276
  }
277
}
278

279
280
281
282
283
284
285
286
287
288
void engine_make_hierarchical_tasks_mapper(void *map_data, int num_elements,
                                           void *extra_data) {
  struct engine *e = (struct engine *)extra_data;

  for (int ind = 0; ind < num_elements; ind++) {
    struct cell *c = &((struct cell *)map_data)[ind];
    engine_make_hierarchical_tasks(e, c);
  }
}

289
#ifdef WITH_MPI
290
/**
Peter W. Draper's avatar
Peter W. Draper committed
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
 * Do the exchange of one type of particles with all the other nodes.
 *
 * @param counts 2D array with the counts of particles to exchange with
 *               each other node.
 * @param parts the particle data to exchange
 * @param new_nr_parts the number of particles this node will have after all
 *                     exchanges have completed.
 * @param sizeofparts sizeof the particle struct.
 * @param alignsize the memory alignment required for this particle type.
 * @param mpi_type the MPI_Datatype for these particles.
 * @param nr_nodes the number of nodes to exchange with.
 * @param nodeID the id of this node.
 *
 * @result new particle data constructed from all the exchanges with the
 *         given alignment.
306
 */
307
static void *engine_do_redistribute(int *counts, char *parts,
308
309
                                    size_t new_nr_parts, size_t sizeofparts,
                                    size_t alignsize, MPI_Datatype mpi_type,
310
                                    int nr_nodes, int nodeID) {
311
312

  /* Allocate a new particle array with some extra margin */
313
  char *parts_new = NULL;
314
315
  if (posix_memalign(
          (void **)&parts_new, alignsize,
316
          sizeofparts * new_nr_parts * engine_redistribute_alloc_margin) != 0)
317
318
319
320
    error("Failed to allocate new particle data.");

  /* Prepare MPI requests for the asynchronous communications */
  MPI_Request *reqs;
321
322
  if ((reqs = (MPI_Request *)malloc(sizeof(MPI_Request) * 2 * nr_nodes)) ==
      NULL)
323
324
    error("Failed to allocate MPI request list.");

325
  /* Only send and receive only "chunk" particles per request. So we need to
326
327
328
   * loop as many times as necessary here. Make 2Gb/sizeofparts so we only
   * send 2Gb packets. */
  const int chunk = INT_MAX / sizeofparts;
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
  int sent = 0;
  int recvd = 0;

  int activenodes = 1;
  while (activenodes) {

    for (int k = 0; k < 2 * nr_nodes; k++) reqs[k] = MPI_REQUEST_NULL;

    /* Emit the sends and recvs for the data. */
    size_t offset_send = sent;
    size_t offset_recv = recvd;
    activenodes = 0;

    for (int k = 0; k < nr_nodes; k++) {

      /* Indices in the count arrays of the node of interest */
      const int ind_send = nodeID * nr_nodes + k;
      const int ind_recv = k * nr_nodes + nodeID;

      /* Are we sending any data this loop? */
      int sending = counts[ind_send] - sent;
      if (sending > 0) {
        activenodes++;
352
        if (sending > chunk) sending = chunk;
353
354
355
356

        /* If the send and receive is local then just copy. */
        if (k == nodeID) {
          int receiving = counts[ind_recv] - recvd;
357
          if (receiving > chunk) receiving = chunk;
358
          memcpy(&parts_new[offset_recv * sizeofparts],
359
                 &parts[offset_send * sizeofparts], sizeofparts * receiving);
360
361
        } else {
          /* Otherwise send it. */
362
363
364
          int res =
              MPI_Isend(&parts[offset_send * sizeofparts], sending, mpi_type, k,
                        ind_send, MPI_COMM_WORLD, &reqs[2 * k + 0]);
365
366
367
368
          if (res != MPI_SUCCESS)
            mpi_error(res, "Failed to isend parts to node %i.", k);
        }
      }
369

370
      /* If we're sending to this node, then move past it to next. */
371
      if (counts[ind_send] > 0) offset_send += counts[ind_send];
372

373
374
375
376
377
378
      /* Are we receiving any data from this node? Note already done if coming
       * from this node. */
      if (k != nodeID) {
        int receiving = counts[ind_recv] - recvd;
        if (receiving > 0) {
          activenodes++;
379
          if (receiving > chunk) receiving = chunk;
380
381
382
383
384
385
          int res = MPI_Irecv(&parts_new[offset_recv * sizeofparts], receiving,
                              mpi_type, k, ind_recv, MPI_COMM_WORLD,
                              &reqs[2 * k + 1]);
          if (res != MPI_SUCCESS)
            mpi_error(res, "Failed to emit irecv of parts from node %i.", k);
        }
386
387
      }

388
      /* If we're receiving from this node, then move past it to next. */
389
      if (counts[ind_recv] > 0) offset_recv += counts[ind_recv];
390
391
    }

392
393
394
395
396
397
398
    /* Wait for all the sends and recvs to tumble in. */
    MPI_Status stats[2 * nr_nodes];
    int res;
    if ((res = MPI_Waitall(2 * nr_nodes, reqs, stats)) != MPI_SUCCESS) {
      for (int k = 0; k < 2 * nr_nodes; k++) {
        char buff[MPI_MAX_ERROR_STRING];
        MPI_Error_string(stats[k].MPI_ERROR, buff, &res);
399
400
        message("request from source %i, tag %i has error '%s'.",
                stats[k].MPI_SOURCE, stats[k].MPI_TAG, buff);
401
402
      }
      error("Failed during waitall for part data.");
403
    }
404
405
406
407

    /* Move to next chunks. */
    sent += chunk;
    recvd += chunk;
408
409
410
411
412
413
414
415
416
417
  }

  /* Free temps. */
  free(reqs);

  /* And return new memory. */
  return parts_new;
}
#endif

418
/**
419
 * @brief Redistribute the particles amongst the nodes according
420
421
 *      to their cell's node IDs.
 *
422
423
424
425
 * The strategy here is as follows:
 * 1) Each node counts the number of particles it has to send to each other
 * node.
 * 2) The number of particles of each type is then exchanged.
426
427
428
429
 * 3) The particles to send are placed in a temporary buffer in which the
 * part-gpart links are preserved.
 * 4) Each node allocates enough space for the new particles.
 * 5) (Asynchronous) communications are issued to transfer the data.
430
431
 *
 *
432
433
 * @param e The #engine.
 */
434
void engine_redistribute(struct engine *e) {
435

436
#ifdef WITH_MPI
437

438
439
  const int nr_nodes = e->nr_nodes;
  const int nodeID = e->nodeID;
440
  struct space *s = e->s;
441
  struct cell *cells = s->cells_top;
442
  const int nr_cells = s->nr_cells;
443
  const int *cdim = s->cdim;
444
  const double iwidth[3] = {s->iwidth[0], s->iwidth[1], s->iwidth[2]};
445
446
447
  const double dim[3] = {s->dim[0], s->dim[1], s->dim[2]};
  struct part *parts = s->parts;
  struct gpart *gparts = s->gparts;
448
  struct spart *sparts = s->sparts;
449
  ticks tic = getticks();
450

451
  /* Allocate temporary arrays to store the counts of particles to be sent
452
453
   * and the destination of each particle */
  int *counts;
454
  if ((counts = (int *)malloc(sizeof(int) * nr_nodes * nr_nodes)) == NULL)
455
    error("Failed to allocate counts temporary buffer.");
456
  bzero(counts, sizeof(int) * nr_nodes * nr_nodes);
457

458
  int *dest;
459
  if ((dest = (int *)malloc(sizeof(int) * s->nr_parts)) == NULL)
460
461
    error("Failed to allocate dest temporary buffer.");

462
  /* Get destination of each particle */
463
  for (size_t k = 0; k < s->nr_parts; k++) {
464
465

    /* Periodic boundary conditions */
466
    for (int j = 0; j < 3; j++) {
467
468
469
470
471
      if (parts[k].x[j] < 0.0)
        parts[k].x[j] += dim[j];
      else if (parts[k].x[j] >= dim[j])
        parts[k].x[j] -= dim[j];
    }
James Willis's avatar
James Willis committed
472
473
474
    const int cid =
        cell_getid(cdim, parts[k].x[0] * iwidth[0], parts[k].x[1] * iwidth[1],
                   parts[k].x[2] * iwidth[2]);
475
476
#ifdef SWIFT_DEBUG_CHECKS
    if (cid < 0 || cid >= s->nr_cells)
477
      error("Bad cell id %i for part %zu at [%.3e,%.3e,%.3e].", cid, k,
478
479
480
            parts[k].x[0], parts[k].x[1], parts[k].x[2]);
#endif

481
    dest[k] = cells[cid].nodeID;
482
483

    /* The counts array is indexed as count[from * nr_nodes + to]. */
484
485
    counts[nodeID * nr_nodes + dest[k]] += 1;
  }
486
487

  /* Sort the particles according to their cell index. */
Matthieu Schaller's avatar
Matthieu Schaller committed
488
  if (s->nr_parts > 0)
489
    space_parts_sort(s, dest, s->nr_parts, 0, nr_nodes - 1, e->verbose);
490

491
492
#ifdef SWIFT_DEBUG_CHECKS
  /* Verify that the part have been sorted correctly. */
493
494
495
496
  for (size_t k = 0; k < s->nr_parts; k++) {
    const struct part *p = &s->parts[k];

    /* New cell index */
497
    const int new_cid =
498
499
500
501
        cell_getid(s->cdim, p->x[0] * s->iwidth[0], p->x[1] * s->iwidth[1],
                   p->x[2] * s->iwidth[2]);

    /* New cell of this part */
502
503
    const struct cell *c = &s->cells_top[new_cid];
    const int new_node = c->nodeID;
504

505
506
    if (dest[k] != new_node)
      error("part's new node index not matching sorted index.");
507
508
509
510
511

    if (p->x[0] < c->loc[0] || p->x[0] > c->loc[0] + c->width[0] ||
        p->x[1] < c->loc[1] || p->x[1] > c->loc[1] + c->width[1] ||
        p->x[2] < c->loc[2] || p->x[2] > c->loc[2] + c->width[2])
      error("part not sorted into the right top-level cell!");
512
513
514
  }
#endif

515
  /* We need to re-link the gpart partners of parts. */
516
517
518
519
520
521
  if (s->nr_parts > 0) {
    int current_dest = dest[0];
    size_t count_this_dest = 0;
    for (size_t k = 0; k < s->nr_parts; ++k) {
      if (s->parts[k].gpart != NULL) {

522
523
524
        /* As the addresses will be invalidated by the communications, we will
         * instead store the absolute index from the start of the sub-array of
         * particles to be sent to a given node.
525
         * Recall that gparts without partners have a positive id.
526
         * We will restore the pointers on the receiving node later on. */
527
528
529
530
        if (dest[k] != current_dest) {
          current_dest = dest[k];
          count_this_dest = 0;
        }
531

532
#ifdef SWIFT_DEBUG_CHECKS
533
        if (s->parts[k].gpart->id_or_neg_offset > 0)
534
535
          error("Trying to link a partnerless gpart !");
#endif
536

537
        s->parts[k].gpart->id_or_neg_offset = -count_this_dest;
538
        count_this_dest++;
539
540
541
      }
    }
  }
542
  free(dest);
543

544
  /* Get destination of each s-particle */
545
546
547
548
549
550
551
552
553
  int *s_counts;
  if ((s_counts = (int *)malloc(sizeof(int) * nr_nodes * nr_nodes)) == NULL)
    error("Failed to allocate s_counts temporary buffer.");
  bzero(s_counts, sizeof(int) * nr_nodes * nr_nodes);

  int *s_dest;
  if ((s_dest = (int *)malloc(sizeof(int) * s->nr_sparts)) == NULL)
    error("Failed to allocate s_dest temporary buffer.");

554
555
556
557
558
559
560
561
562
563
564
565
566
567
  for (size_t k = 0; k < s->nr_sparts; k++) {

    /* Periodic boundary conditions */
    for (int j = 0; j < 3; j++) {
      if (sparts[k].x[j] < 0.0)
        sparts[k].x[j] += dim[j];
      else if (sparts[k].x[j] >= dim[j])
        sparts[k].x[j] -= dim[j];
    }
    const int cid =
        cell_getid(cdim, sparts[k].x[0] * iwidth[0], sparts[k].x[1] * iwidth[1],
                   sparts[k].x[2] * iwidth[2]);
#ifdef SWIFT_DEBUG_CHECKS
    if (cid < 0 || cid >= s->nr_cells)
568
      error("Bad cell id %i for spart %zu at [%.3e,%.3e,%.3e].", cid, k,
569
570
571
572
573
574
575
576
577
578
            sparts[k].x[0], sparts[k].x[1], sparts[k].x[2]);
#endif

    s_dest[k] = cells[cid].nodeID;

    /* The counts array is indexed as count[from * nr_nodes + to]. */
    s_counts[nodeID * nr_nodes + s_dest[k]] += 1;
  }

  /* Sort the particles according to their cell index. */
Matthieu Schaller's avatar
Matthieu Schaller committed
579
  if (s->nr_sparts > 0)
580
    space_sparts_sort(s, s_dest, s->nr_sparts, 0, nr_nodes - 1, e->verbose);
581

582
583
#ifdef SWIFT_DEBUG_CHECKS
  /* Verify that the spart have been sorted correctly. */
584
585
586
587
  for (size_t k = 0; k < s->nr_sparts; k++) {
    const struct spart *sp = &s->sparts[k];

    /* New cell index */
588
    const int new_cid =
589
590
591
592
        cell_getid(s->cdim, sp->x[0] * s->iwidth[0], sp->x[1] * s->iwidth[1],
                   sp->x[2] * s->iwidth[2]);

    /* New cell of this spart */
593
594
    const struct cell *c = &s->cells_top[new_cid];
    const int new_node = c->nodeID;
595

596
597
    if (s_dest[k] != new_node)
      error("spart's new node index not matching sorted index.");
598
599
600
601
602

    if (sp->x[0] < c->loc[0] || sp->x[0] > c->loc[0] + c->width[0] ||
        sp->x[1] < c->loc[1] || sp->x[1] > c->loc[1] + c->width[1] ||
        sp->x[2] < c->loc[2] || sp->x[2] > c->loc[2] + c->width[2])
      error("spart not sorted into the right top-level cell!");
603
604
605
  }
#endif

606
  /* We need to re-link the gpart partners of sparts. */
607
608
609
  if (s->nr_sparts > 0) {
    int current_dest = s_dest[0];
    size_t count_this_dest = 0;
610
    for (size_t k = 0; k < s->nr_sparts; ++k) {
611
612
613
614
615
      if (s->sparts[k].gpart != NULL) {

        /* As the addresses will be invalidated by the communications, we will
         * instead store the absolute index from the start of the sub-array of
         * particles to be sent to a given node.
616
         * Recall that gparts without partners have a positive id.
617
618
619
620
621
622
623
         * We will restore the pointers on the receiving node later on. */
        if (s_dest[k] != current_dest) {
          current_dest = s_dest[k];
          count_this_dest = 0;
        }

#ifdef SWIFT_DEBUG_CHECKS
624
        if (s->sparts[k].gpart->id_or_neg_offset > 0)
625
626
627
628
629
630
631
632
633
          error("Trying to link a partnerless gpart !");
#endif

        s->sparts[k].gpart->id_or_neg_offset = -count_this_dest;
        count_this_dest++;
      }
    }
  }

634
635
  free(s_dest);

636
  /* Get destination of each g-particle */
637
638
639
640
641
642
643
644
645
  int *g_counts;
  if ((g_counts = (int *)malloc(sizeof(int) * nr_nodes * nr_nodes)) == NULL)
    error("Failed to allocate g_gcount temporary buffer.");
  bzero(g_counts, sizeof(int) * nr_nodes * nr_nodes);

  int *g_dest;
  if ((g_dest = (int *)malloc(sizeof(int) * s->nr_gparts)) == NULL)
    error("Failed to allocate g_dest temporary buffer.");

646
  for (size_t k = 0; k < s->nr_gparts; k++) {
647
648

    /* Periodic boundary conditions */
649
    for (int j = 0; j < 3; j++) {
650
651
652
653
      if (gparts[k].x[j] < 0.0)
        gparts[k].x[j] += dim[j];
      else if (gparts[k].x[j] >= dim[j])
        gparts[k].x[j] -= dim[j];
654
    }
James Willis's avatar
James Willis committed
655
656
657
    const int cid =
        cell_getid(cdim, gparts[k].x[0] * iwidth[0], gparts[k].x[1] * iwidth[1],
                   gparts[k].x[2] * iwidth[2]);
658
659
#ifdef SWIFT_DEBUG_CHECKS
    if (cid < 0 || cid >= s->nr_cells)
660
      error("Bad cell id %i for gpart %zu at [%.3e,%.3e,%.3e].", cid, k,
661
662
663
            gparts[k].x[0], gparts[k].x[1], gparts[k].x[2]);
#endif

664
    g_dest[k] = cells[cid].nodeID;
665
666

    /* The counts array is indexed as count[from * nr_nodes + to]. */
667
    g_counts[nodeID * nr_nodes + g_dest[k]] += 1;
668
  }
669
670

  /* Sort the gparticles according to their cell index. */
Matthieu Schaller's avatar
Matthieu Schaller committed
671
  if (s->nr_gparts > 0)
672
    space_gparts_sort(s, g_dest, s->nr_gparts, 0, nr_nodes - 1, e->verbose);
673

674
675
#ifdef SWIFT_DEBUG_CHECKS
  /* Verify that the gpart have been sorted correctly. */
676
677
678
679
  for (size_t k = 0; k < s->nr_gparts; k++) {
    const struct gpart *gp = &s->gparts[k];

    /* New cell index */
680
    const int new_cid =
681
682
683
684
        cell_getid(s->cdim, gp->x[0] * s->iwidth[0], gp->x[1] * s->iwidth[1],
                   gp->x[2] * s->iwidth[2]);

    /* New cell of this gpart */
685
686
    const struct cell *c = &s->cells_top[new_cid];
    const int new_node = c->nodeID;
687

688
    if (g_dest[k] != new_node)
689
690
      error("gpart's new node index not matching sorted index (%d != %d).",
            g_dest[k], new_node);
691
692
693
694
695

    if (gp->x[0] < c->loc[0] || gp->x[0] > c->loc[0] + c->width[0] ||
        gp->x[1] < c->loc[1] || gp->x[1] > c->loc[1] + c->width[1] ||
        gp->x[2] < c->loc[2] || gp->x[2] > c->loc[2] + c->width[2])
      error("gpart not sorted into the right top-level cell!");
696
697
698
  }
#endif

699
700
  free(g_dest);

701
702
703
704
705
  /* Get all the counts from all the nodes. */
  if (MPI_Allreduce(MPI_IN_PLACE, counts, nr_nodes * nr_nodes, MPI_INT, MPI_SUM,
                    MPI_COMM_WORLD) != MPI_SUCCESS)
    error("Failed to allreduce particle transfer counts.");

706
  /* Get all the s_counts from all the nodes. */
707
708
709
710
711
712
713
714
715
  if (MPI_Allreduce(MPI_IN_PLACE, g_counts, nr_nodes * nr_nodes, MPI_INT,
                    MPI_SUM, MPI_COMM_WORLD) != MPI_SUCCESS)
    error("Failed to allreduce gparticle transfer counts.");

  /* Get all the g_counts from all the nodes. */
  if (MPI_Allreduce(MPI_IN_PLACE, s_counts, nr_nodes * nr_nodes, MPI_INT,
                    MPI_SUM, MPI_COMM_WORLD) != MPI_SUCCESS)
    error("Failed to allreduce sparticle transfer counts.");

Peter W. Draper's avatar
Peter W. Draper committed
716
  /* Report how many particles will be moved. */
717
718
  if (e->verbose) {
    if (e->nodeID == 0) {
719
720
      size_t total = 0, g_total = 0, s_total = 0;
      size_t unmoved = 0, g_unmoved = 0, s_unmoved = 0;
721
      for (int p = 0, r = 0; p < nr_nodes; p++) {
722
        for (int n = 0; n < nr_nodes; n++) {
723
          total += counts[r];
724
725
          g_total += g_counts[r];
          s_total += s_counts[r];
726
          if (p == n) {
727
728
729
730
            unmoved += counts[r];
            g_unmoved += g_counts[r];
            s_unmoved += s_counts[r];
          }
731
732
733
          r++;
        }
      }
Matthieu Schaller's avatar
Matthieu Schaller committed
734
735
736
737
738
739
740
741
742
743
744
      if (total > 0)
        message("%ld of %ld (%.2f%%) of particles moved", total - unmoved,
                total, 100.0 * (double)(total - unmoved) / (double)total);
      if (g_total > 0)
        message("%ld of %ld (%.2f%%) of g-particles moved", g_total - g_unmoved,
                g_total,
                100.0 * (double)(g_total - g_unmoved) / (double)g_total);
      if (s_total > 0)
        message("%ld of %ld (%.2f%%) of s-particles moved", s_total - s_unmoved,
                s_total,
                100.0 * (double)(s_total - s_unmoved) / (double)s_total);
745
    }
746
747
  }

Peter W. Draper's avatar
Peter W. Draper committed
748
749
750
  /* Now each node knows how many parts, sparts and gparts will be transferred
   * to every other node.
   * Get the new numbers of particles for this node. */
751
  size_t nr_parts = 0, nr_gparts = 0, nr_sparts = 0;
752
  for (int k = 0; k < nr_nodes; k++) nr_parts += counts[k * nr_nodes + nodeID];
753
754
  for (int k = 0; k < nr_nodes; k++)
    nr_gparts += g_counts[k * nr_nodes + nodeID];
755
756
  for (int k = 0; k < nr_nodes; k++)
    nr_sparts += s_counts[k * nr_nodes + nodeID];
757

Peter W. Draper's avatar
Peter W. Draper committed
758
759
760
761
  /* Now exchange the particles, type by type to keep the memory required
   * under control. */

  /* SPH particles. */
762
  void *new_parts = engine_do_redistribute(counts, (char *)s->parts, nr_parts,
763
764
765
                                           sizeof(struct part), part_align,
                                           part_mpi_type, nr_nodes, nodeID);
  free(s->parts);
766
  s->parts = (struct part *)new_parts;
767
768
  s->nr_parts = nr_parts;
  s->size_parts = engine_redistribute_alloc_margin * nr_parts;
769

Peter W. Draper's avatar
Peter W. Draper committed
770
  /* Extra SPH particle properties. */
771
  new_parts = engine_do_redistribute(counts, (char *)s->xparts, nr_parts,
772
773
774
                                     sizeof(struct xpart), xpart_align,
                                     xpart_mpi_type, nr_nodes, nodeID);
  free(s->xparts);
775
  s->xparts = (struct xpart *)new_parts;
776

Peter W. Draper's avatar
Peter W. Draper committed
777
  /* Gravity particles. */
778
  new_parts = engine_do_redistribute(g_counts, (char *)s->gparts, nr_gparts,
779
780
781
                                     sizeof(struct gpart), gpart_align,
                                     gpart_mpi_type, nr_nodes, nodeID);
  free(s->gparts);
782
  s->gparts = (struct gpart *)new_parts;
783
784
  s->nr_gparts = nr_gparts;
  s->size_gparts = engine_redistribute_alloc_margin * nr_gparts;
785

Peter W. Draper's avatar
Peter W. Draper committed
786
  /* Star particles. */
787
  new_parts = engine_do_redistribute(s_counts, (char *)s->sparts, nr_sparts,
788
789
790
                                     sizeof(struct spart), spart_align,
                                     spart_mpi_type, nr_nodes, nodeID);
  free(s->sparts);
791
  s->sparts = (struct spart *)new_parts;
792
793
  s->nr_sparts = nr_sparts;
  s->size_sparts = engine_redistribute_alloc_margin * nr_sparts;
794

795
796
797
798
799
  /* All particles have now arrived. Time for some final operations on the
     stuff we just received */

  /* Restore the part<->gpart and spart<->gpart links */
  size_t offset_parts = 0, offset_sparts = 0, offset_gparts = 0;
800
801
802
803
804
  for (int node = 0; node < nr_nodes; ++node) {

    const int ind_recv = node * nr_nodes + nodeID;
    const size_t count_parts = counts[ind_recv];
    const size_t count_gparts = g_counts[ind_recv];
805
    const size_t count_sparts = s_counts[ind_recv];
806
807
808
809

    /* Loop over the gparts received from that node */
    for (size_t k = offset_gparts; k < offset_gparts + count_gparts; ++k) {

810
      /* Does this gpart have a gas partner ? */
811
      if (s->gparts[k].type == swift_type_gas) {
812

Matthieu Schaller's avatar
Style    
Matthieu Schaller committed
813
        const ptrdiff_t partner_index =
814
            offset_parts - s->gparts[k].id_or_neg_offset;
815
816

        /* Re-link */
817
818
        s->gparts[k].id_or_neg_offset = -partner_index;
        s->parts[partner_index].gpart = &s->gparts[k];
819
      }
820
821

      /* Does this gpart have a star partner ? */
822
      if (s->gparts[k].type == swift_type_star) {
823
824

        const ptrdiff_t partner_index =
825
            offset_sparts - s->gparts[k].id_or_neg_offset;
826
827

        /* Re-link */
828
829
        s->gparts[k].id_or_neg_offset = -partner_index;
        s->sparts[partner_index].gpart = &s->gparts[k];
830
      }
831
832
833
834
    }

    offset_parts += count_parts;
    offset_gparts += count_gparts;
835
    offset_sparts += count_sparts;
836
837
  }

838
839
840
841
842
  /* Clean up the counts now we done. */
  free(counts);
  free(g_counts);
  free(s_counts);

843
#ifdef SWIFT_DEBUG_CHECKS
844
  /* Verify that all parts are in the right place. */
845
  for (size_t k = 0; k < nr_parts; k++) {
846
847
848
    const int cid =
        cell_getid(cdim, s->parts[k].x[0] * iwidth[0],
                   s->parts[k].x[1] * iwidth[1], s->parts[k].x[2] * iwidth[2]);
849
    if (cells[cid].nodeID != nodeID)
850
      error("Received particle (%zu) that does not belong here (nodeID=%i).", k,
851
852
            cells[cid].nodeID);
  }
853
  for (size_t k = 0; k < nr_gparts; k++) {
854
855
856
    const int cid = cell_getid(cdim, s->gparts[k].x[0] * iwidth[0],
                               s->gparts[k].x[1] * iwidth[1],
                               s->gparts[k].x[2] * iwidth[2]);
857
858
859
860
861
    if (cells[cid].nodeID != nodeID)
      error("Received g-particle (%zu) that does not belong here (nodeID=%i).",
            k, cells[cid].nodeID);
  }
  for (size_t k = 0; k < nr_sparts; k++) {
862
863
864
    const int cid = cell_getid(cdim, s->sparts[k].x[0] * iwidth[0],
                               s->sparts[k].x[1] * iwidth[1],
                               s->sparts[k].x[2] * iwidth[2]);
865
866
867
868
    if (cells[cid].nodeID != nodeID)
      error("Received s-particle (%zu) that does not belong here (nodeID=%i).",
            k, cells[cid].nodeID);
  }
869

870
  /* Verify that the links are correct */
871
  part_verify_links(s->parts, s->gparts, s->sparts, nr_parts, nr_gparts,
872
                    nr_sparts, e->verbose);
873
#endif
874

875
876
877
878
879
  /* Be verbose about what just happened. */
  if (e->verbose) {
    int my_cells = 0;
    for (int k = 0; k < nr_cells; k++)
      if (cells[k].nodeID == nodeID) my_cells += 1;
880
881
    message("node %i now has %zu parts, %zu sparts and %zu gparts in %i cells.",
            nodeID, nr_parts, nr_sparts, nr_gparts, my_cells);
882
883
  }

884
885
886
  if (e->verbose)
    message("took %.3f %s.", clocks_from_ticks(getticks() - tic),
            clocks_getunit());
887
#else
888
  error("SWIFT was not compiled with MPI support.");
889
890
#endif
}
891

892
/**
893
 * @brief Repartition the cells amongst the nodes.
894
895
896
 *
 * @param e The #engine.
 */
897
void engine_repartition(struct engine *e) {