Repository navigation
Expand file tree
/
Copy pathljs.cpp
More file actions
634 lines (536 loc) · 19.3 KB
/
Copy pathljs.cpp
File metadata and controls
634 lines (536 loc) · 19.3 KB
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
446
447
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
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
/* ----------------------------------------------------------------------
miniMD is a simple, parallel molecular dynamics (MD) code. miniMD is
an MD microapplication in the Mantevo project at Sandia National
Laboratories ( http://www.mantevo.org ). The primary
authors of miniMD are Steve Plimpton (sjplimp@sandia.gov), Paul Crozier
(pscrozi@sandia.gov) and Christian Trott (crtrott@sandia.gov).
Copyright (2008) Sandia Corporation. Under the terms of Contract
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
certain rights in this software. This library 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 library 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
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this software; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
USA. See also: http://www.gnu.org/licenses/lgpl.txt .
For questions, contact Paul S. Crozier (pscrozi@sandia.gov) or
Christian Trott (crtrott@sandia.gov).
Please read the accompanying README and LICENSE files.
---------------------------------------------------------------------- */
#include "mpi.h"
#include "stdio.h"
#include "stdlib.h"
#include "atom.h"
#include "comm.h"
#include "force.h"
#include "integrate.h"
#include "ljs.h"
#include "neighbor.h"
#include "offload.h"
#include "openmp.h"
#include "string.h"
#include "thermo.h"
#include "threadData.h"
#include "timer.h"
#include "util.h"
#include "variant.h"
#ifdef OFFLOAD_NVPTX
#include "nvToolsExt.h"
#endif
#define MAXLINE 256
int input(In &, const char *);
void create_box(Atom &, int, int, int, double);
int create_atoms(Atom &, int, int, int, double);
void create_velocity(double, Atom &, Thermo &);
void output(In &, Atom &, Force *, Neighbor &, Comm &, Thermo &, Integrate &, Timer &, int);
void finalize_setup(const Atom &);
int read_lammps_data(Atom &atom, Comm &comm, Neighbor &neighbor, Integrate &integrate, Thermo &thermo, char *file, int units);
int main(int argc, char **argv)
{
In in;
in.datafile = NULL;
int me = 0; // local MPI rank
int nprocs = 1; // number of MPI ranks
int num_threads = 1; // number of OpenMP threads
int num_steps = -1; // number of timesteps (if -1 use value from lj.in)
int system_size = -1; // size of the system (if -1 use value from lj.in)
int nx = -1;
int ny = -1;
int nz = -1;
int check_safeexchange = 0; // if 1 complain if atom moves further than 1 subdomain length between exchanges
int do_safeexchange = 0; // if 1 use safe exchange mode [allows exchange over multiple subdomains]
int use_sse = 0; // setting for SSE variant of miniMD only
int screen_yaml = 0; // print yaml output to screen also
int yaml_output = 0; // print yaml output
int halfneigh = 1; // 1: use half neighborlist; 0: use full neighborlist; -1: use original miniMD version half neighborlist force
int teams = 1;
int device = 0;
int neighbor_size = -1;
char *input_file = NULL;
int ghost_newton = 1;
int sort = -1;
int ntypes = 4;
char *check_output = 0;
for(int i = 0; i < argc; i++)
{
if((strcmp(argv[i], "-i") == 0) || (strcmp(argv[i], "--input_file") == 0))
{
input_file = argv[++i];
continue;
}
if((strcmp(argv[i], "--print-user-max-team-size") == 0))
{
printf("%d\n", get_user_max_team_size());
exit(EXIT_SUCCESS);
}
}
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &me);
MPI_Comm_size(MPI_COMM_WORLD, &nprocs);
int error = 0;
if(input_file == NULL)
{
error = input(in, "in.lj.miniMD");
}
else
{
error = input(in, input_file);
}
if(error)
{
MPI_Finalize();
exit(0);
}
srand(5413);
for(int i = 0; i < argc; i++)
{
if((strcmp(argv[i], "-t") == 0) || (strcmp(argv[i], "--num_threads") == 0))
{
num_threads = atoi(argv[++i]);
continue;
}
if(strcmp(argv[i], "--check-output") == 0)
{
check_output = strdup(argv[++i]);
continue;
}
if((strcmp(argv[i], "--teams") == 0))
{
teams = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-n") == 0) || (strcmp(argv[i], "--nsteps") == 0))
{
num_steps = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-s") == 0) || (strcmp(argv[i], "--size") == 0))
{
system_size = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-nx") == 0))
{
nx = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-ny") == 0))
{
ny = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-nz") == 0))
{
nz = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "--ntypes") == 0))
{
ntypes = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-b") == 0) || (strcmp(argv[i], "--neigh_bins") == 0))
{
neighbor_size = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "--half_neigh") == 0))
{
halfneigh = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-sse") == 0))
{
use_sse = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "--check_exchange") == 0))
{
check_safeexchange = 1;
continue;
}
if((strcmp(argv[i], "--safe_exchange") == 0))
{
do_safeexchange = 1;
continue;
}
if((strcmp(argv[i], "--sort") == 0))
{
sort = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-o") == 0) || (strcmp(argv[i], "--yaml_output") == 0))
{
yaml_output = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "--yaml_screen") == 0))
{
screen_yaml = 1;
continue;
}
if((strcmp(argv[i], "-f") == 0) || (strcmp(argv[i], "--data_file") == 0))
{
if(in.datafile == NULL)
{
in.datafile = new char[1000];
}
strcpy(in.datafile, argv[++i]);
continue;
}
if((strcmp(argv[i], "-u") == 0) || (strcmp(argv[i], "--units") == 0))
{
in.units = strcmp(argv[++i], "metal") == 0 ? 1 : 0;
continue;
}
if((strcmp(argv[i], "-p") == 0) || (strcmp(argv[i], "--force") == 0))
{
in.forcetype = strcmp(argv[++i], "eam") == 0 ? FORCEEAM : FORCELJ;
continue;
}
if((strcmp(argv[i], "-gn") == 0) || (strcmp(argv[i], "--ghost_newton") == 0))
{
ghost_newton = atoi(argv[++i]);
continue;
}
if((strcmp(argv[i], "-h") == 0) || (strcmp(argv[i], "--help") == 0))
{
printf("\n-----------------------------------------------------------------------------------------------------------\n");
printf("-------------" VARIANT_STRING "--------------------\n");
printf("-------------------------------------------------------------------------------------------------------------\n\n");
printf("miniMD is a simple, parallel molecular dynamics (MD) code,\n"
"which is part of the Mantevo project at Sandia National\n"
"Laboratories ( http://www.mantevo.org ).\n"
"The original authors of miniMD are Steve Plimpton (sjplimp@sandia.gov) ,\n"
"Paul Crozier (pscrozi@sandia.gov) with current\n"
"versions written by Christian Trott (crtrott@sandia.gov).\n\n");
printf("Commandline Options:\n");
printf("\n Execution configuration:\n");
printf("\t--teams <nteams>: set number of thread-teams used per MPI rank (default 1)\n");
printf("\t-t / --num_threads <threads>: set number of threads per thread-team (default 1)\n");
printf("\t--half_neigh <int>: use half neighborlists (default 1)\n"
"\t 0: full neighborlist\n"
"\t 1: half neighborlist\n"
"\t -1: original miniMD half neighborlist force (not OpenMP safe)\n");
printf("\t-d / --device <int>: choose device to use (only applicable for GPU execution)\n");
printf("\t-dm / --device_map: map devices to MPI ranks\n");
printf("\t-ng / --num_gpus <int>: give number of GPUs per Node (used in conjuction with -dm\n"
"\t to determine device id: 'id=mpi_rank%%ng' (default 2)\n");
printf("\t--skip_gpu <int>: skip the specified gpu when assigning devices to MPI ranks\n"
"\t used in conjunction with -dm (but must come first in arg list)\n");
printf("\t-sse <sse_version>: use explicit sse intrinsics (use miniMD-SSE variant)\n");
printf("\t-gn / --ghost_newton <int>: set usage of newtons third law for ghost atoms\n"
"\t (only applicable with half neighborlists)\n");
printf("\n Simulation setup:\n");
printf("\t-i / --input_file <string>: set input file to be used (default: in.lj.miniMD)\n");
printf("\t--ntypes <int>: set number of atom types for simulation (default: 4)\n");
printf("\t-n / --nsteps <int>: set number of timesteps for simulation\n");
printf("\t-s / --size <int>: set linear dimension of systembox\n");
printf("\t-nx/-ny/-nz <int>: set linear dimension of systembox in x/y/z direction\n");
printf("\t-b / --neigh_bins <int>: set linear dimension of neighbor bin grid\n");
printf("\t-u / --units <string>: set units (lj or metal), see LAMMPS documentation\n");
printf("\t-p / --force <string>: set interaction model (lj or eam)\n");
printf("\t-f / --data_file <string>: read configuration from LAMMPS data file\n");
printf("\n Miscelaneous:\n");
printf("\t--check_exchange: check whether atoms moved further than subdomain width\n");
printf("\t--check_output <file>: write out final check values to file\n");
printf("\t--safe_exchange: perform exchange communication with all MPI processes\n"
"\t within rcut_neighbor (outer force cutoff)\n");
printf("\t--sort <n>: resort atoms (simple bins) every <n> steps (default: use reneigh frequency; never=0)");
printf("\t-o / --yaml_output <int>: level of yaml output (default 1)\n");
printf("\t--yaml_screen: write yaml output also to screen\n");
printf("\t-h / --help: display this help message\n\n");
printf("---------------------------------------------------------\n\n");
exit(0);
}
}
#ifdef USE_OFFLOAD
if(!check_offload())
{
die("Offload isn't working properly\n");
}
#endif
Atom atom(ntypes);
Neighbor neighbor(ntypes);
Integrate integrate;
Thermo thermo;
Comm comm;
Timer timer;
ThreadData threads;
Force *force = new Force(ntypes);
threads.mpi_me = me;
threads.mpi_num_threads = nprocs;
threads.omp_me = 0;
threads.omp_num_threads = num_threads;
threads.teams = teams;
atom.threads = &threads;
comm.threads = &threads;
force->threads = &threads;
integrate.threads = &threads;
neighbor.threads = &threads;
thermo.threads = &threads;
if(in.forcetype == FORCELJ)
{
for(int i = 0; i < ntypes * ntypes; i++)
{
force->epsilon[i] = in.epsilon;
force->sigma[i] = in.sigma;
force->sigma6[i] = in.sigma * in.sigma * in.sigma * in.sigma * in.sigma * in.sigma;
}
}
neighbor.ghost_newton = ghost_newton;
omp_set_num_threads(num_threads);
neighbor.timer = &timer;
force->timer = &timer;
comm.check_safeexchange = check_safeexchange;
comm.do_safeexchange = do_safeexchange;
force->use_sse = use_sse;
neighbor.halfneigh = halfneigh;
if(halfneigh < 0)
{
force->use_oldcompute = 1;
}
if(use_sse)
{
#ifdef VARIANT_REFERENCE
if(me == 0)
{
fprintf(stderr, "ERROR: Trying to run with -sse with miniMD reference version. Use SSE variant instead. Exiting.\n");
}
MPI_Finalize();
exit(0);
#endif
}
if(num_steps > 0)
{
in.ntimes = num_steps;
}
if(system_size > 0)
{
in.nx = system_size;
in.ny = system_size;
in.nz = system_size;
}
if(nx > 0)
{
in.nx = nx;
if(ny > 0)
{
in.ny = ny;
}
else if(system_size < 0)
{
in.ny = nx;
}
if(nz > 0)
{
in.nz = nz;
}
else if(system_size < 0)
{
in.nz = nx;
}
}
if(neighbor_size > 0)
{
neighbor.nbinx = neighbor_size;
neighbor.nbiny = neighbor_size;
neighbor.nbinz = neighbor_size;
}
if(neighbor_size < 0 && in.datafile == NULL)
{
MMD_float neighscale = 5.0 / 6.0;
neighbor.nbinx = neighscale * in.nx;
neighbor.nbiny = neighscale * in.ny;
neighbor.nbinz = neighscale * in.nz;
}
if(neighbor_size < 0 && in.datafile)
{
neighbor.nbinx = -1;
}
if(neighbor.nbinx == 0)
{
neighbor.nbinx = 1;
}
if(neighbor.nbiny == 0)
{
neighbor.nbiny = 1;
}
if(neighbor.nbinz == 0)
{
neighbor.nbinz = 1;
}
integrate.ntimes = in.ntimes;
integrate.dt = in.dt;
integrate.sort_every = sort > 0 ? sort : (sort < 0 ? in.neigh_every : 0);
neighbor.every = in.neigh_every;
neighbor.cutneigh = in.neigh_cut;
force->cutforce = in.force_cut;
thermo.nstat = in.thermo_nstat;
if(me == 0)
{
printf("# Create System:\n");
}
if(in.datafile)
{
read_lammps_data(atom, comm, neighbor, integrate, thermo, in.datafile, in.units);
MMD_float volume = atom.box.xprd * atom.box.yprd * atom.box.zprd;
in.rho = 1.0 * atom.natoms / volume;
force->setup();
if(in.forcetype == FORCEEAM)
{
atom.mass = force->mass;
}
}
else
{
create_box(atom, in.nx, in.ny, in.nz, in.rho);
comm.setup(neighbor.cutneigh, atom);
neighbor.setup(atom);
integrate.setup();
force->setup();
if(in.forcetype == FORCEEAM)
{
atom.mass = force->mass;
}
create_atoms(atom, in.nx, in.ny, in.nz, in.rho);
thermo.setup(in.rho, integrate, atom, in.units);
create_velocity(in.t_request, atom, thermo);
}
if(in.forcetype == FORCEEAM)
{
die("ERROR: ForceEAM not supported by OpenMP 5.0 version\n");
}
finalize_setup(atom);
if(me == 0)
{
printf("# Done .... \n");
}
if(me == 0)
{
fprintf(stdout, "# " VARIANT_STRING " output ...\n");
fprintf(stdout, "# Run Settings: \n");
fprintf(stdout, "\t# MPI processes: %i\n", neighbor.threads->mpi_num_threads);
fprintf(stdout, "\t# OpenMP threads: %i\n", neighbor.threads->omp_num_threads);
fprintf(stdout, "\t# Inputfile: %s\n", input_file == 0 ? "in.lj.miniMD" : input_file);
fprintf(stdout, "\t# Datafile: %s\n", in.datafile ? in.datafile : "None");
fprintf(stdout, "# Physics Settings: \n");
fprintf(stdout, "\t# ForceStyle: %s\n", in.forcetype == FORCELJ ? "LJ" : "EAM");
fprintf(stdout, "\t# Force Parameters: %2.2lf %2.2lf\n", in.epsilon, in.sigma);
fprintf(stdout, "\t# Units: %s\n", in.units == 0 ? "LJ" : "METAL");
fprintf(stdout, "\t# Atoms: %i\n", atom.natoms);
fprintf(stdout, "\t# Atom types: %i\n", atom.ntypes);
fprintf(stdout, "\t# System size: %2.2lf %2.2lf %2.2lf (unit cells: %i %i %i)\n", atom.box.xprd, atom.box.yprd, atom.box.zprd, in.nx, in.ny, in.nz);
fprintf(stdout, "\t# Density: %lf\n", in.rho);
fprintf(stdout, "\t# Force cutoff: %lf\n", force->cutforce);
fprintf(stdout, "\t# Timestep size: %lf\n", integrate.dt);
fprintf(stdout, "# Technical Settings: \n");
fprintf(stdout, "\t# Neigh cutoff: %lf\n", neighbor.cutneigh);
fprintf(stdout, "\t# Half neighborlists: %i\n", neighbor.halfneigh);
fprintf(stdout, "\t# Neighbor bins: %i %i %i\n", neighbor.nbinx, neighbor.nbiny, neighbor.nbinz);
fprintf(stdout, "\t# Neighbor frequency: %i\n", neighbor.every);
fprintf(stdout, "\t# Sorting frequency: %i\n", integrate.sort_every);
fprintf(stdout, "\t# Thermo frequency: %i\n", thermo.nstat);
fprintf(stdout, "\t# Ghost Newton: %i\n", ghost_newton);
fprintf(stdout, "\t# Use intrinsics: %i\n", force->use_sse);
fprintf(stdout, "\t# Do safe exchange: %i\n", comm.do_safeexchange);
fprintf(stdout, "\t# Size of float: %i\n", ( int )sizeof(MMD_float));
}
comm.exchange(atom);
if(sort > 0)
{
atom.sort(neighbor);
}
comm.borders(atom);
force->evflag = 1;
neighbor.build(atom);
force->compute(atom, neighbor, comm, me);
if(neighbor.halfneigh && neighbor.ghost_newton)
{
comm.reverse_communicate(atom);
}
if(me == 0)
{
printf("# Starting dynamics ...\n");
}
if(me == 0)
{
printf("# Timestep T U P Time\n");
}
thermo.compute(0, atom, neighbor, force, timer, comm);
timer.barrier_start(TIME_TOTAL);
#ifdef OFFLOAD_NVPTX
nvtxRangeId_t id1 = nvtxRangeStartA("Compute range");
#endif
integrate.run(atom, force, neighbor, comm, thermo, timer);
#ifdef OFFLOAD_NVPTX
nvtxRangeEnd(id1);
#endif
timer.barrier_stop(TIME_TOTAL);
int natoms;
MPI_Allreduce(&atom.nlocal, &natoms, 1, MPI_INT, MPI_SUM, MPI_COMM_WORLD);
force->evflag = 1;
force->compute(atom, neighbor, comm, me);
if(neighbor.halfneigh && neighbor.ghost_newton)
{
comm.reverse_communicate(atom);
}
thermo.compute(-1, atom, neighbor, force, timer, comm);
if(me == 0)
{
double time_other = timer.array[TIME_TOTAL] - timer.array[TIME_FORCE] - timer.array[TIME_NEIGH] - timer.array[TIME_COMM];
printf("\n\n");
printf("# Performance Summary:\n");
printf("# Variant MPI_proc OMP_threads nsteps natoms t_total t_force t_neigh t_comm t_other performance perf/thread grep_string t_extra\n");
printf("%s %i %i %i %i %lf %lf %lf %lf %lf %lf %lf PERF_SUMMARY %lf\n\n\n", get_variant_string(), nprocs, num_threads, integrate.ntimes, natoms, timer.array[TIME_TOTAL], timer.array[TIME_FORCE], timer.array[TIME_NEIGH], timer.array[TIME_COMM], time_other, 1.0 * natoms * integrate.ntimes / timer.array[TIME_TOTAL], 1.0 * natoms * integrate.ntimes / timer.array[TIME_TOTAL] / nprocs / num_threads, timer.array[TIME_TEST]);
}
if(yaml_output)
{
output(in, atom, force, neighbor, comm, thermo, integrate, timer, screen_yaml);
}
int status = EXIT_SUCCESS;
if(check_output)
{
FILE *fp = fopen(check_output, "w");
if(!fp)
{
fprintf(stderr, "Couldn't open %s for check output\n", check_output);
status = EXIT_FAILURE;
}
int end = thermo.mstat - 1;
fprintf(fp, "%i %e %e %e %6.3lf\n", integrate.ntimes, thermo.tmparr[end], thermo.engarr[end], thermo.prsarr[end], timer.array[TIME_TOTAL]);
fclose(fp);
printf("Wrote check output to %s\n", check_output);
free(check_output);
}
delete force;
MPI_Barrier(MPI_COMM_WORLD);
MPI_Finalize();
return status;
}