Bug Summary

File:gromacs/mdlib/minimize.c
Location:line 2869, column 57
Description:Array access (from variable 'full_matrix') results in a null pointer dereference

Annotated Source Code

1/*
2 * This file is part of the GROMACS molecular simulation package.
3 *
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2004, The GROMACS development team.
6 * Copyright (c) 2013,2014, by the GROMACS development team, led by
7 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
8 * and including many others, as listed in the AUTHORS file in the
9 * top-level source directory and at http://www.gromacs.org.
10 *
11 * GROMACS is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public License
13 * as published by the Free Software Foundation; either version 2.1
14 * of the License, or (at your option) any later version.
15 *
16 * GROMACS is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
20 *
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with GROMACS; if not, see
23 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
24 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
25 *
26 * If you want to redistribute modifications to GROMACS, please
27 * consider that scientific software is very special. Version
28 * control is crucial - bugs must be traceable. We will be happy to
29 * consider code for inclusion in the official distribution, but
30 * derived work must not be called official GROMACS. Details are found
31 * in the README & COPYING files - if they are missing, get the
32 * official version at http://www.gromacs.org.
33 *
34 * To help us fund GROMACS development, we humbly ask that you cite
35 * the research papers on the package. Check out http://www.gromacs.org.
36 */
37#ifdef HAVE_CONFIG_H1
38#include <config.h>
39#endif
40
41#include <string.h>
42#include <time.h>
43#include <math.h>
44#include "gromacs/utility/cstringutil.h"
45#include "network.h"
46#include "gromacs/utility/smalloc.h"
47#include "nrnb.h"
48#include "force.h"
49#include "macros.h"
50#include "names.h"
51#include "gromacs/utility/fatalerror.h"
52#include "txtdump.h"
53#include "typedefs.h"
54#include "update.h"
55#include "constr.h"
56#include "gromacs/math/vec.h"
57#include "tgroup.h"
58#include "mdebin.h"
59#include "vsite.h"
60#include "force.h"
61#include "mdrun.h"
62#include "md_support.h"
63#include "sim_util.h"
64#include "domdec.h"
65#include "mdatoms.h"
66#include "ns.h"
67#include "mtop_util.h"
68#include "pme.h"
69#include "bondf.h"
70#include "gmx_omp_nthreads.h"
71#include "md_logging.h"
72
73#include "gromacs/fileio/confio.h"
74#include "gromacs/fileio/trajectory_writing.h"
75#include "gromacs/linearalgebra/mtxio.h"
76#include "gromacs/linearalgebra/sparsematrix.h"
77#include "gromacs/timing/wallcycle.h"
78#include "gromacs/timing/walltime_accounting.h"
79#include "gromacs/imd/imd.h"
80
81typedef struct {
82 t_state s;
83 rvec *f;
84 real epot;
85 real fnorm;
86 real fmax;
87 int a_fmax;
88} em_state_t;
89
90static em_state_t *init_em_state()
91{
92 em_state_t *ems;
93
94 snew(ems, 1)(ems) = save_calloc("ems", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 94, (1), sizeof(*(ems)));
95
96 /* does this need to be here? Should the array be declared differently (staticaly)in the state definition? */
97 snew(ems->s.lambda, efptNR)(ems->s.lambda) = save_calloc("ems->s.lambda", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 97, (efptNR), sizeof(*(ems->s.lambda)));
98
99 return ems;
100}
101
102static void print_em_start(FILE *fplog,
103 t_commrec *cr,
104 gmx_walltime_accounting_t walltime_accounting,
105 gmx_wallcycle_t wcycle,
106 const char *name)
107{
108 walltime_accounting_start(walltime_accounting);
109 wallcycle_start(wcycle, ewcRUN);
110 print_start(fplog, cr, walltime_accounting, name);
111}
112static void em_time_end(gmx_walltime_accounting_t walltime_accounting,
113 gmx_wallcycle_t wcycle)
114{
115 wallcycle_stop(wcycle, ewcRUN);
116
117 walltime_accounting_end(walltime_accounting);
118}
119
120static void sp_header(FILE *out, const char *minimizer, real ftol, int nsteps)
121{
122 fprintf(out, "\n");
123 fprintf(out, "%s:\n", minimizer);
124 fprintf(out, " Tolerance (Fmax) = %12.5e\n", ftol);
125 fprintf(out, " Number of steps = %12d\n", nsteps);
126}
127
128static void warn_step(FILE *fp, real ftol, gmx_bool bLastStep, gmx_bool bConstrain)
129{
130 char buffer[2048];
131 if (bLastStep)
132 {
133 sprintf(buffer,
134 "\nEnergy minimization reached the maximum number "
135 "of steps before the forces reached the requested "
136 "precision Fmax < %g.\n", ftol);
137 }
138 else
139 {
140 sprintf(buffer,
141 "\nEnergy minimization has stopped, but the forces have "
142 "not converged to the requested precision Fmax < %g (which "
143 "may not be possible for your system). It stopped "
144 "because the algorithm tried to make a new step whose size "
145 "was too small, or there was no change in the energy since "
146 "last step. Either way, we regard the minimization as "
147 "converged to within the available machine precision, "
148 "given your starting configuration and EM parameters.\n%s%s",
149 ftol,
150 sizeof(real) < sizeof(double) ?
151 "\nDouble precision normally gives you higher accuracy, but "
152 "this is often not needed for preparing to run molecular "
153 "dynamics.\n" :
154 "",
155 bConstrain ?
156 "You might need to increase your constraint accuracy, or turn\n"
157 "off constraints altogether (set constraints = none in mdp file)\n" :
158 "");
159 }
160 fputs(wrap_lines(buffer, 78, 0, FALSE0), fp);
161}
162
163
164
165static void print_converged(FILE *fp, const char *alg, real ftol,
166 gmx_int64_t count, gmx_bool bDone, gmx_int64_t nsteps,
167 real epot, real fmax, int nfmax, real fnorm)
168{
169 char buf[STEPSTRSIZE22];
170
171 if (bDone)
172 {
173 fprintf(fp, "\n%s converged to Fmax < %g in %s steps\n",
174 alg, ftol, gmx_step_str(count, buf));
175 }
176 else if (count < nsteps)
177 {
178 fprintf(fp, "\n%s converged to machine precision in %s steps,\n"
179 "but did not reach the requested Fmax < %g.\n",
180 alg, gmx_step_str(count, buf), ftol);
181 }
182 else
183 {
184 fprintf(fp, "\n%s did not converge to Fmax < %g in %s steps.\n",
185 alg, ftol, gmx_step_str(count, buf));
186 }
187
188#ifdef GMX_DOUBLE
189 fprintf(fp, "Potential Energy = %21.14e\n", epot);
190 fprintf(fp, "Maximum force = %21.14e on atom %d\n", fmax, nfmax+1);
191 fprintf(fp, "Norm of force = %21.14e\n", fnorm);
192#else
193 fprintf(fp, "Potential Energy = %14.7e\n", epot);
194 fprintf(fp, "Maximum force = %14.7e on atom %d\n", fmax, nfmax+1);
195 fprintf(fp, "Norm of force = %14.7e\n", fnorm);
196#endif
197}
198
199static void get_f_norm_max(t_commrec *cr,
200 t_grpopts *opts, t_mdatoms *mdatoms, rvec *f,
201 real *fnorm, real *fmax, int *a_fmax)
202{
203 double fnorm2, *sum;
204 real fmax2, fmax2_0, fam;
205 int la_max, a_max, start, end, i, m, gf;
206
207 /* This routine finds the largest force and returns it.
208 * On parallel machines the global max is taken.
209 */
210 fnorm2 = 0;
211 fmax2 = 0;
212 la_max = -1;
213 gf = 0;
214 start = 0;
215 end = mdatoms->homenr;
216 if (mdatoms->cFREEZE)
217 {
218 for (i = start; i < end; i++)
219 {
220 gf = mdatoms->cFREEZE[i];
221 fam = 0;
222 for (m = 0; m < DIM3; m++)
223 {
224 if (!opts->nFreeze[gf][m])
225 {
226 fam += sqr(f[i][m]);
227 }
228 }
229 fnorm2 += fam;
230 if (fam > fmax2)
231 {
232 fmax2 = fam;
233 la_max = i;
234 }
235 }
236 }
237 else
238 {
239 for (i = start; i < end; i++)
240 {
241 fam = norm2(f[i]);
242 fnorm2 += fam;
243 if (fam > fmax2)
244 {
245 fmax2 = fam;
246 la_max = i;
247 }
248 }
249 }
250
251 if (la_max >= 0 && DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)))
252 {
253 a_max = cr->dd->gatindex[la_max];
254 }
255 else
256 {
257 a_max = la_max;
258 }
259 if (PAR(cr)((cr)->nnodes > 1))
260 {
261 snew(sum, 2*cr->nnodes+1)(sum) = save_calloc("sum", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 261, (2*cr->nnodes+1), sizeof(*(sum)));
262 sum[2*cr->nodeid] = fmax2;
263 sum[2*cr->nodeid+1] = a_max;
264 sum[2*cr->nnodes] = fnorm2;
265 gmx_sumd(2*cr->nnodes+1, sum, cr);
266 fnorm2 = sum[2*cr->nnodes];
267 /* Determine the global maximum */
268 for (i = 0; i < cr->nnodes; i++)
269 {
270 if (sum[2*i] > fmax2)
271 {
272 fmax2 = sum[2*i];
273 a_max = (int)(sum[2*i+1] + 0.5);
274 }
275 }
276 sfree(sum)save_free("sum", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 276, (sum));
277 }
278
279 if (fnorm)
280 {
281 *fnorm = sqrt(fnorm2);
282 }
283 if (fmax)
284 {
285 *fmax = sqrt(fmax2);
286 }
287 if (a_fmax)
288 {
289 *a_fmax = a_max;
290 }
291}
292
293static void get_state_f_norm_max(t_commrec *cr,
294 t_grpopts *opts, t_mdatoms *mdatoms,
295 em_state_t *ems)
296{
297 get_f_norm_max(cr, opts, mdatoms, ems->f, &ems->fnorm, &ems->fmax, &ems->a_fmax);
298}
299
300void init_em(FILE *fplog, const char *title,
301 t_commrec *cr, t_inputrec *ir,
302 t_state *state_global, gmx_mtop_t *top_global,
303 em_state_t *ems, gmx_localtop_t **top,
304 rvec **f, rvec **f_global,
305 t_nrnb *nrnb, rvec mu_tot,
306 t_forcerec *fr, gmx_enerdata_t **enerd,
307 t_graph **graph, t_mdatoms *mdatoms, gmx_global_stat_t *gstat,
308 gmx_vsite_t *vsite, gmx_constr_t constr,
309 int nfile, const t_filenm fnm[],
310 gmx_mdoutf_t *outf, t_mdebin **mdebin,
311 int imdport, unsigned long gmx_unused__attribute__ ((unused)) Flags)
312{
313 int i;
314 real dvdl_constr;
315
316 if (fplog)
317 {
318 fprintf(fplog, "Initiating %s\n", title);
319 }
320
321 state_global->ngtc = 0;
322
323 /* Initialize lambda variables */
324 initialize_lambdas(fplog, ir, &(state_global->fep_state), state_global->lambda, NULL((void*)0));
325
326 init_nrnb(nrnb);
327
328 /* Interactive molecular dynamics */
329 init_IMD(ir, cr, top_global, fplog, 1, state_global->x,
330 nfile, fnm, NULL((void*)0), imdport, Flags);
331
332 if (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)))
333 {
334 *top = dd_init_local_top(top_global);
335
336 dd_init_local_state(cr->dd, state_global, &ems->s);
337
338 *f = NULL((void*)0);
339
340 /* Distribute the charge groups over the nodes from the master node */
341 dd_partition_system(fplog, ir->init_step, cr, TRUE1, 1,
342 state_global, top_global, ir,
343 &ems->s, &ems->f, mdatoms, *top,
344 fr, vsite, NULL((void*)0), constr,
345 nrnb, NULL((void*)0), FALSE0);
346 dd_store_state(cr->dd, &ems->s);
347
348 if (ir->nstfout)
349 {
350 snew(*f_global, top_global->natoms)(*f_global) = save_calloc("*f_global", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 350, (top_global->natoms), sizeof(*(*f_global)));
351 }
352 else
353 {
354 *f_global = NULL((void*)0);
355 }
356 *graph = NULL((void*)0);
357 }
358 else
359 {
360 snew(*f, top_global->natoms)(*f) = save_calloc("*f", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 360, (top_global->natoms), sizeof(*(*f)));
361
362 /* Just copy the state */
363 ems->s = *state_global;
364 snew(ems->s.x, ems->s.nalloc)(ems->s.x) = save_calloc("ems->s.x", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 364, (ems->s.nalloc), sizeof(*(ems->s.x)));
365 snew(ems->f, ems->s.nalloc)(ems->f) = save_calloc("ems->f", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 365, (ems->s.nalloc), sizeof(*(ems->f)));
366 for (i = 0; i < state_global->natoms; i++)
367 {
368 copy_rvec(state_global->x[i], ems->s.x[i]);
369 }
370 copy_mat(state_global->box, ems->s.box);
371
372 *top = gmx_mtop_generate_local_top(top_global, ir);
373 *f_global = *f;
374
375 forcerec_set_excl_load(fr, *top);
376
377 setup_bonded_threading(fr, &(*top)->idef);
378
379 if (ir->ePBC != epbcNONE && !fr->bMolPBC)
380 {
381 *graph = mk_graph(fplog, &((*top)->idef), 0, top_global->natoms, FALSE0, FALSE0);
382 }
383 else
384 {
385 *graph = NULL((void*)0);
386 }
387
388 atoms2md(top_global, ir, 0, NULL((void*)0), top_global->natoms, mdatoms);
389 update_mdatoms(mdatoms, state_global->lambda[efptFEP]);
390
391 if (vsite)
392 {
393 set_vsite_top(vsite, *top, mdatoms, cr);
394 }
395 }
396
397 if (constr)
398 {
399 if (ir->eConstrAlg == econtSHAKE &&
400 gmx_mtop_ftype_count(top_global, F_CONSTR) > 0)
401 {
402 gmx_fatal(FARGS0, "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 402, "Can not do energy minimization with %s, use %s\n",
403 econstr_names[econtSHAKE], econstr_names[econtLINCS]);
404 }
405
406 if (!DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)))
407 {
408 set_constraints(constr, *top, ir, mdatoms, cr);
409 }
410
411 if (!ir->bContinuation)
412 {
413 /* Constrain the starting coordinates */
414 dvdl_constr = 0;
415 constrain(PAR(cr)((cr)->nnodes > 1) ? NULL((void*)0) : fplog, TRUE1, TRUE1, constr, &(*top)->idef,
416 ir, NULL((void*)0), cr, -1, 0, mdatoms,
417 ems->s.x, ems->s.x, NULL((void*)0), fr->bMolPBC, ems->s.box,
418 ems->s.lambda[efptFEP], &dvdl_constr,
419 NULL((void*)0), NULL((void*)0), nrnb, econqCoord, FALSE0, 0, 0);
420 }
421 }
422
423 if (PAR(cr)((cr)->nnodes > 1))
424 {
425 *gstat = global_stat_init(ir);
426 }
427
428 *outf = init_mdoutf(fplog, nfile, fnm, 0, cr, ir, top_global, NULL((void*)0));
429
430 snew(*enerd, 1)(*enerd) = save_calloc("*enerd", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 430, (1), sizeof(*(*enerd)));
431 init_enerdata(top_global->groups.grps[egcENER].nr, ir->fepvals->n_lambda,
432 *enerd);
433
434 if (mdebin != NULL((void*)0))
435 {
436 /* Init bin for energy stuff */
437 *mdebin = init_mdebin(mdoutf_get_fp_ene(*outf), top_global, ir, NULL((void*)0));
438 }
439
440 clear_rvec(mu_tot);
441 calc_shifts(ems->s.box, fr->shift_vec);
442}
443
444static void finish_em(t_commrec *cr, gmx_mdoutf_t outf,
445 gmx_walltime_accounting_t walltime_accounting,
446 gmx_wallcycle_t wcycle)
447{
448 if (!(cr->duty & DUTY_PME(1<<1)))
449 {
450 /* Tell the PME only node to finish */
451 gmx_pme_send_finish(cr);
452 }
453
454 done_mdoutf(outf);
455
456 em_time_end(walltime_accounting, wcycle);
457}
458
459static void swap_em_state(em_state_t *ems1, em_state_t *ems2)
460{
461 em_state_t tmp;
462
463 tmp = *ems1;
464 *ems1 = *ems2;
465 *ems2 = tmp;
466}
467
468static void copy_em_coords(em_state_t *ems, t_state *state)
469{
470 int i;
471
472 for (i = 0; (i < state->natoms); i++)
473 {
474 copy_rvec(ems->s.x[i], state->x[i]);
475 }
476}
477
478static void write_em_traj(FILE *fplog, t_commrec *cr,
479 gmx_mdoutf_t outf,
480 gmx_bool bX, gmx_bool bF, const char *confout,
481 gmx_mtop_t *top_global,
482 t_inputrec *ir, gmx_int64_t step,
483 em_state_t *state,
484 t_state *state_global, rvec *f_global)
485{
486 int mdof_flags;
487 gmx_bool bIMDout = FALSE0;
488
489
490 /* Shall we do IMD output? */
491 if (ir->bIMD)
492 {
493 bIMDout = do_per_step(step, IMD_get_step(ir->imd->setup));
494 }
495
496 if ((bX || bF || bIMDout || confout != NULL((void*)0)) && !DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)))
497 {
498 copy_em_coords(state, state_global);
499 f_global = state->f;
500 }
501
502 mdof_flags = 0;
503 if (bX)
504 {
505 mdof_flags |= MDOF_X(1<<0);
506 }
507 if (bF)
508 {
509 mdof_flags |= MDOF_F(1<<2);
510 }
511
512 /* If we want IMD output, set appropriate MDOF flag */
513 if (ir->bIMD)
514 {
515 mdof_flags |= MDOF_IMD(1<<5);
516 }
517
518 mdoutf_write_to_trajectory_files(fplog, cr, outf, mdof_flags,
519 top_global, step, (double)step,
520 &state->s, state_global, state->f, f_global);
521
522 if (confout != NULL((void*)0) && MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
523 {
524 if (ir->ePBC != epbcNONE && !ir->bPeriodicMols && DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)))
525 {
526 /* Make molecules whole only for confout writing */
527 do_pbc_mtop(fplog, ir->ePBC, state_global->box, top_global,
528 state_global->x);
529 }
530
531 write_sto_conf_mtop(confout,
532 *top_global->name, top_global,
533 state_global->x, NULL((void*)0), ir->ePBC, state_global->box);
534 }
535}
536
537static void do_em_step(t_commrec *cr, t_inputrec *ir, t_mdatoms *md,
538 gmx_bool bMolPBC,
539 em_state_t *ems1, real a, rvec *f, em_state_t *ems2,
540 gmx_constr_t constr, gmx_localtop_t *top,
541 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
542 gmx_int64_t count)
543
544{
545 t_state *s1, *s2;
546 int i;
547 int start, end;
548 rvec *x1, *x2;
549 real dvdl_constr;
550
551 s1 = &ems1->s;
552 s2 = &ems2->s;
553
554 if (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)) && s1->ddp_count != cr->dd->ddp_count)
555 {
556 gmx_incons("state mismatch in do_em_step")_gmx_error("incons", "state mismatch in do_em_step", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 556);
557 }
558
559 s2->flags = s1->flags;
560
561 if (s2->nalloc != s1->nalloc)
562 {
563 s2->nalloc = s1->nalloc;
564 srenew(s2->x, s1->nalloc)(s2->x) = save_realloc("s2->x", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 564, (s2->x), (s1->nalloc), sizeof(*(s2->x)));
565 srenew(ems2->f, s1->nalloc)(ems2->f) = save_realloc("ems2->f", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 565, (ems2->f), (s1->nalloc), sizeof(*(ems2->f)));
566 if (s2->flags & (1<<estCGP))
567 {
568 srenew(s2->cg_p, s1->nalloc)(s2->cg_p) = save_realloc("s2->cg_p", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 568, (s2->cg_p), (s1->nalloc), sizeof(*(s2->cg_p))
);
569 }
570 }
571
572 s2->natoms = s1->natoms;
573 copy_mat(s1->box, s2->box);
574 /* Copy free energy state */
575 for (i = 0; i < efptNR; i++)
576 {
577 s2->lambda[i] = s1->lambda[i];
578 }
579 copy_mat(s1->box, s2->box);
580
581 start = 0;
582 end = md->homenr;
583
584 x1 = s1->x;
585 x2 = s2->x;
586
587#pragma omp parallel num_threads(gmx_omp_nthreads_get(emntUpdate))
588 {
589 int gf, i, m;
590
591 gf = 0;
592#pragma omp for schedule(static) nowait
593 for (i = start; i < end; i++)
594 {
595 if (md->cFREEZE)
596 {
597 gf = md->cFREEZE[i];
598 }
599 for (m = 0; m < DIM3; m++)
600 {
601 if (ir->opts.nFreeze[gf][m])
602 {
603 x2[i][m] = x1[i][m];
604 }
605 else
606 {
607 x2[i][m] = x1[i][m] + a*f[i][m];
608 }
609 }
610 }
611
612 if (s2->flags & (1<<estCGP))
613 {
614 /* Copy the CG p vector */
615 x1 = s1->cg_p;
616 x2 = s2->cg_p;
617#pragma omp for schedule(static) nowait
618 for (i = start; i < end; i++)
619 {
620 copy_rvec(x1[i], x2[i]);
621 }
622 }
623
624 if (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)))
625 {
626 s2->ddp_count = s1->ddp_count;
627 if (s2->cg_gl_nalloc < s1->cg_gl_nalloc)
628 {
629#pragma omp barrier
630 s2->cg_gl_nalloc = s1->cg_gl_nalloc;
631 srenew(s2->cg_gl, s2->cg_gl_nalloc)(s2->cg_gl) = save_realloc("s2->cg_gl", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 631, (s2->cg_gl), (s2->cg_gl_nalloc), sizeof(*(s2->
cg_gl)));
632#pragma omp barrier
633 }
634 s2->ncg_gl = s1->ncg_gl;
635#pragma omp for schedule(static) nowait
636 for (i = 0; i < s2->ncg_gl; i++)
637 {
638 s2->cg_gl[i] = s1->cg_gl[i];
639 }
640 s2->ddp_count_cg_gl = s1->ddp_count_cg_gl;
641 }
642 }
643
644 if (constr)
645 {
646 wallcycle_start(wcycle, ewcCONSTR);
647 dvdl_constr = 0;
648 constrain(NULL((void*)0), TRUE1, TRUE1, constr, &top->idef,
649 ir, NULL((void*)0), cr, count, 0, md,
650 s1->x, s2->x, NULL((void*)0), bMolPBC, s2->box,
651 s2->lambda[efptBONDED], &dvdl_constr,
652 NULL((void*)0), NULL((void*)0), nrnb, econqCoord, FALSE0, 0, 0);
653 wallcycle_stop(wcycle, ewcCONSTR);
654 }
655}
656
657static void em_dd_partition_system(FILE *fplog, int step, t_commrec *cr,
658 gmx_mtop_t *top_global, t_inputrec *ir,
659 em_state_t *ems, gmx_localtop_t *top,
660 t_mdatoms *mdatoms, t_forcerec *fr,
661 gmx_vsite_t *vsite, gmx_constr_t constr,
662 t_nrnb *nrnb, gmx_wallcycle_t wcycle)
663{
664 /* Repartition the domain decomposition */
665 wallcycle_start(wcycle, ewcDOMDEC);
666 dd_partition_system(fplog, step, cr, FALSE0, 1,
667 NULL((void*)0), top_global, ir,
668 &ems->s, &ems->f,
669 mdatoms, top, fr, vsite, NULL((void*)0), constr,
670 nrnb, wcycle, FALSE0);
671 dd_store_state(cr->dd, &ems->s);
672 wallcycle_stop(wcycle, ewcDOMDEC);
673}
674
675static void evaluate_energy(FILE *fplog, t_commrec *cr,
676 gmx_mtop_t *top_global,
677 em_state_t *ems, gmx_localtop_t *top,
678 t_inputrec *inputrec,
679 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
680 gmx_global_stat_t gstat,
681 gmx_vsite_t *vsite, gmx_constr_t constr,
682 t_fcdata *fcd,
683 t_graph *graph, t_mdatoms *mdatoms,
684 t_forcerec *fr, rvec mu_tot,
685 gmx_enerdata_t *enerd, tensor vir, tensor pres,
686 gmx_int64_t count, gmx_bool bFirst)
687{
688 real t;
689 gmx_bool bNS;
690 int nabnsb;
691 tensor force_vir, shake_vir, ekin;
692 real dvdl_constr, prescorr, enercorr, dvdlcorr;
693 real terminate = 0;
694
695 /* Set the time to the initial time, the time does not change during EM */
696 t = inputrec->init_t;
697
698 if (bFirst ||
699 (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)) && ems->s.ddp_count < cr->dd->ddp_count))
700 {
701 /* This is the first state or an old state used before the last ns */
702 bNS = TRUE1;
703 }
704 else
705 {
706 bNS = FALSE0;
707 if (inputrec->nstlist > 0)
708 {
709 bNS = TRUE1;
710 }
711 else if (inputrec->nstlist == -1)
712 {
713 nabnsb = natoms_beyond_ns_buffer(inputrec, fr, &top->cgs, NULL((void*)0), ems->s.x);
714 if (PAR(cr)((cr)->nnodes > 1))
715 {
716 gmx_sumi(1, &nabnsb, cr);
717 }
718 bNS = (nabnsb > 0);
719 }
720 }
721
722 if (vsite)
723 {
724 construct_vsites(vsite, ems->s.x, 1, NULL((void*)0),
725 top->idef.iparams, top->idef.il,
726 fr->ePBC, fr->bMolPBC, cr, ems->s.box);
727 }
728
729 if (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)) && bNS)
730 {
731 /* Repartition the domain decomposition */
732 em_dd_partition_system(fplog, count, cr, top_global, inputrec,
733 ems, top, mdatoms, fr, vsite, constr,
734 nrnb, wcycle);
735 }
736
737 /* Calc force & energy on new trial position */
738 /* do_force always puts the charge groups in the box and shifts again
739 * We do not unshift, so molecules are always whole in congrad.c
740 */
741 do_force(fplog, cr, inputrec,
742 count, nrnb, wcycle, top, &top_global->groups,
743 ems->s.box, ems->s.x, &ems->s.hist,
744 ems->f, force_vir, mdatoms, enerd, fcd,
745 ems->s.lambda, graph, fr, vsite, mu_tot, t, NULL((void*)0), NULL((void*)0), TRUE1,
746 GMX_FORCE_STATECHANGED(1<<0) | GMX_FORCE_ALLFORCES((1<<4) | (1<<6) | (1<<7)) |
747 GMX_FORCE_VIRIAL(1<<8) | GMX_FORCE_ENERGY(1<<9) |
748 (bNS ? GMX_FORCE_NS(1<<2) | GMX_FORCE_DO_LR(1<<11) : 0));
749
750 /* Clear the unused shake virial and pressure */
751 clear_mat(shake_vir);
752 clear_mat(pres);
753
754 /* Communicate stuff when parallel */
755 if (PAR(cr)((cr)->nnodes > 1) && inputrec->eI != eiNM)
756 {
757 wallcycle_start(wcycle, ewcMoveE);
758
759 global_stat(fplog, gstat, cr, enerd, force_vir, shake_vir, mu_tot,
760 inputrec, NULL((void*)0), NULL((void*)0), NULL((void*)0), 1, &terminate,
761 top_global, &ems->s, FALSE0,
762 CGLO_ENERGY(1<<6) |
763 CGLO_PRESSURE(1<<8) |
764 CGLO_CONSTRAINT(1<<9) |
765 CGLO_FIRSTITERATE(1<<11));
766
767 wallcycle_stop(wcycle, ewcMoveE);
768 }
769
770 /* Calculate long range corrections to pressure and energy */
771 calc_dispcorr(fplog, inputrec, fr, count, top_global->natoms, ems->s.box, ems->s.lambda[efptVDW],
772 pres, force_vir, &prescorr, &enercorr, &dvdlcorr);
773 enerd->term[F_DISPCORR] = enercorr;
774 enerd->term[F_EPOT] += enercorr;
775 enerd->term[F_PRES] += prescorr;
776 enerd->term[F_DVDL] += dvdlcorr;
777
778 ems->epot = enerd->term[F_EPOT];
779
780 if (constr)
781 {
782 /* Project out the constraint components of the force */
783 wallcycle_start(wcycle, ewcCONSTR);
784 dvdl_constr = 0;
785 constrain(NULL((void*)0), FALSE0, FALSE0, constr, &top->idef,
786 inputrec, NULL((void*)0), cr, count, 0, mdatoms,
787 ems->s.x, ems->f, ems->f, fr->bMolPBC, ems->s.box,
788 ems->s.lambda[efptBONDED], &dvdl_constr,
789 NULL((void*)0), &shake_vir, nrnb, econqForceDispl, FALSE0, 0, 0);
790 if (fr->bSepDVDL && fplog)
791 {
792 gmx_print_sepdvdl(fplog, "Constraints", t, dvdl_constr);
793 }
794 enerd->term[F_DVDL_CONSTR] += dvdl_constr;
795 m_add(force_vir, shake_vir, vir);
796 wallcycle_stop(wcycle, ewcCONSTR);
797 }
798 else
799 {
800 copy_mat(force_vir, vir);
801 }
802
803 clear_mat(ekin);
804 enerd->term[F_PRES] =
805 calc_pres(fr->ePBC, inputrec->nwall, ems->s.box, ekin, vir, pres);
806
807 sum_dhdl(enerd, ems->s.lambda, inputrec->fepvals);
808
809 if (EI_ENERGY_MINIMIZATION(inputrec->eI)((inputrec->eI) == eiSteep || (inputrec->eI) == eiCG ||
(inputrec->eI) == eiLBFGS))
810 {
811 get_state_f_norm_max(cr, &(inputrec->opts), mdatoms, ems);
812 }
813}
814
815static double reorder_partsum(t_commrec *cr, t_grpopts *opts, t_mdatoms *mdatoms,
816 gmx_mtop_t *mtop,
817 em_state_t *s_min, em_state_t *s_b)
818{
819 rvec *fm, *fb, *fmg;
820 t_block *cgs_gl;
821 int ncg, *cg_gl, *index, c, cg, i, a0, a1, a, gf, m;
822 double partsum;
823 unsigned char *grpnrFREEZE;
824
825 if (debug)
826 {
827 fprintf(debug, "Doing reorder_partsum\n");
828 }
829
830 fm = s_min->f;
831 fb = s_b->f;
832
833 cgs_gl = dd_charge_groups_global(cr->dd);
834 index = cgs_gl->index;
835
836 /* Collect fm in a global vector fmg.
837 * This conflicts with the spirit of domain decomposition,
838 * but to fully optimize this a much more complicated algorithm is required.
839 */
840 snew(fmg, mtop->natoms)(fmg) = save_calloc("fmg", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 840, (mtop->natoms), sizeof(*(fmg)));
841
842 ncg = s_min->s.ncg_gl;
843 cg_gl = s_min->s.cg_gl;
844 i = 0;
845 for (c = 0; c < ncg; c++)
846 {
847 cg = cg_gl[c];
848 a0 = index[cg];
849 a1 = index[cg+1];
850 for (a = a0; a < a1; a++)
851 {
852 copy_rvec(fm[i], fmg[a]);
853 i++;
854 }
855 }
856 gmx_sumgmx_sumf(mtop->natoms*3, fmg[0], cr);
857
858 /* Now we will determine the part of the sum for the cgs in state s_b */
859 ncg = s_b->s.ncg_gl;
860 cg_gl = s_b->s.cg_gl;
861 partsum = 0;
862 i = 0;
863 gf = 0;
864 grpnrFREEZE = mtop->groups.grpnr[egcFREEZE];
865 for (c = 0; c < ncg; c++)
866 {
867 cg = cg_gl[c];
868 a0 = index[cg];
869 a1 = index[cg+1];
870 for (a = a0; a < a1; a++)
871 {
872 if (mdatoms->cFREEZE && grpnrFREEZE)
873 {
874 gf = grpnrFREEZE[i];
875 }
876 for (m = 0; m < DIM3; m++)
877 {
878 if (!opts->nFreeze[gf][m])
879 {
880 partsum += (fb[i][m] - fmg[a][m])*fb[i][m];
881 }
882 }
883 i++;
884 }
885 }
886
887 sfree(fmg)save_free("fmg", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 887, (fmg));
888
889 return partsum;
890}
891
892static real pr_beta(t_commrec *cr, t_grpopts *opts, t_mdatoms *mdatoms,
893 gmx_mtop_t *mtop,
894 em_state_t *s_min, em_state_t *s_b)
895{
896 rvec *fm, *fb;
897 double sum;
898 int gf, i, m;
899
900 /* This is just the classical Polak-Ribiere calculation of beta;
901 * it looks a bit complicated since we take freeze groups into account,
902 * and might have to sum it in parallel runs.
903 */
904
905 if (!DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)) ||
906 (s_min->s.ddp_count == cr->dd->ddp_count &&
907 s_b->s.ddp_count == cr->dd->ddp_count))
908 {
909 fm = s_min->f;
910 fb = s_b->f;
911 sum = 0;
912 gf = 0;
913 /* This part of code can be incorrect with DD,
914 * since the atom ordering in s_b and s_min might differ.
915 */
916 for (i = 0; i < mdatoms->homenr; i++)
917 {
918 if (mdatoms->cFREEZE)
919 {
920 gf = mdatoms->cFREEZE[i];
921 }
922 for (m = 0; m < DIM3; m++)
923 {
924 if (!opts->nFreeze[gf][m])
925 {
926 sum += (fb[i][m] - fm[i][m])*fb[i][m];
927 }
928 }
929 }
930 }
931 else
932 {
933 /* We need to reorder cgs while summing */
934 sum = reorder_partsum(cr, opts, mdatoms, mtop, s_min, s_b);
935 }
936 if (PAR(cr)((cr)->nnodes > 1))
937 {
938 gmx_sumd(1, &sum, cr);
939 }
940
941 return sum/sqr(s_min->fnorm);
942}
943
944double do_cg(FILE *fplog, t_commrec *cr,
945 int nfile, const t_filenm fnm[],
946 const output_env_t gmx_unused__attribute__ ((unused)) oenv, gmx_bool bVerbose, gmx_bool gmx_unused__attribute__ ((unused)) bCompact,
947 int gmx_unused__attribute__ ((unused)) nstglobalcomm,
948 gmx_vsite_t *vsite, gmx_constr_t constr,
949 int gmx_unused__attribute__ ((unused)) stepout,
950 t_inputrec *inputrec,
951 gmx_mtop_t *top_global, t_fcdata *fcd,
952 t_state *state_global,
953 t_mdatoms *mdatoms,
954 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
955 gmx_edsam_t gmx_unused__attribute__ ((unused)) ed,
956 t_forcerec *fr,
957 int gmx_unused__attribute__ ((unused)) repl_ex_nst, int gmx_unused__attribute__ ((unused)) repl_ex_nex, int gmx_unused__attribute__ ((unused)) repl_ex_seed,
958 gmx_membed_t gmx_unused__attribute__ ((unused)) membed,
959 real gmx_unused__attribute__ ((unused)) cpt_period, real gmx_unused__attribute__ ((unused)) max_hours,
960 const char gmx_unused__attribute__ ((unused)) *deviceOptions,
961 int imdport,
962 unsigned long gmx_unused__attribute__ ((unused)) Flags,
963 gmx_walltime_accounting_t walltime_accounting)
964{
965 const char *CG = "Polak-Ribiere Conjugate Gradients";
966
967 em_state_t *s_min, *s_a, *s_b, *s_c;
968 gmx_localtop_t *top;
969 gmx_enerdata_t *enerd;
970 rvec *f;
971 gmx_global_stat_t gstat;
972 t_graph *graph;
973 rvec *f_global, *p, *sf, *sfm;
974 double gpa, gpb, gpc, tmp, sum[2], minstep;
975 real fnormn;
976 real stepsize;
977 real a, b, c, beta = 0.0;
978 real epot_repl = 0;
979 real pnorm;
980 t_mdebin *mdebin;
981 gmx_bool converged, foundlower;
982 rvec mu_tot;
983 gmx_bool do_log = FALSE0, do_ene = FALSE0, do_x, do_f;
984 tensor vir, pres;
985 int number_steps, neval = 0, nstcg = inputrec->nstcgsteep;
986 gmx_mdoutf_t outf;
987 int i, m, gf, step, nminstep;
988 real terminate = 0;
989
990 step = 0;
991
992 s_min = init_em_state();
993 s_a = init_em_state();
994 s_b = init_em_state();
995 s_c = init_em_state();
996
997 /* Init em and store the local state in s_min */
998 init_em(fplog, CG, cr, inputrec,
999 state_global, top_global, s_min, &top, &f, &f_global,
1000 nrnb, mu_tot, fr, &enerd, &graph, mdatoms, &gstat, vsite, constr,
1001 nfile, fnm, &outf, &mdebin, imdport, Flags);
1002
1003 /* Print to log file */
1004 print_em_start(fplog, cr, walltime_accounting, wcycle, CG);
1005
1006 /* Max number of steps */
1007 number_steps = inputrec->nsteps;
1008
1009 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1010 {
1011 sp_header(stderrstderr, CG, inputrec->em_tol, number_steps);
1012 }
1013 if (fplog)
1014 {
1015 sp_header(fplog, CG, inputrec->em_tol, number_steps);
1016 }
1017
1018 /* Call the force routine and some auxiliary (neighboursearching etc.) */
1019 /* do_force always puts the charge groups in the box and shifts again
1020 * We do not unshift, so molecules are always whole in congrad.c
1021 */
1022 evaluate_energy(fplog, cr,
1023 top_global, s_min, top,
1024 inputrec, nrnb, wcycle, gstat,
1025 vsite, constr, fcd, graph, mdatoms, fr,
1026 mu_tot, enerd, vir, pres, -1, TRUE1);
1027 where()_where("/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1027);
1028
1029 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1030 {
1031 /* Copy stuff to the energy bin for easy printing etc. */
1032 upd_mdebin(mdebin, FALSE0, FALSE0, (double)step,
1033 mdatoms->tmass, enerd, &s_min->s, inputrec->fepvals, inputrec->expandedvals, s_min->s.box,
1034 NULL((void*)0), NULL((void*)0), vir, pres, NULL((void*)0), mu_tot, constr);
1035
1036 print_ebin_header(fplog, step, step, s_min->s.lambda[efptFEP]);
1037 print_ebin(mdoutf_get_fp_ene(outf), TRUE1, FALSE0, FALSE0, fplog, step, step, eprNORMAL,
1038 TRUE1, mdebin, fcd, &(top_global->groups), &(inputrec->opts));
1039 }
1040 where()_where("/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1040);
1041
1042 /* Estimate/guess the initial stepsize */
1043 stepsize = inputrec->em_stepsize/s_min->fnorm;
1044
1045 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1046 {
1047 fprintf(stderrstderr, " F-max = %12.5e on atom %d\n",
1048 s_min->fmax, s_min->a_fmax+1);
1049 fprintf(stderrstderr, " F-Norm = %12.5e\n",
1050 s_min->fnorm/sqrt(state_global->natoms));
1051 fprintf(stderrstderr, "\n");
1052 /* and copy to the log file too... */
1053 fprintf(fplog, " F-max = %12.5e on atom %d\n",
1054 s_min->fmax, s_min->a_fmax+1);
1055 fprintf(fplog, " F-Norm = %12.5e\n",
1056 s_min->fnorm/sqrt(state_global->natoms));
1057 fprintf(fplog, "\n");
1058 }
1059 /* Start the loop over CG steps.
1060 * Each successful step is counted, and we continue until
1061 * we either converge or reach the max number of steps.
1062 */
1063 converged = FALSE0;
1064 for (step = 0; (number_steps < 0 || (number_steps >= 0 && step <= number_steps)) && !converged; step++)
1065 {
1066
1067 /* start taking steps in a new direction
1068 * First time we enter the routine, beta=0, and the direction is
1069 * simply the negative gradient.
1070 */
1071
1072 /* Calculate the new direction in p, and the gradient in this direction, gpa */
1073 p = s_min->s.cg_p;
1074 sf = s_min->f;
1075 gpa = 0;
1076 gf = 0;
1077 for (i = 0; i < mdatoms->homenr; i++)
1078 {
1079 if (mdatoms->cFREEZE)
1080 {
1081 gf = mdatoms->cFREEZE[i];
1082 }
1083 for (m = 0; m < DIM3; m++)
1084 {
1085 if (!inputrec->opts.nFreeze[gf][m])
1086 {
1087 p[i][m] = sf[i][m] + beta*p[i][m];
1088 gpa -= p[i][m]*sf[i][m];
1089 /* f is negative gradient, thus the sign */
1090 }
1091 else
1092 {
1093 p[i][m] = 0;
1094 }
1095 }
1096 }
1097
1098 /* Sum the gradient along the line across CPUs */
1099 if (PAR(cr)((cr)->nnodes > 1))
1100 {
1101 gmx_sumd(1, &gpa, cr);
1102 }
1103
1104 /* Calculate the norm of the search vector */
1105 get_f_norm_max(cr, &(inputrec->opts), mdatoms, p, &pnorm, NULL((void*)0), NULL((void*)0));
1106
1107 /* Just in case stepsize reaches zero due to numerical precision... */
1108 if (stepsize <= 0)
1109 {
1110 stepsize = inputrec->em_stepsize/pnorm;
1111 }
1112
1113 /*
1114 * Double check the value of the derivative in the search direction.
1115 * If it is positive it must be due to the old information in the
1116 * CG formula, so just remove that and start over with beta=0.
1117 * This corresponds to a steepest descent step.
1118 */
1119 if (gpa > 0)
1120 {
1121 beta = 0;
1122 step--; /* Don't count this step since we are restarting */
1123 continue; /* Go back to the beginning of the big for-loop */
1124 }
1125
1126 /* Calculate minimum allowed stepsize, before the average (norm)
1127 * relative change in coordinate is smaller than precision
1128 */
1129 minstep = 0;
1130 for (i = 0; i < mdatoms->homenr; i++)
1131 {
1132 for (m = 0; m < DIM3; m++)
1133 {
1134 tmp = fabs(s_min->s.x[i][m]);
1135 if (tmp < 1.0)
1136 {
1137 tmp = 1.0;
1138 }
1139 tmp = p[i][m]/tmp;
1140 minstep += tmp*tmp;
1141 }
1142 }
1143 /* Add up from all CPUs */
1144 if (PAR(cr)((cr)->nnodes > 1))
1145 {
1146 gmx_sumd(1, &minstep, cr);
1147 }
1148
1149 minstep = GMX_REAL_EPS5.96046448E-08/sqrt(minstep/(3*state_global->natoms));
1150
1151 if (stepsize < minstep)
1152 {
1153 converged = TRUE1;
1154 break;
1155 }
1156
1157 /* Write coordinates if necessary */
1158 do_x = do_per_step(step, inputrec->nstxout);
1159 do_f = do_per_step(step, inputrec->nstfout);
1160
1161 write_em_traj(fplog, cr, outf, do_x, do_f, NULL((void*)0),
1162 top_global, inputrec, step,
1163 s_min, state_global, f_global);
1164
1165 /* Take a step downhill.
1166 * In theory, we should minimize the function along this direction.
1167 * That is quite possible, but it turns out to take 5-10 function evaluations
1168 * for each line. However, we dont really need to find the exact minimum -
1169 * it is much better to start a new CG step in a modified direction as soon
1170 * as we are close to it. This will save a lot of energy evaluations.
1171 *
1172 * In practice, we just try to take a single step.
1173 * If it worked (i.e. lowered the energy), we increase the stepsize but
1174 * the continue straight to the next CG step without trying to find any minimum.
1175 * If it didn't work (higher energy), there must be a minimum somewhere between
1176 * the old position and the new one.
1177 *
1178 * Due to the finite numerical accuracy, it turns out that it is a good idea
1179 * to even accept a SMALL increase in energy, if the derivative is still downhill.
1180 * This leads to lower final energies in the tests I've done. / Erik
1181 */
1182 s_a->epot = s_min->epot;
1183 a = 0.0;
1184 c = a + stepsize; /* reference position along line is zero */
1185
1186 if (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)) && s_min->s.ddp_count < cr->dd->ddp_count)
1187 {
1188 em_dd_partition_system(fplog, step, cr, top_global, inputrec,
1189 s_min, top, mdatoms, fr, vsite, constr,
1190 nrnb, wcycle);
1191 }
1192
1193 /* Take a trial step (new coords in s_c) */
1194 do_em_step(cr, inputrec, mdatoms, fr->bMolPBC, s_min, c, s_min->s.cg_p, s_c,
1195 constr, top, nrnb, wcycle, -1);
1196
1197 neval++;
1198 /* Calculate energy for the trial step */
1199 evaluate_energy(fplog, cr,
1200 top_global, s_c, top,
1201 inputrec, nrnb, wcycle, gstat,
1202 vsite, constr, fcd, graph, mdatoms, fr,
1203 mu_tot, enerd, vir, pres, -1, FALSE0);
1204
1205 /* Calc derivative along line */
1206 p = s_c->s.cg_p;
1207 sf = s_c->f;
1208 gpc = 0;
1209 for (i = 0; i < mdatoms->homenr; i++)
1210 {
1211 for (m = 0; m < DIM3; m++)
1212 {
1213 gpc -= p[i][m]*sf[i][m]; /* f is negative gradient, thus the sign */
1214 }
1215 }
1216 /* Sum the gradient along the line across CPUs */
1217 if (PAR(cr)((cr)->nnodes > 1))
1218 {
1219 gmx_sumd(1, &gpc, cr);
1220 }
1221
1222 /* This is the max amount of increase in energy we tolerate */
1223 tmp = sqrt(GMX_REAL_EPS5.96046448E-08)*fabs(s_a->epot);
1224
1225 /* Accept the step if the energy is lower, or if it is not significantly higher
1226 * and the line derivative is still negative.
1227 */
1228 if (s_c->epot < s_a->epot || (gpc < 0 && s_c->epot < (s_a->epot + tmp)))
1229 {
1230 foundlower = TRUE1;
1231 /* Great, we found a better energy. Increase step for next iteration
1232 * if we are still going down, decrease it otherwise
1233 */
1234 if (gpc < 0)
1235 {
1236 stepsize *= 1.618034; /* The golden section */
1237 }
1238 else
1239 {
1240 stepsize *= 0.618034; /* 1/golden section */
1241 }
1242 }
1243 else
1244 {
1245 /* New energy is the same or higher. We will have to do some work
1246 * to find a smaller value in the interval. Take smaller step next time!
1247 */
1248 foundlower = FALSE0;
1249 stepsize *= 0.618034;
1250 }
1251
1252
1253
1254
1255 /* OK, if we didn't find a lower value we will have to locate one now - there must
1256 * be one in the interval [a=0,c].
1257 * The same thing is valid here, though: Don't spend dozens of iterations to find
1258 * the line minimum. We try to interpolate based on the derivative at the endpoints,
1259 * and only continue until we find a lower value. In most cases this means 1-2 iterations.
1260 *
1261 * I also have a safeguard for potentially really patological functions so we never
1262 * take more than 20 steps before we give up ...
1263 *
1264 * If we already found a lower value we just skip this step and continue to the update.
1265 */
1266 if (!foundlower)
1267 {
1268 nminstep = 0;
1269
1270 do
1271 {
1272 /* Select a new trial point.
1273 * If the derivatives at points a & c have different sign we interpolate to zero,
1274 * otherwise just do a bisection.
1275 */
1276 if (gpa < 0 && gpc > 0)
1277 {
1278 b = a + gpa*(a-c)/(gpc-gpa);
1279 }
1280 else
1281 {
1282 b = 0.5*(a+c);
1283 }
1284
1285 /* safeguard if interpolation close to machine accuracy causes errors:
1286 * never go outside the interval
1287 */
1288 if (b <= a || b >= c)
1289 {
1290 b = 0.5*(a+c);
1291 }
1292
1293 if (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)) && s_min->s.ddp_count != cr->dd->ddp_count)
1294 {
1295 /* Reload the old state */
1296 em_dd_partition_system(fplog, -1, cr, top_global, inputrec,
1297 s_min, top, mdatoms, fr, vsite, constr,
1298 nrnb, wcycle);
1299 }
1300
1301 /* Take a trial step to this new point - new coords in s_b */
1302 do_em_step(cr, inputrec, mdatoms, fr->bMolPBC, s_min, b, s_min->s.cg_p, s_b,
1303 constr, top, nrnb, wcycle, -1);
1304
1305 neval++;
1306 /* Calculate energy for the trial step */
1307 evaluate_energy(fplog, cr,
1308 top_global, s_b, top,
1309 inputrec, nrnb, wcycle, gstat,
1310 vsite, constr, fcd, graph, mdatoms, fr,
1311 mu_tot, enerd, vir, pres, -1, FALSE0);
1312
1313 /* p does not change within a step, but since the domain decomposition
1314 * might change, we have to use cg_p of s_b here.
1315 */
1316 p = s_b->s.cg_p;
1317 sf = s_b->f;
1318 gpb = 0;
1319 for (i = 0; i < mdatoms->homenr; i++)
1320 {
1321 for (m = 0; m < DIM3; m++)
1322 {
1323 gpb -= p[i][m]*sf[i][m]; /* f is negative gradient, thus the sign */
1324 }
1325 }
1326 /* Sum the gradient along the line across CPUs */
1327 if (PAR(cr)((cr)->nnodes > 1))
1328 {
1329 gmx_sumd(1, &gpb, cr);
1330 }
1331
1332 if (debug)
1333 {
1334 fprintf(debug, "CGE: EpotA %f EpotB %f EpotC %f gpb %f\n",
1335 s_a->epot, s_b->epot, s_c->epot, gpb);
1336 }
1337
1338 epot_repl = s_b->epot;
1339
1340 /* Keep one of the intervals based on the value of the derivative at the new point */
1341 if (gpb > 0)
1342 {
1343 /* Replace c endpoint with b */
1344 swap_em_state(s_b, s_c);
1345 c = b;
1346 gpc = gpb;
1347 }
1348 else
1349 {
1350 /* Replace a endpoint with b */
1351 swap_em_state(s_b, s_a);
1352 a = b;
1353 gpa = gpb;
1354 }
1355
1356 /*
1357 * Stop search as soon as we find a value smaller than the endpoints.
1358 * Never run more than 20 steps, no matter what.
1359 */
1360 nminstep++;
1361 }
1362 while ((epot_repl > s_a->epot || epot_repl > s_c->epot) &&
1363 (nminstep < 20));
1364
1365 if (fabs(epot_repl - s_min->epot) < fabs(s_min->epot)*GMX_REAL_EPS5.96046448E-08 ||
1366 nminstep >= 20)
1367 {
1368 /* OK. We couldn't find a significantly lower energy.
1369 * If beta==0 this was steepest descent, and then we give up.
1370 * If not, set beta=0 and restart with steepest descent before quitting.
1371 */
1372 if (beta == 0.0)
1373 {
1374 /* Converged */
1375 converged = TRUE1;
1376 break;
1377 }
1378 else
1379 {
1380 /* Reset memory before giving up */
1381 beta = 0.0;
1382 continue;
1383 }
1384 }
1385
1386 /* Select min energy state of A & C, put the best in B.
1387 */
1388 if (s_c->epot < s_a->epot)
1389 {
1390 if (debug)
1391 {
1392 fprintf(debug, "CGE: C (%f) is lower than A (%f), moving C to B\n",
1393 s_c->epot, s_a->epot);
1394 }
1395 swap_em_state(s_b, s_c);
1396 gpb = gpc;
1397 b = c;
1398 }
1399 else
1400 {
1401 if (debug)
1402 {
1403 fprintf(debug, "CGE: A (%f) is lower than C (%f), moving A to B\n",
1404 s_a->epot, s_c->epot);
1405 }
1406 swap_em_state(s_b, s_a);
1407 gpb = gpa;
1408 b = a;
1409 }
1410
1411 }
1412 else
1413 {
1414 if (debug)
1415 {
1416 fprintf(debug, "CGE: Found a lower energy %f, moving C to B\n",
1417 s_c->epot);
1418 }
1419 swap_em_state(s_b, s_c);
1420 gpb = gpc;
1421 b = c;
1422 }
1423
1424 /* new search direction */
1425 /* beta = 0 means forget all memory and restart with steepest descents. */
1426 if (nstcg && ((step % nstcg) == 0))
1427 {
1428 beta = 0.0;
1429 }
1430 else
1431 {
1432 /* s_min->fnorm cannot be zero, because then we would have converged
1433 * and broken out.
1434 */
1435
1436 /* Polak-Ribiere update.
1437 * Change to fnorm2/fnorm2_old for Fletcher-Reeves
1438 */
1439 beta = pr_beta(cr, &inputrec->opts, mdatoms, top_global, s_min, s_b);
1440 }
1441 /* Limit beta to prevent oscillations */
1442 if (fabs(beta) > 5.0)
1443 {
1444 beta = 0.0;
1445 }
1446
1447
1448 /* update positions */
1449 swap_em_state(s_min, s_b);
1450 gpa = gpb;
1451
1452 /* Print it if necessary */
1453 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1454 {
1455 if (bVerbose)
1456 {
1457 fprintf(stderrstderr, "\rStep %d, Epot=%12.6e, Fnorm=%9.3e, Fmax=%9.3e (atom %d)\n",
1458 step, s_min->epot, s_min->fnorm/sqrt(state_global->natoms),
1459 s_min->fmax, s_min->a_fmax+1);
1460 }
1461 /* Store the new (lower) energies */
1462 upd_mdebin(mdebin, FALSE0, FALSE0, (double)step,
1463 mdatoms->tmass, enerd, &s_min->s, inputrec->fepvals, inputrec->expandedvals, s_min->s.box,
1464 NULL((void*)0), NULL((void*)0), vir, pres, NULL((void*)0), mu_tot, constr);
1465
1466 do_log = do_per_step(step, inputrec->nstlog);
1467 do_ene = do_per_step(step, inputrec->nstenergy);
1468
1469 /* Prepare IMD energy record, if bIMD is TRUE. */
1470 IMD_fill_energy_record(inputrec->bIMD, inputrec->imd, enerd, step, TRUE1);
1471
1472 if (do_log)
1473 {
1474 print_ebin_header(fplog, step, step, s_min->s.lambda[efptFEP]);
1475 }
1476 print_ebin(mdoutf_get_fp_ene(outf), do_ene, FALSE0, FALSE0,
1477 do_log ? fplog : NULL((void*)0), step, step, eprNORMAL,
1478 TRUE1, mdebin, fcd, &(top_global->groups), &(inputrec->opts));
1479 }
1480
1481 /* Send energies and positions to the IMD client if bIMD is TRUE. */
1482 if (do_IMD(inputrec->bIMD, step, cr, TRUE1, state_global->box, state_global->x, inputrec, 0, wcycle) && MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1483 {
1484 IMD_send_positions(inputrec->imd);
1485 }
1486
1487 /* Stop when the maximum force lies below tolerance.
1488 * If we have reached machine precision, converged is already set to true.
1489 */
1490 converged = converged || (s_min->fmax < inputrec->em_tol);
1491
1492 } /* End of the loop */
1493
1494 /* IMD cleanup, if bIMD is TRUE. */
1495 IMD_finalize(inputrec->bIMD, inputrec->imd);
1496
1497 if (converged)
1498 {
1499 step--; /* we never took that last step in this case */
1500
1501 }
1502 if (s_min->fmax > inputrec->em_tol)
1503 {
1504 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1505 {
1506 warn_step(stderrstderr, inputrec->em_tol, step-1 == number_steps, FALSE0);
1507 warn_step(fplog, inputrec->em_tol, step-1 == number_steps, FALSE0);
1508 }
1509 converged = FALSE0;
1510 }
1511
1512 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1513 {
1514 /* If we printed energy and/or logfile last step (which was the last step)
1515 * we don't have to do it again, but otherwise print the final values.
1516 */
1517 if (!do_log)
1518 {
1519 /* Write final value to log since we didn't do anything the last step */
1520 print_ebin_header(fplog, step, step, s_min->s.lambda[efptFEP]);
1521 }
1522 if (!do_ene || !do_log)
1523 {
1524 /* Write final energy file entries */
1525 print_ebin(mdoutf_get_fp_ene(outf), !do_ene, FALSE0, FALSE0,
1526 !do_log ? fplog : NULL((void*)0), step, step, eprNORMAL,
1527 TRUE1, mdebin, fcd, &(top_global->groups), &(inputrec->opts));
1528 }
1529 }
1530
1531 /* Print some stuff... */
1532 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1533 {
1534 fprintf(stderrstderr, "\nwriting lowest energy coordinates.\n");
1535 }
1536
1537 /* IMPORTANT!
1538 * For accurate normal mode calculation it is imperative that we
1539 * store the last conformation into the full precision binary trajectory.
1540 *
1541 * However, we should only do it if we did NOT already write this step
1542 * above (which we did if do_x or do_f was true).
1543 */
1544 do_x = !do_per_step(step, inputrec->nstxout);
1545 do_f = (inputrec->nstfout > 0 && !do_per_step(step, inputrec->nstfout));
1546
1547 write_em_traj(fplog, cr, outf, do_x, do_f, ftp2fn(efSTO, nfile, fnm),
1548 top_global, inputrec, step,
1549 s_min, state_global, f_global);
1550
1551 fnormn = s_min->fnorm/sqrt(state_global->natoms);
1552
1553 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1554 {
1555 print_converged(stderrstderr, CG, inputrec->em_tol, step, converged, number_steps,
1556 s_min->epot, s_min->fmax, s_min->a_fmax, fnormn);
1557 print_converged(fplog, CG, inputrec->em_tol, step, converged, number_steps,
1558 s_min->epot, s_min->fmax, s_min->a_fmax, fnormn);
1559
1560 fprintf(fplog, "\nPerformed %d energy evaluations in total.\n", neval);
1561 }
1562
1563 finish_em(cr, outf, walltime_accounting, wcycle);
1564
1565 /* To print the actual number of steps we needed somewhere */
1566 walltime_accounting_set_nsteps_done(walltime_accounting, step);
1567
1568 return 0;
1569} /* That's all folks */
1570
1571
1572double do_lbfgs(FILE *fplog, t_commrec *cr,
1573 int nfile, const t_filenm fnm[],
1574 const output_env_t gmx_unused__attribute__ ((unused)) oenv, gmx_bool bVerbose, gmx_bool gmx_unused__attribute__ ((unused)) bCompact,
1575 int gmx_unused__attribute__ ((unused)) nstglobalcomm,
1576 gmx_vsite_t *vsite, gmx_constr_t constr,
1577 int gmx_unused__attribute__ ((unused)) stepout,
1578 t_inputrec *inputrec,
1579 gmx_mtop_t *top_global, t_fcdata *fcd,
1580 t_state *state,
1581 t_mdatoms *mdatoms,
1582 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
1583 gmx_edsam_t gmx_unused__attribute__ ((unused)) ed,
1584 t_forcerec *fr,
1585 int gmx_unused__attribute__ ((unused)) repl_ex_nst, int gmx_unused__attribute__ ((unused)) repl_ex_nex, int gmx_unused__attribute__ ((unused)) repl_ex_seed,
1586 gmx_membed_t gmx_unused__attribute__ ((unused)) membed,
1587 real gmx_unused__attribute__ ((unused)) cpt_period, real gmx_unused__attribute__ ((unused)) max_hours,
1588 const char gmx_unused__attribute__ ((unused)) *deviceOptions,
1589 int imdport,
1590 unsigned long gmx_unused__attribute__ ((unused)) Flags,
1591 gmx_walltime_accounting_t walltime_accounting)
1592{
1593 static const char *LBFGS = "Low-Memory BFGS Minimizer";
1594 em_state_t ems;
1595 gmx_localtop_t *top;
1596 gmx_enerdata_t *enerd;
1597 rvec *f;
1598 gmx_global_stat_t gstat;
1599 t_graph *graph;
1600 rvec *f_global;
1601 int ncorr, nmaxcorr, point, cp, neval, nminstep;
1602 double stepsize, gpa, gpb, gpc, tmp, minstep;
1603 real *rho, *alpha, *ff, *xx, *p, *s, *lastx, *lastf, **dx, **dg;
1604 real *xa, *xb, *xc, *fa, *fb, *fc, *xtmp, *ftmp;
1605 real a, b, c, maxdelta, delta;
1606 real diag, Epot0, Epot, EpotA, EpotB, EpotC;
1607 real dgdx, dgdg, sq, yr, beta;
1608 t_mdebin *mdebin;
1609 gmx_bool converged, first;
1610 rvec mu_tot;
1611 real fnorm, fmax;
1612 gmx_bool do_log, do_ene, do_x, do_f, foundlower, *frozen;
1613 tensor vir, pres;
1614 int start, end, number_steps;
1615 gmx_mdoutf_t outf;
1616 int i, k, m, n, nfmax, gf, step;
1617 int mdof_flags;
1618 /* not used */
1619 real terminate;
1620
1621 if (PAR(cr)((cr)->nnodes > 1))
1622 {
1623 gmx_fatal(FARGS0, "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1623, "Cannot do parallel L-BFGS Minimization - yet.\n");
1624 }
1625
1626 if (NULL((void*)0) != constr)
1627 {
1628 gmx_fatal(FARGS0, "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1628, "The combination of constraints and L-BFGS minimization is not implemented. Either do not use constraints, or use another minimizer (e.g. steepest descent).");
1629 }
1630
1631 n = 3*state->natoms;
1632 nmaxcorr = inputrec->nbfgscorr;
1633
1634 /* Allocate memory */
1635 /* Use pointers to real so we dont have to loop over both atoms and
1636 * dimensions all the time...
1637 * x/f are allocated as rvec *, so make new x0/f0 pointers-to-real
1638 * that point to the same memory.
1639 */
1640 snew(xa, n)(xa) = save_calloc("xa", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1640, (n), sizeof(*(xa)));
1641 snew(xb, n)(xb) = save_calloc("xb", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1641, (n), sizeof(*(xb)));
1642 snew(xc, n)(xc) = save_calloc("xc", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1642, (n), sizeof(*(xc)));
1643 snew(fa, n)(fa) = save_calloc("fa", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1643, (n), sizeof(*(fa)));
1644 snew(fb, n)(fb) = save_calloc("fb", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1644, (n), sizeof(*(fb)));
1645 snew(fc, n)(fc) = save_calloc("fc", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1645, (n), sizeof(*(fc)));
1646 snew(frozen, n)(frozen) = save_calloc("frozen", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1646, (n), sizeof(*(frozen)));
1647
1648 snew(p, n)(p) = save_calloc("p", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1648, (n), sizeof(*(p)));
1649 snew(lastx, n)(lastx) = save_calloc("lastx", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1649, (n), sizeof(*(lastx)));
1650 snew(lastf, n)(lastf) = save_calloc("lastf", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1650, (n), sizeof(*(lastf)));
1651 snew(rho, nmaxcorr)(rho) = save_calloc("rho", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1651, (nmaxcorr), sizeof(*(rho)));
1652 snew(alpha, nmaxcorr)(alpha) = save_calloc("alpha", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1652, (nmaxcorr), sizeof(*(alpha)));
1653
1654 snew(dx, nmaxcorr)(dx) = save_calloc("dx", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1654, (nmaxcorr), sizeof(*(dx)));
1655 for (i = 0; i < nmaxcorr; i++)
1656 {
1657 snew(dx[i], n)(dx[i]) = save_calloc("dx[i]", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1657, (n), sizeof(*(dx[i])));
1658 }
1659
1660 snew(dg, nmaxcorr)(dg) = save_calloc("dg", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1660, (nmaxcorr), sizeof(*(dg)));
1661 for (i = 0; i < nmaxcorr; i++)
1662 {
1663 snew(dg[i], n)(dg[i]) = save_calloc("dg[i]", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1663, (n), sizeof(*(dg[i])));
1664 }
1665
1666 step = 0;
1667 neval = 0;
1668
1669 /* Init em */
1670 init_em(fplog, LBFGS, cr, inputrec,
1671 state, top_global, &ems, &top, &f, &f_global,
1672 nrnb, mu_tot, fr, &enerd, &graph, mdatoms, &gstat, vsite, constr,
1673 nfile, fnm, &outf, &mdebin, imdport, Flags);
1674 /* Do_lbfgs is not completely updated like do_steep and do_cg,
1675 * so we free some memory again.
1676 */
1677 sfree(ems.s.x)save_free("ems.s.x", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1677, (ems.s.x));
1678 sfree(ems.f)save_free("ems.f", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1678, (ems.f));
1679
1680 xx = (real *)state->x;
1681 ff = (real *)f;
1682
1683 start = 0;
1684 end = mdatoms->homenr;
1685
1686 /* Print to log file */
1687 print_em_start(fplog, cr, walltime_accounting, wcycle, LBFGS);
1688
1689 do_log = do_ene = do_x = do_f = TRUE1;
1690
1691 /* Max number of steps */
1692 number_steps = inputrec->nsteps;
1693
1694 /* Create a 3*natoms index to tell whether each degree of freedom is frozen */
1695 gf = 0;
1696 for (i = start; i < end; i++)
1697 {
1698 if (mdatoms->cFREEZE)
1699 {
1700 gf = mdatoms->cFREEZE[i];
1701 }
1702 for (m = 0; m < DIM3; m++)
1703 {
1704 frozen[3*i+m] = inputrec->opts.nFreeze[gf][m];
1705 }
1706 }
1707 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1708 {
1709 sp_header(stderrstderr, LBFGS, inputrec->em_tol, number_steps);
1710 }
1711 if (fplog)
1712 {
1713 sp_header(fplog, LBFGS, inputrec->em_tol, number_steps);
1714 }
1715
1716 if (vsite)
1717 {
1718 construct_vsites(vsite, state->x, 1, NULL((void*)0),
1719 top->idef.iparams, top->idef.il,
1720 fr->ePBC, fr->bMolPBC, cr, state->box);
1721 }
1722
1723 /* Call the force routine and some auxiliary (neighboursearching etc.) */
1724 /* do_force always puts the charge groups in the box and shifts again
1725 * We do not unshift, so molecules are always whole
1726 */
1727 neval++;
1728 ems.s.x = state->x;
1729 ems.f = f;
1730 evaluate_energy(fplog, cr,
1731 top_global, &ems, top,
1732 inputrec, nrnb, wcycle, gstat,
1733 vsite, constr, fcd, graph, mdatoms, fr,
1734 mu_tot, enerd, vir, pres, -1, TRUE1);
1735 where()_where("/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1735);
1736
1737 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1738 {
1739 /* Copy stuff to the energy bin for easy printing etc. */
1740 upd_mdebin(mdebin, FALSE0, FALSE0, (double)step,
1741 mdatoms->tmass, enerd, state, inputrec->fepvals, inputrec->expandedvals, state->box,
1742 NULL((void*)0), NULL((void*)0), vir, pres, NULL((void*)0), mu_tot, constr);
1743
1744 print_ebin_header(fplog, step, step, state->lambda[efptFEP]);
1745 print_ebin(mdoutf_get_fp_ene(outf), TRUE1, FALSE0, FALSE0, fplog, step, step, eprNORMAL,
1746 TRUE1, mdebin, fcd, &(top_global->groups), &(inputrec->opts));
1747 }
1748 where()_where("/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 1748);
1749
1750 /* This is the starting energy */
1751 Epot = enerd->term[F_EPOT];
1752
1753 fnorm = ems.fnorm;
1754 fmax = ems.fmax;
1755 nfmax = ems.a_fmax;
1756
1757 /* Set the initial step.
1758 * since it will be multiplied by the non-normalized search direction
1759 * vector (force vector the first time), we scale it by the
1760 * norm of the force.
1761 */
1762
1763 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
1764 {
1765 fprintf(stderrstderr, "Using %d BFGS correction steps.\n\n", nmaxcorr);
1766 fprintf(stderrstderr, " F-max = %12.5e on atom %d\n", fmax, nfmax+1);
1767 fprintf(stderrstderr, " F-Norm = %12.5e\n", fnorm/sqrt(state->natoms));
1768 fprintf(stderrstderr, "\n");
1769 /* and copy to the log file too... */
1770 fprintf(fplog, "Using %d BFGS correction steps.\n\n", nmaxcorr);
1771 fprintf(fplog, " F-max = %12.5e on atom %d\n", fmax, nfmax+1);
1772 fprintf(fplog, " F-Norm = %12.5e\n", fnorm/sqrt(state->natoms));
1773 fprintf(fplog, "\n");
1774 }
1775
1776 point = 0;
1777 for (i = 0; i < n; i++)
1778 {
1779 if (!frozen[i])
1780 {
1781 dx[point][i] = ff[i]; /* Initial search direction */
1782 }
1783 else
1784 {
1785 dx[point][i] = 0;
1786 }
1787 }
1788
1789 stepsize = 1.0/fnorm;
1790 converged = FALSE0;
1791
1792 /* Start the loop over BFGS steps.
1793 * Each successful step is counted, and we continue until
1794 * we either converge or reach the max number of steps.
1795 */
1796
1797 ncorr = 0;
1798
1799 /* Set the gradient from the force */
1800 converged = FALSE0;
1801 for (step = 0; (number_steps < 0 || (number_steps >= 0 && step <= number_steps)) && !converged; step++)
1802 {
1803
1804 /* Write coordinates if necessary */
1805 do_x = do_per_step(step, inputrec->nstxout);
1806 do_f = do_per_step(step, inputrec->nstfout);
1807
1808 mdof_flags = 0;
1809 if (do_x)
1810 {
1811 mdof_flags |= MDOF_X(1<<0);
1812 }
1813
1814 if (do_f)
1815 {
1816 mdof_flags |= MDOF_F(1<<2);
1817 }
1818
1819 if (inputrec->bIMD)
1820 {
1821 mdof_flags |= MDOF_IMD(1<<5);
1822 }
1823
1824 mdoutf_write_to_trajectory_files(fplog, cr, outf, mdof_flags,
1825 top_global, step, (real)step, state, state, f, f);
1826
1827 /* Do the linesearching in the direction dx[point][0..(n-1)] */
1828
1829 /* pointer to current direction - point=0 first time here */
1830 s = dx[point];
1831
1832 /* calculate line gradient */
1833 for (gpa = 0, i = 0; i < n; i++)
1834 {
1835 gpa -= s[i]*ff[i];
1836 }
1837
1838 /* Calculate minimum allowed stepsize, before the average (norm)
1839 * relative change in coordinate is smaller than precision
1840 */
1841 for (minstep = 0, i = 0; i < n; i++)
1842 {
1843 tmp = fabs(xx[i]);
1844 if (tmp < 1.0)
1845 {
1846 tmp = 1.0;
1847 }
1848 tmp = s[i]/tmp;
1849 minstep += tmp*tmp;
1850 }
1851 minstep = GMX_REAL_EPS5.96046448E-08/sqrt(minstep/n);
1852
1853 if (stepsize < minstep)
1854 {
1855 converged = TRUE1;
1856 break;
1857 }
1858
1859 /* Store old forces and coordinates */
1860 for (i = 0; i < n; i++)
1861 {
1862 lastx[i] = xx[i];
1863 lastf[i] = ff[i];
1864 }
1865 Epot0 = Epot;
1866
1867 first = TRUE1;
1868
1869 for (i = 0; i < n; i++)
1870 {
1871 xa[i] = xx[i];
1872 }
1873
1874 /* Take a step downhill.
1875 * In theory, we should minimize the function along this direction.
1876 * That is quite possible, but it turns out to take 5-10 function evaluations
1877 * for each line. However, we dont really need to find the exact minimum -
1878 * it is much better to start a new BFGS step in a modified direction as soon
1879 * as we are close to it. This will save a lot of energy evaluations.
1880 *
1881 * In practice, we just try to take a single step.
1882 * If it worked (i.e. lowered the energy), we increase the stepsize but
1883 * the continue straight to the next BFGS step without trying to find any minimum.
1884 * If it didn't work (higher energy), there must be a minimum somewhere between
1885 * the old position and the new one.
1886 *
1887 * Due to the finite numerical accuracy, it turns out that it is a good idea
1888 * to even accept a SMALL increase in energy, if the derivative is still downhill.
1889 * This leads to lower final energies in the tests I've done. / Erik
1890 */
1891 foundlower = FALSE0;
1892 EpotA = Epot0;
1893 a = 0.0;
1894 c = a + stepsize; /* reference position along line is zero */
1895
1896 /* Check stepsize first. We do not allow displacements
1897 * larger than emstep.
1898 */
1899 do
1900 {
1901 c = a + stepsize;
1902 maxdelta = 0;
1903 for (i = 0; i < n; i++)
1904 {
1905 delta = c*s[i];
1906 if (delta > maxdelta)
1907 {
1908 maxdelta = delta;
1909 }
1910 }
1911 if (maxdelta > inputrec->em_stepsize)
1912 {
1913 stepsize *= 0.1;
1914 }
1915 }
1916 while (maxdelta > inputrec->em_stepsize);
1917
1918 /* Take a trial step */
1919 for (i = 0; i < n; i++)
1920 {
1921 xc[i] = lastx[i] + c*s[i];
1922 }
1923
1924 neval++;
1925 /* Calculate energy for the trial step */
1926 ems.s.x = (rvec *)xc;
1927 ems.f = (rvec *)fc;
1928 evaluate_energy(fplog, cr,
1929 top_global, &ems, top,
1930 inputrec, nrnb, wcycle, gstat,
1931 vsite, constr, fcd, graph, mdatoms, fr,
1932 mu_tot, enerd, vir, pres, step, FALSE0);
1933 EpotC = ems.epot;
1934
1935 /* Calc derivative along line */
1936 for (gpc = 0, i = 0; i < n; i++)
1937 {
1938 gpc -= s[i]*fc[i]; /* f is negative gradient, thus the sign */
1939 }
1940 /* Sum the gradient along the line across CPUs */
1941 if (PAR(cr)((cr)->nnodes > 1))
1942 {
1943 gmx_sumd(1, &gpc, cr);
1944 }
1945
1946 /* This is the max amount of increase in energy we tolerate */
1947 tmp = sqrt(GMX_REAL_EPS5.96046448E-08)*fabs(EpotA);
1948
1949 /* Accept the step if the energy is lower, or if it is not significantly higher
1950 * and the line derivative is still negative.
1951 */
1952 if (EpotC < EpotA || (gpc < 0 && EpotC < (EpotA+tmp)))
1953 {
1954 foundlower = TRUE1;
1955 /* Great, we found a better energy. Increase step for next iteration
1956 * if we are still going down, decrease it otherwise
1957 */
1958 if (gpc < 0)
1959 {
1960 stepsize *= 1.618034; /* The golden section */
1961 }
1962 else
1963 {
1964 stepsize *= 0.618034; /* 1/golden section */
1965 }
1966 }
1967 else
1968 {
1969 /* New energy is the same or higher. We will have to do some work
1970 * to find a smaller value in the interval. Take smaller step next time!
1971 */
1972 foundlower = FALSE0;
1973 stepsize *= 0.618034;
1974 }
1975
1976 /* OK, if we didn't find a lower value we will have to locate one now - there must
1977 * be one in the interval [a=0,c].
1978 * The same thing is valid here, though: Don't spend dozens of iterations to find
1979 * the line minimum. We try to interpolate based on the derivative at the endpoints,
1980 * and only continue until we find a lower value. In most cases this means 1-2 iterations.
1981 *
1982 * I also have a safeguard for potentially really patological functions so we never
1983 * take more than 20 steps before we give up ...
1984 *
1985 * If we already found a lower value we just skip this step and continue to the update.
1986 */
1987
1988 if (!foundlower)
1989 {
1990
1991 nminstep = 0;
1992 do
1993 {
1994 /* Select a new trial point.
1995 * If the derivatives at points a & c have different sign we interpolate to zero,
1996 * otherwise just do a bisection.
1997 */
1998
1999 if (gpa < 0 && gpc > 0)
2000 {
2001 b = a + gpa*(a-c)/(gpc-gpa);
2002 }
2003 else
2004 {
2005 b = 0.5*(a+c);
2006 }
2007
2008 /* safeguard if interpolation close to machine accuracy causes errors:
2009 * never go outside the interval
2010 */
2011 if (b <= a || b >= c)
2012 {
2013 b = 0.5*(a+c);
2014 }
2015
2016 /* Take a trial step */
2017 for (i = 0; i < n; i++)
2018 {
2019 xb[i] = lastx[i] + b*s[i];
2020 }
2021
2022 neval++;
2023 /* Calculate energy for the trial step */
2024 ems.s.x = (rvec *)xb;
2025 ems.f = (rvec *)fb;
2026 evaluate_energy(fplog, cr,
2027 top_global, &ems, top,
2028 inputrec, nrnb, wcycle, gstat,
2029 vsite, constr, fcd, graph, mdatoms, fr,
2030 mu_tot, enerd, vir, pres, step, FALSE0);
2031 EpotB = ems.epot;
2032
2033 fnorm = ems.fnorm;
2034
2035 for (gpb = 0, i = 0; i < n; i++)
2036 {
2037 gpb -= s[i]*fb[i]; /* f is negative gradient, thus the sign */
2038
2039 }
2040 /* Sum the gradient along the line across CPUs */
2041 if (PAR(cr)((cr)->nnodes > 1))
2042 {
2043 gmx_sumd(1, &gpb, cr);
2044 }
2045
2046 /* Keep one of the intervals based on the value of the derivative at the new point */
2047 if (gpb > 0)
2048 {
2049 /* Replace c endpoint with b */
2050 EpotC = EpotB;
2051 c = b;
2052 gpc = gpb;
2053 /* swap coord pointers b/c */
2054 xtmp = xb;
2055 ftmp = fb;
2056 xb = xc;
2057 fb = fc;
2058 xc = xtmp;
2059 fc = ftmp;
2060 }
2061 else
2062 {
2063 /* Replace a endpoint with b */
2064 EpotA = EpotB;
2065 a = b;
2066 gpa = gpb;
2067 /* swap coord pointers a/b */
2068 xtmp = xb;
2069 ftmp = fb;
2070 xb = xa;
2071 fb = fa;
2072 xa = xtmp;
2073 fa = ftmp;
2074 }
2075
2076 /*
2077 * Stop search as soon as we find a value smaller than the endpoints,
2078 * or if the tolerance is below machine precision.
2079 * Never run more than 20 steps, no matter what.
2080 */
2081 nminstep++;
2082 }
2083 while ((EpotB > EpotA || EpotB > EpotC) && (nminstep < 20));
2084
2085 if (fabs(EpotB-Epot0) < GMX_REAL_EPS5.96046448E-08 || nminstep >= 20)
2086 {
2087 /* OK. We couldn't find a significantly lower energy.
2088 * If ncorr==0 this was steepest descent, and then we give up.
2089 * If not, reset memory to restart as steepest descent before quitting.
2090 */
2091 if (ncorr == 0)
2092 {
2093 /* Converged */
2094 converged = TRUE1;
2095 break;
2096 }
2097 else
2098 {
2099 /* Reset memory */
2100 ncorr = 0;
2101 /* Search in gradient direction */
2102 for (i = 0; i < n; i++)
2103 {
2104 dx[point][i] = ff[i];
2105 }
2106 /* Reset stepsize */
2107 stepsize = 1.0/fnorm;
2108 continue;
2109 }
2110 }
2111
2112 /* Select min energy state of A & C, put the best in xx/ff/Epot
2113 */
2114 if (EpotC < EpotA)
2115 {
2116 Epot = EpotC;
2117 /* Use state C */
2118 for (i = 0; i < n; i++)
2119 {
2120 xx[i] = xc[i];
2121 ff[i] = fc[i];
2122 }
2123 stepsize = c;
2124 }
2125 else
2126 {
2127 Epot = EpotA;
2128 /* Use state A */
2129 for (i = 0; i < n; i++)
2130 {
2131 xx[i] = xa[i];
2132 ff[i] = fa[i];
2133 }
2134 stepsize = a;
2135 }
2136
2137 }
2138 else
2139 {
2140 /* found lower */
2141 Epot = EpotC;
2142 /* Use state C */
2143 for (i = 0; i < n; i++)
2144 {
2145 xx[i] = xc[i];
2146 ff[i] = fc[i];
2147 }
2148 stepsize = c;
2149 }
2150
2151 /* Update the memory information, and calculate a new
2152 * approximation of the inverse hessian
2153 */
2154
2155 /* Have new data in Epot, xx, ff */
2156 if (ncorr < nmaxcorr)
2157 {
2158 ncorr++;
2159 }
2160
2161 for (i = 0; i < n; i++)
2162 {
2163 dg[point][i] = lastf[i]-ff[i];
2164 dx[point][i] *= stepsize;
2165 }
2166
2167 dgdg = 0;
2168 dgdx = 0;
2169 for (i = 0; i < n; i++)
2170 {
2171 dgdg += dg[point][i]*dg[point][i];
2172 dgdx += dg[point][i]*dx[point][i];
2173 }
2174
2175 diag = dgdx/dgdg;
2176
2177 rho[point] = 1.0/dgdx;
2178 point++;
2179
2180 if (point >= nmaxcorr)
2181 {
2182 point = 0;
2183 }
2184
2185 /* Update */
2186 for (i = 0; i < n; i++)
2187 {
2188 p[i] = ff[i];
2189 }
2190
2191 cp = point;
2192
2193 /* Recursive update. First go back over the memory points */
2194 for (k = 0; k < ncorr; k++)
2195 {
2196 cp--;
2197 if (cp < 0)
2198 {
2199 cp = ncorr-1;
2200 }
2201
2202 sq = 0;
2203 for (i = 0; i < n; i++)
2204 {
2205 sq += dx[cp][i]*p[i];
2206 }
2207
2208 alpha[cp] = rho[cp]*sq;
2209
2210 for (i = 0; i < n; i++)
2211 {
2212 p[i] -= alpha[cp]*dg[cp][i];
2213 }
2214 }
2215
2216 for (i = 0; i < n; i++)
2217 {
2218 p[i] *= diag;
2219 }
2220
2221 /* And then go forward again */
2222 for (k = 0; k < ncorr; k++)
2223 {
2224 yr = 0;
2225 for (i = 0; i < n; i++)
2226 {
2227 yr += p[i]*dg[cp][i];
2228 }
2229
2230 beta = rho[cp]*yr;
2231 beta = alpha[cp]-beta;
2232
2233 for (i = 0; i < n; i++)
2234 {
2235 p[i] += beta*dx[cp][i];
2236 }
2237
2238 cp++;
2239 if (cp >= ncorr)
2240 {
2241 cp = 0;
2242 }
2243 }
2244
2245 for (i = 0; i < n; i++)
2246 {
2247 if (!frozen[i])
2248 {
2249 dx[point][i] = p[i];
2250 }
2251 else
2252 {
2253 dx[point][i] = 0;
2254 }
2255 }
2256
2257 stepsize = 1.0;
2258
2259 /* Test whether the convergence criterion is met */
2260 get_f_norm_max(cr, &(inputrec->opts), mdatoms, f, &fnorm, &fmax, &nfmax);
2261
2262 /* Print it if necessary */
2263 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2264 {
2265 if (bVerbose)
2266 {
2267 fprintf(stderrstderr, "\rStep %d, Epot=%12.6e, Fnorm=%9.3e, Fmax=%9.3e (atom %d)\n",
2268 step, Epot, fnorm/sqrt(state->natoms), fmax, nfmax+1);
2269 }
2270 /* Store the new (lower) energies */
2271 upd_mdebin(mdebin, FALSE0, FALSE0, (double)step,
2272 mdatoms->tmass, enerd, state, inputrec->fepvals, inputrec->expandedvals, state->box,
2273 NULL((void*)0), NULL((void*)0), vir, pres, NULL((void*)0), mu_tot, constr);
2274 do_log = do_per_step(step, inputrec->nstlog);
2275 do_ene = do_per_step(step, inputrec->nstenergy);
2276 if (do_log)
2277 {
2278 print_ebin_header(fplog, step, step, state->lambda[efptFEP]);
2279 }
2280 print_ebin(mdoutf_get_fp_ene(outf), do_ene, FALSE0, FALSE0,
2281 do_log ? fplog : NULL((void*)0), step, step, eprNORMAL,
2282 TRUE1, mdebin, fcd, &(top_global->groups), &(inputrec->opts));
2283 }
2284
2285 /* Send x and E to IMD client, if bIMD is TRUE. */
2286 if (do_IMD(inputrec->bIMD, step, cr, TRUE1, state->box, state->x, inputrec, 0, wcycle) && MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2287 {
2288 IMD_send_positions(inputrec->imd);
2289 }
2290
2291 /* Stop when the maximum force lies below tolerance.
2292 * If we have reached machine precision, converged is already set to true.
2293 */
2294
2295 converged = converged || (fmax < inputrec->em_tol);
2296
2297 } /* End of the loop */
2298
2299 /* IMD cleanup, if bIMD is TRUE. */
2300 IMD_finalize(inputrec->bIMD, inputrec->imd);
2301
2302 if (converged)
2303 {
2304 step--; /* we never took that last step in this case */
2305
2306 }
2307 if (fmax > inputrec->em_tol)
2308 {
2309 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2310 {
2311 warn_step(stderrstderr, inputrec->em_tol, step-1 == number_steps, FALSE0);
2312 warn_step(fplog, inputrec->em_tol, step-1 == number_steps, FALSE0);
2313 }
2314 converged = FALSE0;
2315 }
2316
2317 /* If we printed energy and/or logfile last step (which was the last step)
2318 * we don't have to do it again, but otherwise print the final values.
2319 */
2320 if (!do_log) /* Write final value to log since we didn't do anythin last step */
2321 {
2322 print_ebin_header(fplog, step, step, state->lambda[efptFEP]);
2323 }
2324 if (!do_ene || !do_log) /* Write final energy file entries */
2325 {
2326 print_ebin(mdoutf_get_fp_ene(outf), !do_ene, FALSE0, FALSE0,
2327 !do_log ? fplog : NULL((void*)0), step, step, eprNORMAL,
2328 TRUE1, mdebin, fcd, &(top_global->groups), &(inputrec->opts));
2329 }
2330
2331 /* Print some stuff... */
2332 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2333 {
2334 fprintf(stderrstderr, "\nwriting lowest energy coordinates.\n");
2335 }
2336
2337 /* IMPORTANT!
2338 * For accurate normal mode calculation it is imperative that we
2339 * store the last conformation into the full precision binary trajectory.
2340 *
2341 * However, we should only do it if we did NOT already write this step
2342 * above (which we did if do_x or do_f was true).
2343 */
2344 do_x = !do_per_step(step, inputrec->nstxout);
2345 do_f = !do_per_step(step, inputrec->nstfout);
2346 write_em_traj(fplog, cr, outf, do_x, do_f, ftp2fn(efSTO, nfile, fnm),
2347 top_global, inputrec, step,
2348 &ems, state, f);
2349
2350 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2351 {
2352 print_converged(stderrstderr, LBFGS, inputrec->em_tol, step, converged,
2353 number_steps, Epot, fmax, nfmax, fnorm/sqrt(state->natoms));
2354 print_converged(fplog, LBFGS, inputrec->em_tol, step, converged,
2355 number_steps, Epot, fmax, nfmax, fnorm/sqrt(state->natoms));
2356
2357 fprintf(fplog, "\nPerformed %d energy evaluations in total.\n", neval);
2358 }
2359
2360 finish_em(cr, outf, walltime_accounting, wcycle);
2361
2362 /* To print the actual number of steps we needed somewhere */
2363 walltime_accounting_set_nsteps_done(walltime_accounting, step);
2364
2365 return 0;
2366} /* That's all folks */
2367
2368
2369double do_steep(FILE *fplog, t_commrec *cr,
2370 int nfile, const t_filenm fnm[],
2371 const output_env_t gmx_unused__attribute__ ((unused)) oenv, gmx_bool bVerbose, gmx_bool gmx_unused__attribute__ ((unused)) bCompact,
2372 int gmx_unused__attribute__ ((unused)) nstglobalcomm,
2373 gmx_vsite_t *vsite, gmx_constr_t constr,
2374 int gmx_unused__attribute__ ((unused)) stepout,
2375 t_inputrec *inputrec,
2376 gmx_mtop_t *top_global, t_fcdata *fcd,
2377 t_state *state_global,
2378 t_mdatoms *mdatoms,
2379 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
2380 gmx_edsam_t gmx_unused__attribute__ ((unused)) ed,
2381 t_forcerec *fr,
2382 int gmx_unused__attribute__ ((unused)) repl_ex_nst, int gmx_unused__attribute__ ((unused)) repl_ex_nex, int gmx_unused__attribute__ ((unused)) repl_ex_seed,
2383 gmx_membed_t gmx_unused__attribute__ ((unused)) membed,
2384 real gmx_unused__attribute__ ((unused)) cpt_period, real gmx_unused__attribute__ ((unused)) max_hours,
2385 const char gmx_unused__attribute__ ((unused)) *deviceOptions,
2386 int imdport,
2387 unsigned long gmx_unused__attribute__ ((unused)) Flags,
2388 gmx_walltime_accounting_t walltime_accounting)
2389{
2390 const char *SD = "Steepest Descents";
2391 em_state_t *s_min, *s_try;
2392 rvec *f_global;
2393 gmx_localtop_t *top;
2394 gmx_enerdata_t *enerd;
2395 rvec *f;
2396 gmx_global_stat_t gstat;
2397 t_graph *graph;
2398 real stepsize, constepsize;
2399 real ustep, fnormn;
2400 gmx_mdoutf_t outf;
2401 t_mdebin *mdebin;
2402 gmx_bool bDone, bAbort, do_x, do_f;
2403 tensor vir, pres;
2404 rvec mu_tot;
2405 int nsteps;
2406 int count = 0;
2407 int steps_accepted = 0;
2408 /* not used */
2409 real terminate = 0;
2410
2411 s_min = init_em_state();
2412 s_try = init_em_state();
2413
2414 /* Init em and store the local state in s_try */
2415 init_em(fplog, SD, cr, inputrec,
2416 state_global, top_global, s_try, &top, &f, &f_global,
2417 nrnb, mu_tot, fr, &enerd, &graph, mdatoms, &gstat, vsite, constr,
2418 nfile, fnm, &outf, &mdebin, imdport, Flags);
2419
2420 /* Print to log file */
2421 print_em_start(fplog, cr, walltime_accounting, wcycle, SD);
2422
2423 /* Set variables for stepsize (in nm). This is the largest
2424 * step that we are going to make in any direction.
2425 */
2426 ustep = inputrec->em_stepsize;
2427 stepsize = 0;
2428
2429 /* Max number of steps */
2430 nsteps = inputrec->nsteps;
2431
2432 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2433 {
2434 /* Print to the screen */
2435 sp_header(stderrstderr, SD, inputrec->em_tol, nsteps);
2436 }
2437 if (fplog)
2438 {
2439 sp_header(fplog, SD, inputrec->em_tol, nsteps);
2440 }
2441
2442 /**** HERE STARTS THE LOOP ****
2443 * count is the counter for the number of steps
2444 * bDone will be TRUE when the minimization has converged
2445 * bAbort will be TRUE when nsteps steps have been performed or when
2446 * the stepsize becomes smaller than is reasonable for machine precision
2447 */
2448 count = 0;
2449 bDone = FALSE0;
2450 bAbort = FALSE0;
2451 while (!bDone && !bAbort)
2452 {
2453 bAbort = (nsteps >= 0) && (count == nsteps);
2454
2455 /* set new coordinates, except for first step */
2456 if (count > 0)
2457 {
2458 do_em_step(cr, inputrec, mdatoms, fr->bMolPBC,
2459 s_min, stepsize, s_min->f, s_try,
2460 constr, top, nrnb, wcycle, count);
2461 }
2462
2463 evaluate_energy(fplog, cr,
2464 top_global, s_try, top,
2465 inputrec, nrnb, wcycle, gstat,
2466 vsite, constr, fcd, graph, mdatoms, fr,
2467 mu_tot, enerd, vir, pres, count, count == 0);
2468
2469 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2470 {
2471 print_ebin_header(fplog, count, count, s_try->s.lambda[efptFEP]);
2472 }
2473
2474 if (count == 0)
2475 {
2476 s_min->epot = s_try->epot + 1;
2477 }
2478
2479 /* Print it if necessary */
2480 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2481 {
2482 if (bVerbose)
2483 {
2484 fprintf(stderrstderr, "Step=%5d, Dmax= %6.1e nm, Epot= %12.5e Fmax= %11.5e, atom= %d%c",
2485 count, ustep, s_try->epot, s_try->fmax, s_try->a_fmax+1,
2486 (s_try->epot < s_min->epot) ? '\n' : '\r');
2487 }
2488
2489 if (s_try->epot < s_min->epot)
2490 {
2491 /* Store the new (lower) energies */
2492 upd_mdebin(mdebin, FALSE0, FALSE0, (double)count,
2493 mdatoms->tmass, enerd, &s_try->s, inputrec->fepvals, inputrec->expandedvals,
2494 s_try->s.box, NULL((void*)0), NULL((void*)0), vir, pres, NULL((void*)0), mu_tot, constr);
2495
2496 /* Prepare IMD energy record, if bIMD is TRUE. */
2497 IMD_fill_energy_record(inputrec->bIMD, inputrec->imd, enerd, count, TRUE1);
2498
2499 print_ebin(mdoutf_get_fp_ene(outf), TRUE1,
2500 do_per_step(steps_accepted, inputrec->nstdisreout),
2501 do_per_step(steps_accepted, inputrec->nstorireout),
2502 fplog, count, count, eprNORMAL, TRUE1,
2503 mdebin, fcd, &(top_global->groups), &(inputrec->opts));
2504 fflush(fplog);
2505 }
2506 }
2507
2508 /* Now if the new energy is smaller than the previous...
2509 * or if this is the first step!
2510 * or if we did random steps!
2511 */
2512
2513 if ( (count == 0) || (s_try->epot < s_min->epot) )
2514 {
2515 steps_accepted++;
2516
2517 /* Test whether the convergence criterion is met... */
2518 bDone = (s_try->fmax < inputrec->em_tol);
2519
2520 /* Copy the arrays for force, positions and energy */
2521 /* The 'Min' array always holds the coords and forces of the minimal
2522 sampled energy */
2523 swap_em_state(s_min, s_try);
2524 if (count > 0)
2525 {
2526 ustep *= 1.2;
2527 }
2528
2529 /* Write to trn, if necessary */
2530 do_x = do_per_step(steps_accepted, inputrec->nstxout);
2531 do_f = do_per_step(steps_accepted, inputrec->nstfout);
2532 write_em_traj(fplog, cr, outf, do_x, do_f, NULL((void*)0),
2533 top_global, inputrec, count,
2534 s_min, state_global, f_global);
2535 }
2536 else
2537 {
2538 /* If energy is not smaller make the step smaller... */
2539 ustep *= 0.5;
2540
2541 if (DOMAINDECOMP(cr)(((cr)->dd != ((void*)0)) && ((cr)->nnodes >
1)) && s_min->s.ddp_count != cr->dd->ddp_count)
2542 {
2543 /* Reload the old state */
2544 em_dd_partition_system(fplog, count, cr, top_global, inputrec,
2545 s_min, top, mdatoms, fr, vsite, constr,
2546 nrnb, wcycle);
2547 }
2548 }
2549
2550 /* Determine new step */
2551 stepsize = ustep/s_min->fmax;
2552
2553 /* Check if stepsize is too small, with 1 nm as a characteristic length */
2554#ifdef GMX_DOUBLE
2555 if (count == nsteps || ustep < 1e-12)
2556#else
2557 if (count == nsteps || ustep < 1e-6)
2558#endif
2559 {
2560 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2561 {
2562 warn_step(stderrstderr, inputrec->em_tol, count == nsteps, constr != NULL((void*)0));
2563 warn_step(fplog, inputrec->em_tol, count == nsteps, constr != NULL((void*)0));
2564 }
2565 bAbort = TRUE1;
2566 }
2567
2568 /* Send IMD energies and positions, if bIMD is TRUE. */
2569 if (do_IMD(inputrec->bIMD, count, cr, TRUE1, state_global->box, state_global->x, inputrec, 0, wcycle) && MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2570 {
2571 IMD_send_positions(inputrec->imd);
2572 }
2573
2574 count++;
2575 } /* End of the loop */
2576
2577 /* IMD cleanup, if bIMD is TRUE. */
2578 IMD_finalize(inputrec->bIMD, inputrec->imd);
2579
2580 /* Print some data... */
2581 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2582 {
2583 fprintf(stderrstderr, "\nwriting lowest energy coordinates.\n");
2584 }
2585 write_em_traj(fplog, cr, outf, TRUE1, inputrec->nstfout, ftp2fn(efSTO, nfile, fnm),
2586 top_global, inputrec, count,
2587 s_min, state_global, f_global);
2588
2589 fnormn = s_min->fnorm/sqrt(state_global->natoms);
2590
2591 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2592 {
2593 print_converged(stderrstderr, SD, inputrec->em_tol, count, bDone, nsteps,
2594 s_min->epot, s_min->fmax, s_min->a_fmax, fnormn);
2595 print_converged(fplog, SD, inputrec->em_tol, count, bDone, nsteps,
2596 s_min->epot, s_min->fmax, s_min->a_fmax, fnormn);
2597 }
2598
2599 finish_em(cr, outf, walltime_accounting, wcycle);
2600
2601 /* To print the actual number of steps we needed somewhere */
2602 inputrec->nsteps = count;
2603
2604 walltime_accounting_set_nsteps_done(walltime_accounting, count);
2605
2606 return 0;
2607} /* That's all folks */
2608
2609
2610double do_nm(FILE *fplog, t_commrec *cr,
2611 int nfile, const t_filenm fnm[],
2612 const output_env_t gmx_unused__attribute__ ((unused)) oenv, gmx_bool bVerbose, gmx_bool gmx_unused__attribute__ ((unused)) bCompact,
2613 int gmx_unused__attribute__ ((unused)) nstglobalcomm,
2614 gmx_vsite_t *vsite, gmx_constr_t constr,
2615 int gmx_unused__attribute__ ((unused)) stepout,
2616 t_inputrec *inputrec,
2617 gmx_mtop_t *top_global, t_fcdata *fcd,
2618 t_state *state_global,
2619 t_mdatoms *mdatoms,
2620 t_nrnb *nrnb, gmx_wallcycle_t wcycle,
2621 gmx_edsam_t gmx_unused__attribute__ ((unused)) ed,
2622 t_forcerec *fr,
2623 int gmx_unused__attribute__ ((unused)) repl_ex_nst, int gmx_unused__attribute__ ((unused)) repl_ex_nex, int gmx_unused__attribute__ ((unused)) repl_ex_seed,
2624 gmx_membed_t gmx_unused__attribute__ ((unused)) membed,
2625 real gmx_unused__attribute__ ((unused)) cpt_period, real gmx_unused__attribute__ ((unused)) max_hours,
2626 const char gmx_unused__attribute__ ((unused)) *deviceOptions,
2627 int imdport,
2628 unsigned long gmx_unused__attribute__ ((unused)) Flags,
2629 gmx_walltime_accounting_t walltime_accounting)
2630{
2631 const char *NM = "Normal Mode Analysis";
2632 gmx_mdoutf_t outf;
2633 int natoms, atom, d;
2634 int nnodes, node;
2635 rvec *f_global;
2636 gmx_localtop_t *top;
2637 gmx_enerdata_t *enerd;
2638 rvec *f;
2639 gmx_global_stat_t gstat;
2640 t_graph *graph;
2641 real t, t0, lambda, lam0;
2642 gmx_bool bNS;
2643 tensor vir, pres;
2644 rvec mu_tot;
2645 rvec *fneg, *dfdx;
2646 gmx_bool bSparse; /* use sparse matrix storage format */
2647 size_t sz = 0;
2648 gmx_sparsematrix_t * sparse_matrix = NULL((void*)0);
2649 real * full_matrix = NULL((void*)0);
1
'full_matrix' initialized to a null pointer value
→
2650 em_state_t * state_work;
2651
2652 /* added with respect to mdrun */
2653 int i, j, k, row, col;
2654 real der_range = 10.0*sqrt(GMX_REAL_EPS5.96046448E-08);
2655 real x_min;
2656 real fnorm, fmax;
2657
2658 if (constr != NULL((void*)0))
2
←
Assuming 'constr' is equal to null
→
3
←
Taking false branch
→
2659 {
2660 gmx_fatal(FARGS0, "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 2660, "Constraints present with Normal Mode Analysis, this combination is not supported");
2661 }
2662
2663 state_work = init_em_state();
2664
2665 /* Init em and store the local state in state_minimum */
2666 init_em(fplog, NM, cr, inputrec,
2667 state_global, top_global, state_work, &top,
2668 &f, &f_global,
2669 nrnb, mu_tot, fr, &enerd, &graph, mdatoms, &gstat, vsite, constr,
2670 nfile, fnm, &outf, NULL((void*)0), imdport, Flags);
2671
2672 natoms = top_global->natoms;
2673 snew(fneg, natoms)(fneg) = save_calloc("fneg", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 2673, (natoms), sizeof(*(fneg)));
2674 snew(dfdx, natoms)(dfdx) = save_calloc("dfdx", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 2674, (natoms), sizeof(*(dfdx)));
2675
2676#ifndef GMX_DOUBLE
2677 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
4
←
Taking false branch
→
2678 {
2679 fprintf(stderrstderr,
2680 "NOTE: This version of Gromacs has been compiled in single precision,\n"
2681 " which MIGHT not be accurate enough for normal mode analysis.\n"
2682 " Gromacs now uses sparse matrix storage, so the memory requirements\n"
2683 " are fairly modest even if you recompile in double precision.\n\n");
2684 }
2685#endif
2686
2687 /* Check if we can/should use sparse storage format.
2688 *
2689 * Sparse format is only useful when the Hessian itself is sparse, which it
2690 * will be when we use a cutoff.
2691 * For small systems (n<1000) it is easier to always use full matrix format, though.
2692 */
2693 if (EEL_FULL(fr->eeltype)((((fr->eeltype) == eelPME || (fr->eeltype) == eelPMESWITCH
|| (fr->eeltype) == eelPMEUSER || (fr->eeltype) == eelPMEUSERSWITCH
|| (fr->eeltype) == eelP3M_AD) || (fr->eeltype) == eelEWALD
) || (fr->eeltype) == eelPOISSON) || fr->rlist == 0.0)
2694 {
2695 md_print_info(cr, fplog, "Non-cutoff electrostatics used, forcing full Hessian format.\n");
2696 bSparse = FALSE0;
2697 }
2698 else if (top_global->natoms < 1000)
2699 {
2700 md_print_info(cr, fplog, "Small system size (N=%d), using full Hessian format.\n", top_global->natoms);
2701 bSparse = FALSE0;
2702 }
2703 else
2704 {
2705 md_print_info(cr, fplog, "Using compressed symmetric sparse Hessian format.\n");
2706 bSparse = TRUE1;
2707 }
2708
2709 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
5
←
Taking false branch
→
2710 {
2711 sz = DIM3*top_global->natoms;
2712
2713 fprintf(stderrstderr, "Allocating Hessian memory...\n\n");
2714
2715 if (bSparse)
2716 {
2717 sparse_matrix = gmx_sparsematrix_init(sz);
2718 sparse_matrix->compressed_symmetric = TRUE1;
2719 }
2720 else
2721 {
2722 snew(full_matrix, sz*sz)(full_matrix) = save_calloc("full_matrix", "/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 2722, (sz*sz), sizeof(*(full_matrix)));
2723 }
2724 }
2725
2726 /* Initial values */
2727 t0 = inputrec->init_t;
2728 lam0 = inputrec->fepvals->init_lambda;
2729 t = t0;
2730 lambda = lam0;
2731
2732 init_nrnb(nrnb);
2733
2734 where()_where("/home/alexxy/Develop/gromacs/src/gromacs/mdlib/minimize.c"
, 2734);
2735
2736 /* Write start time and temperature */
2737 print_em_start(fplog, cr, walltime_accounting, wcycle, NM);
2738
2739 /* fudge nr of steps to nr of atoms */
2740 inputrec->nsteps = natoms*2;
2741
2742 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
6
←
Taking false branch
→
2743 {
2744 fprintf(stderrstderr, "starting normal mode calculation '%s'\n%d steps.\n\n",
2745 *(top_global->name), (int)inputrec->nsteps);
2746 }
2747
2748 nnodes = cr->nnodes;
2749
2750 /* Make evaluate_energy do a single node force calculation */
2751 cr->nnodes = 1;
2752 evaluate_energy(fplog, cr,
2753 top_global, state_work, top,
2754 inputrec, nrnb, wcycle, gstat,
2755 vsite, constr, fcd, graph, mdatoms, fr,
2756 mu_tot, enerd, vir, pres, -1, TRUE1);
2757 cr->nnodes = nnodes;
2758
2759 /* if forces are not small, warn user */
2760 get_state_f_norm_max(cr, &(inputrec->opts), mdatoms, state_work);
2761
2762 md_print_info(cr, fplog, "Maximum force:%12.5e\n", state_work->fmax);
2763 if (state_work->fmax > 1.0e-3)
7
←
Taking false branch
→
2764 {
2765 md_print_info(cr, fplog,
2766 "The force is probably not small enough to "
2767 "ensure that you are at a minimum.\n"
2768 "Be aware that negative eigenvalues may occur\n"
2769 "when the resulting matrix is diagonalized.\n\n");
2770 }
2771
2772 /***********************************************************
2773 *
2774 * Loop over all pairs in matrix
2775 *
2776 * do_force called twice. Once with positive and
2777 * once with negative displacement
2778 *
2779 ************************************************************/
2780
2781 /* Steps are divided one by one over the nodes */
2782 for (atom = cr->nodeid; atom < natoms; atom += nnodes)
8
←
Loop condition is true. Entering loop body
→
2783 {
2784
2785 for (d = 0; d < DIM3; d++)
9
←
Loop condition is true. Entering loop body
→
2786 {
2787 x_min = state_work->s.x[atom][d];
2788
2789 state_work->s.x[atom][d] = x_min - der_range;
2790
2791 /* Make evaluate_energy do a single node force calculation */
2792 cr->nnodes = 1;
2793 evaluate_energy(fplog, cr,
2794 top_global, state_work, top,
2795 inputrec, nrnb, wcycle, gstat,
2796 vsite, constr, fcd, graph, mdatoms, fr,
2797 mu_tot, enerd, vir, pres, atom*2, FALSE0);
2798
2799 for (i = 0; i < natoms; i++)
10
←
Assuming 'i' is < 'natoms'
→
11
←
Loop condition is true. Entering loop body
→
12
←
Assuming 'i' is >= 'natoms'
→
13
←
Loop condition is false. Execution continues on line 2804
→
2800 {
2801 copy_rvec(state_work->f[i], fneg[i]);
2802 }
2803
2804 state_work->s.x[atom][d] = x_min + der_range;
2805
2806 evaluate_energy(fplog, cr,
2807 top_global, state_work, top,
2808 inputrec, nrnb, wcycle, gstat,
2809 vsite, constr, fcd, graph, mdatoms, fr,
2810 mu_tot, enerd, vir, pres, atom*2+1, FALSE0);
2811 cr->nnodes = nnodes;
2812
2813 /* x is restored to original */
2814 state_work->s.x[atom][d] = x_min;
2815
2816 for (j = 0; j < natoms; j++)
14
←
Loop condition is true. Entering loop body
→
19
←
Loop condition is false. Execution continues on line 2825
→
2817 {
2818 for (k = 0; (k < DIM3); k++)
15
←
Loop condition is true. Entering loop body
→
16
←
Loop condition is true. Entering loop body
→
17
←
Loop condition is true. Entering loop body
→
18
←
Loop condition is false. Execution continues on line 2816
→
2819 {
2820 dfdx[j][k] =
2821 -(state_work->f[j][k] - fneg[j][k])/(2*der_range);
2822 }
2823 }
2824
2825 if (!MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
20
←
Taking false branch
→
2826 {
2827#ifdef GMX_MPI
2828#ifdef GMX_DOUBLE
2829#define mpi_type MPI_DOUBLETMPI_DOUBLE
2830#else
2831#define mpi_type MPI_FLOATTMPI_FLOAT
2832#endif
2833 MPI_SendtMPI_Send(dfdx[0], natoms*DIM3, mpi_type, MASTERNODE(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)), cr->nodeid,
2834 cr->mpi_comm_mygroup);
2835#endif
2836 }
2837 else
2838 {
2839 for (node = 0; (node < nnodes && atom+node < natoms); node++)
21
←
Loop condition is true. Entering loop body
→
2840 {
2841 if (node > 0)
22
←
Taking false branch
→
2842 {
2843#ifdef GMX_MPI
2844 MPI_Status stat;
2845 MPI_RecvtMPI_Recv(dfdx[0], natoms*DIM3, mpi_type, node, node,
2846 cr->mpi_comm_mygroup, &stat);
2847#undef mpi_type
2848#endif
2849 }
2850
2851 row = (atom + node)*DIM3 + d;
2852
2853 for (j = 0; j < natoms; j++)
23
←
Loop condition is true. Entering loop body
→
2854 {
2855 for (k = 0; k < DIM3; k++)
24
←
Loop condition is true. Entering loop body
→
2856 {
2857 col = j*DIM3 + k;
2858
2859 if (bSparse)
25
←
Taking false branch
→
2860 {
2861 if (col >= row && dfdx[j][k] != 0.0)
2862 {
2863 gmx_sparsematrix_increment_value(sparse_matrix,
2864 row, col, dfdx[j][k]);
2865 }
2866 }
2867 else
2868 {
2869 full_matrix[row*sz+col] = dfdx[j][k];
26
←
Array access (from variable 'full_matrix') results in a null pointer dereference
2870 }
2871 }
2872 }
2873 }
2874 }
2875
2876 if (bVerbose && fplog)
2877 {
2878 fflush(fplog);
2879 }
2880 }
2881 /* write progress */
2882 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)) && bVerbose)
2883 {
2884 fprintf(stderrstderr, "\rFinished step %d out of %d",
2885 min(atom+nnodes, natoms)(((atom+nnodes) < (natoms)) ? (atom+nnodes) : (natoms) ), natoms);
2886 fflush(stderrstderr);
2887 }
2888 }
2889
2890 if (MASTER(cr)(((cr)->nodeid == 0) || !((cr)->nnodes > 1)))
2891 {
2892 fprintf(stderrstderr, "\n\nWriting Hessian...\n");
2893 gmx_mtxio_write(ftp2fn(efMTX, nfile, fnm), sz, sz, full_matrix, sparse_matrix);
2894 }
2895
2896 finish_em(cr, outf, walltime_accounting, wcycle);
2897
2898 walltime_accounting_set_nsteps_done(walltime_accounting, natoms*2);
2899
2900 return 0;
2901}