2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2004, The GROMACS development team.
6 * Copyright (c) 2011,2012,2013,2014, 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.
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.
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.
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.
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.
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.
46 #include "gromacs/legacyheaders/copyrite.h"
47 #include "gromacs/legacyheaders/domdec.h"
48 #include "gromacs/legacyheaders/main.h"
49 #include "gromacs/legacyheaders/names.h"
50 #include "gromacs/legacyheaders/network.h"
51 #include "gromacs/math/units.h"
52 #include "gromacs/math/vec.h"
53 #include "gromacs/random/random.h"
54 #include "gromacs/utility/smalloc.h"
56 #define PROBABILITYCUTOFF 100
57 /* we don't bother evaluating if events are more rare than exp(-100) = 3.7x10^-44 */
60 ereTEMP, ereLAMBDA, ereENDSINGLE, ereTL, ereNR
62 const char *erename[ereNR] = { "temperature", "lambda", "end_single_marker", "temperature and lambda"};
63 /* end_single_marker merely notes the end of single variable replica exchange. All types higher than
64 it are multiple replica exchange methods */
65 /* Eventually, should add 'pressure', 'temperature and pressure', 'lambda_and_pressure', 'temperature_lambda_pressure'?;
66 Let's wait until we feel better about the pressure control methods giving exact ensembles. Right now, we assume constant pressure */
68 typedef struct gmx_repl_ex
88 /* these are helper arrays for replica exchange; allocated here so they
89 don't have to be allocated each time */
97 /* helper arrays to hold the quantities that are exchanged */
106 static gmx_bool repl_quantity(const gmx_multisim_t *ms,
107 struct gmx_repl_ex *re, int ere, real q)
113 snew(qall, ms->nsim);
115 gmx_sum_sim(ms->nsim, qall, ms);
118 for (s = 1; s < ms->nsim; s++)
120 if (qall[s] != qall[0])
128 /* Set the replica exchange type and quantities */
131 snew(re->q[ere], re->nrepl);
132 for (s = 0; s < ms->nsim; s++)
134 re->q[ere][s] = qall[s];
141 gmx_repl_ex_t init_replica_exchange(FILE *fplog,
142 const gmx_multisim_t *ms,
143 const t_state *state,
144 const t_inputrec *ir,
145 int nst, int nex, int init_seed)
149 struct gmx_repl_ex *re;
151 gmx_bool bLambda = FALSE;
153 fprintf(fplog, "\nInitializing Replica Exchange\n");
155 if (ms == NULL || ms->nsim == 1)
157 gmx_fatal(FARGS, "Nothing to exchange with only one replica, maybe you forgot to set the -multi option of mdrun?");
159 if (!EI_DYNAMICS(ir->eI))
161 gmx_fatal(FARGS, "Replica exchange is only supported by dynamical simulations");
162 /* Note that PAR(cr) is defined by cr->nnodes > 1, which is
163 * distinct from MULTISIM(cr). A multi-simulation only runs
164 * with real MPI parallelism, but this does not imply PAR(cr)
167 * Since we are using a dynamical integrator, the only
168 * decomposition is DD, so PAR(cr) and DOMAINDECOMP(cr) are
169 * synonymous. The only way for cr->nnodes > 1 to be true is
170 * if we are using DD. */
176 re->nrepl = ms->nsim;
177 snew(re->q, ereENDSINGLE);
179 fprintf(fplog, "Repl There are %d replicas:\n", re->nrepl);
181 check_multi_int(fplog, ms, state->natoms, "the number of atoms", FALSE);
182 check_multi_int(fplog, ms, ir->eI, "the integrator", FALSE);
183 check_multi_int64(fplog, ms, ir->init_step+ir->nsteps, "init_step+nsteps", FALSE);
184 check_multi_int64(fplog, ms, (ir->init_step+nst-1)/nst,
185 "first exchange step: init_step/-replex", FALSE);
186 check_multi_int(fplog, ms, ir->etc, "the temperature coupling", FALSE);
187 check_multi_int(fplog, ms, ir->opts.ngtc,
188 "the number of temperature coupling groups", FALSE);
189 check_multi_int(fplog, ms, ir->epc, "the pressure coupling", FALSE);
190 check_multi_int(fplog, ms, ir->efep, "free energy", FALSE);
191 check_multi_int(fplog, ms, ir->fepvals->n_lambda, "number of lambda states", FALSE);
193 re->temp = ir->opts.ref_t[0];
194 for (i = 1; (i < ir->opts.ngtc); i++)
196 if (ir->opts.ref_t[i] != re->temp)
198 fprintf(fplog, "\nWARNING: The temperatures of the different temperature coupling groups are not identical\n\n");
199 fprintf(stderr, "\nWARNING: The temperatures of the different temperature coupling groups are not identical\n\n");
204 bTemp = repl_quantity(ms, re, ereTEMP, re->temp);
205 if (ir->efep != efepNO)
207 bLambda = repl_quantity(ms, re, ereLAMBDA, (real)ir->fepvals->init_fep_state);
209 if (re->type == -1) /* nothing was assigned */
211 gmx_fatal(FARGS, "The properties of the %d systems are all the same, there is nothing to exchange", re->nrepl);
213 if (bLambda && bTemp)
220 please_cite(fplog, "Sugita1999a");
221 if (ir->epc != epcNO)
224 fprintf(fplog, "Repl Using Constant Pressure REMD.\n");
225 please_cite(fplog, "Okabe2001a");
227 if (ir->etc == etcBERENDSEN)
229 gmx_fatal(FARGS, "REMD with the %s thermostat does not produce correct potential energy distributions, consider using the %s thermostat instead",
230 ETCOUPLTYPE(ir->etc), ETCOUPLTYPE(etcVRESCALE));
235 if (ir->fepvals->delta_lambda != 0) /* check this? */
237 gmx_fatal(FARGS, "delta_lambda is not zero");
242 snew(re->pres, re->nrepl);
243 if (ir->epct == epctSURFACETENSION)
245 pres = ir->ref_p[ZZ][ZZ];
251 for (i = 0; i < DIM; i++)
253 if (ir->compress[i][i] != 0)
255 pres += ir->ref_p[i][i];
261 re->pres[re->repl] = pres;
262 gmx_sum_sim(re->nrepl, re->pres, ms);
265 /* Make an index for increasing replica order */
266 /* only makes sense if one or the other is varying, not both!
267 if both are varying, we trust the order the person gave. */
268 snew(re->ind, re->nrepl);
269 for (i = 0; i < re->nrepl; i++)
274 if (re->type < ereENDSINGLE)
277 for (i = 0; i < re->nrepl; i++)
279 for (j = i+1; j < re->nrepl; j++)
281 if (re->q[re->type][re->ind[j]] < re->q[re->type][re->ind[i]])
283 /* Unordered replicas are supposed to work, but there
284 * is still an issues somewhere.
285 * Note that at this point still re->ind[i]=i.
287 gmx_fatal(FARGS, "Replicas with indices %d < %d have %ss %g > %g, please order your replicas on increasing %s",
290 re->q[re->type][i], re->q[re->type][j],
294 re->ind[i] = re->ind[j];
297 else if (re->q[re->type][re->ind[j]] == re->q[re->type][re->ind[i]])
299 gmx_fatal(FARGS, "Two replicas have identical %ss", erename[re->type]);
305 /* keep track of all the swaps, starting with the initial placement. */
306 snew(re->allswaps, re->nrepl);
307 for (i = 0; i < re->nrepl; i++)
309 re->allswaps[i] = re->ind[i];
315 fprintf(fplog, "\nReplica exchange in temperature\n");
316 for (i = 0; i < re->nrepl; i++)
318 fprintf(fplog, " %5.1f", re->q[re->type][re->ind[i]]);
320 fprintf(fplog, "\n");
323 fprintf(fplog, "\nReplica exchange in lambda\n");
324 for (i = 0; i < re->nrepl; i++)
326 fprintf(fplog, " %3d", (int)re->q[re->type][re->ind[i]]);
328 fprintf(fplog, "\n");
331 fprintf(fplog, "\nReplica exchange in temperature and lambda state\n");
332 for (i = 0; i < re->nrepl; i++)
334 fprintf(fplog, " %5.1f", re->q[ereTEMP][re->ind[i]]);
336 fprintf(fplog, "\n");
337 for (i = 0; i < re->nrepl; i++)
339 fprintf(fplog, " %5d", (int)re->q[ereLAMBDA][re->ind[i]]);
341 fprintf(fplog, "\n");
344 gmx_incons("Unknown replica exchange quantity");
348 fprintf(fplog, "\nRepl p");
349 for (i = 0; i < re->nrepl; i++)
351 fprintf(fplog, " %5.2f", re->pres[re->ind[i]]);
354 for (i = 0; i < re->nrepl; i++)
356 if ((i > 0) && (re->pres[re->ind[i]] < re->pres[re->ind[i-1]]))
358 fprintf(fplog, "\nWARNING: The reference pressures decrease with increasing temperatures\n\n");
359 fprintf(stderr, "\nWARNING: The reference pressures decrease with increasing temperatures\n\n");
368 re->seed = (int)gmx_rng_make_seed();
374 gmx_sumi_sim(1, &(re->seed), ms);
378 re->seed = init_seed;
380 fprintf(fplog, "\nReplica exchange interval: %d\n", re->nst);
381 fprintf(fplog, "\nReplica random seed: %d\n", re->seed);
382 re->rng = gmx_rng_init(re->seed);
387 snew(re->prob_sum, re->nrepl);
388 snew(re->nexchange, re->nrepl);
389 snew(re->nmoves, re->nrepl);
390 for (i = 0; i < re->nrepl; i++)
392 snew(re->nmoves[i], re->nrepl);
394 fprintf(fplog, "Replica exchange information below: x=exchange, pr=probability\n");
396 /* generate space for the helper functions so we don't have to snew each time */
398 snew(re->destinations, re->nrepl);
399 snew(re->incycle, re->nrepl);
400 snew(re->tmpswap, re->nrepl);
401 snew(re->cyclic, re->nrepl);
402 snew(re->order, re->nrepl);
403 for (i = 0; i < re->nrepl; i++)
405 snew(re->cyclic[i], re->nrepl);
406 snew(re->order[i], re->nrepl);
408 /* allocate space for the functions storing the data for the replicas */
409 /* not all of these arrays needed in all cases, but they don't take
410 up much space, since the max size is nrepl**2 */
411 snew(re->prob, re->nrepl);
412 snew(re->bEx, re->nrepl);
413 snew(re->beta, re->nrepl);
414 snew(re->Vol, re->nrepl);
415 snew(re->Epot, re->nrepl);
416 snew(re->de, re->nrepl);
417 for (i = 0; i < re->nrepl; i++)
419 snew(re->de[i], re->nrepl);
425 static void exchange_reals(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, real *v, int n)
435 MPI_Sendrecv(v, n*sizeof(real),MPI_BYTE,MSRANK(ms,b),0,
436 buf,n*sizeof(real),MPI_BYTE,MSRANK(ms,b),0,
437 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
442 MPI_Isend(v, n*sizeof(real), MPI_BYTE, MSRANK(ms, b), 0,
443 ms->mpi_comm_masters, &mpi_req);
444 MPI_Recv(buf, n*sizeof(real), MPI_BYTE, MSRANK(ms, b), 0,
445 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
446 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
449 for (i = 0; i < n; i++)
458 static void exchange_ints(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, int *v, int n)
468 MPI_Sendrecv(v, n*sizeof(int),MPI_BYTE,MSRANK(ms,b),0,
469 buf,n*sizeof(int),MPI_BYTE,MSRANK(ms,b),0,
470 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
475 MPI_Isend(v, n*sizeof(int), MPI_BYTE, MSRANK(ms, b), 0,
476 ms->mpi_comm_masters, &mpi_req);
477 MPI_Recv(buf, n*sizeof(int), MPI_BYTE, MSRANK(ms, b), 0,
478 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
479 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
482 for (i = 0; i < n; i++)
490 static void exchange_doubles(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, double *v, int n)
500 MPI_Sendrecv(v, n*sizeof(double),MPI_BYTE,MSRANK(ms,b),0,
501 buf,n*sizeof(double),MPI_BYTE,MSRANK(ms,b),0,
502 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
507 MPI_Isend(v, n*sizeof(double), MPI_BYTE, MSRANK(ms, b), 0,
508 ms->mpi_comm_masters, &mpi_req);
509 MPI_Recv(buf, n*sizeof(double), MPI_BYTE, MSRANK(ms, b), 0,
510 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
511 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
514 for (i = 0; i < n; i++)
522 static void exchange_rvecs(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, rvec *v, int n)
532 MPI_Sendrecv(v[0], n*sizeof(rvec),MPI_BYTE,MSRANK(ms,b),0,
533 buf[0],n*sizeof(rvec),MPI_BYTE,MSRANK(ms,b),0,
534 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
539 MPI_Isend(v[0], n*sizeof(rvec), MPI_BYTE, MSRANK(ms, b), 0,
540 ms->mpi_comm_masters, &mpi_req);
541 MPI_Recv(buf[0], n*sizeof(rvec), MPI_BYTE, MSRANK(ms, b), 0,
542 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
543 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
546 for (i = 0; i < n; i++)
548 copy_rvec(buf[i], v[i]);
554 static void exchange_state(const gmx_multisim_t *ms, int b, t_state *state)
556 /* When t_state changes, this code should be updated. */
558 ngtc = state->ngtc * state->nhchainlength;
559 nnhpres = state->nnhpres* state->nhchainlength;
560 exchange_rvecs(ms, b, state->box, DIM);
561 exchange_rvecs(ms, b, state->box_rel, DIM);
562 exchange_rvecs(ms, b, state->boxv, DIM);
563 exchange_reals(ms, b, &(state->veta), 1);
564 exchange_reals(ms, b, &(state->vol0), 1);
565 exchange_rvecs(ms, b, state->svir_prev, DIM);
566 exchange_rvecs(ms, b, state->fvir_prev, DIM);
567 exchange_rvecs(ms, b, state->pres_prev, DIM);
568 exchange_doubles(ms, b, state->nosehoover_xi, ngtc);
569 exchange_doubles(ms, b, state->nosehoover_vxi, ngtc);
570 exchange_doubles(ms, b, state->nhpres_xi, nnhpres);
571 exchange_doubles(ms, b, state->nhpres_vxi, nnhpres);
572 exchange_doubles(ms, b, state->therm_integral, state->ngtc);
573 exchange_rvecs(ms, b, state->x, state->natoms);
574 exchange_rvecs(ms, b, state->v, state->natoms);
575 exchange_rvecs(ms, b, state->sd_X, state->natoms);
578 static void copy_rvecs(rvec *s, rvec *d, int n)
584 for (i = 0; i < n; i++)
586 copy_rvec(s[i], d[i]);
591 static void copy_doubles(const double *s, double *d, int n)
597 for (i = 0; i < n; i++)
604 static void copy_reals(const real *s, real *d, int n)
610 for (i = 0; i < n; i++)
617 static void copy_ints(const int *s, int *d, int n)
623 for (i = 0; i < n; i++)
630 #define scopy_rvecs(v, n) copy_rvecs(state->v, state_local->v, n);
631 #define scopy_doubles(v, n) copy_doubles(state->v, state_local->v, n);
632 #define scopy_reals(v, n) copy_reals(state->v, state_local->v, n);
633 #define scopy_ints(v, n) copy_ints(state->v, state_local->v, n);
635 static void copy_state_nonatomdata(t_state *state, t_state *state_local)
637 /* When t_state changes, this code should be updated. */
639 ngtc = state->ngtc * state->nhchainlength;
640 nnhpres = state->nnhpres* state->nhchainlength;
641 scopy_rvecs(box, DIM);
642 scopy_rvecs(box_rel, DIM);
643 scopy_rvecs(boxv, DIM);
644 state_local->veta = state->veta;
645 state_local->vol0 = state->vol0;
646 scopy_rvecs(svir_prev, DIM);
647 scopy_rvecs(fvir_prev, DIM);
648 scopy_rvecs(pres_prev, DIM);
649 scopy_doubles(nosehoover_xi, ngtc);
650 scopy_doubles(nosehoover_vxi, ngtc);
651 scopy_doubles(nhpres_xi, nnhpres);
652 scopy_doubles(nhpres_vxi, nnhpres);
653 scopy_doubles(therm_integral, state->ngtc);
654 scopy_rvecs(x, state->natoms);
655 scopy_rvecs(v, state->natoms);
656 scopy_rvecs(sd_X, state->natoms);
657 copy_ints(&(state->fep_state), &(state_local->fep_state), 1);
658 scopy_reals(lambda, efptNR);
661 static void scale_velocities(t_state *state, real fac)
667 for (i = 0; i < state->natoms; i++)
669 svmul(fac, state->v[i], state->v[i]);
674 static void print_transition_matrix(FILE *fplog, int n, int **nmoves, int *nattempt)
679 ntot = nattempt[0] + nattempt[1];
680 fprintf(fplog, "\n");
681 fprintf(fplog, "Repl");
682 for (i = 0; i < n; i++)
684 fprintf(fplog, " "); /* put the title closer to the center */
686 fprintf(fplog, "Empirical Transition Matrix\n");
688 fprintf(fplog, "Repl");
689 for (i = 0; i < n; i++)
691 fprintf(fplog, "%8d", (i+1));
693 fprintf(fplog, "\n");
695 for (i = 0; i < n; i++)
697 fprintf(fplog, "Repl");
698 for (j = 0; j < n; j++)
701 if (nmoves[i][j] > 0)
703 Tprint = nmoves[i][j]/(2.0*ntot);
705 fprintf(fplog, "%8.4f", Tprint);
707 fprintf(fplog, "%3d\n", i);
711 static void print_ind(FILE *fplog, const char *leg, int n, int *ind, gmx_bool *bEx)
715 fprintf(fplog, "Repl %2s %2d", leg, ind[0]);
716 for (i = 1; i < n; i++)
718 fprintf(fplog, " %c %2d", (bEx != 0 && bEx[i]) ? 'x' : ' ', ind[i]);
720 fprintf(fplog, "\n");
723 static void print_allswitchind(FILE *fplog, int n, int *pind, int *allswaps, int *tmpswap)
727 for (i = 0; i < n; i++)
729 tmpswap[i] = allswaps[i];
731 for (i = 0; i < n; i++)
733 allswaps[i] = tmpswap[pind[i]];
736 fprintf(fplog, "\nAccepted Exchanges: ");
737 for (i = 0; i < n; i++)
739 fprintf(fplog, "%d ", pind[i]);
741 fprintf(fplog, "\n");
743 /* the "Order After Exchange" is the state label corresponding to the configuration that
744 started in state listed in order, i.e.
749 configuration starting in simulation 3 is now in simulation 0,
750 configuration starting in simulation 0 is now in simulation 1,
751 configuration starting in simulation 1 is now in simulation 2,
752 configuration starting in simulation 2 is now in simulation 3
754 fprintf(fplog, "Order After Exchange: ");
755 for (i = 0; i < n; i++)
757 fprintf(fplog, "%d ", allswaps[i]);
759 fprintf(fplog, "\n\n");
762 static void print_prob(FILE *fplog, const char *leg, int n, real *prob)
767 fprintf(fplog, "Repl %2s ", leg);
768 for (i = 1; i < n; i++)
772 sprintf(buf, "%4.2f", prob[i]);
773 fprintf(fplog, " %3s", buf[0] == '1' ? "1.0" : buf+1);
780 fprintf(fplog, "\n");
783 static void print_count(FILE *fplog, const char *leg, int n, int *count)
787 fprintf(fplog, "Repl %2s ", leg);
788 for (i = 1; i < n; i++)
790 fprintf(fplog, " %4d", count[i]);
792 fprintf(fplog, "\n");
795 static real calc_delta(FILE *fplog, gmx_bool bPrint, struct gmx_repl_ex *re, int a, int b, int ap, int bp)
798 real ediff, dpV, delta = 0;
799 real *Epot = re->Epot;
802 real *beta = re->beta;
804 /* Two cases; we are permuted and not. In all cases, setting ap = a and bp = b will reduce
805 to the non permuted case */
811 * Okabe et. al. Chem. Phys. Lett. 335 (2001) 435-439
813 ediff = Epot[b] - Epot[a];
814 delta = -(beta[bp] - beta[ap])*ediff;
817 /* two cases: when we are permuted, and not. */
819 ediff = E_new - E_old
820 = [H_b(x_a) + H_a(x_b)] - [H_b(x_b) + H_a(x_a)]
821 = [H_b(x_a) - H_a(x_a)] + [H_a(x_b) - H_b(x_b)]
822 = de[b][a] + de[a][b] */
825 ediff = E_new - E_old
826 = [H_bp(x_a) + H_ap(x_b)] - [H_bp(x_b) + H_ap(x_a)]
827 = [H_bp(x_a) - H_ap(x_a)] + [H_ap(x_b) - H_bp(x_b)]
828 = [H_bp(x_a) - H_a(x_a) + H_a(x_a) - H_ap(x_a)] + [H_ap(x_b) - H_b(x_b) + H_b(x_b) - H_bp(x_b)]
829 = [H_bp(x_a) - H_a(x_a)] - [H_ap(x_a) - H_a(x_a)] + [H_ap(x_b) - H_b(x_b)] - H_bp(x_b) - H_b(x_b)]
830 = (de[bp][a] - de[ap][a]) + (de[ap][b] - de[bp][b]) */
831 /* but, in the current code implementation, we flip configurations, not indices . . .
832 So let's examine that.
833 = [H_b(x_ap) - H_a(x_a)] - [H_a(x_ap) - H_a(x_a)] + [H_a(x_bp) - H_b(x_b)] - H_b(x_bp) - H_b(x_b)]
834 = [H_b(x_ap) - H_a(x_ap)] + [H_a(x_bp) - H_b(x_pb)]
835 = (de[b][ap] - de[a][ap]) + (de[a][bp] - de[b][bp]
836 So, if we exchange b<=> bp and a<=> ap, we return to the same result.
837 So the simple solution is to flip the
838 position of perturbed and original indices in the tests.
841 ediff = (de[bp][a] - de[ap][a]) + (de[ap][b] - de[bp][b]);
842 delta = ediff*beta[a]; /* assume all same temperature in this case */
846 /* delta = reduced E_new - reduced E_old
847 = [beta_b H_b(x_a) + beta_a H_a(x_b)] - [beta_b H_b(x_b) + beta_a H_a(x_a)]
848 = [beta_b H_b(x_a) - beta_a H_a(x_a)] + [beta_a H_a(x_b) - beta_b H_b(x_b)]
849 = [beta_b dH_b(x_a) + beta_b H_a(x_a) - beta_a H_a(x_a)] +
850 [beta_a dH_a(x_b) + beta_a H_b(x_b) - beta_b H_b(x_b)]
851 = [beta_b dH_b(x_a) + [beta_a dH_a(x_b) +
852 beta_b (H_a(x_a) - H_b(x_b)]) - beta_a (H_a(x_a) - H_b(x_b))
853 = beta_b dH_b(x_a) + beta_a dH_a(x_b) - (beta_b - beta_a)(H_b(x_b) - H_a(x_a) */
854 /* delta = beta[b]*de[b][a] + beta[a]*de[a][b] - (beta[b] - beta[a])*(Epot[b] - Epot[a]; */
855 /* permuted (big breath!) */
856 /* delta = reduced E_new - reduced E_old
857 = [beta_bp H_bp(x_a) + beta_ap H_ap(x_b)] - [beta_bp H_bp(x_b) + beta_ap H_ap(x_a)]
858 = [beta_bp H_bp(x_a) - beta_ap H_ap(x_a)] + [beta_ap H_ap(x_b) - beta_bp H_bp(x_b)]
859 = [beta_bp H_bp(x_a) - beta_ap H_ap(x_a)] + [beta_ap H_ap(x_b) - beta_bp H_bp(x_b)]
860 - beta_pb H_a(x_a) + beta_ap H_a(x_a) + beta_pb H_a(x_a) - beta_ap H_a(x_a)
861 - beta_ap H_b(x_b) + beta_bp H_b(x_b) + beta_ap H_b(x_b) - beta_bp H_b(x_b)
862 = [(beta_bp H_bp(x_a) - beta_bp H_a(x_a)) - (beta_ap H_ap(x_a) - beta_ap H_a(x_a))] +
863 [(beta_ap H_ap(x_b) - beta_ap H_b(x_b)) - (beta_bp H_bp(x_b) - beta_bp H_b(x_b))]
864 + beta_pb H_a(x_a) - beta_ap H_a(x_a) + beta_ap H_b(x_b) - beta_bp H_b(x_b)
865 = [beta_bp (H_bp(x_a) - H_a(x_a)) - beta_ap (H_ap(x_a) - H_a(x_a))] +
866 [beta_ap (H_ap(x_b) - H_b(x_b)) - beta_bp (H_bp(x_b) - H_b(x_b))]
867 + beta_pb (H_a(x_a) - H_b(x_b)) - beta_ap (H_a(x_a) - H_b(x_b))
868 = ([beta_bp de[bp][a] - beta_ap de[ap][a]) + beta_ap de[ap][b] - beta_bp de[bp][b])
869 + (beta_pb-beta_ap)(H_a(x_a) - H_b(x_b)) */
870 delta = beta[bp]*(de[bp][a] - de[bp][b]) + beta[ap]*(de[ap][b] - de[ap][a]) - (beta[bp]-beta[ap])*(Epot[b]-Epot[a]);
873 gmx_incons("Unknown replica exchange quantity");
877 fprintf(fplog, "Repl %d <-> %d dE_term = %10.3e (kT)\n", a, b, delta);
881 /* revist the calculation for 5.0. Might be some improvements. */
882 dpV = (beta[ap]*re->pres[ap]-beta[bp]*re->pres[bp])*(Vol[b]-Vol[a])/PRESFAC;
885 fprintf(fplog, " dpV = %10.3e d = %10.3e\n", dpV, delta + dpV);
893 test_for_replica_exchange(FILE *fplog,
894 const gmx_multisim_t *ms,
895 struct gmx_repl_ex *re,
896 gmx_enerdata_t *enerd,
901 int m, i, j, a, b, ap, bp, i0, i1, tmp;
903 gmx_bool bPrint, bMultiEx;
904 gmx_bool *bEx = re->bEx;
905 real *prob = re->prob;
906 int *pind = re->destinations; /* permuted index */
907 gmx_bool bEpot = FALSE;
908 gmx_bool bDLambda = FALSE;
909 gmx_bool bVol = FALSE;
911 bMultiEx = (re->nex > 1); /* multiple exchanges at each state */
912 fprintf(fplog, "Replica exchange at step %" GMX_PRId64 " time %.5f\n", step, time);
916 for (i = 0; i < re->nrepl; i++)
921 re->Vol[re->repl] = vol;
923 if ((re->type == ereTEMP || re->type == ereTL))
925 for (i = 0; i < re->nrepl; i++)
930 re->Epot[re->repl] = enerd->term[F_EPOT];
931 /* temperatures of different states*/
932 for (i = 0; i < re->nrepl; i++)
934 re->beta[i] = 1.0/(re->q[ereTEMP][i]*BOLTZ);
939 for (i = 0; i < re->nrepl; i++)
941 re->beta[i] = 1.0/(re->temp*BOLTZ); /* we have a single temperature */
944 if (re->type == ereLAMBDA || re->type == ereTL)
947 /* lambda differences. */
948 /* de[i][j] is the energy of the jth simulation in the ith Hamiltonian
949 minus the energy of the jth simulation in the jth Hamiltonian */
950 for (i = 0; i < re->nrepl; i++)
952 for (j = 0; j < re->nrepl; j++)
957 for (i = 0; i < re->nrepl; i++)
959 re->de[i][re->repl] = (enerd->enerpart_lambda[(int)re->q[ereLAMBDA][i]+1]-enerd->enerpart_lambda[0]);
963 /* now actually do the communication */
966 gmx_sum_sim(re->nrepl, re->Vol, ms);
970 gmx_sum_sim(re->nrepl, re->Epot, ms);
974 for (i = 0; i < re->nrepl; i++)
976 gmx_sum_sim(re->nrepl, re->de[i], ms);
980 /* make a duplicate set of indices for shuffling */
981 for (i = 0; i < re->nrepl; i++)
983 pind[i] = re->ind[i];
988 /* multiple random switch exchange */
990 for (i = 0; i < re->nex + nself; i++)
994 gmx_rng_cycle_2uniform(step, i*2, re->seed, RND_SEED_REPLEX, rnd);
995 /* randomly select a pair */
996 /* in theory, could reduce this by identifying only which switches had a nonneglibible
997 probability of occurring (log p > -100) and only operate on those switches */
998 /* find out which state it is from, and what label that state currently has. Likely
999 more work that useful. */
1000 i0 = (int)(re->nrepl*rnd[0]);
1001 i1 = (int)(re->nrepl*rnd[1]);
1005 continue; /* self-exchange, back up and do it again */
1008 a = re->ind[i0]; /* what are the indices of these states? */
1013 bPrint = FALSE; /* too noisy */
1014 /* calculate the energy difference */
1015 /* if the code changes to flip the STATES, rather than the configurations,
1016 use the commented version of the code */
1017 /* delta = calc_delta(fplog,bPrint,re,a,b,ap,bp); */
1018 delta = calc_delta(fplog, bPrint, re, ap, bp, a, b);
1020 /* we actually only use the first space in the prob and bEx array,
1021 since there are actually many switches between pairs. */
1031 if (delta > PROBABILITYCUTOFF)
1037 prob[0] = exp(-delta);
1039 /* roll a number to determine if accepted */
1040 gmx_rng_cycle_2uniform(step, i*2+1, re->seed, RND_SEED_REPLEX, rnd);
1041 bEx[0] = rnd[0] < prob[0];
1043 re->prob_sum[0] += prob[0];
1047 /* swap the states */
1049 pind[i0] = pind[i1];
1053 re->nattempt[0]++; /* keep track of total permutation trials here */
1054 print_allswitchind(fplog, re->nrepl, pind, re->allswaps, re->tmpswap);
1058 /* standard nearest neighbor replica exchange */
1060 m = (step / re->nst) % 2;
1061 for (i = 1; i < re->nrepl; i++)
1066 bPrint = (re->repl == a || re->repl == b);
1069 delta = calc_delta(fplog, bPrint, re, a, b, a, b);
1080 if (delta > PROBABILITYCUTOFF)
1086 prob[i] = exp(-delta);
1088 /* roll a number to determine if accepted */
1089 gmx_rng_cycle_2uniform(step, i, re->seed, RND_SEED_REPLEX, rnd);
1090 bEx[i] = rnd[0] < prob[i];
1092 re->prob_sum[i] += prob[i];
1096 /* swap these two */
1098 pind[i-1] = pind[i];
1100 re->nexchange[i]++; /* statistics for back compatibility */
1109 /* print some statistics */
1110 print_ind(fplog, "ex", re->nrepl, re->ind, bEx);
1111 print_prob(fplog, "pr", re->nrepl, prob);
1112 fprintf(fplog, "\n");
1116 /* record which moves were made and accepted */
1117 for (i = 0; i < re->nrepl; i++)
1119 re->nmoves[re->ind[i]][pind[i]] += 1;
1120 re->nmoves[pind[i]][re->ind[i]] += 1;
1122 fflush(fplog); /* make sure we can see what the last exchange was */
1125 static void write_debug_x(t_state *state)
1131 for (i = 0; i < state->natoms; i += 10)
1133 fprintf(debug, "dx %5d %10.5f %10.5f %10.5f\n", i, state->x[i][XX], state->x[i][YY], state->x[i][ZZ]);
1139 cyclic_decomposition(const int *destinations,
1148 for (i = 0; i < nrepl; i++)
1152 for (i = 0; i < nrepl; i++) /* one cycle for each replica */
1163 for (j = 0; j < nrepl; j++) /* potentially all cycles are part, but we will break first */
1165 p = destinations[p]; /* start permuting */
1173 break; /* we've reached the original element, the cycle is complete, and we marked the end. */
1177 cyclic[i][c] = p; /* each permutation gives a new member of the cycle */
1183 *nswap = maxlen - 1;
1187 for (i = 0; i < nrepl; i++)
1189 fprintf(debug, "Cycle %d:", i);
1190 for (j = 0; j < nrepl; j++)
1192 if (cyclic[i][j] < 0)
1196 fprintf(debug, "%2d", cyclic[i][j]);
1198 fprintf(debug, "\n");
1205 compute_exchange_order(FILE *fplog,
1213 for (j = 0; j < maxswap; j++)
1215 for (i = 0; i < nrepl; i++)
1217 if (cyclic[i][j+1] >= 0)
1219 order[cyclic[i][j+1]][j] = cyclic[i][j];
1220 order[cyclic[i][j]][j] = cyclic[i][j+1];
1223 for (i = 0; i < nrepl; i++)
1225 if (order[i][j] < 0)
1227 order[i][j] = i; /* if it's not exchanging, it should stay this round*/
1234 fprintf(fplog, "Replica Exchange Order\n");
1235 for (i = 0; i < nrepl; i++)
1237 fprintf(fplog, "Replica %d:", i);
1238 for (j = 0; j < maxswap; j++)
1240 if (order[i][j] < 0)
1244 fprintf(debug, "%2d", order[i][j]);
1246 fprintf(fplog, "\n");
1253 prepare_to_do_exchange(FILE *fplog,
1254 struct gmx_repl_ex *re,
1255 const int replica_id,
1257 gmx_bool *bThisReplicaExchanged)
1260 /* Hold the cyclic decomposition of the (multiple) replica
1262 gmx_bool bAnyReplicaExchanged = FALSE;
1263 *bThisReplicaExchanged = FALSE;
1265 for (i = 0; i < re->nrepl; i++)
1267 if (re->destinations[i] != re->ind[i])
1269 /* only mark as exchanged if the index has been shuffled */
1270 bAnyReplicaExchanged = TRUE;
1274 if (bAnyReplicaExchanged)
1276 /* reinitialize the placeholder arrays */
1277 for (i = 0; i < re->nrepl; i++)
1279 for (j = 0; j < re->nrepl; j++)
1281 re->cyclic[i][j] = -1;
1282 re->order[i][j] = -1;
1286 /* Identify the cyclic decomposition of the permutation (very
1287 * fast if neighbor replica exchange). */
1288 cyclic_decomposition(re->destinations, re->cyclic, re->incycle, re->nrepl, maxswap);
1290 /* Now translate the decomposition into a replica exchange
1291 * order at each step. */
1292 compute_exchange_order(fplog, re->cyclic, re->order, re->nrepl, *maxswap);
1294 /* Did this replica do any exchange at any point? */
1295 for (j = 0; j < *maxswap; j++)
1297 if (replica_id != re->order[replica_id][j])
1299 *bThisReplicaExchanged = TRUE;
1306 gmx_bool replica_exchange(FILE *fplog, const t_commrec *cr, struct gmx_repl_ex *re,
1307 t_state *state, gmx_enerdata_t *enerd,
1308 t_state *state_local, gmx_int64_t step, real time)
1312 int exchange_partner;
1314 /* Number of rounds of exchanges needed to deal with any multiple
1316 /* Where each replica ends up after the exchange attempt(s). */
1317 /* The order in which multiple exchanges will occur. */
1318 gmx_bool bThisReplicaExchanged = FALSE;
1322 replica_id = re->repl;
1323 test_for_replica_exchange(fplog, cr->ms, re, enerd, det(state_local->box), step, time);
1324 prepare_to_do_exchange(fplog, re, replica_id, &maxswap, &bThisReplicaExchanged);
1326 /* Do intra-simulation broadcast so all processors belonging to
1327 * each simulation know whether they need to participate in
1328 * collecting the state. Otherwise, they might as well get on with
1329 * the next thing to do. */
1330 if (DOMAINDECOMP(cr))
1333 MPI_Bcast(&bThisReplicaExchanged, sizeof(gmx_bool), MPI_BYTE, MASTERRANK(cr),
1334 cr->mpi_comm_mygroup);
1338 if (bThisReplicaExchanged)
1340 /* Exchange the states */
1341 /* Collect the global state on the master node */
1342 if (DOMAINDECOMP(cr))
1344 dd_collect_state(cr->dd, state_local, state);
1348 copy_state_nonatomdata(state_local, state);
1353 /* There will be only one swap cycle with standard replica
1354 * exchange, but there may be multiple swap cycles if we
1355 * allow multiple swaps. */
1357 for (j = 0; j < maxswap; j++)
1359 exchange_partner = re->order[replica_id][j];
1361 if (exchange_partner != replica_id)
1363 /* Exchange the global states between the master nodes */
1366 fprintf(debug, "Exchanging %d with %d\n", replica_id, exchange_partner);
1368 exchange_state(cr->ms, exchange_partner, state);
1371 /* For temperature-type replica exchange, we need to scale
1372 * the velocities. */
1373 if (re->type == ereTEMP || re->type == ereTL)
1375 scale_velocities(state, sqrt(re->q[ereTEMP][replica_id]/re->q[ereTEMP][re->destinations[replica_id]]));
1380 /* With domain decomposition the global state is distributed later */
1381 if (!DOMAINDECOMP(cr))
1383 /* Copy the global state to the local state data structure */
1384 copy_state_nonatomdata(state, state_local);
1388 return bThisReplicaExchanged;
1391 void print_replica_exchange_statistics(FILE *fplog, struct gmx_repl_ex *re)
1395 fprintf(fplog, "\nReplica exchange statistics\n");
1399 fprintf(fplog, "Repl %d attempts, %d odd, %d even\n",
1400 re->nattempt[0]+re->nattempt[1], re->nattempt[1], re->nattempt[0]);
1402 fprintf(fplog, "Repl average probabilities:\n");
1403 for (i = 1; i < re->nrepl; i++)
1405 if (re->nattempt[i%2] == 0)
1411 re->prob[i] = re->prob_sum[i]/re->nattempt[i%2];
1414 print_ind(fplog, "", re->nrepl, re->ind, NULL);
1415 print_prob(fplog, "", re->nrepl, re->prob);
1417 fprintf(fplog, "Repl number of exchanges:\n");
1418 print_ind(fplog, "", re->nrepl, re->ind, NULL);
1419 print_count(fplog, "", re->nrepl, re->nexchange);
1421 fprintf(fplog, "Repl average number of exchanges:\n");
1422 for (i = 1; i < re->nrepl; i++)
1424 if (re->nattempt[i%2] == 0)
1430 re->prob[i] = ((real)re->nexchange[i])/re->nattempt[i%2];
1433 print_ind(fplog, "", re->nrepl, re->ind, NULL);
1434 print_prob(fplog, "", re->nrepl, re->prob);
1436 fprintf(fplog, "\n");
1438 /* print the transition matrix */
1439 print_transition_matrix(fplog, re->nrepl, re->nmoves, re->nattempt);