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.
44 #include "gromacs/random/random.h"
45 #include "gromacs/utility/smalloc.h"
52 #include "gromacs/random/random.h"
54 #define PROBABILITYCUTOFF 100
55 /* we don't bother evaluating if events are more rare than exp(-100) = 3.7x10^-44 */
58 ereTEMP, ereLAMBDA, ereENDSINGLE, ereTL, ereNR
60 const char *erename[ereNR] = { "temperature", "lambda", "end_single_marker", "temperature and lambda"};
61 /* end_single_marker merely notes the end of single variable replica exchange. All types higher than
62 it are multiple replica exchange methods */
63 /* Eventually, should add 'pressure', 'temperature and pressure', 'lambda_and_pressure', 'temperature_lambda_pressure'?;
64 Let's wait until we feel better about the pressure control methods giving exact ensembles. Right now, we assume constant pressure */
66 typedef struct gmx_repl_ex
86 /* these are helper arrays for replica exchange; allocated here so they
87 don't have to be allocated each time */
95 /* helper arrays to hold the quantities that are exchanged */
104 static gmx_bool repl_quantity(const gmx_multisim_t *ms,
105 struct gmx_repl_ex *re, int ere, real q)
111 snew(qall, ms->nsim);
113 gmx_sum_sim(ms->nsim, qall, ms);
116 for (s = 1; s < ms->nsim; s++)
118 if (qall[s] != qall[0])
126 /* Set the replica exchange type and quantities */
129 snew(re->q[ere], re->nrepl);
130 for (s = 0; s < ms->nsim; s++)
132 re->q[ere][s] = qall[s];
139 gmx_repl_ex_t init_replica_exchange(FILE *fplog,
140 const gmx_multisim_t *ms,
141 const t_state *state,
142 const t_inputrec *ir,
143 int nst, int nex, int init_seed)
147 struct gmx_repl_ex *re;
149 gmx_bool bLambda = FALSE;
151 fprintf(fplog, "\nInitializing Replica Exchange\n");
153 if (ms == NULL || ms->nsim == 1)
155 gmx_fatal(FARGS, "Nothing to exchange with only one replica, maybe you forgot to set the -multi option of mdrun?");
157 if (!EI_DYNAMICS(ir->eI))
159 gmx_fatal(FARGS, "Replica exchange is only supported by dynamical simulations");
160 /* Note that PAR(cr) is defined by cr->nnodes > 1, which is
161 * distinct from MULTISIM(cr). A multi-simulation only runs
162 * with real MPI parallelism, but this does not imply PAR(cr)
165 * Since we are using a dynamical integrator, the only
166 * decomposition is DD, so PAR(cr) and DOMAINDECOMP(cr) are
167 * synonymous. The only way for cr->nnodes > 1 to be true is
168 * if we are using DD. */
174 re->nrepl = ms->nsim;
175 snew(re->q, ereENDSINGLE);
177 fprintf(fplog, "Repl There are %d replicas:\n", re->nrepl);
179 check_multi_int(fplog, ms, state->natoms, "the number of atoms", FALSE);
180 check_multi_int(fplog, ms, ir->eI, "the integrator", FALSE);
181 check_multi_int64(fplog, ms, ir->init_step+ir->nsteps, "init_step+nsteps", FALSE);
182 check_multi_int64(fplog, ms, (ir->init_step+nst-1)/nst,
183 "first exchange step: init_step/-replex", FALSE);
184 check_multi_int(fplog, ms, ir->etc, "the temperature coupling", FALSE);
185 check_multi_int(fplog, ms, ir->opts.ngtc,
186 "the number of temperature coupling groups", FALSE);
187 check_multi_int(fplog, ms, ir->epc, "the pressure coupling", FALSE);
188 check_multi_int(fplog, ms, ir->efep, "free energy", FALSE);
189 check_multi_int(fplog, ms, ir->fepvals->n_lambda, "number of lambda states", FALSE);
191 re->temp = ir->opts.ref_t[0];
192 for (i = 1; (i < ir->opts.ngtc); i++)
194 if (ir->opts.ref_t[i] != re->temp)
196 fprintf(fplog, "\nWARNING: The temperatures of the different temperature coupling groups are not identical\n\n");
197 fprintf(stderr, "\nWARNING: The temperatures of the different temperature coupling groups are not identical\n\n");
202 bTemp = repl_quantity(ms, re, ereTEMP, re->temp);
203 if (ir->efep != efepNO)
205 bLambda = repl_quantity(ms, re, ereLAMBDA, (real)ir->fepvals->init_fep_state);
207 if (re->type == -1) /* nothing was assigned */
209 gmx_fatal(FARGS, "The properties of the %d systems are all the same, there is nothing to exchange", re->nrepl);
211 if (bLambda && bTemp)
218 please_cite(fplog, "Sugita1999a");
219 if (ir->epc != epcNO)
222 fprintf(fplog, "Repl Using Constant Pressure REMD.\n");
223 please_cite(fplog, "Okabe2001a");
225 if (ir->etc == etcBERENDSEN)
227 gmx_fatal(FARGS, "REMD with the %s thermostat does not produce correct potential energy distributions, consider using the %s thermostat instead",
228 ETCOUPLTYPE(ir->etc), ETCOUPLTYPE(etcVRESCALE));
233 if (ir->fepvals->delta_lambda != 0) /* check this? */
235 gmx_fatal(FARGS, "delta_lambda is not zero");
240 snew(re->pres, re->nrepl);
241 if (ir->epct == epctSURFACETENSION)
243 pres = ir->ref_p[ZZ][ZZ];
249 for (i = 0; i < DIM; i++)
251 if (ir->compress[i][i] != 0)
253 pres += ir->ref_p[i][i];
259 re->pres[re->repl] = pres;
260 gmx_sum_sim(re->nrepl, re->pres, ms);
263 /* Make an index for increasing replica order */
264 /* only makes sense if one or the other is varying, not both!
265 if both are varying, we trust the order the person gave. */
266 snew(re->ind, re->nrepl);
267 for (i = 0; i < re->nrepl; i++)
272 if (re->type < ereENDSINGLE)
275 for (i = 0; i < re->nrepl; i++)
277 for (j = i+1; j < re->nrepl; j++)
279 if (re->q[re->type][re->ind[j]] < re->q[re->type][re->ind[i]])
281 /* Unordered replicas are supposed to work, but there
282 * is still an issues somewhere.
283 * Note that at this point still re->ind[i]=i.
285 gmx_fatal(FARGS, "Replicas with indices %d < %d have %ss %g > %g, please order your replicas on increasing %s",
288 re->q[re->type][i], re->q[re->type][j],
292 re->ind[i] = re->ind[j];
295 else if (re->q[re->type][re->ind[j]] == re->q[re->type][re->ind[i]])
297 gmx_fatal(FARGS, "Two replicas have identical %ss", erename[re->type]);
303 /* keep track of all the swaps, starting with the initial placement. */
304 snew(re->allswaps, re->nrepl);
305 for (i = 0; i < re->nrepl; i++)
307 re->allswaps[i] = re->ind[i];
313 fprintf(fplog, "\nReplica exchange in temperature\n");
314 for (i = 0; i < re->nrepl; i++)
316 fprintf(fplog, " %5.1f", re->q[re->type][re->ind[i]]);
318 fprintf(fplog, "\n");
321 fprintf(fplog, "\nReplica exchange in lambda\n");
322 for (i = 0; i < re->nrepl; i++)
324 fprintf(fplog, " %3d", (int)re->q[re->type][re->ind[i]]);
326 fprintf(fplog, "\n");
329 fprintf(fplog, "\nReplica exchange in temperature and lambda state\n");
330 for (i = 0; i < re->nrepl; i++)
332 fprintf(fplog, " %5.1f", re->q[ereTEMP][re->ind[i]]);
334 fprintf(fplog, "\n");
335 for (i = 0; i < re->nrepl; i++)
337 fprintf(fplog, " %5d", (int)re->q[ereLAMBDA][re->ind[i]]);
339 fprintf(fplog, "\n");
342 gmx_incons("Unknown replica exchange quantity");
346 fprintf(fplog, "\nRepl p");
347 for (i = 0; i < re->nrepl; i++)
349 fprintf(fplog, " %5.2f", re->pres[re->ind[i]]);
352 for (i = 0; i < re->nrepl; i++)
354 if ((i > 0) && (re->pres[re->ind[i]] < re->pres[re->ind[i-1]]))
356 fprintf(fplog, "\nWARNING: The reference pressures decrease with increasing temperatures\n\n");
357 fprintf(stderr, "\nWARNING: The reference pressures decrease with increasing temperatures\n\n");
366 re->seed = (int)gmx_rng_make_seed();
372 gmx_sumi_sim(1, &(re->seed), ms);
376 re->seed = init_seed;
378 fprintf(fplog, "\nReplica exchange interval: %d\n", re->nst);
379 fprintf(fplog, "\nReplica random seed: %d\n", re->seed);
380 re->rng = gmx_rng_init(re->seed);
385 snew(re->prob_sum, re->nrepl);
386 snew(re->nexchange, re->nrepl);
387 snew(re->nmoves, re->nrepl);
388 for (i = 0; i < re->nrepl; i++)
390 snew(re->nmoves[i], re->nrepl);
392 fprintf(fplog, "Replica exchange information below: x=exchange, pr=probability\n");
394 /* generate space for the helper functions so we don't have to snew each time */
396 snew(re->destinations, re->nrepl);
397 snew(re->incycle, re->nrepl);
398 snew(re->tmpswap, re->nrepl);
399 snew(re->cyclic, re->nrepl);
400 snew(re->order, re->nrepl);
401 for (i = 0; i < re->nrepl; i++)
403 snew(re->cyclic[i], re->nrepl);
404 snew(re->order[i], re->nrepl);
406 /* allocate space for the functions storing the data for the replicas */
407 /* not all of these arrays needed in all cases, but they don't take
408 up much space, since the max size is nrepl**2 */
409 snew(re->prob, re->nrepl);
410 snew(re->bEx, re->nrepl);
411 snew(re->beta, re->nrepl);
412 snew(re->Vol, re->nrepl);
413 snew(re->Epot, re->nrepl);
414 snew(re->de, re->nrepl);
415 for (i = 0; i < re->nrepl; i++)
417 snew(re->de[i], re->nrepl);
423 static void exchange_reals(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, real *v, int n)
433 MPI_Sendrecv(v, n*sizeof(real),MPI_BYTE,MSRANK(ms,b),0,
434 buf,n*sizeof(real),MPI_BYTE,MSRANK(ms,b),0,
435 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
440 MPI_Isend(v, n*sizeof(real), MPI_BYTE, MSRANK(ms, b), 0,
441 ms->mpi_comm_masters, &mpi_req);
442 MPI_Recv(buf, n*sizeof(real), MPI_BYTE, MSRANK(ms, b), 0,
443 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
444 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
447 for (i = 0; i < n; i++)
456 static void exchange_ints(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, int *v, int n)
466 MPI_Sendrecv(v, n*sizeof(int),MPI_BYTE,MSRANK(ms,b),0,
467 buf,n*sizeof(int),MPI_BYTE,MSRANK(ms,b),0,
468 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
473 MPI_Isend(v, n*sizeof(int), MPI_BYTE, MSRANK(ms, b), 0,
474 ms->mpi_comm_masters, &mpi_req);
475 MPI_Recv(buf, n*sizeof(int), MPI_BYTE, MSRANK(ms, b), 0,
476 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
477 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
480 for (i = 0; i < n; i++)
488 static void exchange_doubles(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, double *v, int n)
498 MPI_Sendrecv(v, n*sizeof(double),MPI_BYTE,MSRANK(ms,b),0,
499 buf,n*sizeof(double),MPI_BYTE,MSRANK(ms,b),0,
500 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
505 MPI_Isend(v, n*sizeof(double), MPI_BYTE, MSRANK(ms, b), 0,
506 ms->mpi_comm_masters, &mpi_req);
507 MPI_Recv(buf, n*sizeof(double), MPI_BYTE, MSRANK(ms, b), 0,
508 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
509 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
512 for (i = 0; i < n; i++)
520 static void exchange_rvecs(const gmx_multisim_t gmx_unused *ms, int gmx_unused b, rvec *v, int n)
530 MPI_Sendrecv(v[0], n*sizeof(rvec),MPI_BYTE,MSRANK(ms,b),0,
531 buf[0],n*sizeof(rvec),MPI_BYTE,MSRANK(ms,b),0,
532 ms->mpi_comm_masters,MPI_STATUS_IGNORE);
537 MPI_Isend(v[0], n*sizeof(rvec), MPI_BYTE, MSRANK(ms, b), 0,
538 ms->mpi_comm_masters, &mpi_req);
539 MPI_Recv(buf[0], n*sizeof(rvec), MPI_BYTE, MSRANK(ms, b), 0,
540 ms->mpi_comm_masters, MPI_STATUS_IGNORE);
541 MPI_Wait(&mpi_req, MPI_STATUS_IGNORE);
544 for (i = 0; i < n; i++)
546 copy_rvec(buf[i], v[i]);
552 static void exchange_state(const gmx_multisim_t *ms, int b, t_state *state)
554 /* When t_state changes, this code should be updated. */
556 ngtc = state->ngtc * state->nhchainlength;
557 nnhpres = state->nnhpres* state->nhchainlength;
558 exchange_rvecs(ms, b, state->box, DIM);
559 exchange_rvecs(ms, b, state->box_rel, DIM);
560 exchange_rvecs(ms, b, state->boxv, DIM);
561 exchange_reals(ms, b, &(state->veta), 1);
562 exchange_reals(ms, b, &(state->vol0), 1);
563 exchange_rvecs(ms, b, state->svir_prev, DIM);
564 exchange_rvecs(ms, b, state->fvir_prev, DIM);
565 exchange_rvecs(ms, b, state->pres_prev, DIM);
566 exchange_doubles(ms, b, state->nosehoover_xi, ngtc);
567 exchange_doubles(ms, b, state->nosehoover_vxi, ngtc);
568 exchange_doubles(ms, b, state->nhpres_xi, nnhpres);
569 exchange_doubles(ms, b, state->nhpres_vxi, nnhpres);
570 exchange_doubles(ms, b, state->therm_integral, state->ngtc);
571 exchange_rvecs(ms, b, state->x, state->natoms);
572 exchange_rvecs(ms, b, state->v, state->natoms);
573 exchange_rvecs(ms, b, state->sd_X, state->natoms);
576 static void copy_rvecs(rvec *s, rvec *d, int n)
582 for (i = 0; i < n; i++)
584 copy_rvec(s[i], d[i]);
589 static void copy_doubles(const double *s, double *d, int n)
595 for (i = 0; i < n; i++)
602 static void copy_reals(const real *s, real *d, int n)
608 for (i = 0; i < n; i++)
615 static void copy_ints(const int *s, int *d, int n)
621 for (i = 0; i < n; i++)
628 #define scopy_rvecs(v, n) copy_rvecs(state->v, state_local->v, n);
629 #define scopy_doubles(v, n) copy_doubles(state->v, state_local->v, n);
630 #define scopy_reals(v, n) copy_reals(state->v, state_local->v, n);
631 #define scopy_ints(v, n) copy_ints(state->v, state_local->v, n);
633 static void copy_state_nonatomdata(t_state *state, t_state *state_local)
635 /* When t_state changes, this code should be updated. */
637 ngtc = state->ngtc * state->nhchainlength;
638 nnhpres = state->nnhpres* state->nhchainlength;
639 scopy_rvecs(box, DIM);
640 scopy_rvecs(box_rel, DIM);
641 scopy_rvecs(boxv, DIM);
642 state_local->veta = state->veta;
643 state_local->vol0 = state->vol0;
644 scopy_rvecs(svir_prev, DIM);
645 scopy_rvecs(fvir_prev, DIM);
646 scopy_rvecs(pres_prev, DIM);
647 scopy_doubles(nosehoover_xi, ngtc);
648 scopy_doubles(nosehoover_vxi, ngtc);
649 scopy_doubles(nhpres_xi, nnhpres);
650 scopy_doubles(nhpres_vxi, nnhpres);
651 scopy_doubles(therm_integral, state->ngtc);
652 scopy_rvecs(x, state->natoms);
653 scopy_rvecs(v, state->natoms);
654 scopy_rvecs(sd_X, state->natoms);
655 copy_ints(&(state->fep_state), &(state_local->fep_state), 1);
656 scopy_reals(lambda, efptNR);
659 static void scale_velocities(t_state *state, real fac)
665 for (i = 0; i < state->natoms; i++)
667 svmul(fac, state->v[i], state->v[i]);
672 static void print_transition_matrix(FILE *fplog, int n, int **nmoves, int *nattempt)
677 ntot = nattempt[0] + nattempt[1];
678 fprintf(fplog, "\n");
679 fprintf(fplog, "Repl");
680 for (i = 0; i < n; i++)
682 fprintf(fplog, " "); /* put the title closer to the center */
684 fprintf(fplog, "Empirical Transition Matrix\n");
686 fprintf(fplog, "Repl");
687 for (i = 0; i < n; i++)
689 fprintf(fplog, "%8d", (i+1));
691 fprintf(fplog, "\n");
693 for (i = 0; i < n; i++)
695 fprintf(fplog, "Repl");
696 for (j = 0; j < n; j++)
699 if (nmoves[i][j] > 0)
701 Tprint = nmoves[i][j]/(2.0*ntot);
703 fprintf(fplog, "%8.4f", Tprint);
705 fprintf(fplog, "%3d\n", i);
709 static void print_ind(FILE *fplog, const char *leg, int n, int *ind, gmx_bool *bEx)
713 fprintf(fplog, "Repl %2s %2d", leg, ind[0]);
714 for (i = 1; i < n; i++)
716 fprintf(fplog, " %c %2d", (bEx != 0 && bEx[i]) ? 'x' : ' ', ind[i]);
718 fprintf(fplog, "\n");
721 static void print_allswitchind(FILE *fplog, int n, int *pind, int *allswaps, int *tmpswap)
725 for (i = 0; i < n; i++)
727 tmpswap[i] = allswaps[i];
729 for (i = 0; i < n; i++)
731 allswaps[i] = tmpswap[pind[i]];
734 fprintf(fplog, "\nAccepted Exchanges: ");
735 for (i = 0; i < n; i++)
737 fprintf(fplog, "%d ", pind[i]);
739 fprintf(fplog, "\n");
741 /* the "Order After Exchange" is the state label corresponding to the configuration that
742 started in state listed in order, i.e.
747 configuration starting in simulation 3 is now in simulation 0,
748 configuration starting in simulation 0 is now in simulation 1,
749 configuration starting in simulation 1 is now in simulation 2,
750 configuration starting in simulation 2 is now in simulation 3
752 fprintf(fplog, "Order After Exchange: ");
753 for (i = 0; i < n; i++)
755 fprintf(fplog, "%d ", allswaps[i]);
757 fprintf(fplog, "\n\n");
760 static void print_prob(FILE *fplog, const char *leg, int n, real *prob)
765 fprintf(fplog, "Repl %2s ", leg);
766 for (i = 1; i < n; i++)
770 sprintf(buf, "%4.2f", prob[i]);
771 fprintf(fplog, " %3s", buf[0] == '1' ? "1.0" : buf+1);
778 fprintf(fplog, "\n");
781 static void print_count(FILE *fplog, const char *leg, int n, int *count)
785 fprintf(fplog, "Repl %2s ", leg);
786 for (i = 1; i < n; i++)
788 fprintf(fplog, " %4d", count[i]);
790 fprintf(fplog, "\n");
793 static real calc_delta(FILE *fplog, gmx_bool bPrint, struct gmx_repl_ex *re, int a, int b, int ap, int bp)
796 real ediff, dpV, delta = 0;
797 real *Epot = re->Epot;
800 real *beta = re->beta;
802 /* Two cases; we are permuted and not. In all cases, setting ap = a and bp = b will reduce
803 to the non permuted case */
809 * Okabe et. al. Chem. Phys. Lett. 335 (2001) 435-439
811 ediff = Epot[b] - Epot[a];
812 delta = -(beta[bp] - beta[ap])*ediff;
815 /* two cases: when we are permuted, and not. */
817 ediff = E_new - E_old
818 = [H_b(x_a) + H_a(x_b)] - [H_b(x_b) + H_a(x_a)]
819 = [H_b(x_a) - H_a(x_a)] + [H_a(x_b) - H_b(x_b)]
820 = de[b][a] + de[a][b] */
823 ediff = E_new - E_old
824 = [H_bp(x_a) + H_ap(x_b)] - [H_bp(x_b) + H_ap(x_a)]
825 = [H_bp(x_a) - H_ap(x_a)] + [H_ap(x_b) - H_bp(x_b)]
826 = [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)]
827 = [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)]
828 = (de[bp][a] - de[ap][a]) + (de[ap][b] - de[bp][b]) */
829 /* but, in the current code implementation, we flip configurations, not indices . . .
830 So let's examine that.
831 = [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)]
832 = [H_b(x_ap) - H_a(x_ap)] + [H_a(x_bp) - H_b(x_pb)]
833 = (de[b][ap] - de[a][ap]) + (de[a][bp] - de[b][bp]
834 So, if we exchange b<=> bp and a<=> ap, we return to the same result.
835 So the simple solution is to flip the
836 position of perturbed and original indices in the tests.
839 ediff = (de[bp][a] - de[ap][a]) + (de[ap][b] - de[bp][b]);
840 delta = ediff*beta[a]; /* assume all same temperature in this case */
844 /* delta = reduced E_new - reduced E_old
845 = [beta_b H_b(x_a) + beta_a H_a(x_b)] - [beta_b H_b(x_b) + beta_a H_a(x_a)]
846 = [beta_b H_b(x_a) - beta_a H_a(x_a)] + [beta_a H_a(x_b) - beta_b H_b(x_b)]
847 = [beta_b dH_b(x_a) + beta_b H_a(x_a) - beta_a H_a(x_a)] +
848 [beta_a dH_a(x_b) + beta_a H_b(x_b) - beta_b H_b(x_b)]
849 = [beta_b dH_b(x_a) + [beta_a dH_a(x_b) +
850 beta_b (H_a(x_a) - H_b(x_b)]) - beta_a (H_a(x_a) - H_b(x_b))
851 = beta_b dH_b(x_a) + beta_a dH_a(x_b) - (beta_b - beta_a)(H_b(x_b) - H_a(x_a) */
852 /* delta = beta[b]*de[b][a] + beta[a]*de[a][b] - (beta[b] - beta[a])*(Epot[b] - Epot[a]; */
853 /* permuted (big breath!) */
854 /* delta = reduced E_new - reduced E_old
855 = [beta_bp H_bp(x_a) + beta_ap H_ap(x_b)] - [beta_bp H_bp(x_b) + beta_ap H_ap(x_a)]
856 = [beta_bp H_bp(x_a) - beta_ap H_ap(x_a)] + [beta_ap H_ap(x_b) - beta_bp H_bp(x_b)]
857 = [beta_bp H_bp(x_a) - beta_ap H_ap(x_a)] + [beta_ap H_ap(x_b) - beta_bp H_bp(x_b)]
858 - beta_pb H_a(x_a) + beta_ap H_a(x_a) + beta_pb H_a(x_a) - beta_ap H_a(x_a)
859 - beta_ap H_b(x_b) + beta_bp H_b(x_b) + beta_ap H_b(x_b) - beta_bp H_b(x_b)
860 = [(beta_bp H_bp(x_a) - beta_bp H_a(x_a)) - (beta_ap H_ap(x_a) - beta_ap H_a(x_a))] +
861 [(beta_ap H_ap(x_b) - beta_ap H_b(x_b)) - (beta_bp H_bp(x_b) - beta_bp H_b(x_b))]
862 + beta_pb H_a(x_a) - beta_ap H_a(x_a) + beta_ap H_b(x_b) - beta_bp H_b(x_b)
863 = [beta_bp (H_bp(x_a) - H_a(x_a)) - beta_ap (H_ap(x_a) - H_a(x_a))] +
864 [beta_ap (H_ap(x_b) - H_b(x_b)) - beta_bp (H_bp(x_b) - H_b(x_b))]
865 + beta_pb (H_a(x_a) - H_b(x_b)) - beta_ap (H_a(x_a) - H_b(x_b))
866 = ([beta_bp de[bp][a] - beta_ap de[ap][a]) + beta_ap de[ap][b] - beta_bp de[bp][b])
867 + (beta_pb-beta_ap)(H_a(x_a) - H_b(x_b)) */
868 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]);
871 gmx_incons("Unknown replica exchange quantity");
875 fprintf(fplog, "Repl %d <-> %d dE_term = %10.3e (kT)\n", a, b, delta);
879 /* revist the calculation for 5.0. Might be some improvements. */
880 dpV = (beta[ap]*re->pres[ap]-beta[bp]*re->pres[bp])*(Vol[b]-Vol[a])/PRESFAC;
883 fprintf(fplog, " dpV = %10.3e d = %10.3e\n", dpV, delta + dpV);
891 test_for_replica_exchange(FILE *fplog,
892 const gmx_multisim_t *ms,
893 struct gmx_repl_ex *re,
894 gmx_enerdata_t *enerd,
899 int m, i, j, a, b, ap, bp, i0, i1, tmp;
900 real ediff = 0, delta = 0, dpV = 0;
901 gmx_bool bPrint, bMultiEx;
902 gmx_bool *bEx = re->bEx;
903 real *prob = re->prob;
904 int *pind = re->destinations; /* permuted index */
905 gmx_bool bEpot = FALSE;
906 gmx_bool bDLambda = FALSE;
907 gmx_bool bVol = FALSE;
910 bMultiEx = (re->nex > 1); /* multiple exchanges at each state */
911 fprintf(fplog, "Replica exchange at step " "%"GMX_PRId64 " time %.5f\n", step, time);
915 for (i = 0; i < re->nrepl; i++)
920 re->Vol[re->repl] = vol;
922 if ((re->type == ereTEMP || re->type == ereTL))
924 for (i = 0; i < re->nrepl; i++)
929 re->Epot[re->repl] = enerd->term[F_EPOT];
930 /* temperatures of different states*/
931 for (i = 0; i < re->nrepl; i++)
933 re->beta[i] = 1.0/(re->q[ereTEMP][i]*BOLTZ);
938 for (i = 0; i < re->nrepl; i++)
940 re->beta[i] = 1.0/(re->temp*BOLTZ); /* we have a single temperature */
943 if (re->type == ereLAMBDA || re->type == ereTL)
946 /* lambda differences. */
947 /* de[i][j] is the energy of the jth simulation in the ith Hamiltonian
948 minus the energy of the jth simulation in the jth Hamiltonian */
949 for (i = 0; i < re->nrepl; i++)
951 for (j = 0; j < re->nrepl; j++)
956 for (i = 0; i < re->nrepl; i++)
958 re->de[i][re->repl] = (enerd->enerpart_lambda[(int)re->q[ereLAMBDA][i]+1]-enerd->enerpart_lambda[0]);
962 /* now actually do the communication */
965 gmx_sum_sim(re->nrepl, re->Vol, ms);
969 gmx_sum_sim(re->nrepl, re->Epot, ms);
973 for (i = 0; i < re->nrepl; i++)
975 gmx_sum_sim(re->nrepl, re->de[i], ms);
979 /* make a duplicate set of indices for shuffling */
980 for (i = 0; i < re->nrepl; i++)
982 pind[i] = re->ind[i];
987 /* multiple random switch exchange */
989 for (i = 0; i < re->nex + nself; i++)
993 gmx_rng_cycle_2uniform(step, i*2, re->seed, RND_SEED_REPLEX, rnd);
994 /* randomly select a pair */
995 /* in theory, could reduce this by identifying only which switches had a nonneglibible
996 probability of occurring (log p > -100) and only operate on those switches */
997 /* find out which state it is from, and what label that state currently has. Likely
998 more work that useful. */
999 i0 = (int)(re->nrepl*rnd[0]);
1000 i1 = (int)(re->nrepl*rnd[1]);
1004 continue; /* self-exchange, back up and do it again */
1007 a = re->ind[i0]; /* what are the indices of these states? */
1012 bPrint = FALSE; /* too noisy */
1013 /* calculate the energy difference */
1014 /* if the code changes to flip the STATES, rather than the configurations,
1015 use the commented version of the code */
1016 /* delta = calc_delta(fplog,bPrint,re,a,b,ap,bp); */
1017 delta = calc_delta(fplog, bPrint, re, ap, bp, a, b);
1019 /* we actually only use the first space in the prob and bEx array,
1020 since there are actually many switches between pairs. */
1030 if (delta > PROBABILITYCUTOFF)
1036 prob[0] = exp(-delta);
1038 /* roll a number to determine if accepted */
1039 gmx_rng_cycle_2uniform(step, i*2+1, re->seed, RND_SEED_REPLEX, rnd);
1040 bEx[0] = rnd[0] < prob[0];
1042 re->prob_sum[0] += prob[0];
1046 /* swap the states */
1048 pind[i0] = pind[i1];
1052 re->nattempt[0]++; /* keep track of total permutation trials here */
1053 print_allswitchind(fplog, re->nrepl, pind, re->allswaps, re->tmpswap);
1057 /* standard nearest neighbor replica exchange */
1059 m = (step / re->nst) % 2;
1060 for (i = 1; i < re->nrepl; i++)
1065 bPrint = (re->repl == a || re->repl == b);
1068 delta = calc_delta(fplog, bPrint, re, a, b, a, b);
1079 if (delta > PROBABILITYCUTOFF)
1085 prob[i] = exp(-delta);
1087 /* roll a number to determine if accepted */
1088 gmx_rng_cycle_2uniform(step, i, re->seed, RND_SEED_REPLEX, rnd);
1089 bEx[i] = rnd[0] < prob[i];
1091 re->prob_sum[i] += prob[i];
1095 /* swap these two */
1097 pind[i-1] = pind[i];
1099 re->nexchange[i]++; /* statistics for back compatibility */
1108 /* print some statistics */
1109 print_ind(fplog, "ex", re->nrepl, re->ind, bEx);
1110 print_prob(fplog, "pr", re->nrepl, prob);
1111 fprintf(fplog, "\n");
1115 /* record which moves were made and accepted */
1116 for (i = 0; i < re->nrepl; i++)
1118 re->nmoves[re->ind[i]][pind[i]] += 1;
1119 re->nmoves[pind[i]][re->ind[i]] += 1;
1121 fflush(fplog); /* make sure we can see what the last exchange was */
1124 static void write_debug_x(t_state *state)
1130 for (i = 0; i < state->natoms; i += 10)
1132 fprintf(debug, "dx %5d %10.5f %10.5f %10.5f\n", i, state->x[i][XX], state->x[i][YY], state->x[i][ZZ]);
1138 cyclic_decomposition(const int *destinations,
1147 for (i = 0; i < nrepl; i++)
1151 for (i = 0; i < nrepl; i++) /* one cycle for each replica */
1162 for (j = 0; j < nrepl; j++) /* potentially all cycles are part, but we will break first */
1164 p = destinations[p]; /* start permuting */
1172 break; /* we've reached the original element, the cycle is complete, and we marked the end. */
1176 cyclic[i][c] = p; /* each permutation gives a new member of the cycle */
1182 *nswap = maxlen - 1;
1186 for (i = 0; i < nrepl; i++)
1188 fprintf(debug, "Cycle %d:", i);
1189 for (j = 0; j < nrepl; j++)
1191 if (cyclic[i][j] < 0)
1195 fprintf(debug, "%2d", cyclic[i][j]);
1197 fprintf(debug, "\n");
1204 compute_exchange_order(FILE *fplog,
1212 for (j = 0; j < maxswap; j++)
1214 for (i = 0; i < nrepl; i++)
1216 if (cyclic[i][j+1] >= 0)
1218 order[cyclic[i][j+1]][j] = cyclic[i][j];
1219 order[cyclic[i][j]][j] = cyclic[i][j+1];
1222 for (i = 0; i < nrepl; i++)
1224 if (order[i][j] < 0)
1226 order[i][j] = i; /* if it's not exchanging, it should stay this round*/
1233 fprintf(fplog, "Replica Exchange Order\n");
1234 for (i = 0; i < nrepl; i++)
1236 fprintf(fplog, "Replica %d:", i);
1237 for (j = 0; j < maxswap; j++)
1239 if (order[i][j] < 0)
1243 fprintf(debug, "%2d", order[i][j]);
1245 fprintf(fplog, "\n");
1252 prepare_to_do_exchange(FILE *fplog,
1253 struct gmx_repl_ex *re,
1254 const int replica_id,
1256 gmx_bool *bThisReplicaExchanged)
1259 /* Hold the cyclic decomposition of the (multiple) replica
1261 gmx_bool bAnyReplicaExchanged = FALSE;
1262 *bThisReplicaExchanged = FALSE;
1264 for (i = 0; i < re->nrepl; i++)
1266 if (re->destinations[i] != re->ind[i])
1268 /* only mark as exchanged if the index has been shuffled */
1269 bAnyReplicaExchanged = TRUE;
1273 if (bAnyReplicaExchanged)
1275 /* reinitialize the placeholder arrays */
1276 for (i = 0; i < re->nrepl; i++)
1278 for (j = 0; j < re->nrepl; j++)
1280 re->cyclic[i][j] = -1;
1281 re->order[i][j] = -1;
1285 /* Identify the cyclic decomposition of the permutation (very
1286 * fast if neighbor replica exchange). */
1287 cyclic_decomposition(re->destinations, re->cyclic, re->incycle, re->nrepl, maxswap);
1289 /* Now translate the decomposition into a replica exchange
1290 * order at each step. */
1291 compute_exchange_order(fplog, re->cyclic, re->order, re->nrepl, *maxswap);
1293 /* Did this replica do any exchange at any point? */
1294 for (j = 0; j < *maxswap; j++)
1296 if (replica_id != re->order[replica_id][j])
1298 *bThisReplicaExchanged = TRUE;
1305 gmx_bool replica_exchange(FILE *fplog, const t_commrec *cr, struct gmx_repl_ex *re,
1306 t_state *state, gmx_enerdata_t *enerd,
1307 t_state *state_local, gmx_int64_t step, real time)
1311 int exchange_partner;
1313 /* Number of rounds of exchanges needed to deal with any multiple
1315 /* Where each replica ends up after the exchange attempt(s). */
1316 /* The order in which multiple exchanges will occur. */
1317 gmx_bool bThisReplicaExchanged = FALSE;
1321 replica_id = re->repl;
1322 test_for_replica_exchange(fplog, cr->ms, re, enerd, det(state_local->box), step, time);
1323 prepare_to_do_exchange(fplog, re, replica_id, &maxswap, &bThisReplicaExchanged);
1325 /* Do intra-simulation broadcast so all processors belonging to
1326 * each simulation know whether they need to participate in
1327 * collecting the state. Otherwise, they might as well get on with
1328 * the next thing to do. */
1329 if (DOMAINDECOMP(cr))
1332 MPI_Bcast(&bThisReplicaExchanged, sizeof(gmx_bool), MPI_BYTE, MASTERRANK(cr),
1333 cr->mpi_comm_mygroup);
1337 if (bThisReplicaExchanged)
1339 /* Exchange the states */
1340 /* Collect the global state on the master node */
1341 if (DOMAINDECOMP(cr))
1343 dd_collect_state(cr->dd, state_local, state);
1347 copy_state_nonatomdata(state_local, state);
1352 /* There will be only one swap cycle with standard replica
1353 * exchange, but there may be multiple swap cycles if we
1354 * allow multiple swaps. */
1356 for (j = 0; j < maxswap; j++)
1358 exchange_partner = re->order[replica_id][j];
1360 if (exchange_partner != replica_id)
1362 /* Exchange the global states between the master nodes */
1365 fprintf(debug, "Exchanging %d with %d\n", replica_id, exchange_partner);
1367 exchange_state(cr->ms, exchange_partner, state);
1370 /* For temperature-type replica exchange, we need to scale
1371 * the velocities. */
1372 if (re->type == ereTEMP || re->type == ereTL)
1374 scale_velocities(state, sqrt(re->q[ereTEMP][replica_id]/re->q[ereTEMP][re->destinations[replica_id]]));
1379 /* With domain decomposition the global state is distributed later */
1380 if (!DOMAINDECOMP(cr))
1382 /* Copy the global state to the local state data structure */
1383 copy_state_nonatomdata(state, state_local);
1387 return bThisReplicaExchanged;
1390 void print_replica_exchange_statistics(FILE *fplog, struct gmx_repl_ex *re)
1394 fprintf(fplog, "\nReplica exchange statistics\n");
1398 fprintf(fplog, "Repl %d attempts, %d odd, %d even\n",
1399 re->nattempt[0]+re->nattempt[1], re->nattempt[1], re->nattempt[0]);
1401 fprintf(fplog, "Repl average probabilities:\n");
1402 for (i = 1; i < re->nrepl; i++)
1404 if (re->nattempt[i%2] == 0)
1410 re->prob[i] = re->prob_sum[i]/re->nattempt[i%2];
1413 print_ind(fplog, "", re->nrepl, re->ind, NULL);
1414 print_prob(fplog, "", re->nrepl, re->prob);
1416 fprintf(fplog, "Repl number of exchanges:\n");
1417 print_ind(fplog, "", re->nrepl, re->ind, NULL);
1418 print_count(fplog, "", re->nrepl, re->nexchange);
1420 fprintf(fplog, "Repl average number of exchanges:\n");
1421 for (i = 1; i < re->nrepl; i++)
1423 if (re->nattempt[i%2] == 0)
1429 re->prob[i] = ((real)re->nexchange[i])/re->nattempt[i%2];
1432 print_ind(fplog, "", re->nrepl, re->ind, NULL);
1433 print_prob(fplog, "", re->nrepl, re->prob);
1435 fprintf(fplog, "\n");
1437 /* print the transition matrix */
1438 print_transition_matrix(fplog, re->nrepl, re->nmoves, re->nattempt);