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) 2013,2014,2015,2016,2017,2018, 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.
51 #include "gromacs/commandline/filenm.h"
52 #include "gromacs/compat/make_unique.h"
53 #include "gromacs/domdec/domdec_struct.h"
54 #include "gromacs/domdec/localatomset.h"
55 #include "gromacs/domdec/localatomsetmanager.h"
56 #include "gromacs/fileio/gmxfio.h"
57 #include "gromacs/fileio/xvgr.h"
58 #include "gromacs/gmxlib/network.h"
59 #include "gromacs/math/functions.h"
60 #include "gromacs/math/units.h"
61 #include "gromacs/math/vec.h"
62 #include "gromacs/math/vectypes.h"
63 #include "gromacs/mdlib/gmx_omp_nthreads.h"
64 #include "gromacs/mdlib/mdrun.h"
65 #include "gromacs/mdtypes/commrec.h"
66 #include "gromacs/mdtypes/forceoutput.h"
67 #include "gromacs/mdtypes/inputrec.h"
68 #include "gromacs/mdtypes/md_enums.h"
69 #include "gromacs/mdtypes/mdatom.h"
70 #include "gromacs/pbcutil/pbc.h"
71 #include "gromacs/topology/mtop_lookup.h"
72 #include "gromacs/topology/topology.h"
73 #include "gromacs/utility/cstringutil.h"
74 #include "gromacs/utility/exceptions.h"
75 #include "gromacs/utility/fatalerror.h"
76 #include "gromacs/utility/futil.h"
77 #include "gromacs/utility/gmxassert.h"
78 #include "gromacs/utility/mutex.h"
79 #include "gromacs/utility/pleasecite.h"
80 #include "gromacs/utility/real.h"
81 #include "gromacs/utility/smalloc.h"
82 #include "gromacs/utility/strconvert.h"
84 #include "pull_internal.h"
86 static int groupPbcFromParams(const t_pull_group ¶ms)
93 else if (params.pbcatom >= 0)
103 /* NOTE: The params initialization currently copies pointers.
104 * So the lifetime of the source, currently always inputrec,
105 * should not end before that of this object.
106 * This will be fixed when the pointers are replacted by std::vector.
108 pull_group_work_t::pull_group_work_t(const t_pull_group ¶ms,
109 gmx::LocalAtomSet atomSet) :
111 epgrppbc(groupPbcFromParams(params)),
112 needToCalcCom(false),
122 bool pull_coordinate_is_angletype(const t_pull_coord *pcrd)
124 return (pcrd->eGeom == epullgANGLE ||
125 pcrd->eGeom == epullgDIHEDRAL ||
126 pcrd->eGeom == epullgANGLEAXIS);
129 static bool pull_coordinate_is_directional(const t_pull_coord *pcrd)
131 return (pcrd->eGeom == epullgDIR ||
132 pcrd->eGeom == epullgDIRPBC ||
133 pcrd->eGeom == epullgDIRRELATIVE ||
134 pcrd->eGeom == epullgCYL);
137 const char *pull_coordinate_units(const t_pull_coord *pcrd)
139 return pull_coordinate_is_angletype(pcrd) ? "deg" : "nm";
142 double pull_conversion_factor_userinput2internal(const t_pull_coord *pcrd)
144 if (pull_coordinate_is_angletype(pcrd))
154 double pull_conversion_factor_internal2userinput(const t_pull_coord *pcrd)
156 if (pull_coordinate_is_angletype(pcrd))
166 static std::string append_before_extension(const std::string &pathname,
167 const std::string &to_append)
169 /* Appends to_append before last '.' in pathname */
170 size_t extPos = pathname.find_last_of('.');
171 if (extPos == std::string::npos)
173 return pathname + to_append;
177 return pathname.substr(0, extPos) + to_append +
178 pathname.substr(extPos, std::string::npos);
182 static void pull_print_group_x(FILE *out, const ivec dim,
183 const pull_group_work_t *pgrp)
187 for (m = 0; m < DIM; m++)
191 fprintf(out, "\t%g", pgrp->x[m]);
196 static void pull_print_coord_dr_components(FILE *out, const ivec dim, const dvec dr)
198 for (int m = 0; m < DIM; m++)
202 fprintf(out, "\t%g", dr[m]);
207 static void pull_print_coord_dr(FILE *out, const pull_coord_work_t *pcrd,
208 gmx_bool bPrintRefValue,
209 gmx_bool bPrintComponents)
211 double unit_factor = pull_conversion_factor_internal2userinput(&pcrd->params);
213 fprintf(out, "\t%g", pcrd->spatialData.value*unit_factor);
217 fprintf(out, "\t%g", pcrd->value_ref*unit_factor);
220 if (bPrintComponents)
222 pull_print_coord_dr_components(out, pcrd->params.dim, pcrd->spatialData.dr01);
223 if (pcrd->params.ngroup >= 4)
225 pull_print_coord_dr_components(out, pcrd->params.dim, pcrd->spatialData.dr23);
227 if (pcrd->params.ngroup >= 6)
229 pull_print_coord_dr_components(out, pcrd->params.dim, pcrd->spatialData.dr45);
234 static void pull_print_x(FILE *out, struct pull_t *pull, double t)
236 fprintf(out, "%.4f", t);
238 for (size_t c = 0; c < pull->coord.size(); c++)
240 pull_coord_work_t *pcrd;
242 pcrd = &pull->coord[c];
244 pull_print_coord_dr(out, pcrd,
245 pull->params.bPrintRefValue && pcrd->params.eType != epullEXTERNAL,
246 pull->params.bPrintComp);
248 if (pull->params.bPrintCOM)
250 if (pcrd->params.eGeom == epullgCYL)
252 pull_print_group_x(out, pcrd->params.dim, &pull->dyna[c]);
256 pull_print_group_x(out, pcrd->params.dim,
257 &pull->group[pcrd->params.group[0]]);
259 for (int g = 1; g < pcrd->params.ngroup; g++)
261 pull_print_group_x(out, pcrd->params.dim, &pull->group[pcrd->params.group[g]]);
268 static void pull_print_f(FILE *out, const pull_t *pull, double t)
270 fprintf(out, "%.4f", t);
272 for (const pull_coord_work_t &coord : pull->coord)
274 fprintf(out, "\t%g", coord.scalarForce);
279 void pull_print_output(struct pull_t *pull, int64_t step, double time)
281 GMX_ASSERT(pull->numExternalPotentialsStillToBeAppliedThisStep == 0, "pull_print_output called before all external pull potentials have been applied");
283 if ((pull->params.nstxout != 0) && (step % pull->params.nstxout == 0))
285 pull_print_x(pull->out_x, pull, time);
288 if ((pull->params.nstfout != 0) && (step % pull->params.nstfout == 0))
290 pull_print_f(pull->out_f, pull, time);
294 static void set_legend_for_coord_components(const pull_coord_work_t *pcrd, int coord_index, char **setname, int *nsets_ptr)
296 /* Loop over the distance vectors and print their components. Each vector is made up of two consecutive groups. */
297 for (int g = 0; g < pcrd->params.ngroup; g += 2)
299 /* Loop over the components */
300 for (int m = 0; m < DIM; m++)
302 if (pcrd->params.dim[m])
306 if (g == 0 && pcrd->params.ngroup <= 2)
308 /* For the simplest case we print a simplified legend without group indices, just the cooordinate index
309 and which dimensional component it is. */
310 sprintf(legend, "%d d%c", coord_index + 1, 'X' + m);
314 /* Otherwise, print also the group indices (relative to the coordinate, not the global ones). */
315 sprintf(legend, "%d g %d-%d d%c", coord_index + 1, g + 1, g + 2, 'X' + m);
318 setname[*nsets_ptr] = gmx_strdup(legend);
325 static FILE *open_pull_out(const char *fn, struct pull_t *pull,
326 const gmx_output_env_t *oenv,
328 const ContinuationOptions &continuationOptions)
332 char **setname, buf[50];
334 if (continuationOptions.appendFiles)
336 fp = gmx_fio_fopen(fn, "a+");
340 fp = gmx_fio_fopen(fn, "w+");
343 sprintf(buf, "Position (nm%s)", pull->bAngle ? ", deg" : "");
344 xvgr_header(fp, "Pull COM", "Time (ps)", buf,
349 sprintf(buf, "Force (kJ/mol/nm%s)", pull->bAngle ? ", kJ/mol/rad" : "");
350 xvgr_header(fp, "Pull force", "Time (ps)", buf,
354 /* With default mdp options only the actual coordinate value is printed (1),
355 * but optionally the reference value (+ 1),
356 * the group COMs for all the groups (+ ngroups_max*DIM)
357 * and the components of the distance vectors can be printed (+ (ngroups_max/2)*DIM).
359 snew(setname, pull->coord.size()*(1 + 1 + c_pullCoordNgroupMax*DIM + c_pullCoordNgroupMax/2*DIM));
362 for (size_t c = 0; c < pull->coord.size(); c++)
366 /* The order of this legend should match the order of printing
367 * the data in print_pull_x above.
370 /* The pull coord distance */
371 sprintf(buf, "%zu", c+1);
372 setname[nsets] = gmx_strdup(buf);
374 if (pull->params.bPrintRefValue &&
375 pull->coord[c].params.eType != epullEXTERNAL)
377 sprintf(buf, "%zu ref", c+1);
378 setname[nsets] = gmx_strdup(buf);
381 if (pull->params.bPrintComp)
383 set_legend_for_coord_components(&pull->coord[c], c, setname, &nsets);
386 if (pull->params.bPrintCOM)
388 for (int g = 0; g < pull->coord[c].params.ngroup; g++)
390 /* Legend for reference group position */
391 for (m = 0; m < DIM; m++)
393 if (pull->coord[c].params.dim[m])
395 sprintf(buf, "%zu g %d %c", c+1, g + 1, 'X'+m);
396 setname[nsets] = gmx_strdup(buf);
405 /* For the pull force we always only use one scalar */
406 sprintf(buf, "%zu", c+1);
407 setname[nsets] = gmx_strdup(buf);
413 xvgr_legend(fp, nsets, setname, oenv);
415 for (int c = 0; c < nsets; c++)
425 /* Apply forces in a mass weighted fashion for part of the pull group */
426 static void apply_forces_grp_part(const pull_group_work_t *pgrp,
427 int ind_start, int ind_end,
429 const dvec f_pull, int sign, rvec *f)
431 double inv_wm = pgrp->mwscale;
433 auto localAtomIndices = pgrp->atomSet.localIndex();
434 for (int i = ind_start; i < ind_end; i++)
436 int ii = localAtomIndices[i];
437 double wmass = md->massT[ii];
438 if (!pgrp->localWeights.empty())
440 wmass *= pgrp->localWeights[i];
443 for (int d = 0; d < DIM; d++)
445 f[ii][d] += sign * wmass * f_pull[d] * inv_wm;
450 /* Apply forces in a mass weighted fashion */
451 static void apply_forces_grp(const pull_group_work_t *pgrp,
453 const dvec f_pull, int sign, rvec *f,
456 auto localAtomIndices = pgrp->atomSet.localIndex();
458 if (pgrp->params.nat == 1 && pgrp->atomSet.numAtomsLocal() == 1)
460 /* Only one atom and our rank has this atom: we can skip
461 * the mass weighting, which means that this code also works
462 * for mass=0, e.g. with a virtual site.
464 for (int d = 0; d < DIM; d++)
466 f[localAtomIndices[0]][d] += sign*f_pull[d];
471 if (localAtomIndices.size() <= c_pullMaxNumLocalAtomsSingleThreaded)
473 apply_forces_grp_part(pgrp, 0, localAtomIndices.size(), md, f_pull, sign, f);
477 #pragma omp parallel for num_threads(nthreads) schedule(static)
478 for (int th = 0; th < nthreads; th++)
480 int ind_start = (localAtomIndices.size()*(th + 0))/nthreads;
481 int ind_end = (localAtomIndices.size()*(th + 1))/nthreads;
482 apply_forces_grp_part(pgrp, ind_start, ind_end,
483 md, f_pull, sign, f);
489 /* Apply forces in a mass weighted fashion to a cylinder group */
490 static void apply_forces_cyl_grp(const pull_group_work_t *pgrp,
491 const double dv_corr,
493 const dvec f_pull, double f_scal,
495 int gmx_unused nthreads)
497 double inv_wm = pgrp->mwscale;
499 auto localAtomIndices = pgrp->atomSet.localIndex();
501 /* The cylinder group is always a slab in the system, thus large.
502 * Therefore we always thread-parallelize this group.
504 int numAtomsLocal = localAtomIndices.size();
505 #pragma omp parallel for num_threads(nthreads) schedule(static)
506 for (int i = 0; i < numAtomsLocal; i++)
508 double weight = pgrp->localWeights[i];
513 int ii = localAtomIndices[i];
514 double mass = md->massT[ii];
515 /* The stored axial distance from the cylinder center (dv) needs
516 * to be corrected for an offset (dv_corr), which was unknown when
519 double dv_com = pgrp->dv[i] + dv_corr;
521 /* Here we not only add the pull force working along vec (f_pull),
522 * but also a radial component, due to the dependence of the weights
523 * on the radial distance.
525 for (int m = 0; m < DIM; m++)
527 f[ii][m] += sign*inv_wm*(mass*weight*f_pull[m] +
528 pgrp->mdw[i][m]*dv_com*f_scal);
533 /* Apply torque forces in a mass weighted fashion to the groups that define
534 * the pull vector direction for pull coordinate pcrd.
536 static void apply_forces_vec_torque(const struct pull_t *pull,
537 const pull_coord_work_t *pcrd,
541 const PullCoordSpatialData &spatialData = pcrd->spatialData;
543 /* The component inpr along the pull vector is accounted for in the usual
544 * way. Here we account for the component perpendicular to vec.
547 for (int m = 0; m < DIM; m++)
549 inpr += spatialData.dr01[m]*spatialData.vec[m];
551 /* The torque force works along the component of the distance vector
552 * of between the two "usual" pull groups that is perpendicular to
553 * the pull vector. The magnitude of this force is the "usual" scale force
554 * multiplied with the ratio of the distance between the two "usual" pull
555 * groups and the distance between the two groups that define the vector.
558 for (int m = 0; m < DIM; m++)
560 f_perp[m] = (spatialData.dr01[m] - inpr*spatialData.vec[m])/spatialData.vec_len*pcrd->scalarForce;
563 /* Apply the force to the groups defining the vector using opposite signs */
564 apply_forces_grp(&pull->group[pcrd->params.group[2]], md,
565 f_perp, -1, f, pull->nthreads);
566 apply_forces_grp(&pull->group[pcrd->params.group[3]], md,
567 f_perp, 1, f, pull->nthreads);
570 /* Apply forces in a mass weighted fashion */
571 static void apply_forces_coord(struct pull_t * pull, int coord,
572 const PullCoordVectorForces &forces,
573 const t_mdatoms * md,
576 /* Here it would be more efficient to use one large thread-parallel
577 * region instead of potential parallel regions within apply_forces_grp.
578 * But there could be overlap between pull groups and this would lead
582 const pull_coord_work_t &pcrd = pull->coord[coord];
584 if (pcrd.params.eGeom == epullgCYL)
586 apply_forces_cyl_grp(&pull->dyna[coord], pcrd.spatialData.cyl_dev, md,
587 forces.force01, pcrd.scalarForce, -1, f,
590 /* Sum the force along the vector and the radial force */
592 for (int m = 0; m < DIM; m++)
594 f_tot[m] = forces.force01[m] + pcrd.scalarForce*pcrd.spatialData.ffrad[m];
596 apply_forces_grp(&pull->group[pcrd.params.group[1]], md,
597 f_tot, 1, f, pull->nthreads);
601 if (pcrd.params.eGeom == epullgDIRRELATIVE)
603 /* We need to apply the torque forces to the pull groups
604 * that define the pull vector.
606 apply_forces_vec_torque(pull, &pcrd, md, f);
609 if (pull->group[pcrd.params.group[0]].params.nat > 0)
611 apply_forces_grp(&pull->group[pcrd.params.group[0]], md,
612 forces.force01, -1, f, pull->nthreads);
614 apply_forces_grp(&pull->group[pcrd.params.group[1]], md,
615 forces.force01, 1, f, pull->nthreads);
617 if (pcrd.params.ngroup >= 4)
619 apply_forces_grp(&pull->group[pcrd.params.group[2]], md,
620 forces.force23, -1, f, pull->nthreads);
621 apply_forces_grp(&pull->group[pcrd.params.group[3]], md,
622 forces.force23, 1, f, pull->nthreads);
624 if (pcrd.params.ngroup >= 6)
626 apply_forces_grp(&pull->group[pcrd.params.group[4]], md,
627 forces.force45, -1, f, pull->nthreads);
628 apply_forces_grp(&pull->group[pcrd.params.group[5]], md,
629 forces.force45, 1, f, pull->nthreads);
634 real max_pull_distance2(const pull_coord_work_t *pcrd,
637 /* Note that this maximum distance calculation is more complex than
638 * most other cases in GROMACS, since here we have to take care of
639 * distance calculations that don't involve all three dimensions.
640 * For example, we can use distances that are larger than the
641 * box X and Y dimensions for a box that is elongated along Z.
644 real max_d2 = GMX_REAL_MAX;
646 if (pull_coordinate_is_directional(&pcrd->params))
648 /* Directional pulling along along direction pcrd->vec.
649 * Calculating the exact maximum distance is complex and bug-prone.
650 * So we take a safe approach by not allowing distances that
651 * are larger than half the distance between unit cell faces
652 * along dimensions involved in pcrd->vec.
654 for (int m = 0; m < DIM; m++)
656 if (m < pbc->ndim_ePBC && pcrd->spatialData.vec[m] != 0)
658 real imageDistance2 = gmx::square(pbc->box[m][m]);
659 for (int d = m + 1; d < DIM; d++)
661 imageDistance2 -= gmx::square(pbc->box[d][m]);
663 max_d2 = std::min(max_d2, imageDistance2);
669 /* Distance pulling along dimensions with pcrd->params.dim[d]==1.
670 * We use half the minimum box vector length of the dimensions involved.
671 * This is correct for all cases, except for corner cases with
672 * triclinic boxes where e.g. dim[XX]=1 and dim[YY]=0 with
673 * box[YY][XX]!=0 and box[YY][YY] < box[XX][XX]. But since even
674 * in such corner cases the user could get correct results,
675 * depending on the details of the setup, we avoid further
676 * code complications.
678 for (int m = 0; m < DIM; m++)
680 if (m < pbc->ndim_ePBC && pcrd->params.dim[m] != 0)
682 real imageDistance2 = gmx::square(pbc->box[m][m]);
683 for (int d = 0; d < m; d++)
685 if (pcrd->params.dim[d] != 0)
687 imageDistance2 += gmx::square(pbc->box[m][d]);
690 max_d2 = std::min(max_d2, imageDistance2);
698 /* This function returns the distance based on coordinates xg and xref.
699 * Note that the pull coordinate struct pcrd is not modified.
701 static void low_get_pull_coord_dr(const struct pull_t *pull,
702 const pull_coord_work_t *pcrd,
704 dvec xg, dvec xref, double max_dist2,
707 const pull_group_work_t *pgrp0 = &pull->group[pcrd->params.group[0]];
709 /* Only the first group can be an absolute reference, in that case nat=0 */
710 if (pgrp0->params.nat == 0)
712 for (int m = 0; m < DIM; m++)
714 xref[m] = pcrd->params.origin[m];
719 copy_dvec(xref, xrefr);
721 dvec dref = {0, 0, 0};
722 if (pcrd->params.eGeom == epullgDIRPBC)
724 for (int m = 0; m < DIM; m++)
726 dref[m] = pcrd->value_ref*pcrd->spatialData.vec[m];
728 /* Add the reference position, so we use the correct periodic image */
729 dvec_inc(xrefr, dref);
732 pbc_dx_d(pbc, xg, xrefr, dr);
734 bool directional = pull_coordinate_is_directional(&pcrd->params);
736 for (int m = 0; m < DIM; m++)
738 dr[m] *= pcrd->params.dim[m];
739 if (pcrd->params.dim[m] && !(directional && pcrd->spatialData.vec[m] == 0))
744 /* Check if we are close to switching to another periodic image */
745 if (max_dist2 > 0 && dr2 > 0.98*0.98*max_dist2)
747 /* Note that technically there is no issue with switching periodic
748 * image, as pbc_dx_d returns the distance to the closest periodic
749 * image. However in all cases where periodic image switches occur,
750 * the pull results will be useless in practice.
752 gmx_fatal(FARGS, "Distance between pull groups %d and %d (%f nm) is larger than 0.49 times the box size (%f).\n%s",
753 pcrd->params.group[0], pcrd->params.group[1],
754 sqrt(dr2), sqrt(0.98*0.98*max_dist2),
755 pcrd->params.eGeom == epullgDIR ? "You might want to consider using \"pull-geometry = direction-periodic\" instead.\n" : "");
758 if (pcrd->params.eGeom == epullgDIRPBC)
764 /* This function returns the distance based on the contents of the pull struct.
765 * pull->coord[coord_ind].dr, and potentially vector, are updated.
767 static void get_pull_coord_dr(struct pull_t *pull,
771 pull_coord_work_t *pcrd = &pull->coord[coord_ind];
772 PullCoordSpatialData &spatialData = pcrd->spatialData;
775 if (pcrd->params.eGeom == epullgDIRPBC)
781 md2 = static_cast<double>(max_pull_distance2(pcrd, pbc));
784 if (pcrd->params.eGeom == epullgDIRRELATIVE)
786 /* We need to determine the pull vector */
790 const pull_group_work_t &pgrp2 = pull->group[pcrd->params.group[2]];
791 const pull_group_work_t &pgrp3 = pull->group[pcrd->params.group[3]];
793 pbc_dx_d(pbc, pgrp3.x, pgrp2.x, vec);
795 for (m = 0; m < DIM; m++)
797 vec[m] *= pcrd->params.dim[m];
799 spatialData.vec_len = dnorm(vec);
800 for (m = 0; m < DIM; m++)
802 spatialData.vec[m] = vec[m]/spatialData.vec_len;
806 fprintf(debug, "pull coord %d vector: %6.3f %6.3f %6.3f normalized: %6.3f %6.3f %6.3f\n",
808 vec[XX], vec[YY], vec[ZZ],
809 spatialData.vec[XX], spatialData.vec[YY], spatialData.vec[ZZ]);
813 pull_group_work_t *pgrp0 = &pull->group[pcrd->params.group[0]];
814 pull_group_work_t *pgrp1 = &pull->group[pcrd->params.group[1]];
816 low_get_pull_coord_dr(pull, pcrd, pbc,
818 pcrd->params.eGeom == epullgCYL ? pull->dyna[coord_ind].x : pgrp0->x,
822 if (pcrd->params.ngroup >= 4)
824 pull_group_work_t *pgrp2, *pgrp3;
825 pgrp2 = &pull->group[pcrd->params.group[2]];
826 pgrp3 = &pull->group[pcrd->params.group[3]];
828 low_get_pull_coord_dr(pull, pcrd, pbc,
834 if (pcrd->params.ngroup >= 6)
836 pull_group_work_t *pgrp4, *pgrp5;
837 pgrp4 = &pull->group[pcrd->params.group[4]];
838 pgrp5 = &pull->group[pcrd->params.group[5]];
840 low_get_pull_coord_dr(pull, pcrd, pbc,
848 /* Modify x so that it is periodic in [-pi, pi)
849 * It is assumed that x is in [-3pi, 3pi) so that x
850 * needs to be shifted by at most one period.
852 static void make_periodic_2pi(double *x)
864 /* This function should always be used to modify pcrd->value_ref */
865 static void low_set_pull_coord_reference_value(pull_coord_work_t *pcrd,
869 GMX_ASSERT(pcrd->params.eType != epullEXTERNAL, "The pull coord reference value should not be used with type external-potential");
871 if (pcrd->params.eGeom == epullgDIST)
875 gmx_fatal(FARGS, "Pull reference distance for coordinate %d (%f) needs to be non-negative", coord_ind + 1, value_ref);
878 else if (pcrd->params.eGeom == epullgANGLE || pcrd->params.eGeom == epullgANGLEAXIS)
880 if (value_ref < 0 || value_ref > M_PI)
882 gmx_fatal(FARGS, "Pull reference angle for coordinate %d (%f) needs to be in the allowed interval [0,180] deg", coord_ind + 1, value_ref*pull_conversion_factor_internal2userinput(&pcrd->params));
885 else if (pcrd->params.eGeom == epullgDIHEDRAL)
887 /* Allow pulling to be periodic for dihedral angles by remapping the reference value to the interval [-pi, pi). */
888 make_periodic_2pi(&value_ref);
891 pcrd->value_ref = value_ref;
894 static void update_pull_coord_reference_value(pull_coord_work_t *pcrd, int coord_ind, double t)
896 /* With zero rate the reference value is set initially and doesn't change */
897 if (pcrd->params.rate != 0)
899 double value_ref = (pcrd->params.init + pcrd->params.rate*t)*pull_conversion_factor_userinput2internal(&pcrd->params);
900 low_set_pull_coord_reference_value(pcrd, coord_ind, value_ref);
904 /* Returns the dihedral angle. Updates the plane normal vectors m, n. */
905 static double get_dihedral_angle_coord(PullCoordSpatialData *spatialData)
908 dvec dr32; /* store instead of dr23? */
910 dsvmul(-1, spatialData->dr23, dr32);
911 dcprod(spatialData->dr01, dr32, spatialData->planevec_m); /* Normal of first plane */
912 dcprod(dr32, spatialData->dr45, spatialData->planevec_n); /* Normal of second plane */
913 phi = gmx_angle_between_dvecs(spatialData->planevec_m, spatialData->planevec_n);
915 /* Note 1: the sign below is opposite of that in the bondeds or Bekker 1994
916 * because there r_ij = ri - rj, while here dr01 = r_1 - r_0
917 * Note 2: the angle between the plane normal vectors equals pi only when
918 * both planes coincide. Thus, when phi = pi dr01 will lie in both planes and
919 * we get a positive sign below. Thus, the range of the dihedral angle is (-180, 180].
921 sign = (diprod(spatialData->dr01, spatialData->planevec_n) < 0.0) ? 1.0 : -1.0;
925 /* Calculates pull->coord[coord_ind].value.
926 * This function also updates pull->coord[coord_ind].dr.
928 static void get_pull_coord_distance(struct pull_t *pull,
932 pull_coord_work_t *pcrd;
935 pcrd = &pull->coord[coord_ind];
937 get_pull_coord_dr(pull, coord_ind, pbc);
939 PullCoordSpatialData &spatialData = pcrd->spatialData;
941 switch (pcrd->params.eGeom)
944 /* Pull along the vector between the com's */
945 spatialData.value = dnorm(spatialData.dr01);
949 case epullgDIRRELATIVE:
952 spatialData.value = 0;
953 for (m = 0; m < DIM; m++)
955 spatialData.value += spatialData.vec[m]*spatialData.dr01[m];
959 spatialData.value = gmx_angle_between_dvecs(spatialData.dr01, spatialData.dr23);
962 spatialData.value = get_dihedral_angle_coord(&spatialData);
964 case epullgANGLEAXIS:
965 spatialData.value = gmx_angle_between_dvecs(spatialData.dr01, spatialData.vec);
968 gmx_incons("Unsupported pull type in get_pull_coord_distance");
972 /* Returns the deviation from the reference value.
973 * Updates pull->coord[coord_ind].dr, .value and .value_ref.
975 static double get_pull_coord_deviation(struct pull_t *pull,
980 pull_coord_work_t *pcrd;
983 pcrd = &pull->coord[coord_ind];
985 /* Update the reference value before computing the distance,
986 * since it is used in the distance computation with periodic pulling.
988 update_pull_coord_reference_value(pcrd, coord_ind, t);
990 get_pull_coord_distance(pull, coord_ind, pbc);
992 get_pull_coord_distance(pull, coord_ind, pbc);
994 /* Determine the deviation */
995 dev = pcrd->spatialData.value - pcrd->value_ref;
997 /* Check that values are allowed */
998 if (pcrd->params.eGeom == epullgDIST && pcrd->spatialData.value == 0)
1000 /* With no vector we can not determine the direction for the force,
1001 * so we set the force to zero.
1005 else if (pcrd->params.eGeom == epullgDIHEDRAL)
1007 /* The reference value is in [-pi, pi). The coordinate value is in (-pi, pi].
1008 Thus, the unwrapped deviation is here in (-2pi, 2pi].
1009 After making it periodic, the deviation will be in [-pi, pi). */
1010 make_periodic_2pi(&dev);
1016 double get_pull_coord_value(struct pull_t *pull,
1020 get_pull_coord_distance(pull, coord_ind, pbc);
1022 return pull->coord[coord_ind].spatialData.value;
1025 void clear_pull_forces(pull_t *pull)
1027 /* Zeroing the forces is only required for constraint pulling.
1028 * It can happen that multiple constraint steps need to be applied
1029 * and therefore the constraint forces need to be accumulated.
1031 for (pull_coord_work_t &coord : pull->coord)
1033 coord.scalarForce = 0;
1037 /* Apply constraint using SHAKE */
1038 static void do_constraint(struct pull_t *pull, t_pbc *pbc,
1040 gmx_bool bMaster, tensor vir,
1041 double dt, double t)
1044 dvec *r_ij; /* x[i] com of i in prev. step. Obeys constr. -> r_ij[i] */
1045 dvec unc_ij; /* xp[i] com of i this step, before constr. -> unc_ij */
1046 dvec *rnew; /* current 'new' positions of the groups */
1047 double *dr_tot; /* the total update of the coords */
1050 double lambda, rm, invdt = 0;
1051 gmx_bool bConverged_all, bConverged = FALSE;
1052 int niter = 0, ii, j, m, max_iter = 100;
1056 snew(r_ij, pull->coord.size());
1057 snew(dr_tot, pull->coord.size());
1059 snew(rnew, pull->group.size());
1061 /* copy the current unconstrained positions for use in iterations. We
1062 iterate until rinew[i] and rjnew[j] obey the constraints. Then
1063 rinew - pull.x_unc[i] is the correction dr to group i */
1064 for (size_t g = 0; g < pull->group.size(); g++)
1066 copy_dvec(pull->group[g].xp, rnew[g]);
1069 /* Determine the constraint directions from the old positions */
1070 for (size_t c = 0; c < pull->coord.size(); c++)
1072 pull_coord_work_t *pcrd;
1074 pcrd = &pull->coord[c];
1076 if (pcrd->params.eType != epullCONSTRAINT)
1081 /* Note that get_pull_coord_distance also sets pcrd->dr and pcrd->value.
1082 * We don't modify dr and value anymore, so these values are also used
1085 get_pull_coord_distance(pull, c, pbc);
1087 const PullCoordSpatialData &spatialData = pcrd->spatialData;
1090 fprintf(debug, "Pull coord %zu dr %f %f %f\n",
1091 c, spatialData.dr01[XX], spatialData.dr01[YY], spatialData.dr01[ZZ]);
1094 if (pcrd->params.eGeom == epullgDIR ||
1095 pcrd->params.eGeom == epullgDIRPBC)
1097 /* Select the component along vec */
1099 for (m = 0; m < DIM; m++)
1101 a += spatialData.vec[m]*spatialData.dr01[m];
1103 for (m = 0; m < DIM; m++)
1105 r_ij[c][m] = a*spatialData.vec[m];
1110 copy_dvec(spatialData.dr01, r_ij[c]);
1113 if (dnorm2(r_ij[c]) == 0)
1115 gmx_fatal(FARGS, "Distance for pull coordinate %lu is zero with constraint pulling, which is not allowed.", c + 1);
1119 bConverged_all = FALSE;
1120 while (!bConverged_all && niter < max_iter)
1122 bConverged_all = TRUE;
1124 /* loop over all constraints */
1125 for (size_t c = 0; c < pull->coord.size(); c++)
1127 pull_coord_work_t *pcrd;
1128 pull_group_work_t *pgrp0, *pgrp1;
1131 pcrd = &pull->coord[c];
1133 if (pcrd->params.eType != epullCONSTRAINT)
1138 update_pull_coord_reference_value(pcrd, c, t);
1140 pgrp0 = &pull->group[pcrd->params.group[0]];
1141 pgrp1 = &pull->group[pcrd->params.group[1]];
1143 /* Get the current difference vector */
1144 low_get_pull_coord_dr(pull, pcrd, pbc,
1145 rnew[pcrd->params.group[1]],
1146 rnew[pcrd->params.group[0]],
1151 fprintf(debug, "Pull coord %zu, iteration %d\n", c, niter);
1154 rm = 1.0/(pgrp0->invtm + pgrp1->invtm);
1156 switch (pcrd->params.eGeom)
1159 if (pcrd->value_ref <= 0)
1161 gmx_fatal(FARGS, "The pull constraint reference distance for group %zu is <= 0 (%f)", c, pcrd->value_ref);
1165 double q, c_a, c_b, c_c;
1167 c_a = diprod(r_ij[c], r_ij[c]);
1168 c_b = diprod(unc_ij, r_ij[c])*2;
1169 c_c = diprod(unc_ij, unc_ij) - gmx::square(pcrd->value_ref);
1173 q = -0.5*(c_b - std::sqrt(c_b*c_b - 4*c_a*c_c));
1178 q = -0.5*(c_b + std::sqrt(c_b*c_b - 4*c_a*c_c));
1185 "Pull ax^2+bx+c=0: a=%e b=%e c=%e lambda=%e\n",
1186 c_a, c_b, c_c, lambda);
1190 /* The position corrections dr due to the constraints */
1191 dsvmul(-lambda*rm*pgrp1->invtm, r_ij[c], dr1);
1192 dsvmul( lambda*rm*pgrp0->invtm, r_ij[c], dr0);
1193 dr_tot[c] += -lambda*dnorm(r_ij[c]);
1198 /* A 1-dimensional constraint along a vector */
1200 for (m = 0; m < DIM; m++)
1202 vec[m] = pcrd->spatialData.vec[m];
1203 a += unc_ij[m]*vec[m];
1205 /* Select only the component along the vector */
1206 dsvmul(a, vec, unc_ij);
1207 lambda = a - pcrd->value_ref;
1210 fprintf(debug, "Pull inpr %e lambda: %e\n", a, lambda);
1213 /* The position corrections dr due to the constraints */
1214 dsvmul(-lambda*rm*pgrp1->invtm, vec, dr1);
1215 dsvmul( lambda*rm*pgrp0->invtm, vec, dr0);
1216 dr_tot[c] += -lambda;
1219 gmx_incons("Invalid enumeration value for eGeom");
1227 g0 = pcrd->params.group[0];
1228 g1 = pcrd->params.group[1];
1229 low_get_pull_coord_dr(pull, pcrd, pbc, rnew[g1], rnew[g0], -1, tmp);
1230 low_get_pull_coord_dr(pull, pcrd, pbc, dr1, dr0, -1, tmp3);
1232 "Pull cur %8.5f %8.5f %8.5f j:%8.5f %8.5f %8.5f d: %8.5f\n",
1233 rnew[g0][0], rnew[g0][1], rnew[g0][2],
1234 rnew[g1][0], rnew[g1][1], rnew[g1][2], dnorm(tmp));
1236 "Pull ref %8s %8s %8s %8s %8s %8s d: %8.5f\n",
1237 "", "", "", "", "", "", pcrd->value_ref);
1239 "Pull cor %8.5f %8.5f %8.5f j:%8.5f %8.5f %8.5f d: %8.5f\n",
1240 dr0[0], dr0[1], dr0[2],
1241 dr1[0], dr1[1], dr1[2],
1245 /* Update the COMs with dr */
1246 dvec_inc(rnew[pcrd->params.group[1]], dr1);
1247 dvec_inc(rnew[pcrd->params.group[0]], dr0);
1250 /* Check if all constraints are fullfilled now */
1251 for (pull_coord_work_t &coord : pull->coord)
1253 if (coord.params.eType != epullCONSTRAINT)
1258 low_get_pull_coord_dr(pull, &coord, pbc,
1259 rnew[coord.params.group[1]],
1260 rnew[coord.params.group[0]],
1263 switch (coord.params.eGeom)
1267 fabs(dnorm(unc_ij) - coord.value_ref) < pull->params.constr_tol;
1272 for (m = 0; m < DIM; m++)
1274 vec[m] = coord.spatialData.vec[m];
1276 inpr = diprod(unc_ij, vec);
1277 dsvmul(inpr, vec, unc_ij);
1279 fabs(diprod(unc_ij, vec) - coord.value_ref) < pull->params.constr_tol;
1287 fprintf(debug, "Pull constraint not converged: "
1289 "d_ref = %f, current d = %f\n",
1290 coord.params.group[0], coord.params.group[1],
1291 coord.value_ref, dnorm(unc_ij));
1294 bConverged_all = FALSE;
1299 /* if after all constraints are dealt with and bConverged is still TRUE
1300 we're finished, if not we do another iteration */
1302 if (niter > max_iter)
1304 gmx_fatal(FARGS, "Too many iterations for constraint run: %d", niter);
1307 /* DONE ITERATING, NOW UPDATE COORDINATES AND CALC. CONSTRAINT FORCES */
1314 /* update atoms in the groups */
1315 for (size_t g = 0; g < pull->group.size(); g++)
1317 const pull_group_work_t *pgrp;
1320 pgrp = &pull->group[g];
1322 /* get the final constraint displacement dr for group g */
1323 dvec_sub(rnew[g], pgrp->xp, dr);
1325 if (dnorm2(dr) == 0)
1327 /* No displacement, no update necessary */
1331 /* update the atom positions */
1332 auto localAtomIndices = pgrp->atomSet.localIndex();
1334 for (gmx::index j = 0; j < localAtomIndices.size(); j++)
1336 ii = localAtomIndices[j];
1337 if (!pgrp->localWeights.empty())
1339 dsvmul(pgrp->wscale*pgrp->localWeights[j], dr, tmp);
1341 for (m = 0; m < DIM; m++)
1347 for (m = 0; m < DIM; m++)
1349 v[ii][m] += invdt*tmp[m];
1355 /* calculate the constraint forces, used for output and virial only */
1356 for (size_t c = 0; c < pull->coord.size(); c++)
1358 pull_coord_work_t *pcrd;
1360 pcrd = &pull->coord[c];
1362 if (pcrd->params.eType != epullCONSTRAINT)
1367 /* Accumulate the forces, in case we have multiple constraint steps */
1368 double force = dr_tot[c]/((pull->group[pcrd->params.group[0]].invtm + pull->group[pcrd->params.group[1]].invtm)*dt*dt);
1369 pcrd->scalarForce += force;
1371 if (vir != nullptr && pcrd->params.eGeom != epullgDIRPBC && bMaster)
1375 /* Add the pull contribution to the virial */
1376 /* We have already checked above that r_ij[c] != 0 */
1377 f_invr = pcrd->scalarForce/dnorm(r_ij[c]);
1379 for (j = 0; j < DIM; j++)
1381 for (m = 0; m < DIM; m++)
1383 vir[j][m] -= 0.5*f_invr*r_ij[c][j]*r_ij[c][m];
1389 /* finished! I hope. Give back some memory */
1395 static void add_virial_coord_dr(tensor vir, const dvec dr, const dvec f)
1397 for (int j = 0; j < DIM; j++)
1399 for (int m = 0; m < DIM; m++)
1401 vir[j][m] -= 0.5*f[j]*dr[m];
1406 /* Adds the pull contribution to the virial */
1407 static void add_virial_coord(tensor vir,
1408 const pull_coord_work_t &pcrd,
1409 const PullCoordVectorForces &forces)
1411 if (vir != nullptr && pcrd.params.eGeom != epullgDIRPBC)
1413 /* Add the pull contribution for each distance vector to the virial. */
1414 add_virial_coord_dr(vir, pcrd.spatialData.dr01, forces.force01);
1415 if (pcrd.params.ngroup >= 4)
1417 add_virial_coord_dr(vir, pcrd.spatialData.dr23, forces.force23);
1419 if (pcrd.params.ngroup >= 6)
1421 add_virial_coord_dr(vir, pcrd.spatialData.dr45, forces.force45);
1426 static void calc_pull_coord_scalar_force_and_potential(pull_coord_work_t *pcrd,
1427 double dev, real lambda,
1428 real *V, real *dVdl)
1432 k = (1.0 - lambda)*pcrd->params.k + lambda*pcrd->params.kB;
1433 dkdl = pcrd->params.kB - pcrd->params.k;
1435 switch (pcrd->params.eType)
1438 case epullFLATBOTTOM:
1439 case epullFLATBOTTOMHIGH:
1440 /* The only difference between an umbrella and a flat-bottom
1441 * potential is that a flat-bottom is zero above or below
1442 the reference value.
1444 if ((pcrd->params.eType == epullFLATBOTTOM && dev < 0) ||
1445 (pcrd->params.eType == epullFLATBOTTOMHIGH && dev > 0))
1450 pcrd->scalarForce = -k*dev;
1451 *V += 0.5* k*gmx::square(dev);
1452 *dVdl += 0.5*dkdl*gmx::square(dev);
1455 pcrd->scalarForce = -k;
1456 *V += k*pcrd->spatialData.value;
1457 *dVdl += dkdl*pcrd->spatialData.value;
1460 gmx_incons("the scalar pull force should not be calculated internally for pull type external");
1462 gmx_incons("Unsupported pull type in do_pull_pot");
1466 static PullCoordVectorForces
1467 calculateVectorForces(const pull_coord_work_t &pcrd)
1469 const t_pull_coord ¶ms = pcrd.params;
1470 const PullCoordSpatialData &spatialData = pcrd.spatialData;
1472 /* The geometry of the coordinate determines how the scalar force relates to the force on each group */
1473 PullCoordVectorForces forces;
1475 if (params.eGeom == epullgDIST)
1477 double invdr01 = spatialData.value > 0 ? 1./spatialData.value : 0.;
1478 for (int m = 0; m < DIM; m++)
1480 forces.force01[m] = pcrd.scalarForce*spatialData.dr01[m]*invdr01;
1483 else if (params.eGeom == epullgANGLE)
1486 double cos_theta, cos_theta2;
1488 cos_theta = cos(spatialData.value);
1489 cos_theta2 = gmx::square(cos_theta);
1491 /* The force at theta = 0, pi is undefined so we should not apply any force.
1492 * cos(theta) = 1 for theta = 0, pi so this is what we check for below.
1493 * Note that dr01 = 0 or dr23 = 0 evaluates to theta = 0 so we do not
1494 * have to check for this before dividing by their norm below.
1498 /* The forces f01 and f23 can be decomposed into one vector parallel to dr01
1499 * and another vector parallel to dr23:
1500 * f01 = -dV/dtheta*(a*dr23/|dr23| - b*dr01*|dr01|)/|dr01|,
1501 * f23 = -dV/dtheta*(a*dr01/|dr01| - b*dr23*|dr23|)/|dr23|,
1503 double a = -gmx::invsqrt(1 - cos_theta2); /* comes from d/dx acos(x) */
1504 double b = a*cos_theta;
1505 double invdr01 = 1./dnorm(spatialData.dr01);
1506 double invdr23 = 1./dnorm(spatialData.dr23);
1507 dvec normalized_dr01, normalized_dr23;
1508 dsvmul(invdr01, spatialData.dr01, normalized_dr01);
1509 dsvmul(invdr23, spatialData.dr23, normalized_dr23);
1511 for (int m = 0; m < DIM; m++)
1513 /* Here, f_scal is -dV/dtheta */
1514 forces.force01[m] = pcrd.scalarForce*invdr01*(a*normalized_dr23[m] - b*normalized_dr01[m]);
1515 forces.force23[m] = pcrd.scalarForce*invdr23*(a*normalized_dr01[m] - b*normalized_dr23[m]);
1520 /* No forces to apply for ill-defined cases*/
1521 clear_dvec(forces.force01);
1522 clear_dvec(forces.force23);
1525 else if (params.eGeom == epullgANGLEAXIS)
1527 double cos_theta, cos_theta2;
1529 /* The angle-axis force is exactly the same as the angle force (above) except that in
1530 this case the second vector (dr23) is replaced by the pull vector. */
1531 cos_theta = cos(spatialData.value);
1532 cos_theta2 = gmx::square(cos_theta);
1538 dvec normalized_dr01;
1540 invdr01 = 1./dnorm(spatialData.dr01);
1541 dsvmul(invdr01, spatialData.dr01, normalized_dr01);
1542 a = -gmx::invsqrt(1 - cos_theta2); /* comes from d/dx acos(x) */
1545 for (int m = 0; m < DIM; m++)
1547 forces.force01[m] = pcrd.scalarForce*invdr01*(a*spatialData.vec[m] - b*normalized_dr01[m]);
1552 clear_dvec(forces.force01);
1555 else if (params.eGeom == epullgDIHEDRAL)
1557 double m2, n2, tol, sqrdist_32;
1559 /* Note: there is a small difference here compared to the
1560 dihedral force calculations in the bondeds (ref: Bekker 1994).
1561 There rij = ri - rj, while here dr01 = r1 - r0.
1562 However, all distance vectors occur in form of cross or inner products
1563 so that two signs cancel and we end up with the same expressions.
1564 Also, we treat the more general case of 6 groups (0..5) instead of 4 (i, j, k, l).
1566 m2 = diprod(spatialData.planevec_m, spatialData.planevec_m);
1567 n2 = diprod(spatialData.planevec_n, spatialData.planevec_n);
1568 dsvmul(-1, spatialData.dr23, dr32);
1569 sqrdist_32 = diprod(dr32, dr32);
1570 tol = sqrdist_32*GMX_REAL_EPS; /* Avoid tiny angles */
1571 if ((m2 > tol) && (n2 > tol))
1573 double a_01, a_23_01, a_23_45, a_45;
1574 double inv_dist_32, inv_sqrdist_32, dist_32;
1576 inv_dist_32 = gmx::invsqrt(sqrdist_32);
1577 inv_sqrdist_32 = inv_dist_32*inv_dist_32;
1578 dist_32 = sqrdist_32*inv_dist_32;
1580 /* Forces on groups 0, 1 */
1581 a_01 = pcrd.scalarForce*dist_32/m2; /* scalarForce is -dV/dphi */
1582 dsvmul(-a_01, spatialData.planevec_m, forces.force01); /* added sign to get force on group 1, not 0 */
1584 /* Forces on groups 4, 5 */
1585 a_45 = -pcrd.scalarForce*dist_32/n2;
1586 dsvmul(a_45, spatialData.planevec_n, forces.force45); /* force on group 5 */
1588 /* Force on groups 2, 3 (defining the axis) */
1589 a_23_01 = -diprod(spatialData.dr01, dr32)*inv_sqrdist_32;
1590 a_23_45 = -diprod(spatialData.dr45, dr32)*inv_sqrdist_32;
1591 dsvmul(-a_23_01, forces.force01, u); /* added sign to get force from group 0, not 1 */
1592 dsvmul(a_23_45, forces.force45, v);
1593 dvec_sub(u, v, forces.force23); /* force on group 3 */
1597 /* No force to apply for ill-defined cases */
1598 clear_dvec(forces.force01);
1599 clear_dvec(forces.force23);
1600 clear_dvec(forces.force45);
1605 for (int m = 0; m < DIM; m++)
1607 forces.force01[m] = pcrd.scalarForce*spatialData.vec[m];
1615 /* We use a global mutex for locking access to the pull data structure
1616 * during registration of external pull potential providers.
1617 * We could use a different, local mutex for each pull object, but the overhead
1618 * is extremely small here and registration is only done during initialization.
1620 static gmx::Mutex registrationMutex;
1622 using Lock = gmx::lock_guard<gmx::Mutex>;
1624 void register_external_pull_potential(struct pull_t *pull,
1626 const char *provider)
1628 GMX_RELEASE_ASSERT(pull != nullptr, "register_external_pull_potential called before init_pull");
1629 GMX_RELEASE_ASSERT(provider != nullptr, "register_external_pull_potential called with NULL as provider name");
1631 if (coord_index < 0 || coord_index >= static_cast<int>(pull->coord.size()))
1633 gmx_fatal(FARGS, "Module '%s' attempted to register an external potential for pull coordinate %d which is out of the pull coordinate range %d - %zu\n",
1634 provider, coord_index + 1, 1, pull->coord.size());
1637 pull_coord_work_t *pcrd = &pull->coord[coord_index];
1639 if (pcrd->params.eType != epullEXTERNAL)
1641 gmx_fatal(FARGS, "Module '%s' attempted to register an external potential for pull coordinate %d which of type '%s', whereas external potentials are only supported with type '%s'",
1642 provider, coord_index + 1, epull_names[pcrd->params.eType], epull_names[epullEXTERNAL]);
1645 GMX_RELEASE_ASSERT(pcrd->params.externalPotentialProvider != nullptr, "The external potential provider string for a pull coordinate is NULL");
1647 if (gmx_strcasecmp(provider, pcrd->params.externalPotentialProvider) != 0)
1649 gmx_fatal(FARGS, "Module '%s' attempted to register an external potential for pull coordinate %d which expects the external potential to be provided by a module named '%s'",
1650 provider, coord_index + 1, pcrd->params.externalPotentialProvider);
1653 /* Lock to avoid (extremely unlikely) simultaneous reading and writing of
1654 * pcrd->bExternalPotentialProviderHasBeenRegistered and
1655 * pull->numUnregisteredExternalPotentials.
1657 Lock registrationLock(registrationMutex);
1659 if (pcrd->bExternalPotentialProviderHasBeenRegistered)
1661 gmx_fatal(FARGS, "Module '%s' attempted to register an external potential for pull coordinate %d more than once",
1662 provider, coord_index + 1);
1665 pcrd->bExternalPotentialProviderHasBeenRegistered = true;
1666 pull->numUnregisteredExternalPotentials--;
1668 GMX_RELEASE_ASSERT(pull->numUnregisteredExternalPotentials >= 0, "Negative unregistered potentials, the pull code is inconsistent");
1672 static void check_external_potential_registration(const struct pull_t *pull)
1674 if (pull->numUnregisteredExternalPotentials > 0)
1677 for (c = 0; c < pull->coord.size(); c++)
1679 if (pull->coord[c].params.eType == epullEXTERNAL &&
1680 !pull->coord[c].bExternalPotentialProviderHasBeenRegistered)
1686 GMX_RELEASE_ASSERT(c < pull->coord.size(), "Internal inconsistency in the pull potential provider counting");
1688 gmx_fatal(FARGS, "No external provider for external pull potentials have been provided for %d pull coordinates. The first coordinate without provider is number %zu, which expects a module named '%s' to provide the external potential.",
1689 pull->numUnregisteredExternalPotentials,
1690 c + 1, pull->coord[c].params.externalPotentialProvider);
1695 /* Pull takes care of adding the forces of the external potential.
1696 * The external potential module has to make sure that the corresponding
1697 * potential energy is added either to the pull term or to a term
1698 * specific to the external module.
1700 void apply_external_pull_coord_force(struct pull_t *pull,
1703 const t_mdatoms *mdatoms,
1704 gmx::ForceWithVirial *forceWithVirial)
1706 pull_coord_work_t *pcrd;
1708 GMX_ASSERT(coord_index >= 0 && coord_index < static_cast<int>(pull->coord.size()), "apply_external_pull_coord_force called with coord_index out of range");
1710 if (pull->comm.bParticipate)
1712 pcrd = &pull->coord[coord_index];
1714 GMX_RELEASE_ASSERT(pcrd->params.eType == epullEXTERNAL, "The pull force can only be set externally on pull coordinates of external type");
1716 GMX_ASSERT(pcrd->bExternalPotentialProviderHasBeenRegistered, "apply_external_pull_coord_force called for an unregistered pull coordinate");
1719 pcrd->scalarForce = coord_force;
1721 /* Calculate the forces on the pull groups */
1722 PullCoordVectorForces pullCoordForces = calculateVectorForces(*pcrd);
1724 /* Add the forces for this coordinate to the total virial and force */
1725 if (forceWithVirial->computeVirial_ && pull->comm.isMasterRank)
1727 matrix virial = { { 0 } };
1728 add_virial_coord(virial, *pcrd, pullCoordForces);
1729 forceWithVirial->addVirialContribution(virial);
1732 apply_forces_coord(pull, coord_index, pullCoordForces, mdatoms,
1733 as_rvec_array(forceWithVirial->force_.data()));
1736 pull->numExternalPotentialsStillToBeAppliedThisStep--;
1739 /* Calculate the pull potential and scalar force for a pull coordinate.
1740 * Returns the vector forces for the pull coordinate.
1742 static PullCoordVectorForces
1743 do_pull_pot_coord(struct pull_t *pull, int coord_ind, t_pbc *pbc,
1744 double t, real lambda,
1745 real *V, tensor vir, real *dVdl)
1747 pull_coord_work_t &pcrd = pull->coord[coord_ind];
1749 assert(pcrd.params.eType != epullCONSTRAINT);
1751 double dev = get_pull_coord_deviation(pull, coord_ind, pbc, t);
1753 calc_pull_coord_scalar_force_and_potential(&pcrd, dev, lambda, V, dVdl);
1755 PullCoordVectorForces pullCoordForces = calculateVectorForces(pcrd);
1757 add_virial_coord(vir, pcrd, pullCoordForces);
1759 return pullCoordForces;
1762 real pull_potential(struct pull_t *pull, const t_mdatoms *md, t_pbc *pbc,
1763 const t_commrec *cr, double t, real lambda,
1764 const rvec *x, gmx::ForceWithVirial *force, real *dvdlambda)
1768 assert(pull != nullptr);
1770 /* Ideally we should check external potential registration only during
1771 * the initialization phase, but that requires another function call
1772 * that should be done exactly in the right order. So we check here.
1774 check_external_potential_registration(pull);
1776 if (pull->comm.bParticipate)
1780 pull_calc_coms(cr, pull, md, pbc, t, x, nullptr);
1782 rvec *f = as_rvec_array(force->force_.data());
1783 matrix virial = { { 0 } };
1784 const bool computeVirial = (force->computeVirial_ && MASTER(cr));
1785 for (size_t c = 0; c < pull->coord.size(); c++)
1787 /* For external potential the force is assumed to be given by an external module by a call to
1788 apply_pull_coord_external_force */
1789 if (pull->coord[c].params.eType == epullCONSTRAINT || pull->coord[c].params.eType == epullEXTERNAL)
1794 PullCoordVectorForces pullCoordForces =
1795 do_pull_pot_coord(pull, c, pbc, t, lambda,
1797 computeVirial ? virial : nullptr,
1800 /* Distribute the force over the atoms in the pulled groups */
1801 apply_forces_coord(pull, c, pullCoordForces, md, f);
1806 force->addVirialContribution(virial);
1811 GMX_ASSERT(pull->numExternalPotentialsStillToBeAppliedThisStep == 0, "Too few or too many external pull potentials have been applied the previous step");
1812 /* All external pull potentials still need to be applied */
1813 pull->numExternalPotentialsStillToBeAppliedThisStep =
1814 pull->numCoordinatesWithExternalPotential;
1816 return (MASTER(cr) ? V : 0.0);
1819 void pull_constraint(struct pull_t *pull, const t_mdatoms *md, t_pbc *pbc,
1820 const t_commrec *cr, double dt, double t,
1821 rvec *x, rvec *xp, rvec *v, tensor vir)
1823 assert(pull != nullptr);
1825 if (pull->comm.bParticipate)
1827 pull_calc_coms(cr, pull, md, pbc, t, x, xp);
1829 do_constraint(pull, pbc, xp, v, MASTER(cr), vir, dt, t);
1833 void dd_make_local_pull_groups(const t_commrec *cr, struct pull_t *pull)
1837 gmx_bool bMustParticipate;
1843 /* We always make the master node participate, such that it can do i/o,
1844 * add the virial and to simplify MC type extensions people might have.
1846 bMustParticipate = (comm->bParticipateAll || comm->isMasterRank);
1848 for (pull_group_work_t &group : pull->group)
1850 if (group.epgrppbc == epgrppbcCOS || !group.globalWeights.empty())
1852 group.localWeights.resize(group.atomSet.numAtomsLocal());
1853 for (size_t i = 0; i < group.atomSet.numAtomsLocal(); ++i)
1855 group.localWeights[i] = group.globalWeights[group.atomSet.collectiveIndex()[i]];
1859 GMX_ASSERT(bMustParticipate || dd != nullptr, "Either all ranks (including this rank) participate, or we use DD and need to have access to dd here");
1861 /* We should participate if we have pull or pbc atoms */
1862 if (!bMustParticipate &&
1863 (group.atomSet.numAtomsLocal() > 0 ||
1864 (group.epgrppbc == epgrppbcREFAT &&
1865 group.pbcAtomSet->numAtomsLocal() > 0)))
1867 bMustParticipate = TRUE;
1871 if (!comm->bParticipateAll)
1873 /* Keep currently not required ranks in the communicator
1874 * if they needed to participate up to 20 decompositions ago.
1875 * This avoids frequent rebuilds due to atoms jumping back and forth.
1877 const int64_t history_count = 20;
1878 gmx_bool bWillParticipate;
1881 /* Increase the decomposition counter for the current call */
1882 comm->setup_count++;
1884 if (bMustParticipate)
1886 comm->must_count = comm->setup_count;
1891 (comm->bParticipate &&
1892 comm->must_count >= comm->setup_count - history_count);
1894 if (debug && dd != nullptr)
1896 fprintf(debug, "Our DD rank (%3d) pull #atoms>0 or master: %s, will be part %s\n",
1897 dd->rank, gmx::boolToString(bMustParticipate), gmx::boolToString(bWillParticipate));
1900 if (bWillParticipate)
1902 /* Count the number of ranks that we want to have participating */
1904 /* Count the number of ranks that need to be added */
1905 count[1] = comm->bParticipate ? 0 : 1;
1913 /* The cost of this global operation will be less that the cost
1914 * of the extra MPI_Comm_split calls that we can avoid.
1916 gmx_sumi(2, count, cr);
1918 /* If we are missing ranks or if we have 20% more ranks than needed
1919 * we make a new sub-communicator.
1921 if (count[1] > 0 || 6*count[0] < 5*comm->nparticipate)
1925 fprintf(debug, "Creating new pull subcommunicator of size %d\n",
1930 if (comm->mpi_comm_com != MPI_COMM_NULL)
1932 MPI_Comm_free(&comm->mpi_comm_com);
1935 /* This might be an extremely expensive operation, so we try
1936 * to avoid this splitting as much as possible.
1938 assert(dd != nullptr);
1939 MPI_Comm_split(dd->mpi_comm_all, bWillParticipate ? 0 : 1, dd->rank,
1940 &comm->mpi_comm_com);
1943 comm->bParticipate = bWillParticipate;
1944 comm->nparticipate = count[0];
1948 /* Since the PBC of atoms might have changed, we need to update the PBC */
1949 pull->bSetPBCatoms = TRUE;
1952 static void init_pull_group_index(FILE *fplog, const t_commrec *cr,
1953 int g, pull_group_work_t *pg,
1954 gmx_bool bConstraint, const ivec pulldim_con,
1955 const gmx_mtop_t *mtop,
1956 const t_inputrec *ir, real lambda)
1958 /* With EM and BD there are no masses in the integrator.
1959 * But we still want to have the correct mass-weighted COMs.
1960 * So we store the real masses in the weights.
1962 const bool setWeights = (pg->params.nweight > 0 ||
1963 EI_ENERGY_MINIMIZATION(ir->eI) ||
1966 /* In parallel, store we need to extract localWeights from weights at DD time */
1967 std::vector<real> &weights = ((cr && PAR(cr)) ? pg->globalWeights : pg->localWeights);
1969 const gmx_groups_t *groups = &mtop->groups;
1971 /* Count frozen dimensions and (weighted) mass */
1977 for (int i = 0; i < pg->params.nat; i++)
1979 int ii = pg->params.ind[i];
1980 if (bConstraint && ir->opts.nFreeze)
1982 for (int d = 0; d < DIM; d++)
1984 if (pulldim_con[d] == 1 &&
1985 ir->opts.nFreeze[getGroupType(groups, egcFREEZE, ii)][d])
1991 const t_atom &atom = mtopGetAtomParameters(mtop, ii, &molb);
1993 if (ir->efep == efepNO)
1999 m = (1 - lambda)*atom.m + lambda*atom.mB;
2002 if (pg->params.nweight > 0)
2004 w = pg->params.weight[i];
2010 if (EI_ENERGY_MINIMIZATION(ir->eI))
2012 /* Move the mass to the weight */
2016 else if (ir->eI == eiBD)
2019 if (ir->bd_fric != 0.0f)
2021 mbd = ir->bd_fric*ir->delta_t;
2025 if (groups->grpnr[egcTC] == nullptr)
2027 mbd = ir->delta_t/ir->opts.tau_t[0];
2031 mbd = ir->delta_t/ir->opts.tau_t[groups->grpnr[egcTC][ii]];
2039 weights.push_back(w);
2048 /* We can have single atom groups with zero mass with potential pulling
2049 * without cosine weighting.
2051 if (pg->params.nat == 1 && !bConstraint && pg->epgrppbc != epgrppbcCOS)
2053 /* With one atom the mass doesn't matter */
2058 gmx_fatal(FARGS, "The total%s mass of pull group %d is zero",
2059 pg->params.weight ? " weighted" : "", g);
2065 "Pull group %d: %5d atoms, mass %9.3f",
2066 g, pg->params.nat, tmass);
2067 if (pg->params.weight ||
2068 EI_ENERGY_MINIMIZATION(ir->eI) || ir->eI == eiBD)
2070 fprintf(fplog, ", weighted mass %9.3f", wmass*wmass/wwmass);
2072 if (pg->epgrppbc == epgrppbcCOS)
2074 fprintf(fplog, ", cosine weighting will be used");
2076 fprintf(fplog, "\n");
2081 /* A value != 0 signals not frozen, it is updated later */
2087 for (int d = 0; d < DIM; d++)
2089 ndim += pulldim_con[d]*pg->params.nat;
2091 if (fplog && nfrozen > 0 && nfrozen < ndim)
2094 "\nWARNING: In pull group %d some, but not all of the degrees of freedom\n"
2095 " that are subject to pulling are frozen.\n"
2096 " For constraint pulling the whole group will be frozen.\n\n",
2105 init_pull(FILE *fplog, const pull_params_t *pull_params, const t_inputrec *ir,
2106 const gmx_mtop_t *mtop, const t_commrec *cr, gmx::LocalAtomSetManager *atomSets,
2109 struct pull_t *pull;
2114 /* Copy the pull parameters */
2115 pull->params = *pull_params;
2116 /* Avoid pointer copies */
2117 pull->params.group = nullptr;
2118 pull->params.coord = nullptr;
2120 for (int i = 0; i < pull_params->ngroup; ++i)
2122 pull->group.emplace_back(pull_params->group[i], atomSets->add({pull_params->group[i].ind, pull_params->group[i].ind+pull_params->group[i].nat}));
2125 if (cr != nullptr && DOMAINDECOMP(cr))
2127 /* Set up the global to local atom mapping for PBC atoms */
2128 for (pull_group_work_t &group : pull->group)
2130 if (group.epgrppbc == epgrppbcREFAT)
2132 /* pbcAtomSet consists of a single atom */
2133 group.pbcAtomSet = gmx::compat::make_unique<gmx::LocalAtomSet>(atomSets->add({&group.params.pbcatom, &group.params.pbcatom + 1}));
2138 pull->bPotential = FALSE;
2139 pull->bConstraint = FALSE;
2140 pull->bCylinder = FALSE;
2141 pull->bAngle = FALSE;
2143 GMX_RELEASE_ASSERT(pull->group[0].params.nat == 0, "pull group 0 is an absolute reference group and should not contain atoms");
2145 pull->numCoordinatesWithExternalPotential = 0;
2147 for (int c = 0; c < pull->params.ncoord; c++)
2149 /* Construct a pull coordinate, copying all coordinate parameters */
2150 pull->coord.emplace_back(pull_params->coord[c]);
2152 pull_coord_work_t *pcrd = &pull->coord.back();
2154 switch (pcrd->params.eGeom)
2157 case epullgDIRRELATIVE: /* Direction vector is determined at each step */
2159 case epullgDIHEDRAL:
2164 case epullgANGLEAXIS:
2165 copy_rvec_to_dvec(pull_params->coord[c].vec, pcrd->spatialData.vec);
2168 /* We allow reading of newer tpx files with new pull geometries,
2169 * but with the same tpx format, with old code. A new geometry
2170 * only adds a new enum value, which will be out of range for
2171 * old code. The first place we need to generate an error is
2172 * here, since the pull code can't handle this.
2173 * The enum can be used before arriving here only for printing
2174 * the string corresponding to the geometry, which will result
2175 * in printing "UNDEFINED".
2177 gmx_fatal(FARGS, "Pull geometry not supported for pull coordinate %d. The geometry enum %d in the input is larger than that supported by the code (up to %d). You are probably reading a tpr file generated with a newer version of Gromacs with an binary from an older version of Gromacs.",
2178 c + 1, pcrd->params.eGeom, epullgNR - 1);
2181 if (pcrd->params.eType == epullCONSTRAINT)
2183 /* Check restrictions of the constraint pull code */
2184 if (pcrd->params.eGeom == epullgCYL ||
2185 pcrd->params.eGeom == epullgDIRRELATIVE ||
2186 pcrd->params.eGeom == epullgANGLE ||
2187 pcrd->params.eGeom == epullgDIHEDRAL ||
2188 pcrd->params.eGeom == epullgANGLEAXIS)
2190 gmx_fatal(FARGS, "Pulling of type %s can not be combined with geometry %s. Consider using pull type %s.",
2191 epull_names[pcrd->params.eType],
2192 epullg_names[pcrd->params.eGeom],
2193 epull_names[epullUMBRELLA]);
2196 pull->bConstraint = TRUE;
2200 pull->bPotential = TRUE;
2203 if (pcrd->params.eGeom == epullgCYL)
2205 pull->bCylinder = TRUE;
2207 else if (pcrd->params.eGeom == epullgANGLE || pcrd->params.eGeom == epullgDIHEDRAL || pcrd->params.eGeom == epullgANGLEAXIS)
2209 pull->bAngle = TRUE;
2212 /* We only need to calculate the plain COM of a group
2213 * when it is not only used as a cylinder group.
2214 * Also the absolute reference group 0 needs no COM computation.
2216 for (int i = 0; i < pcrd->params.ngroup; i++)
2218 int groupIndex = pcrd->params.group[i];
2219 if (groupIndex > 0 &&
2220 !(pcrd->params.eGeom == epullgCYL && i == 0))
2222 pull->group[groupIndex].needToCalcCom = true;
2226 /* With non-zero rate the reference value is set at every step */
2227 if (pcrd->params.rate == 0)
2229 /* Initialize the constant reference value */
2230 if (pcrd->params.eType != epullEXTERNAL)
2232 low_set_pull_coord_reference_value(pcrd, c, pcrd->params.init*pull_conversion_factor_userinput2internal(&pcrd->params));
2236 /* With an external pull potential, the reference value loses
2237 * it's meaning and should not be used. Setting it to zero
2238 * makes any terms dependent on it disappear.
2239 * The only issue this causes is that with cylinder pulling
2240 * no atoms of the cylinder group within the cylinder radius
2241 * should be more than half a box length away from the COM of
2242 * the pull group along the axial direction.
2244 pcrd->value_ref = 0.0;
2248 if (pcrd->params.eType == epullEXTERNAL)
2250 GMX_RELEASE_ASSERT(pcrd->params.rate == 0, "With an external potential, a pull coordinate should have rate = 0");
2252 /* This external potential needs to be registered later */
2253 pull->numCoordinatesWithExternalPotential++;
2255 pcrd->bExternalPotentialProviderHasBeenRegistered = false;
2258 pull->numUnregisteredExternalPotentials =
2259 pull->numCoordinatesWithExternalPotential;
2260 pull->numExternalPotentialsStillToBeAppliedThisStep = 0;
2262 pull->ePBC = ir->ePBC;
2265 case epbcNONE: pull->npbcdim = 0; break;
2266 case epbcXY: pull->npbcdim = 2; break;
2267 default: pull->npbcdim = 3; break;
2272 gmx_bool bAbs, bCos;
2275 for (const pull_coord_work_t &coord : pull->coord)
2277 if (pull->group[coord.params.group[0]].params.nat == 0 ||
2278 pull->group[coord.params.group[1]].params.nat == 0)
2284 fprintf(fplog, "\n");
2285 if (pull->bPotential)
2287 fprintf(fplog, "Will apply potential COM pulling\n");
2289 if (pull->bConstraint)
2291 fprintf(fplog, "Will apply constraint COM pulling\n");
2293 // Don't include the reference group 0 in output, so we report ngroup-1
2294 int numRealGroups = pull->group.size() - 1;
2295 GMX_RELEASE_ASSERT(numRealGroups > 0, "The reference absolute position pull group should always be present");
2296 fprintf(fplog, "with %zu pull coordinate%s and %d group%s\n",
2297 pull->coord.size(), pull->coord.size() == 1 ? "" : "s",
2298 numRealGroups, numRealGroups == 1 ? "" : "s");
2301 fprintf(fplog, "with an absolute reference\n");
2304 // Don't include the reference group 0 in loop
2305 for (size_t g = 1; g < pull->group.size(); g++)
2307 if (pull->group[g].params.nat > 1 &&
2308 pull->group[g].params.pbcatom < 0)
2310 /* We are using cosine weighting */
2311 fprintf(fplog, "Cosine weighting is used for group %zu\n", g);
2317 please_cite(fplog, "Engin2010");
2321 pull->bRefAt = FALSE;
2323 for (size_t g = 0; g < pull->group.size(); g++)
2325 pull_group_work_t *pgrp;
2327 pgrp = &pull->group[g];
2328 if (pgrp->params.nat > 0)
2330 /* There is an issue when a group is used in multiple coordinates
2331 * and constraints are applied in different dimensions with atoms
2332 * frozen in some, but not all dimensions.
2333 * Since there is only one mass array per group, we can't have
2334 * frozen/non-frozen atoms for different coords at the same time.
2335 * But since this is a very exotic case, we don't check for this.
2336 * A warning is printed in init_pull_group_index.
2339 gmx_bool bConstraint;
2340 ivec pulldim, pulldim_con;
2342 /* Loop over all pull coordinates to see along which dimensions
2343 * this group is pulled and if it is involved in constraints.
2345 bConstraint = FALSE;
2346 clear_ivec(pulldim);
2347 clear_ivec(pulldim_con);
2348 for (const pull_coord_work_t &coord : pull->coord)
2350 gmx_bool bGroupUsed = FALSE;
2351 for (int gi = 0; gi < coord.params.ngroup; gi++)
2353 if (coord.params.group[gi] == static_cast<int>(g))
2361 for (int m = 0; m < DIM; m++)
2363 if (coord.params.dim[m] == 1)
2367 if (coord.params.eType == epullCONSTRAINT)
2377 /* Determine if we need to take PBC into account for calculating
2378 * the COM's of the pull groups.
2380 switch (pgrp->epgrppbc)
2383 pull->bRefAt = TRUE;
2386 if (pgrp->params.weight != nullptr)
2388 gmx_fatal(FARGS, "Pull groups can not have relative weights and cosine weighting at same time");
2390 for (int m = 0; m < DIM; m++)
2392 if (m < pull->npbcdim && pulldim[m] == 1)
2394 if (pull->cosdim >= 0 && pull->cosdim != m)
2396 gmx_fatal(FARGS, "Can only use cosine weighting with pulling in one dimension (use mdp option pull-coord?-dim)");
2406 /* Set the indices */
2407 init_pull_group_index(fplog, cr, g, pgrp,
2408 bConstraint, pulldim_con,
2413 /* Absolute reference, set the inverse mass to zero.
2414 * This is only relevant (and used) with constraint pulling.
2421 /* If we use cylinder coordinates, do some initialising for them */
2422 if (pull->bCylinder)
2424 for (const pull_coord_work_t &coord : pull->coord)
2426 if (coord.params.eGeom == epullgCYL)
2428 if (pull->group[coord.params.group[0]].params.nat == 0)
2430 gmx_fatal(FARGS, "A cylinder pull group is not supported when using absolute reference!\n");
2433 const auto &referenceGroup = pull->group[coord.params.group[0]];
2434 pull->dyna.emplace_back(referenceGroup.params, referenceGroup.atomSet);
2438 /* The gmx_omp_nthreads module might not be initialized here, so max(1,) */
2439 pull->nthreads = std::max(1, gmx_omp_nthreads_get(emntDefault));
2440 snew(pull->sum_com, pull->nthreads);
2445 /* Use a sub-communicator when we have more than 32 ranks, but not
2446 * when we have an external pull potential, since then the external
2447 * potential provider expects each rank to have the coordinate.
2449 comm->bParticipateAll = (cr == nullptr || !DOMAINDECOMP(cr) ||
2450 cr->dd->nnodes <= 32 ||
2451 pull->numCoordinatesWithExternalPotential > 0 ||
2452 getenv("GMX_PULL_PARTICIPATE_ALL") != nullptr);
2453 /* This sub-commicator is not used with comm->bParticipateAll,
2454 * so we can always initialize it to NULL.
2456 comm->mpi_comm_com = MPI_COMM_NULL;
2457 comm->nparticipate = 0;
2458 comm->isMasterRank = (cr == nullptr || MASTER(cr));
2460 /* No MPI: 1 rank: all ranks pull */
2461 comm->bParticipateAll = TRUE;
2462 comm->isMasterRank = true;
2464 comm->bParticipate = comm->bParticipateAll;
2465 comm->setup_count = 0;
2466 comm->must_count = 0;
2468 if (!comm->bParticipateAll && fplog != nullptr)
2470 fprintf(fplog, "Will use a sub-communicator for pull communication\n");
2473 comm->rbuf = nullptr;
2474 comm->dbuf = nullptr;
2475 comm->dbuf_cyl = nullptr;
2477 /* We still need to initialize the PBC reference coordinates */
2478 pull->bSetPBCatoms = TRUE;
2480 pull->out_x = nullptr;
2481 pull->out_f = nullptr;
2486 void init_pull_output_files(pull_t *pull,
2488 const t_filenm fnm[],
2489 const gmx_output_env_t *oenv,
2490 const ContinuationOptions &continuationOptions)
2492 /* Check for px and pf filename collision, if we are writing
2494 std::string px_filename, pf_filename;
2495 std::string px_appended, pf_appended;
2498 px_filename = std::string(opt2fn("-px", nfile, fnm));
2499 pf_filename = std::string(opt2fn("-pf", nfile, fnm));
2501 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
2503 if ((pull->params.nstxout != 0) &&
2504 (pull->params.nstfout != 0) &&
2505 (px_filename == pf_filename))
2507 if (!opt2bSet("-px", nfile, fnm) && !opt2bSet("-pf", nfile, fnm))
2509 /* We are writing both pull files but neither set directly. */
2512 px_appended = append_before_extension(px_filename, "_pullx");
2513 pf_appended = append_before_extension(pf_filename, "_pullf");
2515 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR;
2516 pull->out_x = open_pull_out(px_appended.c_str(), pull, oenv,
2517 TRUE, continuationOptions);
2518 pull->out_f = open_pull_out(pf_appended.c_str(), pull, oenv,
2519 FALSE, continuationOptions);
2524 /* If at least one of -px and -pf is set but the filenames are identical: */
2525 gmx_fatal(FARGS, "Identical pull_x and pull_f output filenames %s",
2526 px_filename.c_str());
2529 if (pull->params.nstxout != 0)
2531 pull->out_x = open_pull_out(opt2fn("-px", nfile, fnm), pull, oenv,
2532 TRUE, continuationOptions);
2534 if (pull->params.nstfout != 0)
2536 pull->out_f = open_pull_out(opt2fn("-pf", nfile, fnm), pull, oenv,
2537 FALSE, continuationOptions);
2541 static void destroy_pull(struct pull_t *pull)
2544 if (pull->comm.mpi_comm_com != MPI_COMM_NULL)
2546 MPI_Comm_free(&pull->comm.mpi_comm_com);
2549 sfree(pull->comm.rbuf);
2550 sfree(pull->comm.dbuf);
2551 sfree(pull->comm.dbuf_cyl);
2556 void finish_pull(struct pull_t *pull)
2558 check_external_potential_registration(pull);
2562 gmx_fio_fclose(pull->out_x);
2566 gmx_fio_fclose(pull->out_f);
2572 gmx_bool pull_have_potential(const struct pull_t *pull)
2574 return pull->bPotential;
2577 gmx_bool pull_have_constraint(const struct pull_t *pull)
2579 return pull->bConstraint;