2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2008, The GROMACS development team.
6 * Copyright (c) 2013,2014,2015,2016,2017 by the GROMACS development team.
7 * Copyright (c) 2018,2019,2020,2021, by the GROMACS development team, led by
8 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
9 * and including many others, as listed in the AUTHORS file in the
10 * top-level source directory and at http://www.gromacs.org.
12 * GROMACS is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public License
14 * as published by the Free Software Foundation; either version 2.1
15 * of the License, or (at your option) any later version.
17 * GROMACS is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with GROMACS; if not, see
24 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
25 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
27 * If you want to redistribute modifications to GROMACS, please
28 * consider that scientific software is very special. Version
29 * control is crucial - bugs must be traceable. We will be happy to
30 * consider code for inclusion in the official distribution, but
31 * derived work must not be called official GROMACS. Details are found
32 * in the README & COPYING files - if they are missing, get the
33 * official version at http://www.gromacs.org.
35 * To help us fund GROMACS development, we humbly ask that you cite
36 * the research papers on the package. Check out http://www.gromacs.org.
50 #include "gromacs/domdec/dlbtiming.h"
51 #include "gromacs/domdec/domdec.h"
52 #include "gromacs/domdec/domdec_struct.h"
53 #include "gromacs/domdec/mdsetup.h"
54 #include "gromacs/gmxlib/network.h"
55 #include "gromacs/math/functions.h"
56 #include "gromacs/math/units.h"
57 #include "gromacs/math/utilities.h"
58 #include "gromacs/math/vec.h"
59 #include "gromacs/math/vecdump.h"
60 #include "gromacs/mdlib/constr.h"
61 #include "gromacs/mdlib/enerdata_utils.h"
62 #include "gromacs/mdlib/force.h"
63 #include "gromacs/mdlib/force_flags.h"
64 #include "gromacs/mdlib/gmx_omp_nthreads.h"
65 #include "gromacs/mdlib/mdatoms.h"
66 #include "gromacs/mdlib/vsite.h"
67 #include "gromacs/mdtypes/commrec.h"
68 #include "gromacs/mdtypes/enerdata.h"
69 #include "gromacs/mdtypes/forcebuffers.h"
70 #include "gromacs/mdtypes/forcerec.h"
71 #include "gromacs/mdtypes/inputrec.h"
72 #include "gromacs/mdtypes/md_enums.h"
73 #include "gromacs/mdtypes/mdatom.h"
74 #include "gromacs/mdtypes/state.h"
75 #include "gromacs/pbcutil/pbc.h"
76 #include "gromacs/topology/ifunc.h"
77 #include "gromacs/topology/mtop_lookup.h"
78 #include "gromacs/topology/mtop_util.h"
79 #include "gromacs/utility/arrayref.h"
80 #include "gromacs/utility/arraysize.h"
81 #include "gromacs/utility/cstringutil.h"
82 #include "gromacs/utility/fatalerror.h"
83 #include "gromacs/utility/gmxassert.h"
86 using gmx::ArrayRefWithPadding;
91 int nnucl = 0; /* The number of nuclei */
92 int shellIndex = -1; /* The shell index */
93 int nucl1 = -1; /* The first nuclei connected to the shell */
94 int nucl2 = -1; /* The second nuclei connected to the shell */
95 int nucl3 = -1; /* The third nuclei connected to the shell */
96 real k = 0; /* force constant */
97 real k_1 = 0; /* 1 over force constant */
98 rvec xold; /* The old shell coordinates */
99 rvec fold; /* The old force on the shell */
100 rvec step; /* Step size for steepest descents */
105 /* Shell counts, indices, parameters and working data */
106 std::vector<t_shell> shell_gl; /* All the shells (for DD only) */
107 std::vector<int> shell_index_gl; /* Global shell index (for DD only) */
108 gmx_bool bInterCG; /* Are there inter charge-group shells? */
109 std::vector<t_shell> shells; /* The local shells */
110 bool predictShells = false; /* Predict shell positions */
111 bool requireInit = false; /* Require initialization of shell positions */
112 int nflexcon = 0; /* The number of flexible constraints */
114 std::array<PaddedHostVector<RVec>, 2> x; /* Coordinate buffers for iterative minimization */
115 std::array<PaddedHostVector<RVec>, 2> f; /* Force buffers for iterative minimization */
117 /* Flexible constraint working data */
118 std::vector<RVec> acc_dir; /* Acceleration direction for flexcon */
119 gmx::PaddedVector<RVec> x_old; /* Old coordinates for flexcon */
120 gmx::PaddedVector<RVec> adir_xnold; /* Work space for init_adir */
121 gmx::PaddedVector<RVec> adir_xnew; /* Work space for init_adir */
122 std::int64_t numForceEvaluations; /* Total number of force evaluations */
123 int numConvergedIterations; /* Total number of iterations that converged */
127 static void pr_shell(FILE* fplog, ArrayRef<const t_shell> shells)
129 fprintf(fplog, "SHELL DATA\n");
130 fprintf(fplog, "%5s %8s %5s %5s %5s\n", "Shell", "Force k", "Nucl1", "Nucl2", "Nucl3");
131 for (const t_shell& shell : shells)
133 fprintf(fplog, "%5d %8.3f %5d", shell.shellIndex, 1.0 / shell.k_1, shell.nucl1);
134 if (shell.nnucl == 2)
136 fprintf(fplog, " %5d\n", shell.nucl2);
138 else if (shell.nnucl == 3)
140 fprintf(fplog, " %5d %5d\n", shell.nucl2, shell.nucl3);
144 fprintf(fplog, "\n");
149 /* TODO The remain call of this function passes non-NULL mass and NULL
150 * mtop, so this routine can be simplified.
152 * The other code path supported doing prediction before the MD loop
153 * started, but even when called, the prediction was always
154 * over-written by a subsequent call in the MD loop, so has been
156 static void predict_shells(FILE* fplog,
160 ArrayRef<const t_shell> shells,
166 real dt_1, fudge, tm, m1, m2, m3;
168 GMX_RELEASE_ASSERT(mass || mtop, "Must have masses or a way to look them up");
170 /* We introduce a fudge factor for performance reasons: with this choice
171 * the initial force on the shells is about a factor of two lower than
181 fprintf(fplog, "RELAX: Using prediction for initial shell placement\n");
193 for (const t_shell& shell : shells)
195 const int s1 = shell.shellIndex;
204 for (m = 0; (m < DIM); m++)
206 x[s1][m] += xOrV[n1][m] * dt_1;
219 /* Not the correct masses with FE, but it is just a prediction... */
220 m1 = mtopGetAtomMass(mtop, n1, &molb);
221 m2 = mtopGetAtomMass(mtop, n2, &molb);
223 tm = dt_1 / (m1 + m2);
224 for (m = 0; (m < DIM); m++)
226 x[s1][m] += (m1 * xOrV[n1][m] + m2 * xOrV[n2][m]) * tm;
241 /* Not the correct masses with FE, but it is just a prediction... */
242 m1 = mtopGetAtomMass(mtop, n1, &molb);
243 m2 = mtopGetAtomMass(mtop, n2, &molb);
244 m3 = mtopGetAtomMass(mtop, n3, &molb);
246 tm = dt_1 / (m1 + m2 + m3);
247 for (m = 0; (m < DIM); m++)
249 x[s1][m] += (m1 * xOrV[n1][m] + m2 * xOrV[n2][m] + m3 * xOrV[n3][m]) * tm;
252 default: gmx_fatal(FARGS, "Shell %d has %d nuclei!", s1, shell.nnucl);
257 gmx_shellfc_t* init_shell_flexcon(FILE* fplog,
258 const gmx_mtop_t* mtop,
261 bool usingDomainDecomposition,
267 int i, j, type, a_offset, mol, ftype, nra;
269 int aS, aN = 0; /* Shell and nucleus */
270 int bondtypes[] = { F_BONDS, F_HARMONIC, F_CUBICBONDS, F_POLARIZATION, F_ANHARM_POL, F_WATER_POL };
271 #define NBT asize(bondtypes)
272 const gmx_ffparams_t* ffparams;
274 const gmx::EnumerationArray<ParticleType, int> numParticles = gmx_mtop_particletype_count(*mtop);
277 /* Print the number of each particle type */
278 for (const auto entry : gmx::keysOf(numParticles))
280 const int number = numParticles[entry];
283 fprintf(fplog, "There are: %d %ss\n", number, enumValueToString(entry));
288 nshell = numParticles[ParticleType::Shell];
290 if (nshell == 0 && nflexcon == 0)
292 /* We're not doing shells or flexible constraints */
296 shfc = new gmx_shellfc_t;
297 shfc->nflexcon = nflexcon;
301 /* Only flexible constraints, no shells.
302 * Note that make_local_shells() does not need to be called.
307 if (nstcalcenergy != 1)
310 "You have nstcalcenergy set to a value (%d) that is different from 1.\nThis is "
311 "not supported in combination with shell particles.\nPlease make a new tpr file.",
314 if (usingDomainDecomposition)
318 "Shell particles are not implemented with domain decomposition, use a single rank");
321 /* We have shells: fill the shell data structure */
323 /* Global system sized array, this should be avoided */
324 std::vector<int> shell_index(mtop->natoms);
327 for (const AtomProxy atomP : AtomRange(*mtop))
329 const t_atom& local = atomP.atom();
330 int i = atomP.globalAtomNumber();
331 if (local.ptype == ParticleType::Shell)
333 shell_index[i] = nshell++;
337 std::vector<t_shell> shell(nshell);
339 ffparams = &mtop->ffparams;
341 /* Now fill the structures */
342 /* TODO: See if we can use update groups that cover shell constructions */
343 shfc->bInterCG = FALSE;
346 for (size_t mb = 0; mb < mtop->molblock.size(); mb++)
348 const gmx_molblock_t* molb = &mtop->molblock[mb];
349 const gmx_moltype_t* molt = &mtop->moltype[molb->type];
351 const t_atom* atom = molt->atoms.atom;
352 for (mol = 0; mol < molb->nmol; mol++)
354 for (j = 0; (j < NBT); j++)
356 const int* ia = molt->ilist[bondtypes[j]].iatoms.data();
357 for (i = 0; (i < molt->ilist[bondtypes[j]].size());)
360 ftype = ffparams->functype[type];
361 nra = interaction_function[ftype].nratoms;
363 /* Check whether we have a bond with a shell */
366 switch (bondtypes[j])
373 if (atom[ia[1]].ptype == ParticleType::Shell)
378 else if (atom[ia[2]].ptype == ParticleType::Shell)
385 aN = ia[4]; /* Dummy */
386 aS = ia[5]; /* Shell */
388 default: gmx_fatal(FARGS, "Death Horror: %s, %d", __FILE__, __LINE__);
395 /* Check whether one of the particles is a shell... */
396 nsi = shell_index[a_offset + aS];
397 if ((nsi < 0) || (nsi >= nshell))
399 gmx_fatal(FARGS, "nsi is %d should be within 0 - %d. aS = %d", nsi, nshell, aS);
401 if (shell[nsi].shellIndex == -1)
403 shell[nsi].shellIndex = a_offset + aS;
406 else if (shell[nsi].shellIndex != a_offset + aS)
408 gmx_fatal(FARGS, "Weird stuff in %s, %d", __FILE__, __LINE__);
411 if (shell[nsi].nucl1 == -1)
413 shell[nsi].nucl1 = a_offset + aN;
415 else if (shell[nsi].nucl2 == -1)
417 shell[nsi].nucl2 = a_offset + aN;
419 else if (shell[nsi].nucl3 == -1)
421 shell[nsi].nucl3 = a_offset + aN;
427 pr_shell(fplog, shell);
429 gmx_fatal(FARGS, "Can not handle more than three bonds per shell\n");
433 /* shell[nsi].bInterCG = TRUE; */
434 shfc->bInterCG = TRUE;
437 switch (bondtypes[j])
441 shell[nsi].k += ffparams->iparams[type].harmonic.krA;
444 shell[nsi].k += ffparams->iparams[type].cubic.kb;
448 if (!gmx_within_tol(qS, atom[aS].qB, GMX_REAL_EPS * 10))
451 "polarize can not be used with qA(%e) != qB(%e) for "
452 "atom %d of molecule block %zu",
458 shell[nsi].k += gmx::square(qS) * gmx::c_one4PiEps0
459 / ffparams->iparams[type].polarize.alpha;
462 if (!gmx_within_tol(qS, atom[aS].qB, GMX_REAL_EPS * 10))
465 "water_pol can not be used with qA(%e) != qB(%e) for "
466 "atom %d of molecule block %zu",
472 alpha = (ffparams->iparams[type].wpol.al_x
473 + ffparams->iparams[type].wpol.al_y
474 + ffparams->iparams[type].wpol.al_z)
476 shell[nsi].k += gmx::square(qS) * gmx::c_one4PiEps0 / alpha;
478 default: gmx_fatal(FARGS, "Death Horror: %s, %d", __FILE__, __LINE__);
486 a_offset += molt->atoms.nr;
488 /* Done with this molecule type */
491 /* Verify whether it's all correct */
494 gmx_fatal(FARGS, "Something weird with shells. They may not be bonded to something");
497 for (i = 0; (i < ns); i++)
499 shell[i].k_1 = 1.0 / shell[i].k;
504 pr_shell(debug, shell);
508 shfc->shell_gl = shell;
509 shfc->shell_index_gl = shell_index;
511 shfc->predictShells = (getenv("GMX_NOPREDICT") == nullptr);
512 shfc->requireInit = false;
513 if (!shfc->predictShells)
517 fprintf(fplog, "\nWill never predict shell positions\n");
522 shfc->requireInit = (getenv("GMX_REQUIRE_SHELL_INIT") != nullptr);
523 if (shfc->requireInit && fplog)
525 fprintf(fplog, "\nWill always initiate shell positions\n");
529 if (shfc->predictShells)
536 "\nNOTE: in the current version shell prediction during the crun is "
539 /* Prediction improves performance, so we should implement either:
540 * 1. communication for the atoms needed for prediction
541 * 2. prediction using the velocities of shells; currently the
542 * shell velocities are zeroed, it's a bit tricky to keep
543 * track of the shell displacements and thus the velocity.
545 shfc->predictShells = false;
549 /* shfc->x is used as a coordinate buffer for the sim_util's `do_force` function, and
550 * when using PME it must be pinned. */
553 for (i = 0; i < 2; i++)
555 changePinningPolicy(&shfc->x[i], gmx::PinningPolicy::PinnedIfSupported);
562 void gmx::make_local_shells(const t_commrec* cr, const t_mdatoms* md, gmx_shellfc_t* shfc)
565 gmx_domdec_t* dd = nullptr;
567 if (DOMAINDECOMP(cr))
571 a1 = dd_numHomeAtoms(*dd);
575 /* Single node: we need all shells, copy them */
576 shfc->shells = shfc->shell_gl;
581 ArrayRef<const int> ind = shfc->shell_index_gl;
583 std::vector<t_shell>& shells = shfc->shells;
585 for (int i = a0; i < a1; i++)
587 if (md->ptype[i] == ParticleType::Shell)
591 shells.push_back(shfc->shell_gl[ind[dd->globalAtomIndices[i]]]);
595 shells.push_back(shfc->shell_gl[ind[i]]);
597 t_shell& shell = shells.back();
599 /* With inter-cg shells we can no do shell prediction,
600 * so we do not need the nuclei numbers.
604 shell.nucl1 = i + shell.nucl1 - shell.shellIndex;
607 shell.nucl2 = i + shell.nucl2 - shell.shellIndex;
611 shell.nucl3 = i + shell.nucl3 - shell.shellIndex;
614 shell.shellIndex = i;
619 static void do_1pos(rvec xnew, const rvec xold, const rvec f, real step)
637 static void do_1pos3(rvec xnew, const rvec xold, const rvec f, const rvec step)
646 dx = f[XX] * step[XX];
647 dy = f[YY] * step[YY];
648 dz = f[ZZ] * step[ZZ];
655 static void directional_sd(ArrayRef<const RVec> xold,
657 ArrayRef<const RVec> acc_dir,
661 const rvec* xo = as_rvec_array(xold.data());
662 rvec* xn = as_rvec_array(xnew.data());
664 for (int i = 0; i < homenr; i++)
666 do_1pos(xn[i], xo[i], acc_dir[i], step);
670 static void shell_pos_sd(ArrayRef<const RVec> xcur,
672 ArrayRef<const RVec> f,
673 ArrayRef<t_shell> shells,
676 const real step_scale_min = 0.8, step_scale_increment = 0.2, step_scale_max = 1.2,
677 step_scale_multiple = (step_scale_max - step_scale_min) / step_scale_increment;
682 real step_min, step_max;
687 for (t_shell& shell : shells)
689 const int ind = shell.shellIndex;
692 for (d = 0; d < DIM; d++)
694 shell.step[d] = shell.k_1;
696 step_min = std::min(step_min, shell.step[d]);
697 step_max = std::max(step_max, shell.step[d]);
703 for (d = 0; d < DIM; d++)
705 dx = xcur[ind][d] - shell.xold[d];
706 df = f[ind][d] - shell.fold[d];
707 /* -dx/df gets used to generate an interpolated value, but would
708 * cause a NaN if df were binary-equal to zero. Values close to
709 * zero won't cause problems (because of the min() and max()), so
710 * just testing for binary inequality is OK. */
714 /* Scale the step size by a factor interpolated from
715 * step_scale_min to step_scale_max, as k_est goes from 0 to
716 * step_scale_multiple * shell.step[d] */
717 shell.step[d] = step_scale_min * shell.step[d]
718 + step_scale_increment
719 * std::min(step_scale_multiple * shell.step[d],
720 std::max(k_est, zero));
725 if (gmx_numzero(dx)) /* 0 == dx */
727 /* Likely this will never happen, but if it does just
728 * don't scale the step. */
732 shell.step[d] *= step_scale_max;
736 step_min = std::min(step_min, shell.step[d]);
737 step_max = std::max(step_max, shell.step[d]);
741 copy_rvec(xcur[ind], shell.xold);
742 copy_rvec(f[ind], shell.fold);
744 do_1pos3(xnew[ind], xcur[ind], f[ind], shell.step);
748 fprintf(debug, "shell = %d\n", ind);
749 pr_rvec(debug, 0, "fshell", f[ind], DIM, TRUE);
750 pr_rvec(debug, 0, "xold", xcur[ind], DIM, TRUE);
751 pr_rvec(debug, 0, "step", shell.step, DIM, TRUE);
752 pr_rvec(debug, 0, "xnew", xnew[ind], DIM, TRUE);
756 printf("step %.3e %.3e\n", step_min, step_max);
760 static void decrease_step_size(ArrayRef<t_shell> shells)
762 for (t_shell& shell : shells)
764 svmul(0.8, shell.step, shell.step);
768 static void print_epot(FILE* fp, int64_t mdstep, int count, real epot, real df, int ndir, real sf_dir)
772 fprintf(fp, "MDStep=%5s/%2d EPot: %12.8e, rmsF: %6.2e", gmx_step_str(mdstep, buf), count, epot, df);
775 fprintf(fp, ", dir. rmsF: %6.2e\n", std::sqrt(sf_dir / ndir));
784 static real rms_force(const t_commrec* cr,
785 ArrayRef<const RVec> force,
786 ArrayRef<const t_shell> shells,
792 const rvec* f = as_rvec_array(force.data());
795 for (const t_shell& shell : shells)
797 buf[0] += norm2(f[shell.shellIndex]);
799 int ntot = shells.ssize();
806 gmx_sumd(4, buf, cr);
807 ntot = gmx::roundToInt(buf[1]);
813 return (ntot ? std::sqrt(buf[0] / ntot) : 0);
816 static void dump_shells(FILE* fp, ArrayRef<RVec> f, real ftol, ArrayRef<const t_shell> shells)
820 ft2 = gmx::square(ftol);
822 for (const t_shell& shell : shells)
824 const int ind = shell.shellIndex;
825 ff2 = iprod(f[ind], f[ind]);
829 "SHELL %5d, force %10.5f %10.5f %10.5f, |f| %10.5f\n",
839 static void init_adir(gmx_shellfc_t* shfc,
840 gmx::Constraints* constr,
841 const t_inputrec* ir,
847 ArrayRefWithPadding<RVec> xOld,
848 ArrayRef<RVec> x_init,
849 ArrayRefWithPadding<RVec> xCurrent,
851 ArrayRef<RVec> acc_dir,
853 ArrayRef<const real> lambda,
859 if (DOMAINDECOMP(cr))
867 shfc->adir_xnold.resizeWithPadding(n);
868 shfc->adir_xnew.resizeWithPadding(n);
869 rvec* xnold = as_rvec_array(shfc->adir_xnold.data());
870 rvec* xnew = as_rvec_array(shfc->adir_xnew.data());
871 rvec* x_old = as_rvec_array(xOld.paddedArrayRef().data());
872 rvec* x = as_rvec_array(xCurrent.paddedArrayRef().data());
874 const ParticleType* ptype = md->ptype;
878 /* Does NOT work with freeze groups (yet) */
879 for (n = 0; n < end; n++)
881 w_dt = md->invmass[n] * dt;
883 for (d = 0; d < DIM; d++)
885 if ((ptype[n] != ParticleType::VSite) && (ptype[n] != ParticleType::Shell))
887 xnold[n][d] = x[n][d] - (x_init[n][d] - x_old[n][d]);
888 xnew[n][d] = 2 * x[n][d] - x_old[n][d] + f[n][d] * w_dt * dt;
892 xnold[n][d] = x[n][d];
893 xnew[n][d] = x[n][d];
897 bool needsLogging = false;
898 bool computeEnergy = false;
899 bool computeVirial = false;
900 constr->apply(needsLogging,
906 shfc->adir_xnold.arrayRefWithPadding(),
909 lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Bonded)],
910 &(dvdlambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Bonded)]),
914 gmx::ConstraintVariable::Positions);
915 constr->apply(needsLogging,
921 shfc->adir_xnew.arrayRefWithPadding(),
924 lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Bonded)],
925 &(dvdlambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Bonded)]),
929 gmx::ConstraintVariable::Positions);
931 for (n = 0; n < end; n++)
933 for (d = 0; d < DIM; d++)
935 xnew[n][d] = -(2 * x[n][d] - xnold[n][d] - xnew[n][d]) / gmx::square(dt)
936 - f[n][d] * md->invmass[n];
938 clear_rvec(acc_dir[n]);
941 /* Project the acceleration on the old bond directions */
942 constr->apply(needsLogging,
948 shfc->adir_xnew.arrayRefWithPadding(),
951 lambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Bonded)],
952 &(dvdlambda[static_cast<int>(FreeEnergyPerturbationCouplingType::Bonded)]),
956 gmx::ConstraintVariable::Deriv_FlexCon);
959 void relax_shell_flexcon(FILE* fplog,
961 const gmx_multisim_t* ms,
963 gmx_enfrot* enforcedRotation,
965 const t_inputrec* inputrec,
966 gmx::ImdSession* imdSession,
970 const gmx_localtop_t* top,
971 gmx::Constraints* constr,
972 gmx_enerdata_t* enerd,
974 ArrayRefWithPadding<RVec> xPadded,
975 ArrayRefWithPadding<RVec> vPadded,
977 ArrayRef<real> lambda,
978 const history_t* hist,
979 gmx::ForceBuffersView* f,
983 gmx_wallcycle_t wcycle,
986 gmx::MdrunScheduleWorkload* runScheduleWork,
989 gmx::VirtualSitesHandler* vsite,
990 const DDBalanceRegionHandler& ddBalanceRegionHandler)
996 int nat, dd_ac0, dd_ac1 = 0, i;
997 int homenr = md->homenr, end = homenr;
998 int d, Min = 0, count = 0;
999 #define Try (1 - Min) /* At start Try = 1 */
1001 const bool bCont = (mdstep == inputrec->init_step) && inputrec->bContinuation;
1002 const bool bInit = (mdstep == inputrec->init_step) || shfc->requireInit;
1003 const real ftol = inputrec->em_tol;
1004 const int number_steps = inputrec->niter;
1005 ArrayRef<t_shell> shells = shfc->shells;
1006 const int nflexcon = shfc->nflexcon;
1008 if (DOMAINDECOMP(cr))
1010 nat = dd_natoms_vsite(*cr->dd);
1013 dd_get_constraint_range(*cr->dd, &dd_ac0, &dd_ac1);
1014 nat = std::max(nat, dd_ac1);
1022 for (i = 0; (i < 2); i++)
1024 shfc->x[i].resizeWithPadding(nat);
1025 shfc->f[i].resizeWithPadding(nat);
1028 /* Create views that we can swap for trail and minimum for positions and forces */
1029 ArrayRefWithPadding<RVec> posWithPadding[2];
1030 ArrayRefWithPadding<RVec> forceWithPadding[2];
1031 ArrayRef<RVec> pos[2];
1032 ArrayRef<RVec> force[2];
1033 for (i = 0; (i < 2); i++)
1035 posWithPadding[i] = shfc->x[i].arrayRefWithPadding();
1036 pos[i] = posWithPadding[i].paddedArrayRef();
1037 forceWithPadding[i] = shfc->f[i].arrayRefWithPadding();
1038 force[i] = forceWithPadding[i].paddedArrayRef();
1041 ArrayRef<RVec> x = xPadded.unpaddedArrayRef();
1042 ArrayRef<RVec> v = vPadded.unpaddedArrayRef();
1044 if (bDoNS && inputrec->pbcType != PbcType::No && !DOMAINDECOMP(cr))
1046 /* This is the only time where the coordinates are used
1047 * before do_force is called, which normally puts all
1048 * charge groups in the box.
1050 put_atoms_in_box_omp(
1051 fr->pbcType, box, x.subArray(0, md->homenr), gmx_omp_nthreads_get(emntDefault));
1056 shfc->acc_dir.resize(nat);
1057 shfc->x_old.resizeWithPadding(nat);
1058 ArrayRef<RVec> x_old = shfc->x_old.arrayRefWithPadding().unpaddedArrayRef();
1059 for (i = 0; i < homenr; i++)
1061 for (d = 0; d < DIM; d++)
1063 x_old[i][d] = x[i][d] - v[i][d] * inputrec->delta_t;
1068 /* Do a prediction of the shell positions, when appropriate.
1069 * Without velocities (EM, NM, BD) we only do initial prediction.
1071 if (shfc->predictShells && !bCont && (EI_STATE_VELOCITY(inputrec->eI) || bInit))
1073 predict_shells(fplog, x, v, inputrec->delta_t, shells, md->massT, nullptr, bInit);
1076 /* Calculate the forces first time around */
1079 pr_rvecs(debug, 0, "x b4 do_force", as_rvec_array(x.data()), homenr);
1081 int shellfc_flags = force_flags | (bVerbose ? GMX_FORCE_ENERGY : 0);
1082 gmx::ForceBuffersView forceViewInit = gmx::ForceBuffersView(forceWithPadding[Min], {}, false);
1109 (bDoNS ? GMX_FORCE_NS : 0) | shellfc_flags,
1110 ddBalanceRegionHandler);
1123 shfc->x_old.arrayRefWithPadding(),
1132 for (i = 0; i < end; i++)
1134 sf_dir += md->massT[i] * norm2(shfc->acc_dir[i]);
1137 accumulatePotentialEnergies(enerd, lambda, inputrec->fepvals.get());
1138 Epot[Min] = enerd->term[F_EPOT];
1140 df[Min] = rms_force(cr, forceWithPadding[Min].paddedArrayRef(), shells, nflexcon, &sf_dir, &Epot[Min]);
1144 fprintf(debug, "df = %g %g\n", df[Min], df[Try]);
1149 pr_rvecs(debug, 0, "force0", as_rvec_array(force[Min].data()), md->nr);
1152 if (!shells.empty() || nflexcon > 0)
1154 /* Copy x to pos[Min] & pos[Try]: during minimization only the
1155 * shell positions are updated, therefore the other particles must
1156 * be set here, in advance.
1158 std::copy(xPadded.paddedArrayRef().begin(),
1159 xPadded.paddedArrayRef().end(),
1160 posWithPadding[Min].paddedArrayRef().begin());
1161 std::copy(xPadded.paddedArrayRef().begin(),
1162 xPadded.paddedArrayRef().end(),
1163 posWithPadding[Try].paddedArrayRef().begin());
1166 if (bVerbose && MASTER(cr))
1168 print_epot(stdout, mdstep, 0, Epot[Min], df[Min], nflexcon, sf_dir);
1173 fprintf(debug, "%17s: %14.10e\n", interaction_function[F_EKIN].longname, enerd->term[F_EKIN]);
1174 fprintf(debug, "%17s: %14.10e\n", interaction_function[F_EPOT].longname, enerd->term[F_EPOT]);
1175 fprintf(debug, "%17s: %14.10e\n", interaction_function[F_ETOT].longname, enerd->term[F_ETOT]);
1176 fprintf(debug, "SHELLSTEP %s\n", gmx_step_str(mdstep, sbuf));
1179 /* First check whether we should do shells, or whether the force is
1180 * low enough even without minimization.
1182 bool bConverged = (df[Min] < ftol);
1184 for (count = 1; (!(bConverged) && (count < number_steps)); count++)
1188 vsite->construct(pos[Min], v, box, gmx::VSiteOperation::PositionsAndVelocities);
1201 shfc->x_old.arrayRefWithPadding(),
1203 posWithPadding[Min],
1210 directional_sd(pos[Min], pos[Try], shfc->acc_dir, end, fr->fc_stepsize);
1213 /* New positions, Steepest descent */
1214 shell_pos_sd(pos[Min], pos[Try], force[Min], shells, count);
1218 pr_rvecs(debug, 0, "RELAX: pos[Min] ", as_rvec_array(pos[Min].data()), homenr);
1219 pr_rvecs(debug, 0, "RELAX: pos[Try] ", as_rvec_array(pos[Try].data()), homenr);
1221 /* Try the new positions */
1222 gmx::ForceBuffersView forceViewTry = gmx::ForceBuffersView(forceWithPadding[Try], {}, false);
1236 posWithPadding[Try],
1250 ddBalanceRegionHandler);
1251 accumulatePotentialEnergies(enerd, lambda, inputrec->fepvals.get());
1254 pr_rvecs(debug, 0, "RELAX: force[Min]", as_rvec_array(force[Min].data()), homenr);
1255 pr_rvecs(debug, 0, "RELAX: force[Try]", as_rvec_array(force[Try].data()), homenr);
1268 shfc->x_old.arrayRefWithPadding(),
1270 posWithPadding[Try],
1277 ArrayRef<const RVec> acc_dir = shfc->acc_dir;
1278 for (i = 0; i < end; i++)
1280 sf_dir += md->massT[i] * norm2(acc_dir[i]);
1284 Epot[Try] = enerd->term[F_EPOT];
1286 df[Try] = rms_force(cr, force[Try], shells, nflexcon, &sf_dir, &Epot[Try]);
1290 fprintf(debug, "df = %g %g\n", df[Min], df[Try]);
1297 pr_rvecs(debug, 0, "F na do_force", as_rvec_array(force[Try].data()), homenr);
1301 fprintf(debug, "SHELL ITER %d\n", count);
1302 dump_shells(debug, force[Try], ftol, shells);
1306 if (bVerbose && MASTER(cr))
1308 print_epot(stdout, mdstep, count, Epot[Try], df[Try], nflexcon, sf_dir);
1311 bConverged = (df[Try] < ftol);
1313 if ((df[Try] < df[Min]))
1317 fprintf(debug, "Swapping Min and Try\n");
1321 /* Correct the velocities for the flexible constraints */
1322 invdt = 1 / inputrec->delta_t;
1323 for (i = 0; i < end; i++)
1325 for (d = 0; d < DIM; d++)
1327 v[i][d] += (pos[Try][i][d] - pos[Min][i][d]) * invdt;
1335 decrease_step_size(shells);
1338 shfc->numForceEvaluations += count;
1341 shfc->numConvergedIterations++;
1343 if (MASTER(cr) && !(bConverged))
1345 /* Note that the energies and virial are incorrect when not converged */
1349 "step %s: EM did not converge in %d iterations, RMS force %6.2e\n",
1350 gmx_step_str(mdstep, sbuf),
1355 "step %s: EM did not converge in %d iterations, RMS force %6.2e\n",
1356 gmx_step_str(mdstep, sbuf),
1361 /* Copy back the coordinates and the forces */
1362 std::copy(pos[Min].begin(), pos[Min].end(), x.data());
1363 std::copy(force[Min].begin(), force[Min].end(), f->force().begin());
1366 void done_shellfc(FILE* fplog, gmx_shellfc_t* shfc, int64_t numSteps)
1368 if (shfc && fplog && numSteps > 0)
1370 double numStepsAsDouble = static_cast<double>(numSteps);
1372 "Fraction of iterations that converged: %.2f %%\n",
1373 (shfc->numConvergedIterations * 100.0) / numStepsAsDouble);
1375 "Average number of force evaluations per MD step: %.2f\n\n",
1376 shfc->numForceEvaluations / numStepsAsDouble);