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, 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.
40 #include "gromacs/legacyheaders/genborn.h"
45 #include "gromacs/domdec/domdec.h"
46 #include "gromacs/fileio/pdbio.h"
47 #include "gromacs/legacyheaders/names.h"
48 #include "gromacs/legacyheaders/network.h"
49 #include "gromacs/legacyheaders/nrnb.h"
50 #include "gromacs/legacyheaders/typedefs.h"
51 #include "gromacs/legacyheaders/types/commrec.h"
52 #include "gromacs/math/units.h"
53 #include "gromacs/math/vec.h"
54 #include "gromacs/mdlib/genborn_allvsall.h"
55 #include "gromacs/pbcutil/ishift.h"
56 #include "gromacs/pbcutil/mshift.h"
57 #include "gromacs/pbcutil/pbc.h"
58 #include "gromacs/topology/mtop_util.h"
59 #include "gromacs/utility/fatalerror.h"
60 #include "gromacs/utility/gmxmpi.h"
61 #include "gromacs/utility/smalloc.h"
71 typedef struct gbtmpnbls {
77 /* This function is exactly the same as the one in listed-forces/bonded.cpp. The reason
78 * it is copied here is that the bonded gb-interactions are evaluated
79 * not in calc_bonds, but rather in calc_gb_forces
81 static int pbc_rvec_sub(const t_pbc *pbc, const rvec xi, const rvec xj, rvec dx)
85 return pbc_dx_aiuc(pbc, xi, xj, dx);
94 static int init_gb_nblist(int natoms, t_nblist *nl)
96 nl->maxnri = natoms*4;
105 /*nl->nltype = nltype;*/
107 srenew(nl->iinr, nl->maxnri);
108 srenew(nl->gid, nl->maxnri);
109 srenew(nl->shift, nl->maxnri);
110 srenew(nl->jindex, nl->maxnri+1);
118 static int init_gb_still(const t_atomtypes *atype, t_idef *idef, t_atoms *atoms,
119 gmx_genborn_t *born, int natoms)
122 int i, j, i1, i2, k, m, nbond, nang, ia, ib, ic, id, nb, idx, idx2, at;
126 real r, ri, rj, ri2, ri3, rj2, r2, r3, r4, rk, ratio, term, h, doffset;
127 real p1, p2, p3, factor, cosine, rab, rbc;
134 snew(born->gpol_still_work, natoms+3);
139 doffset = born->gb_doffset;
141 for (i = 0; i < natoms; i++)
143 born->gpol_globalindex[i] = born->vsolv_globalindex[i] =
144 born->gb_radius_globalindex[i] = 0;
147 /* Compute atomic solvation volumes for Still method */
148 for (i = 0; i < natoms; i++)
150 ri = atype->gb_radius[atoms->atom[i].type];
151 born->gb_radius_globalindex[i] = ri;
153 born->vsolv_globalindex[i] = (4*M_PI/3)*r3;
156 for (j = 0; j < idef->il[F_GB12].nr; j += 3)
158 m = idef->il[F_GB12].iatoms[j];
159 ia = idef->il[F_GB12].iatoms[j+1];
160 ib = idef->il[F_GB12].iatoms[j+2];
162 r = 1.01*idef->iparams[m].gb.st;
164 ri = atype->gb_radius[atoms->atom[ia].type];
165 rj = atype->gb_radius[atoms->atom[ib].type];
171 ratio = (rj2-ri2-r*r)/(2*ri*r);
173 term = (M_PI/3.0)*h*h*(3.0*ri-h);
175 born->vsolv_globalindex[ia] -= term;
177 ratio = (ri2-rj2-r*r)/(2*rj*r);
179 term = (M_PI/3.0)*h*h*(3.0*rj-h);
181 born->vsolv_globalindex[ib] -= term;
184 /* Get the self-, 1-2 and 1-3 polarization energies for analytical Still
187 for (j = 0; j < natoms; j++)
189 if (born->use_globalindex[j] == 1)
191 born->gpol_globalindex[j] = -0.5*ONE_4PI_EPS0/
192 (atype->gb_radius[atoms->atom[j].type]-doffset+STILL_P1);
197 for (j = 0; j < idef->il[F_GB12].nr; j += 3)
199 m = idef->il[F_GB12].iatoms[j];
200 ia = idef->il[F_GB12].iatoms[j+1];
201 ib = idef->il[F_GB12].iatoms[j+2];
203 r = idef->iparams[m].gb.st;
207 born->gpol_globalindex[ia] = born->gpol_globalindex[ia]+
208 STILL_P2*born->vsolv_globalindex[ib]/r4;
209 born->gpol_globalindex[ib] = born->gpol_globalindex[ib]+
210 STILL_P2*born->vsolv_globalindex[ia]/r4;
214 for (j = 0; j < idef->il[F_GB13].nr; j += 3)
216 m = idef->il[F_GB13].iatoms[j];
217 ia = idef->il[F_GB13].iatoms[j+1];
218 ib = idef->il[F_GB13].iatoms[j+2];
220 r = idef->iparams[m].gb.st;
223 born->gpol_globalindex[ia] = born->gpol_globalindex[ia]+
224 STILL_P3*born->vsolv_globalindex[ib]/r4;
225 born->gpol_globalindex[ib] = born->gpol_globalindex[ib]+
226 STILL_P3*born->vsolv_globalindex[ia]/r4;
235 /* Initialize all GB datastructs and compute polarization energies */
236 int init_gb(gmx_genborn_t **p_born,
237 t_forcerec *fr, const t_inputrec *ir,
238 const gmx_mtop_t *mtop, int gb_algorithm)
240 int i, j, m, ai, aj, jj, natoms, nalloc;
241 real rai, sk, p, doffset;
245 gmx_localtop_t *localtop;
247 natoms = mtop->natoms;
249 atoms = gmx_mtop_global_atoms(mtop);
250 localtop = gmx_mtop_generate_local_top(mtop, ir);
257 snew(born->drobc, natoms);
258 snew(born->bRad, natoms);
260 /* Allocate memory for the global data arrays */
261 snew(born->param_globalindex, natoms+3);
262 snew(born->gpol_globalindex, natoms+3);
263 snew(born->vsolv_globalindex, natoms+3);
264 snew(born->gb_radius_globalindex, natoms+3);
265 snew(born->use_globalindex, natoms+3);
267 snew(fr->invsqrta, natoms);
268 snew(fr->dvda, natoms);
271 fr->dadx_rawptr = NULL;
273 born->gpol_still_work = NULL;
274 born->gpol_hct_work = NULL;
276 /* snew(born->asurf,natoms); */
277 /* snew(born->dasurf,natoms); */
279 /* Initialize the gb neighbourlist */
280 init_gb_nblist(natoms, &(fr->gblist));
282 /* Do the Vsites exclusions (if any) */
283 for (i = 0; i < natoms; i++)
285 jj = atoms.atom[i].type;
286 if (mtop->atomtypes.gb_radius[atoms.atom[i].type] > 0)
288 born->use_globalindex[i] = 1;
292 born->use_globalindex[i] = 0;
295 /* If we have a Vsite, put vs_globalindex[i]=0 */
296 if (C6 (fr->nbfp, fr->ntype, jj, jj) == 0 &&
297 C12(fr->nbfp, fr->ntype, jj, jj) == 0 &&
298 atoms.atom[i].q == 0)
300 born->use_globalindex[i] = 0;
304 /* Copy algorithm parameters from inputrecord to local structure */
305 born->obc_alpha = ir->gb_obc_alpha;
306 born->obc_beta = ir->gb_obc_beta;
307 born->obc_gamma = ir->gb_obc_gamma;
308 born->gb_doffset = ir->gb_dielectric_offset;
309 born->gb_epsilon_solvent = ir->gb_epsilon_solvent;
310 born->epsilon_r = ir->epsilon_r;
312 doffset = born->gb_doffset;
314 /* Set the surface tension */
315 born->sa_surface_tension = ir->sa_surface_tension;
317 /* If Still model, initialise the polarisation energies */
318 if (gb_algorithm == egbSTILL)
320 init_gb_still(&(mtop->atomtypes), &(localtop->idef), &atoms,
325 /* If HCT/OBC, precalculate the sk*atype->S_hct factors */
326 else if (gb_algorithm == egbHCT || gb_algorithm == egbOBC)
329 snew(born->gpol_hct_work, natoms+3);
331 for (i = 0; i < natoms; i++)
333 if (born->use_globalindex[i] == 1)
335 rai = mtop->atomtypes.gb_radius[atoms.atom[i].type]-doffset;
336 sk = rai * mtop->atomtypes.S_hct[atoms.atom[i].type];
337 born->param_globalindex[i] = sk;
338 born->gb_radius_globalindex[i] = rai;
342 born->param_globalindex[i] = 0;
343 born->gb_radius_globalindex[i] = 0;
348 /* Allocate memory for work arrays for temporary use */
349 snew(born->work, natoms+4);
350 snew(born->count, natoms);
351 snew(born->nblist_work, natoms);
353 /* Domain decomposition specific stuff */
362 calc_gb_rad_still(t_commrec *cr, t_forcerec *fr, gmx_localtop_t *top,
363 rvec x[], t_nblist *nl,
364 gmx_genborn_t *born, t_mdatoms *md)
366 int i, k, n, nj0, nj1, ai, aj, type;
369 real gpi, dr, dr2, dr4, idr4, rvdw, ratio, ccf, theta, term, rai, raj;
370 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11;
371 real rinv, idr2, idr6, vaj, dccf, cosq, sinq, prod, gpi2;
373 real vai, prod_ai, icf4, icf6;
375 factor = 0.5*ONE_4PI_EPS0;
378 for (i = 0; i < born->nr; i++)
380 born->gpol_still_work[i] = 0;
383 for (i = 0; i < nl->nri; i++)
388 nj1 = nl->jindex[i+1];
390 /* Load shifts for this list */
391 shift = nl->shift[i];
392 shX = fr->shift_vec[shift][0];
393 shY = fr->shift_vec[shift][1];
394 shZ = fr->shift_vec[shift][2];
398 rai = top->atomtypes.gb_radius[md->typeA[ai]];
399 vai = born->vsolv[ai];
400 prod_ai = STILL_P4*vai;
402 /* Load atom i coordinates, add shift vectors */
403 ix1 = shX + x[ai][0];
404 iy1 = shY + x[ai][1];
405 iz1 = shZ + x[ai][2];
407 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
418 dr2 = dx11*dx11+dy11*dy11+dz11*dz11;
419 rinv = gmx_invsqrt(dr2);
424 raj = top->atomtypes.gb_radius[md->typeA[aj]];
428 ratio = dr2 / (rvdw * rvdw);
429 vaj = born->vsolv[aj];
431 if (ratio > STILL_P5INV)
438 theta = ratio*STILL_PIP5;
440 term = 0.5*(1.0-cosq);
442 sinq = 1.0 - cosq*cosq;
443 dccf = 2.0*term*sinq*gmx_invsqrt(sinq)*theta;
448 icf6 = (4*ccf-dccf)*idr6;
449 born->gpol_still_work[aj] += prod_ai*icf4;
452 /* Save ai->aj and aj->ai chain rule terms */
453 fr->dadx[n++] = prod*icf6;
454 fr->dadx[n++] = prod_ai*icf6;
456 born->gpol_still_work[ai] += gpi;
459 /* Parallel summations */
460 if (DOMAINDECOMP(cr))
462 dd_atom_sum_real(cr->dd, born->gpol_still_work);
465 /* Calculate the radii */
466 for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
468 if (born->use[i] != 0)
470 gpi = born->gpol[i]+born->gpol_still_work[i];
472 born->bRad[i] = factor*gmx_invsqrt(gpi2);
473 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
477 /* Extra communication required for DD */
478 if (DOMAINDECOMP(cr))
480 dd_atom_spread_real(cr->dd, born->bRad);
481 dd_atom_spread_real(cr->dd, fr->invsqrta);
490 calc_gb_rad_hct(t_commrec *cr, t_forcerec *fr, gmx_localtop_t *top,
491 rvec x[], t_nblist *nl,
492 gmx_genborn_t *born, t_mdatoms *md)
494 int i, k, n, ai, aj, nj0, nj1, at0, at1;
497 real rai, raj, gpi, dr2, dr, sk, sk_ai, sk2, sk2_ai, lij, uij, diff2, tmp, sum_ai;
498 real rad, min_rad, rinv, rai_inv;
499 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11;
500 real lij2, uij2, lij3, uij3, t1, t2, t3;
501 real lij_inv, dlij, duij, sk2_rinv, prod, log_term;
502 real doffset, raj_inv, dadx_val;
505 doffset = born->gb_doffset;
506 gb_radius = born->gb_radius;
508 for (i = 0; i < born->nr; i++)
510 born->gpol_hct_work[i] = 0;
513 /* Keep the compiler happy */
517 for (i = 0; i < nl->nri; i++)
522 nj1 = nl->jindex[i+1];
524 /* Load shifts for this list */
525 shift = nl->shift[i];
526 shX = fr->shift_vec[shift][0];
527 shY = fr->shift_vec[shift][1];
528 shZ = fr->shift_vec[shift][2];
533 sk_ai = born->param[ai];
534 sk2_ai = sk_ai*sk_ai;
536 /* Load atom i coordinates, add shift vectors */
537 ix1 = shX + x[ai][0];
538 iy1 = shY + x[ai][1];
539 iz1 = shZ + x[ai][2];
543 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
555 dr2 = dx11*dx11+dy11*dy11+dz11*dz11;
556 rinv = gmx_invsqrt(dr2);
559 sk = born->param[aj];
562 /* aj -> ai interaction */
583 lij_inv = gmx_invsqrt(lij2);
586 prod = 0.25*sk2_rinv;
588 log_term = log(uij*lij_inv);
590 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term +
595 tmp = tmp + 2.0 * (rai_inv-lij);
598 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
599 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
600 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
602 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
603 /* fr->dadx[n++] = (dlij*t1+duij*t2+t3)*rinv; */
604 /* rb2 is moved to chainrule */
612 fr->dadx[n++] = dadx_val;
615 /* ai -> aj interaction */
616 if (raj < dr + sk_ai)
618 lij = 1.0/(dr-sk_ai);
631 uij = 1.0/(dr+sk_ai);
637 lij_inv = gmx_invsqrt(lij2);
638 sk2 = sk2_ai; /* sk2_ai = sk_ai * sk_ai in i loop above */
640 prod = 0.25 * sk2_rinv;
642 /* log_term = table_log(uij*lij_inv,born->log_table,
643 LOG_TABLE_ACCURACY); */
644 log_term = log(uij*lij_inv);
646 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term +
651 tmp = tmp + 2.0 * (raj_inv-lij);
655 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
656 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
657 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
659 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
660 /* fr->dadx[n++] = (dlij*t1+duij*t2+t3)*rinv; */ /* rb2 is moved to chainrule */
662 born->gpol_hct_work[aj] += 0.5*tmp;
668 fr->dadx[n++] = dadx_val;
671 born->gpol_hct_work[ai] += sum_ai;
674 /* Parallel summations */
675 if (DOMAINDECOMP(cr))
677 dd_atom_sum_real(cr->dd, born->gpol_hct_work);
680 for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
682 if (born->use[i] != 0)
684 rai = top->atomtypes.gb_radius[md->typeA[i]]-doffset;
685 sum_ai = 1.0/rai - born->gpol_hct_work[i];
686 min_rad = rai + doffset;
689 born->bRad[i] = rad > min_rad ? rad : min_rad;
690 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
694 /* Extra communication required for DD */
695 if (DOMAINDECOMP(cr))
697 dd_atom_spread_real(cr->dd, born->bRad);
698 dd_atom_spread_real(cr->dd, fr->invsqrta);
706 calc_gb_rad_obc(t_commrec *cr, t_forcerec *fr, gmx_localtop_t *top,
707 rvec x[], t_nblist *nl, gmx_genborn_t *born, t_mdatoms *md)
709 int i, k, ai, aj, nj0, nj1, n, at0, at1;
712 real rai, raj, gpi, dr2, dr, sk, sk2, lij, uij, diff2, tmp, sum_ai;
713 real rad, min_rad, sum_ai2, sum_ai3, tsum, tchain, rinv, rai_inv, lij_inv, rai_inv2;
714 real log_term, prod, sk2_rinv, sk_ai, sk2_ai;
715 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11;
716 real lij2, uij2, lij3, uij3, dlij, duij, t1, t2, t3;
717 real doffset, raj_inv, dadx_val;
720 /* Keep the compiler happy */
725 doffset = born->gb_doffset;
726 gb_radius = born->gb_radius;
728 for (i = 0; i < born->nr; i++)
730 born->gpol_hct_work[i] = 0;
733 for (i = 0; i < nl->nri; i++)
738 nj1 = nl->jindex[i+1];
740 /* Load shifts for this list */
741 shift = nl->shift[i];
742 shX = fr->shift_vec[shift][0];
743 shY = fr->shift_vec[shift][1];
744 shZ = fr->shift_vec[shift][2];
749 sk_ai = born->param[ai];
750 sk2_ai = sk_ai*sk_ai;
752 /* Load atom i coordinates, add shift vectors */
753 ix1 = shX + x[ai][0];
754 iy1 = shY + x[ai][1];
755 iz1 = shZ + x[ai][2];
759 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
771 dr2 = dx11*dx11+dy11*dy11+dz11*dz11;
772 rinv = gmx_invsqrt(dr2);
775 /* sk is precalculated in init_gb() */
776 sk = born->param[aj];
779 /* aj -> ai interaction */
799 lij_inv = gmx_invsqrt(lij2);
802 prod = 0.25*sk2_rinv;
804 log_term = log(uij*lij_inv);
806 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term + prod*(-diff2);
810 tmp = tmp + 2.0 * (rai_inv-lij);
814 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
815 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
816 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
818 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
826 fr->dadx[n++] = dadx_val;
828 /* ai -> aj interaction */
829 if (raj < dr + sk_ai)
831 lij = 1.0/(dr-sk_ai);
844 uij = 1.0/(dr+sk_ai);
850 lij_inv = gmx_invsqrt(lij2);
851 sk2 = sk2_ai; /* sk2_ai = sk_ai * sk_ai in i loop above */
853 prod = 0.25 * sk2_rinv;
855 /* log_term = table_log(uij*lij_inv,born->log_table,LOG_TABLE_ACCURACY); */
856 log_term = log(uij*lij_inv);
858 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term + prod*(-diff2);
862 tmp = tmp + 2.0 * (raj_inv-lij);
865 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
866 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
867 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
869 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
871 born->gpol_hct_work[aj] += 0.5*tmp;
878 fr->dadx[n++] = dadx_val;
881 born->gpol_hct_work[ai] += sum_ai;
885 /* Parallel summations */
886 if (DOMAINDECOMP(cr))
888 dd_atom_sum_real(cr->dd, born->gpol_hct_work);
891 for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
893 if (born->use[i] != 0)
895 rai = top->atomtypes.gb_radius[md->typeA[i]];
899 sum_ai = rai * born->gpol_hct_work[i];
900 sum_ai2 = sum_ai * sum_ai;
901 sum_ai3 = sum_ai2 * sum_ai;
903 tsum = tanh(born->obc_alpha*sum_ai-born->obc_beta*sum_ai2+born->obc_gamma*sum_ai3);
904 born->bRad[i] = rai_inv - tsum*rai_inv2;
905 born->bRad[i] = 1.0 / born->bRad[i];
907 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
909 tchain = rai * (born->obc_alpha-2*born->obc_beta*sum_ai+3*born->obc_gamma*sum_ai2);
910 born->drobc[i] = (1.0-tsum*tsum)*tchain*rai_inv2;
914 /* Extra (local) communication required for DD */
915 if (DOMAINDECOMP(cr))
917 dd_atom_spread_real(cr->dd, born->bRad);
918 dd_atom_spread_real(cr->dd, fr->invsqrta);
919 dd_atom_spread_real(cr->dd, born->drobc);
928 int calc_gb_rad(t_commrec *cr, t_forcerec *fr, t_inputrec *ir, gmx_localtop_t *top,
929 rvec x[], t_nblist *nl, gmx_genborn_t *born, t_mdatoms *md, t_nrnb *nrnb)
935 if (fr->bAllvsAll && fr->dadx == NULL)
937 /* We might need up to 8 atoms of padding before and after,
938 * and another 4 units to guarantee SSE alignment.
940 fr->nalloc_dadx = 2*(md->homenr+12)*(md->nr/2+1+12);
941 snew(fr->dadx_rawptr, fr->nalloc_dadx);
942 fr->dadx = (real *) (((size_t) fr->dadx_rawptr + 16) & (~((size_t) 15)));
946 /* In the SSE-enabled gb-loops, when writing to dadx, we
947 * always write 2*4 elements at a time, even in the case with only
948 * 1-3 j particles, where we only really need to write 2*(1-3)
949 * elements. This is because we want dadx to be aligned to a 16-
950 * byte boundary, and being able to use _mm_store/load_ps
952 ndadx = 2 * (nl->nrj + 3*nl->nri);
954 /* First, reallocate the dadx array, we need 3 extra for SSE */
955 if (ndadx + 3 > fr->nalloc_dadx)
957 fr->nalloc_dadx = over_alloc_large(ndadx) + 3;
958 srenew(fr->dadx_rawptr, fr->nalloc_dadx);
959 fr->dadx = (real *) (((size_t) fr->dadx_rawptr + 16) & (~((size_t) 15)));
965 cnt = md->homenr*(md->nr/2+1);
967 if (ir->gb_algorithm == egbSTILL)
969 genborn_allvsall_calc_still_radii(fr, md, born, top, x[0], &fr->AllvsAll_workgb);
970 /* 13 flops in outer loop, 47 flops in inner loop */
971 inc_nrnb(nrnb, eNR_BORN_AVA_RADII_STILL, md->homenr*13+cnt*47);
973 else if (ir->gb_algorithm == egbHCT || ir->gb_algorithm == egbOBC)
975 genborn_allvsall_calc_hct_obc_radii(fr, md, born, ir->gb_algorithm, top, x[0], &fr->AllvsAll_workgb);
976 /* 24 flops in outer loop, 183 in inner */
977 inc_nrnb(nrnb, eNR_BORN_AVA_RADII_HCT_OBC, md->homenr*24+cnt*183);
981 gmx_fatal(FARGS, "Bad gb algorithm for all-vs-all interactions");
986 /* Switch for determining which algorithm to use for Born radii calculation */
989 switch (ir->gb_algorithm)
992 calc_gb_rad_still(cr, fr, top, x, nl, born, md);
995 calc_gb_rad_hct(cr, fr, top, x, nl, born, md);
998 calc_gb_rad_obc(cr, fr, top, x, nl, born, md);
1002 gmx_fatal(FARGS, "Unknown double precision algorithm for Born radii calculation: %d", ir->gb_algorithm);
1007 switch (ir->gb_algorithm)
1010 calc_gb_rad_still(cr, fr, top, x, nl, born, md);
1013 calc_gb_rad_hct(cr, fr, top, x, nl, born, md);
1016 calc_gb_rad_obc(cr, fr, top, x, nl, born, md);
1020 gmx_fatal(FARGS, "Unknown algorithm for Born radii calculation: %d", ir->gb_algorithm);
1023 #endif /* Double or single precision */
1025 if (fr->bAllvsAll == FALSE)
1027 switch (ir->gb_algorithm)
1030 /* 17 flops per outer loop iteration, 47 flops per inner loop */
1031 inc_nrnb(nrnb, eNR_BORN_RADII_STILL, nl->nri*17+nl->nrj*47);
1035 /* 61 (assuming 10 for tanh) flops for outer loop iteration, 183 flops per inner loop */
1036 inc_nrnb(nrnb, eNR_BORN_RADII_HCT_OBC, nl->nri*61+nl->nrj*183);
1049 real gb_bonds_tab(rvec x[], rvec f[], rvec fshift[], real *charge, real *p_gbtabscale,
1050 real *invsqrta, real *dvda, real *GBtab, t_idef *idef, real epsilon_r,
1051 real gb_epsilon_solvent, real facel, const t_pbc *pbc, const t_graph *graph)
1053 int i, j, n0, m, nnn, type, ai, aj;
1059 real isaprod, qq, gbscale, gbtabscale, Y, F, Geps, Heps2, Fp, VV, FF, rt, eps, eps2;
1060 real vgb, fgb, vcoul, fijC, dvdatmp, fscal, dvdaj;
1066 t_iatom *forceatoms;
1068 /* Scale the electrostatics by gb_epsilon_solvent */
1069 facel = facel * ((1.0/epsilon_r) - 1.0/gb_epsilon_solvent);
1071 gbtabscale = *p_gbtabscale;
1074 for (j = F_GB12; j <= F_GB14; j++)
1076 forceatoms = idef->il[j].iatoms;
1078 for (i = 0; i < idef->il[j].nr; )
1080 /* To avoid reading in the interaction type, we just increment i to pass over
1081 * the types in the forceatoms array, this saves some memory accesses
1084 ai = forceatoms[i++];
1085 aj = forceatoms[i++];
1087 ki = pbc_rvec_sub(pbc, x[ai], x[aj], dx);
1088 rsq11 = iprod(dx, dx);
1090 isai = invsqrta[ai];
1091 iq = (-1)*facel*charge[ai];
1093 rinv11 = gmx_invsqrt(rsq11);
1094 isaj = invsqrta[aj];
1095 isaprod = isai*isaj;
1096 qq = isaprod*iq*charge[aj];
1097 gbscale = isaprod*gbtabscale;
1106 Geps = eps*GBtab[nnn+2];
1107 Heps2 = eps2*GBtab[nnn+3];
1110 FF = Fp+Geps+2.0*Heps2;
1112 fijC = qq*FF*gbscale;
1113 dvdatmp = -(vgb+fijC*r)*0.5;
1114 dvda[aj] = dvda[aj] + dvdatmp*isaj*isaj;
1115 dvda[ai] = dvda[ai] + dvdatmp*isai*isai;
1116 vctot = vctot + vgb;
1117 fgb = -(fijC)*rinv11;
1121 ivec_sub(SHIFT_IVEC(graph, ai), SHIFT_IVEC(graph, aj), dt);
1125 for (m = 0; (m < DIM); m++) /* 15 */
1130 fshift[ki][m] += fscal;
1131 fshift[CENTRAL][m] -= fscal;
1139 real calc_gb_selfcorrections(t_commrec *cr, int natoms,
1140 real *charge, gmx_genborn_t *born, real *dvda, double facel)
1142 int i, ai, at0, at1;
1143 real rai, e, derb, q, q2, fi, rai_inv, vtot;
1145 if (DOMAINDECOMP(cr))
1148 at1 = cr->dd->nat_home;
1157 /* Scale the electrostatics by gb_epsilon_solvent */
1158 facel = facel * ((1.0/born->epsilon_r) - 1.0/born->gb_epsilon_solvent);
1162 /* Apply self corrections */
1163 for (i = at0; i < at1; i++)
1167 if (born->use[ai] == 1)
1169 rai = born->bRad[ai];
1175 derb = 0.5*e*rai_inv*rai_inv;
1176 dvda[ai] += derb*rai;
1185 real calc_gb_nonpolar(t_commrec *cr, t_forcerec *fr, int natoms, gmx_genborn_t *born, gmx_localtop_t *top,
1186 real *dvda, t_mdatoms *md)
1188 int ai, i, at0, at1;
1189 real e, es, rai, rbi, term, probe, tmp, factor;
1190 real rbi_inv, rbi_inv2;
1192 /* To keep the compiler happy */
1195 if (DOMAINDECOMP(cr))
1198 at1 = cr->dd->nat_home;
1206 /* factor is the surface tension */
1207 factor = born->sa_surface_tension;
1210 // The surface tension factor is 0.0049 for Still model, 0.0054 for HCT/OBC
1211 if(gb_algorithm==egbSTILL)
1213 factor=0.0049*100*CAL2JOULE;
1217 factor=0.0054*100*CAL2JOULE;
1220 /* if(gb_algorithm==egbHCT || gb_algorithm==egbOBC) */
1226 for (i = at0; i < at1; i++)
1230 if (born->use[ai] == 1)
1232 rai = top->atomtypes.gb_radius[md->typeA[ai]];
1233 rbi_inv = fr->invsqrta[ai];
1234 rbi_inv2 = rbi_inv * rbi_inv;
1235 tmp = (rai*rbi_inv2)*(rai*rbi_inv2);
1237 e = factor*term*(rai+probe)*(rai+probe)*tmp;
1238 dvda[ai] = dvda[ai] - 6*e*rbi_inv2;
1248 real calc_gb_chainrule(int natoms, t_nblist *nl, real *dadx, real *dvda, rvec x[], rvec t[], rvec fshift[],
1249 rvec shift_vec[], int gb_algorithm, gmx_genborn_t *born)
1251 int i, k, n, ai, aj, nj0, nj1, n0, n1;
1254 real fgb, fij, rb2, rbi, fix1, fiy1, fiz1;
1255 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11, rsq11;
1256 real rinv11, tx, ty, tz, rbai, rbaj, fgb_ai;
1266 if (gb_algorithm == egbSTILL)
1268 for (i = n0; i < n1; i++)
1270 rbi = born->bRad[i];
1271 rb[i] = (2 * rbi * rbi * dvda[i])/ONE_4PI_EPS0;
1274 else if (gb_algorithm == egbHCT)
1276 for (i = n0; i < n1; i++)
1278 rbi = born->bRad[i];
1279 rb[i] = rbi * rbi * dvda[i];
1282 else if (gb_algorithm == egbOBC)
1284 for (i = n0; i < n1; i++)
1286 rbi = born->bRad[i];
1287 rb[i] = rbi * rbi * born->drobc[i] * dvda[i];
1291 for (i = 0; i < nl->nri; i++)
1295 nj0 = nl->jindex[i];
1296 nj1 = nl->jindex[i+1];
1298 /* Load shifts for this list */
1299 shift = nl->shift[i];
1300 shX = shift_vec[shift][0];
1301 shY = shift_vec[shift][1];
1302 shZ = shift_vec[shift][2];
1304 /* Load atom i coordinates, add shift vectors */
1305 ix1 = shX + x[ai][0];
1306 iy1 = shY + x[ai][1];
1307 iz1 = shZ + x[ai][2];
1315 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
1329 fgb = rbai*dadx[n++];
1330 fgb_ai = rbaj*dadx[n++];
1332 /* Total force between ai and aj is the sum of ai->aj and aj->ai */
1343 /* Update force on atom aj */
1344 t[aj][0] = t[aj][0] - tx;
1345 t[aj][1] = t[aj][1] - ty;
1346 t[aj][2] = t[aj][2] - tz;
1349 /* Update force and shift forces on atom ai */
1350 t[ai][0] = t[ai][0] + fix1;
1351 t[ai][1] = t[ai][1] + fiy1;
1352 t[ai][2] = t[ai][2] + fiz1;
1354 fshift[shift][0] = fshift[shift][0] + fix1;
1355 fshift[shift][1] = fshift[shift][1] + fiy1;
1356 fshift[shift][2] = fshift[shift][2] + fiz1;
1365 calc_gb_forces(t_commrec *cr, t_mdatoms *md, gmx_genborn_t *born, gmx_localtop_t *top,
1366 rvec x[], rvec f[], t_forcerec *fr, t_idef *idef, int gb_algorithm, int sa_algorithm, t_nrnb *nrnb,
1367 const t_pbc *pbc, const t_graph *graph, gmx_enerdata_t *enerd)
1374 const t_pbc *pbc_null;
1385 if (sa_algorithm == esaAPPROX)
1387 /* Do a simple ACE type approximation for the non-polar solvation */
1388 enerd->term[F_NPSOLVATION] += calc_gb_nonpolar(cr, fr, born->nr, born, top, fr->dvda, md);
1391 /* Calculate the bonded GB-interactions using either table or analytical formula */
1392 enerd->term[F_GBPOL] += gb_bonds_tab(x, f, fr->fshift, md->chargeA, &(fr->gbtabscale),
1393 fr->invsqrta, fr->dvda, fr->gbtab.data, idef, born->epsilon_r, born->gb_epsilon_solvent, fr->epsfac, pbc_null, graph);
1395 /* Calculate self corrections to the GB energies - currently only A state used! (FIXME) */
1396 enerd->term[F_GBPOL] += calc_gb_selfcorrections(cr, born->nr, md->chargeA, born, fr->dvda, fr->epsfac);
1398 /* If parallel, sum the derivative of the potential w.r.t the born radii */
1399 if (DOMAINDECOMP(cr))
1401 dd_atom_sum_real(cr->dd, fr->dvda);
1402 dd_atom_spread_real(cr->dd, fr->dvda);
1407 genborn_allvsall_calc_chainrule(fr, md, born, x[0], f[0], gb_algorithm, fr->AllvsAll_workgb);
1408 cnt = md->homenr*(md->nr/2+1);
1409 /* 9 flops for outer loop, 15 for inner */
1410 inc_nrnb(nrnb, eNR_BORN_AVA_CHAINRULE, md->homenr*9+cnt*15);
1414 calc_gb_chainrule(fr->natoms_force, &(fr->gblist), fr->dadx, fr->dvda,
1415 x, f, fr->fshift, fr->shift_vec, gb_algorithm, born);
1419 /* 9 flops for outer loop, 15 for inner */
1420 inc_nrnb(nrnb, eNR_BORN_CHAINRULE, fr->gblist.nri*9+fr->gblist.nrj*15);
1424 static void add_j_to_gblist(gbtmpnbl_t *list, int aj)
1426 if (list->naj >= list->aj_nalloc)
1428 list->aj_nalloc = over_alloc_large(list->naj+1);
1429 srenew(list->aj, list->aj_nalloc);
1432 list->aj[list->naj++] = aj;
1435 static gbtmpnbl_t *find_gbtmplist(struct gbtmpnbls *lists, int shift)
1439 /* Search the list with the same shift, if there is one */
1441 while (ind < lists->nlist && shift != lists->list[ind].shift)
1445 if (ind == lists->nlist)
1447 if (lists->nlist == lists->list_nalloc)
1449 lists->list_nalloc++;
1450 srenew(lists->list, lists->list_nalloc);
1451 for (i = lists->nlist; i < lists->list_nalloc; i++)
1453 lists->list[i].aj = NULL;
1454 lists->list[i].aj_nalloc = 0;
1459 lists->list[lists->nlist].shift = shift;
1460 lists->list[lists->nlist].naj = 0;
1464 return &lists->list[ind];
1467 static void add_bondeds_to_gblist(t_ilist *il,
1468 gmx_bool bMolPBC, t_pbc *pbc, t_graph *g, rvec *x,
1469 struct gbtmpnbls *nls)
1471 int ind, j, ai, aj, shift, found;
1477 for (ind = 0; ind < il->nr; ind += 3)
1479 ai = il->iatoms[ind+1];
1480 aj = il->iatoms[ind+2];
1485 rvec_sub(x[ai], x[aj], dx);
1486 ivec_sub(SHIFT_IVEC(g, ai), SHIFT_IVEC(g, aj), dt);
1487 shift = IVEC2IS(dt);
1491 shift = pbc_dx_aiuc(pbc, x[ai], x[aj], dx);
1494 /* Find the list for this shift or create one */
1495 list = find_gbtmplist(&nls[ai], shift);
1499 /* So that we do not add the same bond twice.
1500 * This happens with some constraints between 1-3 atoms
1501 * that are in the bond-list but should not be in the GB nb-list */
1502 for (j = 0; j < list->naj; j++)
1504 if (list->aj[j] == aj)
1514 gmx_incons("ai == aj");
1517 add_j_to_gblist(list, aj);
1523 compare_int (const void * a, const void * b)
1525 return ( *(int*)a - *(int*)b );
1530 int make_gb_nblist(t_commrec *cr, int gb_algorithm,
1531 rvec x[], matrix box,
1532 t_forcerec *fr, t_idef *idef, t_graph *graph, gmx_genborn_t *born)
1534 int i, l, ii, j, k, n, nj0, nj1, ai, aj, at0, at1, found, shift, s;
1539 struct gbtmpnbls *nls;
1540 gbtmpnbl_t *list = NULL;
1542 set_pbc(&pbc, fr->ePBC, box);
1543 nls = born->nblist_work;
1545 for (i = 0; i < born->nr; i++)
1552 set_pbc_dd(&pbc, fr->ePBC, cr->dd, TRUE, box);
1555 switch (gb_algorithm)
1559 /* Loop over 1-2, 1-3 and 1-4 interactions */
1560 for (j = F_GB12; j <= F_GB14; j++)
1562 add_bondeds_to_gblist(&idef->il[j], fr->bMolPBC, &pbc, graph, x, nls);
1566 /* Loop over 1-4 interactions */
1567 add_bondeds_to_gblist(&idef->il[F_GB14], fr->bMolPBC, &pbc, graph, x, nls);
1570 gmx_incons("Unknown GB algorithm");
1573 /* Loop over the VDWQQ and VDW nblists to set up the nonbonded part of the GB list */
1574 for (n = 0; (n < fr->nnblists); n++)
1576 for (i = 0; (i < eNL_NR); i++)
1578 nblist = &(fr->nblists[n].nlist_sr[i]);
1580 if (nblist->nri > 0 && (i == eNL_VDWQQ || i == eNL_QQ))
1582 for (j = 0; j < nblist->nri; j++)
1584 ai = nblist->iinr[j];
1585 shift = nblist->shift[j];
1587 /* Find the list for this shift or create one */
1588 list = find_gbtmplist(&nls[ai], shift);
1590 nj0 = nblist->jindex[j];
1591 nj1 = nblist->jindex[j+1];
1593 /* Add all the j-atoms in the non-bonded list to the GB list */
1594 for (k = nj0; k < nj1; k++)
1596 add_j_to_gblist(list, nblist->jjnr[k]);
1603 /* Zero out some counters */
1607 fr->gblist.jindex[0] = fr->gblist.nri;
1609 for (i = 0; i < fr->natoms_force; i++)
1611 for (s = 0; s < nls[i].nlist; s++)
1613 list = &nls[i].list[s];
1615 /* Only add those atoms that actually have neighbours */
1616 if (born->use[i] != 0)
1618 fr->gblist.iinr[fr->gblist.nri] = i;
1619 fr->gblist.shift[fr->gblist.nri] = list->shift;
1622 for (k = 0; k < list->naj; k++)
1624 /* Memory allocation for jjnr */
1625 if (fr->gblist.nrj >= fr->gblist.maxnrj)
1627 fr->gblist.maxnrj += over_alloc_large(fr->gblist.maxnrj);
1631 fprintf(debug, "Increasing GB neighbourlist j size to %d\n", fr->gblist.maxnrj);
1634 srenew(fr->gblist.jjnr, fr->gblist.maxnrj);
1638 if (i == list->aj[k])
1640 gmx_incons("i == list->aj[k]");
1642 fr->gblist.jjnr[fr->gblist.nrj++] = list->aj[k];
1645 fr->gblist.jindex[fr->gblist.nri] = fr->gblist.nrj;
1652 for (i = 0; i < fr->gblist.nri; i++)
1654 nj0 = fr->gblist.jindex[i];
1655 nj1 = fr->gblist.jindex[i+1];
1656 ai = fr->gblist.iinr[i];
1659 for (j = nj0; j < nj1; j++)
1661 if (fr->gblist.jjnr[j] < ai)
1663 fr->gblist.jjnr[j] += fr->natoms_force;
1666 qsort(fr->gblist.jjnr+nj0, nj1-nj0, sizeof(int), compare_int);
1668 for (j = nj0; j < nj1; j++)
1670 if (fr->gblist.jjnr[j] >= fr->natoms_force)
1672 fr->gblist.jjnr[j] -= fr->natoms_force;
1682 void make_local_gb(const t_commrec *cr, gmx_genborn_t *born, int gb_algorithm)
1685 gmx_domdec_t *dd = NULL;
1687 if (DOMAINDECOMP(cr))
1695 /* Single node, just copy pointers and return */
1696 if (gb_algorithm == egbSTILL)
1698 born->gpol = born->gpol_globalindex;
1699 born->vsolv = born->vsolv_globalindex;
1700 born->gb_radius = born->gb_radius_globalindex;
1704 born->param = born->param_globalindex;
1705 born->gb_radius = born->gb_radius_globalindex;
1708 born->use = born->use_globalindex;
1713 /* Reallocation of local arrays if necessary */
1714 /* fr->natoms_force is equal to dd->nat_tot */
1715 if (DOMAINDECOMP(cr) && dd->nat_tot > born->nalloc)
1719 nalloc = dd->nat_tot;
1721 /* Arrays specific to different gb algorithms */
1722 if (gb_algorithm == egbSTILL)
1724 srenew(born->gpol, nalloc+3);
1725 srenew(born->vsolv, nalloc+3);
1726 srenew(born->gb_radius, nalloc+3);
1727 for (i = born->nalloc; (i < nalloc+3); i++)
1731 born->gb_radius[i] = 0;
1736 srenew(born->param, nalloc+3);
1737 srenew(born->gb_radius, nalloc+3);
1738 for (i = born->nalloc; (i < nalloc+3); i++)
1741 born->gb_radius[i] = 0;
1745 /* All gb-algorithms use the array for vsites exclusions */
1746 srenew(born->use, nalloc+3);
1747 for (i = born->nalloc; (i < nalloc+3); i++)
1752 born->nalloc = nalloc;
1755 /* With dd, copy algorithm specific arrays */
1756 if (gb_algorithm == egbSTILL)
1758 for (i = at0; i < at1; i++)
1760 born->gpol[i] = born->gpol_globalindex[dd->gatindex[i]];
1761 born->vsolv[i] = born->vsolv_globalindex[dd->gatindex[i]];
1762 born->gb_radius[i] = born->gb_radius_globalindex[dd->gatindex[i]];
1763 born->use[i] = born->use_globalindex[dd->gatindex[i]];
1768 for (i = at0; i < at1; i++)
1770 born->param[i] = born->param_globalindex[dd->gatindex[i]];
1771 born->gb_radius[i] = born->gb_radius_globalindex[dd->gatindex[i]];
1772 born->use[i] = born->use_globalindex[dd->gatindex[i]];