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"
48 #include "gromacs/domdec/domdec.h"
49 #include "gromacs/fileio/pdbio.h"
50 #include "gromacs/legacyheaders/names.h"
51 #include "gromacs/legacyheaders/network.h"
52 #include "gromacs/legacyheaders/nrnb.h"
53 #include "gromacs/legacyheaders/typedefs.h"
54 #include "gromacs/legacyheaders/types/commrec.h"
55 #include "gromacs/math/units.h"
56 #include "gromacs/math/vec.h"
57 #include "gromacs/mdlib/genborn_allvsall.h"
58 #include "gromacs/pbcutil/ishift.h"
59 #include "gromacs/pbcutil/mshift.h"
60 #include "gromacs/pbcutil/pbc.h"
61 #include "gromacs/topology/mtop_util.h"
62 #include "gromacs/utility/fatalerror.h"
63 #include "gromacs/utility/gmxmpi.h"
64 #include "gromacs/utility/smalloc.h"
74 typedef struct gbtmpnbls {
80 /* This function is exactly the same as the one in listed-forces/bonded.cpp. The reason
81 * it is copied here is that the bonded gb-interactions are evaluated
82 * not in calc_bonds, but rather in calc_gb_forces
84 static int pbc_rvec_sub(const t_pbc *pbc, const rvec xi, const rvec xj, rvec dx)
88 return pbc_dx_aiuc(pbc, xi, xj, dx);
97 static int init_gb_nblist(int natoms, t_nblist *nl)
99 nl->maxnri = natoms*4;
108 /*nl->nltype = nltype;*/
110 srenew(nl->iinr, nl->maxnri);
111 srenew(nl->gid, nl->maxnri);
112 srenew(nl->shift, nl->maxnri);
113 srenew(nl->jindex, nl->maxnri+1);
121 static int init_gb_still(const t_atomtypes *atype, t_idef *idef, t_atoms *atoms,
122 gmx_genborn_t *born, int natoms)
126 real r, ri, rj, ri2, rj2, r3, r4, ratio, term, h, doffset;
133 snew(born->gpol_still_work, natoms+3);
135 doffset = born->gb_doffset;
137 for (i = 0; i < natoms; i++)
139 born->gpol_globalindex[i] = born->vsolv_globalindex[i] =
140 born->gb_radius_globalindex[i] = 0;
143 /* Compute atomic solvation volumes for Still method */
144 for (i = 0; i < natoms; i++)
146 ri = atype->gb_radius[atoms->atom[i].type];
147 born->gb_radius_globalindex[i] = ri;
149 born->vsolv_globalindex[i] = (4*M_PI/3)*r3;
152 for (j = 0; j < idef->il[F_GB12].nr; j += 3)
154 m = idef->il[F_GB12].iatoms[j];
155 ia = idef->il[F_GB12].iatoms[j+1];
156 ib = idef->il[F_GB12].iatoms[j+2];
158 r = 1.01*idef->iparams[m].gb.st;
160 ri = atype->gb_radius[atoms->atom[ia].type];
161 rj = atype->gb_radius[atoms->atom[ib].type];
166 ratio = (rj2-ri2-r*r)/(2*ri*r);
168 term = (M_PI/3.0)*h*h*(3.0*ri-h);
170 born->vsolv_globalindex[ia] -= term;
172 ratio = (ri2-rj2-r*r)/(2*rj*r);
174 term = (M_PI/3.0)*h*h*(3.0*rj-h);
176 born->vsolv_globalindex[ib] -= term;
179 /* Get the self-, 1-2 and 1-3 polarization energies for analytical Still
182 for (j = 0; j < natoms; j++)
184 if (born->use_globalindex[j] == 1)
186 born->gpol_globalindex[j] = -0.5*ONE_4PI_EPS0/
187 (atype->gb_radius[atoms->atom[j].type]-doffset+STILL_P1);
192 for (j = 0; j < idef->il[F_GB12].nr; j += 3)
194 m = idef->il[F_GB12].iatoms[j];
195 ia = idef->il[F_GB12].iatoms[j+1];
196 ib = idef->il[F_GB12].iatoms[j+2];
198 r = idef->iparams[m].gb.st;
202 born->gpol_globalindex[ia] = born->gpol_globalindex[ia]+
203 STILL_P2*born->vsolv_globalindex[ib]/r4;
204 born->gpol_globalindex[ib] = born->gpol_globalindex[ib]+
205 STILL_P2*born->vsolv_globalindex[ia]/r4;
209 for (j = 0; j < idef->il[F_GB13].nr; j += 3)
211 m = idef->il[F_GB13].iatoms[j];
212 ia = idef->il[F_GB13].iatoms[j+1];
213 ib = idef->il[F_GB13].iatoms[j+2];
215 r = idef->iparams[m].gb.st;
218 born->gpol_globalindex[ia] = born->gpol_globalindex[ia]+
219 STILL_P3*born->vsolv_globalindex[ib]/r4;
220 born->gpol_globalindex[ib] = born->gpol_globalindex[ib]+
221 STILL_P3*born->vsolv_globalindex[ia]/r4;
230 /* Initialize all GB datastructs and compute polarization energies */
231 int init_gb(gmx_genborn_t **p_born,
232 t_forcerec *fr, const t_inputrec *ir,
233 const gmx_mtop_t *mtop, int gb_algorithm)
236 real rai, sk, doffset;
240 gmx_localtop_t *localtop;
242 natoms = mtop->natoms;
244 atoms = gmx_mtop_global_atoms(mtop);
245 localtop = gmx_mtop_generate_local_top(mtop, ir);
252 snew(born->drobc, natoms);
253 snew(born->bRad, natoms);
255 /* Allocate memory for the global data arrays */
256 snew(born->param_globalindex, natoms+3);
257 snew(born->gpol_globalindex, natoms+3);
258 snew(born->vsolv_globalindex, natoms+3);
259 snew(born->gb_radius_globalindex, natoms+3);
260 snew(born->use_globalindex, natoms+3);
262 snew(fr->invsqrta, natoms);
263 snew(fr->dvda, natoms);
266 fr->dadx_rawptr = NULL;
268 born->gpol_still_work = NULL;
269 born->gpol_hct_work = NULL;
271 /* snew(born->asurf,natoms); */
272 /* snew(born->dasurf,natoms); */
274 /* Initialize the gb neighbourlist */
275 init_gb_nblist(natoms, &(fr->gblist));
277 /* Do the Vsites exclusions (if any) */
278 for (i = 0; i < natoms; i++)
280 jj = atoms.atom[i].type;
281 if (mtop->atomtypes.gb_radius[atoms.atom[i].type] > 0)
283 born->use_globalindex[i] = 1;
287 born->use_globalindex[i] = 0;
290 /* If we have a Vsite, put vs_globalindex[i]=0 */
291 if (C6 (fr->nbfp, fr->ntype, jj, jj) == 0 &&
292 C12(fr->nbfp, fr->ntype, jj, jj) == 0 &&
293 atoms.atom[i].q == 0)
295 born->use_globalindex[i] = 0;
299 /* Copy algorithm parameters from inputrecord to local structure */
300 born->obc_alpha = ir->gb_obc_alpha;
301 born->obc_beta = ir->gb_obc_beta;
302 born->obc_gamma = ir->gb_obc_gamma;
303 born->gb_doffset = ir->gb_dielectric_offset;
304 born->gb_epsilon_solvent = ir->gb_epsilon_solvent;
305 born->epsilon_r = ir->epsilon_r;
307 doffset = born->gb_doffset;
309 /* Set the surface tension */
310 born->sa_surface_tension = ir->sa_surface_tension;
312 /* If Still model, initialise the polarisation energies */
313 if (gb_algorithm == egbSTILL)
315 init_gb_still(&(mtop->atomtypes), &(localtop->idef), &atoms,
320 /* If HCT/OBC, precalculate the sk*atype->S_hct factors */
321 else if (gb_algorithm == egbHCT || gb_algorithm == egbOBC)
324 snew(born->gpol_hct_work, natoms+3);
326 for (i = 0; i < natoms; i++)
328 if (born->use_globalindex[i] == 1)
330 rai = mtop->atomtypes.gb_radius[atoms.atom[i].type]-doffset;
331 sk = rai * mtop->atomtypes.S_hct[atoms.atom[i].type];
332 born->param_globalindex[i] = sk;
333 born->gb_radius_globalindex[i] = rai;
337 born->param_globalindex[i] = 0;
338 born->gb_radius_globalindex[i] = 0;
343 /* Allocate memory for work arrays for temporary use */
344 snew(born->work, natoms+4);
345 snew(born->count, natoms);
346 snew(born->nblist_work, natoms);
348 /* Domain decomposition specific stuff */
357 calc_gb_rad_still(t_commrec *cr, t_forcerec *fr, gmx_localtop_t *top,
358 rvec x[], t_nblist *nl,
359 gmx_genborn_t *born, t_mdatoms *md)
361 int i, k, n, nj0, nj1, ai, aj;
364 real gpi, dr2, idr4, rvdw, ratio, ccf, theta, term, rai, raj;
365 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11;
366 real rinv, idr2, idr6, vaj, dccf, cosq, sinq, prod, gpi2;
368 real vai, prod_ai, icf4, icf6;
370 factor = 0.5*ONE_4PI_EPS0;
373 for (i = 0; i < born->nr; i++)
375 born->gpol_still_work[i] = 0;
378 for (i = 0; i < nl->nri; i++)
383 nj1 = nl->jindex[i+1];
385 /* Load shifts for this list */
386 shift = nl->shift[i];
387 shX = fr->shift_vec[shift][0];
388 shY = fr->shift_vec[shift][1];
389 shZ = fr->shift_vec[shift][2];
393 rai = top->atomtypes.gb_radius[md->typeA[ai]];
394 vai = born->vsolv[ai];
395 prod_ai = STILL_P4*vai;
397 /* Load atom i coordinates, add shift vectors */
398 ix1 = shX + x[ai][0];
399 iy1 = shY + x[ai][1];
400 iz1 = shZ + x[ai][2];
402 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
413 dr2 = dx11*dx11+dy11*dy11+dz11*dz11;
414 rinv = gmx_invsqrt(dr2);
419 raj = top->atomtypes.gb_radius[md->typeA[aj]];
423 ratio = dr2 / (rvdw * rvdw);
424 vaj = born->vsolv[aj];
426 if (ratio > STILL_P5INV)
433 theta = ratio*STILL_PIP5;
435 term = 0.5*(1.0-cosq);
437 sinq = 1.0 - cosq*cosq;
438 dccf = 2.0*term*sinq*gmx_invsqrt(sinq)*theta;
443 icf6 = (4*ccf-dccf)*idr6;
444 born->gpol_still_work[aj] += prod_ai*icf4;
447 /* Save ai->aj and aj->ai chain rule terms */
448 fr->dadx[n++] = prod*icf6;
449 fr->dadx[n++] = prod_ai*icf6;
451 born->gpol_still_work[ai] += gpi;
454 /* Parallel summations */
455 if (DOMAINDECOMP(cr))
457 dd_atom_sum_real(cr->dd, born->gpol_still_work);
460 /* Calculate the radii */
461 for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
463 if (born->use[i] != 0)
465 gpi = born->gpol[i]+born->gpol_still_work[i];
467 born->bRad[i] = factor*gmx_invsqrt(gpi2);
468 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
472 /* Extra communication required for DD */
473 if (DOMAINDECOMP(cr))
475 dd_atom_spread_real(cr->dd, born->bRad);
476 dd_atom_spread_real(cr->dd, fr->invsqrta);
485 calc_gb_rad_hct(t_commrec *cr, t_forcerec *fr, gmx_localtop_t *top,
486 rvec x[], t_nblist *nl,
487 gmx_genborn_t *born, t_mdatoms *md)
489 int i, k, n, ai, aj, nj0, nj1;
492 real rai, raj, dr2, dr, sk, sk_ai, sk2, sk2_ai, lij, uij, diff2, tmp, sum_ai;
493 real rad, min_rad, rinv, rai_inv;
494 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11;
495 real lij2, uij2, lij3, uij3, t1, t2, t3;
496 real lij_inv, dlij, sk2_rinv, prod, log_term;
497 real doffset, raj_inv, dadx_val;
500 doffset = born->gb_doffset;
501 gb_radius = born->gb_radius;
503 for (i = 0; i < born->nr; i++)
505 born->gpol_hct_work[i] = 0;
508 /* Keep the compiler happy */
511 for (i = 0; i < nl->nri; i++)
516 nj1 = nl->jindex[i+1];
518 /* Load shifts for this list */
519 shift = nl->shift[i];
520 shX = fr->shift_vec[shift][0];
521 shY = fr->shift_vec[shift][1];
522 shZ = fr->shift_vec[shift][2];
527 sk_ai = born->param[ai];
528 sk2_ai = sk_ai*sk_ai;
530 /* Load atom i coordinates, add shift vectors */
531 ix1 = shX + x[ai][0];
532 iy1 = shY + x[ai][1];
533 iz1 = shZ + x[ai][2];
537 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
549 dr2 = dx11*dx11+dy11*dy11+dz11*dz11;
550 rinv = gmx_invsqrt(dr2);
553 sk = born->param[aj];
556 /* aj -> ai interaction */
577 lij_inv = gmx_invsqrt(lij2);
580 prod = 0.25*sk2_rinv;
582 log_term = std::log(uij*lij_inv);
584 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term +
589 tmp = tmp + 2.0 * (rai_inv-lij);
592 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
593 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
594 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
596 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
597 /* fr->dadx[n++] = (dlij*t1+duij*t2+t3)*rinv; */
598 /* rb2 is moved to chainrule */
606 fr->dadx[n++] = dadx_val;
609 /* ai -> aj interaction */
610 if (raj < dr + sk_ai)
612 lij = 1.0/(dr-sk_ai);
625 uij = 1.0/(dr+sk_ai);
631 lij_inv = gmx_invsqrt(lij2);
632 sk2 = sk2_ai; /* sk2_ai = sk_ai * sk_ai in i loop above */
634 prod = 0.25 * sk2_rinv;
636 /* log_term = table_log(uij*lij_inv,born->log_table,
637 LOG_TABLE_ACCURACY); */
638 log_term = std::log(uij*lij_inv);
640 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term +
645 tmp = tmp + 2.0 * (raj_inv-lij);
649 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
650 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
651 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
653 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
654 /* fr->dadx[n++] = (dlij*t1+duij*t2+t3)*rinv; */ /* rb2 is moved to chainrule */
656 born->gpol_hct_work[aj] += 0.5*tmp;
662 fr->dadx[n++] = dadx_val;
665 born->gpol_hct_work[ai] += sum_ai;
668 /* Parallel summations */
669 if (DOMAINDECOMP(cr))
671 dd_atom_sum_real(cr->dd, born->gpol_hct_work);
674 for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
676 if (born->use[i] != 0)
678 rai = top->atomtypes.gb_radius[md->typeA[i]]-doffset;
679 sum_ai = 1.0/rai - born->gpol_hct_work[i];
680 min_rad = rai + doffset;
683 born->bRad[i] = std::max(rad, min_rad);
684 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
688 /* Extra communication required for DD */
689 if (DOMAINDECOMP(cr))
691 dd_atom_spread_real(cr->dd, born->bRad);
692 dd_atom_spread_real(cr->dd, fr->invsqrta);
700 calc_gb_rad_obc(t_commrec *cr, t_forcerec *fr, gmx_localtop_t *top,
701 rvec x[], t_nblist *nl, gmx_genborn_t *born, t_mdatoms *md)
703 int i, k, ai, aj, nj0, nj1, n;
706 real rai, raj, dr2, dr, sk, sk2, lij, uij, diff2, tmp, sum_ai;
707 real sum_ai2, sum_ai3, tsum, tchain, rinv, rai_inv, lij_inv, rai_inv2;
708 real log_term, prod, sk2_rinv, sk_ai, sk2_ai;
709 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11;
710 real lij2, uij2, lij3, uij3, dlij, t1, t2, t3;
711 real doffset, raj_inv, dadx_val;
714 /* Keep the compiler happy */
717 doffset = born->gb_doffset;
718 gb_radius = born->gb_radius;
720 for (i = 0; i < born->nr; i++)
722 born->gpol_hct_work[i] = 0;
725 for (i = 0; i < nl->nri; i++)
730 nj1 = nl->jindex[i+1];
732 /* Load shifts for this list */
733 shift = nl->shift[i];
734 shX = fr->shift_vec[shift][0];
735 shY = fr->shift_vec[shift][1];
736 shZ = fr->shift_vec[shift][2];
741 sk_ai = born->param[ai];
742 sk2_ai = sk_ai*sk_ai;
744 /* Load atom i coordinates, add shift vectors */
745 ix1 = shX + x[ai][0];
746 iy1 = shY + x[ai][1];
747 iz1 = shZ + x[ai][2];
751 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
763 dr2 = dx11*dx11+dy11*dy11+dz11*dz11;
764 rinv = gmx_invsqrt(dr2);
767 /* sk is precalculated in init_gb() */
768 sk = born->param[aj];
771 /* aj -> ai interaction */
791 lij_inv = gmx_invsqrt(lij2);
794 prod = 0.25*sk2_rinv;
796 log_term = std::log(uij*lij_inv);
798 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term + prod*(-diff2);
802 tmp = tmp + 2.0 * (rai_inv-lij);
806 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
807 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
808 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
810 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
818 fr->dadx[n++] = dadx_val;
820 /* ai -> aj interaction */
821 if (raj < dr + sk_ai)
823 lij = 1.0/(dr-sk_ai);
836 uij = 1.0/(dr+sk_ai);
842 lij_inv = gmx_invsqrt(lij2);
843 sk2 = sk2_ai; /* sk2_ai = sk_ai * sk_ai in i loop above */
845 prod = 0.25 * sk2_rinv;
847 /* log_term = table_log(uij*lij_inv,born->log_table,LOG_TABLE_ACCURACY); */
848 log_term = std::log(uij*lij_inv);
850 tmp = lij-uij + 0.25*dr*diff2 + (0.5*rinv)*log_term + prod*(-diff2);
854 tmp = tmp + 2.0 * (raj_inv-lij);
857 t1 = 0.5*lij2 + prod*lij3 - 0.25*(lij*rinv+lij3*dr);
858 t2 = -0.5*uij2 - 0.25*sk2_rinv*uij3 + 0.25*(uij*rinv+uij3*dr);
859 t3 = 0.125*(1.0+sk2_rinv*rinv)*(-diff2)+0.25*log_term*rinv*rinv;
861 dadx_val = (dlij*t1+t2+t3)*rinv; /* rb2 is moved to chainrule */
863 born->gpol_hct_work[aj] += 0.5*tmp;
870 fr->dadx[n++] = dadx_val;
873 born->gpol_hct_work[ai] += sum_ai;
877 /* Parallel summations */
878 if (DOMAINDECOMP(cr))
880 dd_atom_sum_real(cr->dd, born->gpol_hct_work);
883 for (i = 0; i < fr->natoms_force; i++) /* PELA born->nr */
885 if (born->use[i] != 0)
887 rai = top->atomtypes.gb_radius[md->typeA[i]];
891 sum_ai = rai * born->gpol_hct_work[i];
892 sum_ai2 = sum_ai * sum_ai;
893 sum_ai3 = sum_ai2 * sum_ai;
895 tsum = tanh(born->obc_alpha*sum_ai-born->obc_beta*sum_ai2+born->obc_gamma*sum_ai3);
896 born->bRad[i] = rai_inv - tsum*rai_inv2;
897 born->bRad[i] = 1.0 / born->bRad[i];
899 fr->invsqrta[i] = gmx_invsqrt(born->bRad[i]);
901 tchain = rai * (born->obc_alpha-2*born->obc_beta*sum_ai+3*born->obc_gamma*sum_ai2);
902 born->drobc[i] = (1.0-tsum*tsum)*tchain*rai_inv2;
906 /* Extra (local) communication required for DD */
907 if (DOMAINDECOMP(cr))
909 dd_atom_spread_real(cr->dd, born->bRad);
910 dd_atom_spread_real(cr->dd, fr->invsqrta);
911 dd_atom_spread_real(cr->dd, born->drobc);
920 int calc_gb_rad(t_commrec *cr, t_forcerec *fr, t_inputrec *ir, gmx_localtop_t *top,
921 rvec x[], t_nblist *nl, gmx_genborn_t *born, t_mdatoms *md, t_nrnb *nrnb)
926 if (fr->bAllvsAll && fr->dadx == NULL)
928 /* We might need up to 8 atoms of padding before and after,
929 * and another 4 units to guarantee SSE alignment.
931 fr->nalloc_dadx = 2*(md->homenr+12)*(md->nr/2+1+12);
932 snew(fr->dadx_rawptr, fr->nalloc_dadx);
933 fr->dadx = (real *) (((size_t) fr->dadx_rawptr + 16) & (~((size_t) 15)));
937 /* In the SSE-enabled gb-loops, when writing to dadx, we
938 * always write 2*4 elements at a time, even in the case with only
939 * 1-3 j particles, where we only really need to write 2*(1-3)
940 * elements. This is because we want dadx to be aligned to a 16-
941 * byte boundary, and being able to use _mm_store/load_ps
943 ndadx = 2 * (nl->nrj + 3*nl->nri);
945 /* First, reallocate the dadx array, we need 3 extra for SSE */
946 if (ndadx + 3 > fr->nalloc_dadx)
948 fr->nalloc_dadx = over_alloc_large(ndadx) + 3;
949 srenew(fr->dadx_rawptr, fr->nalloc_dadx);
950 fr->dadx = (real *) (((size_t) fr->dadx_rawptr + 16) & (~((size_t) 15)));
956 cnt = md->homenr*(md->nr/2+1);
958 if (ir->gb_algorithm == egbSTILL)
960 genborn_allvsall_calc_still_radii(fr, md, born, top, x[0], &fr->AllvsAll_workgb);
961 /* 13 flops in outer loop, 47 flops in inner loop */
962 inc_nrnb(nrnb, eNR_BORN_AVA_RADII_STILL, md->homenr*13+cnt*47);
964 else if (ir->gb_algorithm == egbHCT || ir->gb_algorithm == egbOBC)
966 genborn_allvsall_calc_hct_obc_radii(fr, md, born, ir->gb_algorithm, top, x[0], &fr->AllvsAll_workgb);
967 /* 24 flops in outer loop, 183 in inner */
968 inc_nrnb(nrnb, eNR_BORN_AVA_RADII_HCT_OBC, md->homenr*24+cnt*183);
972 gmx_fatal(FARGS, "Bad gb algorithm for all-vs-all interactions");
977 /* Switch for determining which algorithm to use for Born radii calculation */
980 switch (ir->gb_algorithm)
983 calc_gb_rad_still(cr, fr, top, x, nl, born, md);
986 calc_gb_rad_hct(cr, fr, top, x, nl, born, md);
989 calc_gb_rad_obc(cr, fr, top, x, nl, born, md);
993 gmx_fatal(FARGS, "Unknown double precision algorithm for Born radii calculation: %d", ir->gb_algorithm);
998 switch (ir->gb_algorithm)
1001 calc_gb_rad_still(cr, fr, top, x, nl, born, md);
1004 calc_gb_rad_hct(cr, fr, top, x, nl, born, md);
1007 calc_gb_rad_obc(cr, fr, top, x, nl, born, md);
1011 gmx_fatal(FARGS, "Unknown algorithm for Born radii calculation: %d", ir->gb_algorithm);
1014 #endif /* Double or single precision */
1016 if (fr->bAllvsAll == FALSE)
1018 switch (ir->gb_algorithm)
1021 /* 17 flops per outer loop iteration, 47 flops per inner loop */
1022 inc_nrnb(nrnb, eNR_BORN_RADII_STILL, nl->nri*17+nl->nrj*47);
1026 /* 61 (assuming 10 for tanh) flops for outer loop iteration, 183 flops per inner loop */
1027 inc_nrnb(nrnb, eNR_BORN_RADII_HCT_OBC, nl->nri*61+nl->nrj*183);
1040 real gb_bonds_tab(rvec x[], rvec f[], rvec fshift[], real *charge, real *p_gbtabscale,
1041 real *invsqrta, real *dvda, real *GBtab, t_idef *idef, real epsilon_r,
1042 real gb_epsilon_solvent, real facel, const t_pbc *pbc, const t_graph *graph)
1044 int i, j, n0, m, nnn, ai, aj;
1050 real isaprod, qq, gbscale, gbtabscale, Y, F, Geps, Heps2, Fp, VV, FF, rt, eps, eps2;
1051 real vgb, fgb, fijC, dvdatmp, fscal;
1057 t_iatom *forceatoms;
1059 /* Scale the electrostatics by gb_epsilon_solvent */
1060 facel = facel * ((1.0/epsilon_r) - 1.0/gb_epsilon_solvent);
1062 gbtabscale = *p_gbtabscale;
1065 for (j = F_GB12; j <= F_GB14; j++)
1067 forceatoms = idef->il[j].iatoms;
1069 for (i = 0; i < idef->il[j].nr; )
1071 /* To avoid reading in the interaction type, we just increment i to pass over
1072 * the types in the forceatoms array, this saves some memory accesses
1075 ai = forceatoms[i++];
1076 aj = forceatoms[i++];
1078 ki = pbc_rvec_sub(pbc, x[ai], x[aj], dx);
1079 rsq11 = iprod(dx, dx);
1081 isai = invsqrta[ai];
1082 iq = (-1)*facel*charge[ai];
1084 rinv11 = gmx_invsqrt(rsq11);
1085 isaj = invsqrta[aj];
1086 isaprod = isai*isaj;
1087 qq = isaprod*iq*charge[aj];
1088 gbscale = isaprod*gbtabscale;
1091 n0 = static_cast<int>(rt);
1097 Geps = eps*GBtab[nnn+2];
1098 Heps2 = eps2*GBtab[nnn+3];
1101 FF = Fp+Geps+2.0*Heps2;
1103 fijC = qq*FF*gbscale;
1104 dvdatmp = -(vgb+fijC*r)*0.5;
1105 dvda[aj] = dvda[aj] + dvdatmp*isaj*isaj;
1106 dvda[ai] = dvda[ai] + dvdatmp*isai*isai;
1107 vctot = vctot + vgb;
1108 fgb = -(fijC)*rinv11;
1112 ivec_sub(SHIFT_IVEC(graph, ai), SHIFT_IVEC(graph, aj), dt);
1116 for (m = 0; (m < DIM); m++) /* 15 */
1121 fshift[ki][m] += fscal;
1122 fshift[CENTRAL][m] -= fscal;
1130 real calc_gb_selfcorrections(t_commrec *cr, int natoms,
1131 real *charge, gmx_genborn_t *born, real *dvda, double facel)
1133 int i, ai, at0, at1;
1134 real rai, e, derb, q, q2, fi, rai_inv, vtot;
1136 if (DOMAINDECOMP(cr))
1139 at1 = cr->dd->nat_home;
1148 /* Scale the electrostatics by gb_epsilon_solvent */
1149 facel = facel * ((1.0/born->epsilon_r) - 1.0/born->gb_epsilon_solvent);
1153 /* Apply self corrections */
1154 for (i = at0; i < at1; i++)
1158 if (born->use[ai] == 1)
1160 rai = born->bRad[ai];
1166 derb = 0.5*e*rai_inv*rai_inv;
1167 dvda[ai] += derb*rai;
1176 real calc_gb_nonpolar(t_commrec *cr, t_forcerec *fr, int natoms, gmx_genborn_t *born, gmx_localtop_t *top,
1177 real *dvda, t_mdatoms *md)
1179 int ai, i, at0, at1;
1180 real e, es, rai, term, probe, tmp, factor;
1181 real rbi_inv, rbi_inv2;
1183 if (DOMAINDECOMP(cr))
1186 at1 = cr->dd->nat_home;
1194 /* factor is the surface tension */
1195 factor = born->sa_surface_tension;
1201 for (i = at0; i < at1; i++)
1205 if (born->use[ai] == 1)
1207 rai = top->atomtypes.gb_radius[md->typeA[ai]];
1208 rbi_inv = fr->invsqrta[ai];
1209 rbi_inv2 = rbi_inv * rbi_inv;
1210 tmp = (rai*rbi_inv2)*(rai*rbi_inv2);
1212 e = factor*term*(rai+probe)*(rai+probe)*tmp;
1213 dvda[ai] = dvda[ai] - 6*e*rbi_inv2;
1223 real calc_gb_chainrule(int natoms, t_nblist *nl, real *dadx, real *dvda, rvec x[], rvec t[], rvec fshift[],
1224 rvec shift_vec[], int gb_algorithm, gmx_genborn_t *born)
1226 int i, k, n, ai, aj, nj0, nj1, n0, n1;
1229 real fgb, rbi, fix1, fiy1, fiz1;
1230 real ix1, iy1, iz1, jx1, jy1, jz1, dx11, dy11, dz11;
1231 real tx, ty, tz, rbai, rbaj, fgb_ai;
1240 if (gb_algorithm == egbSTILL)
1242 for (i = n0; i < n1; i++)
1244 rbi = born->bRad[i];
1245 rb[i] = (2 * rbi * rbi * dvda[i])/ONE_4PI_EPS0;
1248 else if (gb_algorithm == egbHCT)
1250 for (i = n0; i < n1; i++)
1252 rbi = born->bRad[i];
1253 rb[i] = rbi * rbi * dvda[i];
1256 else if (gb_algorithm == egbOBC)
1258 for (i = n0; i < n1; i++)
1260 rbi = born->bRad[i];
1261 rb[i] = rbi * rbi * born->drobc[i] * dvda[i];
1265 for (i = 0; i < nl->nri; i++)
1269 nj0 = nl->jindex[i];
1270 nj1 = nl->jindex[i+1];
1272 /* Load shifts for this list */
1273 shift = nl->shift[i];
1274 shX = shift_vec[shift][0];
1275 shY = shift_vec[shift][1];
1276 shZ = shift_vec[shift][2];
1278 /* Load atom i coordinates, add shift vectors */
1279 ix1 = shX + x[ai][0];
1280 iy1 = shY + x[ai][1];
1281 iz1 = shZ + x[ai][2];
1289 for (k = nj0; k < nj1 && nl->jjnr[k] >= 0; k++)
1303 fgb = rbai*dadx[n++];
1304 fgb_ai = rbaj*dadx[n++];
1306 /* Total force between ai and aj is the sum of ai->aj and aj->ai */
1317 /* Update force on atom aj */
1318 t[aj][0] = t[aj][0] - tx;
1319 t[aj][1] = t[aj][1] - ty;
1320 t[aj][2] = t[aj][2] - tz;
1323 /* Update force and shift forces on atom ai */
1324 t[ai][0] = t[ai][0] + fix1;
1325 t[ai][1] = t[ai][1] + fiy1;
1326 t[ai][2] = t[ai][2] + fiz1;
1328 fshift[shift][0] = fshift[shift][0] + fix1;
1329 fshift[shift][1] = fshift[shift][1] + fiy1;
1330 fshift[shift][2] = fshift[shift][2] + fiz1;
1339 calc_gb_forces(t_commrec *cr, t_mdatoms *md, gmx_genborn_t *born, gmx_localtop_t *top,
1340 rvec x[], rvec f[], t_forcerec *fr, t_idef *idef, int gb_algorithm, int sa_algorithm, t_nrnb *nrnb,
1341 const t_pbc *pbc, const t_graph *graph, gmx_enerdata_t *enerd)
1346 const t_pbc *pbc_null;
1357 if (sa_algorithm == esaAPPROX)
1359 /* Do a simple ACE type approximation for the non-polar solvation */
1360 enerd->term[F_NPSOLVATION] += calc_gb_nonpolar(cr, fr, born->nr, born, top, fr->dvda, md);
1363 /* Calculate the bonded GB-interactions using either table or analytical formula */
1364 enerd->term[F_GBPOL] += gb_bonds_tab(x, f, fr->fshift, md->chargeA, &(fr->gbtabscale),
1365 fr->invsqrta, fr->dvda, fr->gbtab.data, idef, born->epsilon_r, born->gb_epsilon_solvent, fr->epsfac, pbc_null, graph);
1367 /* Calculate self corrections to the GB energies - currently only A state used! (FIXME) */
1368 enerd->term[F_GBPOL] += calc_gb_selfcorrections(cr, born->nr, md->chargeA, born, fr->dvda, fr->epsfac);
1370 /* If parallel, sum the derivative of the potential w.r.t the born radii */
1371 if (DOMAINDECOMP(cr))
1373 dd_atom_sum_real(cr->dd, fr->dvda);
1374 dd_atom_spread_real(cr->dd, fr->dvda);
1379 genborn_allvsall_calc_chainrule(fr, md, born, x[0], f[0], gb_algorithm, fr->AllvsAll_workgb);
1380 cnt = md->homenr*(md->nr/2+1);
1381 /* 9 flops for outer loop, 15 for inner */
1382 inc_nrnb(nrnb, eNR_BORN_AVA_CHAINRULE, md->homenr*9+cnt*15);
1386 calc_gb_chainrule(fr->natoms_force, &(fr->gblist), fr->dadx, fr->dvda,
1387 x, f, fr->fshift, fr->shift_vec, gb_algorithm, born);
1391 /* 9 flops for outer loop, 15 for inner */
1392 inc_nrnb(nrnb, eNR_BORN_CHAINRULE, fr->gblist.nri*9+fr->gblist.nrj*15);
1396 static void add_j_to_gblist(gbtmpnbl_t *list, int aj)
1398 if (list->naj >= list->aj_nalloc)
1400 list->aj_nalloc = over_alloc_large(list->naj+1);
1401 srenew(list->aj, list->aj_nalloc);
1404 list->aj[list->naj++] = aj;
1407 static gbtmpnbl_t *find_gbtmplist(struct gbtmpnbls *lists, int shift)
1411 /* Search the list with the same shift, if there is one */
1413 while (ind < lists->nlist && shift != lists->list[ind].shift)
1417 if (ind == lists->nlist)
1419 if (lists->nlist == lists->list_nalloc)
1421 lists->list_nalloc++;
1422 srenew(lists->list, lists->list_nalloc);
1423 for (i = lists->nlist; i < lists->list_nalloc; i++)
1425 lists->list[i].aj = NULL;
1426 lists->list[i].aj_nalloc = 0;
1431 lists->list[lists->nlist].shift = shift;
1432 lists->list[lists->nlist].naj = 0;
1436 return &lists->list[ind];
1439 static void add_bondeds_to_gblist(t_ilist *il,
1440 gmx_bool bMolPBC, t_pbc *pbc, t_graph *g, rvec *x,
1441 struct gbtmpnbls *nls)
1443 int ind, j, ai, aj, found;
1448 for (ind = 0; ind < il->nr; ind += 3)
1450 ai = il->iatoms[ind+1];
1451 aj = il->iatoms[ind+2];
1453 int shift = CENTRAL;
1456 rvec_sub(x[ai], x[aj], dx);
1457 ivec_sub(SHIFT_IVEC(g, ai), SHIFT_IVEC(g, aj), dt);
1458 shift = IVEC2IS(dt);
1462 shift = pbc_dx_aiuc(pbc, x[ai], x[aj], dx);
1465 /* Find the list for this shift or create one */
1466 list = find_gbtmplist(&nls[ai], shift);
1470 /* So that we do not add the same bond twice.
1471 * This happens with some constraints between 1-3 atoms
1472 * that are in the bond-list but should not be in the GB nb-list */
1473 for (j = 0; j < list->naj; j++)
1475 if (list->aj[j] == aj)
1485 gmx_incons("ai == aj");
1488 add_j_to_gblist(list, aj);
1494 compare_int (const void * a, const void * b)
1496 return ( *(int*)a - *(int*)b );
1501 int make_gb_nblist(t_commrec *cr, int gb_algorithm,
1502 rvec x[], matrix box,
1503 t_forcerec *fr, t_idef *idef, t_graph *graph, gmx_genborn_t *born)
1505 int i, j, k, n, nj0, nj1, ai, shift, s;
1509 struct gbtmpnbls *nls;
1510 gbtmpnbl_t *list = NULL;
1512 set_pbc(&pbc, fr->ePBC, box);
1513 nls = born->nblist_work;
1515 for (i = 0; i < born->nr; i++)
1522 set_pbc_dd(&pbc, fr->ePBC, cr->dd, TRUE, box);
1525 switch (gb_algorithm)
1529 /* Loop over 1-2, 1-3 and 1-4 interactions */
1530 for (j = F_GB12; j <= F_GB14; j++)
1532 add_bondeds_to_gblist(&idef->il[j], fr->bMolPBC, &pbc, graph, x, nls);
1536 /* Loop over 1-4 interactions */
1537 add_bondeds_to_gblist(&idef->il[F_GB14], fr->bMolPBC, &pbc, graph, x, nls);
1540 gmx_incons("Unknown GB algorithm");
1543 /* Loop over the VDWQQ and VDW nblists to set up the nonbonded part of the GB list */
1544 for (n = 0; (n < fr->nnblists); n++)
1546 for (i = 0; (i < eNL_NR); i++)
1548 nblist = &(fr->nblists[n].nlist_sr[i]);
1550 if (nblist->nri > 0 && (i == eNL_VDWQQ || i == eNL_QQ))
1552 for (j = 0; j < nblist->nri; j++)
1554 ai = nblist->iinr[j];
1555 shift = nblist->shift[j];
1557 /* Find the list for this shift or create one */
1558 list = find_gbtmplist(&nls[ai], shift);
1560 nj0 = nblist->jindex[j];
1561 nj1 = nblist->jindex[j+1];
1563 /* Add all the j-atoms in the non-bonded list to the GB list */
1564 for (k = nj0; k < nj1; k++)
1566 add_j_to_gblist(list, nblist->jjnr[k]);
1573 /* Zero out some counters */
1577 fr->gblist.jindex[0] = fr->gblist.nri;
1579 for (i = 0; i < fr->natoms_force; i++)
1581 for (s = 0; s < nls[i].nlist; s++)
1583 list = &nls[i].list[s];
1585 /* Only add those atoms that actually have neighbours */
1586 if (born->use[i] != 0)
1588 fr->gblist.iinr[fr->gblist.nri] = i;
1589 fr->gblist.shift[fr->gblist.nri] = list->shift;
1592 for (k = 0; k < list->naj; k++)
1594 /* Memory allocation for jjnr */
1595 if (fr->gblist.nrj >= fr->gblist.maxnrj)
1597 fr->gblist.maxnrj += over_alloc_large(fr->gblist.maxnrj);
1601 fprintf(debug, "Increasing GB neighbourlist j size to %d\n", fr->gblist.maxnrj);
1604 srenew(fr->gblist.jjnr, fr->gblist.maxnrj);
1608 if (i == list->aj[k])
1610 gmx_incons("i == list->aj[k]");
1612 fr->gblist.jjnr[fr->gblist.nrj++] = list->aj[k];
1615 fr->gblist.jindex[fr->gblist.nri] = fr->gblist.nrj;
1622 for (i = 0; i < fr->gblist.nri; i++)
1624 nj0 = fr->gblist.jindex[i];
1625 nj1 = fr->gblist.jindex[i+1];
1626 ai = fr->gblist.iinr[i];
1629 for (j = nj0; j < nj1; j++)
1631 if (fr->gblist.jjnr[j] < ai)
1633 fr->gblist.jjnr[j] += fr->natoms_force;
1636 qsort(fr->gblist.jjnr+nj0, nj1-nj0, sizeof(int), compare_int);
1638 for (j = nj0; j < nj1; j++)
1640 if (fr->gblist.jjnr[j] >= fr->natoms_force)
1642 fr->gblist.jjnr[j] -= fr->natoms_force;
1652 void make_local_gb(const t_commrec *cr, gmx_genborn_t *born, int gb_algorithm)
1655 gmx_domdec_t *dd = NULL;
1657 if (DOMAINDECOMP(cr))
1665 /* Single node, just copy pointers and return */
1666 if (gb_algorithm == egbSTILL)
1668 born->gpol = born->gpol_globalindex;
1669 born->vsolv = born->vsolv_globalindex;
1670 born->gb_radius = born->gb_radius_globalindex;
1674 born->param = born->param_globalindex;
1675 born->gb_radius = born->gb_radius_globalindex;
1678 born->use = born->use_globalindex;
1683 /* Reallocation of local arrays if necessary */
1684 /* fr->natoms_force is equal to dd->nat_tot */
1685 if (DOMAINDECOMP(cr) && dd->nat_tot > born->nalloc)
1689 nalloc = dd->nat_tot;
1691 /* Arrays specific to different gb algorithms */
1692 if (gb_algorithm == egbSTILL)
1694 srenew(born->gpol, nalloc+3);
1695 srenew(born->vsolv, nalloc+3);
1696 srenew(born->gb_radius, nalloc+3);
1697 for (i = born->nalloc; (i < nalloc+3); i++)
1701 born->gb_radius[i] = 0;
1706 srenew(born->param, nalloc+3);
1707 srenew(born->gb_radius, nalloc+3);
1708 for (i = born->nalloc; (i < nalloc+3); i++)
1711 born->gb_radius[i] = 0;
1715 /* All gb-algorithms use the array for vsites exclusions */
1716 srenew(born->use, nalloc+3);
1717 for (i = born->nalloc; (i < nalloc+3); i++)
1722 born->nalloc = nalloc;
1725 /* With dd, copy algorithm specific arrays */
1726 if (gb_algorithm == egbSTILL)
1728 for (i = at0; i < at1; i++)
1730 born->gpol[i] = born->gpol_globalindex[dd->gatindex[i]];
1731 born->vsolv[i] = born->vsolv_globalindex[dd->gatindex[i]];
1732 born->gb_radius[i] = born->gb_radius_globalindex[dd->gatindex[i]];
1733 born->use[i] = born->use_globalindex[dd->gatindex[i]];
1738 for (i = at0; i < at1; i++)
1740 born->param[i] = born->param_globalindex[dd->gatindex[i]];
1741 born->gb_radius[i] = born->gb_radius_globalindex[dd->gatindex[i]];
1742 born->use[i] = born->use_globalindex[dd->gatindex[i]];