1 /* -*- mode: c; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4; c-file-style: "stroustrup"; -*-
4 * This source code is part of
8 * GROningen MAchine for Chemical Simulations
10 * Written by David van der Spoel, Erik Lindahl, Berk Hess, and others.
11 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
12 * Copyright (c) 2001-2008, The GROMACS development team,
13 * check out http://www.gromacs.org for more information.
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version 2
18 * of the License, or (at your option) any later version.
20 * If you want to redistribute modifications, please consider that
21 * scientific software is very special. Version control is crucial -
22 * bugs must be traceable. We will be happy to consider code for
23 * inclusion in the official distribution, but derived work must not
24 * be called official GROMACS. Details are found in the README & COPYING
25 * files - if they are missing, get the official version at www.gromacs.org.
27 * To help us fund GROMACS development, we humbly ask that you cite
28 * the papers on the package - you can find them in the top README file.
30 * For more info, check our website at http://www.gromacs.org
33 * Gallium Rubidium Oxygen Manganese Argon Carbon Silicon
44 #include "mtop_util.h"
45 #include "types/commrec.h"
46 #include "nbnxn_search.h"
47 #include "nbnxn_consts.h"
50 /* Computational cost of bonded, non-bonded and PME calculations.
51 * This will be machine dependent.
52 * The numbers here are accurate for Intel Core2 and AMD Athlon 64
53 * in single precision. In double precision PME mesh is slightly cheaper,
54 * although not so much that the numbers need to be adjusted.
57 /* Cost of a pair interaction in the "group" cut-off scheme" */
59 #define C_GR_QLJ_CUT 1.5
60 #define C_GR_QLJ_TAB 2.0
61 #define C_GR_LJ_CUT 1.0
62 #define C_GR_LJ_TAB 1.75
63 /* Cost of 1 water with one Q/LJ atom */
64 #define C_GR_QLJW_CUT 2.0
65 #define C_GR_QLJW_TAB 2.25
66 /* Cost of 1 water with one Q atom or with 1/3 water (LJ negligible) */
69 /* Cost of a pair interaction in the "Verlet" cut-off scheme" */
71 #define C_VT_QLJ_RF 0.40
72 #define C_VT_Q_RF 0.30
73 #define C_VT_QLJ_TAB 0.55
74 #define C_VT_Q_TAB 0.50
76 /* Cost of PME, with all components running with SSE instructions */
77 /* Cost of particle reordering and redistribution */
78 #define C_PME_REDIST 12.0
79 /* Cost of q spreading and force interpolation per charge (mainly memory) */
80 #define C_PME_SPREAD 0.30
81 /* Cost of fft's, will be multiplied with N log(N) */
82 #define C_PME_FFT 0.20
83 /* Cost of pme_solve, will be multiplied with N */
84 #define C_PME_SOLVE 0.50
86 /* Cost of a bonded interaction divided by the number of (pbc_)dx nrequired */
89 int n_bonded_dx(gmx_mtop_t *mtop,gmx_bool bExcl)
91 int mb,nmol,ftype,ndxb,ndx_excl;
95 /* Count the number of pbc_rvec_sub calls required for bonded interactions.
96 * This number is also roughly proportional to the computational cost.
100 for(mb=0; mb<mtop->nmolblock; mb++) {
101 molt = &mtop->moltype[mtop->molblock[mb].type];
102 nmol = mtop->molblock[mb].nmol;
103 for(ftype=0; ftype<F_NRE; ftype++) {
104 if (interaction_function[ftype].flags & IF_BOND) {
106 case F_POSRES: ndxb = 1; break;
107 case F_CONNBONDS: ndxb = 0; break;
108 default: ndxb = NRAL(ftype) - 1; break;
110 ndx += nmol*ndxb*molt->ilist[ftype].nr/(1 + NRAL(ftype));
114 ndx_excl += nmol*(molt->excls.nra - molt->atoms.nr)/2;
121 fprintf(debug,"ndx bonded %d exclusions %d\n",ndx,ndx_excl);
128 static void pp_group_load(gmx_mtop_t *mtop,t_inputrec *ir,matrix box,
131 gmx_bool *bChargePerturbed)
134 int mb,nmol,atnr,cg,a,a0,ncqlj,ncq,nclj;
135 gmx_bool bBHAM,bLJcut,bWater,bQ,bLJ;
137 float fq,fqlj,flj,fljtab,fqljw,fqw;
141 bBHAM = (mtop->ffparams.functype[0] == F_BHAM);
143 bLJcut = ((ir->vdwtype == evdwCUT) && !bBHAM);
145 /* Computational cost of bonded, non-bonded and PME calculations.
146 * This will be machine dependent.
147 * The numbers here are accurate for Intel Core2 and AMD Athlon 64
148 * in single precision. In double precision PME mesh is slightly cheaper,
149 * although not so much that the numbers need to be adjusted.
152 fqlj = (bLJcut ? C_GR_QLJ_CUT : C_GR_QLJ_TAB);
153 flj = (bLJcut ? C_GR_LJ_CUT : C_GR_LJ_TAB);
154 /* Cost of 1 water with one Q/LJ atom */
155 fqljw = (bLJcut ? C_GR_QLJW_CUT : C_GR_QLJW_TAB);
156 /* Cost of 1 water with one Q atom or with 1/3 water (LJ negligible) */
159 iparams = mtop->ffparams.iparams;
160 atnr = mtop->ffparams.atnr;
165 *bChargePerturbed = FALSE;
166 for(mb=0; mb<mtop->nmolblock; mb++)
168 molt = &mtop->moltype[mtop->molblock[mb].type];
169 atom = molt->atoms.atom;
170 nmol = mtop->molblock[mb].nmol;
172 for(cg=0; cg<molt->cgs.nr; cg++)
179 while (a < molt->cgs.index[cg+1])
181 bQ = (atom[a].q != 0 || atom[a].qB != 0);
182 bLJ = (iparams[(atnr+1)*atom[a].type].lj.c6 != 0 ||
183 iparams[(atnr+1)*atom[a].type].lj.c12 != 0);
184 if (atom[a].q != atom[a].qB)
186 *bChargePerturbed = TRUE;
188 /* This if this atom fits into water optimization */
189 if (!((a == a0 && bQ && bLJ) ||
190 (a == a0+1 && bQ && !bLJ) ||
191 (a == a0+2 && bQ && !bLJ && atom[a].q == atom[a-1].q) ||
192 (a == a0+3 && !bQ && bLJ)))
224 *nq_tot = nq + nqlj + nw*3;
228 fprintf(debug,"nw %d nqlj %d nq %d nlj %d\n",nw,nqlj,nq,nlj);
231 /* For the PP non-bonded cost it is (unrealistically) assumed
232 * that all atoms are distributed homogeneously in space.
233 * Factor 3 is used because a water molecule has 3 atoms
234 * (and TIP4P effectively has 3 interactions with (water) atoms)).
236 *cost_pp = 0.5*(fqljw*nw*nqlj +
237 fqw *nw*(3*nw + nq) +
239 fq *nq*(3*nw + nqlj + nq) +
240 flj *nlj*(nw + nqlj + nlj))
241 *4/3*M_PI*ir->rlist*ir->rlist*ir->rlist/det(box);
244 static void pp_verlet_load(gmx_mtop_t *mtop,t_inputrec *ir,matrix box,
247 gmx_bool *bChargePerturbed)
250 int mb,nmol,atnr,cg,a,a0,nqlj,nq,nlj;
257 bQRF = (EEL_RF(ir->coulombtype) || ir->coulombtype == eelCUT);
259 iparams = mtop->ffparams.iparams;
260 atnr = mtop->ffparams.atnr;
263 *bChargePerturbed = FALSE;
264 for(mb=0; mb<mtop->nmolblock; mb++)
266 molt = &mtop->moltype[mtop->molblock[mb].type];
267 atom = molt->atoms.atom;
268 nmol = mtop->molblock[mb].nmol;
270 for(a=0; a<molt->atoms.nr; a++)
272 if (atom[a].q != 0 || atom[a].qB != 0)
274 if (iparams[(atnr+1)*atom[a].type].lj.c6 != 0 ||
275 iparams[(atnr+1)*atom[a].type].lj.c12 != 0)
284 if (atom[a].q != atom[a].qB)
286 *bChargePerturbed = TRUE;
291 nlj = mtop->natoms - nqlj - nq;
295 /* Effective cut-off for cluster pair list of 4x4 atoms */
296 r_eff = ir->rlist + nbnxn_get_rlist_effective_inc(NBNXN_CPU_CLUSTER_I_SIZE,mtop->natoms/det(box));
300 fprintf(debug,"nqlj %d nq %d nlj %d rlist %.3f r_eff %.3f\n",
301 nqlj,nq,nlj,ir->rlist,r_eff);
304 /* For the PP non-bonded cost it is (unrealistically) assumed
305 * that all atoms are distributed homogeneously in space.
307 /* Convert mtop->natoms to double to avoid int overflow */
309 *cost_pp = 0.5*(nqlj*nat*(bQRF ? C_VT_QLJ_RF : C_VT_QLJ_TAB) +
310 nq*nat*(bQRF ? C_VT_Q_RF : C_VT_Q_TAB) +
312 *4/3*M_PI*r_eff*r_eff*r_eff/det(box);
315 float pme_load_estimate(gmx_mtop_t *mtop,t_inputrec *ir,matrix box)
318 int mb,nmol,atnr,cg,a,a0,nq_tot;
319 gmx_bool bBHAM,bLJcut,bChargePerturbed,bWater,bQ,bLJ;
320 double cost_bond,cost_pp,cost_redist,cost_spread,cost_fft,cost_solve,cost_pme;
325 /* Computational cost of bonded, non-bonded and PME calculations.
326 * This will be machine dependent.
327 * The numbers here are accurate for Intel Core2 and AMD Athlon 64
328 * in single precision. In double precision PME mesh is slightly cheaper,
329 * although not so much that the numbers need to be adjusted.
332 iparams = mtop->ffparams.iparams;
333 atnr = mtop->ffparams.atnr;
335 cost_bond = C_BOND*n_bonded_dx(mtop,TRUE);
337 if (ir->cutoff_scheme == ecutsGROUP)
339 pp_group_load(mtop,ir,box,&nq_tot,&cost_pp,&bChargePerturbed);
343 pp_verlet_load(mtop,ir,box,&nq_tot,&cost_pp,&bChargePerturbed);
346 cost_redist = C_PME_REDIST*nq_tot;
347 cost_spread = C_PME_SPREAD*nq_tot*pow(ir->pme_order,3);
348 cost_fft = C_PME_FFT*ir->nkx*ir->nky*ir->nkz*log(ir->nkx*ir->nky*ir->nkz);
349 cost_solve = C_PME_SOLVE*ir->nkx*ir->nky*ir->nkz;
351 if (ir->efep != efepNO && bChargePerturbed) {
352 /* All PME work, except redist & spline coefficient calculation, doubles */
358 cost_pme = cost_redist + cost_spread + cost_fft + cost_solve;
360 ratio = cost_pme/(cost_bond + cost_pp + cost_pme);
370 cost_bond,cost_pp,cost_redist,cost_spread,cost_fft,cost_solve);
372 fprintf(debug,"Estimate for relative PME load: %.3f\n",ratio);