Fix segmentation fault in minimize
[gromacs.git] / src / gromacs / gmxlib / nonbonded / nb_kernel_c / nb_kernel_ElecEw_VdwCSTab_GeomW4W4_c.cpp
blob9eb413e9134b38176f4a91c195c636efbcce6f39
1 /*
2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014.2015,2017,2018, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS c kernel generator.
38 #include "gmxpre.h"
40 #include "config.h"
42 #include <math.h>
44 #include "../nb_kernel.h"
45 #include "gromacs/gmxlib/nrnb.h"
48 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_c
49 * Electrostatics interaction: Ewald
50 * VdW interaction: CubicSplineTable
51 * Geometry: Water4-Water4
52 * Calculate force/pot: PotentialAndForce
54 void
55 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_c
56 (t_nblist * gmx_restrict nlist,
57 rvec * gmx_restrict xx,
58 rvec * gmx_restrict ff,
59 struct t_forcerec * gmx_restrict fr,
60 t_mdatoms * gmx_restrict mdatoms,
61 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
62 t_nrnb * gmx_restrict nrnb)
64 int i_shift_offset,i_coord_offset,j_coord_offset;
65 int j_index_start,j_index_end;
66 int nri,inr,ggid,iidx,jidx,jnr,outeriter,inneriter;
67 real shX,shY,shZ,tx,ty,tz,fscal,rcutoff,rcutoff2;
68 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
69 real *shiftvec,*fshift,*x,*f;
70 int vdwioffset0;
71 real ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
72 int vdwioffset1;
73 real ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
74 int vdwioffset2;
75 real ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
76 int vdwioffset3;
77 real ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
78 int vdwjidx0;
79 real jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
80 int vdwjidx1;
81 real jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
82 int vdwjidx2;
83 real jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
84 int vdwjidx3;
85 real jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
86 real dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00,cexp1_00,cexp2_00;
87 real dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11,cexp1_11,cexp2_11;
88 real dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12,cexp1_12,cexp2_12;
89 real dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13,cexp1_13,cexp2_13;
90 real dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21,cexp1_21,cexp2_21;
91 real dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22,cexp1_22,cexp2_22;
92 real dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23,cexp1_23,cexp2_23;
93 real dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31,cexp1_31,cexp2_31;
94 real dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32,cexp1_32,cexp2_32;
95 real dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33,cexp1_33,cexp2_33;
96 real velec,felec,velecsum,facel,crf,krf,krf2;
97 real *charge;
98 int nvdwtype;
99 real rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,br,vvdwexp,sh_vdw_invrcut6;
100 int *vdwtype;
101 real *vdwparam;
102 int vfitab;
103 real rt,vfeps,vftabscale,Y,F,Geps,Heps2,Fp,VV,FF;
104 real *vftab;
105 int ewitab;
106 real ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace;
107 real *ewtab;
109 x = xx[0];
110 f = ff[0];
112 nri = nlist->nri;
113 iinr = nlist->iinr;
114 jindex = nlist->jindex;
115 jjnr = nlist->jjnr;
116 shiftidx = nlist->shift;
117 gid = nlist->gid;
118 shiftvec = fr->shift_vec[0];
119 fshift = fr->fshift[0];
120 facel = fr->ic->epsfac;
121 charge = mdatoms->chargeA;
122 nvdwtype = fr->ntype;
123 vdwparam = fr->nbfp;
124 vdwtype = mdatoms->typeA;
126 vftab = kernel_data->table_vdw->data;
127 vftabscale = kernel_data->table_vdw->scale;
129 sh_ewald = fr->ic->sh_ewald;
130 ewtab = fr->ic->tabq_coul_FDV0;
131 ewtabscale = fr->ic->tabq_scale;
132 ewtabhalfspace = 0.5/ewtabscale;
134 /* Setup water-specific parameters */
135 inr = nlist->iinr[0];
136 iq1 = facel*charge[inr+1];
137 iq2 = facel*charge[inr+2];
138 iq3 = facel*charge[inr+3];
139 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
141 jq1 = charge[inr+1];
142 jq2 = charge[inr+2];
143 jq3 = charge[inr+3];
144 vdwjidx0 = 2*vdwtype[inr+0];
145 c6_00 = vdwparam[vdwioffset0+vdwjidx0];
146 c12_00 = vdwparam[vdwioffset0+vdwjidx0+1];
147 qq11 = iq1*jq1;
148 qq12 = iq1*jq2;
149 qq13 = iq1*jq3;
150 qq21 = iq2*jq1;
151 qq22 = iq2*jq2;
152 qq23 = iq2*jq3;
153 qq31 = iq3*jq1;
154 qq32 = iq3*jq2;
155 qq33 = iq3*jq3;
157 outeriter = 0;
158 inneriter = 0;
160 /* Start outer loop over neighborlists */
161 for(iidx=0; iidx<nri; iidx++)
163 /* Load shift vector for this list */
164 i_shift_offset = DIM*shiftidx[iidx];
165 shX = shiftvec[i_shift_offset+XX];
166 shY = shiftvec[i_shift_offset+YY];
167 shZ = shiftvec[i_shift_offset+ZZ];
169 /* Load limits for loop over neighbors */
170 j_index_start = jindex[iidx];
171 j_index_end = jindex[iidx+1];
173 /* Get outer coordinate index */
174 inr = iinr[iidx];
175 i_coord_offset = DIM*inr;
177 /* Load i particle coords and add shift vector */
178 ix0 = shX + x[i_coord_offset+DIM*0+XX];
179 iy0 = shY + x[i_coord_offset+DIM*0+YY];
180 iz0 = shZ + x[i_coord_offset+DIM*0+ZZ];
181 ix1 = shX + x[i_coord_offset+DIM*1+XX];
182 iy1 = shY + x[i_coord_offset+DIM*1+YY];
183 iz1 = shZ + x[i_coord_offset+DIM*1+ZZ];
184 ix2 = shX + x[i_coord_offset+DIM*2+XX];
185 iy2 = shY + x[i_coord_offset+DIM*2+YY];
186 iz2 = shZ + x[i_coord_offset+DIM*2+ZZ];
187 ix3 = shX + x[i_coord_offset+DIM*3+XX];
188 iy3 = shY + x[i_coord_offset+DIM*3+YY];
189 iz3 = shZ + x[i_coord_offset+DIM*3+ZZ];
191 fix0 = 0.0;
192 fiy0 = 0.0;
193 fiz0 = 0.0;
194 fix1 = 0.0;
195 fiy1 = 0.0;
196 fiz1 = 0.0;
197 fix2 = 0.0;
198 fiy2 = 0.0;
199 fiz2 = 0.0;
200 fix3 = 0.0;
201 fiy3 = 0.0;
202 fiz3 = 0.0;
204 /* Reset potential sums */
205 velecsum = 0.0;
206 vvdwsum = 0.0;
208 /* Start inner kernel loop */
209 for(jidx=j_index_start; jidx<j_index_end; jidx++)
211 /* Get j neighbor index, and coordinate index */
212 jnr = jjnr[jidx];
213 j_coord_offset = DIM*jnr;
215 /* load j atom coordinates */
216 jx0 = x[j_coord_offset+DIM*0+XX];
217 jy0 = x[j_coord_offset+DIM*0+YY];
218 jz0 = x[j_coord_offset+DIM*0+ZZ];
219 jx1 = x[j_coord_offset+DIM*1+XX];
220 jy1 = x[j_coord_offset+DIM*1+YY];
221 jz1 = x[j_coord_offset+DIM*1+ZZ];
222 jx2 = x[j_coord_offset+DIM*2+XX];
223 jy2 = x[j_coord_offset+DIM*2+YY];
224 jz2 = x[j_coord_offset+DIM*2+ZZ];
225 jx3 = x[j_coord_offset+DIM*3+XX];
226 jy3 = x[j_coord_offset+DIM*3+YY];
227 jz3 = x[j_coord_offset+DIM*3+ZZ];
229 /* Calculate displacement vector */
230 dx00 = ix0 - jx0;
231 dy00 = iy0 - jy0;
232 dz00 = iz0 - jz0;
233 dx11 = ix1 - jx1;
234 dy11 = iy1 - jy1;
235 dz11 = iz1 - jz1;
236 dx12 = ix1 - jx2;
237 dy12 = iy1 - jy2;
238 dz12 = iz1 - jz2;
239 dx13 = ix1 - jx3;
240 dy13 = iy1 - jy3;
241 dz13 = iz1 - jz3;
242 dx21 = ix2 - jx1;
243 dy21 = iy2 - jy1;
244 dz21 = iz2 - jz1;
245 dx22 = ix2 - jx2;
246 dy22 = iy2 - jy2;
247 dz22 = iz2 - jz2;
248 dx23 = ix2 - jx3;
249 dy23 = iy2 - jy3;
250 dz23 = iz2 - jz3;
251 dx31 = ix3 - jx1;
252 dy31 = iy3 - jy1;
253 dz31 = iz3 - jz1;
254 dx32 = ix3 - jx2;
255 dy32 = iy3 - jy2;
256 dz32 = iz3 - jz2;
257 dx33 = ix3 - jx3;
258 dy33 = iy3 - jy3;
259 dz33 = iz3 - jz3;
261 /* Calculate squared distance and things based on it */
262 rsq00 = dx00*dx00+dy00*dy00+dz00*dz00;
263 rsq11 = dx11*dx11+dy11*dy11+dz11*dz11;
264 rsq12 = dx12*dx12+dy12*dy12+dz12*dz12;
265 rsq13 = dx13*dx13+dy13*dy13+dz13*dz13;
266 rsq21 = dx21*dx21+dy21*dy21+dz21*dz21;
267 rsq22 = dx22*dx22+dy22*dy22+dz22*dz22;
268 rsq23 = dx23*dx23+dy23*dy23+dz23*dz23;
269 rsq31 = dx31*dx31+dy31*dy31+dz31*dz31;
270 rsq32 = dx32*dx32+dy32*dy32+dz32*dz32;
271 rsq33 = dx33*dx33+dy33*dy33+dz33*dz33;
273 rinv00 = 1.0/sqrt(rsq00);
274 rinv11 = 1.0/sqrt(rsq11);
275 rinv12 = 1.0/sqrt(rsq12);
276 rinv13 = 1.0/sqrt(rsq13);
277 rinv21 = 1.0/sqrt(rsq21);
278 rinv22 = 1.0/sqrt(rsq22);
279 rinv23 = 1.0/sqrt(rsq23);
280 rinv31 = 1.0/sqrt(rsq31);
281 rinv32 = 1.0/sqrt(rsq32);
282 rinv33 = 1.0/sqrt(rsq33);
284 rinvsq11 = rinv11*rinv11;
285 rinvsq12 = rinv12*rinv12;
286 rinvsq13 = rinv13*rinv13;
287 rinvsq21 = rinv21*rinv21;
288 rinvsq22 = rinv22*rinv22;
289 rinvsq23 = rinv23*rinv23;
290 rinvsq31 = rinv31*rinv31;
291 rinvsq32 = rinv32*rinv32;
292 rinvsq33 = rinv33*rinv33;
294 /**************************
295 * CALCULATE INTERACTIONS *
296 **************************/
298 r00 = rsq00*rinv00;
300 /* Calculate table index by multiplying r with table scale and truncate to integer */
301 rt = r00*vftabscale;
302 vfitab = rt;
303 vfeps = rt-vfitab;
304 vfitab = 2*4*vfitab;
306 /* CUBIC SPLINE TABLE DISPERSION */
307 vfitab += 0;
308 Y = vftab[vfitab];
309 F = vftab[vfitab+1];
310 Geps = vfeps*vftab[vfitab+2];
311 Heps2 = vfeps*vfeps*vftab[vfitab+3];
312 Fp = F+Geps+Heps2;
313 VV = Y+vfeps*Fp;
314 vvdw6 = c6_00*VV;
315 FF = Fp+Geps+2.0*Heps2;
316 fvdw6 = c6_00*FF;
318 /* CUBIC SPLINE TABLE REPULSION */
319 Y = vftab[vfitab+4];
320 F = vftab[vfitab+5];
321 Geps = vfeps*vftab[vfitab+6];
322 Heps2 = vfeps*vfeps*vftab[vfitab+7];
323 Fp = F+Geps+Heps2;
324 VV = Y+vfeps*Fp;
325 vvdw12 = c12_00*VV;
326 FF = Fp+Geps+2.0*Heps2;
327 fvdw12 = c12_00*FF;
328 vvdw = vvdw12+vvdw6;
329 fvdw = -(fvdw6+fvdw12)*vftabscale*rinv00;
331 /* Update potential sums from outer loop */
332 vvdwsum += vvdw;
334 fscal = fvdw;
336 /* Calculate temporary vectorial force */
337 tx = fscal*dx00;
338 ty = fscal*dy00;
339 tz = fscal*dz00;
341 /* Update vectorial force */
342 fix0 += tx;
343 fiy0 += ty;
344 fiz0 += tz;
345 f[j_coord_offset+DIM*0+XX] -= tx;
346 f[j_coord_offset+DIM*0+YY] -= ty;
347 f[j_coord_offset+DIM*0+ZZ] -= tz;
349 /**************************
350 * CALCULATE INTERACTIONS *
351 **************************/
353 r11 = rsq11*rinv11;
355 /* EWALD ELECTROSTATICS */
357 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
358 ewrt = r11*ewtabscale;
359 ewitab = ewrt;
360 eweps = ewrt-ewitab;
361 ewitab = 4*ewitab;
362 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
363 velec = qq11*(rinv11-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
364 felec = qq11*rinv11*(rinvsq11-felec);
366 /* Update potential sums from outer loop */
367 velecsum += velec;
369 fscal = felec;
371 /* Calculate temporary vectorial force */
372 tx = fscal*dx11;
373 ty = fscal*dy11;
374 tz = fscal*dz11;
376 /* Update vectorial force */
377 fix1 += tx;
378 fiy1 += ty;
379 fiz1 += tz;
380 f[j_coord_offset+DIM*1+XX] -= tx;
381 f[j_coord_offset+DIM*1+YY] -= ty;
382 f[j_coord_offset+DIM*1+ZZ] -= tz;
384 /**************************
385 * CALCULATE INTERACTIONS *
386 **************************/
388 r12 = rsq12*rinv12;
390 /* EWALD ELECTROSTATICS */
392 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
393 ewrt = r12*ewtabscale;
394 ewitab = ewrt;
395 eweps = ewrt-ewitab;
396 ewitab = 4*ewitab;
397 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
398 velec = qq12*(rinv12-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
399 felec = qq12*rinv12*(rinvsq12-felec);
401 /* Update potential sums from outer loop */
402 velecsum += velec;
404 fscal = felec;
406 /* Calculate temporary vectorial force */
407 tx = fscal*dx12;
408 ty = fscal*dy12;
409 tz = fscal*dz12;
411 /* Update vectorial force */
412 fix1 += tx;
413 fiy1 += ty;
414 fiz1 += tz;
415 f[j_coord_offset+DIM*2+XX] -= tx;
416 f[j_coord_offset+DIM*2+YY] -= ty;
417 f[j_coord_offset+DIM*2+ZZ] -= tz;
419 /**************************
420 * CALCULATE INTERACTIONS *
421 **************************/
423 r13 = rsq13*rinv13;
425 /* EWALD ELECTROSTATICS */
427 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
428 ewrt = r13*ewtabscale;
429 ewitab = ewrt;
430 eweps = ewrt-ewitab;
431 ewitab = 4*ewitab;
432 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
433 velec = qq13*(rinv13-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
434 felec = qq13*rinv13*(rinvsq13-felec);
436 /* Update potential sums from outer loop */
437 velecsum += velec;
439 fscal = felec;
441 /* Calculate temporary vectorial force */
442 tx = fscal*dx13;
443 ty = fscal*dy13;
444 tz = fscal*dz13;
446 /* Update vectorial force */
447 fix1 += tx;
448 fiy1 += ty;
449 fiz1 += tz;
450 f[j_coord_offset+DIM*3+XX] -= tx;
451 f[j_coord_offset+DIM*3+YY] -= ty;
452 f[j_coord_offset+DIM*3+ZZ] -= tz;
454 /**************************
455 * CALCULATE INTERACTIONS *
456 **************************/
458 r21 = rsq21*rinv21;
460 /* EWALD ELECTROSTATICS */
462 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
463 ewrt = r21*ewtabscale;
464 ewitab = ewrt;
465 eweps = ewrt-ewitab;
466 ewitab = 4*ewitab;
467 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
468 velec = qq21*(rinv21-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
469 felec = qq21*rinv21*(rinvsq21-felec);
471 /* Update potential sums from outer loop */
472 velecsum += velec;
474 fscal = felec;
476 /* Calculate temporary vectorial force */
477 tx = fscal*dx21;
478 ty = fscal*dy21;
479 tz = fscal*dz21;
481 /* Update vectorial force */
482 fix2 += tx;
483 fiy2 += ty;
484 fiz2 += tz;
485 f[j_coord_offset+DIM*1+XX] -= tx;
486 f[j_coord_offset+DIM*1+YY] -= ty;
487 f[j_coord_offset+DIM*1+ZZ] -= tz;
489 /**************************
490 * CALCULATE INTERACTIONS *
491 **************************/
493 r22 = rsq22*rinv22;
495 /* EWALD ELECTROSTATICS */
497 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
498 ewrt = r22*ewtabscale;
499 ewitab = ewrt;
500 eweps = ewrt-ewitab;
501 ewitab = 4*ewitab;
502 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
503 velec = qq22*(rinv22-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
504 felec = qq22*rinv22*(rinvsq22-felec);
506 /* Update potential sums from outer loop */
507 velecsum += velec;
509 fscal = felec;
511 /* Calculate temporary vectorial force */
512 tx = fscal*dx22;
513 ty = fscal*dy22;
514 tz = fscal*dz22;
516 /* Update vectorial force */
517 fix2 += tx;
518 fiy2 += ty;
519 fiz2 += tz;
520 f[j_coord_offset+DIM*2+XX] -= tx;
521 f[j_coord_offset+DIM*2+YY] -= ty;
522 f[j_coord_offset+DIM*2+ZZ] -= tz;
524 /**************************
525 * CALCULATE INTERACTIONS *
526 **************************/
528 r23 = rsq23*rinv23;
530 /* EWALD ELECTROSTATICS */
532 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
533 ewrt = r23*ewtabscale;
534 ewitab = ewrt;
535 eweps = ewrt-ewitab;
536 ewitab = 4*ewitab;
537 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
538 velec = qq23*(rinv23-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
539 felec = qq23*rinv23*(rinvsq23-felec);
541 /* Update potential sums from outer loop */
542 velecsum += velec;
544 fscal = felec;
546 /* Calculate temporary vectorial force */
547 tx = fscal*dx23;
548 ty = fscal*dy23;
549 tz = fscal*dz23;
551 /* Update vectorial force */
552 fix2 += tx;
553 fiy2 += ty;
554 fiz2 += tz;
555 f[j_coord_offset+DIM*3+XX] -= tx;
556 f[j_coord_offset+DIM*3+YY] -= ty;
557 f[j_coord_offset+DIM*3+ZZ] -= tz;
559 /**************************
560 * CALCULATE INTERACTIONS *
561 **************************/
563 r31 = rsq31*rinv31;
565 /* EWALD ELECTROSTATICS */
567 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
568 ewrt = r31*ewtabscale;
569 ewitab = ewrt;
570 eweps = ewrt-ewitab;
571 ewitab = 4*ewitab;
572 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
573 velec = qq31*(rinv31-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
574 felec = qq31*rinv31*(rinvsq31-felec);
576 /* Update potential sums from outer loop */
577 velecsum += velec;
579 fscal = felec;
581 /* Calculate temporary vectorial force */
582 tx = fscal*dx31;
583 ty = fscal*dy31;
584 tz = fscal*dz31;
586 /* Update vectorial force */
587 fix3 += tx;
588 fiy3 += ty;
589 fiz3 += tz;
590 f[j_coord_offset+DIM*1+XX] -= tx;
591 f[j_coord_offset+DIM*1+YY] -= ty;
592 f[j_coord_offset+DIM*1+ZZ] -= tz;
594 /**************************
595 * CALCULATE INTERACTIONS *
596 **************************/
598 r32 = rsq32*rinv32;
600 /* EWALD ELECTROSTATICS */
602 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
603 ewrt = r32*ewtabscale;
604 ewitab = ewrt;
605 eweps = ewrt-ewitab;
606 ewitab = 4*ewitab;
607 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
608 velec = qq32*(rinv32-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
609 felec = qq32*rinv32*(rinvsq32-felec);
611 /* Update potential sums from outer loop */
612 velecsum += velec;
614 fscal = felec;
616 /* Calculate temporary vectorial force */
617 tx = fscal*dx32;
618 ty = fscal*dy32;
619 tz = fscal*dz32;
621 /* Update vectorial force */
622 fix3 += tx;
623 fiy3 += ty;
624 fiz3 += tz;
625 f[j_coord_offset+DIM*2+XX] -= tx;
626 f[j_coord_offset+DIM*2+YY] -= ty;
627 f[j_coord_offset+DIM*2+ZZ] -= tz;
629 /**************************
630 * CALCULATE INTERACTIONS *
631 **************************/
633 r33 = rsq33*rinv33;
635 /* EWALD ELECTROSTATICS */
637 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
638 ewrt = r33*ewtabscale;
639 ewitab = ewrt;
640 eweps = ewrt-ewitab;
641 ewitab = 4*ewitab;
642 felec = ewtab[ewitab]+eweps*ewtab[ewitab+1];
643 velec = qq33*(rinv33-(ewtab[ewitab+2]-ewtabhalfspace*eweps*(ewtab[ewitab]+felec)));
644 felec = qq33*rinv33*(rinvsq33-felec);
646 /* Update potential sums from outer loop */
647 velecsum += velec;
649 fscal = felec;
651 /* Calculate temporary vectorial force */
652 tx = fscal*dx33;
653 ty = fscal*dy33;
654 tz = fscal*dz33;
656 /* Update vectorial force */
657 fix3 += tx;
658 fiy3 += ty;
659 fiz3 += tz;
660 f[j_coord_offset+DIM*3+XX] -= tx;
661 f[j_coord_offset+DIM*3+YY] -= ty;
662 f[j_coord_offset+DIM*3+ZZ] -= tz;
664 /* Inner loop uses 415 flops */
666 /* End of innermost loop */
668 tx = ty = tz = 0;
669 f[i_coord_offset+DIM*0+XX] += fix0;
670 f[i_coord_offset+DIM*0+YY] += fiy0;
671 f[i_coord_offset+DIM*0+ZZ] += fiz0;
672 tx += fix0;
673 ty += fiy0;
674 tz += fiz0;
675 f[i_coord_offset+DIM*1+XX] += fix1;
676 f[i_coord_offset+DIM*1+YY] += fiy1;
677 f[i_coord_offset+DIM*1+ZZ] += fiz1;
678 tx += fix1;
679 ty += fiy1;
680 tz += fiz1;
681 f[i_coord_offset+DIM*2+XX] += fix2;
682 f[i_coord_offset+DIM*2+YY] += fiy2;
683 f[i_coord_offset+DIM*2+ZZ] += fiz2;
684 tx += fix2;
685 ty += fiy2;
686 tz += fiz2;
687 f[i_coord_offset+DIM*3+XX] += fix3;
688 f[i_coord_offset+DIM*3+YY] += fiy3;
689 f[i_coord_offset+DIM*3+ZZ] += fiz3;
690 tx += fix3;
691 ty += fiy3;
692 tz += fiz3;
693 fshift[i_shift_offset+XX] += tx;
694 fshift[i_shift_offset+YY] += ty;
695 fshift[i_shift_offset+ZZ] += tz;
697 ggid = gid[iidx];
698 /* Update potential energies */
699 kernel_data->energygrp_elec[ggid] += velecsum;
700 kernel_data->energygrp_vdw[ggid] += vvdwsum;
702 /* Increment number of inner iterations */
703 inneriter += j_index_end - j_index_start;
705 /* Outer loop uses 41 flops */
708 /* Increment number of outer iterations */
709 outeriter += nri;
711 /* Update outer/inner flops */
713 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*41 + inneriter*415);
716 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_c
717 * Electrostatics interaction: Ewald
718 * VdW interaction: CubicSplineTable
719 * Geometry: Water4-Water4
720 * Calculate force/pot: Force
722 void
723 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_c
724 (t_nblist * gmx_restrict nlist,
725 rvec * gmx_restrict xx,
726 rvec * gmx_restrict ff,
727 struct t_forcerec * gmx_restrict fr,
728 t_mdatoms * gmx_restrict mdatoms,
729 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
730 t_nrnb * gmx_restrict nrnb)
732 int i_shift_offset,i_coord_offset,j_coord_offset;
733 int j_index_start,j_index_end;
734 int nri,inr,ggid,iidx,jidx,jnr,outeriter,inneriter;
735 real shX,shY,shZ,tx,ty,tz,fscal,rcutoff,rcutoff2;
736 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
737 real *shiftvec,*fshift,*x,*f;
738 int vdwioffset0;
739 real ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
740 int vdwioffset1;
741 real ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
742 int vdwioffset2;
743 real ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
744 int vdwioffset3;
745 real ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
746 int vdwjidx0;
747 real jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
748 int vdwjidx1;
749 real jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
750 int vdwjidx2;
751 real jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
752 int vdwjidx3;
753 real jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
754 real dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00,cexp1_00,cexp2_00;
755 real dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11,cexp1_11,cexp2_11;
756 real dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12,cexp1_12,cexp2_12;
757 real dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13,cexp1_13,cexp2_13;
758 real dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21,cexp1_21,cexp2_21;
759 real dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22,cexp1_22,cexp2_22;
760 real dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23,cexp1_23,cexp2_23;
761 real dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31,cexp1_31,cexp2_31;
762 real dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32,cexp1_32,cexp2_32;
763 real dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33,cexp1_33,cexp2_33;
764 real velec,felec,velecsum,facel,crf,krf,krf2;
765 real *charge;
766 int nvdwtype;
767 real rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,br,vvdwexp,sh_vdw_invrcut6;
768 int *vdwtype;
769 real *vdwparam;
770 int vfitab;
771 real rt,vfeps,vftabscale,Y,F,Geps,Heps2,Fp,VV,FF;
772 real *vftab;
773 int ewitab;
774 real ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace;
775 real *ewtab;
777 x = xx[0];
778 f = ff[0];
780 nri = nlist->nri;
781 iinr = nlist->iinr;
782 jindex = nlist->jindex;
783 jjnr = nlist->jjnr;
784 shiftidx = nlist->shift;
785 gid = nlist->gid;
786 shiftvec = fr->shift_vec[0];
787 fshift = fr->fshift[0];
788 facel = fr->ic->epsfac;
789 charge = mdatoms->chargeA;
790 nvdwtype = fr->ntype;
791 vdwparam = fr->nbfp;
792 vdwtype = mdatoms->typeA;
794 vftab = kernel_data->table_vdw->data;
795 vftabscale = kernel_data->table_vdw->scale;
797 sh_ewald = fr->ic->sh_ewald;
798 ewtab = fr->ic->tabq_coul_F;
799 ewtabscale = fr->ic->tabq_scale;
800 ewtabhalfspace = 0.5/ewtabscale;
802 /* Setup water-specific parameters */
803 inr = nlist->iinr[0];
804 iq1 = facel*charge[inr+1];
805 iq2 = facel*charge[inr+2];
806 iq3 = facel*charge[inr+3];
807 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
809 jq1 = charge[inr+1];
810 jq2 = charge[inr+2];
811 jq3 = charge[inr+3];
812 vdwjidx0 = 2*vdwtype[inr+0];
813 c6_00 = vdwparam[vdwioffset0+vdwjidx0];
814 c12_00 = vdwparam[vdwioffset0+vdwjidx0+1];
815 qq11 = iq1*jq1;
816 qq12 = iq1*jq2;
817 qq13 = iq1*jq3;
818 qq21 = iq2*jq1;
819 qq22 = iq2*jq2;
820 qq23 = iq2*jq3;
821 qq31 = iq3*jq1;
822 qq32 = iq3*jq2;
823 qq33 = iq3*jq3;
825 outeriter = 0;
826 inneriter = 0;
828 /* Start outer loop over neighborlists */
829 for(iidx=0; iidx<nri; iidx++)
831 /* Load shift vector for this list */
832 i_shift_offset = DIM*shiftidx[iidx];
833 shX = shiftvec[i_shift_offset+XX];
834 shY = shiftvec[i_shift_offset+YY];
835 shZ = shiftvec[i_shift_offset+ZZ];
837 /* Load limits for loop over neighbors */
838 j_index_start = jindex[iidx];
839 j_index_end = jindex[iidx+1];
841 /* Get outer coordinate index */
842 inr = iinr[iidx];
843 i_coord_offset = DIM*inr;
845 /* Load i particle coords and add shift vector */
846 ix0 = shX + x[i_coord_offset+DIM*0+XX];
847 iy0 = shY + x[i_coord_offset+DIM*0+YY];
848 iz0 = shZ + x[i_coord_offset+DIM*0+ZZ];
849 ix1 = shX + x[i_coord_offset+DIM*1+XX];
850 iy1 = shY + x[i_coord_offset+DIM*1+YY];
851 iz1 = shZ + x[i_coord_offset+DIM*1+ZZ];
852 ix2 = shX + x[i_coord_offset+DIM*2+XX];
853 iy2 = shY + x[i_coord_offset+DIM*2+YY];
854 iz2 = shZ + x[i_coord_offset+DIM*2+ZZ];
855 ix3 = shX + x[i_coord_offset+DIM*3+XX];
856 iy3 = shY + x[i_coord_offset+DIM*3+YY];
857 iz3 = shZ + x[i_coord_offset+DIM*3+ZZ];
859 fix0 = 0.0;
860 fiy0 = 0.0;
861 fiz0 = 0.0;
862 fix1 = 0.0;
863 fiy1 = 0.0;
864 fiz1 = 0.0;
865 fix2 = 0.0;
866 fiy2 = 0.0;
867 fiz2 = 0.0;
868 fix3 = 0.0;
869 fiy3 = 0.0;
870 fiz3 = 0.0;
872 /* Start inner kernel loop */
873 for(jidx=j_index_start; jidx<j_index_end; jidx++)
875 /* Get j neighbor index, and coordinate index */
876 jnr = jjnr[jidx];
877 j_coord_offset = DIM*jnr;
879 /* load j atom coordinates */
880 jx0 = x[j_coord_offset+DIM*0+XX];
881 jy0 = x[j_coord_offset+DIM*0+YY];
882 jz0 = x[j_coord_offset+DIM*0+ZZ];
883 jx1 = x[j_coord_offset+DIM*1+XX];
884 jy1 = x[j_coord_offset+DIM*1+YY];
885 jz1 = x[j_coord_offset+DIM*1+ZZ];
886 jx2 = x[j_coord_offset+DIM*2+XX];
887 jy2 = x[j_coord_offset+DIM*2+YY];
888 jz2 = x[j_coord_offset+DIM*2+ZZ];
889 jx3 = x[j_coord_offset+DIM*3+XX];
890 jy3 = x[j_coord_offset+DIM*3+YY];
891 jz3 = x[j_coord_offset+DIM*3+ZZ];
893 /* Calculate displacement vector */
894 dx00 = ix0 - jx0;
895 dy00 = iy0 - jy0;
896 dz00 = iz0 - jz0;
897 dx11 = ix1 - jx1;
898 dy11 = iy1 - jy1;
899 dz11 = iz1 - jz1;
900 dx12 = ix1 - jx2;
901 dy12 = iy1 - jy2;
902 dz12 = iz1 - jz2;
903 dx13 = ix1 - jx3;
904 dy13 = iy1 - jy3;
905 dz13 = iz1 - jz3;
906 dx21 = ix2 - jx1;
907 dy21 = iy2 - jy1;
908 dz21 = iz2 - jz1;
909 dx22 = ix2 - jx2;
910 dy22 = iy2 - jy2;
911 dz22 = iz2 - jz2;
912 dx23 = ix2 - jx3;
913 dy23 = iy2 - jy3;
914 dz23 = iz2 - jz3;
915 dx31 = ix3 - jx1;
916 dy31 = iy3 - jy1;
917 dz31 = iz3 - jz1;
918 dx32 = ix3 - jx2;
919 dy32 = iy3 - jy2;
920 dz32 = iz3 - jz2;
921 dx33 = ix3 - jx3;
922 dy33 = iy3 - jy3;
923 dz33 = iz3 - jz3;
925 /* Calculate squared distance and things based on it */
926 rsq00 = dx00*dx00+dy00*dy00+dz00*dz00;
927 rsq11 = dx11*dx11+dy11*dy11+dz11*dz11;
928 rsq12 = dx12*dx12+dy12*dy12+dz12*dz12;
929 rsq13 = dx13*dx13+dy13*dy13+dz13*dz13;
930 rsq21 = dx21*dx21+dy21*dy21+dz21*dz21;
931 rsq22 = dx22*dx22+dy22*dy22+dz22*dz22;
932 rsq23 = dx23*dx23+dy23*dy23+dz23*dz23;
933 rsq31 = dx31*dx31+dy31*dy31+dz31*dz31;
934 rsq32 = dx32*dx32+dy32*dy32+dz32*dz32;
935 rsq33 = dx33*dx33+dy33*dy33+dz33*dz33;
937 rinv00 = 1.0/sqrt(rsq00);
938 rinv11 = 1.0/sqrt(rsq11);
939 rinv12 = 1.0/sqrt(rsq12);
940 rinv13 = 1.0/sqrt(rsq13);
941 rinv21 = 1.0/sqrt(rsq21);
942 rinv22 = 1.0/sqrt(rsq22);
943 rinv23 = 1.0/sqrt(rsq23);
944 rinv31 = 1.0/sqrt(rsq31);
945 rinv32 = 1.0/sqrt(rsq32);
946 rinv33 = 1.0/sqrt(rsq33);
948 rinvsq11 = rinv11*rinv11;
949 rinvsq12 = rinv12*rinv12;
950 rinvsq13 = rinv13*rinv13;
951 rinvsq21 = rinv21*rinv21;
952 rinvsq22 = rinv22*rinv22;
953 rinvsq23 = rinv23*rinv23;
954 rinvsq31 = rinv31*rinv31;
955 rinvsq32 = rinv32*rinv32;
956 rinvsq33 = rinv33*rinv33;
958 /**************************
959 * CALCULATE INTERACTIONS *
960 **************************/
962 r00 = rsq00*rinv00;
964 /* Calculate table index by multiplying r with table scale and truncate to integer */
965 rt = r00*vftabscale;
966 vfitab = rt;
967 vfeps = rt-vfitab;
968 vfitab = 2*4*vfitab;
970 /* CUBIC SPLINE TABLE DISPERSION */
971 vfitab += 0;
972 F = vftab[vfitab+1];
973 Geps = vfeps*vftab[vfitab+2];
974 Heps2 = vfeps*vfeps*vftab[vfitab+3];
975 Fp = F+Geps+Heps2;
976 FF = Fp+Geps+2.0*Heps2;
977 fvdw6 = c6_00*FF;
979 /* CUBIC SPLINE TABLE REPULSION */
980 F = vftab[vfitab+5];
981 Geps = vfeps*vftab[vfitab+6];
982 Heps2 = vfeps*vfeps*vftab[vfitab+7];
983 Fp = F+Geps+Heps2;
984 FF = Fp+Geps+2.0*Heps2;
985 fvdw12 = c12_00*FF;
986 fvdw = -(fvdw6+fvdw12)*vftabscale*rinv00;
988 fscal = fvdw;
990 /* Calculate temporary vectorial force */
991 tx = fscal*dx00;
992 ty = fscal*dy00;
993 tz = fscal*dz00;
995 /* Update vectorial force */
996 fix0 += tx;
997 fiy0 += ty;
998 fiz0 += tz;
999 f[j_coord_offset+DIM*0+XX] -= tx;
1000 f[j_coord_offset+DIM*0+YY] -= ty;
1001 f[j_coord_offset+DIM*0+ZZ] -= tz;
1003 /**************************
1004 * CALCULATE INTERACTIONS *
1005 **************************/
1007 r11 = rsq11*rinv11;
1009 /* EWALD ELECTROSTATICS */
1011 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1012 ewrt = r11*ewtabscale;
1013 ewitab = ewrt;
1014 eweps = ewrt-ewitab;
1015 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1016 felec = qq11*rinv11*(rinvsq11-felec);
1018 fscal = felec;
1020 /* Calculate temporary vectorial force */
1021 tx = fscal*dx11;
1022 ty = fscal*dy11;
1023 tz = fscal*dz11;
1025 /* Update vectorial force */
1026 fix1 += tx;
1027 fiy1 += ty;
1028 fiz1 += tz;
1029 f[j_coord_offset+DIM*1+XX] -= tx;
1030 f[j_coord_offset+DIM*1+YY] -= ty;
1031 f[j_coord_offset+DIM*1+ZZ] -= tz;
1033 /**************************
1034 * CALCULATE INTERACTIONS *
1035 **************************/
1037 r12 = rsq12*rinv12;
1039 /* EWALD ELECTROSTATICS */
1041 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1042 ewrt = r12*ewtabscale;
1043 ewitab = ewrt;
1044 eweps = ewrt-ewitab;
1045 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1046 felec = qq12*rinv12*(rinvsq12-felec);
1048 fscal = felec;
1050 /* Calculate temporary vectorial force */
1051 tx = fscal*dx12;
1052 ty = fscal*dy12;
1053 tz = fscal*dz12;
1055 /* Update vectorial force */
1056 fix1 += tx;
1057 fiy1 += ty;
1058 fiz1 += tz;
1059 f[j_coord_offset+DIM*2+XX] -= tx;
1060 f[j_coord_offset+DIM*2+YY] -= ty;
1061 f[j_coord_offset+DIM*2+ZZ] -= tz;
1063 /**************************
1064 * CALCULATE INTERACTIONS *
1065 **************************/
1067 r13 = rsq13*rinv13;
1069 /* EWALD ELECTROSTATICS */
1071 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1072 ewrt = r13*ewtabscale;
1073 ewitab = ewrt;
1074 eweps = ewrt-ewitab;
1075 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1076 felec = qq13*rinv13*(rinvsq13-felec);
1078 fscal = felec;
1080 /* Calculate temporary vectorial force */
1081 tx = fscal*dx13;
1082 ty = fscal*dy13;
1083 tz = fscal*dz13;
1085 /* Update vectorial force */
1086 fix1 += tx;
1087 fiy1 += ty;
1088 fiz1 += tz;
1089 f[j_coord_offset+DIM*3+XX] -= tx;
1090 f[j_coord_offset+DIM*3+YY] -= ty;
1091 f[j_coord_offset+DIM*3+ZZ] -= tz;
1093 /**************************
1094 * CALCULATE INTERACTIONS *
1095 **************************/
1097 r21 = rsq21*rinv21;
1099 /* EWALD ELECTROSTATICS */
1101 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1102 ewrt = r21*ewtabscale;
1103 ewitab = ewrt;
1104 eweps = ewrt-ewitab;
1105 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1106 felec = qq21*rinv21*(rinvsq21-felec);
1108 fscal = felec;
1110 /* Calculate temporary vectorial force */
1111 tx = fscal*dx21;
1112 ty = fscal*dy21;
1113 tz = fscal*dz21;
1115 /* Update vectorial force */
1116 fix2 += tx;
1117 fiy2 += ty;
1118 fiz2 += tz;
1119 f[j_coord_offset+DIM*1+XX] -= tx;
1120 f[j_coord_offset+DIM*1+YY] -= ty;
1121 f[j_coord_offset+DIM*1+ZZ] -= tz;
1123 /**************************
1124 * CALCULATE INTERACTIONS *
1125 **************************/
1127 r22 = rsq22*rinv22;
1129 /* EWALD ELECTROSTATICS */
1131 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1132 ewrt = r22*ewtabscale;
1133 ewitab = ewrt;
1134 eweps = ewrt-ewitab;
1135 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1136 felec = qq22*rinv22*(rinvsq22-felec);
1138 fscal = felec;
1140 /* Calculate temporary vectorial force */
1141 tx = fscal*dx22;
1142 ty = fscal*dy22;
1143 tz = fscal*dz22;
1145 /* Update vectorial force */
1146 fix2 += tx;
1147 fiy2 += ty;
1148 fiz2 += tz;
1149 f[j_coord_offset+DIM*2+XX] -= tx;
1150 f[j_coord_offset+DIM*2+YY] -= ty;
1151 f[j_coord_offset+DIM*2+ZZ] -= tz;
1153 /**************************
1154 * CALCULATE INTERACTIONS *
1155 **************************/
1157 r23 = rsq23*rinv23;
1159 /* EWALD ELECTROSTATICS */
1161 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1162 ewrt = r23*ewtabscale;
1163 ewitab = ewrt;
1164 eweps = ewrt-ewitab;
1165 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1166 felec = qq23*rinv23*(rinvsq23-felec);
1168 fscal = felec;
1170 /* Calculate temporary vectorial force */
1171 tx = fscal*dx23;
1172 ty = fscal*dy23;
1173 tz = fscal*dz23;
1175 /* Update vectorial force */
1176 fix2 += tx;
1177 fiy2 += ty;
1178 fiz2 += tz;
1179 f[j_coord_offset+DIM*3+XX] -= tx;
1180 f[j_coord_offset+DIM*3+YY] -= ty;
1181 f[j_coord_offset+DIM*3+ZZ] -= tz;
1183 /**************************
1184 * CALCULATE INTERACTIONS *
1185 **************************/
1187 r31 = rsq31*rinv31;
1189 /* EWALD ELECTROSTATICS */
1191 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1192 ewrt = r31*ewtabscale;
1193 ewitab = ewrt;
1194 eweps = ewrt-ewitab;
1195 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1196 felec = qq31*rinv31*(rinvsq31-felec);
1198 fscal = felec;
1200 /* Calculate temporary vectorial force */
1201 tx = fscal*dx31;
1202 ty = fscal*dy31;
1203 tz = fscal*dz31;
1205 /* Update vectorial force */
1206 fix3 += tx;
1207 fiy3 += ty;
1208 fiz3 += tz;
1209 f[j_coord_offset+DIM*1+XX] -= tx;
1210 f[j_coord_offset+DIM*1+YY] -= ty;
1211 f[j_coord_offset+DIM*1+ZZ] -= tz;
1213 /**************************
1214 * CALCULATE INTERACTIONS *
1215 **************************/
1217 r32 = rsq32*rinv32;
1219 /* EWALD ELECTROSTATICS */
1221 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1222 ewrt = r32*ewtabscale;
1223 ewitab = ewrt;
1224 eweps = ewrt-ewitab;
1225 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1226 felec = qq32*rinv32*(rinvsq32-felec);
1228 fscal = felec;
1230 /* Calculate temporary vectorial force */
1231 tx = fscal*dx32;
1232 ty = fscal*dy32;
1233 tz = fscal*dz32;
1235 /* Update vectorial force */
1236 fix3 += tx;
1237 fiy3 += ty;
1238 fiz3 += tz;
1239 f[j_coord_offset+DIM*2+XX] -= tx;
1240 f[j_coord_offset+DIM*2+YY] -= ty;
1241 f[j_coord_offset+DIM*2+ZZ] -= tz;
1243 /**************************
1244 * CALCULATE INTERACTIONS *
1245 **************************/
1247 r33 = rsq33*rinv33;
1249 /* EWALD ELECTROSTATICS */
1251 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1252 ewrt = r33*ewtabscale;
1253 ewitab = ewrt;
1254 eweps = ewrt-ewitab;
1255 felec = (1.0-eweps)*ewtab[ewitab]+eweps*ewtab[ewitab+1];
1256 felec = qq33*rinv33*(rinvsq33-felec);
1258 fscal = felec;
1260 /* Calculate temporary vectorial force */
1261 tx = fscal*dx33;
1262 ty = fscal*dy33;
1263 tz = fscal*dz33;
1265 /* Update vectorial force */
1266 fix3 += tx;
1267 fiy3 += ty;
1268 fiz3 += tz;
1269 f[j_coord_offset+DIM*3+XX] -= tx;
1270 f[j_coord_offset+DIM*3+YY] -= ty;
1271 f[j_coord_offset+DIM*3+ZZ] -= tz;
1273 /* Inner loop uses 344 flops */
1275 /* End of innermost loop */
1277 tx = ty = tz = 0;
1278 f[i_coord_offset+DIM*0+XX] += fix0;
1279 f[i_coord_offset+DIM*0+YY] += fiy0;
1280 f[i_coord_offset+DIM*0+ZZ] += fiz0;
1281 tx += fix0;
1282 ty += fiy0;
1283 tz += fiz0;
1284 f[i_coord_offset+DIM*1+XX] += fix1;
1285 f[i_coord_offset+DIM*1+YY] += fiy1;
1286 f[i_coord_offset+DIM*1+ZZ] += fiz1;
1287 tx += fix1;
1288 ty += fiy1;
1289 tz += fiz1;
1290 f[i_coord_offset+DIM*2+XX] += fix2;
1291 f[i_coord_offset+DIM*2+YY] += fiy2;
1292 f[i_coord_offset+DIM*2+ZZ] += fiz2;
1293 tx += fix2;
1294 ty += fiy2;
1295 tz += fiz2;
1296 f[i_coord_offset+DIM*3+XX] += fix3;
1297 f[i_coord_offset+DIM*3+YY] += fiy3;
1298 f[i_coord_offset+DIM*3+ZZ] += fiz3;
1299 tx += fix3;
1300 ty += fiy3;
1301 tz += fiz3;
1302 fshift[i_shift_offset+XX] += tx;
1303 fshift[i_shift_offset+YY] += ty;
1304 fshift[i_shift_offset+ZZ] += tz;
1306 /* Increment number of inner iterations */
1307 inneriter += j_index_end - j_index_start;
1309 /* Outer loop uses 39 flops */
1312 /* Increment number of outer iterations */
1313 outeriter += nri;
1315 /* Update outer/inner flops */
1317 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*39 + inneriter*344);