2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, 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 sparc64_hpc_ace_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "kernelutil_sparc64_hpc_ace_double.h"
50 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
51 * Electrostatics interaction: Ewald
52 * VdW interaction: None
53 * Geometry: Water3-Particle
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecEw_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 t_forcerec * gmx_restrict fr,
62 t_mdatoms * gmx_restrict mdatoms,
63 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
64 t_nrnb * gmx_restrict nrnb)
66 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
67 * just 0 for non-waters.
68 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
69 * jnr indices corresponding to data put in the four positions in the SIMD register.
71 int i_shift_offset,i_coord_offset,outeriter,inneriter;
72 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int j_coord_offsetA,j_coord_offsetB;
75 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
77 real *shiftvec,*fshift,*x,*f;
78 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
80 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
82 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
84 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
85 int vdwjidx0A,vdwjidx0B;
86 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
87 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
88 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
89 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
90 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
92 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
95 _fjsp_v2r8 dummy_mask,cutoff_mask;
96 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
97 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
98 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
105 jindex = nlist->jindex;
107 shiftidx = nlist->shift;
109 shiftvec = fr->shift_vec[0];
110 fshift = fr->fshift[0];
111 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
112 charge = mdatoms->chargeA;
114 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
115 ewtab = fr->ic->tabq_coul_FDV0;
116 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
117 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
119 /* Setup water-specific parameters */
120 inr = nlist->iinr[0];
121 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
122 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
123 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
125 /* Avoid stupid compiler warnings */
133 /* Start outer loop over neighborlists */
134 for(iidx=0; iidx<nri; iidx++)
136 /* Load shift vector for this list */
137 i_shift_offset = DIM*shiftidx[iidx];
139 /* Load limits for loop over neighbors */
140 j_index_start = jindex[iidx];
141 j_index_end = jindex[iidx+1];
143 /* Get outer coordinate index */
145 i_coord_offset = DIM*inr;
147 /* Load i particle coords and add shift vector */
148 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
149 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
151 fix0 = _fjsp_setzero_v2r8();
152 fiy0 = _fjsp_setzero_v2r8();
153 fiz0 = _fjsp_setzero_v2r8();
154 fix1 = _fjsp_setzero_v2r8();
155 fiy1 = _fjsp_setzero_v2r8();
156 fiz1 = _fjsp_setzero_v2r8();
157 fix2 = _fjsp_setzero_v2r8();
158 fiy2 = _fjsp_setzero_v2r8();
159 fiz2 = _fjsp_setzero_v2r8();
161 /* Reset potential sums */
162 velecsum = _fjsp_setzero_v2r8();
164 /* Start inner kernel loop */
165 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
168 /* Get j neighbor index, and coordinate index */
171 j_coord_offsetA = DIM*jnrA;
172 j_coord_offsetB = DIM*jnrB;
174 /* load j atom coordinates */
175 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
178 /* Calculate displacement vector */
179 dx00 = _fjsp_sub_v2r8(ix0,jx0);
180 dy00 = _fjsp_sub_v2r8(iy0,jy0);
181 dz00 = _fjsp_sub_v2r8(iz0,jz0);
182 dx10 = _fjsp_sub_v2r8(ix1,jx0);
183 dy10 = _fjsp_sub_v2r8(iy1,jy0);
184 dz10 = _fjsp_sub_v2r8(iz1,jz0);
185 dx20 = _fjsp_sub_v2r8(ix2,jx0);
186 dy20 = _fjsp_sub_v2r8(iy2,jy0);
187 dz20 = _fjsp_sub_v2r8(iz2,jz0);
189 /* Calculate squared distance and things based on it */
190 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
191 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
192 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
194 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
195 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
196 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
198 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
199 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
200 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
202 /* Load parameters for j particles */
203 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
205 fjx0 = _fjsp_setzero_v2r8();
206 fjy0 = _fjsp_setzero_v2r8();
207 fjz0 = _fjsp_setzero_v2r8();
209 /**************************
210 * CALCULATE INTERACTIONS *
211 **************************/
213 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
215 /* Compute parameters for interactions between i and j atoms */
216 qq00 = _fjsp_mul_v2r8(iq0,jq0);
218 /* EWALD ELECTROSTATICS */
220 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
221 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
222 itab_tmp = _fjsp_dtox_v2r8(ewrt);
223 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
224 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
226 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
227 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
228 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
229 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
230 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
231 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
232 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
233 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
234 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
235 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
237 /* Update potential sum for this i atom from the interaction with this j atom. */
238 velecsum = _fjsp_add_v2r8(velecsum,velec);
242 /* Update vectorial force */
243 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
244 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
245 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
247 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
248 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
249 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
251 /**************************
252 * CALCULATE INTERACTIONS *
253 **************************/
255 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
257 /* Compute parameters for interactions between i and j atoms */
258 qq10 = _fjsp_mul_v2r8(iq1,jq0);
260 /* EWALD ELECTROSTATICS */
262 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
263 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
264 itab_tmp = _fjsp_dtox_v2r8(ewrt);
265 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
266 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
268 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
269 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
270 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
271 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
272 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
273 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
274 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
275 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
276 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
277 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
279 /* Update potential sum for this i atom from the interaction with this j atom. */
280 velecsum = _fjsp_add_v2r8(velecsum,velec);
284 /* Update vectorial force */
285 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
286 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
287 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
289 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
290 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
291 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
293 /**************************
294 * CALCULATE INTERACTIONS *
295 **************************/
297 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
299 /* Compute parameters for interactions between i and j atoms */
300 qq20 = _fjsp_mul_v2r8(iq2,jq0);
302 /* EWALD ELECTROSTATICS */
304 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
305 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
306 itab_tmp = _fjsp_dtox_v2r8(ewrt);
307 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
308 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
310 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
311 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
312 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
313 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
314 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
315 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
316 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
317 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
318 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
319 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
321 /* Update potential sum for this i atom from the interaction with this j atom. */
322 velecsum = _fjsp_add_v2r8(velecsum,velec);
326 /* Update vectorial force */
327 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
328 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
329 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
331 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
332 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
333 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
335 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
337 /* Inner loop uses 135 flops */
344 j_coord_offsetA = DIM*jnrA;
346 /* load j atom coordinates */
347 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
350 /* Calculate displacement vector */
351 dx00 = _fjsp_sub_v2r8(ix0,jx0);
352 dy00 = _fjsp_sub_v2r8(iy0,jy0);
353 dz00 = _fjsp_sub_v2r8(iz0,jz0);
354 dx10 = _fjsp_sub_v2r8(ix1,jx0);
355 dy10 = _fjsp_sub_v2r8(iy1,jy0);
356 dz10 = _fjsp_sub_v2r8(iz1,jz0);
357 dx20 = _fjsp_sub_v2r8(ix2,jx0);
358 dy20 = _fjsp_sub_v2r8(iy2,jy0);
359 dz20 = _fjsp_sub_v2r8(iz2,jz0);
361 /* Calculate squared distance and things based on it */
362 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
363 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
364 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
366 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
367 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
368 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
370 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
371 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
372 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
374 /* Load parameters for j particles */
375 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
377 fjx0 = _fjsp_setzero_v2r8();
378 fjy0 = _fjsp_setzero_v2r8();
379 fjz0 = _fjsp_setzero_v2r8();
381 /**************************
382 * CALCULATE INTERACTIONS *
383 **************************/
385 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
387 /* Compute parameters for interactions between i and j atoms */
388 qq00 = _fjsp_mul_v2r8(iq0,jq0);
390 /* EWALD ELECTROSTATICS */
392 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
393 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
394 itab_tmp = _fjsp_dtox_v2r8(ewrt);
395 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
396 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
398 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
399 ewtabD = _fjsp_setzero_v2r8();
400 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
401 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
402 ewtabFn = _fjsp_setzero_v2r8();
403 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
404 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
405 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
406 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
407 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
409 /* Update potential sum for this i atom from the interaction with this j atom. */
410 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
411 velecsum = _fjsp_add_v2r8(velecsum,velec);
415 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
417 /* Update vectorial force */
418 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
419 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
420 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
422 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
423 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
424 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
426 /**************************
427 * CALCULATE INTERACTIONS *
428 **************************/
430 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
432 /* Compute parameters for interactions between i and j atoms */
433 qq10 = _fjsp_mul_v2r8(iq1,jq0);
435 /* EWALD ELECTROSTATICS */
437 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
438 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
439 itab_tmp = _fjsp_dtox_v2r8(ewrt);
440 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
441 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
443 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
444 ewtabD = _fjsp_setzero_v2r8();
445 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
446 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
447 ewtabFn = _fjsp_setzero_v2r8();
448 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
449 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
450 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
451 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
452 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
454 /* Update potential sum for this i atom from the interaction with this j atom. */
455 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
456 velecsum = _fjsp_add_v2r8(velecsum,velec);
460 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
462 /* Update vectorial force */
463 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
464 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
465 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
467 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
468 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
469 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
471 /**************************
472 * CALCULATE INTERACTIONS *
473 **************************/
475 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
477 /* Compute parameters for interactions between i and j atoms */
478 qq20 = _fjsp_mul_v2r8(iq2,jq0);
480 /* EWALD ELECTROSTATICS */
482 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
483 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
484 itab_tmp = _fjsp_dtox_v2r8(ewrt);
485 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
486 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
488 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
489 ewtabD = _fjsp_setzero_v2r8();
490 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
491 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
492 ewtabFn = _fjsp_setzero_v2r8();
493 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
494 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
495 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
496 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
497 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
499 /* Update potential sum for this i atom from the interaction with this j atom. */
500 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
501 velecsum = _fjsp_add_v2r8(velecsum,velec);
505 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
507 /* Update vectorial force */
508 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
509 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
510 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
512 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
513 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
514 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
516 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
518 /* Inner loop uses 135 flops */
521 /* End of innermost loop */
523 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
524 f+i_coord_offset,fshift+i_shift_offset);
527 /* Update potential energies */
528 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
530 /* Increment number of inner iterations */
531 inneriter += j_index_end - j_index_start;
533 /* Outer loop uses 19 flops */
536 /* Increment number of outer iterations */
539 /* Update outer/inner flops */
541 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*135);
544 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
545 * Electrostatics interaction: Ewald
546 * VdW interaction: None
547 * Geometry: Water3-Particle
548 * Calculate force/pot: Force
551 nb_kernel_ElecEw_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
552 (t_nblist * gmx_restrict nlist,
553 rvec * gmx_restrict xx,
554 rvec * gmx_restrict ff,
555 t_forcerec * gmx_restrict fr,
556 t_mdatoms * gmx_restrict mdatoms,
557 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
558 t_nrnb * gmx_restrict nrnb)
560 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
561 * just 0 for non-waters.
562 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
563 * jnr indices corresponding to data put in the four positions in the SIMD register.
565 int i_shift_offset,i_coord_offset,outeriter,inneriter;
566 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
568 int j_coord_offsetA,j_coord_offsetB;
569 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
571 real *shiftvec,*fshift,*x,*f;
572 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
574 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
576 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
578 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
579 int vdwjidx0A,vdwjidx0B;
580 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
581 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
582 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
583 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
584 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
586 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
589 _fjsp_v2r8 dummy_mask,cutoff_mask;
590 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
591 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
592 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
599 jindex = nlist->jindex;
601 shiftidx = nlist->shift;
603 shiftvec = fr->shift_vec[0];
604 fshift = fr->fshift[0];
605 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
606 charge = mdatoms->chargeA;
608 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
609 ewtab = fr->ic->tabq_coul_F;
610 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
611 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
613 /* Setup water-specific parameters */
614 inr = nlist->iinr[0];
615 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
616 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
617 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
619 /* Avoid stupid compiler warnings */
627 /* Start outer loop over neighborlists */
628 for(iidx=0; iidx<nri; iidx++)
630 /* Load shift vector for this list */
631 i_shift_offset = DIM*shiftidx[iidx];
633 /* Load limits for loop over neighbors */
634 j_index_start = jindex[iidx];
635 j_index_end = jindex[iidx+1];
637 /* Get outer coordinate index */
639 i_coord_offset = DIM*inr;
641 /* Load i particle coords and add shift vector */
642 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
643 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
645 fix0 = _fjsp_setzero_v2r8();
646 fiy0 = _fjsp_setzero_v2r8();
647 fiz0 = _fjsp_setzero_v2r8();
648 fix1 = _fjsp_setzero_v2r8();
649 fiy1 = _fjsp_setzero_v2r8();
650 fiz1 = _fjsp_setzero_v2r8();
651 fix2 = _fjsp_setzero_v2r8();
652 fiy2 = _fjsp_setzero_v2r8();
653 fiz2 = _fjsp_setzero_v2r8();
655 /* Start inner kernel loop */
656 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
659 /* Get j neighbor index, and coordinate index */
662 j_coord_offsetA = DIM*jnrA;
663 j_coord_offsetB = DIM*jnrB;
665 /* load j atom coordinates */
666 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
669 /* Calculate displacement vector */
670 dx00 = _fjsp_sub_v2r8(ix0,jx0);
671 dy00 = _fjsp_sub_v2r8(iy0,jy0);
672 dz00 = _fjsp_sub_v2r8(iz0,jz0);
673 dx10 = _fjsp_sub_v2r8(ix1,jx0);
674 dy10 = _fjsp_sub_v2r8(iy1,jy0);
675 dz10 = _fjsp_sub_v2r8(iz1,jz0);
676 dx20 = _fjsp_sub_v2r8(ix2,jx0);
677 dy20 = _fjsp_sub_v2r8(iy2,jy0);
678 dz20 = _fjsp_sub_v2r8(iz2,jz0);
680 /* Calculate squared distance and things based on it */
681 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
682 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
683 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
685 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
686 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
687 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
689 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
690 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
691 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
693 /* Load parameters for j particles */
694 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
696 fjx0 = _fjsp_setzero_v2r8();
697 fjy0 = _fjsp_setzero_v2r8();
698 fjz0 = _fjsp_setzero_v2r8();
700 /**************************
701 * CALCULATE INTERACTIONS *
702 **************************/
704 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
706 /* Compute parameters for interactions between i and j atoms */
707 qq00 = _fjsp_mul_v2r8(iq0,jq0);
709 /* EWALD ELECTROSTATICS */
711 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
712 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
713 itab_tmp = _fjsp_dtox_v2r8(ewrt);
714 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
715 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
717 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
719 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
720 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
724 /* Update vectorial force */
725 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
726 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
727 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
729 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
730 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
731 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
733 /**************************
734 * CALCULATE INTERACTIONS *
735 **************************/
737 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
739 /* Compute parameters for interactions between i and j atoms */
740 qq10 = _fjsp_mul_v2r8(iq1,jq0);
742 /* EWALD ELECTROSTATICS */
744 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
745 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
746 itab_tmp = _fjsp_dtox_v2r8(ewrt);
747 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
748 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
750 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
752 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
753 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
757 /* Update vectorial force */
758 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
759 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
760 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
762 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
763 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
764 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
766 /**************************
767 * CALCULATE INTERACTIONS *
768 **************************/
770 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
772 /* Compute parameters for interactions between i and j atoms */
773 qq20 = _fjsp_mul_v2r8(iq2,jq0);
775 /* EWALD ELECTROSTATICS */
777 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
778 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
779 itab_tmp = _fjsp_dtox_v2r8(ewrt);
780 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
781 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
783 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
785 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
786 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
790 /* Update vectorial force */
791 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
792 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
793 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
795 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
796 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
797 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
799 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
801 /* Inner loop uses 120 flops */
808 j_coord_offsetA = DIM*jnrA;
810 /* load j atom coordinates */
811 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
814 /* Calculate displacement vector */
815 dx00 = _fjsp_sub_v2r8(ix0,jx0);
816 dy00 = _fjsp_sub_v2r8(iy0,jy0);
817 dz00 = _fjsp_sub_v2r8(iz0,jz0);
818 dx10 = _fjsp_sub_v2r8(ix1,jx0);
819 dy10 = _fjsp_sub_v2r8(iy1,jy0);
820 dz10 = _fjsp_sub_v2r8(iz1,jz0);
821 dx20 = _fjsp_sub_v2r8(ix2,jx0);
822 dy20 = _fjsp_sub_v2r8(iy2,jy0);
823 dz20 = _fjsp_sub_v2r8(iz2,jz0);
825 /* Calculate squared distance and things based on it */
826 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
827 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
828 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
830 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
831 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
832 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
834 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
835 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
836 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
838 /* Load parameters for j particles */
839 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
841 fjx0 = _fjsp_setzero_v2r8();
842 fjy0 = _fjsp_setzero_v2r8();
843 fjz0 = _fjsp_setzero_v2r8();
845 /**************************
846 * CALCULATE INTERACTIONS *
847 **************************/
849 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
851 /* Compute parameters for interactions between i and j atoms */
852 qq00 = _fjsp_mul_v2r8(iq0,jq0);
854 /* EWALD ELECTROSTATICS */
856 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
857 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
858 itab_tmp = _fjsp_dtox_v2r8(ewrt);
859 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
860 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
862 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
863 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
864 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
868 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
870 /* Update vectorial force */
871 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
872 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
873 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
875 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
876 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
877 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
879 /**************************
880 * CALCULATE INTERACTIONS *
881 **************************/
883 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
885 /* Compute parameters for interactions between i and j atoms */
886 qq10 = _fjsp_mul_v2r8(iq1,jq0);
888 /* EWALD ELECTROSTATICS */
890 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
891 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
892 itab_tmp = _fjsp_dtox_v2r8(ewrt);
893 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
894 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
896 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
897 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
898 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
902 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
904 /* Update vectorial force */
905 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
906 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
907 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
909 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
910 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
911 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
913 /**************************
914 * CALCULATE INTERACTIONS *
915 **************************/
917 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
919 /* Compute parameters for interactions between i and j atoms */
920 qq20 = _fjsp_mul_v2r8(iq2,jq0);
922 /* EWALD ELECTROSTATICS */
924 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
925 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
926 itab_tmp = _fjsp_dtox_v2r8(ewrt);
927 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
928 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
930 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
931 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
932 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
936 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
938 /* Update vectorial force */
939 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
940 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
941 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
943 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
944 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
945 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
947 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
949 /* Inner loop uses 120 flops */
952 /* End of innermost loop */
954 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
955 f+i_coord_offset,fshift+i_shift_offset);
957 /* Increment number of inner iterations */
958 inneriter += j_index_end - j_index_start;
960 /* Outer loop uses 18 flops */
963 /* Increment number of outer iterations */
966 /* Update outer/inner flops */
968 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*120);