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 "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "kernelutil_sparc64_hpc_ace_double.h"
50 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
51 * Electrostatics interaction: Ewald
52 * VdW interaction: None
53 * Geometry: Water4-Particle
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecEwSh_VdwNone_GeomW4P1_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 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
82 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
84 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
85 int vdwjidx0A,vdwjidx0B;
86 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
87 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
88 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
89 _fjsp_v2r8 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
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 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
122 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
123 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
125 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
126 rcutoff_scalar = fr->rcoulomb;
127 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
128 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
130 /* Avoid stupid compiler warnings */
138 /* Start outer loop over neighborlists */
139 for(iidx=0; iidx<nri; iidx++)
141 /* Load shift vector for this list */
142 i_shift_offset = DIM*shiftidx[iidx];
144 /* Load limits for loop over neighbors */
145 j_index_start = jindex[iidx];
146 j_index_end = jindex[iidx+1];
148 /* Get outer coordinate index */
150 i_coord_offset = DIM*inr;
152 /* Load i particle coords and add shift vector */
153 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
154 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
156 fix1 = _fjsp_setzero_v2r8();
157 fiy1 = _fjsp_setzero_v2r8();
158 fiz1 = _fjsp_setzero_v2r8();
159 fix2 = _fjsp_setzero_v2r8();
160 fiy2 = _fjsp_setzero_v2r8();
161 fiz2 = _fjsp_setzero_v2r8();
162 fix3 = _fjsp_setzero_v2r8();
163 fiy3 = _fjsp_setzero_v2r8();
164 fiz3 = _fjsp_setzero_v2r8();
166 /* Reset potential sums */
167 velecsum = _fjsp_setzero_v2r8();
169 /* Start inner kernel loop */
170 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
173 /* Get j neighbor index, and coordinate index */
176 j_coord_offsetA = DIM*jnrA;
177 j_coord_offsetB = DIM*jnrB;
179 /* load j atom coordinates */
180 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
183 /* Calculate displacement vector */
184 dx10 = _fjsp_sub_v2r8(ix1,jx0);
185 dy10 = _fjsp_sub_v2r8(iy1,jy0);
186 dz10 = _fjsp_sub_v2r8(iz1,jz0);
187 dx20 = _fjsp_sub_v2r8(ix2,jx0);
188 dy20 = _fjsp_sub_v2r8(iy2,jy0);
189 dz20 = _fjsp_sub_v2r8(iz2,jz0);
190 dx30 = _fjsp_sub_v2r8(ix3,jx0);
191 dy30 = _fjsp_sub_v2r8(iy3,jy0);
192 dz30 = _fjsp_sub_v2r8(iz3,jz0);
194 /* Calculate squared distance and things based on it */
195 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
196 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
197 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
199 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
200 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
201 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
203 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
204 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
205 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
207 /* Load parameters for j particles */
208 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
210 fjx0 = _fjsp_setzero_v2r8();
211 fjy0 = _fjsp_setzero_v2r8();
212 fjz0 = _fjsp_setzero_v2r8();
214 /**************************
215 * CALCULATE INTERACTIONS *
216 **************************/
218 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
221 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
223 /* Compute parameters for interactions between i and j atoms */
224 qq10 = _fjsp_mul_v2r8(iq1,jq0);
226 /* EWALD ELECTROSTATICS */
228 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
229 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
230 itab_tmp = _fjsp_dtox_v2r8(ewrt);
231 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
232 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
234 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
235 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
236 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
237 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
238 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
239 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
240 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
241 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
242 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
243 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
245 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
247 /* Update potential sum for this i atom from the interaction with this j atom. */
248 velec = _fjsp_and_v2r8(velec,cutoff_mask);
249 velecsum = _fjsp_add_v2r8(velecsum,velec);
253 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
255 /* Update vectorial force */
256 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
257 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
258 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
260 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
261 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
262 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
266 /**************************
267 * CALCULATE INTERACTIONS *
268 **************************/
270 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
273 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
275 /* Compute parameters for interactions between i and j atoms */
276 qq20 = _fjsp_mul_v2r8(iq2,jq0);
278 /* EWALD ELECTROSTATICS */
280 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
281 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
282 itab_tmp = _fjsp_dtox_v2r8(ewrt);
283 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
284 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
286 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
287 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
288 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
289 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
290 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
291 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
292 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
293 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
294 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
295 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
297 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
299 /* Update potential sum for this i atom from the interaction with this j atom. */
300 velec = _fjsp_and_v2r8(velec,cutoff_mask);
301 velecsum = _fjsp_add_v2r8(velecsum,velec);
305 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
307 /* Update vectorial force */
308 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
309 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
310 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
312 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
313 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
314 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
318 /**************************
319 * CALCULATE INTERACTIONS *
320 **************************/
322 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
325 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
327 /* Compute parameters for interactions between i and j atoms */
328 qq30 = _fjsp_mul_v2r8(iq3,jq0);
330 /* EWALD ELECTROSTATICS */
332 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
333 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
334 itab_tmp = _fjsp_dtox_v2r8(ewrt);
335 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
336 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
338 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
339 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
340 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
341 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
342 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
343 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
344 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
345 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
346 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv30,sh_ewald),velec));
347 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
349 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
351 /* Update potential sum for this i atom from the interaction with this j atom. */
352 velec = _fjsp_and_v2r8(velec,cutoff_mask);
353 velecsum = _fjsp_add_v2r8(velecsum,velec);
357 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
359 /* Update vectorial force */
360 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
361 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
362 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
364 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
365 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
366 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
370 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
372 /* Inner loop uses 150 flops */
379 j_coord_offsetA = DIM*jnrA;
381 /* load j atom coordinates */
382 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
385 /* Calculate displacement vector */
386 dx10 = _fjsp_sub_v2r8(ix1,jx0);
387 dy10 = _fjsp_sub_v2r8(iy1,jy0);
388 dz10 = _fjsp_sub_v2r8(iz1,jz0);
389 dx20 = _fjsp_sub_v2r8(ix2,jx0);
390 dy20 = _fjsp_sub_v2r8(iy2,jy0);
391 dz20 = _fjsp_sub_v2r8(iz2,jz0);
392 dx30 = _fjsp_sub_v2r8(ix3,jx0);
393 dy30 = _fjsp_sub_v2r8(iy3,jy0);
394 dz30 = _fjsp_sub_v2r8(iz3,jz0);
396 /* Calculate squared distance and things based on it */
397 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
398 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
399 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
401 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
402 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
403 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
405 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
406 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
407 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
409 /* Load parameters for j particles */
410 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
412 fjx0 = _fjsp_setzero_v2r8();
413 fjy0 = _fjsp_setzero_v2r8();
414 fjz0 = _fjsp_setzero_v2r8();
416 /**************************
417 * CALCULATE INTERACTIONS *
418 **************************/
420 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
423 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
425 /* Compute parameters for interactions between i and j atoms */
426 qq10 = _fjsp_mul_v2r8(iq1,jq0);
428 /* EWALD ELECTROSTATICS */
430 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
431 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
432 itab_tmp = _fjsp_dtox_v2r8(ewrt);
433 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
434 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
436 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
437 ewtabD = _fjsp_setzero_v2r8();
438 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
439 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
440 ewtabFn = _fjsp_setzero_v2r8();
441 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
442 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
443 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
444 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
445 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
447 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
449 /* Update potential sum for this i atom from the interaction with this j atom. */
450 velec = _fjsp_and_v2r8(velec,cutoff_mask);
451 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
452 velecsum = _fjsp_add_v2r8(velecsum,velec);
456 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
458 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
460 /* Update vectorial force */
461 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
462 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
463 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
465 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
466 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
467 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
471 /**************************
472 * CALCULATE INTERACTIONS *
473 **************************/
475 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
478 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
480 /* Compute parameters for interactions between i and j atoms */
481 qq20 = _fjsp_mul_v2r8(iq2,jq0);
483 /* EWALD ELECTROSTATICS */
485 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
486 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
487 itab_tmp = _fjsp_dtox_v2r8(ewrt);
488 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
489 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
491 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
492 ewtabD = _fjsp_setzero_v2r8();
493 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
494 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
495 ewtabFn = _fjsp_setzero_v2r8();
496 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
497 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
498 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
499 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
500 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
502 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
504 /* Update potential sum for this i atom from the interaction with this j atom. */
505 velec = _fjsp_and_v2r8(velec,cutoff_mask);
506 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
507 velecsum = _fjsp_add_v2r8(velecsum,velec);
511 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
513 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
515 /* Update vectorial force */
516 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
517 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
518 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
520 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
521 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
522 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
526 /**************************
527 * CALCULATE INTERACTIONS *
528 **************************/
530 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
533 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
535 /* Compute parameters for interactions between i and j atoms */
536 qq30 = _fjsp_mul_v2r8(iq3,jq0);
538 /* EWALD ELECTROSTATICS */
540 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
541 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
542 itab_tmp = _fjsp_dtox_v2r8(ewrt);
543 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
544 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
546 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
547 ewtabD = _fjsp_setzero_v2r8();
548 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
549 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
550 ewtabFn = _fjsp_setzero_v2r8();
551 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
552 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
553 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
554 velec = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv30,sh_ewald),velec));
555 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
557 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
559 /* Update potential sum for this i atom from the interaction with this j atom. */
560 velec = _fjsp_and_v2r8(velec,cutoff_mask);
561 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
562 velecsum = _fjsp_add_v2r8(velecsum,velec);
566 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
568 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
570 /* Update vectorial force */
571 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
572 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
573 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
575 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
576 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
577 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
581 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
583 /* Inner loop uses 150 flops */
586 /* End of innermost loop */
588 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
589 f+i_coord_offset+DIM,fshift+i_shift_offset);
592 /* Update potential energies */
593 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
595 /* Increment number of inner iterations */
596 inneriter += j_index_end - j_index_start;
598 /* Outer loop uses 19 flops */
601 /* Increment number of outer iterations */
604 /* Update outer/inner flops */
606 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*150);
609 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
610 * Electrostatics interaction: Ewald
611 * VdW interaction: None
612 * Geometry: Water4-Particle
613 * Calculate force/pot: Force
616 nb_kernel_ElecEwSh_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
617 (t_nblist * gmx_restrict nlist,
618 rvec * gmx_restrict xx,
619 rvec * gmx_restrict ff,
620 t_forcerec * gmx_restrict fr,
621 t_mdatoms * gmx_restrict mdatoms,
622 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
623 t_nrnb * gmx_restrict nrnb)
625 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
626 * just 0 for non-waters.
627 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
628 * jnr indices corresponding to data put in the four positions in the SIMD register.
630 int i_shift_offset,i_coord_offset,outeriter,inneriter;
631 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
633 int j_coord_offsetA,j_coord_offsetB;
634 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
636 real *shiftvec,*fshift,*x,*f;
637 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
639 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
641 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
643 _fjsp_v2r8 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
644 int vdwjidx0A,vdwjidx0B;
645 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
646 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
647 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
648 _fjsp_v2r8 dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
649 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
651 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
654 _fjsp_v2r8 dummy_mask,cutoff_mask;
655 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
656 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
657 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
664 jindex = nlist->jindex;
666 shiftidx = nlist->shift;
668 shiftvec = fr->shift_vec[0];
669 fshift = fr->fshift[0];
670 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
671 charge = mdatoms->chargeA;
673 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
674 ewtab = fr->ic->tabq_coul_F;
675 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
676 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
678 /* Setup water-specific parameters */
679 inr = nlist->iinr[0];
680 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
681 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
682 iq3 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
684 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
685 rcutoff_scalar = fr->rcoulomb;
686 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
687 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
689 /* Avoid stupid compiler warnings */
697 /* Start outer loop over neighborlists */
698 for(iidx=0; iidx<nri; iidx++)
700 /* Load shift vector for this list */
701 i_shift_offset = DIM*shiftidx[iidx];
703 /* Load limits for loop over neighbors */
704 j_index_start = jindex[iidx];
705 j_index_end = jindex[iidx+1];
707 /* Get outer coordinate index */
709 i_coord_offset = DIM*inr;
711 /* Load i particle coords and add shift vector */
712 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
713 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
715 fix1 = _fjsp_setzero_v2r8();
716 fiy1 = _fjsp_setzero_v2r8();
717 fiz1 = _fjsp_setzero_v2r8();
718 fix2 = _fjsp_setzero_v2r8();
719 fiy2 = _fjsp_setzero_v2r8();
720 fiz2 = _fjsp_setzero_v2r8();
721 fix3 = _fjsp_setzero_v2r8();
722 fiy3 = _fjsp_setzero_v2r8();
723 fiz3 = _fjsp_setzero_v2r8();
725 /* Start inner kernel loop */
726 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
729 /* Get j neighbor index, and coordinate index */
732 j_coord_offsetA = DIM*jnrA;
733 j_coord_offsetB = DIM*jnrB;
735 /* load j atom coordinates */
736 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
739 /* Calculate displacement vector */
740 dx10 = _fjsp_sub_v2r8(ix1,jx0);
741 dy10 = _fjsp_sub_v2r8(iy1,jy0);
742 dz10 = _fjsp_sub_v2r8(iz1,jz0);
743 dx20 = _fjsp_sub_v2r8(ix2,jx0);
744 dy20 = _fjsp_sub_v2r8(iy2,jy0);
745 dz20 = _fjsp_sub_v2r8(iz2,jz0);
746 dx30 = _fjsp_sub_v2r8(ix3,jx0);
747 dy30 = _fjsp_sub_v2r8(iy3,jy0);
748 dz30 = _fjsp_sub_v2r8(iz3,jz0);
750 /* Calculate squared distance and things based on it */
751 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
752 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
753 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
755 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
756 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
757 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
759 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
760 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
761 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
763 /* Load parameters for j particles */
764 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
766 fjx0 = _fjsp_setzero_v2r8();
767 fjy0 = _fjsp_setzero_v2r8();
768 fjz0 = _fjsp_setzero_v2r8();
770 /**************************
771 * CALCULATE INTERACTIONS *
772 **************************/
774 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
777 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
779 /* Compute parameters for interactions between i and j atoms */
780 qq10 = _fjsp_mul_v2r8(iq1,jq0);
782 /* EWALD ELECTROSTATICS */
784 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
785 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
786 itab_tmp = _fjsp_dtox_v2r8(ewrt);
787 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
788 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
790 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
792 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
793 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
795 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
799 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
801 /* Update vectorial force */
802 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
803 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
804 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
806 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
807 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
808 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
812 /**************************
813 * CALCULATE INTERACTIONS *
814 **************************/
816 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
819 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
821 /* Compute parameters for interactions between i and j atoms */
822 qq20 = _fjsp_mul_v2r8(iq2,jq0);
824 /* EWALD ELECTROSTATICS */
826 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
827 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
828 itab_tmp = _fjsp_dtox_v2r8(ewrt);
829 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
830 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
832 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
834 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
835 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
837 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
841 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
843 /* Update vectorial force */
844 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
845 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
846 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
848 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
849 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
850 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
854 /**************************
855 * CALCULATE INTERACTIONS *
856 **************************/
858 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
861 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
863 /* Compute parameters for interactions between i and j atoms */
864 qq30 = _fjsp_mul_v2r8(iq3,jq0);
866 /* EWALD ELECTROSTATICS */
868 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
869 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
870 itab_tmp = _fjsp_dtox_v2r8(ewrt);
871 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
872 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
874 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
876 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
877 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
879 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
883 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
885 /* Update vectorial force */
886 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
887 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
888 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
890 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
891 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
892 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
896 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
898 /* Inner loop uses 129 flops */
905 j_coord_offsetA = DIM*jnrA;
907 /* load j atom coordinates */
908 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
911 /* Calculate displacement vector */
912 dx10 = _fjsp_sub_v2r8(ix1,jx0);
913 dy10 = _fjsp_sub_v2r8(iy1,jy0);
914 dz10 = _fjsp_sub_v2r8(iz1,jz0);
915 dx20 = _fjsp_sub_v2r8(ix2,jx0);
916 dy20 = _fjsp_sub_v2r8(iy2,jy0);
917 dz20 = _fjsp_sub_v2r8(iz2,jz0);
918 dx30 = _fjsp_sub_v2r8(ix3,jx0);
919 dy30 = _fjsp_sub_v2r8(iy3,jy0);
920 dz30 = _fjsp_sub_v2r8(iz3,jz0);
922 /* Calculate squared distance and things based on it */
923 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
924 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
925 rsq30 = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
927 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
928 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
929 rinv30 = gmx_fjsp_invsqrt_v2r8(rsq30);
931 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
932 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
933 rinvsq30 = _fjsp_mul_v2r8(rinv30,rinv30);
935 /* Load parameters for j particles */
936 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
938 fjx0 = _fjsp_setzero_v2r8();
939 fjy0 = _fjsp_setzero_v2r8();
940 fjz0 = _fjsp_setzero_v2r8();
942 /**************************
943 * CALCULATE INTERACTIONS *
944 **************************/
946 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
949 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
951 /* Compute parameters for interactions between i and j atoms */
952 qq10 = _fjsp_mul_v2r8(iq1,jq0);
954 /* EWALD ELECTROSTATICS */
956 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
957 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
958 itab_tmp = _fjsp_dtox_v2r8(ewrt);
959 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
960 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
962 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
963 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
964 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
966 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
970 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
972 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
974 /* Update vectorial force */
975 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
976 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
977 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
979 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
980 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
981 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
985 /**************************
986 * CALCULATE INTERACTIONS *
987 **************************/
989 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
992 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
994 /* Compute parameters for interactions between i and j atoms */
995 qq20 = _fjsp_mul_v2r8(iq2,jq0);
997 /* EWALD ELECTROSTATICS */
999 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1000 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
1001 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1002 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1003 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1005 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1006 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1007 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1009 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1013 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1015 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1017 /* Update vectorial force */
1018 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1019 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1020 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1022 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1023 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1024 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1028 /**************************
1029 * CALCULATE INTERACTIONS *
1030 **************************/
1032 if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1035 r30 = _fjsp_mul_v2r8(rsq30,rinv30);
1037 /* Compute parameters for interactions between i and j atoms */
1038 qq30 = _fjsp_mul_v2r8(iq3,jq0);
1040 /* EWALD ELECTROSTATICS */
1042 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1043 ewrt = _fjsp_mul_v2r8(r30,ewtabscale);
1044 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1045 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1046 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1048 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1049 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1050 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1052 cutoff_mask = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1056 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1058 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1060 /* Update vectorial force */
1061 fix3 = _fjsp_madd_v2r8(dx30,fscal,fix3);
1062 fiy3 = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1063 fiz3 = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1065 fjx0 = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1066 fjy0 = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1067 fjz0 = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1071 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1073 /* Inner loop uses 129 flops */
1076 /* End of innermost loop */
1078 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1079 f+i_coord_offset+DIM,fshift+i_shift_offset);
1081 /* Increment number of inner iterations */
1082 inneriter += j_index_end - j_index_start;
1084 /* Outer loop uses 18 flops */
1087 /* Increment number of outer iterations */
1090 /* Update outer/inner flops */
1092 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*129);