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.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
46 #include "gromacs/math/vec.h"
49 #include "kernelutil_sparc64_hpc_ace_double.h"
52 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
53 * Electrostatics interaction: Ewald
54 * VdW interaction: None
55 * Geometry: Water3-Particle
56 * Calculate force/pot: PotentialAndForce
59 nb_kernel_ElecEwSh_VdwNone_GeomW3P1_VF_sparc64_hpc_ace_double
60 (t_nblist * gmx_restrict nlist,
61 rvec * gmx_restrict xx,
62 rvec * gmx_restrict ff,
63 t_forcerec * gmx_restrict fr,
64 t_mdatoms * gmx_restrict mdatoms,
65 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
66 t_nrnb * gmx_restrict nrnb)
68 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
69 * just 0 for non-waters.
70 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
71 * jnr indices corresponding to data put in the four positions in the SIMD register.
73 int i_shift_offset,i_coord_offset,outeriter,inneriter;
74 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int j_coord_offsetA,j_coord_offsetB;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
82 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
84 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
86 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
87 int vdwjidx0A,vdwjidx0B;
88 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
89 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
90 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
91 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
92 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
94 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
97 _fjsp_v2r8 dummy_mask,cutoff_mask;
98 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
99 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
100 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
107 jindex = nlist->jindex;
109 shiftidx = nlist->shift;
111 shiftvec = fr->shift_vec[0];
112 fshift = fr->fshift[0];
113 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
114 charge = mdatoms->chargeA;
116 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
117 ewtab = fr->ic->tabq_coul_FDV0;
118 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
119 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
121 /* Setup water-specific parameters */
122 inr = nlist->iinr[0];
123 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
124 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
125 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
127 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
128 rcutoff_scalar = fr->rcoulomb;
129 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
130 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
132 /* Avoid stupid compiler warnings */
140 /* Start outer loop over neighborlists */
141 for(iidx=0; iidx<nri; iidx++)
143 /* Load shift vector for this list */
144 i_shift_offset = DIM*shiftidx[iidx];
146 /* Load limits for loop over neighbors */
147 j_index_start = jindex[iidx];
148 j_index_end = jindex[iidx+1];
150 /* Get outer coordinate index */
152 i_coord_offset = DIM*inr;
154 /* Load i particle coords and add shift vector */
155 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
156 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
158 fix0 = _fjsp_setzero_v2r8();
159 fiy0 = _fjsp_setzero_v2r8();
160 fiz0 = _fjsp_setzero_v2r8();
161 fix1 = _fjsp_setzero_v2r8();
162 fiy1 = _fjsp_setzero_v2r8();
163 fiz1 = _fjsp_setzero_v2r8();
164 fix2 = _fjsp_setzero_v2r8();
165 fiy2 = _fjsp_setzero_v2r8();
166 fiz2 = _fjsp_setzero_v2r8();
168 /* Reset potential sums */
169 velecsum = _fjsp_setzero_v2r8();
171 /* Start inner kernel loop */
172 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
175 /* Get j neighbor index, and coordinate index */
178 j_coord_offsetA = DIM*jnrA;
179 j_coord_offsetB = DIM*jnrB;
181 /* load j atom coordinates */
182 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
185 /* Calculate displacement vector */
186 dx00 = _fjsp_sub_v2r8(ix0,jx0);
187 dy00 = _fjsp_sub_v2r8(iy0,jy0);
188 dz00 = _fjsp_sub_v2r8(iz0,jz0);
189 dx10 = _fjsp_sub_v2r8(ix1,jx0);
190 dy10 = _fjsp_sub_v2r8(iy1,jy0);
191 dz10 = _fjsp_sub_v2r8(iz1,jz0);
192 dx20 = _fjsp_sub_v2r8(ix2,jx0);
193 dy20 = _fjsp_sub_v2r8(iy2,jy0);
194 dz20 = _fjsp_sub_v2r8(iz2,jz0);
196 /* Calculate squared distance and things based on it */
197 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
198 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
199 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
201 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
202 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
203 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
205 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
206 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
207 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
209 /* Load parameters for j particles */
210 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
212 fjx0 = _fjsp_setzero_v2r8();
213 fjy0 = _fjsp_setzero_v2r8();
214 fjz0 = _fjsp_setzero_v2r8();
216 /**************************
217 * CALCULATE INTERACTIONS *
218 **************************/
220 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
223 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
225 /* Compute parameters for interactions between i and j atoms */
226 qq00 = _fjsp_mul_v2r8(iq0,jq0);
228 /* EWALD ELECTROSTATICS */
230 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
231 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
232 itab_tmp = _fjsp_dtox_v2r8(ewrt);
233 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
234 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
236 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
237 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
238 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
239 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
240 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
241 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
242 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
243 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
244 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
245 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
247 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
249 /* Update potential sum for this i atom from the interaction with this j atom. */
250 velec = _fjsp_and_v2r8(velec,cutoff_mask);
251 velecsum = _fjsp_add_v2r8(velecsum,velec);
255 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
257 /* Update vectorial force */
258 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
259 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
260 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
262 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
263 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
264 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
268 /**************************
269 * CALCULATE INTERACTIONS *
270 **************************/
272 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
275 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
277 /* Compute parameters for interactions between i and j atoms */
278 qq10 = _fjsp_mul_v2r8(iq1,jq0);
280 /* EWALD ELECTROSTATICS */
282 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
283 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
284 itab_tmp = _fjsp_dtox_v2r8(ewrt);
285 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
286 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
288 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
289 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
290 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
291 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
292 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
293 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
294 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
295 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
296 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
297 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
299 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
301 /* Update potential sum for this i atom from the interaction with this j atom. */
302 velec = _fjsp_and_v2r8(velec,cutoff_mask);
303 velecsum = _fjsp_add_v2r8(velecsum,velec);
307 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
309 /* Update vectorial force */
310 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
311 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
312 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
314 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
315 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
316 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
320 /**************************
321 * CALCULATE INTERACTIONS *
322 **************************/
324 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
327 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
329 /* Compute parameters for interactions between i and j atoms */
330 qq20 = _fjsp_mul_v2r8(iq2,jq0);
332 /* EWALD ELECTROSTATICS */
334 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
335 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
336 itab_tmp = _fjsp_dtox_v2r8(ewrt);
337 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
338 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
340 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
341 ewtabD = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
342 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
343 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
344 ewtabFn = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
345 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
346 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
347 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
348 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
349 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
351 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
353 /* Update potential sum for this i atom from the interaction with this j atom. */
354 velec = _fjsp_and_v2r8(velec,cutoff_mask);
355 velecsum = _fjsp_add_v2r8(velecsum,velec);
359 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
361 /* Update vectorial force */
362 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
363 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
364 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
366 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
367 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
368 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
372 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
374 /* Inner loop uses 150 flops */
381 j_coord_offsetA = DIM*jnrA;
383 /* load j atom coordinates */
384 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
387 /* Calculate displacement vector */
388 dx00 = _fjsp_sub_v2r8(ix0,jx0);
389 dy00 = _fjsp_sub_v2r8(iy0,jy0);
390 dz00 = _fjsp_sub_v2r8(iz0,jz0);
391 dx10 = _fjsp_sub_v2r8(ix1,jx0);
392 dy10 = _fjsp_sub_v2r8(iy1,jy0);
393 dz10 = _fjsp_sub_v2r8(iz1,jz0);
394 dx20 = _fjsp_sub_v2r8(ix2,jx0);
395 dy20 = _fjsp_sub_v2r8(iy2,jy0);
396 dz20 = _fjsp_sub_v2r8(iz2,jz0);
398 /* Calculate squared distance and things based on it */
399 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
400 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
401 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
403 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
404 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
405 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
407 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
408 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
409 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
411 /* Load parameters for j particles */
412 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
414 fjx0 = _fjsp_setzero_v2r8();
415 fjy0 = _fjsp_setzero_v2r8();
416 fjz0 = _fjsp_setzero_v2r8();
418 /**************************
419 * CALCULATE INTERACTIONS *
420 **************************/
422 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
425 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
427 /* Compute parameters for interactions between i and j atoms */
428 qq00 = _fjsp_mul_v2r8(iq0,jq0);
430 /* EWALD ELECTROSTATICS */
432 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
433 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
434 itab_tmp = _fjsp_dtox_v2r8(ewrt);
435 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
436 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
438 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
439 ewtabD = _fjsp_setzero_v2r8();
440 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
441 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
442 ewtabFn = _fjsp_setzero_v2r8();
443 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
444 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
445 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
446 velec = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
447 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
449 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
451 /* Update potential sum for this i atom from the interaction with this j atom. */
452 velec = _fjsp_and_v2r8(velec,cutoff_mask);
453 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
454 velecsum = _fjsp_add_v2r8(velecsum,velec);
458 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
460 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
462 /* Update vectorial force */
463 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
464 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
465 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
467 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
468 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
469 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
473 /**************************
474 * CALCULATE INTERACTIONS *
475 **************************/
477 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
480 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
482 /* Compute parameters for interactions between i and j atoms */
483 qq10 = _fjsp_mul_v2r8(iq1,jq0);
485 /* EWALD ELECTROSTATICS */
487 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
488 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
489 itab_tmp = _fjsp_dtox_v2r8(ewrt);
490 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
491 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
493 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
494 ewtabD = _fjsp_setzero_v2r8();
495 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
496 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
497 ewtabFn = _fjsp_setzero_v2r8();
498 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
499 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
500 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
501 velec = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
502 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
504 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
506 /* Update potential sum for this i atom from the interaction with this j atom. */
507 velec = _fjsp_and_v2r8(velec,cutoff_mask);
508 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
509 velecsum = _fjsp_add_v2r8(velecsum,velec);
513 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
515 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
517 /* Update vectorial force */
518 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
519 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
520 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
522 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
523 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
524 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
528 /**************************
529 * CALCULATE INTERACTIONS *
530 **************************/
532 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
535 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
537 /* Compute parameters for interactions between i and j atoms */
538 qq20 = _fjsp_mul_v2r8(iq2,jq0);
540 /* EWALD ELECTROSTATICS */
542 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
543 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
544 itab_tmp = _fjsp_dtox_v2r8(ewrt);
545 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
546 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
548 ewtabF = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
549 ewtabD = _fjsp_setzero_v2r8();
550 GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
551 ewtabV = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
552 ewtabFn = _fjsp_setzero_v2r8();
553 GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
554 felec = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
555 velec = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
556 velec = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
557 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
559 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
561 /* Update potential sum for this i atom from the interaction with this j atom. */
562 velec = _fjsp_and_v2r8(velec,cutoff_mask);
563 velec = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
564 velecsum = _fjsp_add_v2r8(velecsum,velec);
568 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
570 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
572 /* Update vectorial force */
573 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
574 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
575 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
577 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
578 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
579 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
583 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
585 /* Inner loop uses 150 flops */
588 /* End of innermost loop */
590 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
591 f+i_coord_offset,fshift+i_shift_offset);
594 /* Update potential energies */
595 gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
597 /* Increment number of inner iterations */
598 inneriter += j_index_end - j_index_start;
600 /* Outer loop uses 19 flops */
603 /* Increment number of outer iterations */
606 /* Update outer/inner flops */
608 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*150);
611 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
612 * Electrostatics interaction: Ewald
613 * VdW interaction: None
614 * Geometry: Water3-Particle
615 * Calculate force/pot: Force
618 nb_kernel_ElecEwSh_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
619 (t_nblist * gmx_restrict nlist,
620 rvec * gmx_restrict xx,
621 rvec * gmx_restrict ff,
622 t_forcerec * gmx_restrict fr,
623 t_mdatoms * gmx_restrict mdatoms,
624 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
625 t_nrnb * gmx_restrict nrnb)
627 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
628 * just 0 for non-waters.
629 * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
630 * jnr indices corresponding to data put in the four positions in the SIMD register.
632 int i_shift_offset,i_coord_offset,outeriter,inneriter;
633 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
635 int j_coord_offsetA,j_coord_offsetB;
636 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
638 real *shiftvec,*fshift,*x,*f;
639 _fjsp_v2r8 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
641 _fjsp_v2r8 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
643 _fjsp_v2r8 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
645 _fjsp_v2r8 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
646 int vdwjidx0A,vdwjidx0B;
647 _fjsp_v2r8 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
648 _fjsp_v2r8 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
649 _fjsp_v2r8 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
650 _fjsp_v2r8 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
651 _fjsp_v2r8 velec,felec,velecsum,facel,crf,krf,krf2;
653 _fjsp_v2r8 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
656 _fjsp_v2r8 dummy_mask,cutoff_mask;
657 _fjsp_v2r8 one = gmx_fjsp_set1_v2r8(1.0);
658 _fjsp_v2r8 two = gmx_fjsp_set1_v2r8(2.0);
659 union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
666 jindex = nlist->jindex;
668 shiftidx = nlist->shift;
670 shiftvec = fr->shift_vec[0];
671 fshift = fr->fshift[0];
672 facel = gmx_fjsp_set1_v2r8(fr->epsfac);
673 charge = mdatoms->chargeA;
675 sh_ewald = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
676 ewtab = fr->ic->tabq_coul_F;
677 ewtabscale = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
678 ewtabhalfspace = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
680 /* Setup water-specific parameters */
681 inr = nlist->iinr[0];
682 iq0 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
683 iq1 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
684 iq2 = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
686 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
687 rcutoff_scalar = fr->rcoulomb;
688 rcutoff = gmx_fjsp_set1_v2r8(rcutoff_scalar);
689 rcutoff2 = _fjsp_mul_v2r8(rcutoff,rcutoff);
691 /* Avoid stupid compiler warnings */
699 /* Start outer loop over neighborlists */
700 for(iidx=0; iidx<nri; iidx++)
702 /* Load shift vector for this list */
703 i_shift_offset = DIM*shiftidx[iidx];
705 /* Load limits for loop over neighbors */
706 j_index_start = jindex[iidx];
707 j_index_end = jindex[iidx+1];
709 /* Get outer coordinate index */
711 i_coord_offset = DIM*inr;
713 /* Load i particle coords and add shift vector */
714 gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
715 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
717 fix0 = _fjsp_setzero_v2r8();
718 fiy0 = _fjsp_setzero_v2r8();
719 fiz0 = _fjsp_setzero_v2r8();
720 fix1 = _fjsp_setzero_v2r8();
721 fiy1 = _fjsp_setzero_v2r8();
722 fiz1 = _fjsp_setzero_v2r8();
723 fix2 = _fjsp_setzero_v2r8();
724 fiy2 = _fjsp_setzero_v2r8();
725 fiz2 = _fjsp_setzero_v2r8();
727 /* Start inner kernel loop */
728 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
731 /* Get j neighbor index, and coordinate index */
734 j_coord_offsetA = DIM*jnrA;
735 j_coord_offsetB = DIM*jnrB;
737 /* load j atom coordinates */
738 gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
741 /* Calculate displacement vector */
742 dx00 = _fjsp_sub_v2r8(ix0,jx0);
743 dy00 = _fjsp_sub_v2r8(iy0,jy0);
744 dz00 = _fjsp_sub_v2r8(iz0,jz0);
745 dx10 = _fjsp_sub_v2r8(ix1,jx0);
746 dy10 = _fjsp_sub_v2r8(iy1,jy0);
747 dz10 = _fjsp_sub_v2r8(iz1,jz0);
748 dx20 = _fjsp_sub_v2r8(ix2,jx0);
749 dy20 = _fjsp_sub_v2r8(iy2,jy0);
750 dz20 = _fjsp_sub_v2r8(iz2,jz0);
752 /* Calculate squared distance and things based on it */
753 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
754 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
755 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
757 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
758 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
759 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
761 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
762 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
763 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
765 /* Load parameters for j particles */
766 jq0 = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
768 fjx0 = _fjsp_setzero_v2r8();
769 fjy0 = _fjsp_setzero_v2r8();
770 fjz0 = _fjsp_setzero_v2r8();
772 /**************************
773 * CALCULATE INTERACTIONS *
774 **************************/
776 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
779 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
781 /* Compute parameters for interactions between i and j atoms */
782 qq00 = _fjsp_mul_v2r8(iq0,jq0);
784 /* EWALD ELECTROSTATICS */
786 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
787 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
788 itab_tmp = _fjsp_dtox_v2r8(ewrt);
789 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
790 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
792 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
794 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
795 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
797 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
801 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
803 /* Update vectorial force */
804 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
805 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
806 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
808 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
809 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
810 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
814 /**************************
815 * CALCULATE INTERACTIONS *
816 **************************/
818 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
821 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
823 /* Compute parameters for interactions between i and j atoms */
824 qq10 = _fjsp_mul_v2r8(iq1,jq0);
826 /* EWALD ELECTROSTATICS */
828 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
829 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
830 itab_tmp = _fjsp_dtox_v2r8(ewrt);
831 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
832 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
834 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
836 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
837 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
839 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
843 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
845 /* Update vectorial force */
846 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
847 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
848 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
850 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
851 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
852 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
856 /**************************
857 * CALCULATE INTERACTIONS *
858 **************************/
860 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
863 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
865 /* Compute parameters for interactions between i and j atoms */
866 qq20 = _fjsp_mul_v2r8(iq2,jq0);
868 /* EWALD ELECTROSTATICS */
870 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
871 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
872 itab_tmp = _fjsp_dtox_v2r8(ewrt);
873 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
874 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
876 gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
878 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
879 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
881 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
885 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
887 /* Update vectorial force */
888 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
889 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
890 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
892 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
893 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
894 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
898 gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
900 /* Inner loop uses 129 flops */
907 j_coord_offsetA = DIM*jnrA;
909 /* load j atom coordinates */
910 gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
913 /* Calculate displacement vector */
914 dx00 = _fjsp_sub_v2r8(ix0,jx0);
915 dy00 = _fjsp_sub_v2r8(iy0,jy0);
916 dz00 = _fjsp_sub_v2r8(iz0,jz0);
917 dx10 = _fjsp_sub_v2r8(ix1,jx0);
918 dy10 = _fjsp_sub_v2r8(iy1,jy0);
919 dz10 = _fjsp_sub_v2r8(iz1,jz0);
920 dx20 = _fjsp_sub_v2r8(ix2,jx0);
921 dy20 = _fjsp_sub_v2r8(iy2,jy0);
922 dz20 = _fjsp_sub_v2r8(iz2,jz0);
924 /* Calculate squared distance and things based on it */
925 rsq00 = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
926 rsq10 = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
927 rsq20 = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
929 rinv00 = gmx_fjsp_invsqrt_v2r8(rsq00);
930 rinv10 = gmx_fjsp_invsqrt_v2r8(rsq10);
931 rinv20 = gmx_fjsp_invsqrt_v2r8(rsq20);
933 rinvsq00 = _fjsp_mul_v2r8(rinv00,rinv00);
934 rinvsq10 = _fjsp_mul_v2r8(rinv10,rinv10);
935 rinvsq20 = _fjsp_mul_v2r8(rinv20,rinv20);
937 /* Load parameters for j particles */
938 jq0 = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
940 fjx0 = _fjsp_setzero_v2r8();
941 fjy0 = _fjsp_setzero_v2r8();
942 fjz0 = _fjsp_setzero_v2r8();
944 /**************************
945 * CALCULATE INTERACTIONS *
946 **************************/
948 if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
951 r00 = _fjsp_mul_v2r8(rsq00,rinv00);
953 /* Compute parameters for interactions between i and j atoms */
954 qq00 = _fjsp_mul_v2r8(iq0,jq0);
956 /* EWALD ELECTROSTATICS */
958 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
959 ewrt = _fjsp_mul_v2r8(r00,ewtabscale);
960 itab_tmp = _fjsp_dtox_v2r8(ewrt);
961 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
962 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
964 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
965 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
966 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
968 cutoff_mask = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
972 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
974 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
976 /* Update vectorial force */
977 fix0 = _fjsp_madd_v2r8(dx00,fscal,fix0);
978 fiy0 = _fjsp_madd_v2r8(dy00,fscal,fiy0);
979 fiz0 = _fjsp_madd_v2r8(dz00,fscal,fiz0);
981 fjx0 = _fjsp_madd_v2r8(dx00,fscal,fjx0);
982 fjy0 = _fjsp_madd_v2r8(dy00,fscal,fjy0);
983 fjz0 = _fjsp_madd_v2r8(dz00,fscal,fjz0);
987 /**************************
988 * CALCULATE INTERACTIONS *
989 **************************/
991 if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
994 r10 = _fjsp_mul_v2r8(rsq10,rinv10);
996 /* Compute parameters for interactions between i and j atoms */
997 qq10 = _fjsp_mul_v2r8(iq1,jq0);
999 /* EWALD ELECTROSTATICS */
1001 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1002 ewrt = _fjsp_mul_v2r8(r10,ewtabscale);
1003 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1004 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1005 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1007 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1008 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1009 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1011 cutoff_mask = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1015 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1017 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1019 /* Update vectorial force */
1020 fix1 = _fjsp_madd_v2r8(dx10,fscal,fix1);
1021 fiy1 = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1022 fiz1 = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1024 fjx0 = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1025 fjy0 = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1026 fjz0 = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1030 /**************************
1031 * CALCULATE INTERACTIONS *
1032 **************************/
1034 if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1037 r20 = _fjsp_mul_v2r8(rsq20,rinv20);
1039 /* Compute parameters for interactions between i and j atoms */
1040 qq20 = _fjsp_mul_v2r8(iq2,jq0);
1042 /* EWALD ELECTROSTATICS */
1044 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1045 ewrt = _fjsp_mul_v2r8(r20,ewtabscale);
1046 itab_tmp = _fjsp_dtox_v2r8(ewrt);
1047 eweps = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1048 _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1050 gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1051 felec = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1052 felec = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1054 cutoff_mask = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1058 fscal = _fjsp_and_v2r8(fscal,cutoff_mask);
1060 fscal = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1062 /* Update vectorial force */
1063 fix2 = _fjsp_madd_v2r8(dx20,fscal,fix2);
1064 fiy2 = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1065 fiz2 = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1067 fjx0 = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1068 fjy0 = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1069 fjz0 = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1073 gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1075 /* Inner loop uses 129 flops */
1078 /* End of innermost loop */
1080 gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1081 f+i_coord_offset,fshift+i_shift_offset);
1083 /* Increment number of inner iterations */
1084 inneriter += j_index_end - j_index_start;
1086 /* Outer loop uses 18 flops */
1089 /* Increment number of outer iterations */
1092 /* Update outer/inner flops */
1094 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*129);