2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, 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 sse2_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_sse2_double.h"
50 #include "kernelutil_x86_sse2_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwNone_GeomW3P1_VF_sse2_double
54 * Electrostatics interaction: ReactionField
55 * VdW interaction: None
56 * Geometry: Water3-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecRFCut_VdwNone_GeomW3P1_VF_sse2_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
77 int j_coord_offsetA,j_coord_offsetB;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
85 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
87 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
88 int vdwjidx0A,vdwjidx0B;
89 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
90 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
91 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
92 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
93 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
95 __m128d dummy_mask,cutoff_mask;
96 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
97 __m128d one = _mm_set1_pd(1.0);
98 __m128d two = _mm_set1_pd(2.0);
104 jindex = nlist->jindex;
106 shiftidx = nlist->shift;
108 shiftvec = fr->shift_vec[0];
109 fshift = fr->fshift[0];
110 facel = _mm_set1_pd(fr->epsfac);
111 charge = mdatoms->chargeA;
112 krf = _mm_set1_pd(fr->ic->k_rf);
113 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
114 crf = _mm_set1_pd(fr->ic->c_rf);
116 /* Setup water-specific parameters */
117 inr = nlist->iinr[0];
118 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
119 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
120 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
122 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
123 rcutoff_scalar = fr->rcoulomb;
124 rcutoff = _mm_set1_pd(rcutoff_scalar);
125 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
127 /* Avoid stupid compiler warnings */
135 /* Start outer loop over neighborlists */
136 for(iidx=0; iidx<nri; iidx++)
138 /* Load shift vector for this list */
139 i_shift_offset = DIM*shiftidx[iidx];
141 /* Load limits for loop over neighbors */
142 j_index_start = jindex[iidx];
143 j_index_end = jindex[iidx+1];
145 /* Get outer coordinate index */
147 i_coord_offset = DIM*inr;
149 /* Load i particle coords and add shift vector */
150 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
151 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
153 fix0 = _mm_setzero_pd();
154 fiy0 = _mm_setzero_pd();
155 fiz0 = _mm_setzero_pd();
156 fix1 = _mm_setzero_pd();
157 fiy1 = _mm_setzero_pd();
158 fiz1 = _mm_setzero_pd();
159 fix2 = _mm_setzero_pd();
160 fiy2 = _mm_setzero_pd();
161 fiz2 = _mm_setzero_pd();
163 /* Reset potential sums */
164 velecsum = _mm_setzero_pd();
166 /* Start inner kernel loop */
167 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
170 /* Get j neighbor index, and coordinate index */
173 j_coord_offsetA = DIM*jnrA;
174 j_coord_offsetB = DIM*jnrB;
176 /* load j atom coordinates */
177 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
180 /* Calculate displacement vector */
181 dx00 = _mm_sub_pd(ix0,jx0);
182 dy00 = _mm_sub_pd(iy0,jy0);
183 dz00 = _mm_sub_pd(iz0,jz0);
184 dx10 = _mm_sub_pd(ix1,jx0);
185 dy10 = _mm_sub_pd(iy1,jy0);
186 dz10 = _mm_sub_pd(iz1,jz0);
187 dx20 = _mm_sub_pd(ix2,jx0);
188 dy20 = _mm_sub_pd(iy2,jy0);
189 dz20 = _mm_sub_pd(iz2,jz0);
191 /* Calculate squared distance and things based on it */
192 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
193 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
194 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
196 rinv00 = gmx_mm_invsqrt_pd(rsq00);
197 rinv10 = gmx_mm_invsqrt_pd(rsq10);
198 rinv20 = gmx_mm_invsqrt_pd(rsq20);
200 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
201 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
202 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
204 /* Load parameters for j particles */
205 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
207 fjx0 = _mm_setzero_pd();
208 fjy0 = _mm_setzero_pd();
209 fjz0 = _mm_setzero_pd();
211 /**************************
212 * CALCULATE INTERACTIONS *
213 **************************/
215 if (gmx_mm_any_lt(rsq00,rcutoff2))
218 /* Compute parameters for interactions between i and j atoms */
219 qq00 = _mm_mul_pd(iq0,jq0);
221 /* REACTION-FIELD ELECTROSTATICS */
222 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
223 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
225 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
227 /* Update potential sum for this i atom from the interaction with this j atom. */
228 velec = _mm_and_pd(velec,cutoff_mask);
229 velecsum = _mm_add_pd(velecsum,velec);
233 fscal = _mm_and_pd(fscal,cutoff_mask);
235 /* Calculate temporary vectorial force */
236 tx = _mm_mul_pd(fscal,dx00);
237 ty = _mm_mul_pd(fscal,dy00);
238 tz = _mm_mul_pd(fscal,dz00);
240 /* Update vectorial force */
241 fix0 = _mm_add_pd(fix0,tx);
242 fiy0 = _mm_add_pd(fiy0,ty);
243 fiz0 = _mm_add_pd(fiz0,tz);
245 fjx0 = _mm_add_pd(fjx0,tx);
246 fjy0 = _mm_add_pd(fjy0,ty);
247 fjz0 = _mm_add_pd(fjz0,tz);
251 /**************************
252 * CALCULATE INTERACTIONS *
253 **************************/
255 if (gmx_mm_any_lt(rsq10,rcutoff2))
258 /* Compute parameters for interactions between i and j atoms */
259 qq10 = _mm_mul_pd(iq1,jq0);
261 /* REACTION-FIELD ELECTROSTATICS */
262 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
263 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
265 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
267 /* Update potential sum for this i atom from the interaction with this j atom. */
268 velec = _mm_and_pd(velec,cutoff_mask);
269 velecsum = _mm_add_pd(velecsum,velec);
273 fscal = _mm_and_pd(fscal,cutoff_mask);
275 /* Calculate temporary vectorial force */
276 tx = _mm_mul_pd(fscal,dx10);
277 ty = _mm_mul_pd(fscal,dy10);
278 tz = _mm_mul_pd(fscal,dz10);
280 /* Update vectorial force */
281 fix1 = _mm_add_pd(fix1,tx);
282 fiy1 = _mm_add_pd(fiy1,ty);
283 fiz1 = _mm_add_pd(fiz1,tz);
285 fjx0 = _mm_add_pd(fjx0,tx);
286 fjy0 = _mm_add_pd(fjy0,ty);
287 fjz0 = _mm_add_pd(fjz0,tz);
291 /**************************
292 * CALCULATE INTERACTIONS *
293 **************************/
295 if (gmx_mm_any_lt(rsq20,rcutoff2))
298 /* Compute parameters for interactions between i and j atoms */
299 qq20 = _mm_mul_pd(iq2,jq0);
301 /* REACTION-FIELD ELECTROSTATICS */
302 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
303 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
305 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
307 /* Update potential sum for this i atom from the interaction with this j atom. */
308 velec = _mm_and_pd(velec,cutoff_mask);
309 velecsum = _mm_add_pd(velecsum,velec);
313 fscal = _mm_and_pd(fscal,cutoff_mask);
315 /* Calculate temporary vectorial force */
316 tx = _mm_mul_pd(fscal,dx20);
317 ty = _mm_mul_pd(fscal,dy20);
318 tz = _mm_mul_pd(fscal,dz20);
320 /* Update vectorial force */
321 fix2 = _mm_add_pd(fix2,tx);
322 fiy2 = _mm_add_pd(fiy2,ty);
323 fiz2 = _mm_add_pd(fiz2,tz);
325 fjx0 = _mm_add_pd(fjx0,tx);
326 fjy0 = _mm_add_pd(fjy0,ty);
327 fjz0 = _mm_add_pd(fjz0,tz);
331 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
333 /* Inner loop uses 111 flops */
340 j_coord_offsetA = DIM*jnrA;
342 /* load j atom coordinates */
343 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
346 /* Calculate displacement vector */
347 dx00 = _mm_sub_pd(ix0,jx0);
348 dy00 = _mm_sub_pd(iy0,jy0);
349 dz00 = _mm_sub_pd(iz0,jz0);
350 dx10 = _mm_sub_pd(ix1,jx0);
351 dy10 = _mm_sub_pd(iy1,jy0);
352 dz10 = _mm_sub_pd(iz1,jz0);
353 dx20 = _mm_sub_pd(ix2,jx0);
354 dy20 = _mm_sub_pd(iy2,jy0);
355 dz20 = _mm_sub_pd(iz2,jz0);
357 /* Calculate squared distance and things based on it */
358 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
359 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
360 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
362 rinv00 = gmx_mm_invsqrt_pd(rsq00);
363 rinv10 = gmx_mm_invsqrt_pd(rsq10);
364 rinv20 = gmx_mm_invsqrt_pd(rsq20);
366 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
367 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
368 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
370 /* Load parameters for j particles */
371 jq0 = _mm_load_sd(charge+jnrA+0);
373 fjx0 = _mm_setzero_pd();
374 fjy0 = _mm_setzero_pd();
375 fjz0 = _mm_setzero_pd();
377 /**************************
378 * CALCULATE INTERACTIONS *
379 **************************/
381 if (gmx_mm_any_lt(rsq00,rcutoff2))
384 /* Compute parameters for interactions between i and j atoms */
385 qq00 = _mm_mul_pd(iq0,jq0);
387 /* REACTION-FIELD ELECTROSTATICS */
388 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
389 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
391 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
393 /* Update potential sum for this i atom from the interaction with this j atom. */
394 velec = _mm_and_pd(velec,cutoff_mask);
395 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
396 velecsum = _mm_add_pd(velecsum,velec);
400 fscal = _mm_and_pd(fscal,cutoff_mask);
402 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
404 /* Calculate temporary vectorial force */
405 tx = _mm_mul_pd(fscal,dx00);
406 ty = _mm_mul_pd(fscal,dy00);
407 tz = _mm_mul_pd(fscal,dz00);
409 /* Update vectorial force */
410 fix0 = _mm_add_pd(fix0,tx);
411 fiy0 = _mm_add_pd(fiy0,ty);
412 fiz0 = _mm_add_pd(fiz0,tz);
414 fjx0 = _mm_add_pd(fjx0,tx);
415 fjy0 = _mm_add_pd(fjy0,ty);
416 fjz0 = _mm_add_pd(fjz0,tz);
420 /**************************
421 * CALCULATE INTERACTIONS *
422 **************************/
424 if (gmx_mm_any_lt(rsq10,rcutoff2))
427 /* Compute parameters for interactions between i and j atoms */
428 qq10 = _mm_mul_pd(iq1,jq0);
430 /* REACTION-FIELD ELECTROSTATICS */
431 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
432 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
434 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
436 /* Update potential sum for this i atom from the interaction with this j atom. */
437 velec = _mm_and_pd(velec,cutoff_mask);
438 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
439 velecsum = _mm_add_pd(velecsum,velec);
443 fscal = _mm_and_pd(fscal,cutoff_mask);
445 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
447 /* Calculate temporary vectorial force */
448 tx = _mm_mul_pd(fscal,dx10);
449 ty = _mm_mul_pd(fscal,dy10);
450 tz = _mm_mul_pd(fscal,dz10);
452 /* Update vectorial force */
453 fix1 = _mm_add_pd(fix1,tx);
454 fiy1 = _mm_add_pd(fiy1,ty);
455 fiz1 = _mm_add_pd(fiz1,tz);
457 fjx0 = _mm_add_pd(fjx0,tx);
458 fjy0 = _mm_add_pd(fjy0,ty);
459 fjz0 = _mm_add_pd(fjz0,tz);
463 /**************************
464 * CALCULATE INTERACTIONS *
465 **************************/
467 if (gmx_mm_any_lt(rsq20,rcutoff2))
470 /* Compute parameters for interactions between i and j atoms */
471 qq20 = _mm_mul_pd(iq2,jq0);
473 /* REACTION-FIELD ELECTROSTATICS */
474 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
475 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
477 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
479 /* Update potential sum for this i atom from the interaction with this j atom. */
480 velec = _mm_and_pd(velec,cutoff_mask);
481 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
482 velecsum = _mm_add_pd(velecsum,velec);
486 fscal = _mm_and_pd(fscal,cutoff_mask);
488 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
490 /* Calculate temporary vectorial force */
491 tx = _mm_mul_pd(fscal,dx20);
492 ty = _mm_mul_pd(fscal,dy20);
493 tz = _mm_mul_pd(fscal,dz20);
495 /* Update vectorial force */
496 fix2 = _mm_add_pd(fix2,tx);
497 fiy2 = _mm_add_pd(fiy2,ty);
498 fiz2 = _mm_add_pd(fiz2,tz);
500 fjx0 = _mm_add_pd(fjx0,tx);
501 fjy0 = _mm_add_pd(fjy0,ty);
502 fjz0 = _mm_add_pd(fjz0,tz);
506 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
508 /* Inner loop uses 111 flops */
511 /* End of innermost loop */
513 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
514 f+i_coord_offset,fshift+i_shift_offset);
517 /* Update potential energies */
518 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
520 /* Increment number of inner iterations */
521 inneriter += j_index_end - j_index_start;
523 /* Outer loop uses 19 flops */
526 /* Increment number of outer iterations */
529 /* Update outer/inner flops */
531 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*111);
534 * Gromacs nonbonded kernel: nb_kernel_ElecRFCut_VdwNone_GeomW3P1_F_sse2_double
535 * Electrostatics interaction: ReactionField
536 * VdW interaction: None
537 * Geometry: Water3-Particle
538 * Calculate force/pot: Force
541 nb_kernel_ElecRFCut_VdwNone_GeomW3P1_F_sse2_double
542 (t_nblist * gmx_restrict nlist,
543 rvec * gmx_restrict xx,
544 rvec * gmx_restrict ff,
545 t_forcerec * gmx_restrict fr,
546 t_mdatoms * gmx_restrict mdatoms,
547 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
548 t_nrnb * gmx_restrict nrnb)
550 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
551 * just 0 for non-waters.
552 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
553 * jnr indices corresponding to data put in the four positions in the SIMD register.
555 int i_shift_offset,i_coord_offset,outeriter,inneriter;
556 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
558 int j_coord_offsetA,j_coord_offsetB;
559 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
561 real *shiftvec,*fshift,*x,*f;
562 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
564 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
566 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
568 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
569 int vdwjidx0A,vdwjidx0B;
570 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
571 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
572 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
573 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
574 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
576 __m128d dummy_mask,cutoff_mask;
577 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
578 __m128d one = _mm_set1_pd(1.0);
579 __m128d two = _mm_set1_pd(2.0);
585 jindex = nlist->jindex;
587 shiftidx = nlist->shift;
589 shiftvec = fr->shift_vec[0];
590 fshift = fr->fshift[0];
591 facel = _mm_set1_pd(fr->epsfac);
592 charge = mdatoms->chargeA;
593 krf = _mm_set1_pd(fr->ic->k_rf);
594 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
595 crf = _mm_set1_pd(fr->ic->c_rf);
597 /* Setup water-specific parameters */
598 inr = nlist->iinr[0];
599 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
600 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
601 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
603 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
604 rcutoff_scalar = fr->rcoulomb;
605 rcutoff = _mm_set1_pd(rcutoff_scalar);
606 rcutoff2 = _mm_mul_pd(rcutoff,rcutoff);
608 /* Avoid stupid compiler warnings */
616 /* Start outer loop over neighborlists */
617 for(iidx=0; iidx<nri; iidx++)
619 /* Load shift vector for this list */
620 i_shift_offset = DIM*shiftidx[iidx];
622 /* Load limits for loop over neighbors */
623 j_index_start = jindex[iidx];
624 j_index_end = jindex[iidx+1];
626 /* Get outer coordinate index */
628 i_coord_offset = DIM*inr;
630 /* Load i particle coords and add shift vector */
631 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
632 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
634 fix0 = _mm_setzero_pd();
635 fiy0 = _mm_setzero_pd();
636 fiz0 = _mm_setzero_pd();
637 fix1 = _mm_setzero_pd();
638 fiy1 = _mm_setzero_pd();
639 fiz1 = _mm_setzero_pd();
640 fix2 = _mm_setzero_pd();
641 fiy2 = _mm_setzero_pd();
642 fiz2 = _mm_setzero_pd();
644 /* Start inner kernel loop */
645 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
648 /* Get j neighbor index, and coordinate index */
651 j_coord_offsetA = DIM*jnrA;
652 j_coord_offsetB = DIM*jnrB;
654 /* load j atom coordinates */
655 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
658 /* Calculate displacement vector */
659 dx00 = _mm_sub_pd(ix0,jx0);
660 dy00 = _mm_sub_pd(iy0,jy0);
661 dz00 = _mm_sub_pd(iz0,jz0);
662 dx10 = _mm_sub_pd(ix1,jx0);
663 dy10 = _mm_sub_pd(iy1,jy0);
664 dz10 = _mm_sub_pd(iz1,jz0);
665 dx20 = _mm_sub_pd(ix2,jx0);
666 dy20 = _mm_sub_pd(iy2,jy0);
667 dz20 = _mm_sub_pd(iz2,jz0);
669 /* Calculate squared distance and things based on it */
670 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
671 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
672 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
674 rinv00 = gmx_mm_invsqrt_pd(rsq00);
675 rinv10 = gmx_mm_invsqrt_pd(rsq10);
676 rinv20 = gmx_mm_invsqrt_pd(rsq20);
678 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
679 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
680 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
682 /* Load parameters for j particles */
683 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
685 fjx0 = _mm_setzero_pd();
686 fjy0 = _mm_setzero_pd();
687 fjz0 = _mm_setzero_pd();
689 /**************************
690 * CALCULATE INTERACTIONS *
691 **************************/
693 if (gmx_mm_any_lt(rsq00,rcutoff2))
696 /* Compute parameters for interactions between i and j atoms */
697 qq00 = _mm_mul_pd(iq0,jq0);
699 /* REACTION-FIELD ELECTROSTATICS */
700 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
702 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
706 fscal = _mm_and_pd(fscal,cutoff_mask);
708 /* Calculate temporary vectorial force */
709 tx = _mm_mul_pd(fscal,dx00);
710 ty = _mm_mul_pd(fscal,dy00);
711 tz = _mm_mul_pd(fscal,dz00);
713 /* Update vectorial force */
714 fix0 = _mm_add_pd(fix0,tx);
715 fiy0 = _mm_add_pd(fiy0,ty);
716 fiz0 = _mm_add_pd(fiz0,tz);
718 fjx0 = _mm_add_pd(fjx0,tx);
719 fjy0 = _mm_add_pd(fjy0,ty);
720 fjz0 = _mm_add_pd(fjz0,tz);
724 /**************************
725 * CALCULATE INTERACTIONS *
726 **************************/
728 if (gmx_mm_any_lt(rsq10,rcutoff2))
731 /* Compute parameters for interactions between i and j atoms */
732 qq10 = _mm_mul_pd(iq1,jq0);
734 /* REACTION-FIELD ELECTROSTATICS */
735 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
737 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
741 fscal = _mm_and_pd(fscal,cutoff_mask);
743 /* Calculate temporary vectorial force */
744 tx = _mm_mul_pd(fscal,dx10);
745 ty = _mm_mul_pd(fscal,dy10);
746 tz = _mm_mul_pd(fscal,dz10);
748 /* Update vectorial force */
749 fix1 = _mm_add_pd(fix1,tx);
750 fiy1 = _mm_add_pd(fiy1,ty);
751 fiz1 = _mm_add_pd(fiz1,tz);
753 fjx0 = _mm_add_pd(fjx0,tx);
754 fjy0 = _mm_add_pd(fjy0,ty);
755 fjz0 = _mm_add_pd(fjz0,tz);
759 /**************************
760 * CALCULATE INTERACTIONS *
761 **************************/
763 if (gmx_mm_any_lt(rsq20,rcutoff2))
766 /* Compute parameters for interactions between i and j atoms */
767 qq20 = _mm_mul_pd(iq2,jq0);
769 /* REACTION-FIELD ELECTROSTATICS */
770 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
772 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
776 fscal = _mm_and_pd(fscal,cutoff_mask);
778 /* Calculate temporary vectorial force */
779 tx = _mm_mul_pd(fscal,dx20);
780 ty = _mm_mul_pd(fscal,dy20);
781 tz = _mm_mul_pd(fscal,dz20);
783 /* Update vectorial force */
784 fix2 = _mm_add_pd(fix2,tx);
785 fiy2 = _mm_add_pd(fiy2,ty);
786 fiz2 = _mm_add_pd(fiz2,tz);
788 fjx0 = _mm_add_pd(fjx0,tx);
789 fjy0 = _mm_add_pd(fjy0,ty);
790 fjz0 = _mm_add_pd(fjz0,tz);
794 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
796 /* Inner loop uses 93 flops */
803 j_coord_offsetA = DIM*jnrA;
805 /* load j atom coordinates */
806 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
809 /* Calculate displacement vector */
810 dx00 = _mm_sub_pd(ix0,jx0);
811 dy00 = _mm_sub_pd(iy0,jy0);
812 dz00 = _mm_sub_pd(iz0,jz0);
813 dx10 = _mm_sub_pd(ix1,jx0);
814 dy10 = _mm_sub_pd(iy1,jy0);
815 dz10 = _mm_sub_pd(iz1,jz0);
816 dx20 = _mm_sub_pd(ix2,jx0);
817 dy20 = _mm_sub_pd(iy2,jy0);
818 dz20 = _mm_sub_pd(iz2,jz0);
820 /* Calculate squared distance and things based on it */
821 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
822 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
823 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
825 rinv00 = gmx_mm_invsqrt_pd(rsq00);
826 rinv10 = gmx_mm_invsqrt_pd(rsq10);
827 rinv20 = gmx_mm_invsqrt_pd(rsq20);
829 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
830 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
831 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
833 /* Load parameters for j particles */
834 jq0 = _mm_load_sd(charge+jnrA+0);
836 fjx0 = _mm_setzero_pd();
837 fjy0 = _mm_setzero_pd();
838 fjz0 = _mm_setzero_pd();
840 /**************************
841 * CALCULATE INTERACTIONS *
842 **************************/
844 if (gmx_mm_any_lt(rsq00,rcutoff2))
847 /* Compute parameters for interactions between i and j atoms */
848 qq00 = _mm_mul_pd(iq0,jq0);
850 /* REACTION-FIELD ELECTROSTATICS */
851 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
853 cutoff_mask = _mm_cmplt_pd(rsq00,rcutoff2);
857 fscal = _mm_and_pd(fscal,cutoff_mask);
859 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
861 /* Calculate temporary vectorial force */
862 tx = _mm_mul_pd(fscal,dx00);
863 ty = _mm_mul_pd(fscal,dy00);
864 tz = _mm_mul_pd(fscal,dz00);
866 /* Update vectorial force */
867 fix0 = _mm_add_pd(fix0,tx);
868 fiy0 = _mm_add_pd(fiy0,ty);
869 fiz0 = _mm_add_pd(fiz0,tz);
871 fjx0 = _mm_add_pd(fjx0,tx);
872 fjy0 = _mm_add_pd(fjy0,ty);
873 fjz0 = _mm_add_pd(fjz0,tz);
877 /**************************
878 * CALCULATE INTERACTIONS *
879 **************************/
881 if (gmx_mm_any_lt(rsq10,rcutoff2))
884 /* Compute parameters for interactions between i and j atoms */
885 qq10 = _mm_mul_pd(iq1,jq0);
887 /* REACTION-FIELD ELECTROSTATICS */
888 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
890 cutoff_mask = _mm_cmplt_pd(rsq10,rcutoff2);
894 fscal = _mm_and_pd(fscal,cutoff_mask);
896 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
898 /* Calculate temporary vectorial force */
899 tx = _mm_mul_pd(fscal,dx10);
900 ty = _mm_mul_pd(fscal,dy10);
901 tz = _mm_mul_pd(fscal,dz10);
903 /* Update vectorial force */
904 fix1 = _mm_add_pd(fix1,tx);
905 fiy1 = _mm_add_pd(fiy1,ty);
906 fiz1 = _mm_add_pd(fiz1,tz);
908 fjx0 = _mm_add_pd(fjx0,tx);
909 fjy0 = _mm_add_pd(fjy0,ty);
910 fjz0 = _mm_add_pd(fjz0,tz);
914 /**************************
915 * CALCULATE INTERACTIONS *
916 **************************/
918 if (gmx_mm_any_lt(rsq20,rcutoff2))
921 /* Compute parameters for interactions between i and j atoms */
922 qq20 = _mm_mul_pd(iq2,jq0);
924 /* REACTION-FIELD ELECTROSTATICS */
925 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
927 cutoff_mask = _mm_cmplt_pd(rsq20,rcutoff2);
931 fscal = _mm_and_pd(fscal,cutoff_mask);
933 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
935 /* Calculate temporary vectorial force */
936 tx = _mm_mul_pd(fscal,dx20);
937 ty = _mm_mul_pd(fscal,dy20);
938 tz = _mm_mul_pd(fscal,dz20);
940 /* Update vectorial force */
941 fix2 = _mm_add_pd(fix2,tx);
942 fiy2 = _mm_add_pd(fiy2,ty);
943 fiz2 = _mm_add_pd(fiz2,tz);
945 fjx0 = _mm_add_pd(fjx0,tx);
946 fjy0 = _mm_add_pd(fjy0,ty);
947 fjz0 = _mm_add_pd(fjz0,tz);
951 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
953 /* Inner loop uses 93 flops */
956 /* End of innermost loop */
958 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
959 f+i_coord_offset,fshift+i_shift_offset);
961 /* Increment number of inner iterations */
962 inneriter += j_index_end - j_index_start;
964 /* Outer loop uses 18 flops */
967 /* Increment number of outer iterations */
970 /* Update outer/inner flops */
972 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*93);