2 * Note: this file was generated by the Gromacs sse2_double kernel generator.
4 * This source code is part of
8 * Copyright (c) 2001-2012, The GROMACS Development Team
10 * Gromacs is a library for molecular simulation and trajectory analysis,
11 * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
12 * a full list of developers and information, check out http://www.gromacs.org
14 * This program is free software; you can redistribute it and/or modify it under
15 * the terms of the GNU Lesser General Public License as published by the Free
16 * Software Foundation; either version 2 of the License, or (at your option) any
19 * To help fund GROMACS development, we humbly ask that you cite
20 * the papers people have written on it - you can find them on the website.
28 #include "../nb_kernel.h"
29 #include "types/simple.h"
33 #include "gmx_math_x86_sse2_double.h"
34 #include "kernelutil_x86_sse2_double.h"
37 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW4P1_VF_sse2_double
38 * Electrostatics interaction: ReactionField
39 * VdW interaction: None
40 * Geometry: Water4-Particle
41 * Calculate force/pot: PotentialAndForce
44 nb_kernel_ElecRF_VdwNone_GeomW4P1_VF_sse2_double
45 (t_nblist * gmx_restrict nlist,
46 rvec * gmx_restrict xx,
47 rvec * gmx_restrict ff,
48 t_forcerec * gmx_restrict fr,
49 t_mdatoms * gmx_restrict mdatoms,
50 nb_kernel_data_t * gmx_restrict kernel_data,
51 t_nrnb * gmx_restrict nrnb)
53 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
54 * just 0 for non-waters.
55 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
56 * jnr indices corresponding to data put in the four positions in the SIMD register.
58 int i_shift_offset,i_coord_offset,outeriter,inneriter;
59 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
61 int j_coord_offsetA,j_coord_offsetB;
62 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
64 real *shiftvec,*fshift,*x,*f;
65 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
67 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
69 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
71 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
72 int vdwjidx0A,vdwjidx0B;
73 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
74 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
75 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
76 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
77 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
79 __m128d dummy_mask,cutoff_mask;
80 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
81 __m128d one = _mm_set1_pd(1.0);
82 __m128d two = _mm_set1_pd(2.0);
88 jindex = nlist->jindex;
90 shiftidx = nlist->shift;
92 shiftvec = fr->shift_vec[0];
93 fshift = fr->fshift[0];
94 facel = _mm_set1_pd(fr->epsfac);
95 charge = mdatoms->chargeA;
96 krf = _mm_set1_pd(fr->ic->k_rf);
97 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
98 crf = _mm_set1_pd(fr->ic->c_rf);
100 /* Setup water-specific parameters */
101 inr = nlist->iinr[0];
102 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
103 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
104 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
106 /* Avoid stupid compiler warnings */
114 /* Start outer loop over neighborlists */
115 for(iidx=0; iidx<nri; iidx++)
117 /* Load shift vector for this list */
118 i_shift_offset = DIM*shiftidx[iidx];
120 /* Load limits for loop over neighbors */
121 j_index_start = jindex[iidx];
122 j_index_end = jindex[iidx+1];
124 /* Get outer coordinate index */
126 i_coord_offset = DIM*inr;
128 /* Load i particle coords and add shift vector */
129 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
130 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
132 fix1 = _mm_setzero_pd();
133 fiy1 = _mm_setzero_pd();
134 fiz1 = _mm_setzero_pd();
135 fix2 = _mm_setzero_pd();
136 fiy2 = _mm_setzero_pd();
137 fiz2 = _mm_setzero_pd();
138 fix3 = _mm_setzero_pd();
139 fiy3 = _mm_setzero_pd();
140 fiz3 = _mm_setzero_pd();
142 /* Reset potential sums */
143 velecsum = _mm_setzero_pd();
145 /* Start inner kernel loop */
146 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
149 /* Get j neighbor index, and coordinate index */
152 j_coord_offsetA = DIM*jnrA;
153 j_coord_offsetB = DIM*jnrB;
155 /* load j atom coordinates */
156 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
159 /* Calculate displacement vector */
160 dx10 = _mm_sub_pd(ix1,jx0);
161 dy10 = _mm_sub_pd(iy1,jy0);
162 dz10 = _mm_sub_pd(iz1,jz0);
163 dx20 = _mm_sub_pd(ix2,jx0);
164 dy20 = _mm_sub_pd(iy2,jy0);
165 dz20 = _mm_sub_pd(iz2,jz0);
166 dx30 = _mm_sub_pd(ix3,jx0);
167 dy30 = _mm_sub_pd(iy3,jy0);
168 dz30 = _mm_sub_pd(iz3,jz0);
170 /* Calculate squared distance and things based on it */
171 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
172 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
173 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
175 rinv10 = gmx_mm_invsqrt_pd(rsq10);
176 rinv20 = gmx_mm_invsqrt_pd(rsq20);
177 rinv30 = gmx_mm_invsqrt_pd(rsq30);
179 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
180 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
181 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
183 /* Load parameters for j particles */
184 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
186 fjx0 = _mm_setzero_pd();
187 fjy0 = _mm_setzero_pd();
188 fjz0 = _mm_setzero_pd();
190 /**************************
191 * CALCULATE INTERACTIONS *
192 **************************/
194 /* Compute parameters for interactions between i and j atoms */
195 qq10 = _mm_mul_pd(iq1,jq0);
197 /* REACTION-FIELD ELECTROSTATICS */
198 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
199 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
201 /* Update potential sum for this i atom from the interaction with this j atom. */
202 velecsum = _mm_add_pd(velecsum,velec);
206 /* Calculate temporary vectorial force */
207 tx = _mm_mul_pd(fscal,dx10);
208 ty = _mm_mul_pd(fscal,dy10);
209 tz = _mm_mul_pd(fscal,dz10);
211 /* Update vectorial force */
212 fix1 = _mm_add_pd(fix1,tx);
213 fiy1 = _mm_add_pd(fiy1,ty);
214 fiz1 = _mm_add_pd(fiz1,tz);
216 fjx0 = _mm_add_pd(fjx0,tx);
217 fjy0 = _mm_add_pd(fjy0,ty);
218 fjz0 = _mm_add_pd(fjz0,tz);
220 /**************************
221 * CALCULATE INTERACTIONS *
222 **************************/
224 /* Compute parameters for interactions between i and j atoms */
225 qq20 = _mm_mul_pd(iq2,jq0);
227 /* REACTION-FIELD ELECTROSTATICS */
228 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
229 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
231 /* Update potential sum for this i atom from the interaction with this j atom. */
232 velecsum = _mm_add_pd(velecsum,velec);
236 /* Calculate temporary vectorial force */
237 tx = _mm_mul_pd(fscal,dx20);
238 ty = _mm_mul_pd(fscal,dy20);
239 tz = _mm_mul_pd(fscal,dz20);
241 /* Update vectorial force */
242 fix2 = _mm_add_pd(fix2,tx);
243 fiy2 = _mm_add_pd(fiy2,ty);
244 fiz2 = _mm_add_pd(fiz2,tz);
246 fjx0 = _mm_add_pd(fjx0,tx);
247 fjy0 = _mm_add_pd(fjy0,ty);
248 fjz0 = _mm_add_pd(fjz0,tz);
250 /**************************
251 * CALCULATE INTERACTIONS *
252 **************************/
254 /* Compute parameters for interactions between i and j atoms */
255 qq30 = _mm_mul_pd(iq3,jq0);
257 /* REACTION-FIELD ELECTROSTATICS */
258 velec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_add_pd(rinv30,_mm_mul_pd(krf,rsq30)),crf));
259 felec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_mul_pd(rinv30,rinvsq30),krf2));
261 /* Update potential sum for this i atom from the interaction with this j atom. */
262 velecsum = _mm_add_pd(velecsum,velec);
266 /* Calculate temporary vectorial force */
267 tx = _mm_mul_pd(fscal,dx30);
268 ty = _mm_mul_pd(fscal,dy30);
269 tz = _mm_mul_pd(fscal,dz30);
271 /* Update vectorial force */
272 fix3 = _mm_add_pd(fix3,tx);
273 fiy3 = _mm_add_pd(fiy3,ty);
274 fiz3 = _mm_add_pd(fiz3,tz);
276 fjx0 = _mm_add_pd(fjx0,tx);
277 fjy0 = _mm_add_pd(fjy0,ty);
278 fjz0 = _mm_add_pd(fjz0,tz);
280 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
282 /* Inner loop uses 99 flops */
289 j_coord_offsetA = DIM*jnrA;
291 /* load j atom coordinates */
292 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
295 /* Calculate displacement vector */
296 dx10 = _mm_sub_pd(ix1,jx0);
297 dy10 = _mm_sub_pd(iy1,jy0);
298 dz10 = _mm_sub_pd(iz1,jz0);
299 dx20 = _mm_sub_pd(ix2,jx0);
300 dy20 = _mm_sub_pd(iy2,jy0);
301 dz20 = _mm_sub_pd(iz2,jz0);
302 dx30 = _mm_sub_pd(ix3,jx0);
303 dy30 = _mm_sub_pd(iy3,jy0);
304 dz30 = _mm_sub_pd(iz3,jz0);
306 /* Calculate squared distance and things based on it */
307 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
308 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
309 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
311 rinv10 = gmx_mm_invsqrt_pd(rsq10);
312 rinv20 = gmx_mm_invsqrt_pd(rsq20);
313 rinv30 = gmx_mm_invsqrt_pd(rsq30);
315 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
316 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
317 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
319 /* Load parameters for j particles */
320 jq0 = _mm_load_sd(charge+jnrA+0);
322 fjx0 = _mm_setzero_pd();
323 fjy0 = _mm_setzero_pd();
324 fjz0 = _mm_setzero_pd();
326 /**************************
327 * CALCULATE INTERACTIONS *
328 **************************/
330 /* Compute parameters for interactions between i and j atoms */
331 qq10 = _mm_mul_pd(iq1,jq0);
333 /* REACTION-FIELD ELECTROSTATICS */
334 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
335 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
337 /* Update potential sum for this i atom from the interaction with this j atom. */
338 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
339 velecsum = _mm_add_pd(velecsum,velec);
343 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
345 /* Calculate temporary vectorial force */
346 tx = _mm_mul_pd(fscal,dx10);
347 ty = _mm_mul_pd(fscal,dy10);
348 tz = _mm_mul_pd(fscal,dz10);
350 /* Update vectorial force */
351 fix1 = _mm_add_pd(fix1,tx);
352 fiy1 = _mm_add_pd(fiy1,ty);
353 fiz1 = _mm_add_pd(fiz1,tz);
355 fjx0 = _mm_add_pd(fjx0,tx);
356 fjy0 = _mm_add_pd(fjy0,ty);
357 fjz0 = _mm_add_pd(fjz0,tz);
359 /**************************
360 * CALCULATE INTERACTIONS *
361 **************************/
363 /* Compute parameters for interactions between i and j atoms */
364 qq20 = _mm_mul_pd(iq2,jq0);
366 /* REACTION-FIELD ELECTROSTATICS */
367 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
368 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
370 /* Update potential sum for this i atom from the interaction with this j atom. */
371 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
372 velecsum = _mm_add_pd(velecsum,velec);
376 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
378 /* Calculate temporary vectorial force */
379 tx = _mm_mul_pd(fscal,dx20);
380 ty = _mm_mul_pd(fscal,dy20);
381 tz = _mm_mul_pd(fscal,dz20);
383 /* Update vectorial force */
384 fix2 = _mm_add_pd(fix2,tx);
385 fiy2 = _mm_add_pd(fiy2,ty);
386 fiz2 = _mm_add_pd(fiz2,tz);
388 fjx0 = _mm_add_pd(fjx0,tx);
389 fjy0 = _mm_add_pd(fjy0,ty);
390 fjz0 = _mm_add_pd(fjz0,tz);
392 /**************************
393 * CALCULATE INTERACTIONS *
394 **************************/
396 /* Compute parameters for interactions between i and j atoms */
397 qq30 = _mm_mul_pd(iq3,jq0);
399 /* REACTION-FIELD ELECTROSTATICS */
400 velec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_add_pd(rinv30,_mm_mul_pd(krf,rsq30)),crf));
401 felec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_mul_pd(rinv30,rinvsq30),krf2));
403 /* Update potential sum for this i atom from the interaction with this j atom. */
404 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
405 velecsum = _mm_add_pd(velecsum,velec);
409 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
411 /* Calculate temporary vectorial force */
412 tx = _mm_mul_pd(fscal,dx30);
413 ty = _mm_mul_pd(fscal,dy30);
414 tz = _mm_mul_pd(fscal,dz30);
416 /* Update vectorial force */
417 fix3 = _mm_add_pd(fix3,tx);
418 fiy3 = _mm_add_pd(fiy3,ty);
419 fiz3 = _mm_add_pd(fiz3,tz);
421 fjx0 = _mm_add_pd(fjx0,tx);
422 fjy0 = _mm_add_pd(fjy0,ty);
423 fjz0 = _mm_add_pd(fjz0,tz);
425 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
427 /* Inner loop uses 99 flops */
430 /* End of innermost loop */
432 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
433 f+i_coord_offset+DIM,fshift+i_shift_offset);
436 /* Update potential energies */
437 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
439 /* Increment number of inner iterations */
440 inneriter += j_index_end - j_index_start;
442 /* Outer loop uses 19 flops */
445 /* Increment number of outer iterations */
448 /* Update outer/inner flops */
450 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*99);
453 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW4P1_F_sse2_double
454 * Electrostatics interaction: ReactionField
455 * VdW interaction: None
456 * Geometry: Water4-Particle
457 * Calculate force/pot: Force
460 nb_kernel_ElecRF_VdwNone_GeomW4P1_F_sse2_double
461 (t_nblist * gmx_restrict nlist,
462 rvec * gmx_restrict xx,
463 rvec * gmx_restrict ff,
464 t_forcerec * gmx_restrict fr,
465 t_mdatoms * gmx_restrict mdatoms,
466 nb_kernel_data_t * gmx_restrict kernel_data,
467 t_nrnb * gmx_restrict nrnb)
469 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
470 * just 0 for non-waters.
471 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
472 * jnr indices corresponding to data put in the four positions in the SIMD register.
474 int i_shift_offset,i_coord_offset,outeriter,inneriter;
475 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
477 int j_coord_offsetA,j_coord_offsetB;
478 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
480 real *shiftvec,*fshift,*x,*f;
481 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
483 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
485 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
487 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
488 int vdwjidx0A,vdwjidx0B;
489 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
490 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
491 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
492 __m128d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
493 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
495 __m128d dummy_mask,cutoff_mask;
496 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
497 __m128d one = _mm_set1_pd(1.0);
498 __m128d two = _mm_set1_pd(2.0);
504 jindex = nlist->jindex;
506 shiftidx = nlist->shift;
508 shiftvec = fr->shift_vec[0];
509 fshift = fr->fshift[0];
510 facel = _mm_set1_pd(fr->epsfac);
511 charge = mdatoms->chargeA;
512 krf = _mm_set1_pd(fr->ic->k_rf);
513 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
514 crf = _mm_set1_pd(fr->ic->c_rf);
516 /* Setup water-specific parameters */
517 inr = nlist->iinr[0];
518 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
519 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
520 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
522 /* Avoid stupid compiler warnings */
530 /* Start outer loop over neighborlists */
531 for(iidx=0; iidx<nri; iidx++)
533 /* Load shift vector for this list */
534 i_shift_offset = DIM*shiftidx[iidx];
536 /* Load limits for loop over neighbors */
537 j_index_start = jindex[iidx];
538 j_index_end = jindex[iidx+1];
540 /* Get outer coordinate index */
542 i_coord_offset = DIM*inr;
544 /* Load i particle coords and add shift vector */
545 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
546 &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
548 fix1 = _mm_setzero_pd();
549 fiy1 = _mm_setzero_pd();
550 fiz1 = _mm_setzero_pd();
551 fix2 = _mm_setzero_pd();
552 fiy2 = _mm_setzero_pd();
553 fiz2 = _mm_setzero_pd();
554 fix3 = _mm_setzero_pd();
555 fiy3 = _mm_setzero_pd();
556 fiz3 = _mm_setzero_pd();
558 /* Start inner kernel loop */
559 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
562 /* Get j neighbor index, and coordinate index */
565 j_coord_offsetA = DIM*jnrA;
566 j_coord_offsetB = DIM*jnrB;
568 /* load j atom coordinates */
569 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
572 /* Calculate displacement vector */
573 dx10 = _mm_sub_pd(ix1,jx0);
574 dy10 = _mm_sub_pd(iy1,jy0);
575 dz10 = _mm_sub_pd(iz1,jz0);
576 dx20 = _mm_sub_pd(ix2,jx0);
577 dy20 = _mm_sub_pd(iy2,jy0);
578 dz20 = _mm_sub_pd(iz2,jz0);
579 dx30 = _mm_sub_pd(ix3,jx0);
580 dy30 = _mm_sub_pd(iy3,jy0);
581 dz30 = _mm_sub_pd(iz3,jz0);
583 /* Calculate squared distance and things based on it */
584 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
585 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
586 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
588 rinv10 = gmx_mm_invsqrt_pd(rsq10);
589 rinv20 = gmx_mm_invsqrt_pd(rsq20);
590 rinv30 = gmx_mm_invsqrt_pd(rsq30);
592 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
593 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
594 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
596 /* Load parameters for j particles */
597 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
599 fjx0 = _mm_setzero_pd();
600 fjy0 = _mm_setzero_pd();
601 fjz0 = _mm_setzero_pd();
603 /**************************
604 * CALCULATE INTERACTIONS *
605 **************************/
607 /* Compute parameters for interactions between i and j atoms */
608 qq10 = _mm_mul_pd(iq1,jq0);
610 /* REACTION-FIELD ELECTROSTATICS */
611 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
615 /* Calculate temporary vectorial force */
616 tx = _mm_mul_pd(fscal,dx10);
617 ty = _mm_mul_pd(fscal,dy10);
618 tz = _mm_mul_pd(fscal,dz10);
620 /* Update vectorial force */
621 fix1 = _mm_add_pd(fix1,tx);
622 fiy1 = _mm_add_pd(fiy1,ty);
623 fiz1 = _mm_add_pd(fiz1,tz);
625 fjx0 = _mm_add_pd(fjx0,tx);
626 fjy0 = _mm_add_pd(fjy0,ty);
627 fjz0 = _mm_add_pd(fjz0,tz);
629 /**************************
630 * CALCULATE INTERACTIONS *
631 **************************/
633 /* Compute parameters for interactions between i and j atoms */
634 qq20 = _mm_mul_pd(iq2,jq0);
636 /* REACTION-FIELD ELECTROSTATICS */
637 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
641 /* Calculate temporary vectorial force */
642 tx = _mm_mul_pd(fscal,dx20);
643 ty = _mm_mul_pd(fscal,dy20);
644 tz = _mm_mul_pd(fscal,dz20);
646 /* Update vectorial force */
647 fix2 = _mm_add_pd(fix2,tx);
648 fiy2 = _mm_add_pd(fiy2,ty);
649 fiz2 = _mm_add_pd(fiz2,tz);
651 fjx0 = _mm_add_pd(fjx0,tx);
652 fjy0 = _mm_add_pd(fjy0,ty);
653 fjz0 = _mm_add_pd(fjz0,tz);
655 /**************************
656 * CALCULATE INTERACTIONS *
657 **************************/
659 /* Compute parameters for interactions between i and j atoms */
660 qq30 = _mm_mul_pd(iq3,jq0);
662 /* REACTION-FIELD ELECTROSTATICS */
663 felec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_mul_pd(rinv30,rinvsq30),krf2));
667 /* Calculate temporary vectorial force */
668 tx = _mm_mul_pd(fscal,dx30);
669 ty = _mm_mul_pd(fscal,dy30);
670 tz = _mm_mul_pd(fscal,dz30);
672 /* Update vectorial force */
673 fix3 = _mm_add_pd(fix3,tx);
674 fiy3 = _mm_add_pd(fiy3,ty);
675 fiz3 = _mm_add_pd(fiz3,tz);
677 fjx0 = _mm_add_pd(fjx0,tx);
678 fjy0 = _mm_add_pd(fjy0,ty);
679 fjz0 = _mm_add_pd(fjz0,tz);
681 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
683 /* Inner loop uses 84 flops */
690 j_coord_offsetA = DIM*jnrA;
692 /* load j atom coordinates */
693 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
696 /* Calculate displacement vector */
697 dx10 = _mm_sub_pd(ix1,jx0);
698 dy10 = _mm_sub_pd(iy1,jy0);
699 dz10 = _mm_sub_pd(iz1,jz0);
700 dx20 = _mm_sub_pd(ix2,jx0);
701 dy20 = _mm_sub_pd(iy2,jy0);
702 dz20 = _mm_sub_pd(iz2,jz0);
703 dx30 = _mm_sub_pd(ix3,jx0);
704 dy30 = _mm_sub_pd(iy3,jy0);
705 dz30 = _mm_sub_pd(iz3,jz0);
707 /* Calculate squared distance and things based on it */
708 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
709 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
710 rsq30 = gmx_mm_calc_rsq_pd(dx30,dy30,dz30);
712 rinv10 = gmx_mm_invsqrt_pd(rsq10);
713 rinv20 = gmx_mm_invsqrt_pd(rsq20);
714 rinv30 = gmx_mm_invsqrt_pd(rsq30);
716 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
717 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
718 rinvsq30 = _mm_mul_pd(rinv30,rinv30);
720 /* Load parameters for j particles */
721 jq0 = _mm_load_sd(charge+jnrA+0);
723 fjx0 = _mm_setzero_pd();
724 fjy0 = _mm_setzero_pd();
725 fjz0 = _mm_setzero_pd();
727 /**************************
728 * CALCULATE INTERACTIONS *
729 **************************/
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));
739 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
741 /* Calculate temporary vectorial force */
742 tx = _mm_mul_pd(fscal,dx10);
743 ty = _mm_mul_pd(fscal,dy10);
744 tz = _mm_mul_pd(fscal,dz10);
746 /* Update vectorial force */
747 fix1 = _mm_add_pd(fix1,tx);
748 fiy1 = _mm_add_pd(fiy1,ty);
749 fiz1 = _mm_add_pd(fiz1,tz);
751 fjx0 = _mm_add_pd(fjx0,tx);
752 fjy0 = _mm_add_pd(fjy0,ty);
753 fjz0 = _mm_add_pd(fjz0,tz);
755 /**************************
756 * CALCULATE INTERACTIONS *
757 **************************/
759 /* Compute parameters for interactions between i and j atoms */
760 qq20 = _mm_mul_pd(iq2,jq0);
762 /* REACTION-FIELD ELECTROSTATICS */
763 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
767 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
769 /* Calculate temporary vectorial force */
770 tx = _mm_mul_pd(fscal,dx20);
771 ty = _mm_mul_pd(fscal,dy20);
772 tz = _mm_mul_pd(fscal,dz20);
774 /* Update vectorial force */
775 fix2 = _mm_add_pd(fix2,tx);
776 fiy2 = _mm_add_pd(fiy2,ty);
777 fiz2 = _mm_add_pd(fiz2,tz);
779 fjx0 = _mm_add_pd(fjx0,tx);
780 fjy0 = _mm_add_pd(fjy0,ty);
781 fjz0 = _mm_add_pd(fjz0,tz);
783 /**************************
784 * CALCULATE INTERACTIONS *
785 **************************/
787 /* Compute parameters for interactions between i and j atoms */
788 qq30 = _mm_mul_pd(iq3,jq0);
790 /* REACTION-FIELD ELECTROSTATICS */
791 felec = _mm_mul_pd(qq30,_mm_sub_pd(_mm_mul_pd(rinv30,rinvsq30),krf2));
795 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
797 /* Calculate temporary vectorial force */
798 tx = _mm_mul_pd(fscal,dx30);
799 ty = _mm_mul_pd(fscal,dy30);
800 tz = _mm_mul_pd(fscal,dz30);
802 /* Update vectorial force */
803 fix3 = _mm_add_pd(fix3,tx);
804 fiy3 = _mm_add_pd(fiy3,ty);
805 fiz3 = _mm_add_pd(fiz3,tz);
807 fjx0 = _mm_add_pd(fjx0,tx);
808 fjy0 = _mm_add_pd(fjy0,ty);
809 fjz0 = _mm_add_pd(fjz0,tz);
811 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
813 /* Inner loop uses 84 flops */
816 /* End of innermost loop */
818 gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
819 f+i_coord_offset+DIM,fshift+i_shift_offset);
821 /* Increment number of inner iterations */
822 inneriter += j_index_end - j_index_start;
824 /* Outer loop uses 18 flops */
827 /* Increment number of outer iterations */
830 /* Update outer/inner flops */
832 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*84);