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 sse4_1_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "gromacs/simd/math_x86_sse4_1_double.h"
48 #include "kernelutil_x86_sse4_1_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomW3P1_VF_sse4_1_double
52 * Electrostatics interaction: Coulomb
53 * VdW interaction: None
54 * Geometry: Water3-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecCoul_VdwNone_GeomW3P1_VF_sse4_1_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
86 int vdwjidx0A,vdwjidx0B;
87 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
88 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
89 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
90 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
91 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
93 __m128d dummy_mask,cutoff_mask;
94 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
95 __m128d one = _mm_set1_pd(1.0);
96 __m128d two = _mm_set1_pd(2.0);
102 jindex = nlist->jindex;
104 shiftidx = nlist->shift;
106 shiftvec = fr->shift_vec[0];
107 fshift = fr->fshift[0];
108 facel = _mm_set1_pd(fr->epsfac);
109 charge = mdatoms->chargeA;
111 /* Setup water-specific parameters */
112 inr = nlist->iinr[0];
113 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
114 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
115 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
117 /* Avoid stupid compiler warnings */
125 /* Start outer loop over neighborlists */
126 for(iidx=0; iidx<nri; iidx++)
128 /* Load shift vector for this list */
129 i_shift_offset = DIM*shiftidx[iidx];
131 /* Load limits for loop over neighbors */
132 j_index_start = jindex[iidx];
133 j_index_end = jindex[iidx+1];
135 /* Get outer coordinate index */
137 i_coord_offset = DIM*inr;
139 /* Load i particle coords and add shift vector */
140 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
141 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
143 fix0 = _mm_setzero_pd();
144 fiy0 = _mm_setzero_pd();
145 fiz0 = _mm_setzero_pd();
146 fix1 = _mm_setzero_pd();
147 fiy1 = _mm_setzero_pd();
148 fiz1 = _mm_setzero_pd();
149 fix2 = _mm_setzero_pd();
150 fiy2 = _mm_setzero_pd();
151 fiz2 = _mm_setzero_pd();
153 /* Reset potential sums */
154 velecsum = _mm_setzero_pd();
156 /* Start inner kernel loop */
157 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
160 /* Get j neighbor index, and coordinate index */
163 j_coord_offsetA = DIM*jnrA;
164 j_coord_offsetB = DIM*jnrB;
166 /* load j atom coordinates */
167 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
170 /* Calculate displacement vector */
171 dx00 = _mm_sub_pd(ix0,jx0);
172 dy00 = _mm_sub_pd(iy0,jy0);
173 dz00 = _mm_sub_pd(iz0,jz0);
174 dx10 = _mm_sub_pd(ix1,jx0);
175 dy10 = _mm_sub_pd(iy1,jy0);
176 dz10 = _mm_sub_pd(iz1,jz0);
177 dx20 = _mm_sub_pd(ix2,jx0);
178 dy20 = _mm_sub_pd(iy2,jy0);
179 dz20 = _mm_sub_pd(iz2,jz0);
181 /* Calculate squared distance and things based on it */
182 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
183 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
184 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
186 rinv00 = gmx_mm_invsqrt_pd(rsq00);
187 rinv10 = gmx_mm_invsqrt_pd(rsq10);
188 rinv20 = gmx_mm_invsqrt_pd(rsq20);
190 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
191 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
192 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
194 /* Load parameters for j particles */
195 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
197 fjx0 = _mm_setzero_pd();
198 fjy0 = _mm_setzero_pd();
199 fjz0 = _mm_setzero_pd();
201 /**************************
202 * CALCULATE INTERACTIONS *
203 **************************/
205 /* Compute parameters for interactions between i and j atoms */
206 qq00 = _mm_mul_pd(iq0,jq0);
208 /* COULOMB ELECTROSTATICS */
209 velec = _mm_mul_pd(qq00,rinv00);
210 felec = _mm_mul_pd(velec,rinvsq00);
212 /* Update potential sum for this i atom from the interaction with this j atom. */
213 velecsum = _mm_add_pd(velecsum,velec);
217 /* Calculate temporary vectorial force */
218 tx = _mm_mul_pd(fscal,dx00);
219 ty = _mm_mul_pd(fscal,dy00);
220 tz = _mm_mul_pd(fscal,dz00);
222 /* Update vectorial force */
223 fix0 = _mm_add_pd(fix0,tx);
224 fiy0 = _mm_add_pd(fiy0,ty);
225 fiz0 = _mm_add_pd(fiz0,tz);
227 fjx0 = _mm_add_pd(fjx0,tx);
228 fjy0 = _mm_add_pd(fjy0,ty);
229 fjz0 = _mm_add_pd(fjz0,tz);
231 /**************************
232 * CALCULATE INTERACTIONS *
233 **************************/
235 /* Compute parameters for interactions between i and j atoms */
236 qq10 = _mm_mul_pd(iq1,jq0);
238 /* COULOMB ELECTROSTATICS */
239 velec = _mm_mul_pd(qq10,rinv10);
240 felec = _mm_mul_pd(velec,rinvsq10);
242 /* Update potential sum for this i atom from the interaction with this j atom. */
243 velecsum = _mm_add_pd(velecsum,velec);
247 /* Calculate temporary vectorial force */
248 tx = _mm_mul_pd(fscal,dx10);
249 ty = _mm_mul_pd(fscal,dy10);
250 tz = _mm_mul_pd(fscal,dz10);
252 /* Update vectorial force */
253 fix1 = _mm_add_pd(fix1,tx);
254 fiy1 = _mm_add_pd(fiy1,ty);
255 fiz1 = _mm_add_pd(fiz1,tz);
257 fjx0 = _mm_add_pd(fjx0,tx);
258 fjy0 = _mm_add_pd(fjy0,ty);
259 fjz0 = _mm_add_pd(fjz0,tz);
261 /**************************
262 * CALCULATE INTERACTIONS *
263 **************************/
265 /* Compute parameters for interactions between i and j atoms */
266 qq20 = _mm_mul_pd(iq2,jq0);
268 /* COULOMB ELECTROSTATICS */
269 velec = _mm_mul_pd(qq20,rinv20);
270 felec = _mm_mul_pd(velec,rinvsq20);
272 /* Update potential sum for this i atom from the interaction with this j atom. */
273 velecsum = _mm_add_pd(velecsum,velec);
277 /* Calculate temporary vectorial force */
278 tx = _mm_mul_pd(fscal,dx20);
279 ty = _mm_mul_pd(fscal,dy20);
280 tz = _mm_mul_pd(fscal,dz20);
282 /* Update vectorial force */
283 fix2 = _mm_add_pd(fix2,tx);
284 fiy2 = _mm_add_pd(fiy2,ty);
285 fiz2 = _mm_add_pd(fiz2,tz);
287 fjx0 = _mm_add_pd(fjx0,tx);
288 fjy0 = _mm_add_pd(fjy0,ty);
289 fjz0 = _mm_add_pd(fjz0,tz);
291 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
293 /* Inner loop uses 87 flops */
300 j_coord_offsetA = DIM*jnrA;
302 /* load j atom coordinates */
303 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
306 /* Calculate displacement vector */
307 dx00 = _mm_sub_pd(ix0,jx0);
308 dy00 = _mm_sub_pd(iy0,jy0);
309 dz00 = _mm_sub_pd(iz0,jz0);
310 dx10 = _mm_sub_pd(ix1,jx0);
311 dy10 = _mm_sub_pd(iy1,jy0);
312 dz10 = _mm_sub_pd(iz1,jz0);
313 dx20 = _mm_sub_pd(ix2,jx0);
314 dy20 = _mm_sub_pd(iy2,jy0);
315 dz20 = _mm_sub_pd(iz2,jz0);
317 /* Calculate squared distance and things based on it */
318 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
319 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
320 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
322 rinv00 = gmx_mm_invsqrt_pd(rsq00);
323 rinv10 = gmx_mm_invsqrt_pd(rsq10);
324 rinv20 = gmx_mm_invsqrt_pd(rsq20);
326 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
327 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
328 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
330 /* Load parameters for j particles */
331 jq0 = _mm_load_sd(charge+jnrA+0);
333 fjx0 = _mm_setzero_pd();
334 fjy0 = _mm_setzero_pd();
335 fjz0 = _mm_setzero_pd();
337 /**************************
338 * CALCULATE INTERACTIONS *
339 **************************/
341 /* Compute parameters for interactions between i and j atoms */
342 qq00 = _mm_mul_pd(iq0,jq0);
344 /* COULOMB ELECTROSTATICS */
345 velec = _mm_mul_pd(qq00,rinv00);
346 felec = _mm_mul_pd(velec,rinvsq00);
348 /* Update potential sum for this i atom from the interaction with this j atom. */
349 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
350 velecsum = _mm_add_pd(velecsum,velec);
354 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
356 /* Calculate temporary vectorial force */
357 tx = _mm_mul_pd(fscal,dx00);
358 ty = _mm_mul_pd(fscal,dy00);
359 tz = _mm_mul_pd(fscal,dz00);
361 /* Update vectorial force */
362 fix0 = _mm_add_pd(fix0,tx);
363 fiy0 = _mm_add_pd(fiy0,ty);
364 fiz0 = _mm_add_pd(fiz0,tz);
366 fjx0 = _mm_add_pd(fjx0,tx);
367 fjy0 = _mm_add_pd(fjy0,ty);
368 fjz0 = _mm_add_pd(fjz0,tz);
370 /**************************
371 * CALCULATE INTERACTIONS *
372 **************************/
374 /* Compute parameters for interactions between i and j atoms */
375 qq10 = _mm_mul_pd(iq1,jq0);
377 /* COULOMB ELECTROSTATICS */
378 velec = _mm_mul_pd(qq10,rinv10);
379 felec = _mm_mul_pd(velec,rinvsq10);
381 /* Update potential sum for this i atom from the interaction with this j atom. */
382 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
383 velecsum = _mm_add_pd(velecsum,velec);
387 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
389 /* Calculate temporary vectorial force */
390 tx = _mm_mul_pd(fscal,dx10);
391 ty = _mm_mul_pd(fscal,dy10);
392 tz = _mm_mul_pd(fscal,dz10);
394 /* Update vectorial force */
395 fix1 = _mm_add_pd(fix1,tx);
396 fiy1 = _mm_add_pd(fiy1,ty);
397 fiz1 = _mm_add_pd(fiz1,tz);
399 fjx0 = _mm_add_pd(fjx0,tx);
400 fjy0 = _mm_add_pd(fjy0,ty);
401 fjz0 = _mm_add_pd(fjz0,tz);
403 /**************************
404 * CALCULATE INTERACTIONS *
405 **************************/
407 /* Compute parameters for interactions between i and j atoms */
408 qq20 = _mm_mul_pd(iq2,jq0);
410 /* COULOMB ELECTROSTATICS */
411 velec = _mm_mul_pd(qq20,rinv20);
412 felec = _mm_mul_pd(velec,rinvsq20);
414 /* Update potential sum for this i atom from the interaction with this j atom. */
415 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
416 velecsum = _mm_add_pd(velecsum,velec);
420 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
422 /* Calculate temporary vectorial force */
423 tx = _mm_mul_pd(fscal,dx20);
424 ty = _mm_mul_pd(fscal,dy20);
425 tz = _mm_mul_pd(fscal,dz20);
427 /* Update vectorial force */
428 fix2 = _mm_add_pd(fix2,tx);
429 fiy2 = _mm_add_pd(fiy2,ty);
430 fiz2 = _mm_add_pd(fiz2,tz);
432 fjx0 = _mm_add_pd(fjx0,tx);
433 fjy0 = _mm_add_pd(fjy0,ty);
434 fjz0 = _mm_add_pd(fjz0,tz);
436 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
438 /* Inner loop uses 87 flops */
441 /* End of innermost loop */
443 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
444 f+i_coord_offset,fshift+i_shift_offset);
447 /* Update potential energies */
448 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
450 /* Increment number of inner iterations */
451 inneriter += j_index_end - j_index_start;
453 /* Outer loop uses 19 flops */
456 /* Increment number of outer iterations */
459 /* Update outer/inner flops */
461 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*87);
464 * Gromacs nonbonded kernel: nb_kernel_ElecCoul_VdwNone_GeomW3P1_F_sse4_1_double
465 * Electrostatics interaction: Coulomb
466 * VdW interaction: None
467 * Geometry: Water3-Particle
468 * Calculate force/pot: Force
471 nb_kernel_ElecCoul_VdwNone_GeomW3P1_F_sse4_1_double
472 (t_nblist * gmx_restrict nlist,
473 rvec * gmx_restrict xx,
474 rvec * gmx_restrict ff,
475 t_forcerec * gmx_restrict fr,
476 t_mdatoms * gmx_restrict mdatoms,
477 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
478 t_nrnb * gmx_restrict nrnb)
480 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
481 * just 0 for non-waters.
482 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
483 * jnr indices corresponding to data put in the four positions in the SIMD register.
485 int i_shift_offset,i_coord_offset,outeriter,inneriter;
486 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
488 int j_coord_offsetA,j_coord_offsetB;
489 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
491 real *shiftvec,*fshift,*x,*f;
492 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
494 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
496 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
498 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
499 int vdwjidx0A,vdwjidx0B;
500 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
501 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
502 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
503 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
504 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
506 __m128d dummy_mask,cutoff_mask;
507 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
508 __m128d one = _mm_set1_pd(1.0);
509 __m128d two = _mm_set1_pd(2.0);
515 jindex = nlist->jindex;
517 shiftidx = nlist->shift;
519 shiftvec = fr->shift_vec[0];
520 fshift = fr->fshift[0];
521 facel = _mm_set1_pd(fr->epsfac);
522 charge = mdatoms->chargeA;
524 /* Setup water-specific parameters */
525 inr = nlist->iinr[0];
526 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
527 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
528 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
530 /* Avoid stupid compiler warnings */
538 /* Start outer loop over neighborlists */
539 for(iidx=0; iidx<nri; iidx++)
541 /* Load shift vector for this list */
542 i_shift_offset = DIM*shiftidx[iidx];
544 /* Load limits for loop over neighbors */
545 j_index_start = jindex[iidx];
546 j_index_end = jindex[iidx+1];
548 /* Get outer coordinate index */
550 i_coord_offset = DIM*inr;
552 /* Load i particle coords and add shift vector */
553 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
554 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
556 fix0 = _mm_setzero_pd();
557 fiy0 = _mm_setzero_pd();
558 fiz0 = _mm_setzero_pd();
559 fix1 = _mm_setzero_pd();
560 fiy1 = _mm_setzero_pd();
561 fiz1 = _mm_setzero_pd();
562 fix2 = _mm_setzero_pd();
563 fiy2 = _mm_setzero_pd();
564 fiz2 = _mm_setzero_pd();
566 /* Start inner kernel loop */
567 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
570 /* Get j neighbor index, and coordinate index */
573 j_coord_offsetA = DIM*jnrA;
574 j_coord_offsetB = DIM*jnrB;
576 /* load j atom coordinates */
577 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
580 /* Calculate displacement vector */
581 dx00 = _mm_sub_pd(ix0,jx0);
582 dy00 = _mm_sub_pd(iy0,jy0);
583 dz00 = _mm_sub_pd(iz0,jz0);
584 dx10 = _mm_sub_pd(ix1,jx0);
585 dy10 = _mm_sub_pd(iy1,jy0);
586 dz10 = _mm_sub_pd(iz1,jz0);
587 dx20 = _mm_sub_pd(ix2,jx0);
588 dy20 = _mm_sub_pd(iy2,jy0);
589 dz20 = _mm_sub_pd(iz2,jz0);
591 /* Calculate squared distance and things based on it */
592 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
593 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
594 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
596 rinv00 = gmx_mm_invsqrt_pd(rsq00);
597 rinv10 = gmx_mm_invsqrt_pd(rsq10);
598 rinv20 = gmx_mm_invsqrt_pd(rsq20);
600 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
601 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
602 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
604 /* Load parameters for j particles */
605 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
607 fjx0 = _mm_setzero_pd();
608 fjy0 = _mm_setzero_pd();
609 fjz0 = _mm_setzero_pd();
611 /**************************
612 * CALCULATE INTERACTIONS *
613 **************************/
615 /* Compute parameters for interactions between i and j atoms */
616 qq00 = _mm_mul_pd(iq0,jq0);
618 /* COULOMB ELECTROSTATICS */
619 velec = _mm_mul_pd(qq00,rinv00);
620 felec = _mm_mul_pd(velec,rinvsq00);
624 /* Calculate temporary vectorial force */
625 tx = _mm_mul_pd(fscal,dx00);
626 ty = _mm_mul_pd(fscal,dy00);
627 tz = _mm_mul_pd(fscal,dz00);
629 /* Update vectorial force */
630 fix0 = _mm_add_pd(fix0,tx);
631 fiy0 = _mm_add_pd(fiy0,ty);
632 fiz0 = _mm_add_pd(fiz0,tz);
634 fjx0 = _mm_add_pd(fjx0,tx);
635 fjy0 = _mm_add_pd(fjy0,ty);
636 fjz0 = _mm_add_pd(fjz0,tz);
638 /**************************
639 * CALCULATE INTERACTIONS *
640 **************************/
642 /* Compute parameters for interactions between i and j atoms */
643 qq10 = _mm_mul_pd(iq1,jq0);
645 /* COULOMB ELECTROSTATICS */
646 velec = _mm_mul_pd(qq10,rinv10);
647 felec = _mm_mul_pd(velec,rinvsq10);
651 /* Calculate temporary vectorial force */
652 tx = _mm_mul_pd(fscal,dx10);
653 ty = _mm_mul_pd(fscal,dy10);
654 tz = _mm_mul_pd(fscal,dz10);
656 /* Update vectorial force */
657 fix1 = _mm_add_pd(fix1,tx);
658 fiy1 = _mm_add_pd(fiy1,ty);
659 fiz1 = _mm_add_pd(fiz1,tz);
661 fjx0 = _mm_add_pd(fjx0,tx);
662 fjy0 = _mm_add_pd(fjy0,ty);
663 fjz0 = _mm_add_pd(fjz0,tz);
665 /**************************
666 * CALCULATE INTERACTIONS *
667 **************************/
669 /* Compute parameters for interactions between i and j atoms */
670 qq20 = _mm_mul_pd(iq2,jq0);
672 /* COULOMB ELECTROSTATICS */
673 velec = _mm_mul_pd(qq20,rinv20);
674 felec = _mm_mul_pd(velec,rinvsq20);
678 /* Calculate temporary vectorial force */
679 tx = _mm_mul_pd(fscal,dx20);
680 ty = _mm_mul_pd(fscal,dy20);
681 tz = _mm_mul_pd(fscal,dz20);
683 /* Update vectorial force */
684 fix2 = _mm_add_pd(fix2,tx);
685 fiy2 = _mm_add_pd(fiy2,ty);
686 fiz2 = _mm_add_pd(fiz2,tz);
688 fjx0 = _mm_add_pd(fjx0,tx);
689 fjy0 = _mm_add_pd(fjy0,ty);
690 fjz0 = _mm_add_pd(fjz0,tz);
692 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
694 /* Inner loop uses 84 flops */
701 j_coord_offsetA = DIM*jnrA;
703 /* load j atom coordinates */
704 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
707 /* Calculate displacement vector */
708 dx00 = _mm_sub_pd(ix0,jx0);
709 dy00 = _mm_sub_pd(iy0,jy0);
710 dz00 = _mm_sub_pd(iz0,jz0);
711 dx10 = _mm_sub_pd(ix1,jx0);
712 dy10 = _mm_sub_pd(iy1,jy0);
713 dz10 = _mm_sub_pd(iz1,jz0);
714 dx20 = _mm_sub_pd(ix2,jx0);
715 dy20 = _mm_sub_pd(iy2,jy0);
716 dz20 = _mm_sub_pd(iz2,jz0);
718 /* Calculate squared distance and things based on it */
719 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
720 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
721 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
723 rinv00 = gmx_mm_invsqrt_pd(rsq00);
724 rinv10 = gmx_mm_invsqrt_pd(rsq10);
725 rinv20 = gmx_mm_invsqrt_pd(rsq20);
727 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
728 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
729 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
731 /* Load parameters for j particles */
732 jq0 = _mm_load_sd(charge+jnrA+0);
734 fjx0 = _mm_setzero_pd();
735 fjy0 = _mm_setzero_pd();
736 fjz0 = _mm_setzero_pd();
738 /**************************
739 * CALCULATE INTERACTIONS *
740 **************************/
742 /* Compute parameters for interactions between i and j atoms */
743 qq00 = _mm_mul_pd(iq0,jq0);
745 /* COULOMB ELECTROSTATICS */
746 velec = _mm_mul_pd(qq00,rinv00);
747 felec = _mm_mul_pd(velec,rinvsq00);
751 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
753 /* Calculate temporary vectorial force */
754 tx = _mm_mul_pd(fscal,dx00);
755 ty = _mm_mul_pd(fscal,dy00);
756 tz = _mm_mul_pd(fscal,dz00);
758 /* Update vectorial force */
759 fix0 = _mm_add_pd(fix0,tx);
760 fiy0 = _mm_add_pd(fiy0,ty);
761 fiz0 = _mm_add_pd(fiz0,tz);
763 fjx0 = _mm_add_pd(fjx0,tx);
764 fjy0 = _mm_add_pd(fjy0,ty);
765 fjz0 = _mm_add_pd(fjz0,tz);
767 /**************************
768 * CALCULATE INTERACTIONS *
769 **************************/
771 /* Compute parameters for interactions between i and j atoms */
772 qq10 = _mm_mul_pd(iq1,jq0);
774 /* COULOMB ELECTROSTATICS */
775 velec = _mm_mul_pd(qq10,rinv10);
776 felec = _mm_mul_pd(velec,rinvsq10);
780 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
782 /* Calculate temporary vectorial force */
783 tx = _mm_mul_pd(fscal,dx10);
784 ty = _mm_mul_pd(fscal,dy10);
785 tz = _mm_mul_pd(fscal,dz10);
787 /* Update vectorial force */
788 fix1 = _mm_add_pd(fix1,tx);
789 fiy1 = _mm_add_pd(fiy1,ty);
790 fiz1 = _mm_add_pd(fiz1,tz);
792 fjx0 = _mm_add_pd(fjx0,tx);
793 fjy0 = _mm_add_pd(fjy0,ty);
794 fjz0 = _mm_add_pd(fjz0,tz);
796 /**************************
797 * CALCULATE INTERACTIONS *
798 **************************/
800 /* Compute parameters for interactions between i and j atoms */
801 qq20 = _mm_mul_pd(iq2,jq0);
803 /* COULOMB ELECTROSTATICS */
804 velec = _mm_mul_pd(qq20,rinv20);
805 felec = _mm_mul_pd(velec,rinvsq20);
809 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
811 /* Calculate temporary vectorial force */
812 tx = _mm_mul_pd(fscal,dx20);
813 ty = _mm_mul_pd(fscal,dy20);
814 tz = _mm_mul_pd(fscal,dz20);
816 /* Update vectorial force */
817 fix2 = _mm_add_pd(fix2,tx);
818 fiy2 = _mm_add_pd(fiy2,ty);
819 fiz2 = _mm_add_pd(fiz2,tz);
821 fjx0 = _mm_add_pd(fjx0,tx);
822 fjy0 = _mm_add_pd(fjy0,ty);
823 fjz0 = _mm_add_pd(fjz0,tz);
825 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
827 /* Inner loop uses 84 flops */
830 /* End of innermost loop */
832 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
833 f+i_coord_offset,fshift+i_shift_offset);
835 /* Increment number of inner iterations */
836 inneriter += j_index_end - j_index_start;
838 /* Outer loop uses 18 flops */
841 /* Increment number of outer iterations */
844 /* Update outer/inner flops */
846 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*84);