2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015,2017,2018, 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 avx_256_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "gromacs/gmxlib/nrnb.h"
47 #include "kernelutil_x86_avx_256_double.h"
50 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_avx_256_double
51 * Electrostatics interaction: ReactionField
52 * VdW interaction: None
53 * Geometry: Water3-Water3
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_avx_256_double
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 struct t_forcerec * gmx_restrict fr,
62 t_mdatoms * gmx_restrict mdatoms,
63 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
64 t_nrnb * gmx_restrict nrnb)
66 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
67 * just 0 for non-waters.
68 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
69 * jnr indices corresponding to data put in the four positions in the SIMD register.
71 int i_shift_offset,i_coord_offset,outeriter,inneriter;
72 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
73 int jnrA,jnrB,jnrC,jnrD;
74 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
75 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
76 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
77 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
79 real *shiftvec,*fshift,*x,*f;
80 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
82 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 real * vdwioffsetptr0;
84 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
85 real * vdwioffsetptr1;
86 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
87 real * vdwioffsetptr2;
88 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
89 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
90 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
91 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
92 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
93 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
94 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
95 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
96 __m256d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
97 __m256d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
98 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
99 __m256d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
100 __m256d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
101 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
102 __m256d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
103 __m256d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
104 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
106 __m256d dummy_mask,cutoff_mask;
107 __m128 tmpmask0,tmpmask1;
108 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
109 __m256d one = _mm256_set1_pd(1.0);
110 __m256d two = _mm256_set1_pd(2.0);
116 jindex = nlist->jindex;
118 shiftidx = nlist->shift;
120 shiftvec = fr->shift_vec[0];
121 fshift = fr->fshift[0];
122 facel = _mm256_set1_pd(fr->ic->epsfac);
123 charge = mdatoms->chargeA;
124 krf = _mm256_set1_pd(fr->ic->k_rf);
125 krf2 = _mm256_set1_pd(fr->ic->k_rf*2.0);
126 crf = _mm256_set1_pd(fr->ic->c_rf);
128 /* Setup water-specific parameters */
129 inr = nlist->iinr[0];
130 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
131 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
132 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
134 jq0 = _mm256_set1_pd(charge[inr+0]);
135 jq1 = _mm256_set1_pd(charge[inr+1]);
136 jq2 = _mm256_set1_pd(charge[inr+2]);
137 qq00 = _mm256_mul_pd(iq0,jq0);
138 qq01 = _mm256_mul_pd(iq0,jq1);
139 qq02 = _mm256_mul_pd(iq0,jq2);
140 qq10 = _mm256_mul_pd(iq1,jq0);
141 qq11 = _mm256_mul_pd(iq1,jq1);
142 qq12 = _mm256_mul_pd(iq1,jq2);
143 qq20 = _mm256_mul_pd(iq2,jq0);
144 qq21 = _mm256_mul_pd(iq2,jq1);
145 qq22 = _mm256_mul_pd(iq2,jq2);
147 /* Avoid stupid compiler warnings */
148 jnrA = jnrB = jnrC = jnrD = 0;
157 for(iidx=0;iidx<4*DIM;iidx++)
162 /* Start outer loop over neighborlists */
163 for(iidx=0; iidx<nri; iidx++)
165 /* Load shift vector for this list */
166 i_shift_offset = DIM*shiftidx[iidx];
168 /* Load limits for loop over neighbors */
169 j_index_start = jindex[iidx];
170 j_index_end = jindex[iidx+1];
172 /* Get outer coordinate index */
174 i_coord_offset = DIM*inr;
176 /* Load i particle coords and add shift vector */
177 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
178 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
180 fix0 = _mm256_setzero_pd();
181 fiy0 = _mm256_setzero_pd();
182 fiz0 = _mm256_setzero_pd();
183 fix1 = _mm256_setzero_pd();
184 fiy1 = _mm256_setzero_pd();
185 fiz1 = _mm256_setzero_pd();
186 fix2 = _mm256_setzero_pd();
187 fiy2 = _mm256_setzero_pd();
188 fiz2 = _mm256_setzero_pd();
190 /* Reset potential sums */
191 velecsum = _mm256_setzero_pd();
193 /* Start inner kernel loop */
194 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
197 /* Get j neighbor index, and coordinate index */
202 j_coord_offsetA = DIM*jnrA;
203 j_coord_offsetB = DIM*jnrB;
204 j_coord_offsetC = DIM*jnrC;
205 j_coord_offsetD = DIM*jnrD;
207 /* load j atom coordinates */
208 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
209 x+j_coord_offsetC,x+j_coord_offsetD,
210 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
212 /* Calculate displacement vector */
213 dx00 = _mm256_sub_pd(ix0,jx0);
214 dy00 = _mm256_sub_pd(iy0,jy0);
215 dz00 = _mm256_sub_pd(iz0,jz0);
216 dx01 = _mm256_sub_pd(ix0,jx1);
217 dy01 = _mm256_sub_pd(iy0,jy1);
218 dz01 = _mm256_sub_pd(iz0,jz1);
219 dx02 = _mm256_sub_pd(ix0,jx2);
220 dy02 = _mm256_sub_pd(iy0,jy2);
221 dz02 = _mm256_sub_pd(iz0,jz2);
222 dx10 = _mm256_sub_pd(ix1,jx0);
223 dy10 = _mm256_sub_pd(iy1,jy0);
224 dz10 = _mm256_sub_pd(iz1,jz0);
225 dx11 = _mm256_sub_pd(ix1,jx1);
226 dy11 = _mm256_sub_pd(iy1,jy1);
227 dz11 = _mm256_sub_pd(iz1,jz1);
228 dx12 = _mm256_sub_pd(ix1,jx2);
229 dy12 = _mm256_sub_pd(iy1,jy2);
230 dz12 = _mm256_sub_pd(iz1,jz2);
231 dx20 = _mm256_sub_pd(ix2,jx0);
232 dy20 = _mm256_sub_pd(iy2,jy0);
233 dz20 = _mm256_sub_pd(iz2,jz0);
234 dx21 = _mm256_sub_pd(ix2,jx1);
235 dy21 = _mm256_sub_pd(iy2,jy1);
236 dz21 = _mm256_sub_pd(iz2,jz1);
237 dx22 = _mm256_sub_pd(ix2,jx2);
238 dy22 = _mm256_sub_pd(iy2,jy2);
239 dz22 = _mm256_sub_pd(iz2,jz2);
241 /* Calculate squared distance and things based on it */
242 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
243 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
244 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
245 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
246 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
247 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
248 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
249 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
250 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
252 rinv00 = avx256_invsqrt_d(rsq00);
253 rinv01 = avx256_invsqrt_d(rsq01);
254 rinv02 = avx256_invsqrt_d(rsq02);
255 rinv10 = avx256_invsqrt_d(rsq10);
256 rinv11 = avx256_invsqrt_d(rsq11);
257 rinv12 = avx256_invsqrt_d(rsq12);
258 rinv20 = avx256_invsqrt_d(rsq20);
259 rinv21 = avx256_invsqrt_d(rsq21);
260 rinv22 = avx256_invsqrt_d(rsq22);
262 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
263 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
264 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
265 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
266 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
267 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
268 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
269 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
270 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
272 fjx0 = _mm256_setzero_pd();
273 fjy0 = _mm256_setzero_pd();
274 fjz0 = _mm256_setzero_pd();
275 fjx1 = _mm256_setzero_pd();
276 fjy1 = _mm256_setzero_pd();
277 fjz1 = _mm256_setzero_pd();
278 fjx2 = _mm256_setzero_pd();
279 fjy2 = _mm256_setzero_pd();
280 fjz2 = _mm256_setzero_pd();
282 /**************************
283 * CALCULATE INTERACTIONS *
284 **************************/
286 /* REACTION-FIELD ELECTROSTATICS */
287 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
288 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
290 /* Update potential sum for this i atom from the interaction with this j atom. */
291 velecsum = _mm256_add_pd(velecsum,velec);
295 /* Calculate temporary vectorial force */
296 tx = _mm256_mul_pd(fscal,dx00);
297 ty = _mm256_mul_pd(fscal,dy00);
298 tz = _mm256_mul_pd(fscal,dz00);
300 /* Update vectorial force */
301 fix0 = _mm256_add_pd(fix0,tx);
302 fiy0 = _mm256_add_pd(fiy0,ty);
303 fiz0 = _mm256_add_pd(fiz0,tz);
305 fjx0 = _mm256_add_pd(fjx0,tx);
306 fjy0 = _mm256_add_pd(fjy0,ty);
307 fjz0 = _mm256_add_pd(fjz0,tz);
309 /**************************
310 * CALCULATE INTERACTIONS *
311 **************************/
313 /* REACTION-FIELD ELECTROSTATICS */
314 velec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
315 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
317 /* Update potential sum for this i atom from the interaction with this j atom. */
318 velecsum = _mm256_add_pd(velecsum,velec);
322 /* Calculate temporary vectorial force */
323 tx = _mm256_mul_pd(fscal,dx01);
324 ty = _mm256_mul_pd(fscal,dy01);
325 tz = _mm256_mul_pd(fscal,dz01);
327 /* Update vectorial force */
328 fix0 = _mm256_add_pd(fix0,tx);
329 fiy0 = _mm256_add_pd(fiy0,ty);
330 fiz0 = _mm256_add_pd(fiz0,tz);
332 fjx1 = _mm256_add_pd(fjx1,tx);
333 fjy1 = _mm256_add_pd(fjy1,ty);
334 fjz1 = _mm256_add_pd(fjz1,tz);
336 /**************************
337 * CALCULATE INTERACTIONS *
338 **************************/
340 /* REACTION-FIELD ELECTROSTATICS */
341 velec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
342 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
344 /* Update potential sum for this i atom from the interaction with this j atom. */
345 velecsum = _mm256_add_pd(velecsum,velec);
349 /* Calculate temporary vectorial force */
350 tx = _mm256_mul_pd(fscal,dx02);
351 ty = _mm256_mul_pd(fscal,dy02);
352 tz = _mm256_mul_pd(fscal,dz02);
354 /* Update vectorial force */
355 fix0 = _mm256_add_pd(fix0,tx);
356 fiy0 = _mm256_add_pd(fiy0,ty);
357 fiz0 = _mm256_add_pd(fiz0,tz);
359 fjx2 = _mm256_add_pd(fjx2,tx);
360 fjy2 = _mm256_add_pd(fjy2,ty);
361 fjz2 = _mm256_add_pd(fjz2,tz);
363 /**************************
364 * CALCULATE INTERACTIONS *
365 **************************/
367 /* REACTION-FIELD ELECTROSTATICS */
368 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
369 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
371 /* Update potential sum for this i atom from the interaction with this j atom. */
372 velecsum = _mm256_add_pd(velecsum,velec);
376 /* Calculate temporary vectorial force */
377 tx = _mm256_mul_pd(fscal,dx10);
378 ty = _mm256_mul_pd(fscal,dy10);
379 tz = _mm256_mul_pd(fscal,dz10);
381 /* Update vectorial force */
382 fix1 = _mm256_add_pd(fix1,tx);
383 fiy1 = _mm256_add_pd(fiy1,ty);
384 fiz1 = _mm256_add_pd(fiz1,tz);
386 fjx0 = _mm256_add_pd(fjx0,tx);
387 fjy0 = _mm256_add_pd(fjy0,ty);
388 fjz0 = _mm256_add_pd(fjz0,tz);
390 /**************************
391 * CALCULATE INTERACTIONS *
392 **************************/
394 /* REACTION-FIELD ELECTROSTATICS */
395 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
396 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
398 /* Update potential sum for this i atom from the interaction with this j atom. */
399 velecsum = _mm256_add_pd(velecsum,velec);
403 /* Calculate temporary vectorial force */
404 tx = _mm256_mul_pd(fscal,dx11);
405 ty = _mm256_mul_pd(fscal,dy11);
406 tz = _mm256_mul_pd(fscal,dz11);
408 /* Update vectorial force */
409 fix1 = _mm256_add_pd(fix1,tx);
410 fiy1 = _mm256_add_pd(fiy1,ty);
411 fiz1 = _mm256_add_pd(fiz1,tz);
413 fjx1 = _mm256_add_pd(fjx1,tx);
414 fjy1 = _mm256_add_pd(fjy1,ty);
415 fjz1 = _mm256_add_pd(fjz1,tz);
417 /**************************
418 * CALCULATE INTERACTIONS *
419 **************************/
421 /* REACTION-FIELD ELECTROSTATICS */
422 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
423 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
425 /* Update potential sum for this i atom from the interaction with this j atom. */
426 velecsum = _mm256_add_pd(velecsum,velec);
430 /* Calculate temporary vectorial force */
431 tx = _mm256_mul_pd(fscal,dx12);
432 ty = _mm256_mul_pd(fscal,dy12);
433 tz = _mm256_mul_pd(fscal,dz12);
435 /* Update vectorial force */
436 fix1 = _mm256_add_pd(fix1,tx);
437 fiy1 = _mm256_add_pd(fiy1,ty);
438 fiz1 = _mm256_add_pd(fiz1,tz);
440 fjx2 = _mm256_add_pd(fjx2,tx);
441 fjy2 = _mm256_add_pd(fjy2,ty);
442 fjz2 = _mm256_add_pd(fjz2,tz);
444 /**************************
445 * CALCULATE INTERACTIONS *
446 **************************/
448 /* REACTION-FIELD ELECTROSTATICS */
449 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
450 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
452 /* Update potential sum for this i atom from the interaction with this j atom. */
453 velecsum = _mm256_add_pd(velecsum,velec);
457 /* Calculate temporary vectorial force */
458 tx = _mm256_mul_pd(fscal,dx20);
459 ty = _mm256_mul_pd(fscal,dy20);
460 tz = _mm256_mul_pd(fscal,dz20);
462 /* Update vectorial force */
463 fix2 = _mm256_add_pd(fix2,tx);
464 fiy2 = _mm256_add_pd(fiy2,ty);
465 fiz2 = _mm256_add_pd(fiz2,tz);
467 fjx0 = _mm256_add_pd(fjx0,tx);
468 fjy0 = _mm256_add_pd(fjy0,ty);
469 fjz0 = _mm256_add_pd(fjz0,tz);
471 /**************************
472 * CALCULATE INTERACTIONS *
473 **************************/
475 /* REACTION-FIELD ELECTROSTATICS */
476 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
477 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
479 /* Update potential sum for this i atom from the interaction with this j atom. */
480 velecsum = _mm256_add_pd(velecsum,velec);
484 /* Calculate temporary vectorial force */
485 tx = _mm256_mul_pd(fscal,dx21);
486 ty = _mm256_mul_pd(fscal,dy21);
487 tz = _mm256_mul_pd(fscal,dz21);
489 /* Update vectorial force */
490 fix2 = _mm256_add_pd(fix2,tx);
491 fiy2 = _mm256_add_pd(fiy2,ty);
492 fiz2 = _mm256_add_pd(fiz2,tz);
494 fjx1 = _mm256_add_pd(fjx1,tx);
495 fjy1 = _mm256_add_pd(fjy1,ty);
496 fjz1 = _mm256_add_pd(fjz1,tz);
498 /**************************
499 * CALCULATE INTERACTIONS *
500 **************************/
502 /* REACTION-FIELD ELECTROSTATICS */
503 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
504 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
506 /* Update potential sum for this i atom from the interaction with this j atom. */
507 velecsum = _mm256_add_pd(velecsum,velec);
511 /* Calculate temporary vectorial force */
512 tx = _mm256_mul_pd(fscal,dx22);
513 ty = _mm256_mul_pd(fscal,dy22);
514 tz = _mm256_mul_pd(fscal,dz22);
516 /* Update vectorial force */
517 fix2 = _mm256_add_pd(fix2,tx);
518 fiy2 = _mm256_add_pd(fiy2,ty);
519 fiz2 = _mm256_add_pd(fiz2,tz);
521 fjx2 = _mm256_add_pd(fjx2,tx);
522 fjy2 = _mm256_add_pd(fjy2,ty);
523 fjz2 = _mm256_add_pd(fjz2,tz);
525 fjptrA = f+j_coord_offsetA;
526 fjptrB = f+j_coord_offsetB;
527 fjptrC = f+j_coord_offsetC;
528 fjptrD = f+j_coord_offsetD;
530 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
531 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
533 /* Inner loop uses 288 flops */
539 /* Get j neighbor index, and coordinate index */
540 jnrlistA = jjnr[jidx];
541 jnrlistB = jjnr[jidx+1];
542 jnrlistC = jjnr[jidx+2];
543 jnrlistD = jjnr[jidx+3];
544 /* Sign of each element will be negative for non-real atoms.
545 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
546 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
548 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
550 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
551 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
552 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
554 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
555 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
556 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
557 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
558 j_coord_offsetA = DIM*jnrA;
559 j_coord_offsetB = DIM*jnrB;
560 j_coord_offsetC = DIM*jnrC;
561 j_coord_offsetD = DIM*jnrD;
563 /* load j atom coordinates */
564 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
565 x+j_coord_offsetC,x+j_coord_offsetD,
566 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
568 /* Calculate displacement vector */
569 dx00 = _mm256_sub_pd(ix0,jx0);
570 dy00 = _mm256_sub_pd(iy0,jy0);
571 dz00 = _mm256_sub_pd(iz0,jz0);
572 dx01 = _mm256_sub_pd(ix0,jx1);
573 dy01 = _mm256_sub_pd(iy0,jy1);
574 dz01 = _mm256_sub_pd(iz0,jz1);
575 dx02 = _mm256_sub_pd(ix0,jx2);
576 dy02 = _mm256_sub_pd(iy0,jy2);
577 dz02 = _mm256_sub_pd(iz0,jz2);
578 dx10 = _mm256_sub_pd(ix1,jx0);
579 dy10 = _mm256_sub_pd(iy1,jy0);
580 dz10 = _mm256_sub_pd(iz1,jz0);
581 dx11 = _mm256_sub_pd(ix1,jx1);
582 dy11 = _mm256_sub_pd(iy1,jy1);
583 dz11 = _mm256_sub_pd(iz1,jz1);
584 dx12 = _mm256_sub_pd(ix1,jx2);
585 dy12 = _mm256_sub_pd(iy1,jy2);
586 dz12 = _mm256_sub_pd(iz1,jz2);
587 dx20 = _mm256_sub_pd(ix2,jx0);
588 dy20 = _mm256_sub_pd(iy2,jy0);
589 dz20 = _mm256_sub_pd(iz2,jz0);
590 dx21 = _mm256_sub_pd(ix2,jx1);
591 dy21 = _mm256_sub_pd(iy2,jy1);
592 dz21 = _mm256_sub_pd(iz2,jz1);
593 dx22 = _mm256_sub_pd(ix2,jx2);
594 dy22 = _mm256_sub_pd(iy2,jy2);
595 dz22 = _mm256_sub_pd(iz2,jz2);
597 /* Calculate squared distance and things based on it */
598 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
599 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
600 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
601 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
602 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
603 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
604 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
605 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
606 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
608 rinv00 = avx256_invsqrt_d(rsq00);
609 rinv01 = avx256_invsqrt_d(rsq01);
610 rinv02 = avx256_invsqrt_d(rsq02);
611 rinv10 = avx256_invsqrt_d(rsq10);
612 rinv11 = avx256_invsqrt_d(rsq11);
613 rinv12 = avx256_invsqrt_d(rsq12);
614 rinv20 = avx256_invsqrt_d(rsq20);
615 rinv21 = avx256_invsqrt_d(rsq21);
616 rinv22 = avx256_invsqrt_d(rsq22);
618 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
619 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
620 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
621 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
622 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
623 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
624 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
625 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
626 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
628 fjx0 = _mm256_setzero_pd();
629 fjy0 = _mm256_setzero_pd();
630 fjz0 = _mm256_setzero_pd();
631 fjx1 = _mm256_setzero_pd();
632 fjy1 = _mm256_setzero_pd();
633 fjz1 = _mm256_setzero_pd();
634 fjx2 = _mm256_setzero_pd();
635 fjy2 = _mm256_setzero_pd();
636 fjz2 = _mm256_setzero_pd();
638 /**************************
639 * CALCULATE INTERACTIONS *
640 **************************/
642 /* REACTION-FIELD ELECTROSTATICS */
643 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
644 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
646 /* Update potential sum for this i atom from the interaction with this j atom. */
647 velec = _mm256_andnot_pd(dummy_mask,velec);
648 velecsum = _mm256_add_pd(velecsum,velec);
652 fscal = _mm256_andnot_pd(dummy_mask,fscal);
654 /* Calculate temporary vectorial force */
655 tx = _mm256_mul_pd(fscal,dx00);
656 ty = _mm256_mul_pd(fscal,dy00);
657 tz = _mm256_mul_pd(fscal,dz00);
659 /* Update vectorial force */
660 fix0 = _mm256_add_pd(fix0,tx);
661 fiy0 = _mm256_add_pd(fiy0,ty);
662 fiz0 = _mm256_add_pd(fiz0,tz);
664 fjx0 = _mm256_add_pd(fjx0,tx);
665 fjy0 = _mm256_add_pd(fjy0,ty);
666 fjz0 = _mm256_add_pd(fjz0,tz);
668 /**************************
669 * CALCULATE INTERACTIONS *
670 **************************/
672 /* REACTION-FIELD ELECTROSTATICS */
673 velec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
674 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
676 /* Update potential sum for this i atom from the interaction with this j atom. */
677 velec = _mm256_andnot_pd(dummy_mask,velec);
678 velecsum = _mm256_add_pd(velecsum,velec);
682 fscal = _mm256_andnot_pd(dummy_mask,fscal);
684 /* Calculate temporary vectorial force */
685 tx = _mm256_mul_pd(fscal,dx01);
686 ty = _mm256_mul_pd(fscal,dy01);
687 tz = _mm256_mul_pd(fscal,dz01);
689 /* Update vectorial force */
690 fix0 = _mm256_add_pd(fix0,tx);
691 fiy0 = _mm256_add_pd(fiy0,ty);
692 fiz0 = _mm256_add_pd(fiz0,tz);
694 fjx1 = _mm256_add_pd(fjx1,tx);
695 fjy1 = _mm256_add_pd(fjy1,ty);
696 fjz1 = _mm256_add_pd(fjz1,tz);
698 /**************************
699 * CALCULATE INTERACTIONS *
700 **************************/
702 /* REACTION-FIELD ELECTROSTATICS */
703 velec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
704 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
706 /* Update potential sum for this i atom from the interaction with this j atom. */
707 velec = _mm256_andnot_pd(dummy_mask,velec);
708 velecsum = _mm256_add_pd(velecsum,velec);
712 fscal = _mm256_andnot_pd(dummy_mask,fscal);
714 /* Calculate temporary vectorial force */
715 tx = _mm256_mul_pd(fscal,dx02);
716 ty = _mm256_mul_pd(fscal,dy02);
717 tz = _mm256_mul_pd(fscal,dz02);
719 /* Update vectorial force */
720 fix0 = _mm256_add_pd(fix0,tx);
721 fiy0 = _mm256_add_pd(fiy0,ty);
722 fiz0 = _mm256_add_pd(fiz0,tz);
724 fjx2 = _mm256_add_pd(fjx2,tx);
725 fjy2 = _mm256_add_pd(fjy2,ty);
726 fjz2 = _mm256_add_pd(fjz2,tz);
728 /**************************
729 * CALCULATE INTERACTIONS *
730 **************************/
732 /* REACTION-FIELD ELECTROSTATICS */
733 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
734 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
736 /* Update potential sum for this i atom from the interaction with this j atom. */
737 velec = _mm256_andnot_pd(dummy_mask,velec);
738 velecsum = _mm256_add_pd(velecsum,velec);
742 fscal = _mm256_andnot_pd(dummy_mask,fscal);
744 /* Calculate temporary vectorial force */
745 tx = _mm256_mul_pd(fscal,dx10);
746 ty = _mm256_mul_pd(fscal,dy10);
747 tz = _mm256_mul_pd(fscal,dz10);
749 /* Update vectorial force */
750 fix1 = _mm256_add_pd(fix1,tx);
751 fiy1 = _mm256_add_pd(fiy1,ty);
752 fiz1 = _mm256_add_pd(fiz1,tz);
754 fjx0 = _mm256_add_pd(fjx0,tx);
755 fjy0 = _mm256_add_pd(fjy0,ty);
756 fjz0 = _mm256_add_pd(fjz0,tz);
758 /**************************
759 * CALCULATE INTERACTIONS *
760 **************************/
762 /* REACTION-FIELD ELECTROSTATICS */
763 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
764 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
766 /* Update potential sum for this i atom from the interaction with this j atom. */
767 velec = _mm256_andnot_pd(dummy_mask,velec);
768 velecsum = _mm256_add_pd(velecsum,velec);
772 fscal = _mm256_andnot_pd(dummy_mask,fscal);
774 /* Calculate temporary vectorial force */
775 tx = _mm256_mul_pd(fscal,dx11);
776 ty = _mm256_mul_pd(fscal,dy11);
777 tz = _mm256_mul_pd(fscal,dz11);
779 /* Update vectorial force */
780 fix1 = _mm256_add_pd(fix1,tx);
781 fiy1 = _mm256_add_pd(fiy1,ty);
782 fiz1 = _mm256_add_pd(fiz1,tz);
784 fjx1 = _mm256_add_pd(fjx1,tx);
785 fjy1 = _mm256_add_pd(fjy1,ty);
786 fjz1 = _mm256_add_pd(fjz1,tz);
788 /**************************
789 * CALCULATE INTERACTIONS *
790 **************************/
792 /* REACTION-FIELD ELECTROSTATICS */
793 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
794 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
796 /* Update potential sum for this i atom from the interaction with this j atom. */
797 velec = _mm256_andnot_pd(dummy_mask,velec);
798 velecsum = _mm256_add_pd(velecsum,velec);
802 fscal = _mm256_andnot_pd(dummy_mask,fscal);
804 /* Calculate temporary vectorial force */
805 tx = _mm256_mul_pd(fscal,dx12);
806 ty = _mm256_mul_pd(fscal,dy12);
807 tz = _mm256_mul_pd(fscal,dz12);
809 /* Update vectorial force */
810 fix1 = _mm256_add_pd(fix1,tx);
811 fiy1 = _mm256_add_pd(fiy1,ty);
812 fiz1 = _mm256_add_pd(fiz1,tz);
814 fjx2 = _mm256_add_pd(fjx2,tx);
815 fjy2 = _mm256_add_pd(fjy2,ty);
816 fjz2 = _mm256_add_pd(fjz2,tz);
818 /**************************
819 * CALCULATE INTERACTIONS *
820 **************************/
822 /* REACTION-FIELD ELECTROSTATICS */
823 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
824 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
826 /* Update potential sum for this i atom from the interaction with this j atom. */
827 velec = _mm256_andnot_pd(dummy_mask,velec);
828 velecsum = _mm256_add_pd(velecsum,velec);
832 fscal = _mm256_andnot_pd(dummy_mask,fscal);
834 /* Calculate temporary vectorial force */
835 tx = _mm256_mul_pd(fscal,dx20);
836 ty = _mm256_mul_pd(fscal,dy20);
837 tz = _mm256_mul_pd(fscal,dz20);
839 /* Update vectorial force */
840 fix2 = _mm256_add_pd(fix2,tx);
841 fiy2 = _mm256_add_pd(fiy2,ty);
842 fiz2 = _mm256_add_pd(fiz2,tz);
844 fjx0 = _mm256_add_pd(fjx0,tx);
845 fjy0 = _mm256_add_pd(fjy0,ty);
846 fjz0 = _mm256_add_pd(fjz0,tz);
848 /**************************
849 * CALCULATE INTERACTIONS *
850 **************************/
852 /* REACTION-FIELD ELECTROSTATICS */
853 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
854 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
856 /* Update potential sum for this i atom from the interaction with this j atom. */
857 velec = _mm256_andnot_pd(dummy_mask,velec);
858 velecsum = _mm256_add_pd(velecsum,velec);
862 fscal = _mm256_andnot_pd(dummy_mask,fscal);
864 /* Calculate temporary vectorial force */
865 tx = _mm256_mul_pd(fscal,dx21);
866 ty = _mm256_mul_pd(fscal,dy21);
867 tz = _mm256_mul_pd(fscal,dz21);
869 /* Update vectorial force */
870 fix2 = _mm256_add_pd(fix2,tx);
871 fiy2 = _mm256_add_pd(fiy2,ty);
872 fiz2 = _mm256_add_pd(fiz2,tz);
874 fjx1 = _mm256_add_pd(fjx1,tx);
875 fjy1 = _mm256_add_pd(fjy1,ty);
876 fjz1 = _mm256_add_pd(fjz1,tz);
878 /**************************
879 * CALCULATE INTERACTIONS *
880 **************************/
882 /* REACTION-FIELD ELECTROSTATICS */
883 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
884 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
886 /* Update potential sum for this i atom from the interaction with this j atom. */
887 velec = _mm256_andnot_pd(dummy_mask,velec);
888 velecsum = _mm256_add_pd(velecsum,velec);
892 fscal = _mm256_andnot_pd(dummy_mask,fscal);
894 /* Calculate temporary vectorial force */
895 tx = _mm256_mul_pd(fscal,dx22);
896 ty = _mm256_mul_pd(fscal,dy22);
897 tz = _mm256_mul_pd(fscal,dz22);
899 /* Update vectorial force */
900 fix2 = _mm256_add_pd(fix2,tx);
901 fiy2 = _mm256_add_pd(fiy2,ty);
902 fiz2 = _mm256_add_pd(fiz2,tz);
904 fjx2 = _mm256_add_pd(fjx2,tx);
905 fjy2 = _mm256_add_pd(fjy2,ty);
906 fjz2 = _mm256_add_pd(fjz2,tz);
908 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
909 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
910 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
911 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
913 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
914 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
916 /* Inner loop uses 288 flops */
919 /* End of innermost loop */
921 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
922 f+i_coord_offset,fshift+i_shift_offset);
925 /* Update potential energies */
926 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
928 /* Increment number of inner iterations */
929 inneriter += j_index_end - j_index_start;
931 /* Outer loop uses 19 flops */
934 /* Increment number of outer iterations */
937 /* Update outer/inner flops */
939 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*288);
942 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_double
943 * Electrostatics interaction: ReactionField
944 * VdW interaction: None
945 * Geometry: Water3-Water3
946 * Calculate force/pot: Force
949 nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_double
950 (t_nblist * gmx_restrict nlist,
951 rvec * gmx_restrict xx,
952 rvec * gmx_restrict ff,
953 struct t_forcerec * gmx_restrict fr,
954 t_mdatoms * gmx_restrict mdatoms,
955 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
956 t_nrnb * gmx_restrict nrnb)
958 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
959 * just 0 for non-waters.
960 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
961 * jnr indices corresponding to data put in the four positions in the SIMD register.
963 int i_shift_offset,i_coord_offset,outeriter,inneriter;
964 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
965 int jnrA,jnrB,jnrC,jnrD;
966 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
967 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
968 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
969 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
971 real *shiftvec,*fshift,*x,*f;
972 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
974 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
975 real * vdwioffsetptr0;
976 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
977 real * vdwioffsetptr1;
978 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
979 real * vdwioffsetptr2;
980 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
981 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
982 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
983 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
984 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
985 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
986 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
987 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
988 __m256d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
989 __m256d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
990 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
991 __m256d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
992 __m256d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
993 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
994 __m256d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
995 __m256d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
996 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
998 __m256d dummy_mask,cutoff_mask;
999 __m128 tmpmask0,tmpmask1;
1000 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1001 __m256d one = _mm256_set1_pd(1.0);
1002 __m256d two = _mm256_set1_pd(2.0);
1008 jindex = nlist->jindex;
1010 shiftidx = nlist->shift;
1012 shiftvec = fr->shift_vec[0];
1013 fshift = fr->fshift[0];
1014 facel = _mm256_set1_pd(fr->ic->epsfac);
1015 charge = mdatoms->chargeA;
1016 krf = _mm256_set1_pd(fr->ic->k_rf);
1017 krf2 = _mm256_set1_pd(fr->ic->k_rf*2.0);
1018 crf = _mm256_set1_pd(fr->ic->c_rf);
1020 /* Setup water-specific parameters */
1021 inr = nlist->iinr[0];
1022 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1023 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1024 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1026 jq0 = _mm256_set1_pd(charge[inr+0]);
1027 jq1 = _mm256_set1_pd(charge[inr+1]);
1028 jq2 = _mm256_set1_pd(charge[inr+2]);
1029 qq00 = _mm256_mul_pd(iq0,jq0);
1030 qq01 = _mm256_mul_pd(iq0,jq1);
1031 qq02 = _mm256_mul_pd(iq0,jq2);
1032 qq10 = _mm256_mul_pd(iq1,jq0);
1033 qq11 = _mm256_mul_pd(iq1,jq1);
1034 qq12 = _mm256_mul_pd(iq1,jq2);
1035 qq20 = _mm256_mul_pd(iq2,jq0);
1036 qq21 = _mm256_mul_pd(iq2,jq1);
1037 qq22 = _mm256_mul_pd(iq2,jq2);
1039 /* Avoid stupid compiler warnings */
1040 jnrA = jnrB = jnrC = jnrD = 0;
1041 j_coord_offsetA = 0;
1042 j_coord_offsetB = 0;
1043 j_coord_offsetC = 0;
1044 j_coord_offsetD = 0;
1049 for(iidx=0;iidx<4*DIM;iidx++)
1051 scratch[iidx] = 0.0;
1054 /* Start outer loop over neighborlists */
1055 for(iidx=0; iidx<nri; iidx++)
1057 /* Load shift vector for this list */
1058 i_shift_offset = DIM*shiftidx[iidx];
1060 /* Load limits for loop over neighbors */
1061 j_index_start = jindex[iidx];
1062 j_index_end = jindex[iidx+1];
1064 /* Get outer coordinate index */
1066 i_coord_offset = DIM*inr;
1068 /* Load i particle coords and add shift vector */
1069 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1070 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1072 fix0 = _mm256_setzero_pd();
1073 fiy0 = _mm256_setzero_pd();
1074 fiz0 = _mm256_setzero_pd();
1075 fix1 = _mm256_setzero_pd();
1076 fiy1 = _mm256_setzero_pd();
1077 fiz1 = _mm256_setzero_pd();
1078 fix2 = _mm256_setzero_pd();
1079 fiy2 = _mm256_setzero_pd();
1080 fiz2 = _mm256_setzero_pd();
1082 /* Start inner kernel loop */
1083 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1086 /* Get j neighbor index, and coordinate index */
1088 jnrB = jjnr[jidx+1];
1089 jnrC = jjnr[jidx+2];
1090 jnrD = jjnr[jidx+3];
1091 j_coord_offsetA = DIM*jnrA;
1092 j_coord_offsetB = DIM*jnrB;
1093 j_coord_offsetC = DIM*jnrC;
1094 j_coord_offsetD = DIM*jnrD;
1096 /* load j atom coordinates */
1097 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1098 x+j_coord_offsetC,x+j_coord_offsetD,
1099 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1101 /* Calculate displacement vector */
1102 dx00 = _mm256_sub_pd(ix0,jx0);
1103 dy00 = _mm256_sub_pd(iy0,jy0);
1104 dz00 = _mm256_sub_pd(iz0,jz0);
1105 dx01 = _mm256_sub_pd(ix0,jx1);
1106 dy01 = _mm256_sub_pd(iy0,jy1);
1107 dz01 = _mm256_sub_pd(iz0,jz1);
1108 dx02 = _mm256_sub_pd(ix0,jx2);
1109 dy02 = _mm256_sub_pd(iy0,jy2);
1110 dz02 = _mm256_sub_pd(iz0,jz2);
1111 dx10 = _mm256_sub_pd(ix1,jx0);
1112 dy10 = _mm256_sub_pd(iy1,jy0);
1113 dz10 = _mm256_sub_pd(iz1,jz0);
1114 dx11 = _mm256_sub_pd(ix1,jx1);
1115 dy11 = _mm256_sub_pd(iy1,jy1);
1116 dz11 = _mm256_sub_pd(iz1,jz1);
1117 dx12 = _mm256_sub_pd(ix1,jx2);
1118 dy12 = _mm256_sub_pd(iy1,jy2);
1119 dz12 = _mm256_sub_pd(iz1,jz2);
1120 dx20 = _mm256_sub_pd(ix2,jx0);
1121 dy20 = _mm256_sub_pd(iy2,jy0);
1122 dz20 = _mm256_sub_pd(iz2,jz0);
1123 dx21 = _mm256_sub_pd(ix2,jx1);
1124 dy21 = _mm256_sub_pd(iy2,jy1);
1125 dz21 = _mm256_sub_pd(iz2,jz1);
1126 dx22 = _mm256_sub_pd(ix2,jx2);
1127 dy22 = _mm256_sub_pd(iy2,jy2);
1128 dz22 = _mm256_sub_pd(iz2,jz2);
1130 /* Calculate squared distance and things based on it */
1131 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1132 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1133 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1134 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1135 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1136 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1137 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1138 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1139 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1141 rinv00 = avx256_invsqrt_d(rsq00);
1142 rinv01 = avx256_invsqrt_d(rsq01);
1143 rinv02 = avx256_invsqrt_d(rsq02);
1144 rinv10 = avx256_invsqrt_d(rsq10);
1145 rinv11 = avx256_invsqrt_d(rsq11);
1146 rinv12 = avx256_invsqrt_d(rsq12);
1147 rinv20 = avx256_invsqrt_d(rsq20);
1148 rinv21 = avx256_invsqrt_d(rsq21);
1149 rinv22 = avx256_invsqrt_d(rsq22);
1151 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1152 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
1153 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
1154 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1155 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
1156 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
1157 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1158 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
1159 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
1161 fjx0 = _mm256_setzero_pd();
1162 fjy0 = _mm256_setzero_pd();
1163 fjz0 = _mm256_setzero_pd();
1164 fjx1 = _mm256_setzero_pd();
1165 fjy1 = _mm256_setzero_pd();
1166 fjz1 = _mm256_setzero_pd();
1167 fjx2 = _mm256_setzero_pd();
1168 fjy2 = _mm256_setzero_pd();
1169 fjz2 = _mm256_setzero_pd();
1171 /**************************
1172 * CALCULATE INTERACTIONS *
1173 **************************/
1175 /* REACTION-FIELD ELECTROSTATICS */
1176 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1180 /* Calculate temporary vectorial force */
1181 tx = _mm256_mul_pd(fscal,dx00);
1182 ty = _mm256_mul_pd(fscal,dy00);
1183 tz = _mm256_mul_pd(fscal,dz00);
1185 /* Update vectorial force */
1186 fix0 = _mm256_add_pd(fix0,tx);
1187 fiy0 = _mm256_add_pd(fiy0,ty);
1188 fiz0 = _mm256_add_pd(fiz0,tz);
1190 fjx0 = _mm256_add_pd(fjx0,tx);
1191 fjy0 = _mm256_add_pd(fjy0,ty);
1192 fjz0 = _mm256_add_pd(fjz0,tz);
1194 /**************************
1195 * CALCULATE INTERACTIONS *
1196 **************************/
1198 /* REACTION-FIELD ELECTROSTATICS */
1199 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1203 /* Calculate temporary vectorial force */
1204 tx = _mm256_mul_pd(fscal,dx01);
1205 ty = _mm256_mul_pd(fscal,dy01);
1206 tz = _mm256_mul_pd(fscal,dz01);
1208 /* Update vectorial force */
1209 fix0 = _mm256_add_pd(fix0,tx);
1210 fiy0 = _mm256_add_pd(fiy0,ty);
1211 fiz0 = _mm256_add_pd(fiz0,tz);
1213 fjx1 = _mm256_add_pd(fjx1,tx);
1214 fjy1 = _mm256_add_pd(fjy1,ty);
1215 fjz1 = _mm256_add_pd(fjz1,tz);
1217 /**************************
1218 * CALCULATE INTERACTIONS *
1219 **************************/
1221 /* REACTION-FIELD ELECTROSTATICS */
1222 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1226 /* Calculate temporary vectorial force */
1227 tx = _mm256_mul_pd(fscal,dx02);
1228 ty = _mm256_mul_pd(fscal,dy02);
1229 tz = _mm256_mul_pd(fscal,dz02);
1231 /* Update vectorial force */
1232 fix0 = _mm256_add_pd(fix0,tx);
1233 fiy0 = _mm256_add_pd(fiy0,ty);
1234 fiz0 = _mm256_add_pd(fiz0,tz);
1236 fjx2 = _mm256_add_pd(fjx2,tx);
1237 fjy2 = _mm256_add_pd(fjy2,ty);
1238 fjz2 = _mm256_add_pd(fjz2,tz);
1240 /**************************
1241 * CALCULATE INTERACTIONS *
1242 **************************/
1244 /* REACTION-FIELD ELECTROSTATICS */
1245 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1249 /* Calculate temporary vectorial force */
1250 tx = _mm256_mul_pd(fscal,dx10);
1251 ty = _mm256_mul_pd(fscal,dy10);
1252 tz = _mm256_mul_pd(fscal,dz10);
1254 /* Update vectorial force */
1255 fix1 = _mm256_add_pd(fix1,tx);
1256 fiy1 = _mm256_add_pd(fiy1,ty);
1257 fiz1 = _mm256_add_pd(fiz1,tz);
1259 fjx0 = _mm256_add_pd(fjx0,tx);
1260 fjy0 = _mm256_add_pd(fjy0,ty);
1261 fjz0 = _mm256_add_pd(fjz0,tz);
1263 /**************************
1264 * CALCULATE INTERACTIONS *
1265 **************************/
1267 /* REACTION-FIELD ELECTROSTATICS */
1268 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1272 /* Calculate temporary vectorial force */
1273 tx = _mm256_mul_pd(fscal,dx11);
1274 ty = _mm256_mul_pd(fscal,dy11);
1275 tz = _mm256_mul_pd(fscal,dz11);
1277 /* Update vectorial force */
1278 fix1 = _mm256_add_pd(fix1,tx);
1279 fiy1 = _mm256_add_pd(fiy1,ty);
1280 fiz1 = _mm256_add_pd(fiz1,tz);
1282 fjx1 = _mm256_add_pd(fjx1,tx);
1283 fjy1 = _mm256_add_pd(fjy1,ty);
1284 fjz1 = _mm256_add_pd(fjz1,tz);
1286 /**************************
1287 * CALCULATE INTERACTIONS *
1288 **************************/
1290 /* REACTION-FIELD ELECTROSTATICS */
1291 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1295 /* Calculate temporary vectorial force */
1296 tx = _mm256_mul_pd(fscal,dx12);
1297 ty = _mm256_mul_pd(fscal,dy12);
1298 tz = _mm256_mul_pd(fscal,dz12);
1300 /* Update vectorial force */
1301 fix1 = _mm256_add_pd(fix1,tx);
1302 fiy1 = _mm256_add_pd(fiy1,ty);
1303 fiz1 = _mm256_add_pd(fiz1,tz);
1305 fjx2 = _mm256_add_pd(fjx2,tx);
1306 fjy2 = _mm256_add_pd(fjy2,ty);
1307 fjz2 = _mm256_add_pd(fjz2,tz);
1309 /**************************
1310 * CALCULATE INTERACTIONS *
1311 **************************/
1313 /* REACTION-FIELD ELECTROSTATICS */
1314 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1318 /* Calculate temporary vectorial force */
1319 tx = _mm256_mul_pd(fscal,dx20);
1320 ty = _mm256_mul_pd(fscal,dy20);
1321 tz = _mm256_mul_pd(fscal,dz20);
1323 /* Update vectorial force */
1324 fix2 = _mm256_add_pd(fix2,tx);
1325 fiy2 = _mm256_add_pd(fiy2,ty);
1326 fiz2 = _mm256_add_pd(fiz2,tz);
1328 fjx0 = _mm256_add_pd(fjx0,tx);
1329 fjy0 = _mm256_add_pd(fjy0,ty);
1330 fjz0 = _mm256_add_pd(fjz0,tz);
1332 /**************************
1333 * CALCULATE INTERACTIONS *
1334 **************************/
1336 /* REACTION-FIELD ELECTROSTATICS */
1337 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1341 /* Calculate temporary vectorial force */
1342 tx = _mm256_mul_pd(fscal,dx21);
1343 ty = _mm256_mul_pd(fscal,dy21);
1344 tz = _mm256_mul_pd(fscal,dz21);
1346 /* Update vectorial force */
1347 fix2 = _mm256_add_pd(fix2,tx);
1348 fiy2 = _mm256_add_pd(fiy2,ty);
1349 fiz2 = _mm256_add_pd(fiz2,tz);
1351 fjx1 = _mm256_add_pd(fjx1,tx);
1352 fjy1 = _mm256_add_pd(fjy1,ty);
1353 fjz1 = _mm256_add_pd(fjz1,tz);
1355 /**************************
1356 * CALCULATE INTERACTIONS *
1357 **************************/
1359 /* REACTION-FIELD ELECTROSTATICS */
1360 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1364 /* Calculate temporary vectorial force */
1365 tx = _mm256_mul_pd(fscal,dx22);
1366 ty = _mm256_mul_pd(fscal,dy22);
1367 tz = _mm256_mul_pd(fscal,dz22);
1369 /* Update vectorial force */
1370 fix2 = _mm256_add_pd(fix2,tx);
1371 fiy2 = _mm256_add_pd(fiy2,ty);
1372 fiz2 = _mm256_add_pd(fiz2,tz);
1374 fjx2 = _mm256_add_pd(fjx2,tx);
1375 fjy2 = _mm256_add_pd(fjy2,ty);
1376 fjz2 = _mm256_add_pd(fjz2,tz);
1378 fjptrA = f+j_coord_offsetA;
1379 fjptrB = f+j_coord_offsetB;
1380 fjptrC = f+j_coord_offsetC;
1381 fjptrD = f+j_coord_offsetD;
1383 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1384 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1386 /* Inner loop uses 243 flops */
1389 if(jidx<j_index_end)
1392 /* Get j neighbor index, and coordinate index */
1393 jnrlistA = jjnr[jidx];
1394 jnrlistB = jjnr[jidx+1];
1395 jnrlistC = jjnr[jidx+2];
1396 jnrlistD = jjnr[jidx+3];
1397 /* Sign of each element will be negative for non-real atoms.
1398 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1399 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1401 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1403 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1404 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1405 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1407 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1408 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1409 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1410 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1411 j_coord_offsetA = DIM*jnrA;
1412 j_coord_offsetB = DIM*jnrB;
1413 j_coord_offsetC = DIM*jnrC;
1414 j_coord_offsetD = DIM*jnrD;
1416 /* load j atom coordinates */
1417 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1418 x+j_coord_offsetC,x+j_coord_offsetD,
1419 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1421 /* Calculate displacement vector */
1422 dx00 = _mm256_sub_pd(ix0,jx0);
1423 dy00 = _mm256_sub_pd(iy0,jy0);
1424 dz00 = _mm256_sub_pd(iz0,jz0);
1425 dx01 = _mm256_sub_pd(ix0,jx1);
1426 dy01 = _mm256_sub_pd(iy0,jy1);
1427 dz01 = _mm256_sub_pd(iz0,jz1);
1428 dx02 = _mm256_sub_pd(ix0,jx2);
1429 dy02 = _mm256_sub_pd(iy0,jy2);
1430 dz02 = _mm256_sub_pd(iz0,jz2);
1431 dx10 = _mm256_sub_pd(ix1,jx0);
1432 dy10 = _mm256_sub_pd(iy1,jy0);
1433 dz10 = _mm256_sub_pd(iz1,jz0);
1434 dx11 = _mm256_sub_pd(ix1,jx1);
1435 dy11 = _mm256_sub_pd(iy1,jy1);
1436 dz11 = _mm256_sub_pd(iz1,jz1);
1437 dx12 = _mm256_sub_pd(ix1,jx2);
1438 dy12 = _mm256_sub_pd(iy1,jy2);
1439 dz12 = _mm256_sub_pd(iz1,jz2);
1440 dx20 = _mm256_sub_pd(ix2,jx0);
1441 dy20 = _mm256_sub_pd(iy2,jy0);
1442 dz20 = _mm256_sub_pd(iz2,jz0);
1443 dx21 = _mm256_sub_pd(ix2,jx1);
1444 dy21 = _mm256_sub_pd(iy2,jy1);
1445 dz21 = _mm256_sub_pd(iz2,jz1);
1446 dx22 = _mm256_sub_pd(ix2,jx2);
1447 dy22 = _mm256_sub_pd(iy2,jy2);
1448 dz22 = _mm256_sub_pd(iz2,jz2);
1450 /* Calculate squared distance and things based on it */
1451 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1452 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1453 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1454 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1455 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1456 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1457 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1458 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1459 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1461 rinv00 = avx256_invsqrt_d(rsq00);
1462 rinv01 = avx256_invsqrt_d(rsq01);
1463 rinv02 = avx256_invsqrt_d(rsq02);
1464 rinv10 = avx256_invsqrt_d(rsq10);
1465 rinv11 = avx256_invsqrt_d(rsq11);
1466 rinv12 = avx256_invsqrt_d(rsq12);
1467 rinv20 = avx256_invsqrt_d(rsq20);
1468 rinv21 = avx256_invsqrt_d(rsq21);
1469 rinv22 = avx256_invsqrt_d(rsq22);
1471 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1472 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
1473 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
1474 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1475 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
1476 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
1477 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1478 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
1479 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
1481 fjx0 = _mm256_setzero_pd();
1482 fjy0 = _mm256_setzero_pd();
1483 fjz0 = _mm256_setzero_pd();
1484 fjx1 = _mm256_setzero_pd();
1485 fjy1 = _mm256_setzero_pd();
1486 fjz1 = _mm256_setzero_pd();
1487 fjx2 = _mm256_setzero_pd();
1488 fjy2 = _mm256_setzero_pd();
1489 fjz2 = _mm256_setzero_pd();
1491 /**************************
1492 * CALCULATE INTERACTIONS *
1493 **************************/
1495 /* REACTION-FIELD ELECTROSTATICS */
1496 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1500 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1502 /* Calculate temporary vectorial force */
1503 tx = _mm256_mul_pd(fscal,dx00);
1504 ty = _mm256_mul_pd(fscal,dy00);
1505 tz = _mm256_mul_pd(fscal,dz00);
1507 /* Update vectorial force */
1508 fix0 = _mm256_add_pd(fix0,tx);
1509 fiy0 = _mm256_add_pd(fiy0,ty);
1510 fiz0 = _mm256_add_pd(fiz0,tz);
1512 fjx0 = _mm256_add_pd(fjx0,tx);
1513 fjy0 = _mm256_add_pd(fjy0,ty);
1514 fjz0 = _mm256_add_pd(fjz0,tz);
1516 /**************************
1517 * CALCULATE INTERACTIONS *
1518 **************************/
1520 /* REACTION-FIELD ELECTROSTATICS */
1521 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1525 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1527 /* Calculate temporary vectorial force */
1528 tx = _mm256_mul_pd(fscal,dx01);
1529 ty = _mm256_mul_pd(fscal,dy01);
1530 tz = _mm256_mul_pd(fscal,dz01);
1532 /* Update vectorial force */
1533 fix0 = _mm256_add_pd(fix0,tx);
1534 fiy0 = _mm256_add_pd(fiy0,ty);
1535 fiz0 = _mm256_add_pd(fiz0,tz);
1537 fjx1 = _mm256_add_pd(fjx1,tx);
1538 fjy1 = _mm256_add_pd(fjy1,ty);
1539 fjz1 = _mm256_add_pd(fjz1,tz);
1541 /**************************
1542 * CALCULATE INTERACTIONS *
1543 **************************/
1545 /* REACTION-FIELD ELECTROSTATICS */
1546 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1550 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1552 /* Calculate temporary vectorial force */
1553 tx = _mm256_mul_pd(fscal,dx02);
1554 ty = _mm256_mul_pd(fscal,dy02);
1555 tz = _mm256_mul_pd(fscal,dz02);
1557 /* Update vectorial force */
1558 fix0 = _mm256_add_pd(fix0,tx);
1559 fiy0 = _mm256_add_pd(fiy0,ty);
1560 fiz0 = _mm256_add_pd(fiz0,tz);
1562 fjx2 = _mm256_add_pd(fjx2,tx);
1563 fjy2 = _mm256_add_pd(fjy2,ty);
1564 fjz2 = _mm256_add_pd(fjz2,tz);
1566 /**************************
1567 * CALCULATE INTERACTIONS *
1568 **************************/
1570 /* REACTION-FIELD ELECTROSTATICS */
1571 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1575 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1577 /* Calculate temporary vectorial force */
1578 tx = _mm256_mul_pd(fscal,dx10);
1579 ty = _mm256_mul_pd(fscal,dy10);
1580 tz = _mm256_mul_pd(fscal,dz10);
1582 /* Update vectorial force */
1583 fix1 = _mm256_add_pd(fix1,tx);
1584 fiy1 = _mm256_add_pd(fiy1,ty);
1585 fiz1 = _mm256_add_pd(fiz1,tz);
1587 fjx0 = _mm256_add_pd(fjx0,tx);
1588 fjy0 = _mm256_add_pd(fjy0,ty);
1589 fjz0 = _mm256_add_pd(fjz0,tz);
1591 /**************************
1592 * CALCULATE INTERACTIONS *
1593 **************************/
1595 /* REACTION-FIELD ELECTROSTATICS */
1596 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1600 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1602 /* Calculate temporary vectorial force */
1603 tx = _mm256_mul_pd(fscal,dx11);
1604 ty = _mm256_mul_pd(fscal,dy11);
1605 tz = _mm256_mul_pd(fscal,dz11);
1607 /* Update vectorial force */
1608 fix1 = _mm256_add_pd(fix1,tx);
1609 fiy1 = _mm256_add_pd(fiy1,ty);
1610 fiz1 = _mm256_add_pd(fiz1,tz);
1612 fjx1 = _mm256_add_pd(fjx1,tx);
1613 fjy1 = _mm256_add_pd(fjy1,ty);
1614 fjz1 = _mm256_add_pd(fjz1,tz);
1616 /**************************
1617 * CALCULATE INTERACTIONS *
1618 **************************/
1620 /* REACTION-FIELD ELECTROSTATICS */
1621 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1625 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1627 /* Calculate temporary vectorial force */
1628 tx = _mm256_mul_pd(fscal,dx12);
1629 ty = _mm256_mul_pd(fscal,dy12);
1630 tz = _mm256_mul_pd(fscal,dz12);
1632 /* Update vectorial force */
1633 fix1 = _mm256_add_pd(fix1,tx);
1634 fiy1 = _mm256_add_pd(fiy1,ty);
1635 fiz1 = _mm256_add_pd(fiz1,tz);
1637 fjx2 = _mm256_add_pd(fjx2,tx);
1638 fjy2 = _mm256_add_pd(fjy2,ty);
1639 fjz2 = _mm256_add_pd(fjz2,tz);
1641 /**************************
1642 * CALCULATE INTERACTIONS *
1643 **************************/
1645 /* REACTION-FIELD ELECTROSTATICS */
1646 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1650 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1652 /* Calculate temporary vectorial force */
1653 tx = _mm256_mul_pd(fscal,dx20);
1654 ty = _mm256_mul_pd(fscal,dy20);
1655 tz = _mm256_mul_pd(fscal,dz20);
1657 /* Update vectorial force */
1658 fix2 = _mm256_add_pd(fix2,tx);
1659 fiy2 = _mm256_add_pd(fiy2,ty);
1660 fiz2 = _mm256_add_pd(fiz2,tz);
1662 fjx0 = _mm256_add_pd(fjx0,tx);
1663 fjy0 = _mm256_add_pd(fjy0,ty);
1664 fjz0 = _mm256_add_pd(fjz0,tz);
1666 /**************************
1667 * CALCULATE INTERACTIONS *
1668 **************************/
1670 /* REACTION-FIELD ELECTROSTATICS */
1671 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1675 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1677 /* Calculate temporary vectorial force */
1678 tx = _mm256_mul_pd(fscal,dx21);
1679 ty = _mm256_mul_pd(fscal,dy21);
1680 tz = _mm256_mul_pd(fscal,dz21);
1682 /* Update vectorial force */
1683 fix2 = _mm256_add_pd(fix2,tx);
1684 fiy2 = _mm256_add_pd(fiy2,ty);
1685 fiz2 = _mm256_add_pd(fiz2,tz);
1687 fjx1 = _mm256_add_pd(fjx1,tx);
1688 fjy1 = _mm256_add_pd(fjy1,ty);
1689 fjz1 = _mm256_add_pd(fjz1,tz);
1691 /**************************
1692 * CALCULATE INTERACTIONS *
1693 **************************/
1695 /* REACTION-FIELD ELECTROSTATICS */
1696 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1700 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1702 /* Calculate temporary vectorial force */
1703 tx = _mm256_mul_pd(fscal,dx22);
1704 ty = _mm256_mul_pd(fscal,dy22);
1705 tz = _mm256_mul_pd(fscal,dz22);
1707 /* Update vectorial force */
1708 fix2 = _mm256_add_pd(fix2,tx);
1709 fiy2 = _mm256_add_pd(fiy2,ty);
1710 fiz2 = _mm256_add_pd(fiz2,tz);
1712 fjx2 = _mm256_add_pd(fjx2,tx);
1713 fjy2 = _mm256_add_pd(fjy2,ty);
1714 fjz2 = _mm256_add_pd(fjz2,tz);
1716 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1717 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1718 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1719 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1721 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1722 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1724 /* Inner loop uses 243 flops */
1727 /* End of innermost loop */
1729 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1730 f+i_coord_offset,fshift+i_shift_offset);
1732 /* Increment number of inner iterations */
1733 inneriter += j_index_end - j_index_start;
1735 /* Outer loop uses 18 flops */
1738 /* Increment number of outer iterations */
1741 /* Update outer/inner flops */
1743 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*243);