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 avx_256_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_avx_256_double.h"
48 #include "kernelutil_x86_avx_256_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_avx_256_double
52 * Electrostatics interaction: ReactionField
53 * VdW interaction: None
54 * Geometry: Water3-Water3
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_avx_256_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,C,D refer to j loop unrolling done with AVX, e.g. for the four 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;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
76 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
77 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
83 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
84 real * vdwioffsetptr0;
85 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
86 real * vdwioffsetptr1;
87 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
88 real * vdwioffsetptr2;
89 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
90 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
91 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
92 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
93 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
94 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
95 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
96 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
97 __m256d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
98 __m256d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
99 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
100 __m256d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
101 __m256d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
102 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
103 __m256d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
104 __m256d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
105 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
107 __m256d dummy_mask,cutoff_mask;
108 __m128 tmpmask0,tmpmask1;
109 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
110 __m256d one = _mm256_set1_pd(1.0);
111 __m256d two = _mm256_set1_pd(2.0);
117 jindex = nlist->jindex;
119 shiftidx = nlist->shift;
121 shiftvec = fr->shift_vec[0];
122 fshift = fr->fshift[0];
123 facel = _mm256_set1_pd(fr->epsfac);
124 charge = mdatoms->chargeA;
125 krf = _mm256_set1_pd(fr->ic->k_rf);
126 krf2 = _mm256_set1_pd(fr->ic->k_rf*2.0);
127 crf = _mm256_set1_pd(fr->ic->c_rf);
129 /* Setup water-specific parameters */
130 inr = nlist->iinr[0];
131 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
132 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
133 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
135 jq0 = _mm256_set1_pd(charge[inr+0]);
136 jq1 = _mm256_set1_pd(charge[inr+1]);
137 jq2 = _mm256_set1_pd(charge[inr+2]);
138 qq00 = _mm256_mul_pd(iq0,jq0);
139 qq01 = _mm256_mul_pd(iq0,jq1);
140 qq02 = _mm256_mul_pd(iq0,jq2);
141 qq10 = _mm256_mul_pd(iq1,jq0);
142 qq11 = _mm256_mul_pd(iq1,jq1);
143 qq12 = _mm256_mul_pd(iq1,jq2);
144 qq20 = _mm256_mul_pd(iq2,jq0);
145 qq21 = _mm256_mul_pd(iq2,jq1);
146 qq22 = _mm256_mul_pd(iq2,jq2);
148 /* Avoid stupid compiler warnings */
149 jnrA = jnrB = jnrC = jnrD = 0;
158 for(iidx=0;iidx<4*DIM;iidx++)
163 /* Start outer loop over neighborlists */
164 for(iidx=0; iidx<nri; iidx++)
166 /* Load shift vector for this list */
167 i_shift_offset = DIM*shiftidx[iidx];
169 /* Load limits for loop over neighbors */
170 j_index_start = jindex[iidx];
171 j_index_end = jindex[iidx+1];
173 /* Get outer coordinate index */
175 i_coord_offset = DIM*inr;
177 /* Load i particle coords and add shift vector */
178 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
179 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
181 fix0 = _mm256_setzero_pd();
182 fiy0 = _mm256_setzero_pd();
183 fiz0 = _mm256_setzero_pd();
184 fix1 = _mm256_setzero_pd();
185 fiy1 = _mm256_setzero_pd();
186 fiz1 = _mm256_setzero_pd();
187 fix2 = _mm256_setzero_pd();
188 fiy2 = _mm256_setzero_pd();
189 fiz2 = _mm256_setzero_pd();
191 /* Reset potential sums */
192 velecsum = _mm256_setzero_pd();
194 /* Start inner kernel loop */
195 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
198 /* Get j neighbor index, and coordinate index */
203 j_coord_offsetA = DIM*jnrA;
204 j_coord_offsetB = DIM*jnrB;
205 j_coord_offsetC = DIM*jnrC;
206 j_coord_offsetD = DIM*jnrD;
208 /* load j atom coordinates */
209 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
210 x+j_coord_offsetC,x+j_coord_offsetD,
211 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
213 /* Calculate displacement vector */
214 dx00 = _mm256_sub_pd(ix0,jx0);
215 dy00 = _mm256_sub_pd(iy0,jy0);
216 dz00 = _mm256_sub_pd(iz0,jz0);
217 dx01 = _mm256_sub_pd(ix0,jx1);
218 dy01 = _mm256_sub_pd(iy0,jy1);
219 dz01 = _mm256_sub_pd(iz0,jz1);
220 dx02 = _mm256_sub_pd(ix0,jx2);
221 dy02 = _mm256_sub_pd(iy0,jy2);
222 dz02 = _mm256_sub_pd(iz0,jz2);
223 dx10 = _mm256_sub_pd(ix1,jx0);
224 dy10 = _mm256_sub_pd(iy1,jy0);
225 dz10 = _mm256_sub_pd(iz1,jz0);
226 dx11 = _mm256_sub_pd(ix1,jx1);
227 dy11 = _mm256_sub_pd(iy1,jy1);
228 dz11 = _mm256_sub_pd(iz1,jz1);
229 dx12 = _mm256_sub_pd(ix1,jx2);
230 dy12 = _mm256_sub_pd(iy1,jy2);
231 dz12 = _mm256_sub_pd(iz1,jz2);
232 dx20 = _mm256_sub_pd(ix2,jx0);
233 dy20 = _mm256_sub_pd(iy2,jy0);
234 dz20 = _mm256_sub_pd(iz2,jz0);
235 dx21 = _mm256_sub_pd(ix2,jx1);
236 dy21 = _mm256_sub_pd(iy2,jy1);
237 dz21 = _mm256_sub_pd(iz2,jz1);
238 dx22 = _mm256_sub_pd(ix2,jx2);
239 dy22 = _mm256_sub_pd(iy2,jy2);
240 dz22 = _mm256_sub_pd(iz2,jz2);
242 /* Calculate squared distance and things based on it */
243 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
244 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
245 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
246 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
247 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
248 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
249 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
250 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
251 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
253 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
254 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
255 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
256 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
257 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
258 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
259 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
260 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
261 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
263 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
264 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
265 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
266 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
267 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
268 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
269 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
270 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
271 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
273 fjx0 = _mm256_setzero_pd();
274 fjy0 = _mm256_setzero_pd();
275 fjz0 = _mm256_setzero_pd();
276 fjx1 = _mm256_setzero_pd();
277 fjy1 = _mm256_setzero_pd();
278 fjz1 = _mm256_setzero_pd();
279 fjx2 = _mm256_setzero_pd();
280 fjy2 = _mm256_setzero_pd();
281 fjz2 = _mm256_setzero_pd();
283 /**************************
284 * CALCULATE INTERACTIONS *
285 **************************/
287 /* REACTION-FIELD ELECTROSTATICS */
288 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
289 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
291 /* Update potential sum for this i atom from the interaction with this j atom. */
292 velecsum = _mm256_add_pd(velecsum,velec);
296 /* Calculate temporary vectorial force */
297 tx = _mm256_mul_pd(fscal,dx00);
298 ty = _mm256_mul_pd(fscal,dy00);
299 tz = _mm256_mul_pd(fscal,dz00);
301 /* Update vectorial force */
302 fix0 = _mm256_add_pd(fix0,tx);
303 fiy0 = _mm256_add_pd(fiy0,ty);
304 fiz0 = _mm256_add_pd(fiz0,tz);
306 fjx0 = _mm256_add_pd(fjx0,tx);
307 fjy0 = _mm256_add_pd(fjy0,ty);
308 fjz0 = _mm256_add_pd(fjz0,tz);
310 /**************************
311 * CALCULATE INTERACTIONS *
312 **************************/
314 /* REACTION-FIELD ELECTROSTATICS */
315 velec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
316 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
318 /* Update potential sum for this i atom from the interaction with this j atom. */
319 velecsum = _mm256_add_pd(velecsum,velec);
323 /* Calculate temporary vectorial force */
324 tx = _mm256_mul_pd(fscal,dx01);
325 ty = _mm256_mul_pd(fscal,dy01);
326 tz = _mm256_mul_pd(fscal,dz01);
328 /* Update vectorial force */
329 fix0 = _mm256_add_pd(fix0,tx);
330 fiy0 = _mm256_add_pd(fiy0,ty);
331 fiz0 = _mm256_add_pd(fiz0,tz);
333 fjx1 = _mm256_add_pd(fjx1,tx);
334 fjy1 = _mm256_add_pd(fjy1,ty);
335 fjz1 = _mm256_add_pd(fjz1,tz);
337 /**************************
338 * CALCULATE INTERACTIONS *
339 **************************/
341 /* REACTION-FIELD ELECTROSTATICS */
342 velec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
343 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
345 /* Update potential sum for this i atom from the interaction with this j atom. */
346 velecsum = _mm256_add_pd(velecsum,velec);
350 /* Calculate temporary vectorial force */
351 tx = _mm256_mul_pd(fscal,dx02);
352 ty = _mm256_mul_pd(fscal,dy02);
353 tz = _mm256_mul_pd(fscal,dz02);
355 /* Update vectorial force */
356 fix0 = _mm256_add_pd(fix0,tx);
357 fiy0 = _mm256_add_pd(fiy0,ty);
358 fiz0 = _mm256_add_pd(fiz0,tz);
360 fjx2 = _mm256_add_pd(fjx2,tx);
361 fjy2 = _mm256_add_pd(fjy2,ty);
362 fjz2 = _mm256_add_pd(fjz2,tz);
364 /**************************
365 * CALCULATE INTERACTIONS *
366 **************************/
368 /* REACTION-FIELD ELECTROSTATICS */
369 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
370 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
372 /* Update potential sum for this i atom from the interaction with this j atom. */
373 velecsum = _mm256_add_pd(velecsum,velec);
377 /* Calculate temporary vectorial force */
378 tx = _mm256_mul_pd(fscal,dx10);
379 ty = _mm256_mul_pd(fscal,dy10);
380 tz = _mm256_mul_pd(fscal,dz10);
382 /* Update vectorial force */
383 fix1 = _mm256_add_pd(fix1,tx);
384 fiy1 = _mm256_add_pd(fiy1,ty);
385 fiz1 = _mm256_add_pd(fiz1,tz);
387 fjx0 = _mm256_add_pd(fjx0,tx);
388 fjy0 = _mm256_add_pd(fjy0,ty);
389 fjz0 = _mm256_add_pd(fjz0,tz);
391 /**************************
392 * CALCULATE INTERACTIONS *
393 **************************/
395 /* REACTION-FIELD ELECTROSTATICS */
396 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
397 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
399 /* Update potential sum for this i atom from the interaction with this j atom. */
400 velecsum = _mm256_add_pd(velecsum,velec);
404 /* Calculate temporary vectorial force */
405 tx = _mm256_mul_pd(fscal,dx11);
406 ty = _mm256_mul_pd(fscal,dy11);
407 tz = _mm256_mul_pd(fscal,dz11);
409 /* Update vectorial force */
410 fix1 = _mm256_add_pd(fix1,tx);
411 fiy1 = _mm256_add_pd(fiy1,ty);
412 fiz1 = _mm256_add_pd(fiz1,tz);
414 fjx1 = _mm256_add_pd(fjx1,tx);
415 fjy1 = _mm256_add_pd(fjy1,ty);
416 fjz1 = _mm256_add_pd(fjz1,tz);
418 /**************************
419 * CALCULATE INTERACTIONS *
420 **************************/
422 /* REACTION-FIELD ELECTROSTATICS */
423 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
424 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
426 /* Update potential sum for this i atom from the interaction with this j atom. */
427 velecsum = _mm256_add_pd(velecsum,velec);
431 /* Calculate temporary vectorial force */
432 tx = _mm256_mul_pd(fscal,dx12);
433 ty = _mm256_mul_pd(fscal,dy12);
434 tz = _mm256_mul_pd(fscal,dz12);
436 /* Update vectorial force */
437 fix1 = _mm256_add_pd(fix1,tx);
438 fiy1 = _mm256_add_pd(fiy1,ty);
439 fiz1 = _mm256_add_pd(fiz1,tz);
441 fjx2 = _mm256_add_pd(fjx2,tx);
442 fjy2 = _mm256_add_pd(fjy2,ty);
443 fjz2 = _mm256_add_pd(fjz2,tz);
445 /**************************
446 * CALCULATE INTERACTIONS *
447 **************************/
449 /* REACTION-FIELD ELECTROSTATICS */
450 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
451 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
453 /* Update potential sum for this i atom from the interaction with this j atom. */
454 velecsum = _mm256_add_pd(velecsum,velec);
458 /* Calculate temporary vectorial force */
459 tx = _mm256_mul_pd(fscal,dx20);
460 ty = _mm256_mul_pd(fscal,dy20);
461 tz = _mm256_mul_pd(fscal,dz20);
463 /* Update vectorial force */
464 fix2 = _mm256_add_pd(fix2,tx);
465 fiy2 = _mm256_add_pd(fiy2,ty);
466 fiz2 = _mm256_add_pd(fiz2,tz);
468 fjx0 = _mm256_add_pd(fjx0,tx);
469 fjy0 = _mm256_add_pd(fjy0,ty);
470 fjz0 = _mm256_add_pd(fjz0,tz);
472 /**************************
473 * CALCULATE INTERACTIONS *
474 **************************/
476 /* REACTION-FIELD ELECTROSTATICS */
477 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
478 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
480 /* Update potential sum for this i atom from the interaction with this j atom. */
481 velecsum = _mm256_add_pd(velecsum,velec);
485 /* Calculate temporary vectorial force */
486 tx = _mm256_mul_pd(fscal,dx21);
487 ty = _mm256_mul_pd(fscal,dy21);
488 tz = _mm256_mul_pd(fscal,dz21);
490 /* Update vectorial force */
491 fix2 = _mm256_add_pd(fix2,tx);
492 fiy2 = _mm256_add_pd(fiy2,ty);
493 fiz2 = _mm256_add_pd(fiz2,tz);
495 fjx1 = _mm256_add_pd(fjx1,tx);
496 fjy1 = _mm256_add_pd(fjy1,ty);
497 fjz1 = _mm256_add_pd(fjz1,tz);
499 /**************************
500 * CALCULATE INTERACTIONS *
501 **************************/
503 /* REACTION-FIELD ELECTROSTATICS */
504 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
505 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
507 /* Update potential sum for this i atom from the interaction with this j atom. */
508 velecsum = _mm256_add_pd(velecsum,velec);
512 /* Calculate temporary vectorial force */
513 tx = _mm256_mul_pd(fscal,dx22);
514 ty = _mm256_mul_pd(fscal,dy22);
515 tz = _mm256_mul_pd(fscal,dz22);
517 /* Update vectorial force */
518 fix2 = _mm256_add_pd(fix2,tx);
519 fiy2 = _mm256_add_pd(fiy2,ty);
520 fiz2 = _mm256_add_pd(fiz2,tz);
522 fjx2 = _mm256_add_pd(fjx2,tx);
523 fjy2 = _mm256_add_pd(fjy2,ty);
524 fjz2 = _mm256_add_pd(fjz2,tz);
526 fjptrA = f+j_coord_offsetA;
527 fjptrB = f+j_coord_offsetB;
528 fjptrC = f+j_coord_offsetC;
529 fjptrD = f+j_coord_offsetD;
531 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
532 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
534 /* Inner loop uses 288 flops */
540 /* Get j neighbor index, and coordinate index */
541 jnrlistA = jjnr[jidx];
542 jnrlistB = jjnr[jidx+1];
543 jnrlistC = jjnr[jidx+2];
544 jnrlistD = jjnr[jidx+3];
545 /* Sign of each element will be negative for non-real atoms.
546 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
547 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
549 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
551 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
552 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
553 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
555 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
556 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
557 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
558 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
559 j_coord_offsetA = DIM*jnrA;
560 j_coord_offsetB = DIM*jnrB;
561 j_coord_offsetC = DIM*jnrC;
562 j_coord_offsetD = DIM*jnrD;
564 /* load j atom coordinates */
565 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
566 x+j_coord_offsetC,x+j_coord_offsetD,
567 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
569 /* Calculate displacement vector */
570 dx00 = _mm256_sub_pd(ix0,jx0);
571 dy00 = _mm256_sub_pd(iy0,jy0);
572 dz00 = _mm256_sub_pd(iz0,jz0);
573 dx01 = _mm256_sub_pd(ix0,jx1);
574 dy01 = _mm256_sub_pd(iy0,jy1);
575 dz01 = _mm256_sub_pd(iz0,jz1);
576 dx02 = _mm256_sub_pd(ix0,jx2);
577 dy02 = _mm256_sub_pd(iy0,jy2);
578 dz02 = _mm256_sub_pd(iz0,jz2);
579 dx10 = _mm256_sub_pd(ix1,jx0);
580 dy10 = _mm256_sub_pd(iy1,jy0);
581 dz10 = _mm256_sub_pd(iz1,jz0);
582 dx11 = _mm256_sub_pd(ix1,jx1);
583 dy11 = _mm256_sub_pd(iy1,jy1);
584 dz11 = _mm256_sub_pd(iz1,jz1);
585 dx12 = _mm256_sub_pd(ix1,jx2);
586 dy12 = _mm256_sub_pd(iy1,jy2);
587 dz12 = _mm256_sub_pd(iz1,jz2);
588 dx20 = _mm256_sub_pd(ix2,jx0);
589 dy20 = _mm256_sub_pd(iy2,jy0);
590 dz20 = _mm256_sub_pd(iz2,jz0);
591 dx21 = _mm256_sub_pd(ix2,jx1);
592 dy21 = _mm256_sub_pd(iy2,jy1);
593 dz21 = _mm256_sub_pd(iz2,jz1);
594 dx22 = _mm256_sub_pd(ix2,jx2);
595 dy22 = _mm256_sub_pd(iy2,jy2);
596 dz22 = _mm256_sub_pd(iz2,jz2);
598 /* Calculate squared distance and things based on it */
599 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
600 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
601 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
602 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
603 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
604 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
605 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
606 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
607 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
609 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
610 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
611 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
612 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
613 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
614 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
615 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
616 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
617 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
619 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
620 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
621 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
622 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
623 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
624 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
625 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
626 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
627 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
629 fjx0 = _mm256_setzero_pd();
630 fjy0 = _mm256_setzero_pd();
631 fjz0 = _mm256_setzero_pd();
632 fjx1 = _mm256_setzero_pd();
633 fjy1 = _mm256_setzero_pd();
634 fjz1 = _mm256_setzero_pd();
635 fjx2 = _mm256_setzero_pd();
636 fjy2 = _mm256_setzero_pd();
637 fjz2 = _mm256_setzero_pd();
639 /**************************
640 * CALCULATE INTERACTIONS *
641 **************************/
643 /* REACTION-FIELD ELECTROSTATICS */
644 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
645 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
647 /* Update potential sum for this i atom from the interaction with this j atom. */
648 velec = _mm256_andnot_pd(dummy_mask,velec);
649 velecsum = _mm256_add_pd(velecsum,velec);
653 fscal = _mm256_andnot_pd(dummy_mask,fscal);
655 /* Calculate temporary vectorial force */
656 tx = _mm256_mul_pd(fscal,dx00);
657 ty = _mm256_mul_pd(fscal,dy00);
658 tz = _mm256_mul_pd(fscal,dz00);
660 /* Update vectorial force */
661 fix0 = _mm256_add_pd(fix0,tx);
662 fiy0 = _mm256_add_pd(fiy0,ty);
663 fiz0 = _mm256_add_pd(fiz0,tz);
665 fjx0 = _mm256_add_pd(fjx0,tx);
666 fjy0 = _mm256_add_pd(fjy0,ty);
667 fjz0 = _mm256_add_pd(fjz0,tz);
669 /**************************
670 * CALCULATE INTERACTIONS *
671 **************************/
673 /* REACTION-FIELD ELECTROSTATICS */
674 velec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
675 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
677 /* Update potential sum for this i atom from the interaction with this j atom. */
678 velec = _mm256_andnot_pd(dummy_mask,velec);
679 velecsum = _mm256_add_pd(velecsum,velec);
683 fscal = _mm256_andnot_pd(dummy_mask,fscal);
685 /* Calculate temporary vectorial force */
686 tx = _mm256_mul_pd(fscal,dx01);
687 ty = _mm256_mul_pd(fscal,dy01);
688 tz = _mm256_mul_pd(fscal,dz01);
690 /* Update vectorial force */
691 fix0 = _mm256_add_pd(fix0,tx);
692 fiy0 = _mm256_add_pd(fiy0,ty);
693 fiz0 = _mm256_add_pd(fiz0,tz);
695 fjx1 = _mm256_add_pd(fjx1,tx);
696 fjy1 = _mm256_add_pd(fjy1,ty);
697 fjz1 = _mm256_add_pd(fjz1,tz);
699 /**************************
700 * CALCULATE INTERACTIONS *
701 **************************/
703 /* REACTION-FIELD ELECTROSTATICS */
704 velec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
705 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
707 /* Update potential sum for this i atom from the interaction with this j atom. */
708 velec = _mm256_andnot_pd(dummy_mask,velec);
709 velecsum = _mm256_add_pd(velecsum,velec);
713 fscal = _mm256_andnot_pd(dummy_mask,fscal);
715 /* Calculate temporary vectorial force */
716 tx = _mm256_mul_pd(fscal,dx02);
717 ty = _mm256_mul_pd(fscal,dy02);
718 tz = _mm256_mul_pd(fscal,dz02);
720 /* Update vectorial force */
721 fix0 = _mm256_add_pd(fix0,tx);
722 fiy0 = _mm256_add_pd(fiy0,ty);
723 fiz0 = _mm256_add_pd(fiz0,tz);
725 fjx2 = _mm256_add_pd(fjx2,tx);
726 fjy2 = _mm256_add_pd(fjy2,ty);
727 fjz2 = _mm256_add_pd(fjz2,tz);
729 /**************************
730 * CALCULATE INTERACTIONS *
731 **************************/
733 /* REACTION-FIELD ELECTROSTATICS */
734 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
735 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
737 /* Update potential sum for this i atom from the interaction with this j atom. */
738 velec = _mm256_andnot_pd(dummy_mask,velec);
739 velecsum = _mm256_add_pd(velecsum,velec);
743 fscal = _mm256_andnot_pd(dummy_mask,fscal);
745 /* Calculate temporary vectorial force */
746 tx = _mm256_mul_pd(fscal,dx10);
747 ty = _mm256_mul_pd(fscal,dy10);
748 tz = _mm256_mul_pd(fscal,dz10);
750 /* Update vectorial force */
751 fix1 = _mm256_add_pd(fix1,tx);
752 fiy1 = _mm256_add_pd(fiy1,ty);
753 fiz1 = _mm256_add_pd(fiz1,tz);
755 fjx0 = _mm256_add_pd(fjx0,tx);
756 fjy0 = _mm256_add_pd(fjy0,ty);
757 fjz0 = _mm256_add_pd(fjz0,tz);
759 /**************************
760 * CALCULATE INTERACTIONS *
761 **************************/
763 /* REACTION-FIELD ELECTROSTATICS */
764 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
765 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
767 /* Update potential sum for this i atom from the interaction with this j atom. */
768 velec = _mm256_andnot_pd(dummy_mask,velec);
769 velecsum = _mm256_add_pd(velecsum,velec);
773 fscal = _mm256_andnot_pd(dummy_mask,fscal);
775 /* Calculate temporary vectorial force */
776 tx = _mm256_mul_pd(fscal,dx11);
777 ty = _mm256_mul_pd(fscal,dy11);
778 tz = _mm256_mul_pd(fscal,dz11);
780 /* Update vectorial force */
781 fix1 = _mm256_add_pd(fix1,tx);
782 fiy1 = _mm256_add_pd(fiy1,ty);
783 fiz1 = _mm256_add_pd(fiz1,tz);
785 fjx1 = _mm256_add_pd(fjx1,tx);
786 fjy1 = _mm256_add_pd(fjy1,ty);
787 fjz1 = _mm256_add_pd(fjz1,tz);
789 /**************************
790 * CALCULATE INTERACTIONS *
791 **************************/
793 /* REACTION-FIELD ELECTROSTATICS */
794 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
795 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
797 /* Update potential sum for this i atom from the interaction with this j atom. */
798 velec = _mm256_andnot_pd(dummy_mask,velec);
799 velecsum = _mm256_add_pd(velecsum,velec);
803 fscal = _mm256_andnot_pd(dummy_mask,fscal);
805 /* Calculate temporary vectorial force */
806 tx = _mm256_mul_pd(fscal,dx12);
807 ty = _mm256_mul_pd(fscal,dy12);
808 tz = _mm256_mul_pd(fscal,dz12);
810 /* Update vectorial force */
811 fix1 = _mm256_add_pd(fix1,tx);
812 fiy1 = _mm256_add_pd(fiy1,ty);
813 fiz1 = _mm256_add_pd(fiz1,tz);
815 fjx2 = _mm256_add_pd(fjx2,tx);
816 fjy2 = _mm256_add_pd(fjy2,ty);
817 fjz2 = _mm256_add_pd(fjz2,tz);
819 /**************************
820 * CALCULATE INTERACTIONS *
821 **************************/
823 /* REACTION-FIELD ELECTROSTATICS */
824 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
825 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
827 /* Update potential sum for this i atom from the interaction with this j atom. */
828 velec = _mm256_andnot_pd(dummy_mask,velec);
829 velecsum = _mm256_add_pd(velecsum,velec);
833 fscal = _mm256_andnot_pd(dummy_mask,fscal);
835 /* Calculate temporary vectorial force */
836 tx = _mm256_mul_pd(fscal,dx20);
837 ty = _mm256_mul_pd(fscal,dy20);
838 tz = _mm256_mul_pd(fscal,dz20);
840 /* Update vectorial force */
841 fix2 = _mm256_add_pd(fix2,tx);
842 fiy2 = _mm256_add_pd(fiy2,ty);
843 fiz2 = _mm256_add_pd(fiz2,tz);
845 fjx0 = _mm256_add_pd(fjx0,tx);
846 fjy0 = _mm256_add_pd(fjy0,ty);
847 fjz0 = _mm256_add_pd(fjz0,tz);
849 /**************************
850 * CALCULATE INTERACTIONS *
851 **************************/
853 /* REACTION-FIELD ELECTROSTATICS */
854 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
855 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
857 /* Update potential sum for this i atom from the interaction with this j atom. */
858 velec = _mm256_andnot_pd(dummy_mask,velec);
859 velecsum = _mm256_add_pd(velecsum,velec);
863 fscal = _mm256_andnot_pd(dummy_mask,fscal);
865 /* Calculate temporary vectorial force */
866 tx = _mm256_mul_pd(fscal,dx21);
867 ty = _mm256_mul_pd(fscal,dy21);
868 tz = _mm256_mul_pd(fscal,dz21);
870 /* Update vectorial force */
871 fix2 = _mm256_add_pd(fix2,tx);
872 fiy2 = _mm256_add_pd(fiy2,ty);
873 fiz2 = _mm256_add_pd(fiz2,tz);
875 fjx1 = _mm256_add_pd(fjx1,tx);
876 fjy1 = _mm256_add_pd(fjy1,ty);
877 fjz1 = _mm256_add_pd(fjz1,tz);
879 /**************************
880 * CALCULATE INTERACTIONS *
881 **************************/
883 /* REACTION-FIELD ELECTROSTATICS */
884 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
885 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
887 /* Update potential sum for this i atom from the interaction with this j atom. */
888 velec = _mm256_andnot_pd(dummy_mask,velec);
889 velecsum = _mm256_add_pd(velecsum,velec);
893 fscal = _mm256_andnot_pd(dummy_mask,fscal);
895 /* Calculate temporary vectorial force */
896 tx = _mm256_mul_pd(fscal,dx22);
897 ty = _mm256_mul_pd(fscal,dy22);
898 tz = _mm256_mul_pd(fscal,dz22);
900 /* Update vectorial force */
901 fix2 = _mm256_add_pd(fix2,tx);
902 fiy2 = _mm256_add_pd(fiy2,ty);
903 fiz2 = _mm256_add_pd(fiz2,tz);
905 fjx2 = _mm256_add_pd(fjx2,tx);
906 fjy2 = _mm256_add_pd(fjy2,ty);
907 fjz2 = _mm256_add_pd(fjz2,tz);
909 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
910 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
911 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
912 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
914 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
915 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
917 /* Inner loop uses 288 flops */
920 /* End of innermost loop */
922 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
923 f+i_coord_offset,fshift+i_shift_offset);
926 /* Update potential energies */
927 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
929 /* Increment number of inner iterations */
930 inneriter += j_index_end - j_index_start;
932 /* Outer loop uses 19 flops */
935 /* Increment number of outer iterations */
938 /* Update outer/inner flops */
940 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*288);
943 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_double
944 * Electrostatics interaction: ReactionField
945 * VdW interaction: None
946 * Geometry: Water3-Water3
947 * Calculate force/pot: Force
950 nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_double
951 (t_nblist * gmx_restrict nlist,
952 rvec * gmx_restrict xx,
953 rvec * gmx_restrict ff,
954 t_forcerec * gmx_restrict fr,
955 t_mdatoms * gmx_restrict mdatoms,
956 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
957 t_nrnb * gmx_restrict nrnb)
959 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
960 * just 0 for non-waters.
961 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
962 * jnr indices corresponding to data put in the four positions in the SIMD register.
964 int i_shift_offset,i_coord_offset,outeriter,inneriter;
965 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
966 int jnrA,jnrB,jnrC,jnrD;
967 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
968 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
969 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
970 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
972 real *shiftvec,*fshift,*x,*f;
973 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
975 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
976 real * vdwioffsetptr0;
977 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
978 real * vdwioffsetptr1;
979 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
980 real * vdwioffsetptr2;
981 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
982 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
983 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
984 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
985 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
986 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
987 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
988 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
989 __m256d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
990 __m256d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
991 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
992 __m256d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
993 __m256d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
994 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
995 __m256d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
996 __m256d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
997 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
999 __m256d dummy_mask,cutoff_mask;
1000 __m128 tmpmask0,tmpmask1;
1001 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1002 __m256d one = _mm256_set1_pd(1.0);
1003 __m256d two = _mm256_set1_pd(2.0);
1009 jindex = nlist->jindex;
1011 shiftidx = nlist->shift;
1013 shiftvec = fr->shift_vec[0];
1014 fshift = fr->fshift[0];
1015 facel = _mm256_set1_pd(fr->epsfac);
1016 charge = mdatoms->chargeA;
1017 krf = _mm256_set1_pd(fr->ic->k_rf);
1018 krf2 = _mm256_set1_pd(fr->ic->k_rf*2.0);
1019 crf = _mm256_set1_pd(fr->ic->c_rf);
1021 /* Setup water-specific parameters */
1022 inr = nlist->iinr[0];
1023 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1024 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1025 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1027 jq0 = _mm256_set1_pd(charge[inr+0]);
1028 jq1 = _mm256_set1_pd(charge[inr+1]);
1029 jq2 = _mm256_set1_pd(charge[inr+2]);
1030 qq00 = _mm256_mul_pd(iq0,jq0);
1031 qq01 = _mm256_mul_pd(iq0,jq1);
1032 qq02 = _mm256_mul_pd(iq0,jq2);
1033 qq10 = _mm256_mul_pd(iq1,jq0);
1034 qq11 = _mm256_mul_pd(iq1,jq1);
1035 qq12 = _mm256_mul_pd(iq1,jq2);
1036 qq20 = _mm256_mul_pd(iq2,jq0);
1037 qq21 = _mm256_mul_pd(iq2,jq1);
1038 qq22 = _mm256_mul_pd(iq2,jq2);
1040 /* Avoid stupid compiler warnings */
1041 jnrA = jnrB = jnrC = jnrD = 0;
1042 j_coord_offsetA = 0;
1043 j_coord_offsetB = 0;
1044 j_coord_offsetC = 0;
1045 j_coord_offsetD = 0;
1050 for(iidx=0;iidx<4*DIM;iidx++)
1052 scratch[iidx] = 0.0;
1055 /* Start outer loop over neighborlists */
1056 for(iidx=0; iidx<nri; iidx++)
1058 /* Load shift vector for this list */
1059 i_shift_offset = DIM*shiftidx[iidx];
1061 /* Load limits for loop over neighbors */
1062 j_index_start = jindex[iidx];
1063 j_index_end = jindex[iidx+1];
1065 /* Get outer coordinate index */
1067 i_coord_offset = DIM*inr;
1069 /* Load i particle coords and add shift vector */
1070 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1071 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1073 fix0 = _mm256_setzero_pd();
1074 fiy0 = _mm256_setzero_pd();
1075 fiz0 = _mm256_setzero_pd();
1076 fix1 = _mm256_setzero_pd();
1077 fiy1 = _mm256_setzero_pd();
1078 fiz1 = _mm256_setzero_pd();
1079 fix2 = _mm256_setzero_pd();
1080 fiy2 = _mm256_setzero_pd();
1081 fiz2 = _mm256_setzero_pd();
1083 /* Start inner kernel loop */
1084 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1087 /* Get j neighbor index, and coordinate index */
1089 jnrB = jjnr[jidx+1];
1090 jnrC = jjnr[jidx+2];
1091 jnrD = jjnr[jidx+3];
1092 j_coord_offsetA = DIM*jnrA;
1093 j_coord_offsetB = DIM*jnrB;
1094 j_coord_offsetC = DIM*jnrC;
1095 j_coord_offsetD = DIM*jnrD;
1097 /* load j atom coordinates */
1098 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1099 x+j_coord_offsetC,x+j_coord_offsetD,
1100 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1102 /* Calculate displacement vector */
1103 dx00 = _mm256_sub_pd(ix0,jx0);
1104 dy00 = _mm256_sub_pd(iy0,jy0);
1105 dz00 = _mm256_sub_pd(iz0,jz0);
1106 dx01 = _mm256_sub_pd(ix0,jx1);
1107 dy01 = _mm256_sub_pd(iy0,jy1);
1108 dz01 = _mm256_sub_pd(iz0,jz1);
1109 dx02 = _mm256_sub_pd(ix0,jx2);
1110 dy02 = _mm256_sub_pd(iy0,jy2);
1111 dz02 = _mm256_sub_pd(iz0,jz2);
1112 dx10 = _mm256_sub_pd(ix1,jx0);
1113 dy10 = _mm256_sub_pd(iy1,jy0);
1114 dz10 = _mm256_sub_pd(iz1,jz0);
1115 dx11 = _mm256_sub_pd(ix1,jx1);
1116 dy11 = _mm256_sub_pd(iy1,jy1);
1117 dz11 = _mm256_sub_pd(iz1,jz1);
1118 dx12 = _mm256_sub_pd(ix1,jx2);
1119 dy12 = _mm256_sub_pd(iy1,jy2);
1120 dz12 = _mm256_sub_pd(iz1,jz2);
1121 dx20 = _mm256_sub_pd(ix2,jx0);
1122 dy20 = _mm256_sub_pd(iy2,jy0);
1123 dz20 = _mm256_sub_pd(iz2,jz0);
1124 dx21 = _mm256_sub_pd(ix2,jx1);
1125 dy21 = _mm256_sub_pd(iy2,jy1);
1126 dz21 = _mm256_sub_pd(iz2,jz1);
1127 dx22 = _mm256_sub_pd(ix2,jx2);
1128 dy22 = _mm256_sub_pd(iy2,jy2);
1129 dz22 = _mm256_sub_pd(iz2,jz2);
1131 /* Calculate squared distance and things based on it */
1132 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1133 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1134 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1135 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1136 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1137 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1138 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1139 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1140 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1142 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1143 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
1144 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
1145 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1146 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
1147 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
1148 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1149 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
1150 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
1152 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1153 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
1154 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
1155 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1156 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
1157 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
1158 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1159 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
1160 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
1162 fjx0 = _mm256_setzero_pd();
1163 fjy0 = _mm256_setzero_pd();
1164 fjz0 = _mm256_setzero_pd();
1165 fjx1 = _mm256_setzero_pd();
1166 fjy1 = _mm256_setzero_pd();
1167 fjz1 = _mm256_setzero_pd();
1168 fjx2 = _mm256_setzero_pd();
1169 fjy2 = _mm256_setzero_pd();
1170 fjz2 = _mm256_setzero_pd();
1172 /**************************
1173 * CALCULATE INTERACTIONS *
1174 **************************/
1176 /* REACTION-FIELD ELECTROSTATICS */
1177 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1181 /* Calculate temporary vectorial force */
1182 tx = _mm256_mul_pd(fscal,dx00);
1183 ty = _mm256_mul_pd(fscal,dy00);
1184 tz = _mm256_mul_pd(fscal,dz00);
1186 /* Update vectorial force */
1187 fix0 = _mm256_add_pd(fix0,tx);
1188 fiy0 = _mm256_add_pd(fiy0,ty);
1189 fiz0 = _mm256_add_pd(fiz0,tz);
1191 fjx0 = _mm256_add_pd(fjx0,tx);
1192 fjy0 = _mm256_add_pd(fjy0,ty);
1193 fjz0 = _mm256_add_pd(fjz0,tz);
1195 /**************************
1196 * CALCULATE INTERACTIONS *
1197 **************************/
1199 /* REACTION-FIELD ELECTROSTATICS */
1200 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1204 /* Calculate temporary vectorial force */
1205 tx = _mm256_mul_pd(fscal,dx01);
1206 ty = _mm256_mul_pd(fscal,dy01);
1207 tz = _mm256_mul_pd(fscal,dz01);
1209 /* Update vectorial force */
1210 fix0 = _mm256_add_pd(fix0,tx);
1211 fiy0 = _mm256_add_pd(fiy0,ty);
1212 fiz0 = _mm256_add_pd(fiz0,tz);
1214 fjx1 = _mm256_add_pd(fjx1,tx);
1215 fjy1 = _mm256_add_pd(fjy1,ty);
1216 fjz1 = _mm256_add_pd(fjz1,tz);
1218 /**************************
1219 * CALCULATE INTERACTIONS *
1220 **************************/
1222 /* REACTION-FIELD ELECTROSTATICS */
1223 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1227 /* Calculate temporary vectorial force */
1228 tx = _mm256_mul_pd(fscal,dx02);
1229 ty = _mm256_mul_pd(fscal,dy02);
1230 tz = _mm256_mul_pd(fscal,dz02);
1232 /* Update vectorial force */
1233 fix0 = _mm256_add_pd(fix0,tx);
1234 fiy0 = _mm256_add_pd(fiy0,ty);
1235 fiz0 = _mm256_add_pd(fiz0,tz);
1237 fjx2 = _mm256_add_pd(fjx2,tx);
1238 fjy2 = _mm256_add_pd(fjy2,ty);
1239 fjz2 = _mm256_add_pd(fjz2,tz);
1241 /**************************
1242 * CALCULATE INTERACTIONS *
1243 **************************/
1245 /* REACTION-FIELD ELECTROSTATICS */
1246 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1250 /* Calculate temporary vectorial force */
1251 tx = _mm256_mul_pd(fscal,dx10);
1252 ty = _mm256_mul_pd(fscal,dy10);
1253 tz = _mm256_mul_pd(fscal,dz10);
1255 /* Update vectorial force */
1256 fix1 = _mm256_add_pd(fix1,tx);
1257 fiy1 = _mm256_add_pd(fiy1,ty);
1258 fiz1 = _mm256_add_pd(fiz1,tz);
1260 fjx0 = _mm256_add_pd(fjx0,tx);
1261 fjy0 = _mm256_add_pd(fjy0,ty);
1262 fjz0 = _mm256_add_pd(fjz0,tz);
1264 /**************************
1265 * CALCULATE INTERACTIONS *
1266 **************************/
1268 /* REACTION-FIELD ELECTROSTATICS */
1269 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1273 /* Calculate temporary vectorial force */
1274 tx = _mm256_mul_pd(fscal,dx11);
1275 ty = _mm256_mul_pd(fscal,dy11);
1276 tz = _mm256_mul_pd(fscal,dz11);
1278 /* Update vectorial force */
1279 fix1 = _mm256_add_pd(fix1,tx);
1280 fiy1 = _mm256_add_pd(fiy1,ty);
1281 fiz1 = _mm256_add_pd(fiz1,tz);
1283 fjx1 = _mm256_add_pd(fjx1,tx);
1284 fjy1 = _mm256_add_pd(fjy1,ty);
1285 fjz1 = _mm256_add_pd(fjz1,tz);
1287 /**************************
1288 * CALCULATE INTERACTIONS *
1289 **************************/
1291 /* REACTION-FIELD ELECTROSTATICS */
1292 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1296 /* Calculate temporary vectorial force */
1297 tx = _mm256_mul_pd(fscal,dx12);
1298 ty = _mm256_mul_pd(fscal,dy12);
1299 tz = _mm256_mul_pd(fscal,dz12);
1301 /* Update vectorial force */
1302 fix1 = _mm256_add_pd(fix1,tx);
1303 fiy1 = _mm256_add_pd(fiy1,ty);
1304 fiz1 = _mm256_add_pd(fiz1,tz);
1306 fjx2 = _mm256_add_pd(fjx2,tx);
1307 fjy2 = _mm256_add_pd(fjy2,ty);
1308 fjz2 = _mm256_add_pd(fjz2,tz);
1310 /**************************
1311 * CALCULATE INTERACTIONS *
1312 **************************/
1314 /* REACTION-FIELD ELECTROSTATICS */
1315 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1319 /* Calculate temporary vectorial force */
1320 tx = _mm256_mul_pd(fscal,dx20);
1321 ty = _mm256_mul_pd(fscal,dy20);
1322 tz = _mm256_mul_pd(fscal,dz20);
1324 /* Update vectorial force */
1325 fix2 = _mm256_add_pd(fix2,tx);
1326 fiy2 = _mm256_add_pd(fiy2,ty);
1327 fiz2 = _mm256_add_pd(fiz2,tz);
1329 fjx0 = _mm256_add_pd(fjx0,tx);
1330 fjy0 = _mm256_add_pd(fjy0,ty);
1331 fjz0 = _mm256_add_pd(fjz0,tz);
1333 /**************************
1334 * CALCULATE INTERACTIONS *
1335 **************************/
1337 /* REACTION-FIELD ELECTROSTATICS */
1338 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1342 /* Calculate temporary vectorial force */
1343 tx = _mm256_mul_pd(fscal,dx21);
1344 ty = _mm256_mul_pd(fscal,dy21);
1345 tz = _mm256_mul_pd(fscal,dz21);
1347 /* Update vectorial force */
1348 fix2 = _mm256_add_pd(fix2,tx);
1349 fiy2 = _mm256_add_pd(fiy2,ty);
1350 fiz2 = _mm256_add_pd(fiz2,tz);
1352 fjx1 = _mm256_add_pd(fjx1,tx);
1353 fjy1 = _mm256_add_pd(fjy1,ty);
1354 fjz1 = _mm256_add_pd(fjz1,tz);
1356 /**************************
1357 * CALCULATE INTERACTIONS *
1358 **************************/
1360 /* REACTION-FIELD ELECTROSTATICS */
1361 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1365 /* Calculate temporary vectorial force */
1366 tx = _mm256_mul_pd(fscal,dx22);
1367 ty = _mm256_mul_pd(fscal,dy22);
1368 tz = _mm256_mul_pd(fscal,dz22);
1370 /* Update vectorial force */
1371 fix2 = _mm256_add_pd(fix2,tx);
1372 fiy2 = _mm256_add_pd(fiy2,ty);
1373 fiz2 = _mm256_add_pd(fiz2,tz);
1375 fjx2 = _mm256_add_pd(fjx2,tx);
1376 fjy2 = _mm256_add_pd(fjy2,ty);
1377 fjz2 = _mm256_add_pd(fjz2,tz);
1379 fjptrA = f+j_coord_offsetA;
1380 fjptrB = f+j_coord_offsetB;
1381 fjptrC = f+j_coord_offsetC;
1382 fjptrD = f+j_coord_offsetD;
1384 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1385 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1387 /* Inner loop uses 243 flops */
1390 if(jidx<j_index_end)
1393 /* Get j neighbor index, and coordinate index */
1394 jnrlistA = jjnr[jidx];
1395 jnrlistB = jjnr[jidx+1];
1396 jnrlistC = jjnr[jidx+2];
1397 jnrlistD = jjnr[jidx+3];
1398 /* Sign of each element will be negative for non-real atoms.
1399 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1400 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1402 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1404 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1405 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1406 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1408 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1409 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1410 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1411 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1412 j_coord_offsetA = DIM*jnrA;
1413 j_coord_offsetB = DIM*jnrB;
1414 j_coord_offsetC = DIM*jnrC;
1415 j_coord_offsetD = DIM*jnrD;
1417 /* load j atom coordinates */
1418 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1419 x+j_coord_offsetC,x+j_coord_offsetD,
1420 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1422 /* Calculate displacement vector */
1423 dx00 = _mm256_sub_pd(ix0,jx0);
1424 dy00 = _mm256_sub_pd(iy0,jy0);
1425 dz00 = _mm256_sub_pd(iz0,jz0);
1426 dx01 = _mm256_sub_pd(ix0,jx1);
1427 dy01 = _mm256_sub_pd(iy0,jy1);
1428 dz01 = _mm256_sub_pd(iz0,jz1);
1429 dx02 = _mm256_sub_pd(ix0,jx2);
1430 dy02 = _mm256_sub_pd(iy0,jy2);
1431 dz02 = _mm256_sub_pd(iz0,jz2);
1432 dx10 = _mm256_sub_pd(ix1,jx0);
1433 dy10 = _mm256_sub_pd(iy1,jy0);
1434 dz10 = _mm256_sub_pd(iz1,jz0);
1435 dx11 = _mm256_sub_pd(ix1,jx1);
1436 dy11 = _mm256_sub_pd(iy1,jy1);
1437 dz11 = _mm256_sub_pd(iz1,jz1);
1438 dx12 = _mm256_sub_pd(ix1,jx2);
1439 dy12 = _mm256_sub_pd(iy1,jy2);
1440 dz12 = _mm256_sub_pd(iz1,jz2);
1441 dx20 = _mm256_sub_pd(ix2,jx0);
1442 dy20 = _mm256_sub_pd(iy2,jy0);
1443 dz20 = _mm256_sub_pd(iz2,jz0);
1444 dx21 = _mm256_sub_pd(ix2,jx1);
1445 dy21 = _mm256_sub_pd(iy2,jy1);
1446 dz21 = _mm256_sub_pd(iz2,jz1);
1447 dx22 = _mm256_sub_pd(ix2,jx2);
1448 dy22 = _mm256_sub_pd(iy2,jy2);
1449 dz22 = _mm256_sub_pd(iz2,jz2);
1451 /* Calculate squared distance and things based on it */
1452 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1453 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1454 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1455 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1456 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1457 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1458 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1459 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1460 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1462 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1463 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
1464 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
1465 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1466 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
1467 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
1468 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1469 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
1470 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
1472 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1473 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
1474 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
1475 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1476 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
1477 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
1478 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1479 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
1480 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
1482 fjx0 = _mm256_setzero_pd();
1483 fjy0 = _mm256_setzero_pd();
1484 fjz0 = _mm256_setzero_pd();
1485 fjx1 = _mm256_setzero_pd();
1486 fjy1 = _mm256_setzero_pd();
1487 fjz1 = _mm256_setzero_pd();
1488 fjx2 = _mm256_setzero_pd();
1489 fjy2 = _mm256_setzero_pd();
1490 fjz2 = _mm256_setzero_pd();
1492 /**************************
1493 * CALCULATE INTERACTIONS *
1494 **************************/
1496 /* REACTION-FIELD ELECTROSTATICS */
1497 felec = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1501 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1503 /* Calculate temporary vectorial force */
1504 tx = _mm256_mul_pd(fscal,dx00);
1505 ty = _mm256_mul_pd(fscal,dy00);
1506 tz = _mm256_mul_pd(fscal,dz00);
1508 /* Update vectorial force */
1509 fix0 = _mm256_add_pd(fix0,tx);
1510 fiy0 = _mm256_add_pd(fiy0,ty);
1511 fiz0 = _mm256_add_pd(fiz0,tz);
1513 fjx0 = _mm256_add_pd(fjx0,tx);
1514 fjy0 = _mm256_add_pd(fjy0,ty);
1515 fjz0 = _mm256_add_pd(fjz0,tz);
1517 /**************************
1518 * CALCULATE INTERACTIONS *
1519 **************************/
1521 /* REACTION-FIELD ELECTROSTATICS */
1522 felec = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1526 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1528 /* Calculate temporary vectorial force */
1529 tx = _mm256_mul_pd(fscal,dx01);
1530 ty = _mm256_mul_pd(fscal,dy01);
1531 tz = _mm256_mul_pd(fscal,dz01);
1533 /* Update vectorial force */
1534 fix0 = _mm256_add_pd(fix0,tx);
1535 fiy0 = _mm256_add_pd(fiy0,ty);
1536 fiz0 = _mm256_add_pd(fiz0,tz);
1538 fjx1 = _mm256_add_pd(fjx1,tx);
1539 fjy1 = _mm256_add_pd(fjy1,ty);
1540 fjz1 = _mm256_add_pd(fjz1,tz);
1542 /**************************
1543 * CALCULATE INTERACTIONS *
1544 **************************/
1546 /* REACTION-FIELD ELECTROSTATICS */
1547 felec = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1551 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1553 /* Calculate temporary vectorial force */
1554 tx = _mm256_mul_pd(fscal,dx02);
1555 ty = _mm256_mul_pd(fscal,dy02);
1556 tz = _mm256_mul_pd(fscal,dz02);
1558 /* Update vectorial force */
1559 fix0 = _mm256_add_pd(fix0,tx);
1560 fiy0 = _mm256_add_pd(fiy0,ty);
1561 fiz0 = _mm256_add_pd(fiz0,tz);
1563 fjx2 = _mm256_add_pd(fjx2,tx);
1564 fjy2 = _mm256_add_pd(fjy2,ty);
1565 fjz2 = _mm256_add_pd(fjz2,tz);
1567 /**************************
1568 * CALCULATE INTERACTIONS *
1569 **************************/
1571 /* REACTION-FIELD ELECTROSTATICS */
1572 felec = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1576 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1578 /* Calculate temporary vectorial force */
1579 tx = _mm256_mul_pd(fscal,dx10);
1580 ty = _mm256_mul_pd(fscal,dy10);
1581 tz = _mm256_mul_pd(fscal,dz10);
1583 /* Update vectorial force */
1584 fix1 = _mm256_add_pd(fix1,tx);
1585 fiy1 = _mm256_add_pd(fiy1,ty);
1586 fiz1 = _mm256_add_pd(fiz1,tz);
1588 fjx0 = _mm256_add_pd(fjx0,tx);
1589 fjy0 = _mm256_add_pd(fjy0,ty);
1590 fjz0 = _mm256_add_pd(fjz0,tz);
1592 /**************************
1593 * CALCULATE INTERACTIONS *
1594 **************************/
1596 /* REACTION-FIELD ELECTROSTATICS */
1597 felec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1601 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1603 /* Calculate temporary vectorial force */
1604 tx = _mm256_mul_pd(fscal,dx11);
1605 ty = _mm256_mul_pd(fscal,dy11);
1606 tz = _mm256_mul_pd(fscal,dz11);
1608 /* Update vectorial force */
1609 fix1 = _mm256_add_pd(fix1,tx);
1610 fiy1 = _mm256_add_pd(fiy1,ty);
1611 fiz1 = _mm256_add_pd(fiz1,tz);
1613 fjx1 = _mm256_add_pd(fjx1,tx);
1614 fjy1 = _mm256_add_pd(fjy1,ty);
1615 fjz1 = _mm256_add_pd(fjz1,tz);
1617 /**************************
1618 * CALCULATE INTERACTIONS *
1619 **************************/
1621 /* REACTION-FIELD ELECTROSTATICS */
1622 felec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1626 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1628 /* Calculate temporary vectorial force */
1629 tx = _mm256_mul_pd(fscal,dx12);
1630 ty = _mm256_mul_pd(fscal,dy12);
1631 tz = _mm256_mul_pd(fscal,dz12);
1633 /* Update vectorial force */
1634 fix1 = _mm256_add_pd(fix1,tx);
1635 fiy1 = _mm256_add_pd(fiy1,ty);
1636 fiz1 = _mm256_add_pd(fiz1,tz);
1638 fjx2 = _mm256_add_pd(fjx2,tx);
1639 fjy2 = _mm256_add_pd(fjy2,ty);
1640 fjz2 = _mm256_add_pd(fjz2,tz);
1642 /**************************
1643 * CALCULATE INTERACTIONS *
1644 **************************/
1646 /* REACTION-FIELD ELECTROSTATICS */
1647 felec = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1651 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1653 /* Calculate temporary vectorial force */
1654 tx = _mm256_mul_pd(fscal,dx20);
1655 ty = _mm256_mul_pd(fscal,dy20);
1656 tz = _mm256_mul_pd(fscal,dz20);
1658 /* Update vectorial force */
1659 fix2 = _mm256_add_pd(fix2,tx);
1660 fiy2 = _mm256_add_pd(fiy2,ty);
1661 fiz2 = _mm256_add_pd(fiz2,tz);
1663 fjx0 = _mm256_add_pd(fjx0,tx);
1664 fjy0 = _mm256_add_pd(fjy0,ty);
1665 fjz0 = _mm256_add_pd(fjz0,tz);
1667 /**************************
1668 * CALCULATE INTERACTIONS *
1669 **************************/
1671 /* REACTION-FIELD ELECTROSTATICS */
1672 felec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1676 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1678 /* Calculate temporary vectorial force */
1679 tx = _mm256_mul_pd(fscal,dx21);
1680 ty = _mm256_mul_pd(fscal,dy21);
1681 tz = _mm256_mul_pd(fscal,dz21);
1683 /* Update vectorial force */
1684 fix2 = _mm256_add_pd(fix2,tx);
1685 fiy2 = _mm256_add_pd(fiy2,ty);
1686 fiz2 = _mm256_add_pd(fiz2,tz);
1688 fjx1 = _mm256_add_pd(fjx1,tx);
1689 fjy1 = _mm256_add_pd(fjy1,ty);
1690 fjz1 = _mm256_add_pd(fjz1,tz);
1692 /**************************
1693 * CALCULATE INTERACTIONS *
1694 **************************/
1696 /* REACTION-FIELD ELECTROSTATICS */
1697 felec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1701 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1703 /* Calculate temporary vectorial force */
1704 tx = _mm256_mul_pd(fscal,dx22);
1705 ty = _mm256_mul_pd(fscal,dy22);
1706 tz = _mm256_mul_pd(fscal,dz22);
1708 /* Update vectorial force */
1709 fix2 = _mm256_add_pd(fix2,tx);
1710 fiy2 = _mm256_add_pd(fiy2,ty);
1711 fiz2 = _mm256_add_pd(fiz2,tz);
1713 fjx2 = _mm256_add_pd(fjx2,tx);
1714 fjy2 = _mm256_add_pd(fjy2,ty);
1715 fjz2 = _mm256_add_pd(fjz2,tz);
1717 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1718 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1719 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1720 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1722 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1723 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1725 /* Inner loop uses 243 flops */
1728 /* End of innermost loop */
1730 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1731 f+i_coord_offset,fshift+i_shift_offset);
1733 /* Increment number of inner iterations */
1734 inneriter += j_index_end - j_index_start;
1736 /* Outer loop uses 18 flops */
1739 /* Increment number of outer iterations */
1742 /* Update outer/inner flops */
1744 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*243);