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_single 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_single.h"
48 #include "kernelutil_x86_avx_256_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_avx_256_single
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_single
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,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight 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 jnrE,jnrF,jnrG,jnrH;
76 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
77 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
78 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
79 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
80 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
82 real *shiftvec,*fshift,*x,*f;
83 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
85 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 real * vdwioffsetptr0;
87 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
88 real * vdwioffsetptr1;
89 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
90 real * vdwioffsetptr2;
91 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
92 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
93 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
94 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
95 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
96 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
97 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
98 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
99 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
100 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
101 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
102 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
103 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
104 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
105 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
106 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
107 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
109 __m256 dummy_mask,cutoff_mask;
110 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
111 __m256 one = _mm256_set1_ps(1.0);
112 __m256 two = _mm256_set1_ps(2.0);
118 jindex = nlist->jindex;
120 shiftidx = nlist->shift;
122 shiftvec = fr->shift_vec[0];
123 fshift = fr->fshift[0];
124 facel = _mm256_set1_ps(fr->epsfac);
125 charge = mdatoms->chargeA;
126 krf = _mm256_set1_ps(fr->ic->k_rf);
127 krf2 = _mm256_set1_ps(fr->ic->k_rf*2.0);
128 crf = _mm256_set1_ps(fr->ic->c_rf);
130 /* Setup water-specific parameters */
131 inr = nlist->iinr[0];
132 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
133 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
134 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
136 jq0 = _mm256_set1_ps(charge[inr+0]);
137 jq1 = _mm256_set1_ps(charge[inr+1]);
138 jq2 = _mm256_set1_ps(charge[inr+2]);
139 qq00 = _mm256_mul_ps(iq0,jq0);
140 qq01 = _mm256_mul_ps(iq0,jq1);
141 qq02 = _mm256_mul_ps(iq0,jq2);
142 qq10 = _mm256_mul_ps(iq1,jq0);
143 qq11 = _mm256_mul_ps(iq1,jq1);
144 qq12 = _mm256_mul_ps(iq1,jq2);
145 qq20 = _mm256_mul_ps(iq2,jq0);
146 qq21 = _mm256_mul_ps(iq2,jq1);
147 qq22 = _mm256_mul_ps(iq2,jq2);
149 /* Avoid stupid compiler warnings */
150 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
163 for(iidx=0;iidx<4*DIM;iidx++)
168 /* Start outer loop over neighborlists */
169 for(iidx=0; iidx<nri; iidx++)
171 /* Load shift vector for this list */
172 i_shift_offset = DIM*shiftidx[iidx];
174 /* Load limits for loop over neighbors */
175 j_index_start = jindex[iidx];
176 j_index_end = jindex[iidx+1];
178 /* Get outer coordinate index */
180 i_coord_offset = DIM*inr;
182 /* Load i particle coords and add shift vector */
183 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
184 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
186 fix0 = _mm256_setzero_ps();
187 fiy0 = _mm256_setzero_ps();
188 fiz0 = _mm256_setzero_ps();
189 fix1 = _mm256_setzero_ps();
190 fiy1 = _mm256_setzero_ps();
191 fiz1 = _mm256_setzero_ps();
192 fix2 = _mm256_setzero_ps();
193 fiy2 = _mm256_setzero_ps();
194 fiz2 = _mm256_setzero_ps();
196 /* Reset potential sums */
197 velecsum = _mm256_setzero_ps();
199 /* Start inner kernel loop */
200 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
203 /* Get j neighbor index, and coordinate index */
212 j_coord_offsetA = DIM*jnrA;
213 j_coord_offsetB = DIM*jnrB;
214 j_coord_offsetC = DIM*jnrC;
215 j_coord_offsetD = DIM*jnrD;
216 j_coord_offsetE = DIM*jnrE;
217 j_coord_offsetF = DIM*jnrF;
218 j_coord_offsetG = DIM*jnrG;
219 j_coord_offsetH = DIM*jnrH;
221 /* load j atom coordinates */
222 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
223 x+j_coord_offsetC,x+j_coord_offsetD,
224 x+j_coord_offsetE,x+j_coord_offsetF,
225 x+j_coord_offsetG,x+j_coord_offsetH,
226 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
228 /* Calculate displacement vector */
229 dx00 = _mm256_sub_ps(ix0,jx0);
230 dy00 = _mm256_sub_ps(iy0,jy0);
231 dz00 = _mm256_sub_ps(iz0,jz0);
232 dx01 = _mm256_sub_ps(ix0,jx1);
233 dy01 = _mm256_sub_ps(iy0,jy1);
234 dz01 = _mm256_sub_ps(iz0,jz1);
235 dx02 = _mm256_sub_ps(ix0,jx2);
236 dy02 = _mm256_sub_ps(iy0,jy2);
237 dz02 = _mm256_sub_ps(iz0,jz2);
238 dx10 = _mm256_sub_ps(ix1,jx0);
239 dy10 = _mm256_sub_ps(iy1,jy0);
240 dz10 = _mm256_sub_ps(iz1,jz0);
241 dx11 = _mm256_sub_ps(ix1,jx1);
242 dy11 = _mm256_sub_ps(iy1,jy1);
243 dz11 = _mm256_sub_ps(iz1,jz1);
244 dx12 = _mm256_sub_ps(ix1,jx2);
245 dy12 = _mm256_sub_ps(iy1,jy2);
246 dz12 = _mm256_sub_ps(iz1,jz2);
247 dx20 = _mm256_sub_ps(ix2,jx0);
248 dy20 = _mm256_sub_ps(iy2,jy0);
249 dz20 = _mm256_sub_ps(iz2,jz0);
250 dx21 = _mm256_sub_ps(ix2,jx1);
251 dy21 = _mm256_sub_ps(iy2,jy1);
252 dz21 = _mm256_sub_ps(iz2,jz1);
253 dx22 = _mm256_sub_ps(ix2,jx2);
254 dy22 = _mm256_sub_ps(iy2,jy2);
255 dz22 = _mm256_sub_ps(iz2,jz2);
257 /* Calculate squared distance and things based on it */
258 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
259 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
260 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
261 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
262 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
263 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
264 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
265 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
266 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
268 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
269 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
270 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
271 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
272 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
273 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
274 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
275 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
276 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
278 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
279 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
280 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
281 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
282 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
283 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
284 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
285 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
286 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
288 fjx0 = _mm256_setzero_ps();
289 fjy0 = _mm256_setzero_ps();
290 fjz0 = _mm256_setzero_ps();
291 fjx1 = _mm256_setzero_ps();
292 fjy1 = _mm256_setzero_ps();
293 fjz1 = _mm256_setzero_ps();
294 fjx2 = _mm256_setzero_ps();
295 fjy2 = _mm256_setzero_ps();
296 fjz2 = _mm256_setzero_ps();
298 /**************************
299 * CALCULATE INTERACTIONS *
300 **************************/
302 /* REACTION-FIELD ELECTROSTATICS */
303 velec = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
304 felec = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
306 /* Update potential sum for this i atom from the interaction with this j atom. */
307 velecsum = _mm256_add_ps(velecsum,velec);
311 /* Calculate temporary vectorial force */
312 tx = _mm256_mul_ps(fscal,dx00);
313 ty = _mm256_mul_ps(fscal,dy00);
314 tz = _mm256_mul_ps(fscal,dz00);
316 /* Update vectorial force */
317 fix0 = _mm256_add_ps(fix0,tx);
318 fiy0 = _mm256_add_ps(fiy0,ty);
319 fiz0 = _mm256_add_ps(fiz0,tz);
321 fjx0 = _mm256_add_ps(fjx0,tx);
322 fjy0 = _mm256_add_ps(fjy0,ty);
323 fjz0 = _mm256_add_ps(fjz0,tz);
325 /**************************
326 * CALCULATE INTERACTIONS *
327 **************************/
329 /* REACTION-FIELD ELECTROSTATICS */
330 velec = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
331 felec = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
333 /* Update potential sum for this i atom from the interaction with this j atom. */
334 velecsum = _mm256_add_ps(velecsum,velec);
338 /* Calculate temporary vectorial force */
339 tx = _mm256_mul_ps(fscal,dx01);
340 ty = _mm256_mul_ps(fscal,dy01);
341 tz = _mm256_mul_ps(fscal,dz01);
343 /* Update vectorial force */
344 fix0 = _mm256_add_ps(fix0,tx);
345 fiy0 = _mm256_add_ps(fiy0,ty);
346 fiz0 = _mm256_add_ps(fiz0,tz);
348 fjx1 = _mm256_add_ps(fjx1,tx);
349 fjy1 = _mm256_add_ps(fjy1,ty);
350 fjz1 = _mm256_add_ps(fjz1,tz);
352 /**************************
353 * CALCULATE INTERACTIONS *
354 **************************/
356 /* REACTION-FIELD ELECTROSTATICS */
357 velec = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
358 felec = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
360 /* Update potential sum for this i atom from the interaction with this j atom. */
361 velecsum = _mm256_add_ps(velecsum,velec);
365 /* Calculate temporary vectorial force */
366 tx = _mm256_mul_ps(fscal,dx02);
367 ty = _mm256_mul_ps(fscal,dy02);
368 tz = _mm256_mul_ps(fscal,dz02);
370 /* Update vectorial force */
371 fix0 = _mm256_add_ps(fix0,tx);
372 fiy0 = _mm256_add_ps(fiy0,ty);
373 fiz0 = _mm256_add_ps(fiz0,tz);
375 fjx2 = _mm256_add_ps(fjx2,tx);
376 fjy2 = _mm256_add_ps(fjy2,ty);
377 fjz2 = _mm256_add_ps(fjz2,tz);
379 /**************************
380 * CALCULATE INTERACTIONS *
381 **************************/
383 /* REACTION-FIELD ELECTROSTATICS */
384 velec = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
385 felec = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
387 /* Update potential sum for this i atom from the interaction with this j atom. */
388 velecsum = _mm256_add_ps(velecsum,velec);
392 /* Calculate temporary vectorial force */
393 tx = _mm256_mul_ps(fscal,dx10);
394 ty = _mm256_mul_ps(fscal,dy10);
395 tz = _mm256_mul_ps(fscal,dz10);
397 /* Update vectorial force */
398 fix1 = _mm256_add_ps(fix1,tx);
399 fiy1 = _mm256_add_ps(fiy1,ty);
400 fiz1 = _mm256_add_ps(fiz1,tz);
402 fjx0 = _mm256_add_ps(fjx0,tx);
403 fjy0 = _mm256_add_ps(fjy0,ty);
404 fjz0 = _mm256_add_ps(fjz0,tz);
406 /**************************
407 * CALCULATE INTERACTIONS *
408 **************************/
410 /* REACTION-FIELD ELECTROSTATICS */
411 velec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
412 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
414 /* Update potential sum for this i atom from the interaction with this j atom. */
415 velecsum = _mm256_add_ps(velecsum,velec);
419 /* Calculate temporary vectorial force */
420 tx = _mm256_mul_ps(fscal,dx11);
421 ty = _mm256_mul_ps(fscal,dy11);
422 tz = _mm256_mul_ps(fscal,dz11);
424 /* Update vectorial force */
425 fix1 = _mm256_add_ps(fix1,tx);
426 fiy1 = _mm256_add_ps(fiy1,ty);
427 fiz1 = _mm256_add_ps(fiz1,tz);
429 fjx1 = _mm256_add_ps(fjx1,tx);
430 fjy1 = _mm256_add_ps(fjy1,ty);
431 fjz1 = _mm256_add_ps(fjz1,tz);
433 /**************************
434 * CALCULATE INTERACTIONS *
435 **************************/
437 /* REACTION-FIELD ELECTROSTATICS */
438 velec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
439 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
441 /* Update potential sum for this i atom from the interaction with this j atom. */
442 velecsum = _mm256_add_ps(velecsum,velec);
446 /* Calculate temporary vectorial force */
447 tx = _mm256_mul_ps(fscal,dx12);
448 ty = _mm256_mul_ps(fscal,dy12);
449 tz = _mm256_mul_ps(fscal,dz12);
451 /* Update vectorial force */
452 fix1 = _mm256_add_ps(fix1,tx);
453 fiy1 = _mm256_add_ps(fiy1,ty);
454 fiz1 = _mm256_add_ps(fiz1,tz);
456 fjx2 = _mm256_add_ps(fjx2,tx);
457 fjy2 = _mm256_add_ps(fjy2,ty);
458 fjz2 = _mm256_add_ps(fjz2,tz);
460 /**************************
461 * CALCULATE INTERACTIONS *
462 **************************/
464 /* REACTION-FIELD ELECTROSTATICS */
465 velec = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
466 felec = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
468 /* Update potential sum for this i atom from the interaction with this j atom. */
469 velecsum = _mm256_add_ps(velecsum,velec);
473 /* Calculate temporary vectorial force */
474 tx = _mm256_mul_ps(fscal,dx20);
475 ty = _mm256_mul_ps(fscal,dy20);
476 tz = _mm256_mul_ps(fscal,dz20);
478 /* Update vectorial force */
479 fix2 = _mm256_add_ps(fix2,tx);
480 fiy2 = _mm256_add_ps(fiy2,ty);
481 fiz2 = _mm256_add_ps(fiz2,tz);
483 fjx0 = _mm256_add_ps(fjx0,tx);
484 fjy0 = _mm256_add_ps(fjy0,ty);
485 fjz0 = _mm256_add_ps(fjz0,tz);
487 /**************************
488 * CALCULATE INTERACTIONS *
489 **************************/
491 /* REACTION-FIELD ELECTROSTATICS */
492 velec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
493 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
495 /* Update potential sum for this i atom from the interaction with this j atom. */
496 velecsum = _mm256_add_ps(velecsum,velec);
500 /* Calculate temporary vectorial force */
501 tx = _mm256_mul_ps(fscal,dx21);
502 ty = _mm256_mul_ps(fscal,dy21);
503 tz = _mm256_mul_ps(fscal,dz21);
505 /* Update vectorial force */
506 fix2 = _mm256_add_ps(fix2,tx);
507 fiy2 = _mm256_add_ps(fiy2,ty);
508 fiz2 = _mm256_add_ps(fiz2,tz);
510 fjx1 = _mm256_add_ps(fjx1,tx);
511 fjy1 = _mm256_add_ps(fjy1,ty);
512 fjz1 = _mm256_add_ps(fjz1,tz);
514 /**************************
515 * CALCULATE INTERACTIONS *
516 **************************/
518 /* REACTION-FIELD ELECTROSTATICS */
519 velec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
520 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
522 /* Update potential sum for this i atom from the interaction with this j atom. */
523 velecsum = _mm256_add_ps(velecsum,velec);
527 /* Calculate temporary vectorial force */
528 tx = _mm256_mul_ps(fscal,dx22);
529 ty = _mm256_mul_ps(fscal,dy22);
530 tz = _mm256_mul_ps(fscal,dz22);
532 /* Update vectorial force */
533 fix2 = _mm256_add_ps(fix2,tx);
534 fiy2 = _mm256_add_ps(fiy2,ty);
535 fiz2 = _mm256_add_ps(fiz2,tz);
537 fjx2 = _mm256_add_ps(fjx2,tx);
538 fjy2 = _mm256_add_ps(fjy2,ty);
539 fjz2 = _mm256_add_ps(fjz2,tz);
541 fjptrA = f+j_coord_offsetA;
542 fjptrB = f+j_coord_offsetB;
543 fjptrC = f+j_coord_offsetC;
544 fjptrD = f+j_coord_offsetD;
545 fjptrE = f+j_coord_offsetE;
546 fjptrF = f+j_coord_offsetF;
547 fjptrG = f+j_coord_offsetG;
548 fjptrH = f+j_coord_offsetH;
550 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
551 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
553 /* Inner loop uses 288 flops */
559 /* Get j neighbor index, and coordinate index */
560 jnrlistA = jjnr[jidx];
561 jnrlistB = jjnr[jidx+1];
562 jnrlistC = jjnr[jidx+2];
563 jnrlistD = jjnr[jidx+3];
564 jnrlistE = jjnr[jidx+4];
565 jnrlistF = jjnr[jidx+5];
566 jnrlistG = jjnr[jidx+6];
567 jnrlistH = jjnr[jidx+7];
568 /* Sign of each element will be negative for non-real atoms.
569 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
570 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
572 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
573 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
575 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
576 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
577 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
578 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
579 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
580 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
581 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
582 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
583 j_coord_offsetA = DIM*jnrA;
584 j_coord_offsetB = DIM*jnrB;
585 j_coord_offsetC = DIM*jnrC;
586 j_coord_offsetD = DIM*jnrD;
587 j_coord_offsetE = DIM*jnrE;
588 j_coord_offsetF = DIM*jnrF;
589 j_coord_offsetG = DIM*jnrG;
590 j_coord_offsetH = DIM*jnrH;
592 /* load j atom coordinates */
593 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
594 x+j_coord_offsetC,x+j_coord_offsetD,
595 x+j_coord_offsetE,x+j_coord_offsetF,
596 x+j_coord_offsetG,x+j_coord_offsetH,
597 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
599 /* Calculate displacement vector */
600 dx00 = _mm256_sub_ps(ix0,jx0);
601 dy00 = _mm256_sub_ps(iy0,jy0);
602 dz00 = _mm256_sub_ps(iz0,jz0);
603 dx01 = _mm256_sub_ps(ix0,jx1);
604 dy01 = _mm256_sub_ps(iy0,jy1);
605 dz01 = _mm256_sub_ps(iz0,jz1);
606 dx02 = _mm256_sub_ps(ix0,jx2);
607 dy02 = _mm256_sub_ps(iy0,jy2);
608 dz02 = _mm256_sub_ps(iz0,jz2);
609 dx10 = _mm256_sub_ps(ix1,jx0);
610 dy10 = _mm256_sub_ps(iy1,jy0);
611 dz10 = _mm256_sub_ps(iz1,jz0);
612 dx11 = _mm256_sub_ps(ix1,jx1);
613 dy11 = _mm256_sub_ps(iy1,jy1);
614 dz11 = _mm256_sub_ps(iz1,jz1);
615 dx12 = _mm256_sub_ps(ix1,jx2);
616 dy12 = _mm256_sub_ps(iy1,jy2);
617 dz12 = _mm256_sub_ps(iz1,jz2);
618 dx20 = _mm256_sub_ps(ix2,jx0);
619 dy20 = _mm256_sub_ps(iy2,jy0);
620 dz20 = _mm256_sub_ps(iz2,jz0);
621 dx21 = _mm256_sub_ps(ix2,jx1);
622 dy21 = _mm256_sub_ps(iy2,jy1);
623 dz21 = _mm256_sub_ps(iz2,jz1);
624 dx22 = _mm256_sub_ps(ix2,jx2);
625 dy22 = _mm256_sub_ps(iy2,jy2);
626 dz22 = _mm256_sub_ps(iz2,jz2);
628 /* Calculate squared distance and things based on it */
629 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
630 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
631 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
632 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
633 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
634 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
635 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
636 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
637 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
639 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
640 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
641 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
642 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
643 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
644 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
645 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
646 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
647 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
649 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
650 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
651 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
652 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
653 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
654 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
655 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
656 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
657 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
659 fjx0 = _mm256_setzero_ps();
660 fjy0 = _mm256_setzero_ps();
661 fjz0 = _mm256_setzero_ps();
662 fjx1 = _mm256_setzero_ps();
663 fjy1 = _mm256_setzero_ps();
664 fjz1 = _mm256_setzero_ps();
665 fjx2 = _mm256_setzero_ps();
666 fjy2 = _mm256_setzero_ps();
667 fjz2 = _mm256_setzero_ps();
669 /**************************
670 * CALCULATE INTERACTIONS *
671 **************************/
673 /* REACTION-FIELD ELECTROSTATICS */
674 velec = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
675 felec = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
677 /* Update potential sum for this i atom from the interaction with this j atom. */
678 velec = _mm256_andnot_ps(dummy_mask,velec);
679 velecsum = _mm256_add_ps(velecsum,velec);
683 fscal = _mm256_andnot_ps(dummy_mask,fscal);
685 /* Calculate temporary vectorial force */
686 tx = _mm256_mul_ps(fscal,dx00);
687 ty = _mm256_mul_ps(fscal,dy00);
688 tz = _mm256_mul_ps(fscal,dz00);
690 /* Update vectorial force */
691 fix0 = _mm256_add_ps(fix0,tx);
692 fiy0 = _mm256_add_ps(fiy0,ty);
693 fiz0 = _mm256_add_ps(fiz0,tz);
695 fjx0 = _mm256_add_ps(fjx0,tx);
696 fjy0 = _mm256_add_ps(fjy0,ty);
697 fjz0 = _mm256_add_ps(fjz0,tz);
699 /**************************
700 * CALCULATE INTERACTIONS *
701 **************************/
703 /* REACTION-FIELD ELECTROSTATICS */
704 velec = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
705 felec = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
707 /* Update potential sum for this i atom from the interaction with this j atom. */
708 velec = _mm256_andnot_ps(dummy_mask,velec);
709 velecsum = _mm256_add_ps(velecsum,velec);
713 fscal = _mm256_andnot_ps(dummy_mask,fscal);
715 /* Calculate temporary vectorial force */
716 tx = _mm256_mul_ps(fscal,dx01);
717 ty = _mm256_mul_ps(fscal,dy01);
718 tz = _mm256_mul_ps(fscal,dz01);
720 /* Update vectorial force */
721 fix0 = _mm256_add_ps(fix0,tx);
722 fiy0 = _mm256_add_ps(fiy0,ty);
723 fiz0 = _mm256_add_ps(fiz0,tz);
725 fjx1 = _mm256_add_ps(fjx1,tx);
726 fjy1 = _mm256_add_ps(fjy1,ty);
727 fjz1 = _mm256_add_ps(fjz1,tz);
729 /**************************
730 * CALCULATE INTERACTIONS *
731 **************************/
733 /* REACTION-FIELD ELECTROSTATICS */
734 velec = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
735 felec = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
737 /* Update potential sum for this i atom from the interaction with this j atom. */
738 velec = _mm256_andnot_ps(dummy_mask,velec);
739 velecsum = _mm256_add_ps(velecsum,velec);
743 fscal = _mm256_andnot_ps(dummy_mask,fscal);
745 /* Calculate temporary vectorial force */
746 tx = _mm256_mul_ps(fscal,dx02);
747 ty = _mm256_mul_ps(fscal,dy02);
748 tz = _mm256_mul_ps(fscal,dz02);
750 /* Update vectorial force */
751 fix0 = _mm256_add_ps(fix0,tx);
752 fiy0 = _mm256_add_ps(fiy0,ty);
753 fiz0 = _mm256_add_ps(fiz0,tz);
755 fjx2 = _mm256_add_ps(fjx2,tx);
756 fjy2 = _mm256_add_ps(fjy2,ty);
757 fjz2 = _mm256_add_ps(fjz2,tz);
759 /**************************
760 * CALCULATE INTERACTIONS *
761 **************************/
763 /* REACTION-FIELD ELECTROSTATICS */
764 velec = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
765 felec = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
767 /* Update potential sum for this i atom from the interaction with this j atom. */
768 velec = _mm256_andnot_ps(dummy_mask,velec);
769 velecsum = _mm256_add_ps(velecsum,velec);
773 fscal = _mm256_andnot_ps(dummy_mask,fscal);
775 /* Calculate temporary vectorial force */
776 tx = _mm256_mul_ps(fscal,dx10);
777 ty = _mm256_mul_ps(fscal,dy10);
778 tz = _mm256_mul_ps(fscal,dz10);
780 /* Update vectorial force */
781 fix1 = _mm256_add_ps(fix1,tx);
782 fiy1 = _mm256_add_ps(fiy1,ty);
783 fiz1 = _mm256_add_ps(fiz1,tz);
785 fjx0 = _mm256_add_ps(fjx0,tx);
786 fjy0 = _mm256_add_ps(fjy0,ty);
787 fjz0 = _mm256_add_ps(fjz0,tz);
789 /**************************
790 * CALCULATE INTERACTIONS *
791 **************************/
793 /* REACTION-FIELD ELECTROSTATICS */
794 velec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
795 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
797 /* Update potential sum for this i atom from the interaction with this j atom. */
798 velec = _mm256_andnot_ps(dummy_mask,velec);
799 velecsum = _mm256_add_ps(velecsum,velec);
803 fscal = _mm256_andnot_ps(dummy_mask,fscal);
805 /* Calculate temporary vectorial force */
806 tx = _mm256_mul_ps(fscal,dx11);
807 ty = _mm256_mul_ps(fscal,dy11);
808 tz = _mm256_mul_ps(fscal,dz11);
810 /* Update vectorial force */
811 fix1 = _mm256_add_ps(fix1,tx);
812 fiy1 = _mm256_add_ps(fiy1,ty);
813 fiz1 = _mm256_add_ps(fiz1,tz);
815 fjx1 = _mm256_add_ps(fjx1,tx);
816 fjy1 = _mm256_add_ps(fjy1,ty);
817 fjz1 = _mm256_add_ps(fjz1,tz);
819 /**************************
820 * CALCULATE INTERACTIONS *
821 **************************/
823 /* REACTION-FIELD ELECTROSTATICS */
824 velec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
825 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
827 /* Update potential sum for this i atom from the interaction with this j atom. */
828 velec = _mm256_andnot_ps(dummy_mask,velec);
829 velecsum = _mm256_add_ps(velecsum,velec);
833 fscal = _mm256_andnot_ps(dummy_mask,fscal);
835 /* Calculate temporary vectorial force */
836 tx = _mm256_mul_ps(fscal,dx12);
837 ty = _mm256_mul_ps(fscal,dy12);
838 tz = _mm256_mul_ps(fscal,dz12);
840 /* Update vectorial force */
841 fix1 = _mm256_add_ps(fix1,tx);
842 fiy1 = _mm256_add_ps(fiy1,ty);
843 fiz1 = _mm256_add_ps(fiz1,tz);
845 fjx2 = _mm256_add_ps(fjx2,tx);
846 fjy2 = _mm256_add_ps(fjy2,ty);
847 fjz2 = _mm256_add_ps(fjz2,tz);
849 /**************************
850 * CALCULATE INTERACTIONS *
851 **************************/
853 /* REACTION-FIELD ELECTROSTATICS */
854 velec = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
855 felec = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
857 /* Update potential sum for this i atom from the interaction with this j atom. */
858 velec = _mm256_andnot_ps(dummy_mask,velec);
859 velecsum = _mm256_add_ps(velecsum,velec);
863 fscal = _mm256_andnot_ps(dummy_mask,fscal);
865 /* Calculate temporary vectorial force */
866 tx = _mm256_mul_ps(fscal,dx20);
867 ty = _mm256_mul_ps(fscal,dy20);
868 tz = _mm256_mul_ps(fscal,dz20);
870 /* Update vectorial force */
871 fix2 = _mm256_add_ps(fix2,tx);
872 fiy2 = _mm256_add_ps(fiy2,ty);
873 fiz2 = _mm256_add_ps(fiz2,tz);
875 fjx0 = _mm256_add_ps(fjx0,tx);
876 fjy0 = _mm256_add_ps(fjy0,ty);
877 fjz0 = _mm256_add_ps(fjz0,tz);
879 /**************************
880 * CALCULATE INTERACTIONS *
881 **************************/
883 /* REACTION-FIELD ELECTROSTATICS */
884 velec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
885 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
887 /* Update potential sum for this i atom from the interaction with this j atom. */
888 velec = _mm256_andnot_ps(dummy_mask,velec);
889 velecsum = _mm256_add_ps(velecsum,velec);
893 fscal = _mm256_andnot_ps(dummy_mask,fscal);
895 /* Calculate temporary vectorial force */
896 tx = _mm256_mul_ps(fscal,dx21);
897 ty = _mm256_mul_ps(fscal,dy21);
898 tz = _mm256_mul_ps(fscal,dz21);
900 /* Update vectorial force */
901 fix2 = _mm256_add_ps(fix2,tx);
902 fiy2 = _mm256_add_ps(fiy2,ty);
903 fiz2 = _mm256_add_ps(fiz2,tz);
905 fjx1 = _mm256_add_ps(fjx1,tx);
906 fjy1 = _mm256_add_ps(fjy1,ty);
907 fjz1 = _mm256_add_ps(fjz1,tz);
909 /**************************
910 * CALCULATE INTERACTIONS *
911 **************************/
913 /* REACTION-FIELD ELECTROSTATICS */
914 velec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
915 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
917 /* Update potential sum for this i atom from the interaction with this j atom. */
918 velec = _mm256_andnot_ps(dummy_mask,velec);
919 velecsum = _mm256_add_ps(velecsum,velec);
923 fscal = _mm256_andnot_ps(dummy_mask,fscal);
925 /* Calculate temporary vectorial force */
926 tx = _mm256_mul_ps(fscal,dx22);
927 ty = _mm256_mul_ps(fscal,dy22);
928 tz = _mm256_mul_ps(fscal,dz22);
930 /* Update vectorial force */
931 fix2 = _mm256_add_ps(fix2,tx);
932 fiy2 = _mm256_add_ps(fiy2,ty);
933 fiz2 = _mm256_add_ps(fiz2,tz);
935 fjx2 = _mm256_add_ps(fjx2,tx);
936 fjy2 = _mm256_add_ps(fjy2,ty);
937 fjz2 = _mm256_add_ps(fjz2,tz);
939 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
940 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
941 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
942 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
943 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
944 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
945 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
946 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
948 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
949 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
951 /* Inner loop uses 288 flops */
954 /* End of innermost loop */
956 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
957 f+i_coord_offset,fshift+i_shift_offset);
960 /* Update potential energies */
961 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
963 /* Increment number of inner iterations */
964 inneriter += j_index_end - j_index_start;
966 /* Outer loop uses 19 flops */
969 /* Increment number of outer iterations */
972 /* Update outer/inner flops */
974 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*288);
977 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_single
978 * Electrostatics interaction: ReactionField
979 * VdW interaction: None
980 * Geometry: Water3-Water3
981 * Calculate force/pot: Force
984 nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_single
985 (t_nblist * gmx_restrict nlist,
986 rvec * gmx_restrict xx,
987 rvec * gmx_restrict ff,
988 t_forcerec * gmx_restrict fr,
989 t_mdatoms * gmx_restrict mdatoms,
990 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
991 t_nrnb * gmx_restrict nrnb)
993 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
994 * just 0 for non-waters.
995 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
996 * jnr indices corresponding to data put in the four positions in the SIMD register.
998 int i_shift_offset,i_coord_offset,outeriter,inneriter;
999 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1000 int jnrA,jnrB,jnrC,jnrD;
1001 int jnrE,jnrF,jnrG,jnrH;
1002 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1003 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1004 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1005 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1006 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1007 real rcutoff_scalar;
1008 real *shiftvec,*fshift,*x,*f;
1009 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1010 real scratch[4*DIM];
1011 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1012 real * vdwioffsetptr0;
1013 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1014 real * vdwioffsetptr1;
1015 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1016 real * vdwioffsetptr2;
1017 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1018 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1019 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1020 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1021 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1022 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1023 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1024 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1025 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1026 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1027 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1028 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1029 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1030 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1031 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1032 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1033 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
1035 __m256 dummy_mask,cutoff_mask;
1036 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1037 __m256 one = _mm256_set1_ps(1.0);
1038 __m256 two = _mm256_set1_ps(2.0);
1044 jindex = nlist->jindex;
1046 shiftidx = nlist->shift;
1048 shiftvec = fr->shift_vec[0];
1049 fshift = fr->fshift[0];
1050 facel = _mm256_set1_ps(fr->epsfac);
1051 charge = mdatoms->chargeA;
1052 krf = _mm256_set1_ps(fr->ic->k_rf);
1053 krf2 = _mm256_set1_ps(fr->ic->k_rf*2.0);
1054 crf = _mm256_set1_ps(fr->ic->c_rf);
1056 /* Setup water-specific parameters */
1057 inr = nlist->iinr[0];
1058 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1059 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1060 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1062 jq0 = _mm256_set1_ps(charge[inr+0]);
1063 jq1 = _mm256_set1_ps(charge[inr+1]);
1064 jq2 = _mm256_set1_ps(charge[inr+2]);
1065 qq00 = _mm256_mul_ps(iq0,jq0);
1066 qq01 = _mm256_mul_ps(iq0,jq1);
1067 qq02 = _mm256_mul_ps(iq0,jq2);
1068 qq10 = _mm256_mul_ps(iq1,jq0);
1069 qq11 = _mm256_mul_ps(iq1,jq1);
1070 qq12 = _mm256_mul_ps(iq1,jq2);
1071 qq20 = _mm256_mul_ps(iq2,jq0);
1072 qq21 = _mm256_mul_ps(iq2,jq1);
1073 qq22 = _mm256_mul_ps(iq2,jq2);
1075 /* Avoid stupid compiler warnings */
1076 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1077 j_coord_offsetA = 0;
1078 j_coord_offsetB = 0;
1079 j_coord_offsetC = 0;
1080 j_coord_offsetD = 0;
1081 j_coord_offsetE = 0;
1082 j_coord_offsetF = 0;
1083 j_coord_offsetG = 0;
1084 j_coord_offsetH = 0;
1089 for(iidx=0;iidx<4*DIM;iidx++)
1091 scratch[iidx] = 0.0;
1094 /* Start outer loop over neighborlists */
1095 for(iidx=0; iidx<nri; iidx++)
1097 /* Load shift vector for this list */
1098 i_shift_offset = DIM*shiftidx[iidx];
1100 /* Load limits for loop over neighbors */
1101 j_index_start = jindex[iidx];
1102 j_index_end = jindex[iidx+1];
1104 /* Get outer coordinate index */
1106 i_coord_offset = DIM*inr;
1108 /* Load i particle coords and add shift vector */
1109 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1110 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1112 fix0 = _mm256_setzero_ps();
1113 fiy0 = _mm256_setzero_ps();
1114 fiz0 = _mm256_setzero_ps();
1115 fix1 = _mm256_setzero_ps();
1116 fiy1 = _mm256_setzero_ps();
1117 fiz1 = _mm256_setzero_ps();
1118 fix2 = _mm256_setzero_ps();
1119 fiy2 = _mm256_setzero_ps();
1120 fiz2 = _mm256_setzero_ps();
1122 /* Start inner kernel loop */
1123 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1126 /* Get j neighbor index, and coordinate index */
1128 jnrB = jjnr[jidx+1];
1129 jnrC = jjnr[jidx+2];
1130 jnrD = jjnr[jidx+3];
1131 jnrE = jjnr[jidx+4];
1132 jnrF = jjnr[jidx+5];
1133 jnrG = jjnr[jidx+6];
1134 jnrH = jjnr[jidx+7];
1135 j_coord_offsetA = DIM*jnrA;
1136 j_coord_offsetB = DIM*jnrB;
1137 j_coord_offsetC = DIM*jnrC;
1138 j_coord_offsetD = DIM*jnrD;
1139 j_coord_offsetE = DIM*jnrE;
1140 j_coord_offsetF = DIM*jnrF;
1141 j_coord_offsetG = DIM*jnrG;
1142 j_coord_offsetH = DIM*jnrH;
1144 /* load j atom coordinates */
1145 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1146 x+j_coord_offsetC,x+j_coord_offsetD,
1147 x+j_coord_offsetE,x+j_coord_offsetF,
1148 x+j_coord_offsetG,x+j_coord_offsetH,
1149 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1151 /* Calculate displacement vector */
1152 dx00 = _mm256_sub_ps(ix0,jx0);
1153 dy00 = _mm256_sub_ps(iy0,jy0);
1154 dz00 = _mm256_sub_ps(iz0,jz0);
1155 dx01 = _mm256_sub_ps(ix0,jx1);
1156 dy01 = _mm256_sub_ps(iy0,jy1);
1157 dz01 = _mm256_sub_ps(iz0,jz1);
1158 dx02 = _mm256_sub_ps(ix0,jx2);
1159 dy02 = _mm256_sub_ps(iy0,jy2);
1160 dz02 = _mm256_sub_ps(iz0,jz2);
1161 dx10 = _mm256_sub_ps(ix1,jx0);
1162 dy10 = _mm256_sub_ps(iy1,jy0);
1163 dz10 = _mm256_sub_ps(iz1,jz0);
1164 dx11 = _mm256_sub_ps(ix1,jx1);
1165 dy11 = _mm256_sub_ps(iy1,jy1);
1166 dz11 = _mm256_sub_ps(iz1,jz1);
1167 dx12 = _mm256_sub_ps(ix1,jx2);
1168 dy12 = _mm256_sub_ps(iy1,jy2);
1169 dz12 = _mm256_sub_ps(iz1,jz2);
1170 dx20 = _mm256_sub_ps(ix2,jx0);
1171 dy20 = _mm256_sub_ps(iy2,jy0);
1172 dz20 = _mm256_sub_ps(iz2,jz0);
1173 dx21 = _mm256_sub_ps(ix2,jx1);
1174 dy21 = _mm256_sub_ps(iy2,jy1);
1175 dz21 = _mm256_sub_ps(iz2,jz1);
1176 dx22 = _mm256_sub_ps(ix2,jx2);
1177 dy22 = _mm256_sub_ps(iy2,jy2);
1178 dz22 = _mm256_sub_ps(iz2,jz2);
1180 /* Calculate squared distance and things based on it */
1181 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1182 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1183 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1184 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1185 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1186 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1187 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1188 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1189 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1191 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1192 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1193 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1194 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1195 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1196 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1197 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1198 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1199 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1201 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1202 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1203 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1204 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1205 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1206 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1207 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1208 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1209 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1211 fjx0 = _mm256_setzero_ps();
1212 fjy0 = _mm256_setzero_ps();
1213 fjz0 = _mm256_setzero_ps();
1214 fjx1 = _mm256_setzero_ps();
1215 fjy1 = _mm256_setzero_ps();
1216 fjz1 = _mm256_setzero_ps();
1217 fjx2 = _mm256_setzero_ps();
1218 fjy2 = _mm256_setzero_ps();
1219 fjz2 = _mm256_setzero_ps();
1221 /**************************
1222 * CALCULATE INTERACTIONS *
1223 **************************/
1225 /* REACTION-FIELD ELECTROSTATICS */
1226 felec = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1230 /* Calculate temporary vectorial force */
1231 tx = _mm256_mul_ps(fscal,dx00);
1232 ty = _mm256_mul_ps(fscal,dy00);
1233 tz = _mm256_mul_ps(fscal,dz00);
1235 /* Update vectorial force */
1236 fix0 = _mm256_add_ps(fix0,tx);
1237 fiy0 = _mm256_add_ps(fiy0,ty);
1238 fiz0 = _mm256_add_ps(fiz0,tz);
1240 fjx0 = _mm256_add_ps(fjx0,tx);
1241 fjy0 = _mm256_add_ps(fjy0,ty);
1242 fjz0 = _mm256_add_ps(fjz0,tz);
1244 /**************************
1245 * CALCULATE INTERACTIONS *
1246 **************************/
1248 /* REACTION-FIELD ELECTROSTATICS */
1249 felec = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
1253 /* Calculate temporary vectorial force */
1254 tx = _mm256_mul_ps(fscal,dx01);
1255 ty = _mm256_mul_ps(fscal,dy01);
1256 tz = _mm256_mul_ps(fscal,dz01);
1258 /* Update vectorial force */
1259 fix0 = _mm256_add_ps(fix0,tx);
1260 fiy0 = _mm256_add_ps(fiy0,ty);
1261 fiz0 = _mm256_add_ps(fiz0,tz);
1263 fjx1 = _mm256_add_ps(fjx1,tx);
1264 fjy1 = _mm256_add_ps(fjy1,ty);
1265 fjz1 = _mm256_add_ps(fjz1,tz);
1267 /**************************
1268 * CALCULATE INTERACTIONS *
1269 **************************/
1271 /* REACTION-FIELD ELECTROSTATICS */
1272 felec = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
1276 /* Calculate temporary vectorial force */
1277 tx = _mm256_mul_ps(fscal,dx02);
1278 ty = _mm256_mul_ps(fscal,dy02);
1279 tz = _mm256_mul_ps(fscal,dz02);
1281 /* Update vectorial force */
1282 fix0 = _mm256_add_ps(fix0,tx);
1283 fiy0 = _mm256_add_ps(fiy0,ty);
1284 fiz0 = _mm256_add_ps(fiz0,tz);
1286 fjx2 = _mm256_add_ps(fjx2,tx);
1287 fjy2 = _mm256_add_ps(fjy2,ty);
1288 fjz2 = _mm256_add_ps(fjz2,tz);
1290 /**************************
1291 * CALCULATE INTERACTIONS *
1292 **************************/
1294 /* REACTION-FIELD ELECTROSTATICS */
1295 felec = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1299 /* Calculate temporary vectorial force */
1300 tx = _mm256_mul_ps(fscal,dx10);
1301 ty = _mm256_mul_ps(fscal,dy10);
1302 tz = _mm256_mul_ps(fscal,dz10);
1304 /* Update vectorial force */
1305 fix1 = _mm256_add_ps(fix1,tx);
1306 fiy1 = _mm256_add_ps(fiy1,ty);
1307 fiz1 = _mm256_add_ps(fiz1,tz);
1309 fjx0 = _mm256_add_ps(fjx0,tx);
1310 fjy0 = _mm256_add_ps(fjy0,ty);
1311 fjz0 = _mm256_add_ps(fjz0,tz);
1313 /**************************
1314 * CALCULATE INTERACTIONS *
1315 **************************/
1317 /* REACTION-FIELD ELECTROSTATICS */
1318 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1322 /* Calculate temporary vectorial force */
1323 tx = _mm256_mul_ps(fscal,dx11);
1324 ty = _mm256_mul_ps(fscal,dy11);
1325 tz = _mm256_mul_ps(fscal,dz11);
1327 /* Update vectorial force */
1328 fix1 = _mm256_add_ps(fix1,tx);
1329 fiy1 = _mm256_add_ps(fiy1,ty);
1330 fiz1 = _mm256_add_ps(fiz1,tz);
1332 fjx1 = _mm256_add_ps(fjx1,tx);
1333 fjy1 = _mm256_add_ps(fjy1,ty);
1334 fjz1 = _mm256_add_ps(fjz1,tz);
1336 /**************************
1337 * CALCULATE INTERACTIONS *
1338 **************************/
1340 /* REACTION-FIELD ELECTROSTATICS */
1341 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1345 /* Calculate temporary vectorial force */
1346 tx = _mm256_mul_ps(fscal,dx12);
1347 ty = _mm256_mul_ps(fscal,dy12);
1348 tz = _mm256_mul_ps(fscal,dz12);
1350 /* Update vectorial force */
1351 fix1 = _mm256_add_ps(fix1,tx);
1352 fiy1 = _mm256_add_ps(fiy1,ty);
1353 fiz1 = _mm256_add_ps(fiz1,tz);
1355 fjx2 = _mm256_add_ps(fjx2,tx);
1356 fjy2 = _mm256_add_ps(fjy2,ty);
1357 fjz2 = _mm256_add_ps(fjz2,tz);
1359 /**************************
1360 * CALCULATE INTERACTIONS *
1361 **************************/
1363 /* REACTION-FIELD ELECTROSTATICS */
1364 felec = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1368 /* Calculate temporary vectorial force */
1369 tx = _mm256_mul_ps(fscal,dx20);
1370 ty = _mm256_mul_ps(fscal,dy20);
1371 tz = _mm256_mul_ps(fscal,dz20);
1373 /* Update vectorial force */
1374 fix2 = _mm256_add_ps(fix2,tx);
1375 fiy2 = _mm256_add_ps(fiy2,ty);
1376 fiz2 = _mm256_add_ps(fiz2,tz);
1378 fjx0 = _mm256_add_ps(fjx0,tx);
1379 fjy0 = _mm256_add_ps(fjy0,ty);
1380 fjz0 = _mm256_add_ps(fjz0,tz);
1382 /**************************
1383 * CALCULATE INTERACTIONS *
1384 **************************/
1386 /* REACTION-FIELD ELECTROSTATICS */
1387 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1391 /* Calculate temporary vectorial force */
1392 tx = _mm256_mul_ps(fscal,dx21);
1393 ty = _mm256_mul_ps(fscal,dy21);
1394 tz = _mm256_mul_ps(fscal,dz21);
1396 /* Update vectorial force */
1397 fix2 = _mm256_add_ps(fix2,tx);
1398 fiy2 = _mm256_add_ps(fiy2,ty);
1399 fiz2 = _mm256_add_ps(fiz2,tz);
1401 fjx1 = _mm256_add_ps(fjx1,tx);
1402 fjy1 = _mm256_add_ps(fjy1,ty);
1403 fjz1 = _mm256_add_ps(fjz1,tz);
1405 /**************************
1406 * CALCULATE INTERACTIONS *
1407 **************************/
1409 /* REACTION-FIELD ELECTROSTATICS */
1410 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1414 /* Calculate temporary vectorial force */
1415 tx = _mm256_mul_ps(fscal,dx22);
1416 ty = _mm256_mul_ps(fscal,dy22);
1417 tz = _mm256_mul_ps(fscal,dz22);
1419 /* Update vectorial force */
1420 fix2 = _mm256_add_ps(fix2,tx);
1421 fiy2 = _mm256_add_ps(fiy2,ty);
1422 fiz2 = _mm256_add_ps(fiz2,tz);
1424 fjx2 = _mm256_add_ps(fjx2,tx);
1425 fjy2 = _mm256_add_ps(fjy2,ty);
1426 fjz2 = _mm256_add_ps(fjz2,tz);
1428 fjptrA = f+j_coord_offsetA;
1429 fjptrB = f+j_coord_offsetB;
1430 fjptrC = f+j_coord_offsetC;
1431 fjptrD = f+j_coord_offsetD;
1432 fjptrE = f+j_coord_offsetE;
1433 fjptrF = f+j_coord_offsetF;
1434 fjptrG = f+j_coord_offsetG;
1435 fjptrH = f+j_coord_offsetH;
1437 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1438 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1440 /* Inner loop uses 243 flops */
1443 if(jidx<j_index_end)
1446 /* Get j neighbor index, and coordinate index */
1447 jnrlistA = jjnr[jidx];
1448 jnrlistB = jjnr[jidx+1];
1449 jnrlistC = jjnr[jidx+2];
1450 jnrlistD = jjnr[jidx+3];
1451 jnrlistE = jjnr[jidx+4];
1452 jnrlistF = jjnr[jidx+5];
1453 jnrlistG = jjnr[jidx+6];
1454 jnrlistH = jjnr[jidx+7];
1455 /* Sign of each element will be negative for non-real atoms.
1456 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1457 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1459 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1460 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1462 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1463 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1464 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1465 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1466 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
1467 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
1468 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
1469 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
1470 j_coord_offsetA = DIM*jnrA;
1471 j_coord_offsetB = DIM*jnrB;
1472 j_coord_offsetC = DIM*jnrC;
1473 j_coord_offsetD = DIM*jnrD;
1474 j_coord_offsetE = DIM*jnrE;
1475 j_coord_offsetF = DIM*jnrF;
1476 j_coord_offsetG = DIM*jnrG;
1477 j_coord_offsetH = DIM*jnrH;
1479 /* load j atom coordinates */
1480 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1481 x+j_coord_offsetC,x+j_coord_offsetD,
1482 x+j_coord_offsetE,x+j_coord_offsetF,
1483 x+j_coord_offsetG,x+j_coord_offsetH,
1484 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1486 /* Calculate displacement vector */
1487 dx00 = _mm256_sub_ps(ix0,jx0);
1488 dy00 = _mm256_sub_ps(iy0,jy0);
1489 dz00 = _mm256_sub_ps(iz0,jz0);
1490 dx01 = _mm256_sub_ps(ix0,jx1);
1491 dy01 = _mm256_sub_ps(iy0,jy1);
1492 dz01 = _mm256_sub_ps(iz0,jz1);
1493 dx02 = _mm256_sub_ps(ix0,jx2);
1494 dy02 = _mm256_sub_ps(iy0,jy2);
1495 dz02 = _mm256_sub_ps(iz0,jz2);
1496 dx10 = _mm256_sub_ps(ix1,jx0);
1497 dy10 = _mm256_sub_ps(iy1,jy0);
1498 dz10 = _mm256_sub_ps(iz1,jz0);
1499 dx11 = _mm256_sub_ps(ix1,jx1);
1500 dy11 = _mm256_sub_ps(iy1,jy1);
1501 dz11 = _mm256_sub_ps(iz1,jz1);
1502 dx12 = _mm256_sub_ps(ix1,jx2);
1503 dy12 = _mm256_sub_ps(iy1,jy2);
1504 dz12 = _mm256_sub_ps(iz1,jz2);
1505 dx20 = _mm256_sub_ps(ix2,jx0);
1506 dy20 = _mm256_sub_ps(iy2,jy0);
1507 dz20 = _mm256_sub_ps(iz2,jz0);
1508 dx21 = _mm256_sub_ps(ix2,jx1);
1509 dy21 = _mm256_sub_ps(iy2,jy1);
1510 dz21 = _mm256_sub_ps(iz2,jz1);
1511 dx22 = _mm256_sub_ps(ix2,jx2);
1512 dy22 = _mm256_sub_ps(iy2,jy2);
1513 dz22 = _mm256_sub_ps(iz2,jz2);
1515 /* Calculate squared distance and things based on it */
1516 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1517 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1518 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1519 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1520 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1521 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1522 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1523 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1524 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1526 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1527 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1528 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1529 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1530 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1531 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1532 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1533 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1534 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1536 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1537 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1538 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1539 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1540 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1541 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1542 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1543 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1544 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1546 fjx0 = _mm256_setzero_ps();
1547 fjy0 = _mm256_setzero_ps();
1548 fjz0 = _mm256_setzero_ps();
1549 fjx1 = _mm256_setzero_ps();
1550 fjy1 = _mm256_setzero_ps();
1551 fjz1 = _mm256_setzero_ps();
1552 fjx2 = _mm256_setzero_ps();
1553 fjy2 = _mm256_setzero_ps();
1554 fjz2 = _mm256_setzero_ps();
1556 /**************************
1557 * CALCULATE INTERACTIONS *
1558 **************************/
1560 /* REACTION-FIELD ELECTROSTATICS */
1561 felec = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1565 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1567 /* Calculate temporary vectorial force */
1568 tx = _mm256_mul_ps(fscal,dx00);
1569 ty = _mm256_mul_ps(fscal,dy00);
1570 tz = _mm256_mul_ps(fscal,dz00);
1572 /* Update vectorial force */
1573 fix0 = _mm256_add_ps(fix0,tx);
1574 fiy0 = _mm256_add_ps(fiy0,ty);
1575 fiz0 = _mm256_add_ps(fiz0,tz);
1577 fjx0 = _mm256_add_ps(fjx0,tx);
1578 fjy0 = _mm256_add_ps(fjy0,ty);
1579 fjz0 = _mm256_add_ps(fjz0,tz);
1581 /**************************
1582 * CALCULATE INTERACTIONS *
1583 **************************/
1585 /* REACTION-FIELD ELECTROSTATICS */
1586 felec = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
1590 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1592 /* Calculate temporary vectorial force */
1593 tx = _mm256_mul_ps(fscal,dx01);
1594 ty = _mm256_mul_ps(fscal,dy01);
1595 tz = _mm256_mul_ps(fscal,dz01);
1597 /* Update vectorial force */
1598 fix0 = _mm256_add_ps(fix0,tx);
1599 fiy0 = _mm256_add_ps(fiy0,ty);
1600 fiz0 = _mm256_add_ps(fiz0,tz);
1602 fjx1 = _mm256_add_ps(fjx1,tx);
1603 fjy1 = _mm256_add_ps(fjy1,ty);
1604 fjz1 = _mm256_add_ps(fjz1,tz);
1606 /**************************
1607 * CALCULATE INTERACTIONS *
1608 **************************/
1610 /* REACTION-FIELD ELECTROSTATICS */
1611 felec = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
1615 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1617 /* Calculate temporary vectorial force */
1618 tx = _mm256_mul_ps(fscal,dx02);
1619 ty = _mm256_mul_ps(fscal,dy02);
1620 tz = _mm256_mul_ps(fscal,dz02);
1622 /* Update vectorial force */
1623 fix0 = _mm256_add_ps(fix0,tx);
1624 fiy0 = _mm256_add_ps(fiy0,ty);
1625 fiz0 = _mm256_add_ps(fiz0,tz);
1627 fjx2 = _mm256_add_ps(fjx2,tx);
1628 fjy2 = _mm256_add_ps(fjy2,ty);
1629 fjz2 = _mm256_add_ps(fjz2,tz);
1631 /**************************
1632 * CALCULATE INTERACTIONS *
1633 **************************/
1635 /* REACTION-FIELD ELECTROSTATICS */
1636 felec = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1640 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1642 /* Calculate temporary vectorial force */
1643 tx = _mm256_mul_ps(fscal,dx10);
1644 ty = _mm256_mul_ps(fscal,dy10);
1645 tz = _mm256_mul_ps(fscal,dz10);
1647 /* Update vectorial force */
1648 fix1 = _mm256_add_ps(fix1,tx);
1649 fiy1 = _mm256_add_ps(fiy1,ty);
1650 fiz1 = _mm256_add_ps(fiz1,tz);
1652 fjx0 = _mm256_add_ps(fjx0,tx);
1653 fjy0 = _mm256_add_ps(fjy0,ty);
1654 fjz0 = _mm256_add_ps(fjz0,tz);
1656 /**************************
1657 * CALCULATE INTERACTIONS *
1658 **************************/
1660 /* REACTION-FIELD ELECTROSTATICS */
1661 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1665 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1667 /* Calculate temporary vectorial force */
1668 tx = _mm256_mul_ps(fscal,dx11);
1669 ty = _mm256_mul_ps(fscal,dy11);
1670 tz = _mm256_mul_ps(fscal,dz11);
1672 /* Update vectorial force */
1673 fix1 = _mm256_add_ps(fix1,tx);
1674 fiy1 = _mm256_add_ps(fiy1,ty);
1675 fiz1 = _mm256_add_ps(fiz1,tz);
1677 fjx1 = _mm256_add_ps(fjx1,tx);
1678 fjy1 = _mm256_add_ps(fjy1,ty);
1679 fjz1 = _mm256_add_ps(fjz1,tz);
1681 /**************************
1682 * CALCULATE INTERACTIONS *
1683 **************************/
1685 /* REACTION-FIELD ELECTROSTATICS */
1686 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1690 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1692 /* Calculate temporary vectorial force */
1693 tx = _mm256_mul_ps(fscal,dx12);
1694 ty = _mm256_mul_ps(fscal,dy12);
1695 tz = _mm256_mul_ps(fscal,dz12);
1697 /* Update vectorial force */
1698 fix1 = _mm256_add_ps(fix1,tx);
1699 fiy1 = _mm256_add_ps(fiy1,ty);
1700 fiz1 = _mm256_add_ps(fiz1,tz);
1702 fjx2 = _mm256_add_ps(fjx2,tx);
1703 fjy2 = _mm256_add_ps(fjy2,ty);
1704 fjz2 = _mm256_add_ps(fjz2,tz);
1706 /**************************
1707 * CALCULATE INTERACTIONS *
1708 **************************/
1710 /* REACTION-FIELD ELECTROSTATICS */
1711 felec = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1715 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1717 /* Calculate temporary vectorial force */
1718 tx = _mm256_mul_ps(fscal,dx20);
1719 ty = _mm256_mul_ps(fscal,dy20);
1720 tz = _mm256_mul_ps(fscal,dz20);
1722 /* Update vectorial force */
1723 fix2 = _mm256_add_ps(fix2,tx);
1724 fiy2 = _mm256_add_ps(fiy2,ty);
1725 fiz2 = _mm256_add_ps(fiz2,tz);
1727 fjx0 = _mm256_add_ps(fjx0,tx);
1728 fjy0 = _mm256_add_ps(fjy0,ty);
1729 fjz0 = _mm256_add_ps(fjz0,tz);
1731 /**************************
1732 * CALCULATE INTERACTIONS *
1733 **************************/
1735 /* REACTION-FIELD ELECTROSTATICS */
1736 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1740 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1742 /* Calculate temporary vectorial force */
1743 tx = _mm256_mul_ps(fscal,dx21);
1744 ty = _mm256_mul_ps(fscal,dy21);
1745 tz = _mm256_mul_ps(fscal,dz21);
1747 /* Update vectorial force */
1748 fix2 = _mm256_add_ps(fix2,tx);
1749 fiy2 = _mm256_add_ps(fiy2,ty);
1750 fiz2 = _mm256_add_ps(fiz2,tz);
1752 fjx1 = _mm256_add_ps(fjx1,tx);
1753 fjy1 = _mm256_add_ps(fjy1,ty);
1754 fjz1 = _mm256_add_ps(fjz1,tz);
1756 /**************************
1757 * CALCULATE INTERACTIONS *
1758 **************************/
1760 /* REACTION-FIELD ELECTROSTATICS */
1761 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1765 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1767 /* Calculate temporary vectorial force */
1768 tx = _mm256_mul_ps(fscal,dx22);
1769 ty = _mm256_mul_ps(fscal,dy22);
1770 tz = _mm256_mul_ps(fscal,dz22);
1772 /* Update vectorial force */
1773 fix2 = _mm256_add_ps(fix2,tx);
1774 fiy2 = _mm256_add_ps(fiy2,ty);
1775 fiz2 = _mm256_add_ps(fiz2,tz);
1777 fjx2 = _mm256_add_ps(fjx2,tx);
1778 fjy2 = _mm256_add_ps(fjy2,ty);
1779 fjz2 = _mm256_add_ps(fjz2,tz);
1781 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1782 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1783 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1784 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1785 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1786 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1787 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1788 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1790 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1791 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1793 /* Inner loop uses 243 flops */
1796 /* End of innermost loop */
1798 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1799 f+i_coord_offset,fshift+i_shift_offset);
1801 /* Increment number of inner iterations */
1802 inneriter += j_index_end - j_index_start;
1804 /* Outer loop uses 18 flops */
1807 /* Increment number of outer iterations */
1810 /* Update outer/inner flops */
1812 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*243);