2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015,2017, 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.
44 #include "../nb_kernel.h"
45 #include "gromacs/gmxlib/nrnb.h"
47 #include "kernelutil_x86_avx_256_single.h"
50 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_avx_256_single
51 * Electrostatics interaction: ReactionField
52 * VdW interaction: LennardJones
53 * Geometry: Water4-Water4
54 * Calculate force/pot: PotentialAndForce
57 nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_avx_256_single
58 (t_nblist * gmx_restrict nlist,
59 rvec * gmx_restrict xx,
60 rvec * gmx_restrict ff,
61 struct t_forcerec * gmx_restrict fr,
62 t_mdatoms * gmx_restrict mdatoms,
63 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
64 t_nrnb * gmx_restrict nrnb)
66 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
67 * just 0 for non-waters.
68 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
69 * jnr indices corresponding to data put in the four positions in the SIMD register.
71 int i_shift_offset,i_coord_offset,outeriter,inneriter;
72 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
73 int jnrA,jnrB,jnrC,jnrD;
74 int jnrE,jnrF,jnrG,jnrH;
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 j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
79 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
81 real *shiftvec,*fshift,*x,*f;
82 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
84 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
85 real * vdwioffsetptr0;
86 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
87 real * vdwioffsetptr1;
88 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
89 real * vdwioffsetptr2;
90 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
91 real * vdwioffsetptr3;
92 __m256 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
93 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
94 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
95 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
96 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
97 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
98 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
99 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
100 __m256 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
101 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
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 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
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 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
108 __m256 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
109 __m256 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
110 __m256 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
111 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
114 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
117 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
118 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
119 __m256 dummy_mask,cutoff_mask;
120 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
121 __m256 one = _mm256_set1_ps(1.0);
122 __m256 two = _mm256_set1_ps(2.0);
128 jindex = nlist->jindex;
130 shiftidx = nlist->shift;
132 shiftvec = fr->shift_vec[0];
133 fshift = fr->fshift[0];
134 facel = _mm256_set1_ps(fr->ic->epsfac);
135 charge = mdatoms->chargeA;
136 krf = _mm256_set1_ps(fr->ic->k_rf);
137 krf2 = _mm256_set1_ps(fr->ic->k_rf*2.0);
138 crf = _mm256_set1_ps(fr->ic->c_rf);
139 nvdwtype = fr->ntype;
141 vdwtype = mdatoms->typeA;
143 /* Setup water-specific parameters */
144 inr = nlist->iinr[0];
145 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
146 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
147 iq3 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
148 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
150 jq1 = _mm256_set1_ps(charge[inr+1]);
151 jq2 = _mm256_set1_ps(charge[inr+2]);
152 jq3 = _mm256_set1_ps(charge[inr+3]);
153 vdwjidx0A = 2*vdwtype[inr+0];
154 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
155 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
156 qq11 = _mm256_mul_ps(iq1,jq1);
157 qq12 = _mm256_mul_ps(iq1,jq2);
158 qq13 = _mm256_mul_ps(iq1,jq3);
159 qq21 = _mm256_mul_ps(iq2,jq1);
160 qq22 = _mm256_mul_ps(iq2,jq2);
161 qq23 = _mm256_mul_ps(iq2,jq3);
162 qq31 = _mm256_mul_ps(iq3,jq1);
163 qq32 = _mm256_mul_ps(iq3,jq2);
164 qq33 = _mm256_mul_ps(iq3,jq3);
166 /* Avoid stupid compiler warnings */
167 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
180 for(iidx=0;iidx<4*DIM;iidx++)
185 /* Start outer loop over neighborlists */
186 for(iidx=0; iidx<nri; iidx++)
188 /* Load shift vector for this list */
189 i_shift_offset = DIM*shiftidx[iidx];
191 /* Load limits for loop over neighbors */
192 j_index_start = jindex[iidx];
193 j_index_end = jindex[iidx+1];
195 /* Get outer coordinate index */
197 i_coord_offset = DIM*inr;
199 /* Load i particle coords and add shift vector */
200 gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
201 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
203 fix0 = _mm256_setzero_ps();
204 fiy0 = _mm256_setzero_ps();
205 fiz0 = _mm256_setzero_ps();
206 fix1 = _mm256_setzero_ps();
207 fiy1 = _mm256_setzero_ps();
208 fiz1 = _mm256_setzero_ps();
209 fix2 = _mm256_setzero_ps();
210 fiy2 = _mm256_setzero_ps();
211 fiz2 = _mm256_setzero_ps();
212 fix3 = _mm256_setzero_ps();
213 fiy3 = _mm256_setzero_ps();
214 fiz3 = _mm256_setzero_ps();
216 /* Reset potential sums */
217 velecsum = _mm256_setzero_ps();
218 vvdwsum = _mm256_setzero_ps();
220 /* Start inner kernel loop */
221 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
224 /* Get j neighbor index, and coordinate index */
233 j_coord_offsetA = DIM*jnrA;
234 j_coord_offsetB = DIM*jnrB;
235 j_coord_offsetC = DIM*jnrC;
236 j_coord_offsetD = DIM*jnrD;
237 j_coord_offsetE = DIM*jnrE;
238 j_coord_offsetF = DIM*jnrF;
239 j_coord_offsetG = DIM*jnrG;
240 j_coord_offsetH = DIM*jnrH;
242 /* load j atom coordinates */
243 gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
244 x+j_coord_offsetC,x+j_coord_offsetD,
245 x+j_coord_offsetE,x+j_coord_offsetF,
246 x+j_coord_offsetG,x+j_coord_offsetH,
247 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
248 &jy2,&jz2,&jx3,&jy3,&jz3);
250 /* Calculate displacement vector */
251 dx00 = _mm256_sub_ps(ix0,jx0);
252 dy00 = _mm256_sub_ps(iy0,jy0);
253 dz00 = _mm256_sub_ps(iz0,jz0);
254 dx11 = _mm256_sub_ps(ix1,jx1);
255 dy11 = _mm256_sub_ps(iy1,jy1);
256 dz11 = _mm256_sub_ps(iz1,jz1);
257 dx12 = _mm256_sub_ps(ix1,jx2);
258 dy12 = _mm256_sub_ps(iy1,jy2);
259 dz12 = _mm256_sub_ps(iz1,jz2);
260 dx13 = _mm256_sub_ps(ix1,jx3);
261 dy13 = _mm256_sub_ps(iy1,jy3);
262 dz13 = _mm256_sub_ps(iz1,jz3);
263 dx21 = _mm256_sub_ps(ix2,jx1);
264 dy21 = _mm256_sub_ps(iy2,jy1);
265 dz21 = _mm256_sub_ps(iz2,jz1);
266 dx22 = _mm256_sub_ps(ix2,jx2);
267 dy22 = _mm256_sub_ps(iy2,jy2);
268 dz22 = _mm256_sub_ps(iz2,jz2);
269 dx23 = _mm256_sub_ps(ix2,jx3);
270 dy23 = _mm256_sub_ps(iy2,jy3);
271 dz23 = _mm256_sub_ps(iz2,jz3);
272 dx31 = _mm256_sub_ps(ix3,jx1);
273 dy31 = _mm256_sub_ps(iy3,jy1);
274 dz31 = _mm256_sub_ps(iz3,jz1);
275 dx32 = _mm256_sub_ps(ix3,jx2);
276 dy32 = _mm256_sub_ps(iy3,jy2);
277 dz32 = _mm256_sub_ps(iz3,jz2);
278 dx33 = _mm256_sub_ps(ix3,jx3);
279 dy33 = _mm256_sub_ps(iy3,jy3);
280 dz33 = _mm256_sub_ps(iz3,jz3);
282 /* Calculate squared distance and things based on it */
283 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
284 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
285 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
286 rsq13 = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
287 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
288 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
289 rsq23 = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
290 rsq31 = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
291 rsq32 = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
292 rsq33 = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
294 rinv11 = avx256_invsqrt_f(rsq11);
295 rinv12 = avx256_invsqrt_f(rsq12);
296 rinv13 = avx256_invsqrt_f(rsq13);
297 rinv21 = avx256_invsqrt_f(rsq21);
298 rinv22 = avx256_invsqrt_f(rsq22);
299 rinv23 = avx256_invsqrt_f(rsq23);
300 rinv31 = avx256_invsqrt_f(rsq31);
301 rinv32 = avx256_invsqrt_f(rsq32);
302 rinv33 = avx256_invsqrt_f(rsq33);
304 rinvsq00 = avx256_inv_f(rsq00);
305 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
306 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
307 rinvsq13 = _mm256_mul_ps(rinv13,rinv13);
308 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
309 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
310 rinvsq23 = _mm256_mul_ps(rinv23,rinv23);
311 rinvsq31 = _mm256_mul_ps(rinv31,rinv31);
312 rinvsq32 = _mm256_mul_ps(rinv32,rinv32);
313 rinvsq33 = _mm256_mul_ps(rinv33,rinv33);
315 fjx0 = _mm256_setzero_ps();
316 fjy0 = _mm256_setzero_ps();
317 fjz0 = _mm256_setzero_ps();
318 fjx1 = _mm256_setzero_ps();
319 fjy1 = _mm256_setzero_ps();
320 fjz1 = _mm256_setzero_ps();
321 fjx2 = _mm256_setzero_ps();
322 fjy2 = _mm256_setzero_ps();
323 fjz2 = _mm256_setzero_ps();
324 fjx3 = _mm256_setzero_ps();
325 fjy3 = _mm256_setzero_ps();
326 fjz3 = _mm256_setzero_ps();
328 /**************************
329 * CALCULATE INTERACTIONS *
330 **************************/
332 /* LENNARD-JONES DISPERSION/REPULSION */
334 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
335 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
336 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
337 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
338 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
340 /* Update potential sum for this i atom from the interaction with this j atom. */
341 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
345 /* Calculate temporary vectorial force */
346 tx = _mm256_mul_ps(fscal,dx00);
347 ty = _mm256_mul_ps(fscal,dy00);
348 tz = _mm256_mul_ps(fscal,dz00);
350 /* Update vectorial force */
351 fix0 = _mm256_add_ps(fix0,tx);
352 fiy0 = _mm256_add_ps(fiy0,ty);
353 fiz0 = _mm256_add_ps(fiz0,tz);
355 fjx0 = _mm256_add_ps(fjx0,tx);
356 fjy0 = _mm256_add_ps(fjy0,ty);
357 fjz0 = _mm256_add_ps(fjz0,tz);
359 /**************************
360 * CALCULATE INTERACTIONS *
361 **************************/
363 /* REACTION-FIELD ELECTROSTATICS */
364 velec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
365 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
367 /* Update potential sum for this i atom from the interaction with this j atom. */
368 velecsum = _mm256_add_ps(velecsum,velec);
372 /* Calculate temporary vectorial force */
373 tx = _mm256_mul_ps(fscal,dx11);
374 ty = _mm256_mul_ps(fscal,dy11);
375 tz = _mm256_mul_ps(fscal,dz11);
377 /* Update vectorial force */
378 fix1 = _mm256_add_ps(fix1,tx);
379 fiy1 = _mm256_add_ps(fiy1,ty);
380 fiz1 = _mm256_add_ps(fiz1,tz);
382 fjx1 = _mm256_add_ps(fjx1,tx);
383 fjy1 = _mm256_add_ps(fjy1,ty);
384 fjz1 = _mm256_add_ps(fjz1,tz);
386 /**************************
387 * CALCULATE INTERACTIONS *
388 **************************/
390 /* REACTION-FIELD ELECTROSTATICS */
391 velec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
392 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
394 /* Update potential sum for this i atom from the interaction with this j atom. */
395 velecsum = _mm256_add_ps(velecsum,velec);
399 /* Calculate temporary vectorial force */
400 tx = _mm256_mul_ps(fscal,dx12);
401 ty = _mm256_mul_ps(fscal,dy12);
402 tz = _mm256_mul_ps(fscal,dz12);
404 /* Update vectorial force */
405 fix1 = _mm256_add_ps(fix1,tx);
406 fiy1 = _mm256_add_ps(fiy1,ty);
407 fiz1 = _mm256_add_ps(fiz1,tz);
409 fjx2 = _mm256_add_ps(fjx2,tx);
410 fjy2 = _mm256_add_ps(fjy2,ty);
411 fjz2 = _mm256_add_ps(fjz2,tz);
413 /**************************
414 * CALCULATE INTERACTIONS *
415 **************************/
417 /* REACTION-FIELD ELECTROSTATICS */
418 velec = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
419 felec = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
421 /* Update potential sum for this i atom from the interaction with this j atom. */
422 velecsum = _mm256_add_ps(velecsum,velec);
426 /* Calculate temporary vectorial force */
427 tx = _mm256_mul_ps(fscal,dx13);
428 ty = _mm256_mul_ps(fscal,dy13);
429 tz = _mm256_mul_ps(fscal,dz13);
431 /* Update vectorial force */
432 fix1 = _mm256_add_ps(fix1,tx);
433 fiy1 = _mm256_add_ps(fiy1,ty);
434 fiz1 = _mm256_add_ps(fiz1,tz);
436 fjx3 = _mm256_add_ps(fjx3,tx);
437 fjy3 = _mm256_add_ps(fjy3,ty);
438 fjz3 = _mm256_add_ps(fjz3,tz);
440 /**************************
441 * CALCULATE INTERACTIONS *
442 **************************/
444 /* REACTION-FIELD ELECTROSTATICS */
445 velec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
446 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
448 /* Update potential sum for this i atom from the interaction with this j atom. */
449 velecsum = _mm256_add_ps(velecsum,velec);
453 /* Calculate temporary vectorial force */
454 tx = _mm256_mul_ps(fscal,dx21);
455 ty = _mm256_mul_ps(fscal,dy21);
456 tz = _mm256_mul_ps(fscal,dz21);
458 /* Update vectorial force */
459 fix2 = _mm256_add_ps(fix2,tx);
460 fiy2 = _mm256_add_ps(fiy2,ty);
461 fiz2 = _mm256_add_ps(fiz2,tz);
463 fjx1 = _mm256_add_ps(fjx1,tx);
464 fjy1 = _mm256_add_ps(fjy1,ty);
465 fjz1 = _mm256_add_ps(fjz1,tz);
467 /**************************
468 * CALCULATE INTERACTIONS *
469 **************************/
471 /* REACTION-FIELD ELECTROSTATICS */
472 velec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
473 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
475 /* Update potential sum for this i atom from the interaction with this j atom. */
476 velecsum = _mm256_add_ps(velecsum,velec);
480 /* Calculate temporary vectorial force */
481 tx = _mm256_mul_ps(fscal,dx22);
482 ty = _mm256_mul_ps(fscal,dy22);
483 tz = _mm256_mul_ps(fscal,dz22);
485 /* Update vectorial force */
486 fix2 = _mm256_add_ps(fix2,tx);
487 fiy2 = _mm256_add_ps(fiy2,ty);
488 fiz2 = _mm256_add_ps(fiz2,tz);
490 fjx2 = _mm256_add_ps(fjx2,tx);
491 fjy2 = _mm256_add_ps(fjy2,ty);
492 fjz2 = _mm256_add_ps(fjz2,tz);
494 /**************************
495 * CALCULATE INTERACTIONS *
496 **************************/
498 /* REACTION-FIELD ELECTROSTATICS */
499 velec = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
500 felec = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
502 /* Update potential sum for this i atom from the interaction with this j atom. */
503 velecsum = _mm256_add_ps(velecsum,velec);
507 /* Calculate temporary vectorial force */
508 tx = _mm256_mul_ps(fscal,dx23);
509 ty = _mm256_mul_ps(fscal,dy23);
510 tz = _mm256_mul_ps(fscal,dz23);
512 /* Update vectorial force */
513 fix2 = _mm256_add_ps(fix2,tx);
514 fiy2 = _mm256_add_ps(fiy2,ty);
515 fiz2 = _mm256_add_ps(fiz2,tz);
517 fjx3 = _mm256_add_ps(fjx3,tx);
518 fjy3 = _mm256_add_ps(fjy3,ty);
519 fjz3 = _mm256_add_ps(fjz3,tz);
521 /**************************
522 * CALCULATE INTERACTIONS *
523 **************************/
525 /* REACTION-FIELD ELECTROSTATICS */
526 velec = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
527 felec = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
529 /* Update potential sum for this i atom from the interaction with this j atom. */
530 velecsum = _mm256_add_ps(velecsum,velec);
534 /* Calculate temporary vectorial force */
535 tx = _mm256_mul_ps(fscal,dx31);
536 ty = _mm256_mul_ps(fscal,dy31);
537 tz = _mm256_mul_ps(fscal,dz31);
539 /* Update vectorial force */
540 fix3 = _mm256_add_ps(fix3,tx);
541 fiy3 = _mm256_add_ps(fiy3,ty);
542 fiz3 = _mm256_add_ps(fiz3,tz);
544 fjx1 = _mm256_add_ps(fjx1,tx);
545 fjy1 = _mm256_add_ps(fjy1,ty);
546 fjz1 = _mm256_add_ps(fjz1,tz);
548 /**************************
549 * CALCULATE INTERACTIONS *
550 **************************/
552 /* REACTION-FIELD ELECTROSTATICS */
553 velec = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
554 felec = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
556 /* Update potential sum for this i atom from the interaction with this j atom. */
557 velecsum = _mm256_add_ps(velecsum,velec);
561 /* Calculate temporary vectorial force */
562 tx = _mm256_mul_ps(fscal,dx32);
563 ty = _mm256_mul_ps(fscal,dy32);
564 tz = _mm256_mul_ps(fscal,dz32);
566 /* Update vectorial force */
567 fix3 = _mm256_add_ps(fix3,tx);
568 fiy3 = _mm256_add_ps(fiy3,ty);
569 fiz3 = _mm256_add_ps(fiz3,tz);
571 fjx2 = _mm256_add_ps(fjx2,tx);
572 fjy2 = _mm256_add_ps(fjy2,ty);
573 fjz2 = _mm256_add_ps(fjz2,tz);
575 /**************************
576 * CALCULATE INTERACTIONS *
577 **************************/
579 /* REACTION-FIELD ELECTROSTATICS */
580 velec = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
581 felec = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
583 /* Update potential sum for this i atom from the interaction with this j atom. */
584 velecsum = _mm256_add_ps(velecsum,velec);
588 /* Calculate temporary vectorial force */
589 tx = _mm256_mul_ps(fscal,dx33);
590 ty = _mm256_mul_ps(fscal,dy33);
591 tz = _mm256_mul_ps(fscal,dz33);
593 /* Update vectorial force */
594 fix3 = _mm256_add_ps(fix3,tx);
595 fiy3 = _mm256_add_ps(fiy3,ty);
596 fiz3 = _mm256_add_ps(fiz3,tz);
598 fjx3 = _mm256_add_ps(fjx3,tx);
599 fjy3 = _mm256_add_ps(fjy3,ty);
600 fjz3 = _mm256_add_ps(fjz3,tz);
602 fjptrA = f+j_coord_offsetA;
603 fjptrB = f+j_coord_offsetB;
604 fjptrC = f+j_coord_offsetC;
605 fjptrD = f+j_coord_offsetD;
606 fjptrE = f+j_coord_offsetE;
607 fjptrF = f+j_coord_offsetF;
608 fjptrG = f+j_coord_offsetG;
609 fjptrH = f+j_coord_offsetH;
611 gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
612 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
613 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
615 /* Inner loop uses 323 flops */
621 /* Get j neighbor index, and coordinate index */
622 jnrlistA = jjnr[jidx];
623 jnrlistB = jjnr[jidx+1];
624 jnrlistC = jjnr[jidx+2];
625 jnrlistD = jjnr[jidx+3];
626 jnrlistE = jjnr[jidx+4];
627 jnrlistF = jjnr[jidx+5];
628 jnrlistG = jjnr[jidx+6];
629 jnrlistH = jjnr[jidx+7];
630 /* Sign of each element will be negative for non-real atoms.
631 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
632 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
634 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
635 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
637 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
638 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
639 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
640 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
641 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
642 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
643 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
644 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
645 j_coord_offsetA = DIM*jnrA;
646 j_coord_offsetB = DIM*jnrB;
647 j_coord_offsetC = DIM*jnrC;
648 j_coord_offsetD = DIM*jnrD;
649 j_coord_offsetE = DIM*jnrE;
650 j_coord_offsetF = DIM*jnrF;
651 j_coord_offsetG = DIM*jnrG;
652 j_coord_offsetH = DIM*jnrH;
654 /* load j atom coordinates */
655 gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
656 x+j_coord_offsetC,x+j_coord_offsetD,
657 x+j_coord_offsetE,x+j_coord_offsetF,
658 x+j_coord_offsetG,x+j_coord_offsetH,
659 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
660 &jy2,&jz2,&jx3,&jy3,&jz3);
662 /* Calculate displacement vector */
663 dx00 = _mm256_sub_ps(ix0,jx0);
664 dy00 = _mm256_sub_ps(iy0,jy0);
665 dz00 = _mm256_sub_ps(iz0,jz0);
666 dx11 = _mm256_sub_ps(ix1,jx1);
667 dy11 = _mm256_sub_ps(iy1,jy1);
668 dz11 = _mm256_sub_ps(iz1,jz1);
669 dx12 = _mm256_sub_ps(ix1,jx2);
670 dy12 = _mm256_sub_ps(iy1,jy2);
671 dz12 = _mm256_sub_ps(iz1,jz2);
672 dx13 = _mm256_sub_ps(ix1,jx3);
673 dy13 = _mm256_sub_ps(iy1,jy3);
674 dz13 = _mm256_sub_ps(iz1,jz3);
675 dx21 = _mm256_sub_ps(ix2,jx1);
676 dy21 = _mm256_sub_ps(iy2,jy1);
677 dz21 = _mm256_sub_ps(iz2,jz1);
678 dx22 = _mm256_sub_ps(ix2,jx2);
679 dy22 = _mm256_sub_ps(iy2,jy2);
680 dz22 = _mm256_sub_ps(iz2,jz2);
681 dx23 = _mm256_sub_ps(ix2,jx3);
682 dy23 = _mm256_sub_ps(iy2,jy3);
683 dz23 = _mm256_sub_ps(iz2,jz3);
684 dx31 = _mm256_sub_ps(ix3,jx1);
685 dy31 = _mm256_sub_ps(iy3,jy1);
686 dz31 = _mm256_sub_ps(iz3,jz1);
687 dx32 = _mm256_sub_ps(ix3,jx2);
688 dy32 = _mm256_sub_ps(iy3,jy2);
689 dz32 = _mm256_sub_ps(iz3,jz2);
690 dx33 = _mm256_sub_ps(ix3,jx3);
691 dy33 = _mm256_sub_ps(iy3,jy3);
692 dz33 = _mm256_sub_ps(iz3,jz3);
694 /* Calculate squared distance and things based on it */
695 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
696 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
697 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
698 rsq13 = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
699 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
700 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
701 rsq23 = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
702 rsq31 = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
703 rsq32 = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
704 rsq33 = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
706 rinv11 = avx256_invsqrt_f(rsq11);
707 rinv12 = avx256_invsqrt_f(rsq12);
708 rinv13 = avx256_invsqrt_f(rsq13);
709 rinv21 = avx256_invsqrt_f(rsq21);
710 rinv22 = avx256_invsqrt_f(rsq22);
711 rinv23 = avx256_invsqrt_f(rsq23);
712 rinv31 = avx256_invsqrt_f(rsq31);
713 rinv32 = avx256_invsqrt_f(rsq32);
714 rinv33 = avx256_invsqrt_f(rsq33);
716 rinvsq00 = avx256_inv_f(rsq00);
717 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
718 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
719 rinvsq13 = _mm256_mul_ps(rinv13,rinv13);
720 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
721 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
722 rinvsq23 = _mm256_mul_ps(rinv23,rinv23);
723 rinvsq31 = _mm256_mul_ps(rinv31,rinv31);
724 rinvsq32 = _mm256_mul_ps(rinv32,rinv32);
725 rinvsq33 = _mm256_mul_ps(rinv33,rinv33);
727 fjx0 = _mm256_setzero_ps();
728 fjy0 = _mm256_setzero_ps();
729 fjz0 = _mm256_setzero_ps();
730 fjx1 = _mm256_setzero_ps();
731 fjy1 = _mm256_setzero_ps();
732 fjz1 = _mm256_setzero_ps();
733 fjx2 = _mm256_setzero_ps();
734 fjy2 = _mm256_setzero_ps();
735 fjz2 = _mm256_setzero_ps();
736 fjx3 = _mm256_setzero_ps();
737 fjy3 = _mm256_setzero_ps();
738 fjz3 = _mm256_setzero_ps();
740 /**************************
741 * CALCULATE INTERACTIONS *
742 **************************/
744 /* LENNARD-JONES DISPERSION/REPULSION */
746 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
747 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
748 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
749 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
750 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
752 /* Update potential sum for this i atom from the interaction with this j atom. */
753 vvdw = _mm256_andnot_ps(dummy_mask,vvdw);
754 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
758 fscal = _mm256_andnot_ps(dummy_mask,fscal);
760 /* Calculate temporary vectorial force */
761 tx = _mm256_mul_ps(fscal,dx00);
762 ty = _mm256_mul_ps(fscal,dy00);
763 tz = _mm256_mul_ps(fscal,dz00);
765 /* Update vectorial force */
766 fix0 = _mm256_add_ps(fix0,tx);
767 fiy0 = _mm256_add_ps(fiy0,ty);
768 fiz0 = _mm256_add_ps(fiz0,tz);
770 fjx0 = _mm256_add_ps(fjx0,tx);
771 fjy0 = _mm256_add_ps(fjy0,ty);
772 fjz0 = _mm256_add_ps(fjz0,tz);
774 /**************************
775 * CALCULATE INTERACTIONS *
776 **************************/
778 /* REACTION-FIELD ELECTROSTATICS */
779 velec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
780 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
782 /* Update potential sum for this i atom from the interaction with this j atom. */
783 velec = _mm256_andnot_ps(dummy_mask,velec);
784 velecsum = _mm256_add_ps(velecsum,velec);
788 fscal = _mm256_andnot_ps(dummy_mask,fscal);
790 /* Calculate temporary vectorial force */
791 tx = _mm256_mul_ps(fscal,dx11);
792 ty = _mm256_mul_ps(fscal,dy11);
793 tz = _mm256_mul_ps(fscal,dz11);
795 /* Update vectorial force */
796 fix1 = _mm256_add_ps(fix1,tx);
797 fiy1 = _mm256_add_ps(fiy1,ty);
798 fiz1 = _mm256_add_ps(fiz1,tz);
800 fjx1 = _mm256_add_ps(fjx1,tx);
801 fjy1 = _mm256_add_ps(fjy1,ty);
802 fjz1 = _mm256_add_ps(fjz1,tz);
804 /**************************
805 * CALCULATE INTERACTIONS *
806 **************************/
808 /* REACTION-FIELD ELECTROSTATICS */
809 velec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
810 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
812 /* Update potential sum for this i atom from the interaction with this j atom. */
813 velec = _mm256_andnot_ps(dummy_mask,velec);
814 velecsum = _mm256_add_ps(velecsum,velec);
818 fscal = _mm256_andnot_ps(dummy_mask,fscal);
820 /* Calculate temporary vectorial force */
821 tx = _mm256_mul_ps(fscal,dx12);
822 ty = _mm256_mul_ps(fscal,dy12);
823 tz = _mm256_mul_ps(fscal,dz12);
825 /* Update vectorial force */
826 fix1 = _mm256_add_ps(fix1,tx);
827 fiy1 = _mm256_add_ps(fiy1,ty);
828 fiz1 = _mm256_add_ps(fiz1,tz);
830 fjx2 = _mm256_add_ps(fjx2,tx);
831 fjy2 = _mm256_add_ps(fjy2,ty);
832 fjz2 = _mm256_add_ps(fjz2,tz);
834 /**************************
835 * CALCULATE INTERACTIONS *
836 **************************/
838 /* REACTION-FIELD ELECTROSTATICS */
839 velec = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
840 felec = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
842 /* Update potential sum for this i atom from the interaction with this j atom. */
843 velec = _mm256_andnot_ps(dummy_mask,velec);
844 velecsum = _mm256_add_ps(velecsum,velec);
848 fscal = _mm256_andnot_ps(dummy_mask,fscal);
850 /* Calculate temporary vectorial force */
851 tx = _mm256_mul_ps(fscal,dx13);
852 ty = _mm256_mul_ps(fscal,dy13);
853 tz = _mm256_mul_ps(fscal,dz13);
855 /* Update vectorial force */
856 fix1 = _mm256_add_ps(fix1,tx);
857 fiy1 = _mm256_add_ps(fiy1,ty);
858 fiz1 = _mm256_add_ps(fiz1,tz);
860 fjx3 = _mm256_add_ps(fjx3,tx);
861 fjy3 = _mm256_add_ps(fjy3,ty);
862 fjz3 = _mm256_add_ps(fjz3,tz);
864 /**************************
865 * CALCULATE INTERACTIONS *
866 **************************/
868 /* REACTION-FIELD ELECTROSTATICS */
869 velec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
870 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
872 /* Update potential sum for this i atom from the interaction with this j atom. */
873 velec = _mm256_andnot_ps(dummy_mask,velec);
874 velecsum = _mm256_add_ps(velecsum,velec);
878 fscal = _mm256_andnot_ps(dummy_mask,fscal);
880 /* Calculate temporary vectorial force */
881 tx = _mm256_mul_ps(fscal,dx21);
882 ty = _mm256_mul_ps(fscal,dy21);
883 tz = _mm256_mul_ps(fscal,dz21);
885 /* Update vectorial force */
886 fix2 = _mm256_add_ps(fix2,tx);
887 fiy2 = _mm256_add_ps(fiy2,ty);
888 fiz2 = _mm256_add_ps(fiz2,tz);
890 fjx1 = _mm256_add_ps(fjx1,tx);
891 fjy1 = _mm256_add_ps(fjy1,ty);
892 fjz1 = _mm256_add_ps(fjz1,tz);
894 /**************************
895 * CALCULATE INTERACTIONS *
896 **************************/
898 /* REACTION-FIELD ELECTROSTATICS */
899 velec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
900 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
902 /* Update potential sum for this i atom from the interaction with this j atom. */
903 velec = _mm256_andnot_ps(dummy_mask,velec);
904 velecsum = _mm256_add_ps(velecsum,velec);
908 fscal = _mm256_andnot_ps(dummy_mask,fscal);
910 /* Calculate temporary vectorial force */
911 tx = _mm256_mul_ps(fscal,dx22);
912 ty = _mm256_mul_ps(fscal,dy22);
913 tz = _mm256_mul_ps(fscal,dz22);
915 /* Update vectorial force */
916 fix2 = _mm256_add_ps(fix2,tx);
917 fiy2 = _mm256_add_ps(fiy2,ty);
918 fiz2 = _mm256_add_ps(fiz2,tz);
920 fjx2 = _mm256_add_ps(fjx2,tx);
921 fjy2 = _mm256_add_ps(fjy2,ty);
922 fjz2 = _mm256_add_ps(fjz2,tz);
924 /**************************
925 * CALCULATE INTERACTIONS *
926 **************************/
928 /* REACTION-FIELD ELECTROSTATICS */
929 velec = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
930 felec = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
932 /* Update potential sum for this i atom from the interaction with this j atom. */
933 velec = _mm256_andnot_ps(dummy_mask,velec);
934 velecsum = _mm256_add_ps(velecsum,velec);
938 fscal = _mm256_andnot_ps(dummy_mask,fscal);
940 /* Calculate temporary vectorial force */
941 tx = _mm256_mul_ps(fscal,dx23);
942 ty = _mm256_mul_ps(fscal,dy23);
943 tz = _mm256_mul_ps(fscal,dz23);
945 /* Update vectorial force */
946 fix2 = _mm256_add_ps(fix2,tx);
947 fiy2 = _mm256_add_ps(fiy2,ty);
948 fiz2 = _mm256_add_ps(fiz2,tz);
950 fjx3 = _mm256_add_ps(fjx3,tx);
951 fjy3 = _mm256_add_ps(fjy3,ty);
952 fjz3 = _mm256_add_ps(fjz3,tz);
954 /**************************
955 * CALCULATE INTERACTIONS *
956 **************************/
958 /* REACTION-FIELD ELECTROSTATICS */
959 velec = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
960 felec = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
962 /* Update potential sum for this i atom from the interaction with this j atom. */
963 velec = _mm256_andnot_ps(dummy_mask,velec);
964 velecsum = _mm256_add_ps(velecsum,velec);
968 fscal = _mm256_andnot_ps(dummy_mask,fscal);
970 /* Calculate temporary vectorial force */
971 tx = _mm256_mul_ps(fscal,dx31);
972 ty = _mm256_mul_ps(fscal,dy31);
973 tz = _mm256_mul_ps(fscal,dz31);
975 /* Update vectorial force */
976 fix3 = _mm256_add_ps(fix3,tx);
977 fiy3 = _mm256_add_ps(fiy3,ty);
978 fiz3 = _mm256_add_ps(fiz3,tz);
980 fjx1 = _mm256_add_ps(fjx1,tx);
981 fjy1 = _mm256_add_ps(fjy1,ty);
982 fjz1 = _mm256_add_ps(fjz1,tz);
984 /**************************
985 * CALCULATE INTERACTIONS *
986 **************************/
988 /* REACTION-FIELD ELECTROSTATICS */
989 velec = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
990 felec = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
992 /* Update potential sum for this i atom from the interaction with this j atom. */
993 velec = _mm256_andnot_ps(dummy_mask,velec);
994 velecsum = _mm256_add_ps(velecsum,velec);
998 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1000 /* Calculate temporary vectorial force */
1001 tx = _mm256_mul_ps(fscal,dx32);
1002 ty = _mm256_mul_ps(fscal,dy32);
1003 tz = _mm256_mul_ps(fscal,dz32);
1005 /* Update vectorial force */
1006 fix3 = _mm256_add_ps(fix3,tx);
1007 fiy3 = _mm256_add_ps(fiy3,ty);
1008 fiz3 = _mm256_add_ps(fiz3,tz);
1010 fjx2 = _mm256_add_ps(fjx2,tx);
1011 fjy2 = _mm256_add_ps(fjy2,ty);
1012 fjz2 = _mm256_add_ps(fjz2,tz);
1014 /**************************
1015 * CALCULATE INTERACTIONS *
1016 **************************/
1018 /* REACTION-FIELD ELECTROSTATICS */
1019 velec = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
1020 felec = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1022 /* Update potential sum for this i atom from the interaction with this j atom. */
1023 velec = _mm256_andnot_ps(dummy_mask,velec);
1024 velecsum = _mm256_add_ps(velecsum,velec);
1028 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1030 /* Calculate temporary vectorial force */
1031 tx = _mm256_mul_ps(fscal,dx33);
1032 ty = _mm256_mul_ps(fscal,dy33);
1033 tz = _mm256_mul_ps(fscal,dz33);
1035 /* Update vectorial force */
1036 fix3 = _mm256_add_ps(fix3,tx);
1037 fiy3 = _mm256_add_ps(fiy3,ty);
1038 fiz3 = _mm256_add_ps(fiz3,tz);
1040 fjx3 = _mm256_add_ps(fjx3,tx);
1041 fjy3 = _mm256_add_ps(fjy3,ty);
1042 fjz3 = _mm256_add_ps(fjz3,tz);
1044 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1045 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1046 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1047 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1048 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1049 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1050 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1051 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1053 gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1054 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1055 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1057 /* Inner loop uses 323 flops */
1060 /* End of innermost loop */
1062 gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1063 f+i_coord_offset,fshift+i_shift_offset);
1066 /* Update potential energies */
1067 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1068 gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1070 /* Increment number of inner iterations */
1071 inneriter += j_index_end - j_index_start;
1073 /* Outer loop uses 26 flops */
1076 /* Increment number of outer iterations */
1079 /* Update outer/inner flops */
1081 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*323);
1084 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_single
1085 * Electrostatics interaction: ReactionField
1086 * VdW interaction: LennardJones
1087 * Geometry: Water4-Water4
1088 * Calculate force/pot: Force
1091 nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_single
1092 (t_nblist * gmx_restrict nlist,
1093 rvec * gmx_restrict xx,
1094 rvec * gmx_restrict ff,
1095 struct t_forcerec * gmx_restrict fr,
1096 t_mdatoms * gmx_restrict mdatoms,
1097 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1098 t_nrnb * gmx_restrict nrnb)
1100 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1101 * just 0 for non-waters.
1102 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1103 * jnr indices corresponding to data put in the four positions in the SIMD register.
1105 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1106 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1107 int jnrA,jnrB,jnrC,jnrD;
1108 int jnrE,jnrF,jnrG,jnrH;
1109 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1110 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1111 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1112 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1113 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1114 real rcutoff_scalar;
1115 real *shiftvec,*fshift,*x,*f;
1116 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1117 real scratch[4*DIM];
1118 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1119 real * vdwioffsetptr0;
1120 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1121 real * vdwioffsetptr1;
1122 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1123 real * vdwioffsetptr2;
1124 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1125 real * vdwioffsetptr3;
1126 __m256 ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1127 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1128 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1129 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1130 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1131 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1132 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1133 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1134 __m256 jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1135 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1136 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1137 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1138 __m256 dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1139 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1140 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1141 __m256 dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1142 __m256 dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1143 __m256 dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1144 __m256 dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1145 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
1148 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1151 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
1152 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
1153 __m256 dummy_mask,cutoff_mask;
1154 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1155 __m256 one = _mm256_set1_ps(1.0);
1156 __m256 two = _mm256_set1_ps(2.0);
1162 jindex = nlist->jindex;
1164 shiftidx = nlist->shift;
1166 shiftvec = fr->shift_vec[0];
1167 fshift = fr->fshift[0];
1168 facel = _mm256_set1_ps(fr->ic->epsfac);
1169 charge = mdatoms->chargeA;
1170 krf = _mm256_set1_ps(fr->ic->k_rf);
1171 krf2 = _mm256_set1_ps(fr->ic->k_rf*2.0);
1172 crf = _mm256_set1_ps(fr->ic->c_rf);
1173 nvdwtype = fr->ntype;
1174 vdwparam = fr->nbfp;
1175 vdwtype = mdatoms->typeA;
1177 /* Setup water-specific parameters */
1178 inr = nlist->iinr[0];
1179 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1180 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1181 iq3 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1182 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
1184 jq1 = _mm256_set1_ps(charge[inr+1]);
1185 jq2 = _mm256_set1_ps(charge[inr+2]);
1186 jq3 = _mm256_set1_ps(charge[inr+3]);
1187 vdwjidx0A = 2*vdwtype[inr+0];
1188 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1189 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1190 qq11 = _mm256_mul_ps(iq1,jq1);
1191 qq12 = _mm256_mul_ps(iq1,jq2);
1192 qq13 = _mm256_mul_ps(iq1,jq3);
1193 qq21 = _mm256_mul_ps(iq2,jq1);
1194 qq22 = _mm256_mul_ps(iq2,jq2);
1195 qq23 = _mm256_mul_ps(iq2,jq3);
1196 qq31 = _mm256_mul_ps(iq3,jq1);
1197 qq32 = _mm256_mul_ps(iq3,jq2);
1198 qq33 = _mm256_mul_ps(iq3,jq3);
1200 /* Avoid stupid compiler warnings */
1201 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1202 j_coord_offsetA = 0;
1203 j_coord_offsetB = 0;
1204 j_coord_offsetC = 0;
1205 j_coord_offsetD = 0;
1206 j_coord_offsetE = 0;
1207 j_coord_offsetF = 0;
1208 j_coord_offsetG = 0;
1209 j_coord_offsetH = 0;
1214 for(iidx=0;iidx<4*DIM;iidx++)
1216 scratch[iidx] = 0.0;
1219 /* Start outer loop over neighborlists */
1220 for(iidx=0; iidx<nri; iidx++)
1222 /* Load shift vector for this list */
1223 i_shift_offset = DIM*shiftidx[iidx];
1225 /* Load limits for loop over neighbors */
1226 j_index_start = jindex[iidx];
1227 j_index_end = jindex[iidx+1];
1229 /* Get outer coordinate index */
1231 i_coord_offset = DIM*inr;
1233 /* Load i particle coords and add shift vector */
1234 gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1235 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1237 fix0 = _mm256_setzero_ps();
1238 fiy0 = _mm256_setzero_ps();
1239 fiz0 = _mm256_setzero_ps();
1240 fix1 = _mm256_setzero_ps();
1241 fiy1 = _mm256_setzero_ps();
1242 fiz1 = _mm256_setzero_ps();
1243 fix2 = _mm256_setzero_ps();
1244 fiy2 = _mm256_setzero_ps();
1245 fiz2 = _mm256_setzero_ps();
1246 fix3 = _mm256_setzero_ps();
1247 fiy3 = _mm256_setzero_ps();
1248 fiz3 = _mm256_setzero_ps();
1250 /* Start inner kernel loop */
1251 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1254 /* Get j neighbor index, and coordinate index */
1256 jnrB = jjnr[jidx+1];
1257 jnrC = jjnr[jidx+2];
1258 jnrD = jjnr[jidx+3];
1259 jnrE = jjnr[jidx+4];
1260 jnrF = jjnr[jidx+5];
1261 jnrG = jjnr[jidx+6];
1262 jnrH = jjnr[jidx+7];
1263 j_coord_offsetA = DIM*jnrA;
1264 j_coord_offsetB = DIM*jnrB;
1265 j_coord_offsetC = DIM*jnrC;
1266 j_coord_offsetD = DIM*jnrD;
1267 j_coord_offsetE = DIM*jnrE;
1268 j_coord_offsetF = DIM*jnrF;
1269 j_coord_offsetG = DIM*jnrG;
1270 j_coord_offsetH = DIM*jnrH;
1272 /* load j atom coordinates */
1273 gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1274 x+j_coord_offsetC,x+j_coord_offsetD,
1275 x+j_coord_offsetE,x+j_coord_offsetF,
1276 x+j_coord_offsetG,x+j_coord_offsetH,
1277 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1278 &jy2,&jz2,&jx3,&jy3,&jz3);
1280 /* Calculate displacement vector */
1281 dx00 = _mm256_sub_ps(ix0,jx0);
1282 dy00 = _mm256_sub_ps(iy0,jy0);
1283 dz00 = _mm256_sub_ps(iz0,jz0);
1284 dx11 = _mm256_sub_ps(ix1,jx1);
1285 dy11 = _mm256_sub_ps(iy1,jy1);
1286 dz11 = _mm256_sub_ps(iz1,jz1);
1287 dx12 = _mm256_sub_ps(ix1,jx2);
1288 dy12 = _mm256_sub_ps(iy1,jy2);
1289 dz12 = _mm256_sub_ps(iz1,jz2);
1290 dx13 = _mm256_sub_ps(ix1,jx3);
1291 dy13 = _mm256_sub_ps(iy1,jy3);
1292 dz13 = _mm256_sub_ps(iz1,jz3);
1293 dx21 = _mm256_sub_ps(ix2,jx1);
1294 dy21 = _mm256_sub_ps(iy2,jy1);
1295 dz21 = _mm256_sub_ps(iz2,jz1);
1296 dx22 = _mm256_sub_ps(ix2,jx2);
1297 dy22 = _mm256_sub_ps(iy2,jy2);
1298 dz22 = _mm256_sub_ps(iz2,jz2);
1299 dx23 = _mm256_sub_ps(ix2,jx3);
1300 dy23 = _mm256_sub_ps(iy2,jy3);
1301 dz23 = _mm256_sub_ps(iz2,jz3);
1302 dx31 = _mm256_sub_ps(ix3,jx1);
1303 dy31 = _mm256_sub_ps(iy3,jy1);
1304 dz31 = _mm256_sub_ps(iz3,jz1);
1305 dx32 = _mm256_sub_ps(ix3,jx2);
1306 dy32 = _mm256_sub_ps(iy3,jy2);
1307 dz32 = _mm256_sub_ps(iz3,jz2);
1308 dx33 = _mm256_sub_ps(ix3,jx3);
1309 dy33 = _mm256_sub_ps(iy3,jy3);
1310 dz33 = _mm256_sub_ps(iz3,jz3);
1312 /* Calculate squared distance and things based on it */
1313 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1314 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1315 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1316 rsq13 = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1317 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1318 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1319 rsq23 = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1320 rsq31 = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1321 rsq32 = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1322 rsq33 = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1324 rinv11 = avx256_invsqrt_f(rsq11);
1325 rinv12 = avx256_invsqrt_f(rsq12);
1326 rinv13 = avx256_invsqrt_f(rsq13);
1327 rinv21 = avx256_invsqrt_f(rsq21);
1328 rinv22 = avx256_invsqrt_f(rsq22);
1329 rinv23 = avx256_invsqrt_f(rsq23);
1330 rinv31 = avx256_invsqrt_f(rsq31);
1331 rinv32 = avx256_invsqrt_f(rsq32);
1332 rinv33 = avx256_invsqrt_f(rsq33);
1334 rinvsq00 = avx256_inv_f(rsq00);
1335 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1336 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1337 rinvsq13 = _mm256_mul_ps(rinv13,rinv13);
1338 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1339 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1340 rinvsq23 = _mm256_mul_ps(rinv23,rinv23);
1341 rinvsq31 = _mm256_mul_ps(rinv31,rinv31);
1342 rinvsq32 = _mm256_mul_ps(rinv32,rinv32);
1343 rinvsq33 = _mm256_mul_ps(rinv33,rinv33);
1345 fjx0 = _mm256_setzero_ps();
1346 fjy0 = _mm256_setzero_ps();
1347 fjz0 = _mm256_setzero_ps();
1348 fjx1 = _mm256_setzero_ps();
1349 fjy1 = _mm256_setzero_ps();
1350 fjz1 = _mm256_setzero_ps();
1351 fjx2 = _mm256_setzero_ps();
1352 fjy2 = _mm256_setzero_ps();
1353 fjz2 = _mm256_setzero_ps();
1354 fjx3 = _mm256_setzero_ps();
1355 fjy3 = _mm256_setzero_ps();
1356 fjz3 = _mm256_setzero_ps();
1358 /**************************
1359 * CALCULATE INTERACTIONS *
1360 **************************/
1362 /* LENNARD-JONES DISPERSION/REPULSION */
1364 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1365 fvdw = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1369 /* Calculate temporary vectorial force */
1370 tx = _mm256_mul_ps(fscal,dx00);
1371 ty = _mm256_mul_ps(fscal,dy00);
1372 tz = _mm256_mul_ps(fscal,dz00);
1374 /* Update vectorial force */
1375 fix0 = _mm256_add_ps(fix0,tx);
1376 fiy0 = _mm256_add_ps(fiy0,ty);
1377 fiz0 = _mm256_add_ps(fiz0,tz);
1379 fjx0 = _mm256_add_ps(fjx0,tx);
1380 fjy0 = _mm256_add_ps(fjy0,ty);
1381 fjz0 = _mm256_add_ps(fjz0,tz);
1383 /**************************
1384 * CALCULATE INTERACTIONS *
1385 **************************/
1387 /* REACTION-FIELD ELECTROSTATICS */
1388 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1392 /* Calculate temporary vectorial force */
1393 tx = _mm256_mul_ps(fscal,dx11);
1394 ty = _mm256_mul_ps(fscal,dy11);
1395 tz = _mm256_mul_ps(fscal,dz11);
1397 /* Update vectorial force */
1398 fix1 = _mm256_add_ps(fix1,tx);
1399 fiy1 = _mm256_add_ps(fiy1,ty);
1400 fiz1 = _mm256_add_ps(fiz1,tz);
1402 fjx1 = _mm256_add_ps(fjx1,tx);
1403 fjy1 = _mm256_add_ps(fjy1,ty);
1404 fjz1 = _mm256_add_ps(fjz1,tz);
1406 /**************************
1407 * CALCULATE INTERACTIONS *
1408 **************************/
1410 /* REACTION-FIELD ELECTROSTATICS */
1411 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1415 /* Calculate temporary vectorial force */
1416 tx = _mm256_mul_ps(fscal,dx12);
1417 ty = _mm256_mul_ps(fscal,dy12);
1418 tz = _mm256_mul_ps(fscal,dz12);
1420 /* Update vectorial force */
1421 fix1 = _mm256_add_ps(fix1,tx);
1422 fiy1 = _mm256_add_ps(fiy1,ty);
1423 fiz1 = _mm256_add_ps(fiz1,tz);
1425 fjx2 = _mm256_add_ps(fjx2,tx);
1426 fjy2 = _mm256_add_ps(fjy2,ty);
1427 fjz2 = _mm256_add_ps(fjz2,tz);
1429 /**************************
1430 * CALCULATE INTERACTIONS *
1431 **************************/
1433 /* REACTION-FIELD ELECTROSTATICS */
1434 felec = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1438 /* Calculate temporary vectorial force */
1439 tx = _mm256_mul_ps(fscal,dx13);
1440 ty = _mm256_mul_ps(fscal,dy13);
1441 tz = _mm256_mul_ps(fscal,dz13);
1443 /* Update vectorial force */
1444 fix1 = _mm256_add_ps(fix1,tx);
1445 fiy1 = _mm256_add_ps(fiy1,ty);
1446 fiz1 = _mm256_add_ps(fiz1,tz);
1448 fjx3 = _mm256_add_ps(fjx3,tx);
1449 fjy3 = _mm256_add_ps(fjy3,ty);
1450 fjz3 = _mm256_add_ps(fjz3,tz);
1452 /**************************
1453 * CALCULATE INTERACTIONS *
1454 **************************/
1456 /* REACTION-FIELD ELECTROSTATICS */
1457 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1461 /* Calculate temporary vectorial force */
1462 tx = _mm256_mul_ps(fscal,dx21);
1463 ty = _mm256_mul_ps(fscal,dy21);
1464 tz = _mm256_mul_ps(fscal,dz21);
1466 /* Update vectorial force */
1467 fix2 = _mm256_add_ps(fix2,tx);
1468 fiy2 = _mm256_add_ps(fiy2,ty);
1469 fiz2 = _mm256_add_ps(fiz2,tz);
1471 fjx1 = _mm256_add_ps(fjx1,tx);
1472 fjy1 = _mm256_add_ps(fjy1,ty);
1473 fjz1 = _mm256_add_ps(fjz1,tz);
1475 /**************************
1476 * CALCULATE INTERACTIONS *
1477 **************************/
1479 /* REACTION-FIELD ELECTROSTATICS */
1480 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1484 /* Calculate temporary vectorial force */
1485 tx = _mm256_mul_ps(fscal,dx22);
1486 ty = _mm256_mul_ps(fscal,dy22);
1487 tz = _mm256_mul_ps(fscal,dz22);
1489 /* Update vectorial force */
1490 fix2 = _mm256_add_ps(fix2,tx);
1491 fiy2 = _mm256_add_ps(fiy2,ty);
1492 fiz2 = _mm256_add_ps(fiz2,tz);
1494 fjx2 = _mm256_add_ps(fjx2,tx);
1495 fjy2 = _mm256_add_ps(fjy2,ty);
1496 fjz2 = _mm256_add_ps(fjz2,tz);
1498 /**************************
1499 * CALCULATE INTERACTIONS *
1500 **************************/
1502 /* REACTION-FIELD ELECTROSTATICS */
1503 felec = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1507 /* Calculate temporary vectorial force */
1508 tx = _mm256_mul_ps(fscal,dx23);
1509 ty = _mm256_mul_ps(fscal,dy23);
1510 tz = _mm256_mul_ps(fscal,dz23);
1512 /* Update vectorial force */
1513 fix2 = _mm256_add_ps(fix2,tx);
1514 fiy2 = _mm256_add_ps(fiy2,ty);
1515 fiz2 = _mm256_add_ps(fiz2,tz);
1517 fjx3 = _mm256_add_ps(fjx3,tx);
1518 fjy3 = _mm256_add_ps(fjy3,ty);
1519 fjz3 = _mm256_add_ps(fjz3,tz);
1521 /**************************
1522 * CALCULATE INTERACTIONS *
1523 **************************/
1525 /* REACTION-FIELD ELECTROSTATICS */
1526 felec = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1530 /* Calculate temporary vectorial force */
1531 tx = _mm256_mul_ps(fscal,dx31);
1532 ty = _mm256_mul_ps(fscal,dy31);
1533 tz = _mm256_mul_ps(fscal,dz31);
1535 /* Update vectorial force */
1536 fix3 = _mm256_add_ps(fix3,tx);
1537 fiy3 = _mm256_add_ps(fiy3,ty);
1538 fiz3 = _mm256_add_ps(fiz3,tz);
1540 fjx1 = _mm256_add_ps(fjx1,tx);
1541 fjy1 = _mm256_add_ps(fjy1,ty);
1542 fjz1 = _mm256_add_ps(fjz1,tz);
1544 /**************************
1545 * CALCULATE INTERACTIONS *
1546 **************************/
1548 /* REACTION-FIELD ELECTROSTATICS */
1549 felec = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1553 /* Calculate temporary vectorial force */
1554 tx = _mm256_mul_ps(fscal,dx32);
1555 ty = _mm256_mul_ps(fscal,dy32);
1556 tz = _mm256_mul_ps(fscal,dz32);
1558 /* Update vectorial force */
1559 fix3 = _mm256_add_ps(fix3,tx);
1560 fiy3 = _mm256_add_ps(fiy3,ty);
1561 fiz3 = _mm256_add_ps(fiz3,tz);
1563 fjx2 = _mm256_add_ps(fjx2,tx);
1564 fjy2 = _mm256_add_ps(fjy2,ty);
1565 fjz2 = _mm256_add_ps(fjz2,tz);
1567 /**************************
1568 * CALCULATE INTERACTIONS *
1569 **************************/
1571 /* REACTION-FIELD ELECTROSTATICS */
1572 felec = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1576 /* Calculate temporary vectorial force */
1577 tx = _mm256_mul_ps(fscal,dx33);
1578 ty = _mm256_mul_ps(fscal,dy33);
1579 tz = _mm256_mul_ps(fscal,dz33);
1581 /* Update vectorial force */
1582 fix3 = _mm256_add_ps(fix3,tx);
1583 fiy3 = _mm256_add_ps(fiy3,ty);
1584 fiz3 = _mm256_add_ps(fiz3,tz);
1586 fjx3 = _mm256_add_ps(fjx3,tx);
1587 fjy3 = _mm256_add_ps(fjy3,ty);
1588 fjz3 = _mm256_add_ps(fjz3,tz);
1590 fjptrA = f+j_coord_offsetA;
1591 fjptrB = f+j_coord_offsetB;
1592 fjptrC = f+j_coord_offsetC;
1593 fjptrD = f+j_coord_offsetD;
1594 fjptrE = f+j_coord_offsetE;
1595 fjptrF = f+j_coord_offsetF;
1596 fjptrG = f+j_coord_offsetG;
1597 fjptrH = f+j_coord_offsetH;
1599 gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1600 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1601 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1603 /* Inner loop uses 273 flops */
1606 if(jidx<j_index_end)
1609 /* Get j neighbor index, and coordinate index */
1610 jnrlistA = jjnr[jidx];
1611 jnrlistB = jjnr[jidx+1];
1612 jnrlistC = jjnr[jidx+2];
1613 jnrlistD = jjnr[jidx+3];
1614 jnrlistE = jjnr[jidx+4];
1615 jnrlistF = jjnr[jidx+5];
1616 jnrlistG = jjnr[jidx+6];
1617 jnrlistH = jjnr[jidx+7];
1618 /* Sign of each element will be negative for non-real atoms.
1619 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1620 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1622 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1623 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1625 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1626 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1627 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1628 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1629 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
1630 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
1631 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
1632 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
1633 j_coord_offsetA = DIM*jnrA;
1634 j_coord_offsetB = DIM*jnrB;
1635 j_coord_offsetC = DIM*jnrC;
1636 j_coord_offsetD = DIM*jnrD;
1637 j_coord_offsetE = DIM*jnrE;
1638 j_coord_offsetF = DIM*jnrF;
1639 j_coord_offsetG = DIM*jnrG;
1640 j_coord_offsetH = DIM*jnrH;
1642 /* load j atom coordinates */
1643 gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1644 x+j_coord_offsetC,x+j_coord_offsetD,
1645 x+j_coord_offsetE,x+j_coord_offsetF,
1646 x+j_coord_offsetG,x+j_coord_offsetH,
1647 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1648 &jy2,&jz2,&jx3,&jy3,&jz3);
1650 /* Calculate displacement vector */
1651 dx00 = _mm256_sub_ps(ix0,jx0);
1652 dy00 = _mm256_sub_ps(iy0,jy0);
1653 dz00 = _mm256_sub_ps(iz0,jz0);
1654 dx11 = _mm256_sub_ps(ix1,jx1);
1655 dy11 = _mm256_sub_ps(iy1,jy1);
1656 dz11 = _mm256_sub_ps(iz1,jz1);
1657 dx12 = _mm256_sub_ps(ix1,jx2);
1658 dy12 = _mm256_sub_ps(iy1,jy2);
1659 dz12 = _mm256_sub_ps(iz1,jz2);
1660 dx13 = _mm256_sub_ps(ix1,jx3);
1661 dy13 = _mm256_sub_ps(iy1,jy3);
1662 dz13 = _mm256_sub_ps(iz1,jz3);
1663 dx21 = _mm256_sub_ps(ix2,jx1);
1664 dy21 = _mm256_sub_ps(iy2,jy1);
1665 dz21 = _mm256_sub_ps(iz2,jz1);
1666 dx22 = _mm256_sub_ps(ix2,jx2);
1667 dy22 = _mm256_sub_ps(iy2,jy2);
1668 dz22 = _mm256_sub_ps(iz2,jz2);
1669 dx23 = _mm256_sub_ps(ix2,jx3);
1670 dy23 = _mm256_sub_ps(iy2,jy3);
1671 dz23 = _mm256_sub_ps(iz2,jz3);
1672 dx31 = _mm256_sub_ps(ix3,jx1);
1673 dy31 = _mm256_sub_ps(iy3,jy1);
1674 dz31 = _mm256_sub_ps(iz3,jz1);
1675 dx32 = _mm256_sub_ps(ix3,jx2);
1676 dy32 = _mm256_sub_ps(iy3,jy2);
1677 dz32 = _mm256_sub_ps(iz3,jz2);
1678 dx33 = _mm256_sub_ps(ix3,jx3);
1679 dy33 = _mm256_sub_ps(iy3,jy3);
1680 dz33 = _mm256_sub_ps(iz3,jz3);
1682 /* Calculate squared distance and things based on it */
1683 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1684 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1685 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1686 rsq13 = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1687 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1688 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1689 rsq23 = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1690 rsq31 = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1691 rsq32 = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1692 rsq33 = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1694 rinv11 = avx256_invsqrt_f(rsq11);
1695 rinv12 = avx256_invsqrt_f(rsq12);
1696 rinv13 = avx256_invsqrt_f(rsq13);
1697 rinv21 = avx256_invsqrt_f(rsq21);
1698 rinv22 = avx256_invsqrt_f(rsq22);
1699 rinv23 = avx256_invsqrt_f(rsq23);
1700 rinv31 = avx256_invsqrt_f(rsq31);
1701 rinv32 = avx256_invsqrt_f(rsq32);
1702 rinv33 = avx256_invsqrt_f(rsq33);
1704 rinvsq00 = avx256_inv_f(rsq00);
1705 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1706 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1707 rinvsq13 = _mm256_mul_ps(rinv13,rinv13);
1708 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1709 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1710 rinvsq23 = _mm256_mul_ps(rinv23,rinv23);
1711 rinvsq31 = _mm256_mul_ps(rinv31,rinv31);
1712 rinvsq32 = _mm256_mul_ps(rinv32,rinv32);
1713 rinvsq33 = _mm256_mul_ps(rinv33,rinv33);
1715 fjx0 = _mm256_setzero_ps();
1716 fjy0 = _mm256_setzero_ps();
1717 fjz0 = _mm256_setzero_ps();
1718 fjx1 = _mm256_setzero_ps();
1719 fjy1 = _mm256_setzero_ps();
1720 fjz1 = _mm256_setzero_ps();
1721 fjx2 = _mm256_setzero_ps();
1722 fjy2 = _mm256_setzero_ps();
1723 fjz2 = _mm256_setzero_ps();
1724 fjx3 = _mm256_setzero_ps();
1725 fjy3 = _mm256_setzero_ps();
1726 fjz3 = _mm256_setzero_ps();
1728 /**************************
1729 * CALCULATE INTERACTIONS *
1730 **************************/
1732 /* LENNARD-JONES DISPERSION/REPULSION */
1734 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1735 fvdw = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1739 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1741 /* Calculate temporary vectorial force */
1742 tx = _mm256_mul_ps(fscal,dx00);
1743 ty = _mm256_mul_ps(fscal,dy00);
1744 tz = _mm256_mul_ps(fscal,dz00);
1746 /* Update vectorial force */
1747 fix0 = _mm256_add_ps(fix0,tx);
1748 fiy0 = _mm256_add_ps(fiy0,ty);
1749 fiz0 = _mm256_add_ps(fiz0,tz);
1751 fjx0 = _mm256_add_ps(fjx0,tx);
1752 fjy0 = _mm256_add_ps(fjy0,ty);
1753 fjz0 = _mm256_add_ps(fjz0,tz);
1755 /**************************
1756 * CALCULATE INTERACTIONS *
1757 **************************/
1759 /* REACTION-FIELD ELECTROSTATICS */
1760 felec = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1764 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1766 /* Calculate temporary vectorial force */
1767 tx = _mm256_mul_ps(fscal,dx11);
1768 ty = _mm256_mul_ps(fscal,dy11);
1769 tz = _mm256_mul_ps(fscal,dz11);
1771 /* Update vectorial force */
1772 fix1 = _mm256_add_ps(fix1,tx);
1773 fiy1 = _mm256_add_ps(fiy1,ty);
1774 fiz1 = _mm256_add_ps(fiz1,tz);
1776 fjx1 = _mm256_add_ps(fjx1,tx);
1777 fjy1 = _mm256_add_ps(fjy1,ty);
1778 fjz1 = _mm256_add_ps(fjz1,tz);
1780 /**************************
1781 * CALCULATE INTERACTIONS *
1782 **************************/
1784 /* REACTION-FIELD ELECTROSTATICS */
1785 felec = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1789 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1791 /* Calculate temporary vectorial force */
1792 tx = _mm256_mul_ps(fscal,dx12);
1793 ty = _mm256_mul_ps(fscal,dy12);
1794 tz = _mm256_mul_ps(fscal,dz12);
1796 /* Update vectorial force */
1797 fix1 = _mm256_add_ps(fix1,tx);
1798 fiy1 = _mm256_add_ps(fiy1,ty);
1799 fiz1 = _mm256_add_ps(fiz1,tz);
1801 fjx2 = _mm256_add_ps(fjx2,tx);
1802 fjy2 = _mm256_add_ps(fjy2,ty);
1803 fjz2 = _mm256_add_ps(fjz2,tz);
1805 /**************************
1806 * CALCULATE INTERACTIONS *
1807 **************************/
1809 /* REACTION-FIELD ELECTROSTATICS */
1810 felec = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1814 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1816 /* Calculate temporary vectorial force */
1817 tx = _mm256_mul_ps(fscal,dx13);
1818 ty = _mm256_mul_ps(fscal,dy13);
1819 tz = _mm256_mul_ps(fscal,dz13);
1821 /* Update vectorial force */
1822 fix1 = _mm256_add_ps(fix1,tx);
1823 fiy1 = _mm256_add_ps(fiy1,ty);
1824 fiz1 = _mm256_add_ps(fiz1,tz);
1826 fjx3 = _mm256_add_ps(fjx3,tx);
1827 fjy3 = _mm256_add_ps(fjy3,ty);
1828 fjz3 = _mm256_add_ps(fjz3,tz);
1830 /**************************
1831 * CALCULATE INTERACTIONS *
1832 **************************/
1834 /* REACTION-FIELD ELECTROSTATICS */
1835 felec = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1839 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1841 /* Calculate temporary vectorial force */
1842 tx = _mm256_mul_ps(fscal,dx21);
1843 ty = _mm256_mul_ps(fscal,dy21);
1844 tz = _mm256_mul_ps(fscal,dz21);
1846 /* Update vectorial force */
1847 fix2 = _mm256_add_ps(fix2,tx);
1848 fiy2 = _mm256_add_ps(fiy2,ty);
1849 fiz2 = _mm256_add_ps(fiz2,tz);
1851 fjx1 = _mm256_add_ps(fjx1,tx);
1852 fjy1 = _mm256_add_ps(fjy1,ty);
1853 fjz1 = _mm256_add_ps(fjz1,tz);
1855 /**************************
1856 * CALCULATE INTERACTIONS *
1857 **************************/
1859 /* REACTION-FIELD ELECTROSTATICS */
1860 felec = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1864 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1866 /* Calculate temporary vectorial force */
1867 tx = _mm256_mul_ps(fscal,dx22);
1868 ty = _mm256_mul_ps(fscal,dy22);
1869 tz = _mm256_mul_ps(fscal,dz22);
1871 /* Update vectorial force */
1872 fix2 = _mm256_add_ps(fix2,tx);
1873 fiy2 = _mm256_add_ps(fiy2,ty);
1874 fiz2 = _mm256_add_ps(fiz2,tz);
1876 fjx2 = _mm256_add_ps(fjx2,tx);
1877 fjy2 = _mm256_add_ps(fjy2,ty);
1878 fjz2 = _mm256_add_ps(fjz2,tz);
1880 /**************************
1881 * CALCULATE INTERACTIONS *
1882 **************************/
1884 /* REACTION-FIELD ELECTROSTATICS */
1885 felec = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1889 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1891 /* Calculate temporary vectorial force */
1892 tx = _mm256_mul_ps(fscal,dx23);
1893 ty = _mm256_mul_ps(fscal,dy23);
1894 tz = _mm256_mul_ps(fscal,dz23);
1896 /* Update vectorial force */
1897 fix2 = _mm256_add_ps(fix2,tx);
1898 fiy2 = _mm256_add_ps(fiy2,ty);
1899 fiz2 = _mm256_add_ps(fiz2,tz);
1901 fjx3 = _mm256_add_ps(fjx3,tx);
1902 fjy3 = _mm256_add_ps(fjy3,ty);
1903 fjz3 = _mm256_add_ps(fjz3,tz);
1905 /**************************
1906 * CALCULATE INTERACTIONS *
1907 **************************/
1909 /* REACTION-FIELD ELECTROSTATICS */
1910 felec = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1914 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1916 /* Calculate temporary vectorial force */
1917 tx = _mm256_mul_ps(fscal,dx31);
1918 ty = _mm256_mul_ps(fscal,dy31);
1919 tz = _mm256_mul_ps(fscal,dz31);
1921 /* Update vectorial force */
1922 fix3 = _mm256_add_ps(fix3,tx);
1923 fiy3 = _mm256_add_ps(fiy3,ty);
1924 fiz3 = _mm256_add_ps(fiz3,tz);
1926 fjx1 = _mm256_add_ps(fjx1,tx);
1927 fjy1 = _mm256_add_ps(fjy1,ty);
1928 fjz1 = _mm256_add_ps(fjz1,tz);
1930 /**************************
1931 * CALCULATE INTERACTIONS *
1932 **************************/
1934 /* REACTION-FIELD ELECTROSTATICS */
1935 felec = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1939 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1941 /* Calculate temporary vectorial force */
1942 tx = _mm256_mul_ps(fscal,dx32);
1943 ty = _mm256_mul_ps(fscal,dy32);
1944 tz = _mm256_mul_ps(fscal,dz32);
1946 /* Update vectorial force */
1947 fix3 = _mm256_add_ps(fix3,tx);
1948 fiy3 = _mm256_add_ps(fiy3,ty);
1949 fiz3 = _mm256_add_ps(fiz3,tz);
1951 fjx2 = _mm256_add_ps(fjx2,tx);
1952 fjy2 = _mm256_add_ps(fjy2,ty);
1953 fjz2 = _mm256_add_ps(fjz2,tz);
1955 /**************************
1956 * CALCULATE INTERACTIONS *
1957 **************************/
1959 /* REACTION-FIELD ELECTROSTATICS */
1960 felec = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1964 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1966 /* Calculate temporary vectorial force */
1967 tx = _mm256_mul_ps(fscal,dx33);
1968 ty = _mm256_mul_ps(fscal,dy33);
1969 tz = _mm256_mul_ps(fscal,dz33);
1971 /* Update vectorial force */
1972 fix3 = _mm256_add_ps(fix3,tx);
1973 fiy3 = _mm256_add_ps(fiy3,ty);
1974 fiz3 = _mm256_add_ps(fiz3,tz);
1976 fjx3 = _mm256_add_ps(fjx3,tx);
1977 fjy3 = _mm256_add_ps(fjy3,ty);
1978 fjz3 = _mm256_add_ps(fjz3,tz);
1980 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1981 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1982 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1983 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1984 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1985 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1986 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1987 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1989 gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1990 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1991 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1993 /* Inner loop uses 273 flops */
1996 /* End of innermost loop */
1998 gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1999 f+i_coord_offset,fshift+i_shift_offset);
2001 /* Increment number of inner iterations */
2002 inneriter += j_index_end - j_index_start;
2004 /* Outer loop uses 24 flops */
2007 /* Increment number of outer iterations */
2010 /* Update outer/inner flops */
2012 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*273);