2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS avx_256_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gromacs/simd/math_x86_avx_256_double.h"
50 #include "kernelutil_x86_avx_256_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_VF_avx_256_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LennardJones
56 * Geometry: Water4-Water4
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_VF_avx_256_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int jnrA,jnrB,jnrC,jnrD;
77 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
78 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
79 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
80 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
82 real *shiftvec,*fshift,*x,*f;
83 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
85 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 real * vdwioffsetptr0;
87 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
88 real * vdwioffsetptr1;
89 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
90 real * vdwioffsetptr2;
91 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
92 real * vdwioffsetptr3;
93 __m256d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
94 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
95 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
96 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
97 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
98 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
99 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
100 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
101 __m256d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
102 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
103 __m256d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
104 __m256d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
105 __m256d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
106 __m256d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
107 __m256d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
108 __m256d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
109 __m256d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
110 __m256d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
111 __m256d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
112 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
115 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
118 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
119 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
121 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
122 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
124 __m256d dummy_mask,cutoff_mask;
125 __m128 tmpmask0,tmpmask1;
126 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
127 __m256d one = _mm256_set1_pd(1.0);
128 __m256d two = _mm256_set1_pd(2.0);
134 jindex = nlist->jindex;
136 shiftidx = nlist->shift;
138 shiftvec = fr->shift_vec[0];
139 fshift = fr->fshift[0];
140 facel = _mm256_set1_pd(fr->epsfac);
141 charge = mdatoms->chargeA;
142 nvdwtype = fr->ntype;
144 vdwtype = mdatoms->typeA;
146 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
147 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
148 beta2 = _mm256_mul_pd(beta,beta);
149 beta3 = _mm256_mul_pd(beta,beta2);
151 ewtab = fr->ic->tabq_coul_FDV0;
152 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
153 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
155 /* Setup water-specific parameters */
156 inr = nlist->iinr[0];
157 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
158 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
159 iq3 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
160 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
162 jq1 = _mm256_set1_pd(charge[inr+1]);
163 jq2 = _mm256_set1_pd(charge[inr+2]);
164 jq3 = _mm256_set1_pd(charge[inr+3]);
165 vdwjidx0A = 2*vdwtype[inr+0];
166 c6_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
167 c12_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
168 qq11 = _mm256_mul_pd(iq1,jq1);
169 qq12 = _mm256_mul_pd(iq1,jq2);
170 qq13 = _mm256_mul_pd(iq1,jq3);
171 qq21 = _mm256_mul_pd(iq2,jq1);
172 qq22 = _mm256_mul_pd(iq2,jq2);
173 qq23 = _mm256_mul_pd(iq2,jq3);
174 qq31 = _mm256_mul_pd(iq3,jq1);
175 qq32 = _mm256_mul_pd(iq3,jq2);
176 qq33 = _mm256_mul_pd(iq3,jq3);
178 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
179 rcutoff_scalar = fr->rcoulomb;
180 rcutoff = _mm256_set1_pd(rcutoff_scalar);
181 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
183 sh_vdw_invrcut6 = _mm256_set1_pd(fr->ic->sh_invrc6);
184 rvdw = _mm256_set1_pd(fr->rvdw);
186 /* Avoid stupid compiler warnings */
187 jnrA = jnrB = jnrC = jnrD = 0;
196 for(iidx=0;iidx<4*DIM;iidx++)
201 /* Start outer loop over neighborlists */
202 for(iidx=0; iidx<nri; iidx++)
204 /* Load shift vector for this list */
205 i_shift_offset = DIM*shiftidx[iidx];
207 /* Load limits for loop over neighbors */
208 j_index_start = jindex[iidx];
209 j_index_end = jindex[iidx+1];
211 /* Get outer coordinate index */
213 i_coord_offset = DIM*inr;
215 /* Load i particle coords and add shift vector */
216 gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
217 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
219 fix0 = _mm256_setzero_pd();
220 fiy0 = _mm256_setzero_pd();
221 fiz0 = _mm256_setzero_pd();
222 fix1 = _mm256_setzero_pd();
223 fiy1 = _mm256_setzero_pd();
224 fiz1 = _mm256_setzero_pd();
225 fix2 = _mm256_setzero_pd();
226 fiy2 = _mm256_setzero_pd();
227 fiz2 = _mm256_setzero_pd();
228 fix3 = _mm256_setzero_pd();
229 fiy3 = _mm256_setzero_pd();
230 fiz3 = _mm256_setzero_pd();
232 /* Reset potential sums */
233 velecsum = _mm256_setzero_pd();
234 vvdwsum = _mm256_setzero_pd();
236 /* Start inner kernel loop */
237 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
240 /* Get j neighbor index, and coordinate index */
245 j_coord_offsetA = DIM*jnrA;
246 j_coord_offsetB = DIM*jnrB;
247 j_coord_offsetC = DIM*jnrC;
248 j_coord_offsetD = DIM*jnrD;
250 /* load j atom coordinates */
251 gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
252 x+j_coord_offsetC,x+j_coord_offsetD,
253 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
254 &jy2,&jz2,&jx3,&jy3,&jz3);
256 /* Calculate displacement vector */
257 dx00 = _mm256_sub_pd(ix0,jx0);
258 dy00 = _mm256_sub_pd(iy0,jy0);
259 dz00 = _mm256_sub_pd(iz0,jz0);
260 dx11 = _mm256_sub_pd(ix1,jx1);
261 dy11 = _mm256_sub_pd(iy1,jy1);
262 dz11 = _mm256_sub_pd(iz1,jz1);
263 dx12 = _mm256_sub_pd(ix1,jx2);
264 dy12 = _mm256_sub_pd(iy1,jy2);
265 dz12 = _mm256_sub_pd(iz1,jz2);
266 dx13 = _mm256_sub_pd(ix1,jx3);
267 dy13 = _mm256_sub_pd(iy1,jy3);
268 dz13 = _mm256_sub_pd(iz1,jz3);
269 dx21 = _mm256_sub_pd(ix2,jx1);
270 dy21 = _mm256_sub_pd(iy2,jy1);
271 dz21 = _mm256_sub_pd(iz2,jz1);
272 dx22 = _mm256_sub_pd(ix2,jx2);
273 dy22 = _mm256_sub_pd(iy2,jy2);
274 dz22 = _mm256_sub_pd(iz2,jz2);
275 dx23 = _mm256_sub_pd(ix2,jx3);
276 dy23 = _mm256_sub_pd(iy2,jy3);
277 dz23 = _mm256_sub_pd(iz2,jz3);
278 dx31 = _mm256_sub_pd(ix3,jx1);
279 dy31 = _mm256_sub_pd(iy3,jy1);
280 dz31 = _mm256_sub_pd(iz3,jz1);
281 dx32 = _mm256_sub_pd(ix3,jx2);
282 dy32 = _mm256_sub_pd(iy3,jy2);
283 dz32 = _mm256_sub_pd(iz3,jz2);
284 dx33 = _mm256_sub_pd(ix3,jx3);
285 dy33 = _mm256_sub_pd(iy3,jy3);
286 dz33 = _mm256_sub_pd(iz3,jz3);
288 /* Calculate squared distance and things based on it */
289 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
290 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
291 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
292 rsq13 = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
293 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
294 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
295 rsq23 = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
296 rsq31 = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
297 rsq32 = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
298 rsq33 = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
300 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
301 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
302 rinv13 = gmx_mm256_invsqrt_pd(rsq13);
303 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
304 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
305 rinv23 = gmx_mm256_invsqrt_pd(rsq23);
306 rinv31 = gmx_mm256_invsqrt_pd(rsq31);
307 rinv32 = gmx_mm256_invsqrt_pd(rsq32);
308 rinv33 = gmx_mm256_invsqrt_pd(rsq33);
310 rinvsq00 = gmx_mm256_inv_pd(rsq00);
311 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
312 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
313 rinvsq13 = _mm256_mul_pd(rinv13,rinv13);
314 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
315 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
316 rinvsq23 = _mm256_mul_pd(rinv23,rinv23);
317 rinvsq31 = _mm256_mul_pd(rinv31,rinv31);
318 rinvsq32 = _mm256_mul_pd(rinv32,rinv32);
319 rinvsq33 = _mm256_mul_pd(rinv33,rinv33);
321 fjx0 = _mm256_setzero_pd();
322 fjy0 = _mm256_setzero_pd();
323 fjz0 = _mm256_setzero_pd();
324 fjx1 = _mm256_setzero_pd();
325 fjy1 = _mm256_setzero_pd();
326 fjz1 = _mm256_setzero_pd();
327 fjx2 = _mm256_setzero_pd();
328 fjy2 = _mm256_setzero_pd();
329 fjz2 = _mm256_setzero_pd();
330 fjx3 = _mm256_setzero_pd();
331 fjy3 = _mm256_setzero_pd();
332 fjz3 = _mm256_setzero_pd();
334 /**************************
335 * CALCULATE INTERACTIONS *
336 **************************/
338 if (gmx_mm256_any_lt(rsq00,rcutoff2))
341 /* LENNARD-JONES DISPERSION/REPULSION */
343 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
344 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
345 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
346 vvdw = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
347 _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
348 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
350 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
352 /* Update potential sum for this i atom from the interaction with this j atom. */
353 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
354 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
358 fscal = _mm256_and_pd(fscal,cutoff_mask);
360 /* Calculate temporary vectorial force */
361 tx = _mm256_mul_pd(fscal,dx00);
362 ty = _mm256_mul_pd(fscal,dy00);
363 tz = _mm256_mul_pd(fscal,dz00);
365 /* Update vectorial force */
366 fix0 = _mm256_add_pd(fix0,tx);
367 fiy0 = _mm256_add_pd(fiy0,ty);
368 fiz0 = _mm256_add_pd(fiz0,tz);
370 fjx0 = _mm256_add_pd(fjx0,tx);
371 fjy0 = _mm256_add_pd(fjy0,ty);
372 fjz0 = _mm256_add_pd(fjz0,tz);
376 /**************************
377 * CALCULATE INTERACTIONS *
378 **************************/
380 if (gmx_mm256_any_lt(rsq11,rcutoff2))
383 r11 = _mm256_mul_pd(rsq11,rinv11);
385 /* EWALD ELECTROSTATICS */
387 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
388 ewrt = _mm256_mul_pd(r11,ewtabscale);
389 ewitab = _mm256_cvttpd_epi32(ewrt);
390 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
391 ewitab = _mm_slli_epi32(ewitab,2);
392 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
393 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
394 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
395 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
396 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
397 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
398 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
399 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_sub_pd(rinv11,sh_ewald),velec));
400 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
402 cutoff_mask = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
404 /* Update potential sum for this i atom from the interaction with this j atom. */
405 velec = _mm256_and_pd(velec,cutoff_mask);
406 velecsum = _mm256_add_pd(velecsum,velec);
410 fscal = _mm256_and_pd(fscal,cutoff_mask);
412 /* Calculate temporary vectorial force */
413 tx = _mm256_mul_pd(fscal,dx11);
414 ty = _mm256_mul_pd(fscal,dy11);
415 tz = _mm256_mul_pd(fscal,dz11);
417 /* Update vectorial force */
418 fix1 = _mm256_add_pd(fix1,tx);
419 fiy1 = _mm256_add_pd(fiy1,ty);
420 fiz1 = _mm256_add_pd(fiz1,tz);
422 fjx1 = _mm256_add_pd(fjx1,tx);
423 fjy1 = _mm256_add_pd(fjy1,ty);
424 fjz1 = _mm256_add_pd(fjz1,tz);
428 /**************************
429 * CALCULATE INTERACTIONS *
430 **************************/
432 if (gmx_mm256_any_lt(rsq12,rcutoff2))
435 r12 = _mm256_mul_pd(rsq12,rinv12);
437 /* EWALD ELECTROSTATICS */
439 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
440 ewrt = _mm256_mul_pd(r12,ewtabscale);
441 ewitab = _mm256_cvttpd_epi32(ewrt);
442 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
443 ewitab = _mm_slli_epi32(ewitab,2);
444 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
445 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
446 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
447 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
448 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
449 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
450 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
451 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_sub_pd(rinv12,sh_ewald),velec));
452 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
454 cutoff_mask = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
456 /* Update potential sum for this i atom from the interaction with this j atom. */
457 velec = _mm256_and_pd(velec,cutoff_mask);
458 velecsum = _mm256_add_pd(velecsum,velec);
462 fscal = _mm256_and_pd(fscal,cutoff_mask);
464 /* Calculate temporary vectorial force */
465 tx = _mm256_mul_pd(fscal,dx12);
466 ty = _mm256_mul_pd(fscal,dy12);
467 tz = _mm256_mul_pd(fscal,dz12);
469 /* Update vectorial force */
470 fix1 = _mm256_add_pd(fix1,tx);
471 fiy1 = _mm256_add_pd(fiy1,ty);
472 fiz1 = _mm256_add_pd(fiz1,tz);
474 fjx2 = _mm256_add_pd(fjx2,tx);
475 fjy2 = _mm256_add_pd(fjy2,ty);
476 fjz2 = _mm256_add_pd(fjz2,tz);
480 /**************************
481 * CALCULATE INTERACTIONS *
482 **************************/
484 if (gmx_mm256_any_lt(rsq13,rcutoff2))
487 r13 = _mm256_mul_pd(rsq13,rinv13);
489 /* EWALD ELECTROSTATICS */
491 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
492 ewrt = _mm256_mul_pd(r13,ewtabscale);
493 ewitab = _mm256_cvttpd_epi32(ewrt);
494 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
495 ewitab = _mm_slli_epi32(ewitab,2);
496 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
497 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
498 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
499 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
500 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
501 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
502 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
503 velec = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_sub_pd(rinv13,sh_ewald),velec));
504 felec = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
506 cutoff_mask = _mm256_cmp_pd(rsq13,rcutoff2,_CMP_LT_OQ);
508 /* Update potential sum for this i atom from the interaction with this j atom. */
509 velec = _mm256_and_pd(velec,cutoff_mask);
510 velecsum = _mm256_add_pd(velecsum,velec);
514 fscal = _mm256_and_pd(fscal,cutoff_mask);
516 /* Calculate temporary vectorial force */
517 tx = _mm256_mul_pd(fscal,dx13);
518 ty = _mm256_mul_pd(fscal,dy13);
519 tz = _mm256_mul_pd(fscal,dz13);
521 /* Update vectorial force */
522 fix1 = _mm256_add_pd(fix1,tx);
523 fiy1 = _mm256_add_pd(fiy1,ty);
524 fiz1 = _mm256_add_pd(fiz1,tz);
526 fjx3 = _mm256_add_pd(fjx3,tx);
527 fjy3 = _mm256_add_pd(fjy3,ty);
528 fjz3 = _mm256_add_pd(fjz3,tz);
532 /**************************
533 * CALCULATE INTERACTIONS *
534 **************************/
536 if (gmx_mm256_any_lt(rsq21,rcutoff2))
539 r21 = _mm256_mul_pd(rsq21,rinv21);
541 /* EWALD ELECTROSTATICS */
543 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
544 ewrt = _mm256_mul_pd(r21,ewtabscale);
545 ewitab = _mm256_cvttpd_epi32(ewrt);
546 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
547 ewitab = _mm_slli_epi32(ewitab,2);
548 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
549 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
550 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
551 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
552 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
553 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
554 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
555 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_sub_pd(rinv21,sh_ewald),velec));
556 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
558 cutoff_mask = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
560 /* Update potential sum for this i atom from the interaction with this j atom. */
561 velec = _mm256_and_pd(velec,cutoff_mask);
562 velecsum = _mm256_add_pd(velecsum,velec);
566 fscal = _mm256_and_pd(fscal,cutoff_mask);
568 /* Calculate temporary vectorial force */
569 tx = _mm256_mul_pd(fscal,dx21);
570 ty = _mm256_mul_pd(fscal,dy21);
571 tz = _mm256_mul_pd(fscal,dz21);
573 /* Update vectorial force */
574 fix2 = _mm256_add_pd(fix2,tx);
575 fiy2 = _mm256_add_pd(fiy2,ty);
576 fiz2 = _mm256_add_pd(fiz2,tz);
578 fjx1 = _mm256_add_pd(fjx1,tx);
579 fjy1 = _mm256_add_pd(fjy1,ty);
580 fjz1 = _mm256_add_pd(fjz1,tz);
584 /**************************
585 * CALCULATE INTERACTIONS *
586 **************************/
588 if (gmx_mm256_any_lt(rsq22,rcutoff2))
591 r22 = _mm256_mul_pd(rsq22,rinv22);
593 /* EWALD ELECTROSTATICS */
595 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
596 ewrt = _mm256_mul_pd(r22,ewtabscale);
597 ewitab = _mm256_cvttpd_epi32(ewrt);
598 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
599 ewitab = _mm_slli_epi32(ewitab,2);
600 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
601 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
602 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
603 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
604 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
605 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
606 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
607 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_sub_pd(rinv22,sh_ewald),velec));
608 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
610 cutoff_mask = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
612 /* Update potential sum for this i atom from the interaction with this j atom. */
613 velec = _mm256_and_pd(velec,cutoff_mask);
614 velecsum = _mm256_add_pd(velecsum,velec);
618 fscal = _mm256_and_pd(fscal,cutoff_mask);
620 /* Calculate temporary vectorial force */
621 tx = _mm256_mul_pd(fscal,dx22);
622 ty = _mm256_mul_pd(fscal,dy22);
623 tz = _mm256_mul_pd(fscal,dz22);
625 /* Update vectorial force */
626 fix2 = _mm256_add_pd(fix2,tx);
627 fiy2 = _mm256_add_pd(fiy2,ty);
628 fiz2 = _mm256_add_pd(fiz2,tz);
630 fjx2 = _mm256_add_pd(fjx2,tx);
631 fjy2 = _mm256_add_pd(fjy2,ty);
632 fjz2 = _mm256_add_pd(fjz2,tz);
636 /**************************
637 * CALCULATE INTERACTIONS *
638 **************************/
640 if (gmx_mm256_any_lt(rsq23,rcutoff2))
643 r23 = _mm256_mul_pd(rsq23,rinv23);
645 /* EWALD ELECTROSTATICS */
647 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
648 ewrt = _mm256_mul_pd(r23,ewtabscale);
649 ewitab = _mm256_cvttpd_epi32(ewrt);
650 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
651 ewitab = _mm_slli_epi32(ewitab,2);
652 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
653 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
654 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
655 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
656 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
657 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
658 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
659 velec = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_sub_pd(rinv23,sh_ewald),velec));
660 felec = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
662 cutoff_mask = _mm256_cmp_pd(rsq23,rcutoff2,_CMP_LT_OQ);
664 /* Update potential sum for this i atom from the interaction with this j atom. */
665 velec = _mm256_and_pd(velec,cutoff_mask);
666 velecsum = _mm256_add_pd(velecsum,velec);
670 fscal = _mm256_and_pd(fscal,cutoff_mask);
672 /* Calculate temporary vectorial force */
673 tx = _mm256_mul_pd(fscal,dx23);
674 ty = _mm256_mul_pd(fscal,dy23);
675 tz = _mm256_mul_pd(fscal,dz23);
677 /* Update vectorial force */
678 fix2 = _mm256_add_pd(fix2,tx);
679 fiy2 = _mm256_add_pd(fiy2,ty);
680 fiz2 = _mm256_add_pd(fiz2,tz);
682 fjx3 = _mm256_add_pd(fjx3,tx);
683 fjy3 = _mm256_add_pd(fjy3,ty);
684 fjz3 = _mm256_add_pd(fjz3,tz);
688 /**************************
689 * CALCULATE INTERACTIONS *
690 **************************/
692 if (gmx_mm256_any_lt(rsq31,rcutoff2))
695 r31 = _mm256_mul_pd(rsq31,rinv31);
697 /* EWALD ELECTROSTATICS */
699 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
700 ewrt = _mm256_mul_pd(r31,ewtabscale);
701 ewitab = _mm256_cvttpd_epi32(ewrt);
702 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
703 ewitab = _mm_slli_epi32(ewitab,2);
704 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
705 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
706 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
707 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
708 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
709 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
710 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
711 velec = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_sub_pd(rinv31,sh_ewald),velec));
712 felec = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
714 cutoff_mask = _mm256_cmp_pd(rsq31,rcutoff2,_CMP_LT_OQ);
716 /* Update potential sum for this i atom from the interaction with this j atom. */
717 velec = _mm256_and_pd(velec,cutoff_mask);
718 velecsum = _mm256_add_pd(velecsum,velec);
722 fscal = _mm256_and_pd(fscal,cutoff_mask);
724 /* Calculate temporary vectorial force */
725 tx = _mm256_mul_pd(fscal,dx31);
726 ty = _mm256_mul_pd(fscal,dy31);
727 tz = _mm256_mul_pd(fscal,dz31);
729 /* Update vectorial force */
730 fix3 = _mm256_add_pd(fix3,tx);
731 fiy3 = _mm256_add_pd(fiy3,ty);
732 fiz3 = _mm256_add_pd(fiz3,tz);
734 fjx1 = _mm256_add_pd(fjx1,tx);
735 fjy1 = _mm256_add_pd(fjy1,ty);
736 fjz1 = _mm256_add_pd(fjz1,tz);
740 /**************************
741 * CALCULATE INTERACTIONS *
742 **************************/
744 if (gmx_mm256_any_lt(rsq32,rcutoff2))
747 r32 = _mm256_mul_pd(rsq32,rinv32);
749 /* EWALD ELECTROSTATICS */
751 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
752 ewrt = _mm256_mul_pd(r32,ewtabscale);
753 ewitab = _mm256_cvttpd_epi32(ewrt);
754 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
755 ewitab = _mm_slli_epi32(ewitab,2);
756 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
757 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
758 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
759 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
760 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
761 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
762 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
763 velec = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_sub_pd(rinv32,sh_ewald),velec));
764 felec = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
766 cutoff_mask = _mm256_cmp_pd(rsq32,rcutoff2,_CMP_LT_OQ);
768 /* Update potential sum for this i atom from the interaction with this j atom. */
769 velec = _mm256_and_pd(velec,cutoff_mask);
770 velecsum = _mm256_add_pd(velecsum,velec);
774 fscal = _mm256_and_pd(fscal,cutoff_mask);
776 /* Calculate temporary vectorial force */
777 tx = _mm256_mul_pd(fscal,dx32);
778 ty = _mm256_mul_pd(fscal,dy32);
779 tz = _mm256_mul_pd(fscal,dz32);
781 /* Update vectorial force */
782 fix3 = _mm256_add_pd(fix3,tx);
783 fiy3 = _mm256_add_pd(fiy3,ty);
784 fiz3 = _mm256_add_pd(fiz3,tz);
786 fjx2 = _mm256_add_pd(fjx2,tx);
787 fjy2 = _mm256_add_pd(fjy2,ty);
788 fjz2 = _mm256_add_pd(fjz2,tz);
792 /**************************
793 * CALCULATE INTERACTIONS *
794 **************************/
796 if (gmx_mm256_any_lt(rsq33,rcutoff2))
799 r33 = _mm256_mul_pd(rsq33,rinv33);
801 /* EWALD ELECTROSTATICS */
803 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
804 ewrt = _mm256_mul_pd(r33,ewtabscale);
805 ewitab = _mm256_cvttpd_epi32(ewrt);
806 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
807 ewitab = _mm_slli_epi32(ewitab,2);
808 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
809 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
810 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
811 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
812 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
813 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
814 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
815 velec = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_sub_pd(rinv33,sh_ewald),velec));
816 felec = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
818 cutoff_mask = _mm256_cmp_pd(rsq33,rcutoff2,_CMP_LT_OQ);
820 /* Update potential sum for this i atom from the interaction with this j atom. */
821 velec = _mm256_and_pd(velec,cutoff_mask);
822 velecsum = _mm256_add_pd(velecsum,velec);
826 fscal = _mm256_and_pd(fscal,cutoff_mask);
828 /* Calculate temporary vectorial force */
829 tx = _mm256_mul_pd(fscal,dx33);
830 ty = _mm256_mul_pd(fscal,dy33);
831 tz = _mm256_mul_pd(fscal,dz33);
833 /* Update vectorial force */
834 fix3 = _mm256_add_pd(fix3,tx);
835 fiy3 = _mm256_add_pd(fiy3,ty);
836 fiz3 = _mm256_add_pd(fiz3,tz);
838 fjx3 = _mm256_add_pd(fjx3,tx);
839 fjy3 = _mm256_add_pd(fjy3,ty);
840 fjz3 = _mm256_add_pd(fjz3,tz);
844 fjptrA = f+j_coord_offsetA;
845 fjptrB = f+j_coord_offsetB;
846 fjptrC = f+j_coord_offsetC;
847 fjptrD = f+j_coord_offsetD;
849 gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
850 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
851 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
853 /* Inner loop uses 458 flops */
859 /* Get j neighbor index, and coordinate index */
860 jnrlistA = jjnr[jidx];
861 jnrlistB = jjnr[jidx+1];
862 jnrlistC = jjnr[jidx+2];
863 jnrlistD = jjnr[jidx+3];
864 /* Sign of each element will be negative for non-real atoms.
865 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
866 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
868 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
870 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
871 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
872 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
874 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
875 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
876 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
877 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
878 j_coord_offsetA = DIM*jnrA;
879 j_coord_offsetB = DIM*jnrB;
880 j_coord_offsetC = DIM*jnrC;
881 j_coord_offsetD = DIM*jnrD;
883 /* load j atom coordinates */
884 gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
885 x+j_coord_offsetC,x+j_coord_offsetD,
886 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
887 &jy2,&jz2,&jx3,&jy3,&jz3);
889 /* Calculate displacement vector */
890 dx00 = _mm256_sub_pd(ix0,jx0);
891 dy00 = _mm256_sub_pd(iy0,jy0);
892 dz00 = _mm256_sub_pd(iz0,jz0);
893 dx11 = _mm256_sub_pd(ix1,jx1);
894 dy11 = _mm256_sub_pd(iy1,jy1);
895 dz11 = _mm256_sub_pd(iz1,jz1);
896 dx12 = _mm256_sub_pd(ix1,jx2);
897 dy12 = _mm256_sub_pd(iy1,jy2);
898 dz12 = _mm256_sub_pd(iz1,jz2);
899 dx13 = _mm256_sub_pd(ix1,jx3);
900 dy13 = _mm256_sub_pd(iy1,jy3);
901 dz13 = _mm256_sub_pd(iz1,jz3);
902 dx21 = _mm256_sub_pd(ix2,jx1);
903 dy21 = _mm256_sub_pd(iy2,jy1);
904 dz21 = _mm256_sub_pd(iz2,jz1);
905 dx22 = _mm256_sub_pd(ix2,jx2);
906 dy22 = _mm256_sub_pd(iy2,jy2);
907 dz22 = _mm256_sub_pd(iz2,jz2);
908 dx23 = _mm256_sub_pd(ix2,jx3);
909 dy23 = _mm256_sub_pd(iy2,jy3);
910 dz23 = _mm256_sub_pd(iz2,jz3);
911 dx31 = _mm256_sub_pd(ix3,jx1);
912 dy31 = _mm256_sub_pd(iy3,jy1);
913 dz31 = _mm256_sub_pd(iz3,jz1);
914 dx32 = _mm256_sub_pd(ix3,jx2);
915 dy32 = _mm256_sub_pd(iy3,jy2);
916 dz32 = _mm256_sub_pd(iz3,jz2);
917 dx33 = _mm256_sub_pd(ix3,jx3);
918 dy33 = _mm256_sub_pd(iy3,jy3);
919 dz33 = _mm256_sub_pd(iz3,jz3);
921 /* Calculate squared distance and things based on it */
922 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
923 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
924 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
925 rsq13 = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
926 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
927 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
928 rsq23 = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
929 rsq31 = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
930 rsq32 = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
931 rsq33 = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
933 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
934 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
935 rinv13 = gmx_mm256_invsqrt_pd(rsq13);
936 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
937 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
938 rinv23 = gmx_mm256_invsqrt_pd(rsq23);
939 rinv31 = gmx_mm256_invsqrt_pd(rsq31);
940 rinv32 = gmx_mm256_invsqrt_pd(rsq32);
941 rinv33 = gmx_mm256_invsqrt_pd(rsq33);
943 rinvsq00 = gmx_mm256_inv_pd(rsq00);
944 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
945 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
946 rinvsq13 = _mm256_mul_pd(rinv13,rinv13);
947 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
948 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
949 rinvsq23 = _mm256_mul_pd(rinv23,rinv23);
950 rinvsq31 = _mm256_mul_pd(rinv31,rinv31);
951 rinvsq32 = _mm256_mul_pd(rinv32,rinv32);
952 rinvsq33 = _mm256_mul_pd(rinv33,rinv33);
954 fjx0 = _mm256_setzero_pd();
955 fjy0 = _mm256_setzero_pd();
956 fjz0 = _mm256_setzero_pd();
957 fjx1 = _mm256_setzero_pd();
958 fjy1 = _mm256_setzero_pd();
959 fjz1 = _mm256_setzero_pd();
960 fjx2 = _mm256_setzero_pd();
961 fjy2 = _mm256_setzero_pd();
962 fjz2 = _mm256_setzero_pd();
963 fjx3 = _mm256_setzero_pd();
964 fjy3 = _mm256_setzero_pd();
965 fjz3 = _mm256_setzero_pd();
967 /**************************
968 * CALCULATE INTERACTIONS *
969 **************************/
971 if (gmx_mm256_any_lt(rsq00,rcutoff2))
974 /* LENNARD-JONES DISPERSION/REPULSION */
976 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
977 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
978 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
979 vvdw = _mm256_sub_pd(_mm256_mul_pd( _mm256_sub_pd(vvdw12 , _mm256_mul_pd(c12_00,_mm256_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
980 _mm256_mul_pd( _mm256_sub_pd(vvdw6,_mm256_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
981 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
983 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
985 /* Update potential sum for this i atom from the interaction with this j atom. */
986 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
987 vvdw = _mm256_andnot_pd(dummy_mask,vvdw);
988 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
992 fscal = _mm256_and_pd(fscal,cutoff_mask);
994 fscal = _mm256_andnot_pd(dummy_mask,fscal);
996 /* Calculate temporary vectorial force */
997 tx = _mm256_mul_pd(fscal,dx00);
998 ty = _mm256_mul_pd(fscal,dy00);
999 tz = _mm256_mul_pd(fscal,dz00);
1001 /* Update vectorial force */
1002 fix0 = _mm256_add_pd(fix0,tx);
1003 fiy0 = _mm256_add_pd(fiy0,ty);
1004 fiz0 = _mm256_add_pd(fiz0,tz);
1006 fjx0 = _mm256_add_pd(fjx0,tx);
1007 fjy0 = _mm256_add_pd(fjy0,ty);
1008 fjz0 = _mm256_add_pd(fjz0,tz);
1012 /**************************
1013 * CALCULATE INTERACTIONS *
1014 **************************/
1016 if (gmx_mm256_any_lt(rsq11,rcutoff2))
1019 r11 = _mm256_mul_pd(rsq11,rinv11);
1020 r11 = _mm256_andnot_pd(dummy_mask,r11);
1022 /* EWALD ELECTROSTATICS */
1024 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1025 ewrt = _mm256_mul_pd(r11,ewtabscale);
1026 ewitab = _mm256_cvttpd_epi32(ewrt);
1027 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1028 ewitab = _mm_slli_epi32(ewitab,2);
1029 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1030 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1031 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1032 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1033 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1034 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1035 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1036 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_sub_pd(rinv11,sh_ewald),velec));
1037 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1039 cutoff_mask = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
1041 /* Update potential sum for this i atom from the interaction with this j atom. */
1042 velec = _mm256_and_pd(velec,cutoff_mask);
1043 velec = _mm256_andnot_pd(dummy_mask,velec);
1044 velecsum = _mm256_add_pd(velecsum,velec);
1048 fscal = _mm256_and_pd(fscal,cutoff_mask);
1050 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1052 /* Calculate temporary vectorial force */
1053 tx = _mm256_mul_pd(fscal,dx11);
1054 ty = _mm256_mul_pd(fscal,dy11);
1055 tz = _mm256_mul_pd(fscal,dz11);
1057 /* Update vectorial force */
1058 fix1 = _mm256_add_pd(fix1,tx);
1059 fiy1 = _mm256_add_pd(fiy1,ty);
1060 fiz1 = _mm256_add_pd(fiz1,tz);
1062 fjx1 = _mm256_add_pd(fjx1,tx);
1063 fjy1 = _mm256_add_pd(fjy1,ty);
1064 fjz1 = _mm256_add_pd(fjz1,tz);
1068 /**************************
1069 * CALCULATE INTERACTIONS *
1070 **************************/
1072 if (gmx_mm256_any_lt(rsq12,rcutoff2))
1075 r12 = _mm256_mul_pd(rsq12,rinv12);
1076 r12 = _mm256_andnot_pd(dummy_mask,r12);
1078 /* EWALD ELECTROSTATICS */
1080 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1081 ewrt = _mm256_mul_pd(r12,ewtabscale);
1082 ewitab = _mm256_cvttpd_epi32(ewrt);
1083 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1084 ewitab = _mm_slli_epi32(ewitab,2);
1085 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1086 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1087 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1088 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1089 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1090 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1091 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1092 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_sub_pd(rinv12,sh_ewald),velec));
1093 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1095 cutoff_mask = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
1097 /* Update potential sum for this i atom from the interaction with this j atom. */
1098 velec = _mm256_and_pd(velec,cutoff_mask);
1099 velec = _mm256_andnot_pd(dummy_mask,velec);
1100 velecsum = _mm256_add_pd(velecsum,velec);
1104 fscal = _mm256_and_pd(fscal,cutoff_mask);
1106 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1108 /* Calculate temporary vectorial force */
1109 tx = _mm256_mul_pd(fscal,dx12);
1110 ty = _mm256_mul_pd(fscal,dy12);
1111 tz = _mm256_mul_pd(fscal,dz12);
1113 /* Update vectorial force */
1114 fix1 = _mm256_add_pd(fix1,tx);
1115 fiy1 = _mm256_add_pd(fiy1,ty);
1116 fiz1 = _mm256_add_pd(fiz1,tz);
1118 fjx2 = _mm256_add_pd(fjx2,tx);
1119 fjy2 = _mm256_add_pd(fjy2,ty);
1120 fjz2 = _mm256_add_pd(fjz2,tz);
1124 /**************************
1125 * CALCULATE INTERACTIONS *
1126 **************************/
1128 if (gmx_mm256_any_lt(rsq13,rcutoff2))
1131 r13 = _mm256_mul_pd(rsq13,rinv13);
1132 r13 = _mm256_andnot_pd(dummy_mask,r13);
1134 /* EWALD ELECTROSTATICS */
1136 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1137 ewrt = _mm256_mul_pd(r13,ewtabscale);
1138 ewitab = _mm256_cvttpd_epi32(ewrt);
1139 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1140 ewitab = _mm_slli_epi32(ewitab,2);
1141 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1142 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1143 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1144 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1145 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1146 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1147 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1148 velec = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_sub_pd(rinv13,sh_ewald),velec));
1149 felec = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
1151 cutoff_mask = _mm256_cmp_pd(rsq13,rcutoff2,_CMP_LT_OQ);
1153 /* Update potential sum for this i atom from the interaction with this j atom. */
1154 velec = _mm256_and_pd(velec,cutoff_mask);
1155 velec = _mm256_andnot_pd(dummy_mask,velec);
1156 velecsum = _mm256_add_pd(velecsum,velec);
1160 fscal = _mm256_and_pd(fscal,cutoff_mask);
1162 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1164 /* Calculate temporary vectorial force */
1165 tx = _mm256_mul_pd(fscal,dx13);
1166 ty = _mm256_mul_pd(fscal,dy13);
1167 tz = _mm256_mul_pd(fscal,dz13);
1169 /* Update vectorial force */
1170 fix1 = _mm256_add_pd(fix1,tx);
1171 fiy1 = _mm256_add_pd(fiy1,ty);
1172 fiz1 = _mm256_add_pd(fiz1,tz);
1174 fjx3 = _mm256_add_pd(fjx3,tx);
1175 fjy3 = _mm256_add_pd(fjy3,ty);
1176 fjz3 = _mm256_add_pd(fjz3,tz);
1180 /**************************
1181 * CALCULATE INTERACTIONS *
1182 **************************/
1184 if (gmx_mm256_any_lt(rsq21,rcutoff2))
1187 r21 = _mm256_mul_pd(rsq21,rinv21);
1188 r21 = _mm256_andnot_pd(dummy_mask,r21);
1190 /* EWALD ELECTROSTATICS */
1192 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1193 ewrt = _mm256_mul_pd(r21,ewtabscale);
1194 ewitab = _mm256_cvttpd_epi32(ewrt);
1195 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1196 ewitab = _mm_slli_epi32(ewitab,2);
1197 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1198 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1199 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1200 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1201 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1202 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1203 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1204 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_sub_pd(rinv21,sh_ewald),velec));
1205 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1207 cutoff_mask = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
1209 /* Update potential sum for this i atom from the interaction with this j atom. */
1210 velec = _mm256_and_pd(velec,cutoff_mask);
1211 velec = _mm256_andnot_pd(dummy_mask,velec);
1212 velecsum = _mm256_add_pd(velecsum,velec);
1216 fscal = _mm256_and_pd(fscal,cutoff_mask);
1218 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1220 /* Calculate temporary vectorial force */
1221 tx = _mm256_mul_pd(fscal,dx21);
1222 ty = _mm256_mul_pd(fscal,dy21);
1223 tz = _mm256_mul_pd(fscal,dz21);
1225 /* Update vectorial force */
1226 fix2 = _mm256_add_pd(fix2,tx);
1227 fiy2 = _mm256_add_pd(fiy2,ty);
1228 fiz2 = _mm256_add_pd(fiz2,tz);
1230 fjx1 = _mm256_add_pd(fjx1,tx);
1231 fjy1 = _mm256_add_pd(fjy1,ty);
1232 fjz1 = _mm256_add_pd(fjz1,tz);
1236 /**************************
1237 * CALCULATE INTERACTIONS *
1238 **************************/
1240 if (gmx_mm256_any_lt(rsq22,rcutoff2))
1243 r22 = _mm256_mul_pd(rsq22,rinv22);
1244 r22 = _mm256_andnot_pd(dummy_mask,r22);
1246 /* EWALD ELECTROSTATICS */
1248 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1249 ewrt = _mm256_mul_pd(r22,ewtabscale);
1250 ewitab = _mm256_cvttpd_epi32(ewrt);
1251 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1252 ewitab = _mm_slli_epi32(ewitab,2);
1253 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1254 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1255 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1256 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1257 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1258 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1259 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1260 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_sub_pd(rinv22,sh_ewald),velec));
1261 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1263 cutoff_mask = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
1265 /* Update potential sum for this i atom from the interaction with this j atom. */
1266 velec = _mm256_and_pd(velec,cutoff_mask);
1267 velec = _mm256_andnot_pd(dummy_mask,velec);
1268 velecsum = _mm256_add_pd(velecsum,velec);
1272 fscal = _mm256_and_pd(fscal,cutoff_mask);
1274 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1276 /* Calculate temporary vectorial force */
1277 tx = _mm256_mul_pd(fscal,dx22);
1278 ty = _mm256_mul_pd(fscal,dy22);
1279 tz = _mm256_mul_pd(fscal,dz22);
1281 /* Update vectorial force */
1282 fix2 = _mm256_add_pd(fix2,tx);
1283 fiy2 = _mm256_add_pd(fiy2,ty);
1284 fiz2 = _mm256_add_pd(fiz2,tz);
1286 fjx2 = _mm256_add_pd(fjx2,tx);
1287 fjy2 = _mm256_add_pd(fjy2,ty);
1288 fjz2 = _mm256_add_pd(fjz2,tz);
1292 /**************************
1293 * CALCULATE INTERACTIONS *
1294 **************************/
1296 if (gmx_mm256_any_lt(rsq23,rcutoff2))
1299 r23 = _mm256_mul_pd(rsq23,rinv23);
1300 r23 = _mm256_andnot_pd(dummy_mask,r23);
1302 /* EWALD ELECTROSTATICS */
1304 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1305 ewrt = _mm256_mul_pd(r23,ewtabscale);
1306 ewitab = _mm256_cvttpd_epi32(ewrt);
1307 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1308 ewitab = _mm_slli_epi32(ewitab,2);
1309 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1310 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1311 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1312 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1313 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1314 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1315 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1316 velec = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_sub_pd(rinv23,sh_ewald),velec));
1317 felec = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
1319 cutoff_mask = _mm256_cmp_pd(rsq23,rcutoff2,_CMP_LT_OQ);
1321 /* Update potential sum for this i atom from the interaction with this j atom. */
1322 velec = _mm256_and_pd(velec,cutoff_mask);
1323 velec = _mm256_andnot_pd(dummy_mask,velec);
1324 velecsum = _mm256_add_pd(velecsum,velec);
1328 fscal = _mm256_and_pd(fscal,cutoff_mask);
1330 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1332 /* Calculate temporary vectorial force */
1333 tx = _mm256_mul_pd(fscal,dx23);
1334 ty = _mm256_mul_pd(fscal,dy23);
1335 tz = _mm256_mul_pd(fscal,dz23);
1337 /* Update vectorial force */
1338 fix2 = _mm256_add_pd(fix2,tx);
1339 fiy2 = _mm256_add_pd(fiy2,ty);
1340 fiz2 = _mm256_add_pd(fiz2,tz);
1342 fjx3 = _mm256_add_pd(fjx3,tx);
1343 fjy3 = _mm256_add_pd(fjy3,ty);
1344 fjz3 = _mm256_add_pd(fjz3,tz);
1348 /**************************
1349 * CALCULATE INTERACTIONS *
1350 **************************/
1352 if (gmx_mm256_any_lt(rsq31,rcutoff2))
1355 r31 = _mm256_mul_pd(rsq31,rinv31);
1356 r31 = _mm256_andnot_pd(dummy_mask,r31);
1358 /* EWALD ELECTROSTATICS */
1360 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1361 ewrt = _mm256_mul_pd(r31,ewtabscale);
1362 ewitab = _mm256_cvttpd_epi32(ewrt);
1363 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1364 ewitab = _mm_slli_epi32(ewitab,2);
1365 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1366 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1367 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1368 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1369 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1370 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1371 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1372 velec = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_sub_pd(rinv31,sh_ewald),velec));
1373 felec = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
1375 cutoff_mask = _mm256_cmp_pd(rsq31,rcutoff2,_CMP_LT_OQ);
1377 /* Update potential sum for this i atom from the interaction with this j atom. */
1378 velec = _mm256_and_pd(velec,cutoff_mask);
1379 velec = _mm256_andnot_pd(dummy_mask,velec);
1380 velecsum = _mm256_add_pd(velecsum,velec);
1384 fscal = _mm256_and_pd(fscal,cutoff_mask);
1386 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1388 /* Calculate temporary vectorial force */
1389 tx = _mm256_mul_pd(fscal,dx31);
1390 ty = _mm256_mul_pd(fscal,dy31);
1391 tz = _mm256_mul_pd(fscal,dz31);
1393 /* Update vectorial force */
1394 fix3 = _mm256_add_pd(fix3,tx);
1395 fiy3 = _mm256_add_pd(fiy3,ty);
1396 fiz3 = _mm256_add_pd(fiz3,tz);
1398 fjx1 = _mm256_add_pd(fjx1,tx);
1399 fjy1 = _mm256_add_pd(fjy1,ty);
1400 fjz1 = _mm256_add_pd(fjz1,tz);
1404 /**************************
1405 * CALCULATE INTERACTIONS *
1406 **************************/
1408 if (gmx_mm256_any_lt(rsq32,rcutoff2))
1411 r32 = _mm256_mul_pd(rsq32,rinv32);
1412 r32 = _mm256_andnot_pd(dummy_mask,r32);
1414 /* EWALD ELECTROSTATICS */
1416 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1417 ewrt = _mm256_mul_pd(r32,ewtabscale);
1418 ewitab = _mm256_cvttpd_epi32(ewrt);
1419 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1420 ewitab = _mm_slli_epi32(ewitab,2);
1421 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1422 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1423 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1424 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1425 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1426 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1427 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1428 velec = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_sub_pd(rinv32,sh_ewald),velec));
1429 felec = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
1431 cutoff_mask = _mm256_cmp_pd(rsq32,rcutoff2,_CMP_LT_OQ);
1433 /* Update potential sum for this i atom from the interaction with this j atom. */
1434 velec = _mm256_and_pd(velec,cutoff_mask);
1435 velec = _mm256_andnot_pd(dummy_mask,velec);
1436 velecsum = _mm256_add_pd(velecsum,velec);
1440 fscal = _mm256_and_pd(fscal,cutoff_mask);
1442 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1444 /* Calculate temporary vectorial force */
1445 tx = _mm256_mul_pd(fscal,dx32);
1446 ty = _mm256_mul_pd(fscal,dy32);
1447 tz = _mm256_mul_pd(fscal,dz32);
1449 /* Update vectorial force */
1450 fix3 = _mm256_add_pd(fix3,tx);
1451 fiy3 = _mm256_add_pd(fiy3,ty);
1452 fiz3 = _mm256_add_pd(fiz3,tz);
1454 fjx2 = _mm256_add_pd(fjx2,tx);
1455 fjy2 = _mm256_add_pd(fjy2,ty);
1456 fjz2 = _mm256_add_pd(fjz2,tz);
1460 /**************************
1461 * CALCULATE INTERACTIONS *
1462 **************************/
1464 if (gmx_mm256_any_lt(rsq33,rcutoff2))
1467 r33 = _mm256_mul_pd(rsq33,rinv33);
1468 r33 = _mm256_andnot_pd(dummy_mask,r33);
1470 /* EWALD ELECTROSTATICS */
1472 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1473 ewrt = _mm256_mul_pd(r33,ewtabscale);
1474 ewitab = _mm256_cvttpd_epi32(ewrt);
1475 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1476 ewitab = _mm_slli_epi32(ewitab,2);
1477 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1478 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1479 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1480 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1481 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1482 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1483 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1484 velec = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_sub_pd(rinv33,sh_ewald),velec));
1485 felec = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
1487 cutoff_mask = _mm256_cmp_pd(rsq33,rcutoff2,_CMP_LT_OQ);
1489 /* Update potential sum for this i atom from the interaction with this j atom. */
1490 velec = _mm256_and_pd(velec,cutoff_mask);
1491 velec = _mm256_andnot_pd(dummy_mask,velec);
1492 velecsum = _mm256_add_pd(velecsum,velec);
1496 fscal = _mm256_and_pd(fscal,cutoff_mask);
1498 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1500 /* Calculate temporary vectorial force */
1501 tx = _mm256_mul_pd(fscal,dx33);
1502 ty = _mm256_mul_pd(fscal,dy33);
1503 tz = _mm256_mul_pd(fscal,dz33);
1505 /* Update vectorial force */
1506 fix3 = _mm256_add_pd(fix3,tx);
1507 fiy3 = _mm256_add_pd(fiy3,ty);
1508 fiz3 = _mm256_add_pd(fiz3,tz);
1510 fjx3 = _mm256_add_pd(fjx3,tx);
1511 fjy3 = _mm256_add_pd(fjy3,ty);
1512 fjz3 = _mm256_add_pd(fjz3,tz);
1516 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1517 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1518 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1519 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1521 gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1522 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1523 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1525 /* Inner loop uses 467 flops */
1528 /* End of innermost loop */
1530 gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1531 f+i_coord_offset,fshift+i_shift_offset);
1534 /* Update potential energies */
1535 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1536 gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1538 /* Increment number of inner iterations */
1539 inneriter += j_index_end - j_index_start;
1541 /* Outer loop uses 26 flops */
1544 /* Increment number of outer iterations */
1547 /* Update outer/inner flops */
1549 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*467);
1552 * Gromacs nonbonded kernel: nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_F_avx_256_double
1553 * Electrostatics interaction: Ewald
1554 * VdW interaction: LennardJones
1555 * Geometry: Water4-Water4
1556 * Calculate force/pot: Force
1559 nb_kernel_ElecEwSh_VdwLJSh_GeomW4W4_F_avx_256_double
1560 (t_nblist * gmx_restrict nlist,
1561 rvec * gmx_restrict xx,
1562 rvec * gmx_restrict ff,
1563 t_forcerec * gmx_restrict fr,
1564 t_mdatoms * gmx_restrict mdatoms,
1565 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1566 t_nrnb * gmx_restrict nrnb)
1568 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1569 * just 0 for non-waters.
1570 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1571 * jnr indices corresponding to data put in the four positions in the SIMD register.
1573 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1574 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1575 int jnrA,jnrB,jnrC,jnrD;
1576 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1577 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1578 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1579 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1580 real rcutoff_scalar;
1581 real *shiftvec,*fshift,*x,*f;
1582 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1583 real scratch[4*DIM];
1584 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1585 real * vdwioffsetptr0;
1586 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1587 real * vdwioffsetptr1;
1588 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1589 real * vdwioffsetptr2;
1590 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1591 real * vdwioffsetptr3;
1592 __m256d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1593 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1594 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1595 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1596 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1597 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1598 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1599 int vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1600 __m256d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1601 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1602 __m256d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1603 __m256d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1604 __m256d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1605 __m256d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1606 __m256d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1607 __m256d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1608 __m256d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1609 __m256d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1610 __m256d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1611 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
1614 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1617 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
1618 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
1620 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1621 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1623 __m256d dummy_mask,cutoff_mask;
1624 __m128 tmpmask0,tmpmask1;
1625 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1626 __m256d one = _mm256_set1_pd(1.0);
1627 __m256d two = _mm256_set1_pd(2.0);
1633 jindex = nlist->jindex;
1635 shiftidx = nlist->shift;
1637 shiftvec = fr->shift_vec[0];
1638 fshift = fr->fshift[0];
1639 facel = _mm256_set1_pd(fr->epsfac);
1640 charge = mdatoms->chargeA;
1641 nvdwtype = fr->ntype;
1642 vdwparam = fr->nbfp;
1643 vdwtype = mdatoms->typeA;
1645 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
1646 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
1647 beta2 = _mm256_mul_pd(beta,beta);
1648 beta3 = _mm256_mul_pd(beta,beta2);
1650 ewtab = fr->ic->tabq_coul_F;
1651 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
1652 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1654 /* Setup water-specific parameters */
1655 inr = nlist->iinr[0];
1656 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1657 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1658 iq3 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1659 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
1661 jq1 = _mm256_set1_pd(charge[inr+1]);
1662 jq2 = _mm256_set1_pd(charge[inr+2]);
1663 jq3 = _mm256_set1_pd(charge[inr+3]);
1664 vdwjidx0A = 2*vdwtype[inr+0];
1665 c6_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1666 c12_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1667 qq11 = _mm256_mul_pd(iq1,jq1);
1668 qq12 = _mm256_mul_pd(iq1,jq2);
1669 qq13 = _mm256_mul_pd(iq1,jq3);
1670 qq21 = _mm256_mul_pd(iq2,jq1);
1671 qq22 = _mm256_mul_pd(iq2,jq2);
1672 qq23 = _mm256_mul_pd(iq2,jq3);
1673 qq31 = _mm256_mul_pd(iq3,jq1);
1674 qq32 = _mm256_mul_pd(iq3,jq2);
1675 qq33 = _mm256_mul_pd(iq3,jq3);
1677 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1678 rcutoff_scalar = fr->rcoulomb;
1679 rcutoff = _mm256_set1_pd(rcutoff_scalar);
1680 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
1682 sh_vdw_invrcut6 = _mm256_set1_pd(fr->ic->sh_invrc6);
1683 rvdw = _mm256_set1_pd(fr->rvdw);
1685 /* Avoid stupid compiler warnings */
1686 jnrA = jnrB = jnrC = jnrD = 0;
1687 j_coord_offsetA = 0;
1688 j_coord_offsetB = 0;
1689 j_coord_offsetC = 0;
1690 j_coord_offsetD = 0;
1695 for(iidx=0;iidx<4*DIM;iidx++)
1697 scratch[iidx] = 0.0;
1700 /* Start outer loop over neighborlists */
1701 for(iidx=0; iidx<nri; iidx++)
1703 /* Load shift vector for this list */
1704 i_shift_offset = DIM*shiftidx[iidx];
1706 /* Load limits for loop over neighbors */
1707 j_index_start = jindex[iidx];
1708 j_index_end = jindex[iidx+1];
1710 /* Get outer coordinate index */
1712 i_coord_offset = DIM*inr;
1714 /* Load i particle coords and add shift vector */
1715 gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1716 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1718 fix0 = _mm256_setzero_pd();
1719 fiy0 = _mm256_setzero_pd();
1720 fiz0 = _mm256_setzero_pd();
1721 fix1 = _mm256_setzero_pd();
1722 fiy1 = _mm256_setzero_pd();
1723 fiz1 = _mm256_setzero_pd();
1724 fix2 = _mm256_setzero_pd();
1725 fiy2 = _mm256_setzero_pd();
1726 fiz2 = _mm256_setzero_pd();
1727 fix3 = _mm256_setzero_pd();
1728 fiy3 = _mm256_setzero_pd();
1729 fiz3 = _mm256_setzero_pd();
1731 /* Start inner kernel loop */
1732 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1735 /* Get j neighbor index, and coordinate index */
1737 jnrB = jjnr[jidx+1];
1738 jnrC = jjnr[jidx+2];
1739 jnrD = jjnr[jidx+3];
1740 j_coord_offsetA = DIM*jnrA;
1741 j_coord_offsetB = DIM*jnrB;
1742 j_coord_offsetC = DIM*jnrC;
1743 j_coord_offsetD = DIM*jnrD;
1745 /* load j atom coordinates */
1746 gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1747 x+j_coord_offsetC,x+j_coord_offsetD,
1748 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1749 &jy2,&jz2,&jx3,&jy3,&jz3);
1751 /* Calculate displacement vector */
1752 dx00 = _mm256_sub_pd(ix0,jx0);
1753 dy00 = _mm256_sub_pd(iy0,jy0);
1754 dz00 = _mm256_sub_pd(iz0,jz0);
1755 dx11 = _mm256_sub_pd(ix1,jx1);
1756 dy11 = _mm256_sub_pd(iy1,jy1);
1757 dz11 = _mm256_sub_pd(iz1,jz1);
1758 dx12 = _mm256_sub_pd(ix1,jx2);
1759 dy12 = _mm256_sub_pd(iy1,jy2);
1760 dz12 = _mm256_sub_pd(iz1,jz2);
1761 dx13 = _mm256_sub_pd(ix1,jx3);
1762 dy13 = _mm256_sub_pd(iy1,jy3);
1763 dz13 = _mm256_sub_pd(iz1,jz3);
1764 dx21 = _mm256_sub_pd(ix2,jx1);
1765 dy21 = _mm256_sub_pd(iy2,jy1);
1766 dz21 = _mm256_sub_pd(iz2,jz1);
1767 dx22 = _mm256_sub_pd(ix2,jx2);
1768 dy22 = _mm256_sub_pd(iy2,jy2);
1769 dz22 = _mm256_sub_pd(iz2,jz2);
1770 dx23 = _mm256_sub_pd(ix2,jx3);
1771 dy23 = _mm256_sub_pd(iy2,jy3);
1772 dz23 = _mm256_sub_pd(iz2,jz3);
1773 dx31 = _mm256_sub_pd(ix3,jx1);
1774 dy31 = _mm256_sub_pd(iy3,jy1);
1775 dz31 = _mm256_sub_pd(iz3,jz1);
1776 dx32 = _mm256_sub_pd(ix3,jx2);
1777 dy32 = _mm256_sub_pd(iy3,jy2);
1778 dz32 = _mm256_sub_pd(iz3,jz2);
1779 dx33 = _mm256_sub_pd(ix3,jx3);
1780 dy33 = _mm256_sub_pd(iy3,jy3);
1781 dz33 = _mm256_sub_pd(iz3,jz3);
1783 /* Calculate squared distance and things based on it */
1784 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1785 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1786 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1787 rsq13 = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1788 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1789 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1790 rsq23 = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1791 rsq31 = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1792 rsq32 = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1793 rsq33 = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1795 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
1796 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
1797 rinv13 = gmx_mm256_invsqrt_pd(rsq13);
1798 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
1799 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
1800 rinv23 = gmx_mm256_invsqrt_pd(rsq23);
1801 rinv31 = gmx_mm256_invsqrt_pd(rsq31);
1802 rinv32 = gmx_mm256_invsqrt_pd(rsq32);
1803 rinv33 = gmx_mm256_invsqrt_pd(rsq33);
1805 rinvsq00 = gmx_mm256_inv_pd(rsq00);
1806 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
1807 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
1808 rinvsq13 = _mm256_mul_pd(rinv13,rinv13);
1809 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
1810 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
1811 rinvsq23 = _mm256_mul_pd(rinv23,rinv23);
1812 rinvsq31 = _mm256_mul_pd(rinv31,rinv31);
1813 rinvsq32 = _mm256_mul_pd(rinv32,rinv32);
1814 rinvsq33 = _mm256_mul_pd(rinv33,rinv33);
1816 fjx0 = _mm256_setzero_pd();
1817 fjy0 = _mm256_setzero_pd();
1818 fjz0 = _mm256_setzero_pd();
1819 fjx1 = _mm256_setzero_pd();
1820 fjy1 = _mm256_setzero_pd();
1821 fjz1 = _mm256_setzero_pd();
1822 fjx2 = _mm256_setzero_pd();
1823 fjy2 = _mm256_setzero_pd();
1824 fjz2 = _mm256_setzero_pd();
1825 fjx3 = _mm256_setzero_pd();
1826 fjy3 = _mm256_setzero_pd();
1827 fjz3 = _mm256_setzero_pd();
1829 /**************************
1830 * CALCULATE INTERACTIONS *
1831 **************************/
1833 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1836 /* LENNARD-JONES DISPERSION/REPULSION */
1838 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1839 fvdw = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1841 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1845 fscal = _mm256_and_pd(fscal,cutoff_mask);
1847 /* Calculate temporary vectorial force */
1848 tx = _mm256_mul_pd(fscal,dx00);
1849 ty = _mm256_mul_pd(fscal,dy00);
1850 tz = _mm256_mul_pd(fscal,dz00);
1852 /* Update vectorial force */
1853 fix0 = _mm256_add_pd(fix0,tx);
1854 fiy0 = _mm256_add_pd(fiy0,ty);
1855 fiz0 = _mm256_add_pd(fiz0,tz);
1857 fjx0 = _mm256_add_pd(fjx0,tx);
1858 fjy0 = _mm256_add_pd(fjy0,ty);
1859 fjz0 = _mm256_add_pd(fjz0,tz);
1863 /**************************
1864 * CALCULATE INTERACTIONS *
1865 **************************/
1867 if (gmx_mm256_any_lt(rsq11,rcutoff2))
1870 r11 = _mm256_mul_pd(rsq11,rinv11);
1872 /* EWALD ELECTROSTATICS */
1874 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1875 ewrt = _mm256_mul_pd(r11,ewtabscale);
1876 ewitab = _mm256_cvttpd_epi32(ewrt);
1877 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1878 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1879 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1881 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1882 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1884 cutoff_mask = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
1888 fscal = _mm256_and_pd(fscal,cutoff_mask);
1890 /* Calculate temporary vectorial force */
1891 tx = _mm256_mul_pd(fscal,dx11);
1892 ty = _mm256_mul_pd(fscal,dy11);
1893 tz = _mm256_mul_pd(fscal,dz11);
1895 /* Update vectorial force */
1896 fix1 = _mm256_add_pd(fix1,tx);
1897 fiy1 = _mm256_add_pd(fiy1,ty);
1898 fiz1 = _mm256_add_pd(fiz1,tz);
1900 fjx1 = _mm256_add_pd(fjx1,tx);
1901 fjy1 = _mm256_add_pd(fjy1,ty);
1902 fjz1 = _mm256_add_pd(fjz1,tz);
1906 /**************************
1907 * CALCULATE INTERACTIONS *
1908 **************************/
1910 if (gmx_mm256_any_lt(rsq12,rcutoff2))
1913 r12 = _mm256_mul_pd(rsq12,rinv12);
1915 /* EWALD ELECTROSTATICS */
1917 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1918 ewrt = _mm256_mul_pd(r12,ewtabscale);
1919 ewitab = _mm256_cvttpd_epi32(ewrt);
1920 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1921 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1922 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1924 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1925 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1927 cutoff_mask = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
1931 fscal = _mm256_and_pd(fscal,cutoff_mask);
1933 /* Calculate temporary vectorial force */
1934 tx = _mm256_mul_pd(fscal,dx12);
1935 ty = _mm256_mul_pd(fscal,dy12);
1936 tz = _mm256_mul_pd(fscal,dz12);
1938 /* Update vectorial force */
1939 fix1 = _mm256_add_pd(fix1,tx);
1940 fiy1 = _mm256_add_pd(fiy1,ty);
1941 fiz1 = _mm256_add_pd(fiz1,tz);
1943 fjx2 = _mm256_add_pd(fjx2,tx);
1944 fjy2 = _mm256_add_pd(fjy2,ty);
1945 fjz2 = _mm256_add_pd(fjz2,tz);
1949 /**************************
1950 * CALCULATE INTERACTIONS *
1951 **************************/
1953 if (gmx_mm256_any_lt(rsq13,rcutoff2))
1956 r13 = _mm256_mul_pd(rsq13,rinv13);
1958 /* EWALD ELECTROSTATICS */
1960 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1961 ewrt = _mm256_mul_pd(r13,ewtabscale);
1962 ewitab = _mm256_cvttpd_epi32(ewrt);
1963 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1964 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1965 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1967 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1968 felec = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
1970 cutoff_mask = _mm256_cmp_pd(rsq13,rcutoff2,_CMP_LT_OQ);
1974 fscal = _mm256_and_pd(fscal,cutoff_mask);
1976 /* Calculate temporary vectorial force */
1977 tx = _mm256_mul_pd(fscal,dx13);
1978 ty = _mm256_mul_pd(fscal,dy13);
1979 tz = _mm256_mul_pd(fscal,dz13);
1981 /* Update vectorial force */
1982 fix1 = _mm256_add_pd(fix1,tx);
1983 fiy1 = _mm256_add_pd(fiy1,ty);
1984 fiz1 = _mm256_add_pd(fiz1,tz);
1986 fjx3 = _mm256_add_pd(fjx3,tx);
1987 fjy3 = _mm256_add_pd(fjy3,ty);
1988 fjz3 = _mm256_add_pd(fjz3,tz);
1992 /**************************
1993 * CALCULATE INTERACTIONS *
1994 **************************/
1996 if (gmx_mm256_any_lt(rsq21,rcutoff2))
1999 r21 = _mm256_mul_pd(rsq21,rinv21);
2001 /* EWALD ELECTROSTATICS */
2003 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2004 ewrt = _mm256_mul_pd(r21,ewtabscale);
2005 ewitab = _mm256_cvttpd_epi32(ewrt);
2006 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2007 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2008 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2010 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2011 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2013 cutoff_mask = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
2017 fscal = _mm256_and_pd(fscal,cutoff_mask);
2019 /* Calculate temporary vectorial force */
2020 tx = _mm256_mul_pd(fscal,dx21);
2021 ty = _mm256_mul_pd(fscal,dy21);
2022 tz = _mm256_mul_pd(fscal,dz21);
2024 /* Update vectorial force */
2025 fix2 = _mm256_add_pd(fix2,tx);
2026 fiy2 = _mm256_add_pd(fiy2,ty);
2027 fiz2 = _mm256_add_pd(fiz2,tz);
2029 fjx1 = _mm256_add_pd(fjx1,tx);
2030 fjy1 = _mm256_add_pd(fjy1,ty);
2031 fjz1 = _mm256_add_pd(fjz1,tz);
2035 /**************************
2036 * CALCULATE INTERACTIONS *
2037 **************************/
2039 if (gmx_mm256_any_lt(rsq22,rcutoff2))
2042 r22 = _mm256_mul_pd(rsq22,rinv22);
2044 /* EWALD ELECTROSTATICS */
2046 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2047 ewrt = _mm256_mul_pd(r22,ewtabscale);
2048 ewitab = _mm256_cvttpd_epi32(ewrt);
2049 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2050 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2051 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2053 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2054 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2056 cutoff_mask = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
2060 fscal = _mm256_and_pd(fscal,cutoff_mask);
2062 /* Calculate temporary vectorial force */
2063 tx = _mm256_mul_pd(fscal,dx22);
2064 ty = _mm256_mul_pd(fscal,dy22);
2065 tz = _mm256_mul_pd(fscal,dz22);
2067 /* Update vectorial force */
2068 fix2 = _mm256_add_pd(fix2,tx);
2069 fiy2 = _mm256_add_pd(fiy2,ty);
2070 fiz2 = _mm256_add_pd(fiz2,tz);
2072 fjx2 = _mm256_add_pd(fjx2,tx);
2073 fjy2 = _mm256_add_pd(fjy2,ty);
2074 fjz2 = _mm256_add_pd(fjz2,tz);
2078 /**************************
2079 * CALCULATE INTERACTIONS *
2080 **************************/
2082 if (gmx_mm256_any_lt(rsq23,rcutoff2))
2085 r23 = _mm256_mul_pd(rsq23,rinv23);
2087 /* EWALD ELECTROSTATICS */
2089 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2090 ewrt = _mm256_mul_pd(r23,ewtabscale);
2091 ewitab = _mm256_cvttpd_epi32(ewrt);
2092 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2093 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2094 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2096 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2097 felec = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
2099 cutoff_mask = _mm256_cmp_pd(rsq23,rcutoff2,_CMP_LT_OQ);
2103 fscal = _mm256_and_pd(fscal,cutoff_mask);
2105 /* Calculate temporary vectorial force */
2106 tx = _mm256_mul_pd(fscal,dx23);
2107 ty = _mm256_mul_pd(fscal,dy23);
2108 tz = _mm256_mul_pd(fscal,dz23);
2110 /* Update vectorial force */
2111 fix2 = _mm256_add_pd(fix2,tx);
2112 fiy2 = _mm256_add_pd(fiy2,ty);
2113 fiz2 = _mm256_add_pd(fiz2,tz);
2115 fjx3 = _mm256_add_pd(fjx3,tx);
2116 fjy3 = _mm256_add_pd(fjy3,ty);
2117 fjz3 = _mm256_add_pd(fjz3,tz);
2121 /**************************
2122 * CALCULATE INTERACTIONS *
2123 **************************/
2125 if (gmx_mm256_any_lt(rsq31,rcutoff2))
2128 r31 = _mm256_mul_pd(rsq31,rinv31);
2130 /* EWALD ELECTROSTATICS */
2132 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2133 ewrt = _mm256_mul_pd(r31,ewtabscale);
2134 ewitab = _mm256_cvttpd_epi32(ewrt);
2135 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2136 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2137 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2139 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2140 felec = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
2142 cutoff_mask = _mm256_cmp_pd(rsq31,rcutoff2,_CMP_LT_OQ);
2146 fscal = _mm256_and_pd(fscal,cutoff_mask);
2148 /* Calculate temporary vectorial force */
2149 tx = _mm256_mul_pd(fscal,dx31);
2150 ty = _mm256_mul_pd(fscal,dy31);
2151 tz = _mm256_mul_pd(fscal,dz31);
2153 /* Update vectorial force */
2154 fix3 = _mm256_add_pd(fix3,tx);
2155 fiy3 = _mm256_add_pd(fiy3,ty);
2156 fiz3 = _mm256_add_pd(fiz3,tz);
2158 fjx1 = _mm256_add_pd(fjx1,tx);
2159 fjy1 = _mm256_add_pd(fjy1,ty);
2160 fjz1 = _mm256_add_pd(fjz1,tz);
2164 /**************************
2165 * CALCULATE INTERACTIONS *
2166 **************************/
2168 if (gmx_mm256_any_lt(rsq32,rcutoff2))
2171 r32 = _mm256_mul_pd(rsq32,rinv32);
2173 /* EWALD ELECTROSTATICS */
2175 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2176 ewrt = _mm256_mul_pd(r32,ewtabscale);
2177 ewitab = _mm256_cvttpd_epi32(ewrt);
2178 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2179 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2180 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2182 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2183 felec = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
2185 cutoff_mask = _mm256_cmp_pd(rsq32,rcutoff2,_CMP_LT_OQ);
2189 fscal = _mm256_and_pd(fscal,cutoff_mask);
2191 /* Calculate temporary vectorial force */
2192 tx = _mm256_mul_pd(fscal,dx32);
2193 ty = _mm256_mul_pd(fscal,dy32);
2194 tz = _mm256_mul_pd(fscal,dz32);
2196 /* Update vectorial force */
2197 fix3 = _mm256_add_pd(fix3,tx);
2198 fiy3 = _mm256_add_pd(fiy3,ty);
2199 fiz3 = _mm256_add_pd(fiz3,tz);
2201 fjx2 = _mm256_add_pd(fjx2,tx);
2202 fjy2 = _mm256_add_pd(fjy2,ty);
2203 fjz2 = _mm256_add_pd(fjz2,tz);
2207 /**************************
2208 * CALCULATE INTERACTIONS *
2209 **************************/
2211 if (gmx_mm256_any_lt(rsq33,rcutoff2))
2214 r33 = _mm256_mul_pd(rsq33,rinv33);
2216 /* EWALD ELECTROSTATICS */
2218 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2219 ewrt = _mm256_mul_pd(r33,ewtabscale);
2220 ewitab = _mm256_cvttpd_epi32(ewrt);
2221 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2222 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2223 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2225 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2226 felec = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
2228 cutoff_mask = _mm256_cmp_pd(rsq33,rcutoff2,_CMP_LT_OQ);
2232 fscal = _mm256_and_pd(fscal,cutoff_mask);
2234 /* Calculate temporary vectorial force */
2235 tx = _mm256_mul_pd(fscal,dx33);
2236 ty = _mm256_mul_pd(fscal,dy33);
2237 tz = _mm256_mul_pd(fscal,dz33);
2239 /* Update vectorial force */
2240 fix3 = _mm256_add_pd(fix3,tx);
2241 fiy3 = _mm256_add_pd(fiy3,ty);
2242 fiz3 = _mm256_add_pd(fiz3,tz);
2244 fjx3 = _mm256_add_pd(fjx3,tx);
2245 fjy3 = _mm256_add_pd(fjy3,ty);
2246 fjz3 = _mm256_add_pd(fjz3,tz);
2250 fjptrA = f+j_coord_offsetA;
2251 fjptrB = f+j_coord_offsetB;
2252 fjptrC = f+j_coord_offsetC;
2253 fjptrD = f+j_coord_offsetD;
2255 gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2256 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2257 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2259 /* Inner loop uses 384 flops */
2262 if(jidx<j_index_end)
2265 /* Get j neighbor index, and coordinate index */
2266 jnrlistA = jjnr[jidx];
2267 jnrlistB = jjnr[jidx+1];
2268 jnrlistC = jjnr[jidx+2];
2269 jnrlistD = jjnr[jidx+3];
2270 /* Sign of each element will be negative for non-real atoms.
2271 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2272 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
2274 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2276 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
2277 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
2278 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
2280 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
2281 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
2282 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
2283 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
2284 j_coord_offsetA = DIM*jnrA;
2285 j_coord_offsetB = DIM*jnrB;
2286 j_coord_offsetC = DIM*jnrC;
2287 j_coord_offsetD = DIM*jnrD;
2289 /* load j atom coordinates */
2290 gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
2291 x+j_coord_offsetC,x+j_coord_offsetD,
2292 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2293 &jy2,&jz2,&jx3,&jy3,&jz3);
2295 /* Calculate displacement vector */
2296 dx00 = _mm256_sub_pd(ix0,jx0);
2297 dy00 = _mm256_sub_pd(iy0,jy0);
2298 dz00 = _mm256_sub_pd(iz0,jz0);
2299 dx11 = _mm256_sub_pd(ix1,jx1);
2300 dy11 = _mm256_sub_pd(iy1,jy1);
2301 dz11 = _mm256_sub_pd(iz1,jz1);
2302 dx12 = _mm256_sub_pd(ix1,jx2);
2303 dy12 = _mm256_sub_pd(iy1,jy2);
2304 dz12 = _mm256_sub_pd(iz1,jz2);
2305 dx13 = _mm256_sub_pd(ix1,jx3);
2306 dy13 = _mm256_sub_pd(iy1,jy3);
2307 dz13 = _mm256_sub_pd(iz1,jz3);
2308 dx21 = _mm256_sub_pd(ix2,jx1);
2309 dy21 = _mm256_sub_pd(iy2,jy1);
2310 dz21 = _mm256_sub_pd(iz2,jz1);
2311 dx22 = _mm256_sub_pd(ix2,jx2);
2312 dy22 = _mm256_sub_pd(iy2,jy2);
2313 dz22 = _mm256_sub_pd(iz2,jz2);
2314 dx23 = _mm256_sub_pd(ix2,jx3);
2315 dy23 = _mm256_sub_pd(iy2,jy3);
2316 dz23 = _mm256_sub_pd(iz2,jz3);
2317 dx31 = _mm256_sub_pd(ix3,jx1);
2318 dy31 = _mm256_sub_pd(iy3,jy1);
2319 dz31 = _mm256_sub_pd(iz3,jz1);
2320 dx32 = _mm256_sub_pd(ix3,jx2);
2321 dy32 = _mm256_sub_pd(iy3,jy2);
2322 dz32 = _mm256_sub_pd(iz3,jz2);
2323 dx33 = _mm256_sub_pd(ix3,jx3);
2324 dy33 = _mm256_sub_pd(iy3,jy3);
2325 dz33 = _mm256_sub_pd(iz3,jz3);
2327 /* Calculate squared distance and things based on it */
2328 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
2329 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
2330 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
2331 rsq13 = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
2332 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
2333 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
2334 rsq23 = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
2335 rsq31 = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
2336 rsq32 = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
2337 rsq33 = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
2339 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
2340 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
2341 rinv13 = gmx_mm256_invsqrt_pd(rsq13);
2342 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
2343 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
2344 rinv23 = gmx_mm256_invsqrt_pd(rsq23);
2345 rinv31 = gmx_mm256_invsqrt_pd(rsq31);
2346 rinv32 = gmx_mm256_invsqrt_pd(rsq32);
2347 rinv33 = gmx_mm256_invsqrt_pd(rsq33);
2349 rinvsq00 = gmx_mm256_inv_pd(rsq00);
2350 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
2351 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
2352 rinvsq13 = _mm256_mul_pd(rinv13,rinv13);
2353 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
2354 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
2355 rinvsq23 = _mm256_mul_pd(rinv23,rinv23);
2356 rinvsq31 = _mm256_mul_pd(rinv31,rinv31);
2357 rinvsq32 = _mm256_mul_pd(rinv32,rinv32);
2358 rinvsq33 = _mm256_mul_pd(rinv33,rinv33);
2360 fjx0 = _mm256_setzero_pd();
2361 fjy0 = _mm256_setzero_pd();
2362 fjz0 = _mm256_setzero_pd();
2363 fjx1 = _mm256_setzero_pd();
2364 fjy1 = _mm256_setzero_pd();
2365 fjz1 = _mm256_setzero_pd();
2366 fjx2 = _mm256_setzero_pd();
2367 fjy2 = _mm256_setzero_pd();
2368 fjz2 = _mm256_setzero_pd();
2369 fjx3 = _mm256_setzero_pd();
2370 fjy3 = _mm256_setzero_pd();
2371 fjz3 = _mm256_setzero_pd();
2373 /**************************
2374 * CALCULATE INTERACTIONS *
2375 **************************/
2377 if (gmx_mm256_any_lt(rsq00,rcutoff2))
2380 /* LENNARD-JONES DISPERSION/REPULSION */
2382 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2383 fvdw = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
2385 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
2389 fscal = _mm256_and_pd(fscal,cutoff_mask);
2391 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2393 /* Calculate temporary vectorial force */
2394 tx = _mm256_mul_pd(fscal,dx00);
2395 ty = _mm256_mul_pd(fscal,dy00);
2396 tz = _mm256_mul_pd(fscal,dz00);
2398 /* Update vectorial force */
2399 fix0 = _mm256_add_pd(fix0,tx);
2400 fiy0 = _mm256_add_pd(fiy0,ty);
2401 fiz0 = _mm256_add_pd(fiz0,tz);
2403 fjx0 = _mm256_add_pd(fjx0,tx);
2404 fjy0 = _mm256_add_pd(fjy0,ty);
2405 fjz0 = _mm256_add_pd(fjz0,tz);
2409 /**************************
2410 * CALCULATE INTERACTIONS *
2411 **************************/
2413 if (gmx_mm256_any_lt(rsq11,rcutoff2))
2416 r11 = _mm256_mul_pd(rsq11,rinv11);
2417 r11 = _mm256_andnot_pd(dummy_mask,r11);
2419 /* EWALD ELECTROSTATICS */
2421 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2422 ewrt = _mm256_mul_pd(r11,ewtabscale);
2423 ewitab = _mm256_cvttpd_epi32(ewrt);
2424 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2425 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2426 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2428 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2429 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
2431 cutoff_mask = _mm256_cmp_pd(rsq11,rcutoff2,_CMP_LT_OQ);
2435 fscal = _mm256_and_pd(fscal,cutoff_mask);
2437 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2439 /* Calculate temporary vectorial force */
2440 tx = _mm256_mul_pd(fscal,dx11);
2441 ty = _mm256_mul_pd(fscal,dy11);
2442 tz = _mm256_mul_pd(fscal,dz11);
2444 /* Update vectorial force */
2445 fix1 = _mm256_add_pd(fix1,tx);
2446 fiy1 = _mm256_add_pd(fiy1,ty);
2447 fiz1 = _mm256_add_pd(fiz1,tz);
2449 fjx1 = _mm256_add_pd(fjx1,tx);
2450 fjy1 = _mm256_add_pd(fjy1,ty);
2451 fjz1 = _mm256_add_pd(fjz1,tz);
2455 /**************************
2456 * CALCULATE INTERACTIONS *
2457 **************************/
2459 if (gmx_mm256_any_lt(rsq12,rcutoff2))
2462 r12 = _mm256_mul_pd(rsq12,rinv12);
2463 r12 = _mm256_andnot_pd(dummy_mask,r12);
2465 /* EWALD ELECTROSTATICS */
2467 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2468 ewrt = _mm256_mul_pd(r12,ewtabscale);
2469 ewitab = _mm256_cvttpd_epi32(ewrt);
2470 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2471 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2472 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2474 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2475 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2477 cutoff_mask = _mm256_cmp_pd(rsq12,rcutoff2,_CMP_LT_OQ);
2481 fscal = _mm256_and_pd(fscal,cutoff_mask);
2483 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2485 /* Calculate temporary vectorial force */
2486 tx = _mm256_mul_pd(fscal,dx12);
2487 ty = _mm256_mul_pd(fscal,dy12);
2488 tz = _mm256_mul_pd(fscal,dz12);
2490 /* Update vectorial force */
2491 fix1 = _mm256_add_pd(fix1,tx);
2492 fiy1 = _mm256_add_pd(fiy1,ty);
2493 fiz1 = _mm256_add_pd(fiz1,tz);
2495 fjx2 = _mm256_add_pd(fjx2,tx);
2496 fjy2 = _mm256_add_pd(fjy2,ty);
2497 fjz2 = _mm256_add_pd(fjz2,tz);
2501 /**************************
2502 * CALCULATE INTERACTIONS *
2503 **************************/
2505 if (gmx_mm256_any_lt(rsq13,rcutoff2))
2508 r13 = _mm256_mul_pd(rsq13,rinv13);
2509 r13 = _mm256_andnot_pd(dummy_mask,r13);
2511 /* EWALD ELECTROSTATICS */
2513 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2514 ewrt = _mm256_mul_pd(r13,ewtabscale);
2515 ewitab = _mm256_cvttpd_epi32(ewrt);
2516 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2517 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2518 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2520 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2521 felec = _mm256_mul_pd(_mm256_mul_pd(qq13,rinv13),_mm256_sub_pd(rinvsq13,felec));
2523 cutoff_mask = _mm256_cmp_pd(rsq13,rcutoff2,_CMP_LT_OQ);
2527 fscal = _mm256_and_pd(fscal,cutoff_mask);
2529 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2531 /* Calculate temporary vectorial force */
2532 tx = _mm256_mul_pd(fscal,dx13);
2533 ty = _mm256_mul_pd(fscal,dy13);
2534 tz = _mm256_mul_pd(fscal,dz13);
2536 /* Update vectorial force */
2537 fix1 = _mm256_add_pd(fix1,tx);
2538 fiy1 = _mm256_add_pd(fiy1,ty);
2539 fiz1 = _mm256_add_pd(fiz1,tz);
2541 fjx3 = _mm256_add_pd(fjx3,tx);
2542 fjy3 = _mm256_add_pd(fjy3,ty);
2543 fjz3 = _mm256_add_pd(fjz3,tz);
2547 /**************************
2548 * CALCULATE INTERACTIONS *
2549 **************************/
2551 if (gmx_mm256_any_lt(rsq21,rcutoff2))
2554 r21 = _mm256_mul_pd(rsq21,rinv21);
2555 r21 = _mm256_andnot_pd(dummy_mask,r21);
2557 /* EWALD ELECTROSTATICS */
2559 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2560 ewrt = _mm256_mul_pd(r21,ewtabscale);
2561 ewitab = _mm256_cvttpd_epi32(ewrt);
2562 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2563 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2564 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2566 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2567 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2569 cutoff_mask = _mm256_cmp_pd(rsq21,rcutoff2,_CMP_LT_OQ);
2573 fscal = _mm256_and_pd(fscal,cutoff_mask);
2575 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2577 /* Calculate temporary vectorial force */
2578 tx = _mm256_mul_pd(fscal,dx21);
2579 ty = _mm256_mul_pd(fscal,dy21);
2580 tz = _mm256_mul_pd(fscal,dz21);
2582 /* Update vectorial force */
2583 fix2 = _mm256_add_pd(fix2,tx);
2584 fiy2 = _mm256_add_pd(fiy2,ty);
2585 fiz2 = _mm256_add_pd(fiz2,tz);
2587 fjx1 = _mm256_add_pd(fjx1,tx);
2588 fjy1 = _mm256_add_pd(fjy1,ty);
2589 fjz1 = _mm256_add_pd(fjz1,tz);
2593 /**************************
2594 * CALCULATE INTERACTIONS *
2595 **************************/
2597 if (gmx_mm256_any_lt(rsq22,rcutoff2))
2600 r22 = _mm256_mul_pd(rsq22,rinv22);
2601 r22 = _mm256_andnot_pd(dummy_mask,r22);
2603 /* EWALD ELECTROSTATICS */
2605 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2606 ewrt = _mm256_mul_pd(r22,ewtabscale);
2607 ewitab = _mm256_cvttpd_epi32(ewrt);
2608 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2609 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2610 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2612 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2613 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2615 cutoff_mask = _mm256_cmp_pd(rsq22,rcutoff2,_CMP_LT_OQ);
2619 fscal = _mm256_and_pd(fscal,cutoff_mask);
2621 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2623 /* Calculate temporary vectorial force */
2624 tx = _mm256_mul_pd(fscal,dx22);
2625 ty = _mm256_mul_pd(fscal,dy22);
2626 tz = _mm256_mul_pd(fscal,dz22);
2628 /* Update vectorial force */
2629 fix2 = _mm256_add_pd(fix2,tx);
2630 fiy2 = _mm256_add_pd(fiy2,ty);
2631 fiz2 = _mm256_add_pd(fiz2,tz);
2633 fjx2 = _mm256_add_pd(fjx2,tx);
2634 fjy2 = _mm256_add_pd(fjy2,ty);
2635 fjz2 = _mm256_add_pd(fjz2,tz);
2639 /**************************
2640 * CALCULATE INTERACTIONS *
2641 **************************/
2643 if (gmx_mm256_any_lt(rsq23,rcutoff2))
2646 r23 = _mm256_mul_pd(rsq23,rinv23);
2647 r23 = _mm256_andnot_pd(dummy_mask,r23);
2649 /* EWALD ELECTROSTATICS */
2651 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2652 ewrt = _mm256_mul_pd(r23,ewtabscale);
2653 ewitab = _mm256_cvttpd_epi32(ewrt);
2654 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2655 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2656 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2658 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2659 felec = _mm256_mul_pd(_mm256_mul_pd(qq23,rinv23),_mm256_sub_pd(rinvsq23,felec));
2661 cutoff_mask = _mm256_cmp_pd(rsq23,rcutoff2,_CMP_LT_OQ);
2665 fscal = _mm256_and_pd(fscal,cutoff_mask);
2667 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2669 /* Calculate temporary vectorial force */
2670 tx = _mm256_mul_pd(fscal,dx23);
2671 ty = _mm256_mul_pd(fscal,dy23);
2672 tz = _mm256_mul_pd(fscal,dz23);
2674 /* Update vectorial force */
2675 fix2 = _mm256_add_pd(fix2,tx);
2676 fiy2 = _mm256_add_pd(fiy2,ty);
2677 fiz2 = _mm256_add_pd(fiz2,tz);
2679 fjx3 = _mm256_add_pd(fjx3,tx);
2680 fjy3 = _mm256_add_pd(fjy3,ty);
2681 fjz3 = _mm256_add_pd(fjz3,tz);
2685 /**************************
2686 * CALCULATE INTERACTIONS *
2687 **************************/
2689 if (gmx_mm256_any_lt(rsq31,rcutoff2))
2692 r31 = _mm256_mul_pd(rsq31,rinv31);
2693 r31 = _mm256_andnot_pd(dummy_mask,r31);
2695 /* EWALD ELECTROSTATICS */
2697 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2698 ewrt = _mm256_mul_pd(r31,ewtabscale);
2699 ewitab = _mm256_cvttpd_epi32(ewrt);
2700 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2701 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2702 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2704 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2705 felec = _mm256_mul_pd(_mm256_mul_pd(qq31,rinv31),_mm256_sub_pd(rinvsq31,felec));
2707 cutoff_mask = _mm256_cmp_pd(rsq31,rcutoff2,_CMP_LT_OQ);
2711 fscal = _mm256_and_pd(fscal,cutoff_mask);
2713 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2715 /* Calculate temporary vectorial force */
2716 tx = _mm256_mul_pd(fscal,dx31);
2717 ty = _mm256_mul_pd(fscal,dy31);
2718 tz = _mm256_mul_pd(fscal,dz31);
2720 /* Update vectorial force */
2721 fix3 = _mm256_add_pd(fix3,tx);
2722 fiy3 = _mm256_add_pd(fiy3,ty);
2723 fiz3 = _mm256_add_pd(fiz3,tz);
2725 fjx1 = _mm256_add_pd(fjx1,tx);
2726 fjy1 = _mm256_add_pd(fjy1,ty);
2727 fjz1 = _mm256_add_pd(fjz1,tz);
2731 /**************************
2732 * CALCULATE INTERACTIONS *
2733 **************************/
2735 if (gmx_mm256_any_lt(rsq32,rcutoff2))
2738 r32 = _mm256_mul_pd(rsq32,rinv32);
2739 r32 = _mm256_andnot_pd(dummy_mask,r32);
2741 /* EWALD ELECTROSTATICS */
2743 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2744 ewrt = _mm256_mul_pd(r32,ewtabscale);
2745 ewitab = _mm256_cvttpd_epi32(ewrt);
2746 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2747 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2748 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2750 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2751 felec = _mm256_mul_pd(_mm256_mul_pd(qq32,rinv32),_mm256_sub_pd(rinvsq32,felec));
2753 cutoff_mask = _mm256_cmp_pd(rsq32,rcutoff2,_CMP_LT_OQ);
2757 fscal = _mm256_and_pd(fscal,cutoff_mask);
2759 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2761 /* Calculate temporary vectorial force */
2762 tx = _mm256_mul_pd(fscal,dx32);
2763 ty = _mm256_mul_pd(fscal,dy32);
2764 tz = _mm256_mul_pd(fscal,dz32);
2766 /* Update vectorial force */
2767 fix3 = _mm256_add_pd(fix3,tx);
2768 fiy3 = _mm256_add_pd(fiy3,ty);
2769 fiz3 = _mm256_add_pd(fiz3,tz);
2771 fjx2 = _mm256_add_pd(fjx2,tx);
2772 fjy2 = _mm256_add_pd(fjy2,ty);
2773 fjz2 = _mm256_add_pd(fjz2,tz);
2777 /**************************
2778 * CALCULATE INTERACTIONS *
2779 **************************/
2781 if (gmx_mm256_any_lt(rsq33,rcutoff2))
2784 r33 = _mm256_mul_pd(rsq33,rinv33);
2785 r33 = _mm256_andnot_pd(dummy_mask,r33);
2787 /* EWALD ELECTROSTATICS */
2789 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2790 ewrt = _mm256_mul_pd(r33,ewtabscale);
2791 ewitab = _mm256_cvttpd_epi32(ewrt);
2792 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2793 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2794 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2796 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2797 felec = _mm256_mul_pd(_mm256_mul_pd(qq33,rinv33),_mm256_sub_pd(rinvsq33,felec));
2799 cutoff_mask = _mm256_cmp_pd(rsq33,rcutoff2,_CMP_LT_OQ);
2803 fscal = _mm256_and_pd(fscal,cutoff_mask);
2805 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2807 /* Calculate temporary vectorial force */
2808 tx = _mm256_mul_pd(fscal,dx33);
2809 ty = _mm256_mul_pd(fscal,dy33);
2810 tz = _mm256_mul_pd(fscal,dz33);
2812 /* Update vectorial force */
2813 fix3 = _mm256_add_pd(fix3,tx);
2814 fiy3 = _mm256_add_pd(fiy3,ty);
2815 fiz3 = _mm256_add_pd(fiz3,tz);
2817 fjx3 = _mm256_add_pd(fjx3,tx);
2818 fjy3 = _mm256_add_pd(fjy3,ty);
2819 fjz3 = _mm256_add_pd(fjz3,tz);
2823 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2824 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2825 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2826 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2828 gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2829 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2830 fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2832 /* Inner loop uses 393 flops */
2835 /* End of innermost loop */
2837 gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2838 f+i_coord_offset,fshift+i_shift_offset);
2840 /* Increment number of inner iterations */
2841 inneriter += j_index_end - j_index_start;
2843 /* Outer loop uses 24 flops */
2846 /* Increment number of outer iterations */
2849 /* Update outer/inner flops */
2851 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*393);