2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS avx_256_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "gromacs/simd/math_x86_avx_256_double.h"
48 #include "kernelutil_x86_avx_256_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3W3_VF_avx_256_double
52 * Electrostatics interaction: Ewald
53 * VdW interaction: LJEwald
54 * Geometry: Water3-Water3
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEw_VdwLJEw_GeomW3W3_VF_avx_256_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
76 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
77 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
83 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
84 real * vdwioffsetptr0;
85 real * vdwgridioffsetptr0;
86 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
87 real * vdwioffsetptr1;
88 real * vdwgridioffsetptr1;
89 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
90 real * vdwioffsetptr2;
91 real * vdwgridioffsetptr2;
92 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
93 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
94 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
95 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
96 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
97 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
98 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
99 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
100 __m256d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
101 __m256d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
102 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
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 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
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 velec,felec,velecsum,facel,crf,krf,krf2;
111 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
114 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
115 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
126 __m256d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
127 __m256d one_half = _mm256_set1_pd(0.5);
128 __m256d minus_one = _mm256_set1_pd(-1.0);
130 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
131 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
133 __m256d dummy_mask,cutoff_mask;
134 __m128 tmpmask0,tmpmask1;
135 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
136 __m256d one = _mm256_set1_pd(1.0);
137 __m256d two = _mm256_set1_pd(2.0);
143 jindex = nlist->jindex;
145 shiftidx = nlist->shift;
147 shiftvec = fr->shift_vec[0];
148 fshift = fr->fshift[0];
149 facel = _mm256_set1_pd(fr->epsfac);
150 charge = mdatoms->chargeA;
151 nvdwtype = fr->ntype;
153 vdwtype = mdatoms->typeA;
154 vdwgridparam = fr->ljpme_c6grid;
155 sh_lj_ewald = _mm256_set1_pd(fr->ic->sh_lj_ewald);
156 ewclj = _mm256_set1_pd(fr->ewaldcoeff_lj);
157 ewclj2 = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
159 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
160 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
161 beta2 = _mm256_mul_pd(beta,beta);
162 beta3 = _mm256_mul_pd(beta,beta2);
164 ewtab = fr->ic->tabq_coul_FDV0;
165 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
166 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
168 /* Setup water-specific parameters */
169 inr = nlist->iinr[0];
170 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
171 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
172 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
173 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
174 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
176 jq0 = _mm256_set1_pd(charge[inr+0]);
177 jq1 = _mm256_set1_pd(charge[inr+1]);
178 jq2 = _mm256_set1_pd(charge[inr+2]);
179 vdwjidx0A = 2*vdwtype[inr+0];
180 qq00 = _mm256_mul_pd(iq0,jq0);
181 c6_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
182 c12_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
183 c6grid_00 = _mm256_set1_pd(vdwgridioffsetptr0[vdwjidx0A]);
184 qq01 = _mm256_mul_pd(iq0,jq1);
185 qq02 = _mm256_mul_pd(iq0,jq2);
186 qq10 = _mm256_mul_pd(iq1,jq0);
187 qq11 = _mm256_mul_pd(iq1,jq1);
188 qq12 = _mm256_mul_pd(iq1,jq2);
189 qq20 = _mm256_mul_pd(iq2,jq0);
190 qq21 = _mm256_mul_pd(iq2,jq1);
191 qq22 = _mm256_mul_pd(iq2,jq2);
193 /* Avoid stupid compiler warnings */
194 jnrA = jnrB = jnrC = jnrD = 0;
203 for(iidx=0;iidx<4*DIM;iidx++)
208 /* Start outer loop over neighborlists */
209 for(iidx=0; iidx<nri; iidx++)
211 /* Load shift vector for this list */
212 i_shift_offset = DIM*shiftidx[iidx];
214 /* Load limits for loop over neighbors */
215 j_index_start = jindex[iidx];
216 j_index_end = jindex[iidx+1];
218 /* Get outer coordinate index */
220 i_coord_offset = DIM*inr;
222 /* Load i particle coords and add shift vector */
223 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
224 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
226 fix0 = _mm256_setzero_pd();
227 fiy0 = _mm256_setzero_pd();
228 fiz0 = _mm256_setzero_pd();
229 fix1 = _mm256_setzero_pd();
230 fiy1 = _mm256_setzero_pd();
231 fiz1 = _mm256_setzero_pd();
232 fix2 = _mm256_setzero_pd();
233 fiy2 = _mm256_setzero_pd();
234 fiz2 = _mm256_setzero_pd();
236 /* Reset potential sums */
237 velecsum = _mm256_setzero_pd();
238 vvdwsum = _mm256_setzero_pd();
240 /* Start inner kernel loop */
241 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
244 /* Get j neighbor index, and coordinate index */
249 j_coord_offsetA = DIM*jnrA;
250 j_coord_offsetB = DIM*jnrB;
251 j_coord_offsetC = DIM*jnrC;
252 j_coord_offsetD = DIM*jnrD;
254 /* load j atom coordinates */
255 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
256 x+j_coord_offsetC,x+j_coord_offsetD,
257 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
259 /* Calculate displacement vector */
260 dx00 = _mm256_sub_pd(ix0,jx0);
261 dy00 = _mm256_sub_pd(iy0,jy0);
262 dz00 = _mm256_sub_pd(iz0,jz0);
263 dx01 = _mm256_sub_pd(ix0,jx1);
264 dy01 = _mm256_sub_pd(iy0,jy1);
265 dz01 = _mm256_sub_pd(iz0,jz1);
266 dx02 = _mm256_sub_pd(ix0,jx2);
267 dy02 = _mm256_sub_pd(iy0,jy2);
268 dz02 = _mm256_sub_pd(iz0,jz2);
269 dx10 = _mm256_sub_pd(ix1,jx0);
270 dy10 = _mm256_sub_pd(iy1,jy0);
271 dz10 = _mm256_sub_pd(iz1,jz0);
272 dx11 = _mm256_sub_pd(ix1,jx1);
273 dy11 = _mm256_sub_pd(iy1,jy1);
274 dz11 = _mm256_sub_pd(iz1,jz1);
275 dx12 = _mm256_sub_pd(ix1,jx2);
276 dy12 = _mm256_sub_pd(iy1,jy2);
277 dz12 = _mm256_sub_pd(iz1,jz2);
278 dx20 = _mm256_sub_pd(ix2,jx0);
279 dy20 = _mm256_sub_pd(iy2,jy0);
280 dz20 = _mm256_sub_pd(iz2,jz0);
281 dx21 = _mm256_sub_pd(ix2,jx1);
282 dy21 = _mm256_sub_pd(iy2,jy1);
283 dz21 = _mm256_sub_pd(iz2,jz1);
284 dx22 = _mm256_sub_pd(ix2,jx2);
285 dy22 = _mm256_sub_pd(iy2,jy2);
286 dz22 = _mm256_sub_pd(iz2,jz2);
288 /* Calculate squared distance and things based on it */
289 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
290 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
291 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
292 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
293 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
294 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
295 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
296 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
297 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
299 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
300 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
301 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
302 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
303 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
304 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
305 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
306 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
307 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
309 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
310 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
311 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
312 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
313 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
314 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
315 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
316 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
317 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
319 fjx0 = _mm256_setzero_pd();
320 fjy0 = _mm256_setzero_pd();
321 fjz0 = _mm256_setzero_pd();
322 fjx1 = _mm256_setzero_pd();
323 fjy1 = _mm256_setzero_pd();
324 fjz1 = _mm256_setzero_pd();
325 fjx2 = _mm256_setzero_pd();
326 fjy2 = _mm256_setzero_pd();
327 fjz2 = _mm256_setzero_pd();
329 /**************************
330 * CALCULATE INTERACTIONS *
331 **************************/
333 r00 = _mm256_mul_pd(rsq00,rinv00);
335 /* EWALD ELECTROSTATICS */
337 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
338 ewrt = _mm256_mul_pd(r00,ewtabscale);
339 ewitab = _mm256_cvttpd_epi32(ewrt);
340 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
341 ewitab = _mm_slli_epi32(ewitab,2);
342 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
343 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
344 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
345 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
346 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
347 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
348 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
349 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
350 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
352 /* Analytical LJ-PME */
353 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
354 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
355 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
356 exponent = gmx_simd_exp_d(ewcljrsq);
357 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
358 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
359 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
360 vvdw6 = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
361 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
362 vvdw = _mm256_sub_pd(_mm256_mul_pd(vvdw12,one_twelfth),_mm256_mul_pd(vvdw6,one_sixth));
363 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
364 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
366 /* Update potential sum for this i atom from the interaction with this j atom. */
367 velecsum = _mm256_add_pd(velecsum,velec);
368 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
370 fscal = _mm256_add_pd(felec,fvdw);
372 /* Calculate temporary vectorial force */
373 tx = _mm256_mul_pd(fscal,dx00);
374 ty = _mm256_mul_pd(fscal,dy00);
375 tz = _mm256_mul_pd(fscal,dz00);
377 /* Update vectorial force */
378 fix0 = _mm256_add_pd(fix0,tx);
379 fiy0 = _mm256_add_pd(fiy0,ty);
380 fiz0 = _mm256_add_pd(fiz0,tz);
382 fjx0 = _mm256_add_pd(fjx0,tx);
383 fjy0 = _mm256_add_pd(fjy0,ty);
384 fjz0 = _mm256_add_pd(fjz0,tz);
386 /**************************
387 * CALCULATE INTERACTIONS *
388 **************************/
390 r01 = _mm256_mul_pd(rsq01,rinv01);
392 /* EWALD ELECTROSTATICS */
394 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
395 ewrt = _mm256_mul_pd(r01,ewtabscale);
396 ewitab = _mm256_cvttpd_epi32(ewrt);
397 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
398 ewitab = _mm_slli_epi32(ewitab,2);
399 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
400 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
401 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
402 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
403 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
404 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
405 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
406 velec = _mm256_mul_pd(qq01,_mm256_sub_pd(rinv01,velec));
407 felec = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
409 /* Update potential sum for this i atom from the interaction with this j atom. */
410 velecsum = _mm256_add_pd(velecsum,velec);
414 /* Calculate temporary vectorial force */
415 tx = _mm256_mul_pd(fscal,dx01);
416 ty = _mm256_mul_pd(fscal,dy01);
417 tz = _mm256_mul_pd(fscal,dz01);
419 /* Update vectorial force */
420 fix0 = _mm256_add_pd(fix0,tx);
421 fiy0 = _mm256_add_pd(fiy0,ty);
422 fiz0 = _mm256_add_pd(fiz0,tz);
424 fjx1 = _mm256_add_pd(fjx1,tx);
425 fjy1 = _mm256_add_pd(fjy1,ty);
426 fjz1 = _mm256_add_pd(fjz1,tz);
428 /**************************
429 * CALCULATE INTERACTIONS *
430 **************************/
432 r02 = _mm256_mul_pd(rsq02,rinv02);
434 /* EWALD ELECTROSTATICS */
436 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
437 ewrt = _mm256_mul_pd(r02,ewtabscale);
438 ewitab = _mm256_cvttpd_epi32(ewrt);
439 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
440 ewitab = _mm_slli_epi32(ewitab,2);
441 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
442 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
443 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
444 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
445 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
446 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
447 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
448 velec = _mm256_mul_pd(qq02,_mm256_sub_pd(rinv02,velec));
449 felec = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
451 /* Update potential sum for this i atom from the interaction with this j atom. */
452 velecsum = _mm256_add_pd(velecsum,velec);
456 /* Calculate temporary vectorial force */
457 tx = _mm256_mul_pd(fscal,dx02);
458 ty = _mm256_mul_pd(fscal,dy02);
459 tz = _mm256_mul_pd(fscal,dz02);
461 /* Update vectorial force */
462 fix0 = _mm256_add_pd(fix0,tx);
463 fiy0 = _mm256_add_pd(fiy0,ty);
464 fiz0 = _mm256_add_pd(fiz0,tz);
466 fjx2 = _mm256_add_pd(fjx2,tx);
467 fjy2 = _mm256_add_pd(fjy2,ty);
468 fjz2 = _mm256_add_pd(fjz2,tz);
470 /**************************
471 * CALCULATE INTERACTIONS *
472 **************************/
474 r10 = _mm256_mul_pd(rsq10,rinv10);
476 /* EWALD ELECTROSTATICS */
478 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
479 ewrt = _mm256_mul_pd(r10,ewtabscale);
480 ewitab = _mm256_cvttpd_epi32(ewrt);
481 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
482 ewitab = _mm_slli_epi32(ewitab,2);
483 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
484 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
485 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
486 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
487 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
488 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
489 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
490 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
491 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
493 /* Update potential sum for this i atom from the interaction with this j atom. */
494 velecsum = _mm256_add_pd(velecsum,velec);
498 /* Calculate temporary vectorial force */
499 tx = _mm256_mul_pd(fscal,dx10);
500 ty = _mm256_mul_pd(fscal,dy10);
501 tz = _mm256_mul_pd(fscal,dz10);
503 /* Update vectorial force */
504 fix1 = _mm256_add_pd(fix1,tx);
505 fiy1 = _mm256_add_pd(fiy1,ty);
506 fiz1 = _mm256_add_pd(fiz1,tz);
508 fjx0 = _mm256_add_pd(fjx0,tx);
509 fjy0 = _mm256_add_pd(fjy0,ty);
510 fjz0 = _mm256_add_pd(fjz0,tz);
512 /**************************
513 * CALCULATE INTERACTIONS *
514 **************************/
516 r11 = _mm256_mul_pd(rsq11,rinv11);
518 /* EWALD ELECTROSTATICS */
520 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
521 ewrt = _mm256_mul_pd(r11,ewtabscale);
522 ewitab = _mm256_cvttpd_epi32(ewrt);
523 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
524 ewitab = _mm_slli_epi32(ewitab,2);
525 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
526 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
527 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
528 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
529 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
530 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
531 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
532 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
533 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
535 /* Update potential sum for this i atom from the interaction with this j atom. */
536 velecsum = _mm256_add_pd(velecsum,velec);
540 /* Calculate temporary vectorial force */
541 tx = _mm256_mul_pd(fscal,dx11);
542 ty = _mm256_mul_pd(fscal,dy11);
543 tz = _mm256_mul_pd(fscal,dz11);
545 /* Update vectorial force */
546 fix1 = _mm256_add_pd(fix1,tx);
547 fiy1 = _mm256_add_pd(fiy1,ty);
548 fiz1 = _mm256_add_pd(fiz1,tz);
550 fjx1 = _mm256_add_pd(fjx1,tx);
551 fjy1 = _mm256_add_pd(fjy1,ty);
552 fjz1 = _mm256_add_pd(fjz1,tz);
554 /**************************
555 * CALCULATE INTERACTIONS *
556 **************************/
558 r12 = _mm256_mul_pd(rsq12,rinv12);
560 /* EWALD ELECTROSTATICS */
562 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
563 ewrt = _mm256_mul_pd(r12,ewtabscale);
564 ewitab = _mm256_cvttpd_epi32(ewrt);
565 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
566 ewitab = _mm_slli_epi32(ewitab,2);
567 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
568 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
569 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
570 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
571 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
572 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
573 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
574 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
575 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
577 /* Update potential sum for this i atom from the interaction with this j atom. */
578 velecsum = _mm256_add_pd(velecsum,velec);
582 /* Calculate temporary vectorial force */
583 tx = _mm256_mul_pd(fscal,dx12);
584 ty = _mm256_mul_pd(fscal,dy12);
585 tz = _mm256_mul_pd(fscal,dz12);
587 /* Update vectorial force */
588 fix1 = _mm256_add_pd(fix1,tx);
589 fiy1 = _mm256_add_pd(fiy1,ty);
590 fiz1 = _mm256_add_pd(fiz1,tz);
592 fjx2 = _mm256_add_pd(fjx2,tx);
593 fjy2 = _mm256_add_pd(fjy2,ty);
594 fjz2 = _mm256_add_pd(fjz2,tz);
596 /**************************
597 * CALCULATE INTERACTIONS *
598 **************************/
600 r20 = _mm256_mul_pd(rsq20,rinv20);
602 /* EWALD ELECTROSTATICS */
604 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
605 ewrt = _mm256_mul_pd(r20,ewtabscale);
606 ewitab = _mm256_cvttpd_epi32(ewrt);
607 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
608 ewitab = _mm_slli_epi32(ewitab,2);
609 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
610 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
611 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
612 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
613 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
614 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
615 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
616 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
617 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
619 /* Update potential sum for this i atom from the interaction with this j atom. */
620 velecsum = _mm256_add_pd(velecsum,velec);
624 /* Calculate temporary vectorial force */
625 tx = _mm256_mul_pd(fscal,dx20);
626 ty = _mm256_mul_pd(fscal,dy20);
627 tz = _mm256_mul_pd(fscal,dz20);
629 /* Update vectorial force */
630 fix2 = _mm256_add_pd(fix2,tx);
631 fiy2 = _mm256_add_pd(fiy2,ty);
632 fiz2 = _mm256_add_pd(fiz2,tz);
634 fjx0 = _mm256_add_pd(fjx0,tx);
635 fjy0 = _mm256_add_pd(fjy0,ty);
636 fjz0 = _mm256_add_pd(fjz0,tz);
638 /**************************
639 * CALCULATE INTERACTIONS *
640 **************************/
642 r21 = _mm256_mul_pd(rsq21,rinv21);
644 /* EWALD ELECTROSTATICS */
646 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
647 ewrt = _mm256_mul_pd(r21,ewtabscale);
648 ewitab = _mm256_cvttpd_epi32(ewrt);
649 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
650 ewitab = _mm_slli_epi32(ewitab,2);
651 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
652 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
653 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
654 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
655 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
656 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
657 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
658 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
659 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
661 /* Update potential sum for this i atom from the interaction with this j atom. */
662 velecsum = _mm256_add_pd(velecsum,velec);
666 /* Calculate temporary vectorial force */
667 tx = _mm256_mul_pd(fscal,dx21);
668 ty = _mm256_mul_pd(fscal,dy21);
669 tz = _mm256_mul_pd(fscal,dz21);
671 /* Update vectorial force */
672 fix2 = _mm256_add_pd(fix2,tx);
673 fiy2 = _mm256_add_pd(fiy2,ty);
674 fiz2 = _mm256_add_pd(fiz2,tz);
676 fjx1 = _mm256_add_pd(fjx1,tx);
677 fjy1 = _mm256_add_pd(fjy1,ty);
678 fjz1 = _mm256_add_pd(fjz1,tz);
680 /**************************
681 * CALCULATE INTERACTIONS *
682 **************************/
684 r22 = _mm256_mul_pd(rsq22,rinv22);
686 /* EWALD ELECTROSTATICS */
688 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
689 ewrt = _mm256_mul_pd(r22,ewtabscale);
690 ewitab = _mm256_cvttpd_epi32(ewrt);
691 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
692 ewitab = _mm_slli_epi32(ewitab,2);
693 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
694 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
695 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
696 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
697 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
698 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
699 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
700 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
701 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
703 /* Update potential sum for this i atom from the interaction with this j atom. */
704 velecsum = _mm256_add_pd(velecsum,velec);
708 /* Calculate temporary vectorial force */
709 tx = _mm256_mul_pd(fscal,dx22);
710 ty = _mm256_mul_pd(fscal,dy22);
711 tz = _mm256_mul_pd(fscal,dz22);
713 /* Update vectorial force */
714 fix2 = _mm256_add_pd(fix2,tx);
715 fiy2 = _mm256_add_pd(fiy2,ty);
716 fiz2 = _mm256_add_pd(fiz2,tz);
718 fjx2 = _mm256_add_pd(fjx2,tx);
719 fjy2 = _mm256_add_pd(fjy2,ty);
720 fjz2 = _mm256_add_pd(fjz2,tz);
722 fjptrA = f+j_coord_offsetA;
723 fjptrB = f+j_coord_offsetB;
724 fjptrC = f+j_coord_offsetC;
725 fjptrD = f+j_coord_offsetD;
727 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
728 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
730 /* Inner loop uses 400 flops */
736 /* Get j neighbor index, and coordinate index */
737 jnrlistA = jjnr[jidx];
738 jnrlistB = jjnr[jidx+1];
739 jnrlistC = jjnr[jidx+2];
740 jnrlistD = jjnr[jidx+3];
741 /* Sign of each element will be negative for non-real atoms.
742 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
743 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
745 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
747 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
748 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
749 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
751 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
752 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
753 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
754 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
755 j_coord_offsetA = DIM*jnrA;
756 j_coord_offsetB = DIM*jnrB;
757 j_coord_offsetC = DIM*jnrC;
758 j_coord_offsetD = DIM*jnrD;
760 /* load j atom coordinates */
761 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
762 x+j_coord_offsetC,x+j_coord_offsetD,
763 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
765 /* Calculate displacement vector */
766 dx00 = _mm256_sub_pd(ix0,jx0);
767 dy00 = _mm256_sub_pd(iy0,jy0);
768 dz00 = _mm256_sub_pd(iz0,jz0);
769 dx01 = _mm256_sub_pd(ix0,jx1);
770 dy01 = _mm256_sub_pd(iy0,jy1);
771 dz01 = _mm256_sub_pd(iz0,jz1);
772 dx02 = _mm256_sub_pd(ix0,jx2);
773 dy02 = _mm256_sub_pd(iy0,jy2);
774 dz02 = _mm256_sub_pd(iz0,jz2);
775 dx10 = _mm256_sub_pd(ix1,jx0);
776 dy10 = _mm256_sub_pd(iy1,jy0);
777 dz10 = _mm256_sub_pd(iz1,jz0);
778 dx11 = _mm256_sub_pd(ix1,jx1);
779 dy11 = _mm256_sub_pd(iy1,jy1);
780 dz11 = _mm256_sub_pd(iz1,jz1);
781 dx12 = _mm256_sub_pd(ix1,jx2);
782 dy12 = _mm256_sub_pd(iy1,jy2);
783 dz12 = _mm256_sub_pd(iz1,jz2);
784 dx20 = _mm256_sub_pd(ix2,jx0);
785 dy20 = _mm256_sub_pd(iy2,jy0);
786 dz20 = _mm256_sub_pd(iz2,jz0);
787 dx21 = _mm256_sub_pd(ix2,jx1);
788 dy21 = _mm256_sub_pd(iy2,jy1);
789 dz21 = _mm256_sub_pd(iz2,jz1);
790 dx22 = _mm256_sub_pd(ix2,jx2);
791 dy22 = _mm256_sub_pd(iy2,jy2);
792 dz22 = _mm256_sub_pd(iz2,jz2);
794 /* Calculate squared distance and things based on it */
795 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
796 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
797 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
798 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
799 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
800 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
801 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
802 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
803 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
805 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
806 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
807 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
808 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
809 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
810 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
811 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
812 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
813 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
815 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
816 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
817 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
818 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
819 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
820 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
821 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
822 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
823 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
825 fjx0 = _mm256_setzero_pd();
826 fjy0 = _mm256_setzero_pd();
827 fjz0 = _mm256_setzero_pd();
828 fjx1 = _mm256_setzero_pd();
829 fjy1 = _mm256_setzero_pd();
830 fjz1 = _mm256_setzero_pd();
831 fjx2 = _mm256_setzero_pd();
832 fjy2 = _mm256_setzero_pd();
833 fjz2 = _mm256_setzero_pd();
835 /**************************
836 * CALCULATE INTERACTIONS *
837 **************************/
839 r00 = _mm256_mul_pd(rsq00,rinv00);
840 r00 = _mm256_andnot_pd(dummy_mask,r00);
842 /* EWALD ELECTROSTATICS */
844 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
845 ewrt = _mm256_mul_pd(r00,ewtabscale);
846 ewitab = _mm256_cvttpd_epi32(ewrt);
847 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
848 ewitab = _mm_slli_epi32(ewitab,2);
849 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
850 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
851 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
852 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
853 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
854 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
855 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
856 velec = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
857 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
859 /* Analytical LJ-PME */
860 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
861 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
862 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
863 exponent = gmx_simd_exp_d(ewcljrsq);
864 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
865 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
866 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
867 vvdw6 = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
868 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
869 vvdw = _mm256_sub_pd(_mm256_mul_pd(vvdw12,one_twelfth),_mm256_mul_pd(vvdw6,one_sixth));
870 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
871 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
873 /* Update potential sum for this i atom from the interaction with this j atom. */
874 velec = _mm256_andnot_pd(dummy_mask,velec);
875 velecsum = _mm256_add_pd(velecsum,velec);
876 vvdw = _mm256_andnot_pd(dummy_mask,vvdw);
877 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
879 fscal = _mm256_add_pd(felec,fvdw);
881 fscal = _mm256_andnot_pd(dummy_mask,fscal);
883 /* Calculate temporary vectorial force */
884 tx = _mm256_mul_pd(fscal,dx00);
885 ty = _mm256_mul_pd(fscal,dy00);
886 tz = _mm256_mul_pd(fscal,dz00);
888 /* Update vectorial force */
889 fix0 = _mm256_add_pd(fix0,tx);
890 fiy0 = _mm256_add_pd(fiy0,ty);
891 fiz0 = _mm256_add_pd(fiz0,tz);
893 fjx0 = _mm256_add_pd(fjx0,tx);
894 fjy0 = _mm256_add_pd(fjy0,ty);
895 fjz0 = _mm256_add_pd(fjz0,tz);
897 /**************************
898 * CALCULATE INTERACTIONS *
899 **************************/
901 r01 = _mm256_mul_pd(rsq01,rinv01);
902 r01 = _mm256_andnot_pd(dummy_mask,r01);
904 /* EWALD ELECTROSTATICS */
906 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
907 ewrt = _mm256_mul_pd(r01,ewtabscale);
908 ewitab = _mm256_cvttpd_epi32(ewrt);
909 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
910 ewitab = _mm_slli_epi32(ewitab,2);
911 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
912 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
913 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
914 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
915 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
916 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
917 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
918 velec = _mm256_mul_pd(qq01,_mm256_sub_pd(rinv01,velec));
919 felec = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
921 /* Update potential sum for this i atom from the interaction with this j atom. */
922 velec = _mm256_andnot_pd(dummy_mask,velec);
923 velecsum = _mm256_add_pd(velecsum,velec);
927 fscal = _mm256_andnot_pd(dummy_mask,fscal);
929 /* Calculate temporary vectorial force */
930 tx = _mm256_mul_pd(fscal,dx01);
931 ty = _mm256_mul_pd(fscal,dy01);
932 tz = _mm256_mul_pd(fscal,dz01);
934 /* Update vectorial force */
935 fix0 = _mm256_add_pd(fix0,tx);
936 fiy0 = _mm256_add_pd(fiy0,ty);
937 fiz0 = _mm256_add_pd(fiz0,tz);
939 fjx1 = _mm256_add_pd(fjx1,tx);
940 fjy1 = _mm256_add_pd(fjy1,ty);
941 fjz1 = _mm256_add_pd(fjz1,tz);
943 /**************************
944 * CALCULATE INTERACTIONS *
945 **************************/
947 r02 = _mm256_mul_pd(rsq02,rinv02);
948 r02 = _mm256_andnot_pd(dummy_mask,r02);
950 /* EWALD ELECTROSTATICS */
952 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
953 ewrt = _mm256_mul_pd(r02,ewtabscale);
954 ewitab = _mm256_cvttpd_epi32(ewrt);
955 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
956 ewitab = _mm_slli_epi32(ewitab,2);
957 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
958 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
959 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
960 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
961 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
962 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
963 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
964 velec = _mm256_mul_pd(qq02,_mm256_sub_pd(rinv02,velec));
965 felec = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
967 /* Update potential sum for this i atom from the interaction with this j atom. */
968 velec = _mm256_andnot_pd(dummy_mask,velec);
969 velecsum = _mm256_add_pd(velecsum,velec);
973 fscal = _mm256_andnot_pd(dummy_mask,fscal);
975 /* Calculate temporary vectorial force */
976 tx = _mm256_mul_pd(fscal,dx02);
977 ty = _mm256_mul_pd(fscal,dy02);
978 tz = _mm256_mul_pd(fscal,dz02);
980 /* Update vectorial force */
981 fix0 = _mm256_add_pd(fix0,tx);
982 fiy0 = _mm256_add_pd(fiy0,ty);
983 fiz0 = _mm256_add_pd(fiz0,tz);
985 fjx2 = _mm256_add_pd(fjx2,tx);
986 fjy2 = _mm256_add_pd(fjy2,ty);
987 fjz2 = _mm256_add_pd(fjz2,tz);
989 /**************************
990 * CALCULATE INTERACTIONS *
991 **************************/
993 r10 = _mm256_mul_pd(rsq10,rinv10);
994 r10 = _mm256_andnot_pd(dummy_mask,r10);
996 /* EWALD ELECTROSTATICS */
998 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
999 ewrt = _mm256_mul_pd(r10,ewtabscale);
1000 ewitab = _mm256_cvttpd_epi32(ewrt);
1001 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1002 ewitab = _mm_slli_epi32(ewitab,2);
1003 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1004 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1005 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1006 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1007 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1008 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1009 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1010 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1011 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1013 /* Update potential sum for this i atom from the interaction with this j atom. */
1014 velec = _mm256_andnot_pd(dummy_mask,velec);
1015 velecsum = _mm256_add_pd(velecsum,velec);
1019 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1021 /* Calculate temporary vectorial force */
1022 tx = _mm256_mul_pd(fscal,dx10);
1023 ty = _mm256_mul_pd(fscal,dy10);
1024 tz = _mm256_mul_pd(fscal,dz10);
1026 /* Update vectorial force */
1027 fix1 = _mm256_add_pd(fix1,tx);
1028 fiy1 = _mm256_add_pd(fiy1,ty);
1029 fiz1 = _mm256_add_pd(fiz1,tz);
1031 fjx0 = _mm256_add_pd(fjx0,tx);
1032 fjy0 = _mm256_add_pd(fjy0,ty);
1033 fjz0 = _mm256_add_pd(fjz0,tz);
1035 /**************************
1036 * CALCULATE INTERACTIONS *
1037 **************************/
1039 r11 = _mm256_mul_pd(rsq11,rinv11);
1040 r11 = _mm256_andnot_pd(dummy_mask,r11);
1042 /* EWALD ELECTROSTATICS */
1044 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1045 ewrt = _mm256_mul_pd(r11,ewtabscale);
1046 ewitab = _mm256_cvttpd_epi32(ewrt);
1047 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1048 ewitab = _mm_slli_epi32(ewitab,2);
1049 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1050 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1051 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1052 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1053 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1054 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1055 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1056 velec = _mm256_mul_pd(qq11,_mm256_sub_pd(rinv11,velec));
1057 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1059 /* Update potential sum for this i atom from the interaction with this j atom. */
1060 velec = _mm256_andnot_pd(dummy_mask,velec);
1061 velecsum = _mm256_add_pd(velecsum,velec);
1065 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1067 /* Calculate temporary vectorial force */
1068 tx = _mm256_mul_pd(fscal,dx11);
1069 ty = _mm256_mul_pd(fscal,dy11);
1070 tz = _mm256_mul_pd(fscal,dz11);
1072 /* Update vectorial force */
1073 fix1 = _mm256_add_pd(fix1,tx);
1074 fiy1 = _mm256_add_pd(fiy1,ty);
1075 fiz1 = _mm256_add_pd(fiz1,tz);
1077 fjx1 = _mm256_add_pd(fjx1,tx);
1078 fjy1 = _mm256_add_pd(fjy1,ty);
1079 fjz1 = _mm256_add_pd(fjz1,tz);
1081 /**************************
1082 * CALCULATE INTERACTIONS *
1083 **************************/
1085 r12 = _mm256_mul_pd(rsq12,rinv12);
1086 r12 = _mm256_andnot_pd(dummy_mask,r12);
1088 /* EWALD ELECTROSTATICS */
1090 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1091 ewrt = _mm256_mul_pd(r12,ewtabscale);
1092 ewitab = _mm256_cvttpd_epi32(ewrt);
1093 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1094 ewitab = _mm_slli_epi32(ewitab,2);
1095 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1096 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1097 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1098 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1099 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1100 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1101 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1102 velec = _mm256_mul_pd(qq12,_mm256_sub_pd(rinv12,velec));
1103 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1105 /* Update potential sum for this i atom from the interaction with this j atom. */
1106 velec = _mm256_andnot_pd(dummy_mask,velec);
1107 velecsum = _mm256_add_pd(velecsum,velec);
1111 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1113 /* Calculate temporary vectorial force */
1114 tx = _mm256_mul_pd(fscal,dx12);
1115 ty = _mm256_mul_pd(fscal,dy12);
1116 tz = _mm256_mul_pd(fscal,dz12);
1118 /* Update vectorial force */
1119 fix1 = _mm256_add_pd(fix1,tx);
1120 fiy1 = _mm256_add_pd(fiy1,ty);
1121 fiz1 = _mm256_add_pd(fiz1,tz);
1123 fjx2 = _mm256_add_pd(fjx2,tx);
1124 fjy2 = _mm256_add_pd(fjy2,ty);
1125 fjz2 = _mm256_add_pd(fjz2,tz);
1127 /**************************
1128 * CALCULATE INTERACTIONS *
1129 **************************/
1131 r20 = _mm256_mul_pd(rsq20,rinv20);
1132 r20 = _mm256_andnot_pd(dummy_mask,r20);
1134 /* EWALD ELECTROSTATICS */
1136 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1137 ewrt = _mm256_mul_pd(r20,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(qq20,_mm256_sub_pd(rinv20,velec));
1149 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1151 /* Update potential sum for this i atom from the interaction with this j atom. */
1152 velec = _mm256_andnot_pd(dummy_mask,velec);
1153 velecsum = _mm256_add_pd(velecsum,velec);
1157 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1159 /* Calculate temporary vectorial force */
1160 tx = _mm256_mul_pd(fscal,dx20);
1161 ty = _mm256_mul_pd(fscal,dy20);
1162 tz = _mm256_mul_pd(fscal,dz20);
1164 /* Update vectorial force */
1165 fix2 = _mm256_add_pd(fix2,tx);
1166 fiy2 = _mm256_add_pd(fiy2,ty);
1167 fiz2 = _mm256_add_pd(fiz2,tz);
1169 fjx0 = _mm256_add_pd(fjx0,tx);
1170 fjy0 = _mm256_add_pd(fjy0,ty);
1171 fjz0 = _mm256_add_pd(fjz0,tz);
1173 /**************************
1174 * CALCULATE INTERACTIONS *
1175 **************************/
1177 r21 = _mm256_mul_pd(rsq21,rinv21);
1178 r21 = _mm256_andnot_pd(dummy_mask,r21);
1180 /* EWALD ELECTROSTATICS */
1182 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1183 ewrt = _mm256_mul_pd(r21,ewtabscale);
1184 ewitab = _mm256_cvttpd_epi32(ewrt);
1185 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1186 ewitab = _mm_slli_epi32(ewitab,2);
1187 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1188 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1189 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1190 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1191 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1192 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1193 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1194 velec = _mm256_mul_pd(qq21,_mm256_sub_pd(rinv21,velec));
1195 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1197 /* Update potential sum for this i atom from the interaction with this j atom. */
1198 velec = _mm256_andnot_pd(dummy_mask,velec);
1199 velecsum = _mm256_add_pd(velecsum,velec);
1203 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1205 /* Calculate temporary vectorial force */
1206 tx = _mm256_mul_pd(fscal,dx21);
1207 ty = _mm256_mul_pd(fscal,dy21);
1208 tz = _mm256_mul_pd(fscal,dz21);
1210 /* Update vectorial force */
1211 fix2 = _mm256_add_pd(fix2,tx);
1212 fiy2 = _mm256_add_pd(fiy2,ty);
1213 fiz2 = _mm256_add_pd(fiz2,tz);
1215 fjx1 = _mm256_add_pd(fjx1,tx);
1216 fjy1 = _mm256_add_pd(fjy1,ty);
1217 fjz1 = _mm256_add_pd(fjz1,tz);
1219 /**************************
1220 * CALCULATE INTERACTIONS *
1221 **************************/
1223 r22 = _mm256_mul_pd(rsq22,rinv22);
1224 r22 = _mm256_andnot_pd(dummy_mask,r22);
1226 /* EWALD ELECTROSTATICS */
1228 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1229 ewrt = _mm256_mul_pd(r22,ewtabscale);
1230 ewitab = _mm256_cvttpd_epi32(ewrt);
1231 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1232 ewitab = _mm_slli_epi32(ewitab,2);
1233 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1234 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1235 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1236 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1237 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1238 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1239 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1240 velec = _mm256_mul_pd(qq22,_mm256_sub_pd(rinv22,velec));
1241 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1243 /* Update potential sum for this i atom from the interaction with this j atom. */
1244 velec = _mm256_andnot_pd(dummy_mask,velec);
1245 velecsum = _mm256_add_pd(velecsum,velec);
1249 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1251 /* Calculate temporary vectorial force */
1252 tx = _mm256_mul_pd(fscal,dx22);
1253 ty = _mm256_mul_pd(fscal,dy22);
1254 tz = _mm256_mul_pd(fscal,dz22);
1256 /* Update vectorial force */
1257 fix2 = _mm256_add_pd(fix2,tx);
1258 fiy2 = _mm256_add_pd(fiy2,ty);
1259 fiz2 = _mm256_add_pd(fiz2,tz);
1261 fjx2 = _mm256_add_pd(fjx2,tx);
1262 fjy2 = _mm256_add_pd(fjy2,ty);
1263 fjz2 = _mm256_add_pd(fjz2,tz);
1265 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1266 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1267 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1268 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1270 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1271 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1273 /* Inner loop uses 409 flops */
1276 /* End of innermost loop */
1278 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1279 f+i_coord_offset,fshift+i_shift_offset);
1282 /* Update potential energies */
1283 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1284 gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1286 /* Increment number of inner iterations */
1287 inneriter += j_index_end - j_index_start;
1289 /* Outer loop uses 20 flops */
1292 /* Increment number of outer iterations */
1295 /* Update outer/inner flops */
1297 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*409);
1300 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_256_double
1301 * Electrostatics interaction: Ewald
1302 * VdW interaction: LJEwald
1303 * Geometry: Water3-Water3
1304 * Calculate force/pot: Force
1307 nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_256_double
1308 (t_nblist * gmx_restrict nlist,
1309 rvec * gmx_restrict xx,
1310 rvec * gmx_restrict ff,
1311 t_forcerec * gmx_restrict fr,
1312 t_mdatoms * gmx_restrict mdatoms,
1313 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1314 t_nrnb * gmx_restrict nrnb)
1316 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1317 * just 0 for non-waters.
1318 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1319 * jnr indices corresponding to data put in the four positions in the SIMD register.
1321 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1322 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1323 int jnrA,jnrB,jnrC,jnrD;
1324 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1325 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1326 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1327 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1328 real rcutoff_scalar;
1329 real *shiftvec,*fshift,*x,*f;
1330 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1331 real scratch[4*DIM];
1332 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1333 real * vdwioffsetptr0;
1334 real * vdwgridioffsetptr0;
1335 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1336 real * vdwioffsetptr1;
1337 real * vdwgridioffsetptr1;
1338 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1339 real * vdwioffsetptr2;
1340 real * vdwgridioffsetptr2;
1341 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1342 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1343 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1344 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1345 __m256d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1346 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1347 __m256d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1348 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1349 __m256d dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1350 __m256d dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1351 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1352 __m256d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1353 __m256d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1354 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1355 __m256d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1356 __m256d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1357 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
1360 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1363 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
1364 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
1375 __m256d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1376 __m256d one_half = _mm256_set1_pd(0.5);
1377 __m256d minus_one = _mm256_set1_pd(-1.0);
1379 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1380 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1382 __m256d dummy_mask,cutoff_mask;
1383 __m128 tmpmask0,tmpmask1;
1384 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1385 __m256d one = _mm256_set1_pd(1.0);
1386 __m256d two = _mm256_set1_pd(2.0);
1392 jindex = nlist->jindex;
1394 shiftidx = nlist->shift;
1396 shiftvec = fr->shift_vec[0];
1397 fshift = fr->fshift[0];
1398 facel = _mm256_set1_pd(fr->epsfac);
1399 charge = mdatoms->chargeA;
1400 nvdwtype = fr->ntype;
1401 vdwparam = fr->nbfp;
1402 vdwtype = mdatoms->typeA;
1403 vdwgridparam = fr->ljpme_c6grid;
1404 sh_lj_ewald = _mm256_set1_pd(fr->ic->sh_lj_ewald);
1405 ewclj = _mm256_set1_pd(fr->ewaldcoeff_lj);
1406 ewclj2 = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
1408 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
1409 beta = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
1410 beta2 = _mm256_mul_pd(beta,beta);
1411 beta3 = _mm256_mul_pd(beta,beta2);
1413 ewtab = fr->ic->tabq_coul_F;
1414 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
1415 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1417 /* Setup water-specific parameters */
1418 inr = nlist->iinr[0];
1419 iq0 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1420 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1421 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1422 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
1423 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
1425 jq0 = _mm256_set1_pd(charge[inr+0]);
1426 jq1 = _mm256_set1_pd(charge[inr+1]);
1427 jq2 = _mm256_set1_pd(charge[inr+2]);
1428 vdwjidx0A = 2*vdwtype[inr+0];
1429 qq00 = _mm256_mul_pd(iq0,jq0);
1430 c6_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1431 c12_00 = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1432 c6grid_00 = _mm256_set1_pd(vdwgridioffsetptr0[vdwjidx0A]);
1433 qq01 = _mm256_mul_pd(iq0,jq1);
1434 qq02 = _mm256_mul_pd(iq0,jq2);
1435 qq10 = _mm256_mul_pd(iq1,jq0);
1436 qq11 = _mm256_mul_pd(iq1,jq1);
1437 qq12 = _mm256_mul_pd(iq1,jq2);
1438 qq20 = _mm256_mul_pd(iq2,jq0);
1439 qq21 = _mm256_mul_pd(iq2,jq1);
1440 qq22 = _mm256_mul_pd(iq2,jq2);
1442 /* Avoid stupid compiler warnings */
1443 jnrA = jnrB = jnrC = jnrD = 0;
1444 j_coord_offsetA = 0;
1445 j_coord_offsetB = 0;
1446 j_coord_offsetC = 0;
1447 j_coord_offsetD = 0;
1452 for(iidx=0;iidx<4*DIM;iidx++)
1454 scratch[iidx] = 0.0;
1457 /* Start outer loop over neighborlists */
1458 for(iidx=0; iidx<nri; iidx++)
1460 /* Load shift vector for this list */
1461 i_shift_offset = DIM*shiftidx[iidx];
1463 /* Load limits for loop over neighbors */
1464 j_index_start = jindex[iidx];
1465 j_index_end = jindex[iidx+1];
1467 /* Get outer coordinate index */
1469 i_coord_offset = DIM*inr;
1471 /* Load i particle coords and add shift vector */
1472 gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1473 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1475 fix0 = _mm256_setzero_pd();
1476 fiy0 = _mm256_setzero_pd();
1477 fiz0 = _mm256_setzero_pd();
1478 fix1 = _mm256_setzero_pd();
1479 fiy1 = _mm256_setzero_pd();
1480 fiz1 = _mm256_setzero_pd();
1481 fix2 = _mm256_setzero_pd();
1482 fiy2 = _mm256_setzero_pd();
1483 fiz2 = _mm256_setzero_pd();
1485 /* Start inner kernel loop */
1486 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1489 /* Get j neighbor index, and coordinate index */
1491 jnrB = jjnr[jidx+1];
1492 jnrC = jjnr[jidx+2];
1493 jnrD = jjnr[jidx+3];
1494 j_coord_offsetA = DIM*jnrA;
1495 j_coord_offsetB = DIM*jnrB;
1496 j_coord_offsetC = DIM*jnrC;
1497 j_coord_offsetD = DIM*jnrD;
1499 /* load j atom coordinates */
1500 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1501 x+j_coord_offsetC,x+j_coord_offsetD,
1502 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1504 /* Calculate displacement vector */
1505 dx00 = _mm256_sub_pd(ix0,jx0);
1506 dy00 = _mm256_sub_pd(iy0,jy0);
1507 dz00 = _mm256_sub_pd(iz0,jz0);
1508 dx01 = _mm256_sub_pd(ix0,jx1);
1509 dy01 = _mm256_sub_pd(iy0,jy1);
1510 dz01 = _mm256_sub_pd(iz0,jz1);
1511 dx02 = _mm256_sub_pd(ix0,jx2);
1512 dy02 = _mm256_sub_pd(iy0,jy2);
1513 dz02 = _mm256_sub_pd(iz0,jz2);
1514 dx10 = _mm256_sub_pd(ix1,jx0);
1515 dy10 = _mm256_sub_pd(iy1,jy0);
1516 dz10 = _mm256_sub_pd(iz1,jz0);
1517 dx11 = _mm256_sub_pd(ix1,jx1);
1518 dy11 = _mm256_sub_pd(iy1,jy1);
1519 dz11 = _mm256_sub_pd(iz1,jz1);
1520 dx12 = _mm256_sub_pd(ix1,jx2);
1521 dy12 = _mm256_sub_pd(iy1,jy2);
1522 dz12 = _mm256_sub_pd(iz1,jz2);
1523 dx20 = _mm256_sub_pd(ix2,jx0);
1524 dy20 = _mm256_sub_pd(iy2,jy0);
1525 dz20 = _mm256_sub_pd(iz2,jz0);
1526 dx21 = _mm256_sub_pd(ix2,jx1);
1527 dy21 = _mm256_sub_pd(iy2,jy1);
1528 dz21 = _mm256_sub_pd(iz2,jz1);
1529 dx22 = _mm256_sub_pd(ix2,jx2);
1530 dy22 = _mm256_sub_pd(iy2,jy2);
1531 dz22 = _mm256_sub_pd(iz2,jz2);
1533 /* Calculate squared distance and things based on it */
1534 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1535 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1536 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1537 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1538 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1539 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1540 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1541 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1542 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1544 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1545 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
1546 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
1547 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1548 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
1549 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
1550 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1551 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
1552 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
1554 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1555 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
1556 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
1557 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1558 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
1559 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
1560 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1561 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
1562 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
1564 fjx0 = _mm256_setzero_pd();
1565 fjy0 = _mm256_setzero_pd();
1566 fjz0 = _mm256_setzero_pd();
1567 fjx1 = _mm256_setzero_pd();
1568 fjy1 = _mm256_setzero_pd();
1569 fjz1 = _mm256_setzero_pd();
1570 fjx2 = _mm256_setzero_pd();
1571 fjy2 = _mm256_setzero_pd();
1572 fjz2 = _mm256_setzero_pd();
1574 /**************************
1575 * CALCULATE INTERACTIONS *
1576 **************************/
1578 r00 = _mm256_mul_pd(rsq00,rinv00);
1580 /* EWALD ELECTROSTATICS */
1582 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1583 ewrt = _mm256_mul_pd(r00,ewtabscale);
1584 ewitab = _mm256_cvttpd_epi32(ewrt);
1585 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1586 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1587 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1589 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1590 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1592 /* Analytical LJ-PME */
1593 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1594 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
1595 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
1596 exponent = gmx_simd_exp_d(ewcljrsq);
1597 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1598 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
1599 /* f6A = 6 * C6grid * (1 - poly) */
1600 f6A = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
1601 /* f6B = C6grid * exponent * beta^6 */
1602 f6B = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
1603 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1604 fvdw = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1606 fscal = _mm256_add_pd(felec,fvdw);
1608 /* Calculate temporary vectorial force */
1609 tx = _mm256_mul_pd(fscal,dx00);
1610 ty = _mm256_mul_pd(fscal,dy00);
1611 tz = _mm256_mul_pd(fscal,dz00);
1613 /* Update vectorial force */
1614 fix0 = _mm256_add_pd(fix0,tx);
1615 fiy0 = _mm256_add_pd(fiy0,ty);
1616 fiz0 = _mm256_add_pd(fiz0,tz);
1618 fjx0 = _mm256_add_pd(fjx0,tx);
1619 fjy0 = _mm256_add_pd(fjy0,ty);
1620 fjz0 = _mm256_add_pd(fjz0,tz);
1622 /**************************
1623 * CALCULATE INTERACTIONS *
1624 **************************/
1626 r01 = _mm256_mul_pd(rsq01,rinv01);
1628 /* EWALD ELECTROSTATICS */
1630 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1631 ewrt = _mm256_mul_pd(r01,ewtabscale);
1632 ewitab = _mm256_cvttpd_epi32(ewrt);
1633 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1634 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1635 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1637 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1638 felec = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
1642 /* Calculate temporary vectorial force */
1643 tx = _mm256_mul_pd(fscal,dx01);
1644 ty = _mm256_mul_pd(fscal,dy01);
1645 tz = _mm256_mul_pd(fscal,dz01);
1647 /* Update vectorial force */
1648 fix0 = _mm256_add_pd(fix0,tx);
1649 fiy0 = _mm256_add_pd(fiy0,ty);
1650 fiz0 = _mm256_add_pd(fiz0,tz);
1652 fjx1 = _mm256_add_pd(fjx1,tx);
1653 fjy1 = _mm256_add_pd(fjy1,ty);
1654 fjz1 = _mm256_add_pd(fjz1,tz);
1656 /**************************
1657 * CALCULATE INTERACTIONS *
1658 **************************/
1660 r02 = _mm256_mul_pd(rsq02,rinv02);
1662 /* EWALD ELECTROSTATICS */
1664 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1665 ewrt = _mm256_mul_pd(r02,ewtabscale);
1666 ewitab = _mm256_cvttpd_epi32(ewrt);
1667 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1668 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1669 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1671 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1672 felec = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
1676 /* Calculate temporary vectorial force */
1677 tx = _mm256_mul_pd(fscal,dx02);
1678 ty = _mm256_mul_pd(fscal,dy02);
1679 tz = _mm256_mul_pd(fscal,dz02);
1681 /* Update vectorial force */
1682 fix0 = _mm256_add_pd(fix0,tx);
1683 fiy0 = _mm256_add_pd(fiy0,ty);
1684 fiz0 = _mm256_add_pd(fiz0,tz);
1686 fjx2 = _mm256_add_pd(fjx2,tx);
1687 fjy2 = _mm256_add_pd(fjy2,ty);
1688 fjz2 = _mm256_add_pd(fjz2,tz);
1690 /**************************
1691 * CALCULATE INTERACTIONS *
1692 **************************/
1694 r10 = _mm256_mul_pd(rsq10,rinv10);
1696 /* EWALD ELECTROSTATICS */
1698 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1699 ewrt = _mm256_mul_pd(r10,ewtabscale);
1700 ewitab = _mm256_cvttpd_epi32(ewrt);
1701 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1702 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1703 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1705 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1706 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1710 /* Calculate temporary vectorial force */
1711 tx = _mm256_mul_pd(fscal,dx10);
1712 ty = _mm256_mul_pd(fscal,dy10);
1713 tz = _mm256_mul_pd(fscal,dz10);
1715 /* Update vectorial force */
1716 fix1 = _mm256_add_pd(fix1,tx);
1717 fiy1 = _mm256_add_pd(fiy1,ty);
1718 fiz1 = _mm256_add_pd(fiz1,tz);
1720 fjx0 = _mm256_add_pd(fjx0,tx);
1721 fjy0 = _mm256_add_pd(fjy0,ty);
1722 fjz0 = _mm256_add_pd(fjz0,tz);
1724 /**************************
1725 * CALCULATE INTERACTIONS *
1726 **************************/
1728 r11 = _mm256_mul_pd(rsq11,rinv11);
1730 /* EWALD ELECTROSTATICS */
1732 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1733 ewrt = _mm256_mul_pd(r11,ewtabscale);
1734 ewitab = _mm256_cvttpd_epi32(ewrt);
1735 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1736 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1737 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1739 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1740 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
1744 /* Calculate temporary vectorial force */
1745 tx = _mm256_mul_pd(fscal,dx11);
1746 ty = _mm256_mul_pd(fscal,dy11);
1747 tz = _mm256_mul_pd(fscal,dz11);
1749 /* Update vectorial force */
1750 fix1 = _mm256_add_pd(fix1,tx);
1751 fiy1 = _mm256_add_pd(fiy1,ty);
1752 fiz1 = _mm256_add_pd(fiz1,tz);
1754 fjx1 = _mm256_add_pd(fjx1,tx);
1755 fjy1 = _mm256_add_pd(fjy1,ty);
1756 fjz1 = _mm256_add_pd(fjz1,tz);
1758 /**************************
1759 * CALCULATE INTERACTIONS *
1760 **************************/
1762 r12 = _mm256_mul_pd(rsq12,rinv12);
1764 /* EWALD ELECTROSTATICS */
1766 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1767 ewrt = _mm256_mul_pd(r12,ewtabscale);
1768 ewitab = _mm256_cvttpd_epi32(ewrt);
1769 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1770 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1771 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1773 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1774 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
1778 /* Calculate temporary vectorial force */
1779 tx = _mm256_mul_pd(fscal,dx12);
1780 ty = _mm256_mul_pd(fscal,dy12);
1781 tz = _mm256_mul_pd(fscal,dz12);
1783 /* Update vectorial force */
1784 fix1 = _mm256_add_pd(fix1,tx);
1785 fiy1 = _mm256_add_pd(fiy1,ty);
1786 fiz1 = _mm256_add_pd(fiz1,tz);
1788 fjx2 = _mm256_add_pd(fjx2,tx);
1789 fjy2 = _mm256_add_pd(fjy2,ty);
1790 fjz2 = _mm256_add_pd(fjz2,tz);
1792 /**************************
1793 * CALCULATE INTERACTIONS *
1794 **************************/
1796 r20 = _mm256_mul_pd(rsq20,rinv20);
1798 /* EWALD ELECTROSTATICS */
1800 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1801 ewrt = _mm256_mul_pd(r20,ewtabscale);
1802 ewitab = _mm256_cvttpd_epi32(ewrt);
1803 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1804 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1805 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1807 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1808 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1812 /* Calculate temporary vectorial force */
1813 tx = _mm256_mul_pd(fscal,dx20);
1814 ty = _mm256_mul_pd(fscal,dy20);
1815 tz = _mm256_mul_pd(fscal,dz20);
1817 /* Update vectorial force */
1818 fix2 = _mm256_add_pd(fix2,tx);
1819 fiy2 = _mm256_add_pd(fiy2,ty);
1820 fiz2 = _mm256_add_pd(fiz2,tz);
1822 fjx0 = _mm256_add_pd(fjx0,tx);
1823 fjy0 = _mm256_add_pd(fjy0,ty);
1824 fjz0 = _mm256_add_pd(fjz0,tz);
1826 /**************************
1827 * CALCULATE INTERACTIONS *
1828 **************************/
1830 r21 = _mm256_mul_pd(rsq21,rinv21);
1832 /* EWALD ELECTROSTATICS */
1834 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1835 ewrt = _mm256_mul_pd(r21,ewtabscale);
1836 ewitab = _mm256_cvttpd_epi32(ewrt);
1837 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1838 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1839 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1841 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1842 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
1846 /* Calculate temporary vectorial force */
1847 tx = _mm256_mul_pd(fscal,dx21);
1848 ty = _mm256_mul_pd(fscal,dy21);
1849 tz = _mm256_mul_pd(fscal,dz21);
1851 /* Update vectorial force */
1852 fix2 = _mm256_add_pd(fix2,tx);
1853 fiy2 = _mm256_add_pd(fiy2,ty);
1854 fiz2 = _mm256_add_pd(fiz2,tz);
1856 fjx1 = _mm256_add_pd(fjx1,tx);
1857 fjy1 = _mm256_add_pd(fjy1,ty);
1858 fjz1 = _mm256_add_pd(fjz1,tz);
1860 /**************************
1861 * CALCULATE INTERACTIONS *
1862 **************************/
1864 r22 = _mm256_mul_pd(rsq22,rinv22);
1866 /* EWALD ELECTROSTATICS */
1868 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1869 ewrt = _mm256_mul_pd(r22,ewtabscale);
1870 ewitab = _mm256_cvttpd_epi32(ewrt);
1871 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1872 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
1873 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
1875 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
1876 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
1880 /* Calculate temporary vectorial force */
1881 tx = _mm256_mul_pd(fscal,dx22);
1882 ty = _mm256_mul_pd(fscal,dy22);
1883 tz = _mm256_mul_pd(fscal,dz22);
1885 /* Update vectorial force */
1886 fix2 = _mm256_add_pd(fix2,tx);
1887 fiy2 = _mm256_add_pd(fiy2,ty);
1888 fiz2 = _mm256_add_pd(fiz2,tz);
1890 fjx2 = _mm256_add_pd(fjx2,tx);
1891 fjy2 = _mm256_add_pd(fjy2,ty);
1892 fjz2 = _mm256_add_pd(fjz2,tz);
1894 fjptrA = f+j_coord_offsetA;
1895 fjptrB = f+j_coord_offsetB;
1896 fjptrC = f+j_coord_offsetC;
1897 fjptrD = f+j_coord_offsetD;
1899 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1900 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1902 /* Inner loop uses 347 flops */
1905 if(jidx<j_index_end)
1908 /* Get j neighbor index, and coordinate index */
1909 jnrlistA = jjnr[jidx];
1910 jnrlistB = jjnr[jidx+1];
1911 jnrlistC = jjnr[jidx+2];
1912 jnrlistD = jjnr[jidx+3];
1913 /* Sign of each element will be negative for non-real atoms.
1914 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1915 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1917 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1919 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1920 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1921 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1923 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1924 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1925 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1926 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1927 j_coord_offsetA = DIM*jnrA;
1928 j_coord_offsetB = DIM*jnrB;
1929 j_coord_offsetC = DIM*jnrC;
1930 j_coord_offsetD = DIM*jnrD;
1932 /* load j atom coordinates */
1933 gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1934 x+j_coord_offsetC,x+j_coord_offsetD,
1935 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1937 /* Calculate displacement vector */
1938 dx00 = _mm256_sub_pd(ix0,jx0);
1939 dy00 = _mm256_sub_pd(iy0,jy0);
1940 dz00 = _mm256_sub_pd(iz0,jz0);
1941 dx01 = _mm256_sub_pd(ix0,jx1);
1942 dy01 = _mm256_sub_pd(iy0,jy1);
1943 dz01 = _mm256_sub_pd(iz0,jz1);
1944 dx02 = _mm256_sub_pd(ix0,jx2);
1945 dy02 = _mm256_sub_pd(iy0,jy2);
1946 dz02 = _mm256_sub_pd(iz0,jz2);
1947 dx10 = _mm256_sub_pd(ix1,jx0);
1948 dy10 = _mm256_sub_pd(iy1,jy0);
1949 dz10 = _mm256_sub_pd(iz1,jz0);
1950 dx11 = _mm256_sub_pd(ix1,jx1);
1951 dy11 = _mm256_sub_pd(iy1,jy1);
1952 dz11 = _mm256_sub_pd(iz1,jz1);
1953 dx12 = _mm256_sub_pd(ix1,jx2);
1954 dy12 = _mm256_sub_pd(iy1,jy2);
1955 dz12 = _mm256_sub_pd(iz1,jz2);
1956 dx20 = _mm256_sub_pd(ix2,jx0);
1957 dy20 = _mm256_sub_pd(iy2,jy0);
1958 dz20 = _mm256_sub_pd(iz2,jz0);
1959 dx21 = _mm256_sub_pd(ix2,jx1);
1960 dy21 = _mm256_sub_pd(iy2,jy1);
1961 dz21 = _mm256_sub_pd(iz2,jz1);
1962 dx22 = _mm256_sub_pd(ix2,jx2);
1963 dy22 = _mm256_sub_pd(iy2,jy2);
1964 dz22 = _mm256_sub_pd(iz2,jz2);
1966 /* Calculate squared distance and things based on it */
1967 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1968 rsq01 = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1969 rsq02 = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1970 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1971 rsq11 = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1972 rsq12 = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1973 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1974 rsq21 = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1975 rsq22 = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1977 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1978 rinv01 = gmx_mm256_invsqrt_pd(rsq01);
1979 rinv02 = gmx_mm256_invsqrt_pd(rsq02);
1980 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1981 rinv11 = gmx_mm256_invsqrt_pd(rsq11);
1982 rinv12 = gmx_mm256_invsqrt_pd(rsq12);
1983 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1984 rinv21 = gmx_mm256_invsqrt_pd(rsq21);
1985 rinv22 = gmx_mm256_invsqrt_pd(rsq22);
1987 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1988 rinvsq01 = _mm256_mul_pd(rinv01,rinv01);
1989 rinvsq02 = _mm256_mul_pd(rinv02,rinv02);
1990 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1991 rinvsq11 = _mm256_mul_pd(rinv11,rinv11);
1992 rinvsq12 = _mm256_mul_pd(rinv12,rinv12);
1993 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1994 rinvsq21 = _mm256_mul_pd(rinv21,rinv21);
1995 rinvsq22 = _mm256_mul_pd(rinv22,rinv22);
1997 fjx0 = _mm256_setzero_pd();
1998 fjy0 = _mm256_setzero_pd();
1999 fjz0 = _mm256_setzero_pd();
2000 fjx1 = _mm256_setzero_pd();
2001 fjy1 = _mm256_setzero_pd();
2002 fjz1 = _mm256_setzero_pd();
2003 fjx2 = _mm256_setzero_pd();
2004 fjy2 = _mm256_setzero_pd();
2005 fjz2 = _mm256_setzero_pd();
2007 /**************************
2008 * CALCULATE INTERACTIONS *
2009 **************************/
2011 r00 = _mm256_mul_pd(rsq00,rinv00);
2012 r00 = _mm256_andnot_pd(dummy_mask,r00);
2014 /* EWALD ELECTROSTATICS */
2016 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2017 ewrt = _mm256_mul_pd(r00,ewtabscale);
2018 ewitab = _mm256_cvttpd_epi32(ewrt);
2019 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2020 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2021 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2023 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2024 felec = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
2026 /* Analytical LJ-PME */
2027 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2028 ewcljrsq = _mm256_mul_pd(ewclj2,rsq00);
2029 ewclj6 = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
2030 exponent = gmx_simd_exp_d(ewcljrsq);
2031 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
2032 poly = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
2033 /* f6A = 6 * C6grid * (1 - poly) */
2034 f6A = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
2035 /* f6B = C6grid * exponent * beta^6 */
2036 f6B = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
2037 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2038 fvdw = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
2040 fscal = _mm256_add_pd(felec,fvdw);
2042 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2044 /* Calculate temporary vectorial force */
2045 tx = _mm256_mul_pd(fscal,dx00);
2046 ty = _mm256_mul_pd(fscal,dy00);
2047 tz = _mm256_mul_pd(fscal,dz00);
2049 /* Update vectorial force */
2050 fix0 = _mm256_add_pd(fix0,tx);
2051 fiy0 = _mm256_add_pd(fiy0,ty);
2052 fiz0 = _mm256_add_pd(fiz0,tz);
2054 fjx0 = _mm256_add_pd(fjx0,tx);
2055 fjy0 = _mm256_add_pd(fjy0,ty);
2056 fjz0 = _mm256_add_pd(fjz0,tz);
2058 /**************************
2059 * CALCULATE INTERACTIONS *
2060 **************************/
2062 r01 = _mm256_mul_pd(rsq01,rinv01);
2063 r01 = _mm256_andnot_pd(dummy_mask,r01);
2065 /* EWALD ELECTROSTATICS */
2067 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2068 ewrt = _mm256_mul_pd(r01,ewtabscale);
2069 ewitab = _mm256_cvttpd_epi32(ewrt);
2070 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2071 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2072 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2074 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2075 felec = _mm256_mul_pd(_mm256_mul_pd(qq01,rinv01),_mm256_sub_pd(rinvsq01,felec));
2079 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2081 /* Calculate temporary vectorial force */
2082 tx = _mm256_mul_pd(fscal,dx01);
2083 ty = _mm256_mul_pd(fscal,dy01);
2084 tz = _mm256_mul_pd(fscal,dz01);
2086 /* Update vectorial force */
2087 fix0 = _mm256_add_pd(fix0,tx);
2088 fiy0 = _mm256_add_pd(fiy0,ty);
2089 fiz0 = _mm256_add_pd(fiz0,tz);
2091 fjx1 = _mm256_add_pd(fjx1,tx);
2092 fjy1 = _mm256_add_pd(fjy1,ty);
2093 fjz1 = _mm256_add_pd(fjz1,tz);
2095 /**************************
2096 * CALCULATE INTERACTIONS *
2097 **************************/
2099 r02 = _mm256_mul_pd(rsq02,rinv02);
2100 r02 = _mm256_andnot_pd(dummy_mask,r02);
2102 /* EWALD ELECTROSTATICS */
2104 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2105 ewrt = _mm256_mul_pd(r02,ewtabscale);
2106 ewitab = _mm256_cvttpd_epi32(ewrt);
2107 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2108 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2109 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2111 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2112 felec = _mm256_mul_pd(_mm256_mul_pd(qq02,rinv02),_mm256_sub_pd(rinvsq02,felec));
2116 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2118 /* Calculate temporary vectorial force */
2119 tx = _mm256_mul_pd(fscal,dx02);
2120 ty = _mm256_mul_pd(fscal,dy02);
2121 tz = _mm256_mul_pd(fscal,dz02);
2123 /* Update vectorial force */
2124 fix0 = _mm256_add_pd(fix0,tx);
2125 fiy0 = _mm256_add_pd(fiy0,ty);
2126 fiz0 = _mm256_add_pd(fiz0,tz);
2128 fjx2 = _mm256_add_pd(fjx2,tx);
2129 fjy2 = _mm256_add_pd(fjy2,ty);
2130 fjz2 = _mm256_add_pd(fjz2,tz);
2132 /**************************
2133 * CALCULATE INTERACTIONS *
2134 **************************/
2136 r10 = _mm256_mul_pd(rsq10,rinv10);
2137 r10 = _mm256_andnot_pd(dummy_mask,r10);
2139 /* EWALD ELECTROSTATICS */
2141 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2142 ewrt = _mm256_mul_pd(r10,ewtabscale);
2143 ewitab = _mm256_cvttpd_epi32(ewrt);
2144 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2145 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2146 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2148 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2149 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
2153 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2155 /* Calculate temporary vectorial force */
2156 tx = _mm256_mul_pd(fscal,dx10);
2157 ty = _mm256_mul_pd(fscal,dy10);
2158 tz = _mm256_mul_pd(fscal,dz10);
2160 /* Update vectorial force */
2161 fix1 = _mm256_add_pd(fix1,tx);
2162 fiy1 = _mm256_add_pd(fiy1,ty);
2163 fiz1 = _mm256_add_pd(fiz1,tz);
2165 fjx0 = _mm256_add_pd(fjx0,tx);
2166 fjy0 = _mm256_add_pd(fjy0,ty);
2167 fjz0 = _mm256_add_pd(fjz0,tz);
2169 /**************************
2170 * CALCULATE INTERACTIONS *
2171 **************************/
2173 r11 = _mm256_mul_pd(rsq11,rinv11);
2174 r11 = _mm256_andnot_pd(dummy_mask,r11);
2176 /* EWALD ELECTROSTATICS */
2178 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2179 ewrt = _mm256_mul_pd(r11,ewtabscale);
2180 ewitab = _mm256_cvttpd_epi32(ewrt);
2181 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2182 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2183 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2185 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2186 felec = _mm256_mul_pd(_mm256_mul_pd(qq11,rinv11),_mm256_sub_pd(rinvsq11,felec));
2190 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2192 /* Calculate temporary vectorial force */
2193 tx = _mm256_mul_pd(fscal,dx11);
2194 ty = _mm256_mul_pd(fscal,dy11);
2195 tz = _mm256_mul_pd(fscal,dz11);
2197 /* Update vectorial force */
2198 fix1 = _mm256_add_pd(fix1,tx);
2199 fiy1 = _mm256_add_pd(fiy1,ty);
2200 fiz1 = _mm256_add_pd(fiz1,tz);
2202 fjx1 = _mm256_add_pd(fjx1,tx);
2203 fjy1 = _mm256_add_pd(fjy1,ty);
2204 fjz1 = _mm256_add_pd(fjz1,tz);
2206 /**************************
2207 * CALCULATE INTERACTIONS *
2208 **************************/
2210 r12 = _mm256_mul_pd(rsq12,rinv12);
2211 r12 = _mm256_andnot_pd(dummy_mask,r12);
2213 /* EWALD ELECTROSTATICS */
2215 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2216 ewrt = _mm256_mul_pd(r12,ewtabscale);
2217 ewitab = _mm256_cvttpd_epi32(ewrt);
2218 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2219 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2220 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2222 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2223 felec = _mm256_mul_pd(_mm256_mul_pd(qq12,rinv12),_mm256_sub_pd(rinvsq12,felec));
2227 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2229 /* Calculate temporary vectorial force */
2230 tx = _mm256_mul_pd(fscal,dx12);
2231 ty = _mm256_mul_pd(fscal,dy12);
2232 tz = _mm256_mul_pd(fscal,dz12);
2234 /* Update vectorial force */
2235 fix1 = _mm256_add_pd(fix1,tx);
2236 fiy1 = _mm256_add_pd(fiy1,ty);
2237 fiz1 = _mm256_add_pd(fiz1,tz);
2239 fjx2 = _mm256_add_pd(fjx2,tx);
2240 fjy2 = _mm256_add_pd(fjy2,ty);
2241 fjz2 = _mm256_add_pd(fjz2,tz);
2243 /**************************
2244 * CALCULATE INTERACTIONS *
2245 **************************/
2247 r20 = _mm256_mul_pd(rsq20,rinv20);
2248 r20 = _mm256_andnot_pd(dummy_mask,r20);
2250 /* EWALD ELECTROSTATICS */
2252 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2253 ewrt = _mm256_mul_pd(r20,ewtabscale);
2254 ewitab = _mm256_cvttpd_epi32(ewrt);
2255 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2256 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2257 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2259 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2260 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
2264 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2266 /* Calculate temporary vectorial force */
2267 tx = _mm256_mul_pd(fscal,dx20);
2268 ty = _mm256_mul_pd(fscal,dy20);
2269 tz = _mm256_mul_pd(fscal,dz20);
2271 /* Update vectorial force */
2272 fix2 = _mm256_add_pd(fix2,tx);
2273 fiy2 = _mm256_add_pd(fiy2,ty);
2274 fiz2 = _mm256_add_pd(fiz2,tz);
2276 fjx0 = _mm256_add_pd(fjx0,tx);
2277 fjy0 = _mm256_add_pd(fjy0,ty);
2278 fjz0 = _mm256_add_pd(fjz0,tz);
2280 /**************************
2281 * CALCULATE INTERACTIONS *
2282 **************************/
2284 r21 = _mm256_mul_pd(rsq21,rinv21);
2285 r21 = _mm256_andnot_pd(dummy_mask,r21);
2287 /* EWALD ELECTROSTATICS */
2289 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2290 ewrt = _mm256_mul_pd(r21,ewtabscale);
2291 ewitab = _mm256_cvttpd_epi32(ewrt);
2292 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2293 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2294 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2296 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2297 felec = _mm256_mul_pd(_mm256_mul_pd(qq21,rinv21),_mm256_sub_pd(rinvsq21,felec));
2301 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2303 /* Calculate temporary vectorial force */
2304 tx = _mm256_mul_pd(fscal,dx21);
2305 ty = _mm256_mul_pd(fscal,dy21);
2306 tz = _mm256_mul_pd(fscal,dz21);
2308 /* Update vectorial force */
2309 fix2 = _mm256_add_pd(fix2,tx);
2310 fiy2 = _mm256_add_pd(fiy2,ty);
2311 fiz2 = _mm256_add_pd(fiz2,tz);
2313 fjx1 = _mm256_add_pd(fjx1,tx);
2314 fjy1 = _mm256_add_pd(fjy1,ty);
2315 fjz1 = _mm256_add_pd(fjz1,tz);
2317 /**************************
2318 * CALCULATE INTERACTIONS *
2319 **************************/
2321 r22 = _mm256_mul_pd(rsq22,rinv22);
2322 r22 = _mm256_andnot_pd(dummy_mask,r22);
2324 /* EWALD ELECTROSTATICS */
2326 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2327 ewrt = _mm256_mul_pd(r22,ewtabscale);
2328 ewitab = _mm256_cvttpd_epi32(ewrt);
2329 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
2330 gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
2331 ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
2333 felec = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
2334 felec = _mm256_mul_pd(_mm256_mul_pd(qq22,rinv22),_mm256_sub_pd(rinvsq22,felec));
2338 fscal = _mm256_andnot_pd(dummy_mask,fscal);
2340 /* Calculate temporary vectorial force */
2341 tx = _mm256_mul_pd(fscal,dx22);
2342 ty = _mm256_mul_pd(fscal,dy22);
2343 tz = _mm256_mul_pd(fscal,dz22);
2345 /* Update vectorial force */
2346 fix2 = _mm256_add_pd(fix2,tx);
2347 fiy2 = _mm256_add_pd(fiy2,ty);
2348 fiz2 = _mm256_add_pd(fiz2,tz);
2350 fjx2 = _mm256_add_pd(fjx2,tx);
2351 fjy2 = _mm256_add_pd(fjy2,ty);
2352 fjz2 = _mm256_add_pd(fjz2,tz);
2354 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2355 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2356 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2357 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2359 gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2360 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2362 /* Inner loop uses 356 flops */
2365 /* End of innermost loop */
2367 gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2368 f+i_coord_offset,fshift+i_shift_offset);
2370 /* Increment number of inner iterations */
2371 inneriter += j_index_end - j_index_start;
2373 /* Outer loop uses 18 flops */
2376 /* Increment number of outer iterations */
2379 /* Update outer/inner flops */
2381 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*356);