2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS avx_256_single kernel generator.
42 #include "../nb_kernel.h"
43 #include "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_single.h"
48 #include "kernelutil_x86_avx_256_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3W3_VF_avx_256_single
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_single
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrE,jnrF,jnrG,jnrH;
76 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
77 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
78 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
79 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
80 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
82 real *shiftvec,*fshift,*x,*f;
83 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
85 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 real * vdwioffsetptr0;
87 real * vdwgridioffsetptr0;
88 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
89 real * vdwioffsetptr1;
90 real * vdwgridioffsetptr1;
91 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
92 real * vdwioffsetptr2;
93 real * vdwgridioffsetptr2;
94 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
95 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
96 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
97 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
98 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
99 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
100 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
101 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
102 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
103 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
104 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
105 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
106 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
107 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
108 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
109 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
110 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
113 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
116 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
117 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
128 __m256 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
129 __m256 one_half = _mm256_set1_ps(0.5);
130 __m256 minus_one = _mm256_set1_ps(-1.0);
132 __m128i ewitab_lo,ewitab_hi;
133 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
134 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
136 __m256 dummy_mask,cutoff_mask;
137 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
138 __m256 one = _mm256_set1_ps(1.0);
139 __m256 two = _mm256_set1_ps(2.0);
145 jindex = nlist->jindex;
147 shiftidx = nlist->shift;
149 shiftvec = fr->shift_vec[0];
150 fshift = fr->fshift[0];
151 facel = _mm256_set1_ps(fr->epsfac);
152 charge = mdatoms->chargeA;
153 nvdwtype = fr->ntype;
155 vdwtype = mdatoms->typeA;
156 vdwgridparam = fr->ljpme_c6grid;
157 sh_lj_ewald = _mm256_set1_ps(fr->ic->sh_lj_ewald);
158 ewclj = _mm256_set1_ps(fr->ewaldcoeff_lj);
159 ewclj2 = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
161 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
162 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
163 beta2 = _mm256_mul_ps(beta,beta);
164 beta3 = _mm256_mul_ps(beta,beta2);
166 ewtab = fr->ic->tabq_coul_FDV0;
167 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
168 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
170 /* Setup water-specific parameters */
171 inr = nlist->iinr[0];
172 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
173 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
174 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
175 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
176 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
178 jq0 = _mm256_set1_ps(charge[inr+0]);
179 jq1 = _mm256_set1_ps(charge[inr+1]);
180 jq2 = _mm256_set1_ps(charge[inr+2]);
181 vdwjidx0A = 2*vdwtype[inr+0];
182 qq00 = _mm256_mul_ps(iq0,jq0);
183 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
184 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
185 c6grid_00 = _mm256_set1_ps(vdwgridioffsetptr0[vdwjidx0A]);
186 qq01 = _mm256_mul_ps(iq0,jq1);
187 qq02 = _mm256_mul_ps(iq0,jq2);
188 qq10 = _mm256_mul_ps(iq1,jq0);
189 qq11 = _mm256_mul_ps(iq1,jq1);
190 qq12 = _mm256_mul_ps(iq1,jq2);
191 qq20 = _mm256_mul_ps(iq2,jq0);
192 qq21 = _mm256_mul_ps(iq2,jq1);
193 qq22 = _mm256_mul_ps(iq2,jq2);
195 /* Avoid stupid compiler warnings */
196 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
209 for(iidx=0;iidx<4*DIM;iidx++)
214 /* Start outer loop over neighborlists */
215 for(iidx=0; iidx<nri; iidx++)
217 /* Load shift vector for this list */
218 i_shift_offset = DIM*shiftidx[iidx];
220 /* Load limits for loop over neighbors */
221 j_index_start = jindex[iidx];
222 j_index_end = jindex[iidx+1];
224 /* Get outer coordinate index */
226 i_coord_offset = DIM*inr;
228 /* Load i particle coords and add shift vector */
229 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
230 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
232 fix0 = _mm256_setzero_ps();
233 fiy0 = _mm256_setzero_ps();
234 fiz0 = _mm256_setzero_ps();
235 fix1 = _mm256_setzero_ps();
236 fiy1 = _mm256_setzero_ps();
237 fiz1 = _mm256_setzero_ps();
238 fix2 = _mm256_setzero_ps();
239 fiy2 = _mm256_setzero_ps();
240 fiz2 = _mm256_setzero_ps();
242 /* Reset potential sums */
243 velecsum = _mm256_setzero_ps();
244 vvdwsum = _mm256_setzero_ps();
246 /* Start inner kernel loop */
247 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
250 /* Get j neighbor index, and coordinate index */
259 j_coord_offsetA = DIM*jnrA;
260 j_coord_offsetB = DIM*jnrB;
261 j_coord_offsetC = DIM*jnrC;
262 j_coord_offsetD = DIM*jnrD;
263 j_coord_offsetE = DIM*jnrE;
264 j_coord_offsetF = DIM*jnrF;
265 j_coord_offsetG = DIM*jnrG;
266 j_coord_offsetH = DIM*jnrH;
268 /* load j atom coordinates */
269 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
270 x+j_coord_offsetC,x+j_coord_offsetD,
271 x+j_coord_offsetE,x+j_coord_offsetF,
272 x+j_coord_offsetG,x+j_coord_offsetH,
273 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
275 /* Calculate displacement vector */
276 dx00 = _mm256_sub_ps(ix0,jx0);
277 dy00 = _mm256_sub_ps(iy0,jy0);
278 dz00 = _mm256_sub_ps(iz0,jz0);
279 dx01 = _mm256_sub_ps(ix0,jx1);
280 dy01 = _mm256_sub_ps(iy0,jy1);
281 dz01 = _mm256_sub_ps(iz0,jz1);
282 dx02 = _mm256_sub_ps(ix0,jx2);
283 dy02 = _mm256_sub_ps(iy0,jy2);
284 dz02 = _mm256_sub_ps(iz0,jz2);
285 dx10 = _mm256_sub_ps(ix1,jx0);
286 dy10 = _mm256_sub_ps(iy1,jy0);
287 dz10 = _mm256_sub_ps(iz1,jz0);
288 dx11 = _mm256_sub_ps(ix1,jx1);
289 dy11 = _mm256_sub_ps(iy1,jy1);
290 dz11 = _mm256_sub_ps(iz1,jz1);
291 dx12 = _mm256_sub_ps(ix1,jx2);
292 dy12 = _mm256_sub_ps(iy1,jy2);
293 dz12 = _mm256_sub_ps(iz1,jz2);
294 dx20 = _mm256_sub_ps(ix2,jx0);
295 dy20 = _mm256_sub_ps(iy2,jy0);
296 dz20 = _mm256_sub_ps(iz2,jz0);
297 dx21 = _mm256_sub_ps(ix2,jx1);
298 dy21 = _mm256_sub_ps(iy2,jy1);
299 dz21 = _mm256_sub_ps(iz2,jz1);
300 dx22 = _mm256_sub_ps(ix2,jx2);
301 dy22 = _mm256_sub_ps(iy2,jy2);
302 dz22 = _mm256_sub_ps(iz2,jz2);
304 /* Calculate squared distance and things based on it */
305 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
306 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
307 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
308 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
309 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
310 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
311 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
312 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
313 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
315 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
316 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
317 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
318 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
319 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
320 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
321 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
322 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
323 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
325 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
326 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
327 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
328 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
329 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
330 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
331 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
332 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
333 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
335 fjx0 = _mm256_setzero_ps();
336 fjy0 = _mm256_setzero_ps();
337 fjz0 = _mm256_setzero_ps();
338 fjx1 = _mm256_setzero_ps();
339 fjy1 = _mm256_setzero_ps();
340 fjz1 = _mm256_setzero_ps();
341 fjx2 = _mm256_setzero_ps();
342 fjy2 = _mm256_setzero_ps();
343 fjz2 = _mm256_setzero_ps();
345 /**************************
346 * CALCULATE INTERACTIONS *
347 **************************/
349 r00 = _mm256_mul_ps(rsq00,rinv00);
351 /* EWALD ELECTROSTATICS */
353 /* Analytical PME correction */
354 zeta2 = _mm256_mul_ps(beta2,rsq00);
355 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
356 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
357 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
358 felec = _mm256_mul_ps(qq00,felec);
359 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
360 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
361 velec = _mm256_sub_ps(rinv00,pmecorrV);
362 velec = _mm256_mul_ps(qq00,velec);
364 /* Analytical LJ-PME */
365 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
366 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
367 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
368 exponent = gmx_simd_exp_r(ewcljrsq);
369 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
370 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
371 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
372 vvdw6 = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
373 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
374 vvdw = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
375 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
376 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
378 /* Update potential sum for this i atom from the interaction with this j atom. */
379 velecsum = _mm256_add_ps(velecsum,velec);
380 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
382 fscal = _mm256_add_ps(felec,fvdw);
384 /* Calculate temporary vectorial force */
385 tx = _mm256_mul_ps(fscal,dx00);
386 ty = _mm256_mul_ps(fscal,dy00);
387 tz = _mm256_mul_ps(fscal,dz00);
389 /* Update vectorial force */
390 fix0 = _mm256_add_ps(fix0,tx);
391 fiy0 = _mm256_add_ps(fiy0,ty);
392 fiz0 = _mm256_add_ps(fiz0,tz);
394 fjx0 = _mm256_add_ps(fjx0,tx);
395 fjy0 = _mm256_add_ps(fjy0,ty);
396 fjz0 = _mm256_add_ps(fjz0,tz);
398 /**************************
399 * CALCULATE INTERACTIONS *
400 **************************/
402 r01 = _mm256_mul_ps(rsq01,rinv01);
404 /* EWALD ELECTROSTATICS */
406 /* Analytical PME correction */
407 zeta2 = _mm256_mul_ps(beta2,rsq01);
408 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
409 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
410 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
411 felec = _mm256_mul_ps(qq01,felec);
412 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
413 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
414 velec = _mm256_sub_ps(rinv01,pmecorrV);
415 velec = _mm256_mul_ps(qq01,velec);
417 /* Update potential sum for this i atom from the interaction with this j atom. */
418 velecsum = _mm256_add_ps(velecsum,velec);
422 /* Calculate temporary vectorial force */
423 tx = _mm256_mul_ps(fscal,dx01);
424 ty = _mm256_mul_ps(fscal,dy01);
425 tz = _mm256_mul_ps(fscal,dz01);
427 /* Update vectorial force */
428 fix0 = _mm256_add_ps(fix0,tx);
429 fiy0 = _mm256_add_ps(fiy0,ty);
430 fiz0 = _mm256_add_ps(fiz0,tz);
432 fjx1 = _mm256_add_ps(fjx1,tx);
433 fjy1 = _mm256_add_ps(fjy1,ty);
434 fjz1 = _mm256_add_ps(fjz1,tz);
436 /**************************
437 * CALCULATE INTERACTIONS *
438 **************************/
440 r02 = _mm256_mul_ps(rsq02,rinv02);
442 /* EWALD ELECTROSTATICS */
444 /* Analytical PME correction */
445 zeta2 = _mm256_mul_ps(beta2,rsq02);
446 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
447 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
448 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
449 felec = _mm256_mul_ps(qq02,felec);
450 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
451 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
452 velec = _mm256_sub_ps(rinv02,pmecorrV);
453 velec = _mm256_mul_ps(qq02,velec);
455 /* Update potential sum for this i atom from the interaction with this j atom. */
456 velecsum = _mm256_add_ps(velecsum,velec);
460 /* Calculate temporary vectorial force */
461 tx = _mm256_mul_ps(fscal,dx02);
462 ty = _mm256_mul_ps(fscal,dy02);
463 tz = _mm256_mul_ps(fscal,dz02);
465 /* Update vectorial force */
466 fix0 = _mm256_add_ps(fix0,tx);
467 fiy0 = _mm256_add_ps(fiy0,ty);
468 fiz0 = _mm256_add_ps(fiz0,tz);
470 fjx2 = _mm256_add_ps(fjx2,tx);
471 fjy2 = _mm256_add_ps(fjy2,ty);
472 fjz2 = _mm256_add_ps(fjz2,tz);
474 /**************************
475 * CALCULATE INTERACTIONS *
476 **************************/
478 r10 = _mm256_mul_ps(rsq10,rinv10);
480 /* EWALD ELECTROSTATICS */
482 /* Analytical PME correction */
483 zeta2 = _mm256_mul_ps(beta2,rsq10);
484 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
485 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
486 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
487 felec = _mm256_mul_ps(qq10,felec);
488 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
489 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
490 velec = _mm256_sub_ps(rinv10,pmecorrV);
491 velec = _mm256_mul_ps(qq10,velec);
493 /* Update potential sum for this i atom from the interaction with this j atom. */
494 velecsum = _mm256_add_ps(velecsum,velec);
498 /* Calculate temporary vectorial force */
499 tx = _mm256_mul_ps(fscal,dx10);
500 ty = _mm256_mul_ps(fscal,dy10);
501 tz = _mm256_mul_ps(fscal,dz10);
503 /* Update vectorial force */
504 fix1 = _mm256_add_ps(fix1,tx);
505 fiy1 = _mm256_add_ps(fiy1,ty);
506 fiz1 = _mm256_add_ps(fiz1,tz);
508 fjx0 = _mm256_add_ps(fjx0,tx);
509 fjy0 = _mm256_add_ps(fjy0,ty);
510 fjz0 = _mm256_add_ps(fjz0,tz);
512 /**************************
513 * CALCULATE INTERACTIONS *
514 **************************/
516 r11 = _mm256_mul_ps(rsq11,rinv11);
518 /* EWALD ELECTROSTATICS */
520 /* Analytical PME correction */
521 zeta2 = _mm256_mul_ps(beta2,rsq11);
522 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
523 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
524 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
525 felec = _mm256_mul_ps(qq11,felec);
526 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
527 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
528 velec = _mm256_sub_ps(rinv11,pmecorrV);
529 velec = _mm256_mul_ps(qq11,velec);
531 /* Update potential sum for this i atom from the interaction with this j atom. */
532 velecsum = _mm256_add_ps(velecsum,velec);
536 /* Calculate temporary vectorial force */
537 tx = _mm256_mul_ps(fscal,dx11);
538 ty = _mm256_mul_ps(fscal,dy11);
539 tz = _mm256_mul_ps(fscal,dz11);
541 /* Update vectorial force */
542 fix1 = _mm256_add_ps(fix1,tx);
543 fiy1 = _mm256_add_ps(fiy1,ty);
544 fiz1 = _mm256_add_ps(fiz1,tz);
546 fjx1 = _mm256_add_ps(fjx1,tx);
547 fjy1 = _mm256_add_ps(fjy1,ty);
548 fjz1 = _mm256_add_ps(fjz1,tz);
550 /**************************
551 * CALCULATE INTERACTIONS *
552 **************************/
554 r12 = _mm256_mul_ps(rsq12,rinv12);
556 /* EWALD ELECTROSTATICS */
558 /* Analytical PME correction */
559 zeta2 = _mm256_mul_ps(beta2,rsq12);
560 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
561 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
562 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
563 felec = _mm256_mul_ps(qq12,felec);
564 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
565 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
566 velec = _mm256_sub_ps(rinv12,pmecorrV);
567 velec = _mm256_mul_ps(qq12,velec);
569 /* Update potential sum for this i atom from the interaction with this j atom. */
570 velecsum = _mm256_add_ps(velecsum,velec);
574 /* Calculate temporary vectorial force */
575 tx = _mm256_mul_ps(fscal,dx12);
576 ty = _mm256_mul_ps(fscal,dy12);
577 tz = _mm256_mul_ps(fscal,dz12);
579 /* Update vectorial force */
580 fix1 = _mm256_add_ps(fix1,tx);
581 fiy1 = _mm256_add_ps(fiy1,ty);
582 fiz1 = _mm256_add_ps(fiz1,tz);
584 fjx2 = _mm256_add_ps(fjx2,tx);
585 fjy2 = _mm256_add_ps(fjy2,ty);
586 fjz2 = _mm256_add_ps(fjz2,tz);
588 /**************************
589 * CALCULATE INTERACTIONS *
590 **************************/
592 r20 = _mm256_mul_ps(rsq20,rinv20);
594 /* EWALD ELECTROSTATICS */
596 /* Analytical PME correction */
597 zeta2 = _mm256_mul_ps(beta2,rsq20);
598 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
599 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
600 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
601 felec = _mm256_mul_ps(qq20,felec);
602 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
603 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
604 velec = _mm256_sub_ps(rinv20,pmecorrV);
605 velec = _mm256_mul_ps(qq20,velec);
607 /* Update potential sum for this i atom from the interaction with this j atom. */
608 velecsum = _mm256_add_ps(velecsum,velec);
612 /* Calculate temporary vectorial force */
613 tx = _mm256_mul_ps(fscal,dx20);
614 ty = _mm256_mul_ps(fscal,dy20);
615 tz = _mm256_mul_ps(fscal,dz20);
617 /* Update vectorial force */
618 fix2 = _mm256_add_ps(fix2,tx);
619 fiy2 = _mm256_add_ps(fiy2,ty);
620 fiz2 = _mm256_add_ps(fiz2,tz);
622 fjx0 = _mm256_add_ps(fjx0,tx);
623 fjy0 = _mm256_add_ps(fjy0,ty);
624 fjz0 = _mm256_add_ps(fjz0,tz);
626 /**************************
627 * CALCULATE INTERACTIONS *
628 **************************/
630 r21 = _mm256_mul_ps(rsq21,rinv21);
632 /* EWALD ELECTROSTATICS */
634 /* Analytical PME correction */
635 zeta2 = _mm256_mul_ps(beta2,rsq21);
636 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
637 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
638 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
639 felec = _mm256_mul_ps(qq21,felec);
640 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
641 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
642 velec = _mm256_sub_ps(rinv21,pmecorrV);
643 velec = _mm256_mul_ps(qq21,velec);
645 /* Update potential sum for this i atom from the interaction with this j atom. */
646 velecsum = _mm256_add_ps(velecsum,velec);
650 /* Calculate temporary vectorial force */
651 tx = _mm256_mul_ps(fscal,dx21);
652 ty = _mm256_mul_ps(fscal,dy21);
653 tz = _mm256_mul_ps(fscal,dz21);
655 /* Update vectorial force */
656 fix2 = _mm256_add_ps(fix2,tx);
657 fiy2 = _mm256_add_ps(fiy2,ty);
658 fiz2 = _mm256_add_ps(fiz2,tz);
660 fjx1 = _mm256_add_ps(fjx1,tx);
661 fjy1 = _mm256_add_ps(fjy1,ty);
662 fjz1 = _mm256_add_ps(fjz1,tz);
664 /**************************
665 * CALCULATE INTERACTIONS *
666 **************************/
668 r22 = _mm256_mul_ps(rsq22,rinv22);
670 /* EWALD ELECTROSTATICS */
672 /* Analytical PME correction */
673 zeta2 = _mm256_mul_ps(beta2,rsq22);
674 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
675 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
676 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
677 felec = _mm256_mul_ps(qq22,felec);
678 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
679 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
680 velec = _mm256_sub_ps(rinv22,pmecorrV);
681 velec = _mm256_mul_ps(qq22,velec);
683 /* Update potential sum for this i atom from the interaction with this j atom. */
684 velecsum = _mm256_add_ps(velecsum,velec);
688 /* Calculate temporary vectorial force */
689 tx = _mm256_mul_ps(fscal,dx22);
690 ty = _mm256_mul_ps(fscal,dy22);
691 tz = _mm256_mul_ps(fscal,dz22);
693 /* Update vectorial force */
694 fix2 = _mm256_add_ps(fix2,tx);
695 fiy2 = _mm256_add_ps(fiy2,ty);
696 fiz2 = _mm256_add_ps(fiz2,tz);
698 fjx2 = _mm256_add_ps(fjx2,tx);
699 fjy2 = _mm256_add_ps(fjy2,ty);
700 fjz2 = _mm256_add_ps(fjz2,tz);
702 fjptrA = f+j_coord_offsetA;
703 fjptrB = f+j_coord_offsetB;
704 fjptrC = f+j_coord_offsetC;
705 fjptrD = f+j_coord_offsetD;
706 fjptrE = f+j_coord_offsetE;
707 fjptrF = f+j_coord_offsetF;
708 fjptrG = f+j_coord_offsetG;
709 fjptrH = f+j_coord_offsetH;
711 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
712 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
714 /* Inner loop uses 784 flops */
720 /* Get j neighbor index, and coordinate index */
721 jnrlistA = jjnr[jidx];
722 jnrlistB = jjnr[jidx+1];
723 jnrlistC = jjnr[jidx+2];
724 jnrlistD = jjnr[jidx+3];
725 jnrlistE = jjnr[jidx+4];
726 jnrlistF = jjnr[jidx+5];
727 jnrlistG = jjnr[jidx+6];
728 jnrlistH = jjnr[jidx+7];
729 /* Sign of each element will be negative for non-real atoms.
730 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
731 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
733 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
734 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
736 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
737 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
738 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
739 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
740 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
741 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
742 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
743 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
744 j_coord_offsetA = DIM*jnrA;
745 j_coord_offsetB = DIM*jnrB;
746 j_coord_offsetC = DIM*jnrC;
747 j_coord_offsetD = DIM*jnrD;
748 j_coord_offsetE = DIM*jnrE;
749 j_coord_offsetF = DIM*jnrF;
750 j_coord_offsetG = DIM*jnrG;
751 j_coord_offsetH = DIM*jnrH;
753 /* load j atom coordinates */
754 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
755 x+j_coord_offsetC,x+j_coord_offsetD,
756 x+j_coord_offsetE,x+j_coord_offsetF,
757 x+j_coord_offsetG,x+j_coord_offsetH,
758 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
760 /* Calculate displacement vector */
761 dx00 = _mm256_sub_ps(ix0,jx0);
762 dy00 = _mm256_sub_ps(iy0,jy0);
763 dz00 = _mm256_sub_ps(iz0,jz0);
764 dx01 = _mm256_sub_ps(ix0,jx1);
765 dy01 = _mm256_sub_ps(iy0,jy1);
766 dz01 = _mm256_sub_ps(iz0,jz1);
767 dx02 = _mm256_sub_ps(ix0,jx2);
768 dy02 = _mm256_sub_ps(iy0,jy2);
769 dz02 = _mm256_sub_ps(iz0,jz2);
770 dx10 = _mm256_sub_ps(ix1,jx0);
771 dy10 = _mm256_sub_ps(iy1,jy0);
772 dz10 = _mm256_sub_ps(iz1,jz0);
773 dx11 = _mm256_sub_ps(ix1,jx1);
774 dy11 = _mm256_sub_ps(iy1,jy1);
775 dz11 = _mm256_sub_ps(iz1,jz1);
776 dx12 = _mm256_sub_ps(ix1,jx2);
777 dy12 = _mm256_sub_ps(iy1,jy2);
778 dz12 = _mm256_sub_ps(iz1,jz2);
779 dx20 = _mm256_sub_ps(ix2,jx0);
780 dy20 = _mm256_sub_ps(iy2,jy0);
781 dz20 = _mm256_sub_ps(iz2,jz0);
782 dx21 = _mm256_sub_ps(ix2,jx1);
783 dy21 = _mm256_sub_ps(iy2,jy1);
784 dz21 = _mm256_sub_ps(iz2,jz1);
785 dx22 = _mm256_sub_ps(ix2,jx2);
786 dy22 = _mm256_sub_ps(iy2,jy2);
787 dz22 = _mm256_sub_ps(iz2,jz2);
789 /* Calculate squared distance and things based on it */
790 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
791 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
792 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
793 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
794 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
795 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
796 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
797 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
798 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
800 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
801 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
802 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
803 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
804 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
805 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
806 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
807 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
808 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
810 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
811 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
812 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
813 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
814 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
815 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
816 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
817 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
818 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
820 fjx0 = _mm256_setzero_ps();
821 fjy0 = _mm256_setzero_ps();
822 fjz0 = _mm256_setzero_ps();
823 fjx1 = _mm256_setzero_ps();
824 fjy1 = _mm256_setzero_ps();
825 fjz1 = _mm256_setzero_ps();
826 fjx2 = _mm256_setzero_ps();
827 fjy2 = _mm256_setzero_ps();
828 fjz2 = _mm256_setzero_ps();
830 /**************************
831 * CALCULATE INTERACTIONS *
832 **************************/
834 r00 = _mm256_mul_ps(rsq00,rinv00);
835 r00 = _mm256_andnot_ps(dummy_mask,r00);
837 /* EWALD ELECTROSTATICS */
839 /* Analytical PME correction */
840 zeta2 = _mm256_mul_ps(beta2,rsq00);
841 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
842 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
843 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
844 felec = _mm256_mul_ps(qq00,felec);
845 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
846 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
847 velec = _mm256_sub_ps(rinv00,pmecorrV);
848 velec = _mm256_mul_ps(qq00,velec);
850 /* Analytical LJ-PME */
851 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
852 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
853 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
854 exponent = gmx_simd_exp_r(ewcljrsq);
855 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
856 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
857 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
858 vvdw6 = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
859 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
860 vvdw = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
861 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
862 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,_mm256_sub_ps(vvdw6,_mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6)))),rinvsq00);
864 /* Update potential sum for this i atom from the interaction with this j atom. */
865 velec = _mm256_andnot_ps(dummy_mask,velec);
866 velecsum = _mm256_add_ps(velecsum,velec);
867 vvdw = _mm256_andnot_ps(dummy_mask,vvdw);
868 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
870 fscal = _mm256_add_ps(felec,fvdw);
872 fscal = _mm256_andnot_ps(dummy_mask,fscal);
874 /* Calculate temporary vectorial force */
875 tx = _mm256_mul_ps(fscal,dx00);
876 ty = _mm256_mul_ps(fscal,dy00);
877 tz = _mm256_mul_ps(fscal,dz00);
879 /* Update vectorial force */
880 fix0 = _mm256_add_ps(fix0,tx);
881 fiy0 = _mm256_add_ps(fiy0,ty);
882 fiz0 = _mm256_add_ps(fiz0,tz);
884 fjx0 = _mm256_add_ps(fjx0,tx);
885 fjy0 = _mm256_add_ps(fjy0,ty);
886 fjz0 = _mm256_add_ps(fjz0,tz);
888 /**************************
889 * CALCULATE INTERACTIONS *
890 **************************/
892 r01 = _mm256_mul_ps(rsq01,rinv01);
893 r01 = _mm256_andnot_ps(dummy_mask,r01);
895 /* EWALD ELECTROSTATICS */
897 /* Analytical PME correction */
898 zeta2 = _mm256_mul_ps(beta2,rsq01);
899 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
900 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
901 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
902 felec = _mm256_mul_ps(qq01,felec);
903 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
904 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
905 velec = _mm256_sub_ps(rinv01,pmecorrV);
906 velec = _mm256_mul_ps(qq01,velec);
908 /* Update potential sum for this i atom from the interaction with this j atom. */
909 velec = _mm256_andnot_ps(dummy_mask,velec);
910 velecsum = _mm256_add_ps(velecsum,velec);
914 fscal = _mm256_andnot_ps(dummy_mask,fscal);
916 /* Calculate temporary vectorial force */
917 tx = _mm256_mul_ps(fscal,dx01);
918 ty = _mm256_mul_ps(fscal,dy01);
919 tz = _mm256_mul_ps(fscal,dz01);
921 /* Update vectorial force */
922 fix0 = _mm256_add_ps(fix0,tx);
923 fiy0 = _mm256_add_ps(fiy0,ty);
924 fiz0 = _mm256_add_ps(fiz0,tz);
926 fjx1 = _mm256_add_ps(fjx1,tx);
927 fjy1 = _mm256_add_ps(fjy1,ty);
928 fjz1 = _mm256_add_ps(fjz1,tz);
930 /**************************
931 * CALCULATE INTERACTIONS *
932 **************************/
934 r02 = _mm256_mul_ps(rsq02,rinv02);
935 r02 = _mm256_andnot_ps(dummy_mask,r02);
937 /* EWALD ELECTROSTATICS */
939 /* Analytical PME correction */
940 zeta2 = _mm256_mul_ps(beta2,rsq02);
941 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
942 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
943 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
944 felec = _mm256_mul_ps(qq02,felec);
945 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
946 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
947 velec = _mm256_sub_ps(rinv02,pmecorrV);
948 velec = _mm256_mul_ps(qq02,velec);
950 /* Update potential sum for this i atom from the interaction with this j atom. */
951 velec = _mm256_andnot_ps(dummy_mask,velec);
952 velecsum = _mm256_add_ps(velecsum,velec);
956 fscal = _mm256_andnot_ps(dummy_mask,fscal);
958 /* Calculate temporary vectorial force */
959 tx = _mm256_mul_ps(fscal,dx02);
960 ty = _mm256_mul_ps(fscal,dy02);
961 tz = _mm256_mul_ps(fscal,dz02);
963 /* Update vectorial force */
964 fix0 = _mm256_add_ps(fix0,tx);
965 fiy0 = _mm256_add_ps(fiy0,ty);
966 fiz0 = _mm256_add_ps(fiz0,tz);
968 fjx2 = _mm256_add_ps(fjx2,tx);
969 fjy2 = _mm256_add_ps(fjy2,ty);
970 fjz2 = _mm256_add_ps(fjz2,tz);
972 /**************************
973 * CALCULATE INTERACTIONS *
974 **************************/
976 r10 = _mm256_mul_ps(rsq10,rinv10);
977 r10 = _mm256_andnot_ps(dummy_mask,r10);
979 /* EWALD ELECTROSTATICS */
981 /* Analytical PME correction */
982 zeta2 = _mm256_mul_ps(beta2,rsq10);
983 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
984 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
985 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
986 felec = _mm256_mul_ps(qq10,felec);
987 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
988 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
989 velec = _mm256_sub_ps(rinv10,pmecorrV);
990 velec = _mm256_mul_ps(qq10,velec);
992 /* Update potential sum for this i atom from the interaction with this j atom. */
993 velec = _mm256_andnot_ps(dummy_mask,velec);
994 velecsum = _mm256_add_ps(velecsum,velec);
998 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1000 /* Calculate temporary vectorial force */
1001 tx = _mm256_mul_ps(fscal,dx10);
1002 ty = _mm256_mul_ps(fscal,dy10);
1003 tz = _mm256_mul_ps(fscal,dz10);
1005 /* Update vectorial force */
1006 fix1 = _mm256_add_ps(fix1,tx);
1007 fiy1 = _mm256_add_ps(fiy1,ty);
1008 fiz1 = _mm256_add_ps(fiz1,tz);
1010 fjx0 = _mm256_add_ps(fjx0,tx);
1011 fjy0 = _mm256_add_ps(fjy0,ty);
1012 fjz0 = _mm256_add_ps(fjz0,tz);
1014 /**************************
1015 * CALCULATE INTERACTIONS *
1016 **************************/
1018 r11 = _mm256_mul_ps(rsq11,rinv11);
1019 r11 = _mm256_andnot_ps(dummy_mask,r11);
1021 /* EWALD ELECTROSTATICS */
1023 /* Analytical PME correction */
1024 zeta2 = _mm256_mul_ps(beta2,rsq11);
1025 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
1026 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1027 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1028 felec = _mm256_mul_ps(qq11,felec);
1029 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1030 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1031 velec = _mm256_sub_ps(rinv11,pmecorrV);
1032 velec = _mm256_mul_ps(qq11,velec);
1034 /* Update potential sum for this i atom from the interaction with this j atom. */
1035 velec = _mm256_andnot_ps(dummy_mask,velec);
1036 velecsum = _mm256_add_ps(velecsum,velec);
1040 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1042 /* Calculate temporary vectorial force */
1043 tx = _mm256_mul_ps(fscal,dx11);
1044 ty = _mm256_mul_ps(fscal,dy11);
1045 tz = _mm256_mul_ps(fscal,dz11);
1047 /* Update vectorial force */
1048 fix1 = _mm256_add_ps(fix1,tx);
1049 fiy1 = _mm256_add_ps(fiy1,ty);
1050 fiz1 = _mm256_add_ps(fiz1,tz);
1052 fjx1 = _mm256_add_ps(fjx1,tx);
1053 fjy1 = _mm256_add_ps(fjy1,ty);
1054 fjz1 = _mm256_add_ps(fjz1,tz);
1056 /**************************
1057 * CALCULATE INTERACTIONS *
1058 **************************/
1060 r12 = _mm256_mul_ps(rsq12,rinv12);
1061 r12 = _mm256_andnot_ps(dummy_mask,r12);
1063 /* EWALD ELECTROSTATICS */
1065 /* Analytical PME correction */
1066 zeta2 = _mm256_mul_ps(beta2,rsq12);
1067 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
1068 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1069 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1070 felec = _mm256_mul_ps(qq12,felec);
1071 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1072 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1073 velec = _mm256_sub_ps(rinv12,pmecorrV);
1074 velec = _mm256_mul_ps(qq12,velec);
1076 /* Update potential sum for this i atom from the interaction with this j atom. */
1077 velec = _mm256_andnot_ps(dummy_mask,velec);
1078 velecsum = _mm256_add_ps(velecsum,velec);
1082 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1084 /* Calculate temporary vectorial force */
1085 tx = _mm256_mul_ps(fscal,dx12);
1086 ty = _mm256_mul_ps(fscal,dy12);
1087 tz = _mm256_mul_ps(fscal,dz12);
1089 /* Update vectorial force */
1090 fix1 = _mm256_add_ps(fix1,tx);
1091 fiy1 = _mm256_add_ps(fiy1,ty);
1092 fiz1 = _mm256_add_ps(fiz1,tz);
1094 fjx2 = _mm256_add_ps(fjx2,tx);
1095 fjy2 = _mm256_add_ps(fjy2,ty);
1096 fjz2 = _mm256_add_ps(fjz2,tz);
1098 /**************************
1099 * CALCULATE INTERACTIONS *
1100 **************************/
1102 r20 = _mm256_mul_ps(rsq20,rinv20);
1103 r20 = _mm256_andnot_ps(dummy_mask,r20);
1105 /* EWALD ELECTROSTATICS */
1107 /* Analytical PME correction */
1108 zeta2 = _mm256_mul_ps(beta2,rsq20);
1109 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1110 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1111 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1112 felec = _mm256_mul_ps(qq20,felec);
1113 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1114 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1115 velec = _mm256_sub_ps(rinv20,pmecorrV);
1116 velec = _mm256_mul_ps(qq20,velec);
1118 /* Update potential sum for this i atom from the interaction with this j atom. */
1119 velec = _mm256_andnot_ps(dummy_mask,velec);
1120 velecsum = _mm256_add_ps(velecsum,velec);
1124 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1126 /* Calculate temporary vectorial force */
1127 tx = _mm256_mul_ps(fscal,dx20);
1128 ty = _mm256_mul_ps(fscal,dy20);
1129 tz = _mm256_mul_ps(fscal,dz20);
1131 /* Update vectorial force */
1132 fix2 = _mm256_add_ps(fix2,tx);
1133 fiy2 = _mm256_add_ps(fiy2,ty);
1134 fiz2 = _mm256_add_ps(fiz2,tz);
1136 fjx0 = _mm256_add_ps(fjx0,tx);
1137 fjy0 = _mm256_add_ps(fjy0,ty);
1138 fjz0 = _mm256_add_ps(fjz0,tz);
1140 /**************************
1141 * CALCULATE INTERACTIONS *
1142 **************************/
1144 r21 = _mm256_mul_ps(rsq21,rinv21);
1145 r21 = _mm256_andnot_ps(dummy_mask,r21);
1147 /* EWALD ELECTROSTATICS */
1149 /* Analytical PME correction */
1150 zeta2 = _mm256_mul_ps(beta2,rsq21);
1151 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
1152 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1153 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1154 felec = _mm256_mul_ps(qq21,felec);
1155 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1156 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1157 velec = _mm256_sub_ps(rinv21,pmecorrV);
1158 velec = _mm256_mul_ps(qq21,velec);
1160 /* Update potential sum for this i atom from the interaction with this j atom. */
1161 velec = _mm256_andnot_ps(dummy_mask,velec);
1162 velecsum = _mm256_add_ps(velecsum,velec);
1166 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1168 /* Calculate temporary vectorial force */
1169 tx = _mm256_mul_ps(fscal,dx21);
1170 ty = _mm256_mul_ps(fscal,dy21);
1171 tz = _mm256_mul_ps(fscal,dz21);
1173 /* Update vectorial force */
1174 fix2 = _mm256_add_ps(fix2,tx);
1175 fiy2 = _mm256_add_ps(fiy2,ty);
1176 fiz2 = _mm256_add_ps(fiz2,tz);
1178 fjx1 = _mm256_add_ps(fjx1,tx);
1179 fjy1 = _mm256_add_ps(fjy1,ty);
1180 fjz1 = _mm256_add_ps(fjz1,tz);
1182 /**************************
1183 * CALCULATE INTERACTIONS *
1184 **************************/
1186 r22 = _mm256_mul_ps(rsq22,rinv22);
1187 r22 = _mm256_andnot_ps(dummy_mask,r22);
1189 /* EWALD ELECTROSTATICS */
1191 /* Analytical PME correction */
1192 zeta2 = _mm256_mul_ps(beta2,rsq22);
1193 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
1194 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1195 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1196 felec = _mm256_mul_ps(qq22,felec);
1197 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1198 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1199 velec = _mm256_sub_ps(rinv22,pmecorrV);
1200 velec = _mm256_mul_ps(qq22,velec);
1202 /* Update potential sum for this i atom from the interaction with this j atom. */
1203 velec = _mm256_andnot_ps(dummy_mask,velec);
1204 velecsum = _mm256_add_ps(velecsum,velec);
1208 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1210 /* Calculate temporary vectorial force */
1211 tx = _mm256_mul_ps(fscal,dx22);
1212 ty = _mm256_mul_ps(fscal,dy22);
1213 tz = _mm256_mul_ps(fscal,dz22);
1215 /* Update vectorial force */
1216 fix2 = _mm256_add_ps(fix2,tx);
1217 fiy2 = _mm256_add_ps(fiy2,ty);
1218 fiz2 = _mm256_add_ps(fiz2,tz);
1220 fjx2 = _mm256_add_ps(fjx2,tx);
1221 fjy2 = _mm256_add_ps(fjy2,ty);
1222 fjz2 = _mm256_add_ps(fjz2,tz);
1224 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1225 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1226 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1227 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1228 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1229 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1230 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1231 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1233 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1234 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1236 /* Inner loop uses 793 flops */
1239 /* End of innermost loop */
1241 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1242 f+i_coord_offset,fshift+i_shift_offset);
1245 /* Update potential energies */
1246 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1247 gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1249 /* Increment number of inner iterations */
1250 inneriter += j_index_end - j_index_start;
1252 /* Outer loop uses 20 flops */
1255 /* Increment number of outer iterations */
1258 /* Update outer/inner flops */
1260 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*793);
1263 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_256_single
1264 * Electrostatics interaction: Ewald
1265 * VdW interaction: LJEwald
1266 * Geometry: Water3-Water3
1267 * Calculate force/pot: Force
1270 nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_256_single
1271 (t_nblist * gmx_restrict nlist,
1272 rvec * gmx_restrict xx,
1273 rvec * gmx_restrict ff,
1274 t_forcerec * gmx_restrict fr,
1275 t_mdatoms * gmx_restrict mdatoms,
1276 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1277 t_nrnb * gmx_restrict nrnb)
1279 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1280 * just 0 for non-waters.
1281 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1282 * jnr indices corresponding to data put in the four positions in the SIMD register.
1284 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1285 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1286 int jnrA,jnrB,jnrC,jnrD;
1287 int jnrE,jnrF,jnrG,jnrH;
1288 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1289 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1290 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1291 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1292 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1293 real rcutoff_scalar;
1294 real *shiftvec,*fshift,*x,*f;
1295 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1296 real scratch[4*DIM];
1297 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1298 real * vdwioffsetptr0;
1299 real * vdwgridioffsetptr0;
1300 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1301 real * vdwioffsetptr1;
1302 real * vdwgridioffsetptr1;
1303 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1304 real * vdwioffsetptr2;
1305 real * vdwgridioffsetptr2;
1306 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1307 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1308 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1309 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1310 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1311 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1312 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1313 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1314 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1315 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1316 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1317 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1318 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1319 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1320 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1321 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1322 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
1325 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1328 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
1329 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
1340 __m256 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1341 __m256 one_half = _mm256_set1_ps(0.5);
1342 __m256 minus_one = _mm256_set1_ps(-1.0);
1344 __m128i ewitab_lo,ewitab_hi;
1345 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1346 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1348 __m256 dummy_mask,cutoff_mask;
1349 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1350 __m256 one = _mm256_set1_ps(1.0);
1351 __m256 two = _mm256_set1_ps(2.0);
1357 jindex = nlist->jindex;
1359 shiftidx = nlist->shift;
1361 shiftvec = fr->shift_vec[0];
1362 fshift = fr->fshift[0];
1363 facel = _mm256_set1_ps(fr->epsfac);
1364 charge = mdatoms->chargeA;
1365 nvdwtype = fr->ntype;
1366 vdwparam = fr->nbfp;
1367 vdwtype = mdatoms->typeA;
1368 vdwgridparam = fr->ljpme_c6grid;
1369 sh_lj_ewald = _mm256_set1_ps(fr->ic->sh_lj_ewald);
1370 ewclj = _mm256_set1_ps(fr->ewaldcoeff_lj);
1371 ewclj2 = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
1373 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
1374 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1375 beta2 = _mm256_mul_ps(beta,beta);
1376 beta3 = _mm256_mul_ps(beta,beta2);
1378 ewtab = fr->ic->tabq_coul_F;
1379 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
1380 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1382 /* Setup water-specific parameters */
1383 inr = nlist->iinr[0];
1384 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1385 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1386 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1387 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
1388 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
1390 jq0 = _mm256_set1_ps(charge[inr+0]);
1391 jq1 = _mm256_set1_ps(charge[inr+1]);
1392 jq2 = _mm256_set1_ps(charge[inr+2]);
1393 vdwjidx0A = 2*vdwtype[inr+0];
1394 qq00 = _mm256_mul_ps(iq0,jq0);
1395 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1396 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1397 c6grid_00 = _mm256_set1_ps(vdwgridioffsetptr0[vdwjidx0A]);
1398 qq01 = _mm256_mul_ps(iq0,jq1);
1399 qq02 = _mm256_mul_ps(iq0,jq2);
1400 qq10 = _mm256_mul_ps(iq1,jq0);
1401 qq11 = _mm256_mul_ps(iq1,jq1);
1402 qq12 = _mm256_mul_ps(iq1,jq2);
1403 qq20 = _mm256_mul_ps(iq2,jq0);
1404 qq21 = _mm256_mul_ps(iq2,jq1);
1405 qq22 = _mm256_mul_ps(iq2,jq2);
1407 /* Avoid stupid compiler warnings */
1408 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1409 j_coord_offsetA = 0;
1410 j_coord_offsetB = 0;
1411 j_coord_offsetC = 0;
1412 j_coord_offsetD = 0;
1413 j_coord_offsetE = 0;
1414 j_coord_offsetF = 0;
1415 j_coord_offsetG = 0;
1416 j_coord_offsetH = 0;
1421 for(iidx=0;iidx<4*DIM;iidx++)
1423 scratch[iidx] = 0.0;
1426 /* Start outer loop over neighborlists */
1427 for(iidx=0; iidx<nri; iidx++)
1429 /* Load shift vector for this list */
1430 i_shift_offset = DIM*shiftidx[iidx];
1432 /* Load limits for loop over neighbors */
1433 j_index_start = jindex[iidx];
1434 j_index_end = jindex[iidx+1];
1436 /* Get outer coordinate index */
1438 i_coord_offset = DIM*inr;
1440 /* Load i particle coords and add shift vector */
1441 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1442 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1444 fix0 = _mm256_setzero_ps();
1445 fiy0 = _mm256_setzero_ps();
1446 fiz0 = _mm256_setzero_ps();
1447 fix1 = _mm256_setzero_ps();
1448 fiy1 = _mm256_setzero_ps();
1449 fiz1 = _mm256_setzero_ps();
1450 fix2 = _mm256_setzero_ps();
1451 fiy2 = _mm256_setzero_ps();
1452 fiz2 = _mm256_setzero_ps();
1454 /* Start inner kernel loop */
1455 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1458 /* Get j neighbor index, and coordinate index */
1460 jnrB = jjnr[jidx+1];
1461 jnrC = jjnr[jidx+2];
1462 jnrD = jjnr[jidx+3];
1463 jnrE = jjnr[jidx+4];
1464 jnrF = jjnr[jidx+5];
1465 jnrG = jjnr[jidx+6];
1466 jnrH = jjnr[jidx+7];
1467 j_coord_offsetA = DIM*jnrA;
1468 j_coord_offsetB = DIM*jnrB;
1469 j_coord_offsetC = DIM*jnrC;
1470 j_coord_offsetD = DIM*jnrD;
1471 j_coord_offsetE = DIM*jnrE;
1472 j_coord_offsetF = DIM*jnrF;
1473 j_coord_offsetG = DIM*jnrG;
1474 j_coord_offsetH = DIM*jnrH;
1476 /* load j atom coordinates */
1477 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1478 x+j_coord_offsetC,x+j_coord_offsetD,
1479 x+j_coord_offsetE,x+j_coord_offsetF,
1480 x+j_coord_offsetG,x+j_coord_offsetH,
1481 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1483 /* Calculate displacement vector */
1484 dx00 = _mm256_sub_ps(ix0,jx0);
1485 dy00 = _mm256_sub_ps(iy0,jy0);
1486 dz00 = _mm256_sub_ps(iz0,jz0);
1487 dx01 = _mm256_sub_ps(ix0,jx1);
1488 dy01 = _mm256_sub_ps(iy0,jy1);
1489 dz01 = _mm256_sub_ps(iz0,jz1);
1490 dx02 = _mm256_sub_ps(ix0,jx2);
1491 dy02 = _mm256_sub_ps(iy0,jy2);
1492 dz02 = _mm256_sub_ps(iz0,jz2);
1493 dx10 = _mm256_sub_ps(ix1,jx0);
1494 dy10 = _mm256_sub_ps(iy1,jy0);
1495 dz10 = _mm256_sub_ps(iz1,jz0);
1496 dx11 = _mm256_sub_ps(ix1,jx1);
1497 dy11 = _mm256_sub_ps(iy1,jy1);
1498 dz11 = _mm256_sub_ps(iz1,jz1);
1499 dx12 = _mm256_sub_ps(ix1,jx2);
1500 dy12 = _mm256_sub_ps(iy1,jy2);
1501 dz12 = _mm256_sub_ps(iz1,jz2);
1502 dx20 = _mm256_sub_ps(ix2,jx0);
1503 dy20 = _mm256_sub_ps(iy2,jy0);
1504 dz20 = _mm256_sub_ps(iz2,jz0);
1505 dx21 = _mm256_sub_ps(ix2,jx1);
1506 dy21 = _mm256_sub_ps(iy2,jy1);
1507 dz21 = _mm256_sub_ps(iz2,jz1);
1508 dx22 = _mm256_sub_ps(ix2,jx2);
1509 dy22 = _mm256_sub_ps(iy2,jy2);
1510 dz22 = _mm256_sub_ps(iz2,jz2);
1512 /* Calculate squared distance and things based on it */
1513 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1514 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1515 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1516 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1517 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1518 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1519 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1520 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1521 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1523 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1524 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1525 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1526 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1527 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1528 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1529 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1530 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1531 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1533 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1534 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1535 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1536 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1537 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1538 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1539 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1540 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1541 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1543 fjx0 = _mm256_setzero_ps();
1544 fjy0 = _mm256_setzero_ps();
1545 fjz0 = _mm256_setzero_ps();
1546 fjx1 = _mm256_setzero_ps();
1547 fjy1 = _mm256_setzero_ps();
1548 fjz1 = _mm256_setzero_ps();
1549 fjx2 = _mm256_setzero_ps();
1550 fjy2 = _mm256_setzero_ps();
1551 fjz2 = _mm256_setzero_ps();
1553 /**************************
1554 * CALCULATE INTERACTIONS *
1555 **************************/
1557 r00 = _mm256_mul_ps(rsq00,rinv00);
1559 /* EWALD ELECTROSTATICS */
1561 /* Analytical PME correction */
1562 zeta2 = _mm256_mul_ps(beta2,rsq00);
1563 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1564 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1565 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1566 felec = _mm256_mul_ps(qq00,felec);
1568 /* Analytical LJ-PME */
1569 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1570 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
1571 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
1572 exponent = gmx_simd_exp_r(ewcljrsq);
1573 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1574 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
1575 /* f6A = 6 * C6grid * (1 - poly) */
1576 f6A = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
1577 /* f6B = C6grid * exponent * beta^6 */
1578 f6B = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
1579 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1580 fvdw = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1582 fscal = _mm256_add_ps(felec,fvdw);
1584 /* Calculate temporary vectorial force */
1585 tx = _mm256_mul_ps(fscal,dx00);
1586 ty = _mm256_mul_ps(fscal,dy00);
1587 tz = _mm256_mul_ps(fscal,dz00);
1589 /* Update vectorial force */
1590 fix0 = _mm256_add_ps(fix0,tx);
1591 fiy0 = _mm256_add_ps(fiy0,ty);
1592 fiz0 = _mm256_add_ps(fiz0,tz);
1594 fjx0 = _mm256_add_ps(fjx0,tx);
1595 fjy0 = _mm256_add_ps(fjy0,ty);
1596 fjz0 = _mm256_add_ps(fjz0,tz);
1598 /**************************
1599 * CALCULATE INTERACTIONS *
1600 **************************/
1602 r01 = _mm256_mul_ps(rsq01,rinv01);
1604 /* EWALD ELECTROSTATICS */
1606 /* Analytical PME correction */
1607 zeta2 = _mm256_mul_ps(beta2,rsq01);
1608 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1609 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1610 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1611 felec = _mm256_mul_ps(qq01,felec);
1615 /* Calculate temporary vectorial force */
1616 tx = _mm256_mul_ps(fscal,dx01);
1617 ty = _mm256_mul_ps(fscal,dy01);
1618 tz = _mm256_mul_ps(fscal,dz01);
1620 /* Update vectorial force */
1621 fix0 = _mm256_add_ps(fix0,tx);
1622 fiy0 = _mm256_add_ps(fiy0,ty);
1623 fiz0 = _mm256_add_ps(fiz0,tz);
1625 fjx1 = _mm256_add_ps(fjx1,tx);
1626 fjy1 = _mm256_add_ps(fjy1,ty);
1627 fjz1 = _mm256_add_ps(fjz1,tz);
1629 /**************************
1630 * CALCULATE INTERACTIONS *
1631 **************************/
1633 r02 = _mm256_mul_ps(rsq02,rinv02);
1635 /* EWALD ELECTROSTATICS */
1637 /* Analytical PME correction */
1638 zeta2 = _mm256_mul_ps(beta2,rsq02);
1639 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
1640 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1641 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1642 felec = _mm256_mul_ps(qq02,felec);
1646 /* Calculate temporary vectorial force */
1647 tx = _mm256_mul_ps(fscal,dx02);
1648 ty = _mm256_mul_ps(fscal,dy02);
1649 tz = _mm256_mul_ps(fscal,dz02);
1651 /* Update vectorial force */
1652 fix0 = _mm256_add_ps(fix0,tx);
1653 fiy0 = _mm256_add_ps(fiy0,ty);
1654 fiz0 = _mm256_add_ps(fiz0,tz);
1656 fjx2 = _mm256_add_ps(fjx2,tx);
1657 fjy2 = _mm256_add_ps(fjy2,ty);
1658 fjz2 = _mm256_add_ps(fjz2,tz);
1660 /**************************
1661 * CALCULATE INTERACTIONS *
1662 **************************/
1664 r10 = _mm256_mul_ps(rsq10,rinv10);
1666 /* EWALD ELECTROSTATICS */
1668 /* Analytical PME correction */
1669 zeta2 = _mm256_mul_ps(beta2,rsq10);
1670 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
1671 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1672 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1673 felec = _mm256_mul_ps(qq10,felec);
1677 /* Calculate temporary vectorial force */
1678 tx = _mm256_mul_ps(fscal,dx10);
1679 ty = _mm256_mul_ps(fscal,dy10);
1680 tz = _mm256_mul_ps(fscal,dz10);
1682 /* Update vectorial force */
1683 fix1 = _mm256_add_ps(fix1,tx);
1684 fiy1 = _mm256_add_ps(fiy1,ty);
1685 fiz1 = _mm256_add_ps(fiz1,tz);
1687 fjx0 = _mm256_add_ps(fjx0,tx);
1688 fjy0 = _mm256_add_ps(fjy0,ty);
1689 fjz0 = _mm256_add_ps(fjz0,tz);
1691 /**************************
1692 * CALCULATE INTERACTIONS *
1693 **************************/
1695 r11 = _mm256_mul_ps(rsq11,rinv11);
1697 /* EWALD ELECTROSTATICS */
1699 /* Analytical PME correction */
1700 zeta2 = _mm256_mul_ps(beta2,rsq11);
1701 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
1702 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1703 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1704 felec = _mm256_mul_ps(qq11,felec);
1708 /* Calculate temporary vectorial force */
1709 tx = _mm256_mul_ps(fscal,dx11);
1710 ty = _mm256_mul_ps(fscal,dy11);
1711 tz = _mm256_mul_ps(fscal,dz11);
1713 /* Update vectorial force */
1714 fix1 = _mm256_add_ps(fix1,tx);
1715 fiy1 = _mm256_add_ps(fiy1,ty);
1716 fiz1 = _mm256_add_ps(fiz1,tz);
1718 fjx1 = _mm256_add_ps(fjx1,tx);
1719 fjy1 = _mm256_add_ps(fjy1,ty);
1720 fjz1 = _mm256_add_ps(fjz1,tz);
1722 /**************************
1723 * CALCULATE INTERACTIONS *
1724 **************************/
1726 r12 = _mm256_mul_ps(rsq12,rinv12);
1728 /* EWALD ELECTROSTATICS */
1730 /* Analytical PME correction */
1731 zeta2 = _mm256_mul_ps(beta2,rsq12);
1732 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
1733 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1734 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1735 felec = _mm256_mul_ps(qq12,felec);
1739 /* Calculate temporary vectorial force */
1740 tx = _mm256_mul_ps(fscal,dx12);
1741 ty = _mm256_mul_ps(fscal,dy12);
1742 tz = _mm256_mul_ps(fscal,dz12);
1744 /* Update vectorial force */
1745 fix1 = _mm256_add_ps(fix1,tx);
1746 fiy1 = _mm256_add_ps(fiy1,ty);
1747 fiz1 = _mm256_add_ps(fiz1,tz);
1749 fjx2 = _mm256_add_ps(fjx2,tx);
1750 fjy2 = _mm256_add_ps(fjy2,ty);
1751 fjz2 = _mm256_add_ps(fjz2,tz);
1753 /**************************
1754 * CALCULATE INTERACTIONS *
1755 **************************/
1757 r20 = _mm256_mul_ps(rsq20,rinv20);
1759 /* EWALD ELECTROSTATICS */
1761 /* Analytical PME correction */
1762 zeta2 = _mm256_mul_ps(beta2,rsq20);
1763 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1764 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1765 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1766 felec = _mm256_mul_ps(qq20,felec);
1770 /* Calculate temporary vectorial force */
1771 tx = _mm256_mul_ps(fscal,dx20);
1772 ty = _mm256_mul_ps(fscal,dy20);
1773 tz = _mm256_mul_ps(fscal,dz20);
1775 /* Update vectorial force */
1776 fix2 = _mm256_add_ps(fix2,tx);
1777 fiy2 = _mm256_add_ps(fiy2,ty);
1778 fiz2 = _mm256_add_ps(fiz2,tz);
1780 fjx0 = _mm256_add_ps(fjx0,tx);
1781 fjy0 = _mm256_add_ps(fjy0,ty);
1782 fjz0 = _mm256_add_ps(fjz0,tz);
1784 /**************************
1785 * CALCULATE INTERACTIONS *
1786 **************************/
1788 r21 = _mm256_mul_ps(rsq21,rinv21);
1790 /* EWALD ELECTROSTATICS */
1792 /* Analytical PME correction */
1793 zeta2 = _mm256_mul_ps(beta2,rsq21);
1794 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
1795 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1796 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1797 felec = _mm256_mul_ps(qq21,felec);
1801 /* Calculate temporary vectorial force */
1802 tx = _mm256_mul_ps(fscal,dx21);
1803 ty = _mm256_mul_ps(fscal,dy21);
1804 tz = _mm256_mul_ps(fscal,dz21);
1806 /* Update vectorial force */
1807 fix2 = _mm256_add_ps(fix2,tx);
1808 fiy2 = _mm256_add_ps(fiy2,ty);
1809 fiz2 = _mm256_add_ps(fiz2,tz);
1811 fjx1 = _mm256_add_ps(fjx1,tx);
1812 fjy1 = _mm256_add_ps(fjy1,ty);
1813 fjz1 = _mm256_add_ps(fjz1,tz);
1815 /**************************
1816 * CALCULATE INTERACTIONS *
1817 **************************/
1819 r22 = _mm256_mul_ps(rsq22,rinv22);
1821 /* EWALD ELECTROSTATICS */
1823 /* Analytical PME correction */
1824 zeta2 = _mm256_mul_ps(beta2,rsq22);
1825 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
1826 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1827 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1828 felec = _mm256_mul_ps(qq22,felec);
1832 /* Calculate temporary vectorial force */
1833 tx = _mm256_mul_ps(fscal,dx22);
1834 ty = _mm256_mul_ps(fscal,dy22);
1835 tz = _mm256_mul_ps(fscal,dz22);
1837 /* Update vectorial force */
1838 fix2 = _mm256_add_ps(fix2,tx);
1839 fiy2 = _mm256_add_ps(fiy2,ty);
1840 fiz2 = _mm256_add_ps(fiz2,tz);
1842 fjx2 = _mm256_add_ps(fjx2,tx);
1843 fjy2 = _mm256_add_ps(fjy2,ty);
1844 fjz2 = _mm256_add_ps(fjz2,tz);
1846 fjptrA = f+j_coord_offsetA;
1847 fjptrB = f+j_coord_offsetB;
1848 fjptrC = f+j_coord_offsetC;
1849 fjptrD = f+j_coord_offsetD;
1850 fjptrE = f+j_coord_offsetE;
1851 fjptrF = f+j_coord_offsetF;
1852 fjptrG = f+j_coord_offsetG;
1853 fjptrH = f+j_coord_offsetH;
1855 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1856 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1858 /* Inner loop uses 527 flops */
1861 if(jidx<j_index_end)
1864 /* Get j neighbor index, and coordinate index */
1865 jnrlistA = jjnr[jidx];
1866 jnrlistB = jjnr[jidx+1];
1867 jnrlistC = jjnr[jidx+2];
1868 jnrlistD = jjnr[jidx+3];
1869 jnrlistE = jjnr[jidx+4];
1870 jnrlistF = jjnr[jidx+5];
1871 jnrlistG = jjnr[jidx+6];
1872 jnrlistH = jjnr[jidx+7];
1873 /* Sign of each element will be negative for non-real atoms.
1874 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1875 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1877 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1878 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1880 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1881 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1882 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1883 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1884 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
1885 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
1886 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
1887 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
1888 j_coord_offsetA = DIM*jnrA;
1889 j_coord_offsetB = DIM*jnrB;
1890 j_coord_offsetC = DIM*jnrC;
1891 j_coord_offsetD = DIM*jnrD;
1892 j_coord_offsetE = DIM*jnrE;
1893 j_coord_offsetF = DIM*jnrF;
1894 j_coord_offsetG = DIM*jnrG;
1895 j_coord_offsetH = DIM*jnrH;
1897 /* load j atom coordinates */
1898 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1899 x+j_coord_offsetC,x+j_coord_offsetD,
1900 x+j_coord_offsetE,x+j_coord_offsetF,
1901 x+j_coord_offsetG,x+j_coord_offsetH,
1902 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1904 /* Calculate displacement vector */
1905 dx00 = _mm256_sub_ps(ix0,jx0);
1906 dy00 = _mm256_sub_ps(iy0,jy0);
1907 dz00 = _mm256_sub_ps(iz0,jz0);
1908 dx01 = _mm256_sub_ps(ix0,jx1);
1909 dy01 = _mm256_sub_ps(iy0,jy1);
1910 dz01 = _mm256_sub_ps(iz0,jz1);
1911 dx02 = _mm256_sub_ps(ix0,jx2);
1912 dy02 = _mm256_sub_ps(iy0,jy2);
1913 dz02 = _mm256_sub_ps(iz0,jz2);
1914 dx10 = _mm256_sub_ps(ix1,jx0);
1915 dy10 = _mm256_sub_ps(iy1,jy0);
1916 dz10 = _mm256_sub_ps(iz1,jz0);
1917 dx11 = _mm256_sub_ps(ix1,jx1);
1918 dy11 = _mm256_sub_ps(iy1,jy1);
1919 dz11 = _mm256_sub_ps(iz1,jz1);
1920 dx12 = _mm256_sub_ps(ix1,jx2);
1921 dy12 = _mm256_sub_ps(iy1,jy2);
1922 dz12 = _mm256_sub_ps(iz1,jz2);
1923 dx20 = _mm256_sub_ps(ix2,jx0);
1924 dy20 = _mm256_sub_ps(iy2,jy0);
1925 dz20 = _mm256_sub_ps(iz2,jz0);
1926 dx21 = _mm256_sub_ps(ix2,jx1);
1927 dy21 = _mm256_sub_ps(iy2,jy1);
1928 dz21 = _mm256_sub_ps(iz2,jz1);
1929 dx22 = _mm256_sub_ps(ix2,jx2);
1930 dy22 = _mm256_sub_ps(iy2,jy2);
1931 dz22 = _mm256_sub_ps(iz2,jz2);
1933 /* Calculate squared distance and things based on it */
1934 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1935 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1936 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1937 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1938 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1939 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1940 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1941 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1942 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1944 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1945 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1946 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1947 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1948 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1949 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1950 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1951 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1952 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1954 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1955 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1956 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1957 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1958 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1959 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1960 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1961 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1962 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1964 fjx0 = _mm256_setzero_ps();
1965 fjy0 = _mm256_setzero_ps();
1966 fjz0 = _mm256_setzero_ps();
1967 fjx1 = _mm256_setzero_ps();
1968 fjy1 = _mm256_setzero_ps();
1969 fjz1 = _mm256_setzero_ps();
1970 fjx2 = _mm256_setzero_ps();
1971 fjy2 = _mm256_setzero_ps();
1972 fjz2 = _mm256_setzero_ps();
1974 /**************************
1975 * CALCULATE INTERACTIONS *
1976 **************************/
1978 r00 = _mm256_mul_ps(rsq00,rinv00);
1979 r00 = _mm256_andnot_ps(dummy_mask,r00);
1981 /* EWALD ELECTROSTATICS */
1983 /* Analytical PME correction */
1984 zeta2 = _mm256_mul_ps(beta2,rsq00);
1985 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1986 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1987 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1988 felec = _mm256_mul_ps(qq00,felec);
1990 /* Analytical LJ-PME */
1991 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1992 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
1993 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
1994 exponent = gmx_simd_exp_r(ewcljrsq);
1995 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1996 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
1997 /* f6A = 6 * C6grid * (1 - poly) */
1998 f6A = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
1999 /* f6B = C6grid * exponent * beta^6 */
2000 f6B = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
2001 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2002 fvdw = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),_mm256_sub_ps(c6_00,f6A)),rinvsix),f6B),rinvsq00);
2004 fscal = _mm256_add_ps(felec,fvdw);
2006 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2008 /* Calculate temporary vectorial force */
2009 tx = _mm256_mul_ps(fscal,dx00);
2010 ty = _mm256_mul_ps(fscal,dy00);
2011 tz = _mm256_mul_ps(fscal,dz00);
2013 /* Update vectorial force */
2014 fix0 = _mm256_add_ps(fix0,tx);
2015 fiy0 = _mm256_add_ps(fiy0,ty);
2016 fiz0 = _mm256_add_ps(fiz0,tz);
2018 fjx0 = _mm256_add_ps(fjx0,tx);
2019 fjy0 = _mm256_add_ps(fjy0,ty);
2020 fjz0 = _mm256_add_ps(fjz0,tz);
2022 /**************************
2023 * CALCULATE INTERACTIONS *
2024 **************************/
2026 r01 = _mm256_mul_ps(rsq01,rinv01);
2027 r01 = _mm256_andnot_ps(dummy_mask,r01);
2029 /* EWALD ELECTROSTATICS */
2031 /* Analytical PME correction */
2032 zeta2 = _mm256_mul_ps(beta2,rsq01);
2033 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
2034 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2035 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2036 felec = _mm256_mul_ps(qq01,felec);
2040 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2042 /* Calculate temporary vectorial force */
2043 tx = _mm256_mul_ps(fscal,dx01);
2044 ty = _mm256_mul_ps(fscal,dy01);
2045 tz = _mm256_mul_ps(fscal,dz01);
2047 /* Update vectorial force */
2048 fix0 = _mm256_add_ps(fix0,tx);
2049 fiy0 = _mm256_add_ps(fiy0,ty);
2050 fiz0 = _mm256_add_ps(fiz0,tz);
2052 fjx1 = _mm256_add_ps(fjx1,tx);
2053 fjy1 = _mm256_add_ps(fjy1,ty);
2054 fjz1 = _mm256_add_ps(fjz1,tz);
2056 /**************************
2057 * CALCULATE INTERACTIONS *
2058 **************************/
2060 r02 = _mm256_mul_ps(rsq02,rinv02);
2061 r02 = _mm256_andnot_ps(dummy_mask,r02);
2063 /* EWALD ELECTROSTATICS */
2065 /* Analytical PME correction */
2066 zeta2 = _mm256_mul_ps(beta2,rsq02);
2067 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
2068 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2069 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2070 felec = _mm256_mul_ps(qq02,felec);
2074 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2076 /* Calculate temporary vectorial force */
2077 tx = _mm256_mul_ps(fscal,dx02);
2078 ty = _mm256_mul_ps(fscal,dy02);
2079 tz = _mm256_mul_ps(fscal,dz02);
2081 /* Update vectorial force */
2082 fix0 = _mm256_add_ps(fix0,tx);
2083 fiy0 = _mm256_add_ps(fiy0,ty);
2084 fiz0 = _mm256_add_ps(fiz0,tz);
2086 fjx2 = _mm256_add_ps(fjx2,tx);
2087 fjy2 = _mm256_add_ps(fjy2,ty);
2088 fjz2 = _mm256_add_ps(fjz2,tz);
2090 /**************************
2091 * CALCULATE INTERACTIONS *
2092 **************************/
2094 r10 = _mm256_mul_ps(rsq10,rinv10);
2095 r10 = _mm256_andnot_ps(dummy_mask,r10);
2097 /* EWALD ELECTROSTATICS */
2099 /* Analytical PME correction */
2100 zeta2 = _mm256_mul_ps(beta2,rsq10);
2101 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2102 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2103 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2104 felec = _mm256_mul_ps(qq10,felec);
2108 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2110 /* Calculate temporary vectorial force */
2111 tx = _mm256_mul_ps(fscal,dx10);
2112 ty = _mm256_mul_ps(fscal,dy10);
2113 tz = _mm256_mul_ps(fscal,dz10);
2115 /* Update vectorial force */
2116 fix1 = _mm256_add_ps(fix1,tx);
2117 fiy1 = _mm256_add_ps(fiy1,ty);
2118 fiz1 = _mm256_add_ps(fiz1,tz);
2120 fjx0 = _mm256_add_ps(fjx0,tx);
2121 fjy0 = _mm256_add_ps(fjy0,ty);
2122 fjz0 = _mm256_add_ps(fjz0,tz);
2124 /**************************
2125 * CALCULATE INTERACTIONS *
2126 **************************/
2128 r11 = _mm256_mul_ps(rsq11,rinv11);
2129 r11 = _mm256_andnot_ps(dummy_mask,r11);
2131 /* EWALD ELECTROSTATICS */
2133 /* Analytical PME correction */
2134 zeta2 = _mm256_mul_ps(beta2,rsq11);
2135 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2136 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2137 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2138 felec = _mm256_mul_ps(qq11,felec);
2142 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2144 /* Calculate temporary vectorial force */
2145 tx = _mm256_mul_ps(fscal,dx11);
2146 ty = _mm256_mul_ps(fscal,dy11);
2147 tz = _mm256_mul_ps(fscal,dz11);
2149 /* Update vectorial force */
2150 fix1 = _mm256_add_ps(fix1,tx);
2151 fiy1 = _mm256_add_ps(fiy1,ty);
2152 fiz1 = _mm256_add_ps(fiz1,tz);
2154 fjx1 = _mm256_add_ps(fjx1,tx);
2155 fjy1 = _mm256_add_ps(fjy1,ty);
2156 fjz1 = _mm256_add_ps(fjz1,tz);
2158 /**************************
2159 * CALCULATE INTERACTIONS *
2160 **************************/
2162 r12 = _mm256_mul_ps(rsq12,rinv12);
2163 r12 = _mm256_andnot_ps(dummy_mask,r12);
2165 /* EWALD ELECTROSTATICS */
2167 /* Analytical PME correction */
2168 zeta2 = _mm256_mul_ps(beta2,rsq12);
2169 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2170 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2171 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2172 felec = _mm256_mul_ps(qq12,felec);
2176 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2178 /* Calculate temporary vectorial force */
2179 tx = _mm256_mul_ps(fscal,dx12);
2180 ty = _mm256_mul_ps(fscal,dy12);
2181 tz = _mm256_mul_ps(fscal,dz12);
2183 /* Update vectorial force */
2184 fix1 = _mm256_add_ps(fix1,tx);
2185 fiy1 = _mm256_add_ps(fiy1,ty);
2186 fiz1 = _mm256_add_ps(fiz1,tz);
2188 fjx2 = _mm256_add_ps(fjx2,tx);
2189 fjy2 = _mm256_add_ps(fjy2,ty);
2190 fjz2 = _mm256_add_ps(fjz2,tz);
2192 /**************************
2193 * CALCULATE INTERACTIONS *
2194 **************************/
2196 r20 = _mm256_mul_ps(rsq20,rinv20);
2197 r20 = _mm256_andnot_ps(dummy_mask,r20);
2199 /* EWALD ELECTROSTATICS */
2201 /* Analytical PME correction */
2202 zeta2 = _mm256_mul_ps(beta2,rsq20);
2203 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
2204 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2205 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2206 felec = _mm256_mul_ps(qq20,felec);
2210 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2212 /* Calculate temporary vectorial force */
2213 tx = _mm256_mul_ps(fscal,dx20);
2214 ty = _mm256_mul_ps(fscal,dy20);
2215 tz = _mm256_mul_ps(fscal,dz20);
2217 /* Update vectorial force */
2218 fix2 = _mm256_add_ps(fix2,tx);
2219 fiy2 = _mm256_add_ps(fiy2,ty);
2220 fiz2 = _mm256_add_ps(fiz2,tz);
2222 fjx0 = _mm256_add_ps(fjx0,tx);
2223 fjy0 = _mm256_add_ps(fjy0,ty);
2224 fjz0 = _mm256_add_ps(fjz0,tz);
2226 /**************************
2227 * CALCULATE INTERACTIONS *
2228 **************************/
2230 r21 = _mm256_mul_ps(rsq21,rinv21);
2231 r21 = _mm256_andnot_ps(dummy_mask,r21);
2233 /* EWALD ELECTROSTATICS */
2235 /* Analytical PME correction */
2236 zeta2 = _mm256_mul_ps(beta2,rsq21);
2237 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2238 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2239 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2240 felec = _mm256_mul_ps(qq21,felec);
2244 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2246 /* Calculate temporary vectorial force */
2247 tx = _mm256_mul_ps(fscal,dx21);
2248 ty = _mm256_mul_ps(fscal,dy21);
2249 tz = _mm256_mul_ps(fscal,dz21);
2251 /* Update vectorial force */
2252 fix2 = _mm256_add_ps(fix2,tx);
2253 fiy2 = _mm256_add_ps(fiy2,ty);
2254 fiz2 = _mm256_add_ps(fiz2,tz);
2256 fjx1 = _mm256_add_ps(fjx1,tx);
2257 fjy1 = _mm256_add_ps(fjy1,ty);
2258 fjz1 = _mm256_add_ps(fjz1,tz);
2260 /**************************
2261 * CALCULATE INTERACTIONS *
2262 **************************/
2264 r22 = _mm256_mul_ps(rsq22,rinv22);
2265 r22 = _mm256_andnot_ps(dummy_mask,r22);
2267 /* EWALD ELECTROSTATICS */
2269 /* Analytical PME correction */
2270 zeta2 = _mm256_mul_ps(beta2,rsq22);
2271 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
2272 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2273 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2274 felec = _mm256_mul_ps(qq22,felec);
2278 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2280 /* Calculate temporary vectorial force */
2281 tx = _mm256_mul_ps(fscal,dx22);
2282 ty = _mm256_mul_ps(fscal,dy22);
2283 tz = _mm256_mul_ps(fscal,dz22);
2285 /* Update vectorial force */
2286 fix2 = _mm256_add_ps(fix2,tx);
2287 fiy2 = _mm256_add_ps(fiy2,ty);
2288 fiz2 = _mm256_add_ps(fiz2,tz);
2290 fjx2 = _mm256_add_ps(fjx2,tx);
2291 fjy2 = _mm256_add_ps(fjy2,ty);
2292 fjz2 = _mm256_add_ps(fjz2,tz);
2294 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2295 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2296 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2297 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2298 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2299 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2300 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2301 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2303 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2304 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2306 /* Inner loop uses 536 flops */
2309 /* End of innermost loop */
2311 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2312 f+i_coord_offset,fshift+i_shift_offset);
2314 /* Increment number of inner iterations */
2315 inneriter += j_index_end - j_index_start;
2317 /* Outer loop uses 18 flops */
2320 /* Increment number of outer iterations */
2323 /* Update outer/inner flops */
2325 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*536);