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.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
46 #include "gromacs/math/vec.h"
49 #include "gromacs/simd/math_x86_avx_256_single.h"
50 #include "kernelutil_x86_avx_256_single.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3W3_VF_avx_256_single
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LJEwald
56 * Geometry: Water3-Water3
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEw_VdwLJEw_GeomW3W3_VF_avx_256_single
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int jnrA,jnrB,jnrC,jnrD;
77 int jnrE,jnrF,jnrG,jnrH;
78 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
79 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
80 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
81 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
82 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
84 real *shiftvec,*fshift,*x,*f;
85 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
87 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
88 real * vdwioffsetptr0;
89 real * vdwgridioffsetptr0;
90 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
91 real * vdwioffsetptr1;
92 real * vdwgridioffsetptr1;
93 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
94 real * vdwioffsetptr2;
95 real * vdwgridioffsetptr2;
96 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
97 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
98 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
99 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
100 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
101 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
102 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
103 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
104 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
105 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
106 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
107 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
108 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
109 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
110 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
111 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
112 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
115 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
118 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
119 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
130 __m256 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
131 __m256 one_half = _mm256_set1_ps(0.5);
132 __m256 minus_one = _mm256_set1_ps(-1.0);
134 __m128i ewitab_lo,ewitab_hi;
135 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
136 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
138 __m256 dummy_mask,cutoff_mask;
139 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
140 __m256 one = _mm256_set1_ps(1.0);
141 __m256 two = _mm256_set1_ps(2.0);
147 jindex = nlist->jindex;
149 shiftidx = nlist->shift;
151 shiftvec = fr->shift_vec[0];
152 fshift = fr->fshift[0];
153 facel = _mm256_set1_ps(fr->epsfac);
154 charge = mdatoms->chargeA;
155 nvdwtype = fr->ntype;
157 vdwtype = mdatoms->typeA;
158 vdwgridparam = fr->ljpme_c6grid;
159 sh_lj_ewald = _mm256_set1_ps(fr->ic->sh_lj_ewald);
160 ewclj = _mm256_set1_ps(fr->ewaldcoeff_lj);
161 ewclj2 = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
163 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
164 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
165 beta2 = _mm256_mul_ps(beta,beta);
166 beta3 = _mm256_mul_ps(beta,beta2);
168 ewtab = fr->ic->tabq_coul_FDV0;
169 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
170 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
172 /* Setup water-specific parameters */
173 inr = nlist->iinr[0];
174 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
175 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
176 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
177 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
178 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
180 jq0 = _mm256_set1_ps(charge[inr+0]);
181 jq1 = _mm256_set1_ps(charge[inr+1]);
182 jq2 = _mm256_set1_ps(charge[inr+2]);
183 vdwjidx0A = 2*vdwtype[inr+0];
184 qq00 = _mm256_mul_ps(iq0,jq0);
185 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
186 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
187 c6grid_00 = _mm256_set1_ps(vdwgridioffsetptr0[vdwjidx0A]);
188 qq01 = _mm256_mul_ps(iq0,jq1);
189 qq02 = _mm256_mul_ps(iq0,jq2);
190 qq10 = _mm256_mul_ps(iq1,jq0);
191 qq11 = _mm256_mul_ps(iq1,jq1);
192 qq12 = _mm256_mul_ps(iq1,jq2);
193 qq20 = _mm256_mul_ps(iq2,jq0);
194 qq21 = _mm256_mul_ps(iq2,jq1);
195 qq22 = _mm256_mul_ps(iq2,jq2);
197 /* Avoid stupid compiler warnings */
198 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
211 for(iidx=0;iidx<4*DIM;iidx++)
216 /* Start outer loop over neighborlists */
217 for(iidx=0; iidx<nri; iidx++)
219 /* Load shift vector for this list */
220 i_shift_offset = DIM*shiftidx[iidx];
222 /* Load limits for loop over neighbors */
223 j_index_start = jindex[iidx];
224 j_index_end = jindex[iidx+1];
226 /* Get outer coordinate index */
228 i_coord_offset = DIM*inr;
230 /* Load i particle coords and add shift vector */
231 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
232 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
234 fix0 = _mm256_setzero_ps();
235 fiy0 = _mm256_setzero_ps();
236 fiz0 = _mm256_setzero_ps();
237 fix1 = _mm256_setzero_ps();
238 fiy1 = _mm256_setzero_ps();
239 fiz1 = _mm256_setzero_ps();
240 fix2 = _mm256_setzero_ps();
241 fiy2 = _mm256_setzero_ps();
242 fiz2 = _mm256_setzero_ps();
244 /* Reset potential sums */
245 velecsum = _mm256_setzero_ps();
246 vvdwsum = _mm256_setzero_ps();
248 /* Start inner kernel loop */
249 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
252 /* Get j neighbor index, and coordinate index */
261 j_coord_offsetA = DIM*jnrA;
262 j_coord_offsetB = DIM*jnrB;
263 j_coord_offsetC = DIM*jnrC;
264 j_coord_offsetD = DIM*jnrD;
265 j_coord_offsetE = DIM*jnrE;
266 j_coord_offsetF = DIM*jnrF;
267 j_coord_offsetG = DIM*jnrG;
268 j_coord_offsetH = DIM*jnrH;
270 /* load j atom coordinates */
271 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
272 x+j_coord_offsetC,x+j_coord_offsetD,
273 x+j_coord_offsetE,x+j_coord_offsetF,
274 x+j_coord_offsetG,x+j_coord_offsetH,
275 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
277 /* Calculate displacement vector */
278 dx00 = _mm256_sub_ps(ix0,jx0);
279 dy00 = _mm256_sub_ps(iy0,jy0);
280 dz00 = _mm256_sub_ps(iz0,jz0);
281 dx01 = _mm256_sub_ps(ix0,jx1);
282 dy01 = _mm256_sub_ps(iy0,jy1);
283 dz01 = _mm256_sub_ps(iz0,jz1);
284 dx02 = _mm256_sub_ps(ix0,jx2);
285 dy02 = _mm256_sub_ps(iy0,jy2);
286 dz02 = _mm256_sub_ps(iz0,jz2);
287 dx10 = _mm256_sub_ps(ix1,jx0);
288 dy10 = _mm256_sub_ps(iy1,jy0);
289 dz10 = _mm256_sub_ps(iz1,jz0);
290 dx11 = _mm256_sub_ps(ix1,jx1);
291 dy11 = _mm256_sub_ps(iy1,jy1);
292 dz11 = _mm256_sub_ps(iz1,jz1);
293 dx12 = _mm256_sub_ps(ix1,jx2);
294 dy12 = _mm256_sub_ps(iy1,jy2);
295 dz12 = _mm256_sub_ps(iz1,jz2);
296 dx20 = _mm256_sub_ps(ix2,jx0);
297 dy20 = _mm256_sub_ps(iy2,jy0);
298 dz20 = _mm256_sub_ps(iz2,jz0);
299 dx21 = _mm256_sub_ps(ix2,jx1);
300 dy21 = _mm256_sub_ps(iy2,jy1);
301 dz21 = _mm256_sub_ps(iz2,jz1);
302 dx22 = _mm256_sub_ps(ix2,jx2);
303 dy22 = _mm256_sub_ps(iy2,jy2);
304 dz22 = _mm256_sub_ps(iz2,jz2);
306 /* Calculate squared distance and things based on it */
307 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
308 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
309 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
310 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
311 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
312 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
313 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
314 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
315 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
317 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
318 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
319 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
320 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
321 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
322 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
323 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
324 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
325 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
327 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
328 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
329 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
330 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
331 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
332 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
333 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
334 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
335 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
337 fjx0 = _mm256_setzero_ps();
338 fjy0 = _mm256_setzero_ps();
339 fjz0 = _mm256_setzero_ps();
340 fjx1 = _mm256_setzero_ps();
341 fjy1 = _mm256_setzero_ps();
342 fjz1 = _mm256_setzero_ps();
343 fjx2 = _mm256_setzero_ps();
344 fjy2 = _mm256_setzero_ps();
345 fjz2 = _mm256_setzero_ps();
347 /**************************
348 * CALCULATE INTERACTIONS *
349 **************************/
351 r00 = _mm256_mul_ps(rsq00,rinv00);
353 /* EWALD ELECTROSTATICS */
355 /* Analytical PME correction */
356 zeta2 = _mm256_mul_ps(beta2,rsq00);
357 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
358 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
359 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
360 felec = _mm256_mul_ps(qq00,felec);
361 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
362 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
363 velec = _mm256_sub_ps(rinv00,pmecorrV);
364 velec = _mm256_mul_ps(qq00,velec);
366 /* Analytical LJ-PME */
367 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
368 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
369 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
370 exponent = gmx_simd_exp_r(ewcljrsq);
371 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
372 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
373 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
374 vvdw6 = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
375 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
376 vvdw = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
377 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
378 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);
380 /* Update potential sum for this i atom from the interaction with this j atom. */
381 velecsum = _mm256_add_ps(velecsum,velec);
382 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
384 fscal = _mm256_add_ps(felec,fvdw);
386 /* Calculate temporary vectorial force */
387 tx = _mm256_mul_ps(fscal,dx00);
388 ty = _mm256_mul_ps(fscal,dy00);
389 tz = _mm256_mul_ps(fscal,dz00);
391 /* Update vectorial force */
392 fix0 = _mm256_add_ps(fix0,tx);
393 fiy0 = _mm256_add_ps(fiy0,ty);
394 fiz0 = _mm256_add_ps(fiz0,tz);
396 fjx0 = _mm256_add_ps(fjx0,tx);
397 fjy0 = _mm256_add_ps(fjy0,ty);
398 fjz0 = _mm256_add_ps(fjz0,tz);
400 /**************************
401 * CALCULATE INTERACTIONS *
402 **************************/
404 r01 = _mm256_mul_ps(rsq01,rinv01);
406 /* EWALD ELECTROSTATICS */
408 /* Analytical PME correction */
409 zeta2 = _mm256_mul_ps(beta2,rsq01);
410 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
411 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
412 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
413 felec = _mm256_mul_ps(qq01,felec);
414 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
415 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
416 velec = _mm256_sub_ps(rinv01,pmecorrV);
417 velec = _mm256_mul_ps(qq01,velec);
419 /* Update potential sum for this i atom from the interaction with this j atom. */
420 velecsum = _mm256_add_ps(velecsum,velec);
424 /* Calculate temporary vectorial force */
425 tx = _mm256_mul_ps(fscal,dx01);
426 ty = _mm256_mul_ps(fscal,dy01);
427 tz = _mm256_mul_ps(fscal,dz01);
429 /* Update vectorial force */
430 fix0 = _mm256_add_ps(fix0,tx);
431 fiy0 = _mm256_add_ps(fiy0,ty);
432 fiz0 = _mm256_add_ps(fiz0,tz);
434 fjx1 = _mm256_add_ps(fjx1,tx);
435 fjy1 = _mm256_add_ps(fjy1,ty);
436 fjz1 = _mm256_add_ps(fjz1,tz);
438 /**************************
439 * CALCULATE INTERACTIONS *
440 **************************/
442 r02 = _mm256_mul_ps(rsq02,rinv02);
444 /* EWALD ELECTROSTATICS */
446 /* Analytical PME correction */
447 zeta2 = _mm256_mul_ps(beta2,rsq02);
448 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
449 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
450 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
451 felec = _mm256_mul_ps(qq02,felec);
452 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
453 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
454 velec = _mm256_sub_ps(rinv02,pmecorrV);
455 velec = _mm256_mul_ps(qq02,velec);
457 /* Update potential sum for this i atom from the interaction with this j atom. */
458 velecsum = _mm256_add_ps(velecsum,velec);
462 /* Calculate temporary vectorial force */
463 tx = _mm256_mul_ps(fscal,dx02);
464 ty = _mm256_mul_ps(fscal,dy02);
465 tz = _mm256_mul_ps(fscal,dz02);
467 /* Update vectorial force */
468 fix0 = _mm256_add_ps(fix0,tx);
469 fiy0 = _mm256_add_ps(fiy0,ty);
470 fiz0 = _mm256_add_ps(fiz0,tz);
472 fjx2 = _mm256_add_ps(fjx2,tx);
473 fjy2 = _mm256_add_ps(fjy2,ty);
474 fjz2 = _mm256_add_ps(fjz2,tz);
476 /**************************
477 * CALCULATE INTERACTIONS *
478 **************************/
480 r10 = _mm256_mul_ps(rsq10,rinv10);
482 /* EWALD ELECTROSTATICS */
484 /* Analytical PME correction */
485 zeta2 = _mm256_mul_ps(beta2,rsq10);
486 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
487 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
488 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
489 felec = _mm256_mul_ps(qq10,felec);
490 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
491 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
492 velec = _mm256_sub_ps(rinv10,pmecorrV);
493 velec = _mm256_mul_ps(qq10,velec);
495 /* Update potential sum for this i atom from the interaction with this j atom. */
496 velecsum = _mm256_add_ps(velecsum,velec);
500 /* Calculate temporary vectorial force */
501 tx = _mm256_mul_ps(fscal,dx10);
502 ty = _mm256_mul_ps(fscal,dy10);
503 tz = _mm256_mul_ps(fscal,dz10);
505 /* Update vectorial force */
506 fix1 = _mm256_add_ps(fix1,tx);
507 fiy1 = _mm256_add_ps(fiy1,ty);
508 fiz1 = _mm256_add_ps(fiz1,tz);
510 fjx0 = _mm256_add_ps(fjx0,tx);
511 fjy0 = _mm256_add_ps(fjy0,ty);
512 fjz0 = _mm256_add_ps(fjz0,tz);
514 /**************************
515 * CALCULATE INTERACTIONS *
516 **************************/
518 r11 = _mm256_mul_ps(rsq11,rinv11);
520 /* EWALD ELECTROSTATICS */
522 /* Analytical PME correction */
523 zeta2 = _mm256_mul_ps(beta2,rsq11);
524 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
525 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
526 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
527 felec = _mm256_mul_ps(qq11,felec);
528 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
529 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
530 velec = _mm256_sub_ps(rinv11,pmecorrV);
531 velec = _mm256_mul_ps(qq11,velec);
533 /* Update potential sum for this i atom from the interaction with this j atom. */
534 velecsum = _mm256_add_ps(velecsum,velec);
538 /* Calculate temporary vectorial force */
539 tx = _mm256_mul_ps(fscal,dx11);
540 ty = _mm256_mul_ps(fscal,dy11);
541 tz = _mm256_mul_ps(fscal,dz11);
543 /* Update vectorial force */
544 fix1 = _mm256_add_ps(fix1,tx);
545 fiy1 = _mm256_add_ps(fiy1,ty);
546 fiz1 = _mm256_add_ps(fiz1,tz);
548 fjx1 = _mm256_add_ps(fjx1,tx);
549 fjy1 = _mm256_add_ps(fjy1,ty);
550 fjz1 = _mm256_add_ps(fjz1,tz);
552 /**************************
553 * CALCULATE INTERACTIONS *
554 **************************/
556 r12 = _mm256_mul_ps(rsq12,rinv12);
558 /* EWALD ELECTROSTATICS */
560 /* Analytical PME correction */
561 zeta2 = _mm256_mul_ps(beta2,rsq12);
562 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
563 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
564 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
565 felec = _mm256_mul_ps(qq12,felec);
566 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
567 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
568 velec = _mm256_sub_ps(rinv12,pmecorrV);
569 velec = _mm256_mul_ps(qq12,velec);
571 /* Update potential sum for this i atom from the interaction with this j atom. */
572 velecsum = _mm256_add_ps(velecsum,velec);
576 /* Calculate temporary vectorial force */
577 tx = _mm256_mul_ps(fscal,dx12);
578 ty = _mm256_mul_ps(fscal,dy12);
579 tz = _mm256_mul_ps(fscal,dz12);
581 /* Update vectorial force */
582 fix1 = _mm256_add_ps(fix1,tx);
583 fiy1 = _mm256_add_ps(fiy1,ty);
584 fiz1 = _mm256_add_ps(fiz1,tz);
586 fjx2 = _mm256_add_ps(fjx2,tx);
587 fjy2 = _mm256_add_ps(fjy2,ty);
588 fjz2 = _mm256_add_ps(fjz2,tz);
590 /**************************
591 * CALCULATE INTERACTIONS *
592 **************************/
594 r20 = _mm256_mul_ps(rsq20,rinv20);
596 /* EWALD ELECTROSTATICS */
598 /* Analytical PME correction */
599 zeta2 = _mm256_mul_ps(beta2,rsq20);
600 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
601 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
602 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
603 felec = _mm256_mul_ps(qq20,felec);
604 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
605 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
606 velec = _mm256_sub_ps(rinv20,pmecorrV);
607 velec = _mm256_mul_ps(qq20,velec);
609 /* Update potential sum for this i atom from the interaction with this j atom. */
610 velecsum = _mm256_add_ps(velecsum,velec);
614 /* Calculate temporary vectorial force */
615 tx = _mm256_mul_ps(fscal,dx20);
616 ty = _mm256_mul_ps(fscal,dy20);
617 tz = _mm256_mul_ps(fscal,dz20);
619 /* Update vectorial force */
620 fix2 = _mm256_add_ps(fix2,tx);
621 fiy2 = _mm256_add_ps(fiy2,ty);
622 fiz2 = _mm256_add_ps(fiz2,tz);
624 fjx0 = _mm256_add_ps(fjx0,tx);
625 fjy0 = _mm256_add_ps(fjy0,ty);
626 fjz0 = _mm256_add_ps(fjz0,tz);
628 /**************************
629 * CALCULATE INTERACTIONS *
630 **************************/
632 r21 = _mm256_mul_ps(rsq21,rinv21);
634 /* EWALD ELECTROSTATICS */
636 /* Analytical PME correction */
637 zeta2 = _mm256_mul_ps(beta2,rsq21);
638 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
639 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
640 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
641 felec = _mm256_mul_ps(qq21,felec);
642 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
643 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
644 velec = _mm256_sub_ps(rinv21,pmecorrV);
645 velec = _mm256_mul_ps(qq21,velec);
647 /* Update potential sum for this i atom from the interaction with this j atom. */
648 velecsum = _mm256_add_ps(velecsum,velec);
652 /* Calculate temporary vectorial force */
653 tx = _mm256_mul_ps(fscal,dx21);
654 ty = _mm256_mul_ps(fscal,dy21);
655 tz = _mm256_mul_ps(fscal,dz21);
657 /* Update vectorial force */
658 fix2 = _mm256_add_ps(fix2,tx);
659 fiy2 = _mm256_add_ps(fiy2,ty);
660 fiz2 = _mm256_add_ps(fiz2,tz);
662 fjx1 = _mm256_add_ps(fjx1,tx);
663 fjy1 = _mm256_add_ps(fjy1,ty);
664 fjz1 = _mm256_add_ps(fjz1,tz);
666 /**************************
667 * CALCULATE INTERACTIONS *
668 **************************/
670 r22 = _mm256_mul_ps(rsq22,rinv22);
672 /* EWALD ELECTROSTATICS */
674 /* Analytical PME correction */
675 zeta2 = _mm256_mul_ps(beta2,rsq22);
676 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
677 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
678 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
679 felec = _mm256_mul_ps(qq22,felec);
680 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
681 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
682 velec = _mm256_sub_ps(rinv22,pmecorrV);
683 velec = _mm256_mul_ps(qq22,velec);
685 /* Update potential sum for this i atom from the interaction with this j atom. */
686 velecsum = _mm256_add_ps(velecsum,velec);
690 /* Calculate temporary vectorial force */
691 tx = _mm256_mul_ps(fscal,dx22);
692 ty = _mm256_mul_ps(fscal,dy22);
693 tz = _mm256_mul_ps(fscal,dz22);
695 /* Update vectorial force */
696 fix2 = _mm256_add_ps(fix2,tx);
697 fiy2 = _mm256_add_ps(fiy2,ty);
698 fiz2 = _mm256_add_ps(fiz2,tz);
700 fjx2 = _mm256_add_ps(fjx2,tx);
701 fjy2 = _mm256_add_ps(fjy2,ty);
702 fjz2 = _mm256_add_ps(fjz2,tz);
704 fjptrA = f+j_coord_offsetA;
705 fjptrB = f+j_coord_offsetB;
706 fjptrC = f+j_coord_offsetC;
707 fjptrD = f+j_coord_offsetD;
708 fjptrE = f+j_coord_offsetE;
709 fjptrF = f+j_coord_offsetF;
710 fjptrG = f+j_coord_offsetG;
711 fjptrH = f+j_coord_offsetH;
713 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
714 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
716 /* Inner loop uses 784 flops */
722 /* Get j neighbor index, and coordinate index */
723 jnrlistA = jjnr[jidx];
724 jnrlistB = jjnr[jidx+1];
725 jnrlistC = jjnr[jidx+2];
726 jnrlistD = jjnr[jidx+3];
727 jnrlistE = jjnr[jidx+4];
728 jnrlistF = jjnr[jidx+5];
729 jnrlistG = jjnr[jidx+6];
730 jnrlistH = jjnr[jidx+7];
731 /* Sign of each element will be negative for non-real atoms.
732 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
733 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
735 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
736 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
738 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
739 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
740 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
741 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
742 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
743 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
744 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
745 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
746 j_coord_offsetA = DIM*jnrA;
747 j_coord_offsetB = DIM*jnrB;
748 j_coord_offsetC = DIM*jnrC;
749 j_coord_offsetD = DIM*jnrD;
750 j_coord_offsetE = DIM*jnrE;
751 j_coord_offsetF = DIM*jnrF;
752 j_coord_offsetG = DIM*jnrG;
753 j_coord_offsetH = DIM*jnrH;
755 /* load j atom coordinates */
756 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
757 x+j_coord_offsetC,x+j_coord_offsetD,
758 x+j_coord_offsetE,x+j_coord_offsetF,
759 x+j_coord_offsetG,x+j_coord_offsetH,
760 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
762 /* Calculate displacement vector */
763 dx00 = _mm256_sub_ps(ix0,jx0);
764 dy00 = _mm256_sub_ps(iy0,jy0);
765 dz00 = _mm256_sub_ps(iz0,jz0);
766 dx01 = _mm256_sub_ps(ix0,jx1);
767 dy01 = _mm256_sub_ps(iy0,jy1);
768 dz01 = _mm256_sub_ps(iz0,jz1);
769 dx02 = _mm256_sub_ps(ix0,jx2);
770 dy02 = _mm256_sub_ps(iy0,jy2);
771 dz02 = _mm256_sub_ps(iz0,jz2);
772 dx10 = _mm256_sub_ps(ix1,jx0);
773 dy10 = _mm256_sub_ps(iy1,jy0);
774 dz10 = _mm256_sub_ps(iz1,jz0);
775 dx11 = _mm256_sub_ps(ix1,jx1);
776 dy11 = _mm256_sub_ps(iy1,jy1);
777 dz11 = _mm256_sub_ps(iz1,jz1);
778 dx12 = _mm256_sub_ps(ix1,jx2);
779 dy12 = _mm256_sub_ps(iy1,jy2);
780 dz12 = _mm256_sub_ps(iz1,jz2);
781 dx20 = _mm256_sub_ps(ix2,jx0);
782 dy20 = _mm256_sub_ps(iy2,jy0);
783 dz20 = _mm256_sub_ps(iz2,jz0);
784 dx21 = _mm256_sub_ps(ix2,jx1);
785 dy21 = _mm256_sub_ps(iy2,jy1);
786 dz21 = _mm256_sub_ps(iz2,jz1);
787 dx22 = _mm256_sub_ps(ix2,jx2);
788 dy22 = _mm256_sub_ps(iy2,jy2);
789 dz22 = _mm256_sub_ps(iz2,jz2);
791 /* Calculate squared distance and things based on it */
792 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
793 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
794 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
795 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
796 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
797 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
798 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
799 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
800 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
802 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
803 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
804 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
805 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
806 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
807 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
808 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
809 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
810 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
812 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
813 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
814 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
815 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
816 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
817 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
818 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
819 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
820 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
822 fjx0 = _mm256_setzero_ps();
823 fjy0 = _mm256_setzero_ps();
824 fjz0 = _mm256_setzero_ps();
825 fjx1 = _mm256_setzero_ps();
826 fjy1 = _mm256_setzero_ps();
827 fjz1 = _mm256_setzero_ps();
828 fjx2 = _mm256_setzero_ps();
829 fjy2 = _mm256_setzero_ps();
830 fjz2 = _mm256_setzero_ps();
832 /**************************
833 * CALCULATE INTERACTIONS *
834 **************************/
836 r00 = _mm256_mul_ps(rsq00,rinv00);
837 r00 = _mm256_andnot_ps(dummy_mask,r00);
839 /* EWALD ELECTROSTATICS */
841 /* Analytical PME correction */
842 zeta2 = _mm256_mul_ps(beta2,rsq00);
843 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
844 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
845 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
846 felec = _mm256_mul_ps(qq00,felec);
847 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
848 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
849 velec = _mm256_sub_ps(rinv00,pmecorrV);
850 velec = _mm256_mul_ps(qq00,velec);
852 /* Analytical LJ-PME */
853 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
854 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
855 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
856 exponent = gmx_simd_exp_r(ewcljrsq);
857 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
858 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
859 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
860 vvdw6 = _mm256_mul_ps(_mm256_sub_ps(c6_00,_mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly))),rinvsix);
861 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
862 vvdw = _mm256_sub_ps(_mm256_mul_ps(vvdw12,one_twelfth),_mm256_mul_ps(vvdw6,one_sixth));
863 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
864 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);
866 /* Update potential sum for this i atom from the interaction with this j atom. */
867 velec = _mm256_andnot_ps(dummy_mask,velec);
868 velecsum = _mm256_add_ps(velecsum,velec);
869 vvdw = _mm256_andnot_ps(dummy_mask,vvdw);
870 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
872 fscal = _mm256_add_ps(felec,fvdw);
874 fscal = _mm256_andnot_ps(dummy_mask,fscal);
876 /* Calculate temporary vectorial force */
877 tx = _mm256_mul_ps(fscal,dx00);
878 ty = _mm256_mul_ps(fscal,dy00);
879 tz = _mm256_mul_ps(fscal,dz00);
881 /* Update vectorial force */
882 fix0 = _mm256_add_ps(fix0,tx);
883 fiy0 = _mm256_add_ps(fiy0,ty);
884 fiz0 = _mm256_add_ps(fiz0,tz);
886 fjx0 = _mm256_add_ps(fjx0,tx);
887 fjy0 = _mm256_add_ps(fjy0,ty);
888 fjz0 = _mm256_add_ps(fjz0,tz);
890 /**************************
891 * CALCULATE INTERACTIONS *
892 **************************/
894 r01 = _mm256_mul_ps(rsq01,rinv01);
895 r01 = _mm256_andnot_ps(dummy_mask,r01);
897 /* EWALD ELECTROSTATICS */
899 /* Analytical PME correction */
900 zeta2 = _mm256_mul_ps(beta2,rsq01);
901 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
902 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
903 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
904 felec = _mm256_mul_ps(qq01,felec);
905 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
906 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
907 velec = _mm256_sub_ps(rinv01,pmecorrV);
908 velec = _mm256_mul_ps(qq01,velec);
910 /* Update potential sum for this i atom from the interaction with this j atom. */
911 velec = _mm256_andnot_ps(dummy_mask,velec);
912 velecsum = _mm256_add_ps(velecsum,velec);
916 fscal = _mm256_andnot_ps(dummy_mask,fscal);
918 /* Calculate temporary vectorial force */
919 tx = _mm256_mul_ps(fscal,dx01);
920 ty = _mm256_mul_ps(fscal,dy01);
921 tz = _mm256_mul_ps(fscal,dz01);
923 /* Update vectorial force */
924 fix0 = _mm256_add_ps(fix0,tx);
925 fiy0 = _mm256_add_ps(fiy0,ty);
926 fiz0 = _mm256_add_ps(fiz0,tz);
928 fjx1 = _mm256_add_ps(fjx1,tx);
929 fjy1 = _mm256_add_ps(fjy1,ty);
930 fjz1 = _mm256_add_ps(fjz1,tz);
932 /**************************
933 * CALCULATE INTERACTIONS *
934 **************************/
936 r02 = _mm256_mul_ps(rsq02,rinv02);
937 r02 = _mm256_andnot_ps(dummy_mask,r02);
939 /* EWALD ELECTROSTATICS */
941 /* Analytical PME correction */
942 zeta2 = _mm256_mul_ps(beta2,rsq02);
943 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
944 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
945 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
946 felec = _mm256_mul_ps(qq02,felec);
947 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
948 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
949 velec = _mm256_sub_ps(rinv02,pmecorrV);
950 velec = _mm256_mul_ps(qq02,velec);
952 /* Update potential sum for this i atom from the interaction with this j atom. */
953 velec = _mm256_andnot_ps(dummy_mask,velec);
954 velecsum = _mm256_add_ps(velecsum,velec);
958 fscal = _mm256_andnot_ps(dummy_mask,fscal);
960 /* Calculate temporary vectorial force */
961 tx = _mm256_mul_ps(fscal,dx02);
962 ty = _mm256_mul_ps(fscal,dy02);
963 tz = _mm256_mul_ps(fscal,dz02);
965 /* Update vectorial force */
966 fix0 = _mm256_add_ps(fix0,tx);
967 fiy0 = _mm256_add_ps(fiy0,ty);
968 fiz0 = _mm256_add_ps(fiz0,tz);
970 fjx2 = _mm256_add_ps(fjx2,tx);
971 fjy2 = _mm256_add_ps(fjy2,ty);
972 fjz2 = _mm256_add_ps(fjz2,tz);
974 /**************************
975 * CALCULATE INTERACTIONS *
976 **************************/
978 r10 = _mm256_mul_ps(rsq10,rinv10);
979 r10 = _mm256_andnot_ps(dummy_mask,r10);
981 /* EWALD ELECTROSTATICS */
983 /* Analytical PME correction */
984 zeta2 = _mm256_mul_ps(beta2,rsq10);
985 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
986 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
987 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
988 felec = _mm256_mul_ps(qq10,felec);
989 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
990 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
991 velec = _mm256_sub_ps(rinv10,pmecorrV);
992 velec = _mm256_mul_ps(qq10,velec);
994 /* Update potential sum for this i atom from the interaction with this j atom. */
995 velec = _mm256_andnot_ps(dummy_mask,velec);
996 velecsum = _mm256_add_ps(velecsum,velec);
1000 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1002 /* Calculate temporary vectorial force */
1003 tx = _mm256_mul_ps(fscal,dx10);
1004 ty = _mm256_mul_ps(fscal,dy10);
1005 tz = _mm256_mul_ps(fscal,dz10);
1007 /* Update vectorial force */
1008 fix1 = _mm256_add_ps(fix1,tx);
1009 fiy1 = _mm256_add_ps(fiy1,ty);
1010 fiz1 = _mm256_add_ps(fiz1,tz);
1012 fjx0 = _mm256_add_ps(fjx0,tx);
1013 fjy0 = _mm256_add_ps(fjy0,ty);
1014 fjz0 = _mm256_add_ps(fjz0,tz);
1016 /**************************
1017 * CALCULATE INTERACTIONS *
1018 **************************/
1020 r11 = _mm256_mul_ps(rsq11,rinv11);
1021 r11 = _mm256_andnot_ps(dummy_mask,r11);
1023 /* EWALD ELECTROSTATICS */
1025 /* Analytical PME correction */
1026 zeta2 = _mm256_mul_ps(beta2,rsq11);
1027 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
1028 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1029 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1030 felec = _mm256_mul_ps(qq11,felec);
1031 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1032 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1033 velec = _mm256_sub_ps(rinv11,pmecorrV);
1034 velec = _mm256_mul_ps(qq11,velec);
1036 /* Update potential sum for this i atom from the interaction with this j atom. */
1037 velec = _mm256_andnot_ps(dummy_mask,velec);
1038 velecsum = _mm256_add_ps(velecsum,velec);
1042 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1044 /* Calculate temporary vectorial force */
1045 tx = _mm256_mul_ps(fscal,dx11);
1046 ty = _mm256_mul_ps(fscal,dy11);
1047 tz = _mm256_mul_ps(fscal,dz11);
1049 /* Update vectorial force */
1050 fix1 = _mm256_add_ps(fix1,tx);
1051 fiy1 = _mm256_add_ps(fiy1,ty);
1052 fiz1 = _mm256_add_ps(fiz1,tz);
1054 fjx1 = _mm256_add_ps(fjx1,tx);
1055 fjy1 = _mm256_add_ps(fjy1,ty);
1056 fjz1 = _mm256_add_ps(fjz1,tz);
1058 /**************************
1059 * CALCULATE INTERACTIONS *
1060 **************************/
1062 r12 = _mm256_mul_ps(rsq12,rinv12);
1063 r12 = _mm256_andnot_ps(dummy_mask,r12);
1065 /* EWALD ELECTROSTATICS */
1067 /* Analytical PME correction */
1068 zeta2 = _mm256_mul_ps(beta2,rsq12);
1069 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
1070 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1071 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1072 felec = _mm256_mul_ps(qq12,felec);
1073 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1074 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1075 velec = _mm256_sub_ps(rinv12,pmecorrV);
1076 velec = _mm256_mul_ps(qq12,velec);
1078 /* Update potential sum for this i atom from the interaction with this j atom. */
1079 velec = _mm256_andnot_ps(dummy_mask,velec);
1080 velecsum = _mm256_add_ps(velecsum,velec);
1084 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1086 /* Calculate temporary vectorial force */
1087 tx = _mm256_mul_ps(fscal,dx12);
1088 ty = _mm256_mul_ps(fscal,dy12);
1089 tz = _mm256_mul_ps(fscal,dz12);
1091 /* Update vectorial force */
1092 fix1 = _mm256_add_ps(fix1,tx);
1093 fiy1 = _mm256_add_ps(fiy1,ty);
1094 fiz1 = _mm256_add_ps(fiz1,tz);
1096 fjx2 = _mm256_add_ps(fjx2,tx);
1097 fjy2 = _mm256_add_ps(fjy2,ty);
1098 fjz2 = _mm256_add_ps(fjz2,tz);
1100 /**************************
1101 * CALCULATE INTERACTIONS *
1102 **************************/
1104 r20 = _mm256_mul_ps(rsq20,rinv20);
1105 r20 = _mm256_andnot_ps(dummy_mask,r20);
1107 /* EWALD ELECTROSTATICS */
1109 /* Analytical PME correction */
1110 zeta2 = _mm256_mul_ps(beta2,rsq20);
1111 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1112 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1113 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1114 felec = _mm256_mul_ps(qq20,felec);
1115 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1116 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1117 velec = _mm256_sub_ps(rinv20,pmecorrV);
1118 velec = _mm256_mul_ps(qq20,velec);
1120 /* Update potential sum for this i atom from the interaction with this j atom. */
1121 velec = _mm256_andnot_ps(dummy_mask,velec);
1122 velecsum = _mm256_add_ps(velecsum,velec);
1126 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1128 /* Calculate temporary vectorial force */
1129 tx = _mm256_mul_ps(fscal,dx20);
1130 ty = _mm256_mul_ps(fscal,dy20);
1131 tz = _mm256_mul_ps(fscal,dz20);
1133 /* Update vectorial force */
1134 fix2 = _mm256_add_ps(fix2,tx);
1135 fiy2 = _mm256_add_ps(fiy2,ty);
1136 fiz2 = _mm256_add_ps(fiz2,tz);
1138 fjx0 = _mm256_add_ps(fjx0,tx);
1139 fjy0 = _mm256_add_ps(fjy0,ty);
1140 fjz0 = _mm256_add_ps(fjz0,tz);
1142 /**************************
1143 * CALCULATE INTERACTIONS *
1144 **************************/
1146 r21 = _mm256_mul_ps(rsq21,rinv21);
1147 r21 = _mm256_andnot_ps(dummy_mask,r21);
1149 /* EWALD ELECTROSTATICS */
1151 /* Analytical PME correction */
1152 zeta2 = _mm256_mul_ps(beta2,rsq21);
1153 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
1154 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1155 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1156 felec = _mm256_mul_ps(qq21,felec);
1157 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1158 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1159 velec = _mm256_sub_ps(rinv21,pmecorrV);
1160 velec = _mm256_mul_ps(qq21,velec);
1162 /* Update potential sum for this i atom from the interaction with this j atom. */
1163 velec = _mm256_andnot_ps(dummy_mask,velec);
1164 velecsum = _mm256_add_ps(velecsum,velec);
1168 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1170 /* Calculate temporary vectorial force */
1171 tx = _mm256_mul_ps(fscal,dx21);
1172 ty = _mm256_mul_ps(fscal,dy21);
1173 tz = _mm256_mul_ps(fscal,dz21);
1175 /* Update vectorial force */
1176 fix2 = _mm256_add_ps(fix2,tx);
1177 fiy2 = _mm256_add_ps(fiy2,ty);
1178 fiz2 = _mm256_add_ps(fiz2,tz);
1180 fjx1 = _mm256_add_ps(fjx1,tx);
1181 fjy1 = _mm256_add_ps(fjy1,ty);
1182 fjz1 = _mm256_add_ps(fjz1,tz);
1184 /**************************
1185 * CALCULATE INTERACTIONS *
1186 **************************/
1188 r22 = _mm256_mul_ps(rsq22,rinv22);
1189 r22 = _mm256_andnot_ps(dummy_mask,r22);
1191 /* EWALD ELECTROSTATICS */
1193 /* Analytical PME correction */
1194 zeta2 = _mm256_mul_ps(beta2,rsq22);
1195 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
1196 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1197 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1198 felec = _mm256_mul_ps(qq22,felec);
1199 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1200 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1201 velec = _mm256_sub_ps(rinv22,pmecorrV);
1202 velec = _mm256_mul_ps(qq22,velec);
1204 /* Update potential sum for this i atom from the interaction with this j atom. */
1205 velec = _mm256_andnot_ps(dummy_mask,velec);
1206 velecsum = _mm256_add_ps(velecsum,velec);
1210 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1212 /* Calculate temporary vectorial force */
1213 tx = _mm256_mul_ps(fscal,dx22);
1214 ty = _mm256_mul_ps(fscal,dy22);
1215 tz = _mm256_mul_ps(fscal,dz22);
1217 /* Update vectorial force */
1218 fix2 = _mm256_add_ps(fix2,tx);
1219 fiy2 = _mm256_add_ps(fiy2,ty);
1220 fiz2 = _mm256_add_ps(fiz2,tz);
1222 fjx2 = _mm256_add_ps(fjx2,tx);
1223 fjy2 = _mm256_add_ps(fjy2,ty);
1224 fjz2 = _mm256_add_ps(fjz2,tz);
1226 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1227 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1228 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1229 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1230 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1231 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1232 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1233 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1235 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1236 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1238 /* Inner loop uses 793 flops */
1241 /* End of innermost loop */
1243 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1244 f+i_coord_offset,fshift+i_shift_offset);
1247 /* Update potential energies */
1248 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1249 gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1251 /* Increment number of inner iterations */
1252 inneriter += j_index_end - j_index_start;
1254 /* Outer loop uses 20 flops */
1257 /* Increment number of outer iterations */
1260 /* Update outer/inner flops */
1262 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*793);
1265 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_256_single
1266 * Electrostatics interaction: Ewald
1267 * VdW interaction: LJEwald
1268 * Geometry: Water3-Water3
1269 * Calculate force/pot: Force
1272 nb_kernel_ElecEw_VdwLJEw_GeomW3W3_F_avx_256_single
1273 (t_nblist * gmx_restrict nlist,
1274 rvec * gmx_restrict xx,
1275 rvec * gmx_restrict ff,
1276 t_forcerec * gmx_restrict fr,
1277 t_mdatoms * gmx_restrict mdatoms,
1278 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1279 t_nrnb * gmx_restrict nrnb)
1281 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1282 * just 0 for non-waters.
1283 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1284 * jnr indices corresponding to data put in the four positions in the SIMD register.
1286 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1287 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1288 int jnrA,jnrB,jnrC,jnrD;
1289 int jnrE,jnrF,jnrG,jnrH;
1290 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1291 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1292 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1293 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1294 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1295 real rcutoff_scalar;
1296 real *shiftvec,*fshift,*x,*f;
1297 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1298 real scratch[4*DIM];
1299 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1300 real * vdwioffsetptr0;
1301 real * vdwgridioffsetptr0;
1302 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1303 real * vdwioffsetptr1;
1304 real * vdwgridioffsetptr1;
1305 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1306 real * vdwioffsetptr2;
1307 real * vdwgridioffsetptr2;
1308 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1309 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1310 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1311 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1312 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1313 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1314 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1315 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1316 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1317 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1318 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1319 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1320 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1321 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1322 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1323 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1324 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
1327 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1330 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
1331 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
1342 __m256 ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1343 __m256 one_half = _mm256_set1_ps(0.5);
1344 __m256 minus_one = _mm256_set1_ps(-1.0);
1346 __m128i ewitab_lo,ewitab_hi;
1347 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1348 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1350 __m256 dummy_mask,cutoff_mask;
1351 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1352 __m256 one = _mm256_set1_ps(1.0);
1353 __m256 two = _mm256_set1_ps(2.0);
1359 jindex = nlist->jindex;
1361 shiftidx = nlist->shift;
1363 shiftvec = fr->shift_vec[0];
1364 fshift = fr->fshift[0];
1365 facel = _mm256_set1_ps(fr->epsfac);
1366 charge = mdatoms->chargeA;
1367 nvdwtype = fr->ntype;
1368 vdwparam = fr->nbfp;
1369 vdwtype = mdatoms->typeA;
1370 vdwgridparam = fr->ljpme_c6grid;
1371 sh_lj_ewald = _mm256_set1_ps(fr->ic->sh_lj_ewald);
1372 ewclj = _mm256_set1_ps(fr->ewaldcoeff_lj);
1373 ewclj2 = _mm256_mul_ps(minus_one,_mm256_mul_ps(ewclj,ewclj));
1375 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
1376 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1377 beta2 = _mm256_mul_ps(beta,beta);
1378 beta3 = _mm256_mul_ps(beta,beta2);
1380 ewtab = fr->ic->tabq_coul_F;
1381 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
1382 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1384 /* Setup water-specific parameters */
1385 inr = nlist->iinr[0];
1386 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1387 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1388 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1389 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
1390 vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
1392 jq0 = _mm256_set1_ps(charge[inr+0]);
1393 jq1 = _mm256_set1_ps(charge[inr+1]);
1394 jq2 = _mm256_set1_ps(charge[inr+2]);
1395 vdwjidx0A = 2*vdwtype[inr+0];
1396 qq00 = _mm256_mul_ps(iq0,jq0);
1397 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1398 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1399 c6grid_00 = _mm256_set1_ps(vdwgridioffsetptr0[vdwjidx0A]);
1400 qq01 = _mm256_mul_ps(iq0,jq1);
1401 qq02 = _mm256_mul_ps(iq0,jq2);
1402 qq10 = _mm256_mul_ps(iq1,jq0);
1403 qq11 = _mm256_mul_ps(iq1,jq1);
1404 qq12 = _mm256_mul_ps(iq1,jq2);
1405 qq20 = _mm256_mul_ps(iq2,jq0);
1406 qq21 = _mm256_mul_ps(iq2,jq1);
1407 qq22 = _mm256_mul_ps(iq2,jq2);
1409 /* Avoid stupid compiler warnings */
1410 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1411 j_coord_offsetA = 0;
1412 j_coord_offsetB = 0;
1413 j_coord_offsetC = 0;
1414 j_coord_offsetD = 0;
1415 j_coord_offsetE = 0;
1416 j_coord_offsetF = 0;
1417 j_coord_offsetG = 0;
1418 j_coord_offsetH = 0;
1423 for(iidx=0;iidx<4*DIM;iidx++)
1425 scratch[iidx] = 0.0;
1428 /* Start outer loop over neighborlists */
1429 for(iidx=0; iidx<nri; iidx++)
1431 /* Load shift vector for this list */
1432 i_shift_offset = DIM*shiftidx[iidx];
1434 /* Load limits for loop over neighbors */
1435 j_index_start = jindex[iidx];
1436 j_index_end = jindex[iidx+1];
1438 /* Get outer coordinate index */
1440 i_coord_offset = DIM*inr;
1442 /* Load i particle coords and add shift vector */
1443 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1444 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1446 fix0 = _mm256_setzero_ps();
1447 fiy0 = _mm256_setzero_ps();
1448 fiz0 = _mm256_setzero_ps();
1449 fix1 = _mm256_setzero_ps();
1450 fiy1 = _mm256_setzero_ps();
1451 fiz1 = _mm256_setzero_ps();
1452 fix2 = _mm256_setzero_ps();
1453 fiy2 = _mm256_setzero_ps();
1454 fiz2 = _mm256_setzero_ps();
1456 /* Start inner kernel loop */
1457 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1460 /* Get j neighbor index, and coordinate index */
1462 jnrB = jjnr[jidx+1];
1463 jnrC = jjnr[jidx+2];
1464 jnrD = jjnr[jidx+3];
1465 jnrE = jjnr[jidx+4];
1466 jnrF = jjnr[jidx+5];
1467 jnrG = jjnr[jidx+6];
1468 jnrH = jjnr[jidx+7];
1469 j_coord_offsetA = DIM*jnrA;
1470 j_coord_offsetB = DIM*jnrB;
1471 j_coord_offsetC = DIM*jnrC;
1472 j_coord_offsetD = DIM*jnrD;
1473 j_coord_offsetE = DIM*jnrE;
1474 j_coord_offsetF = DIM*jnrF;
1475 j_coord_offsetG = DIM*jnrG;
1476 j_coord_offsetH = DIM*jnrH;
1478 /* load j atom coordinates */
1479 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1480 x+j_coord_offsetC,x+j_coord_offsetD,
1481 x+j_coord_offsetE,x+j_coord_offsetF,
1482 x+j_coord_offsetG,x+j_coord_offsetH,
1483 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1485 /* Calculate displacement vector */
1486 dx00 = _mm256_sub_ps(ix0,jx0);
1487 dy00 = _mm256_sub_ps(iy0,jy0);
1488 dz00 = _mm256_sub_ps(iz0,jz0);
1489 dx01 = _mm256_sub_ps(ix0,jx1);
1490 dy01 = _mm256_sub_ps(iy0,jy1);
1491 dz01 = _mm256_sub_ps(iz0,jz1);
1492 dx02 = _mm256_sub_ps(ix0,jx2);
1493 dy02 = _mm256_sub_ps(iy0,jy2);
1494 dz02 = _mm256_sub_ps(iz0,jz2);
1495 dx10 = _mm256_sub_ps(ix1,jx0);
1496 dy10 = _mm256_sub_ps(iy1,jy0);
1497 dz10 = _mm256_sub_ps(iz1,jz0);
1498 dx11 = _mm256_sub_ps(ix1,jx1);
1499 dy11 = _mm256_sub_ps(iy1,jy1);
1500 dz11 = _mm256_sub_ps(iz1,jz1);
1501 dx12 = _mm256_sub_ps(ix1,jx2);
1502 dy12 = _mm256_sub_ps(iy1,jy2);
1503 dz12 = _mm256_sub_ps(iz1,jz2);
1504 dx20 = _mm256_sub_ps(ix2,jx0);
1505 dy20 = _mm256_sub_ps(iy2,jy0);
1506 dz20 = _mm256_sub_ps(iz2,jz0);
1507 dx21 = _mm256_sub_ps(ix2,jx1);
1508 dy21 = _mm256_sub_ps(iy2,jy1);
1509 dz21 = _mm256_sub_ps(iz2,jz1);
1510 dx22 = _mm256_sub_ps(ix2,jx2);
1511 dy22 = _mm256_sub_ps(iy2,jy2);
1512 dz22 = _mm256_sub_ps(iz2,jz2);
1514 /* Calculate squared distance and things based on it */
1515 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1516 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1517 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1518 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1519 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1520 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1521 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1522 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1523 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1525 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1526 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1527 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1528 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1529 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1530 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1531 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1532 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1533 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1535 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1536 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1537 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1538 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1539 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1540 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1541 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1542 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1543 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1545 fjx0 = _mm256_setzero_ps();
1546 fjy0 = _mm256_setzero_ps();
1547 fjz0 = _mm256_setzero_ps();
1548 fjx1 = _mm256_setzero_ps();
1549 fjy1 = _mm256_setzero_ps();
1550 fjz1 = _mm256_setzero_ps();
1551 fjx2 = _mm256_setzero_ps();
1552 fjy2 = _mm256_setzero_ps();
1553 fjz2 = _mm256_setzero_ps();
1555 /**************************
1556 * CALCULATE INTERACTIONS *
1557 **************************/
1559 r00 = _mm256_mul_ps(rsq00,rinv00);
1561 /* EWALD ELECTROSTATICS */
1563 /* Analytical PME correction */
1564 zeta2 = _mm256_mul_ps(beta2,rsq00);
1565 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1566 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1567 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1568 felec = _mm256_mul_ps(qq00,felec);
1570 /* Analytical LJ-PME */
1571 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1572 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
1573 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
1574 exponent = gmx_simd_exp_r(ewcljrsq);
1575 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1576 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
1577 /* f6A = 6 * C6grid * (1 - poly) */
1578 f6A = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
1579 /* f6B = C6grid * exponent * beta^6 */
1580 f6B = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
1581 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1582 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);
1584 fscal = _mm256_add_ps(felec,fvdw);
1586 /* Calculate temporary vectorial force */
1587 tx = _mm256_mul_ps(fscal,dx00);
1588 ty = _mm256_mul_ps(fscal,dy00);
1589 tz = _mm256_mul_ps(fscal,dz00);
1591 /* Update vectorial force */
1592 fix0 = _mm256_add_ps(fix0,tx);
1593 fiy0 = _mm256_add_ps(fiy0,ty);
1594 fiz0 = _mm256_add_ps(fiz0,tz);
1596 fjx0 = _mm256_add_ps(fjx0,tx);
1597 fjy0 = _mm256_add_ps(fjy0,ty);
1598 fjz0 = _mm256_add_ps(fjz0,tz);
1600 /**************************
1601 * CALCULATE INTERACTIONS *
1602 **************************/
1604 r01 = _mm256_mul_ps(rsq01,rinv01);
1606 /* EWALD ELECTROSTATICS */
1608 /* Analytical PME correction */
1609 zeta2 = _mm256_mul_ps(beta2,rsq01);
1610 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1611 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1612 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1613 felec = _mm256_mul_ps(qq01,felec);
1617 /* Calculate temporary vectorial force */
1618 tx = _mm256_mul_ps(fscal,dx01);
1619 ty = _mm256_mul_ps(fscal,dy01);
1620 tz = _mm256_mul_ps(fscal,dz01);
1622 /* Update vectorial force */
1623 fix0 = _mm256_add_ps(fix0,tx);
1624 fiy0 = _mm256_add_ps(fiy0,ty);
1625 fiz0 = _mm256_add_ps(fiz0,tz);
1627 fjx1 = _mm256_add_ps(fjx1,tx);
1628 fjy1 = _mm256_add_ps(fjy1,ty);
1629 fjz1 = _mm256_add_ps(fjz1,tz);
1631 /**************************
1632 * CALCULATE INTERACTIONS *
1633 **************************/
1635 r02 = _mm256_mul_ps(rsq02,rinv02);
1637 /* EWALD ELECTROSTATICS */
1639 /* Analytical PME correction */
1640 zeta2 = _mm256_mul_ps(beta2,rsq02);
1641 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
1642 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1643 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1644 felec = _mm256_mul_ps(qq02,felec);
1648 /* Calculate temporary vectorial force */
1649 tx = _mm256_mul_ps(fscal,dx02);
1650 ty = _mm256_mul_ps(fscal,dy02);
1651 tz = _mm256_mul_ps(fscal,dz02);
1653 /* Update vectorial force */
1654 fix0 = _mm256_add_ps(fix0,tx);
1655 fiy0 = _mm256_add_ps(fiy0,ty);
1656 fiz0 = _mm256_add_ps(fiz0,tz);
1658 fjx2 = _mm256_add_ps(fjx2,tx);
1659 fjy2 = _mm256_add_ps(fjy2,ty);
1660 fjz2 = _mm256_add_ps(fjz2,tz);
1662 /**************************
1663 * CALCULATE INTERACTIONS *
1664 **************************/
1666 r10 = _mm256_mul_ps(rsq10,rinv10);
1668 /* EWALD ELECTROSTATICS */
1670 /* Analytical PME correction */
1671 zeta2 = _mm256_mul_ps(beta2,rsq10);
1672 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
1673 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1674 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1675 felec = _mm256_mul_ps(qq10,felec);
1679 /* Calculate temporary vectorial force */
1680 tx = _mm256_mul_ps(fscal,dx10);
1681 ty = _mm256_mul_ps(fscal,dy10);
1682 tz = _mm256_mul_ps(fscal,dz10);
1684 /* Update vectorial force */
1685 fix1 = _mm256_add_ps(fix1,tx);
1686 fiy1 = _mm256_add_ps(fiy1,ty);
1687 fiz1 = _mm256_add_ps(fiz1,tz);
1689 fjx0 = _mm256_add_ps(fjx0,tx);
1690 fjy0 = _mm256_add_ps(fjy0,ty);
1691 fjz0 = _mm256_add_ps(fjz0,tz);
1693 /**************************
1694 * CALCULATE INTERACTIONS *
1695 **************************/
1697 r11 = _mm256_mul_ps(rsq11,rinv11);
1699 /* EWALD ELECTROSTATICS */
1701 /* Analytical PME correction */
1702 zeta2 = _mm256_mul_ps(beta2,rsq11);
1703 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
1704 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1705 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1706 felec = _mm256_mul_ps(qq11,felec);
1710 /* Calculate temporary vectorial force */
1711 tx = _mm256_mul_ps(fscal,dx11);
1712 ty = _mm256_mul_ps(fscal,dy11);
1713 tz = _mm256_mul_ps(fscal,dz11);
1715 /* Update vectorial force */
1716 fix1 = _mm256_add_ps(fix1,tx);
1717 fiy1 = _mm256_add_ps(fiy1,ty);
1718 fiz1 = _mm256_add_ps(fiz1,tz);
1720 fjx1 = _mm256_add_ps(fjx1,tx);
1721 fjy1 = _mm256_add_ps(fjy1,ty);
1722 fjz1 = _mm256_add_ps(fjz1,tz);
1724 /**************************
1725 * CALCULATE INTERACTIONS *
1726 **************************/
1728 r12 = _mm256_mul_ps(rsq12,rinv12);
1730 /* EWALD ELECTROSTATICS */
1732 /* Analytical PME correction */
1733 zeta2 = _mm256_mul_ps(beta2,rsq12);
1734 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
1735 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1736 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1737 felec = _mm256_mul_ps(qq12,felec);
1741 /* Calculate temporary vectorial force */
1742 tx = _mm256_mul_ps(fscal,dx12);
1743 ty = _mm256_mul_ps(fscal,dy12);
1744 tz = _mm256_mul_ps(fscal,dz12);
1746 /* Update vectorial force */
1747 fix1 = _mm256_add_ps(fix1,tx);
1748 fiy1 = _mm256_add_ps(fiy1,ty);
1749 fiz1 = _mm256_add_ps(fiz1,tz);
1751 fjx2 = _mm256_add_ps(fjx2,tx);
1752 fjy2 = _mm256_add_ps(fjy2,ty);
1753 fjz2 = _mm256_add_ps(fjz2,tz);
1755 /**************************
1756 * CALCULATE INTERACTIONS *
1757 **************************/
1759 r20 = _mm256_mul_ps(rsq20,rinv20);
1761 /* EWALD ELECTROSTATICS */
1763 /* Analytical PME correction */
1764 zeta2 = _mm256_mul_ps(beta2,rsq20);
1765 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1766 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1767 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1768 felec = _mm256_mul_ps(qq20,felec);
1772 /* Calculate temporary vectorial force */
1773 tx = _mm256_mul_ps(fscal,dx20);
1774 ty = _mm256_mul_ps(fscal,dy20);
1775 tz = _mm256_mul_ps(fscal,dz20);
1777 /* Update vectorial force */
1778 fix2 = _mm256_add_ps(fix2,tx);
1779 fiy2 = _mm256_add_ps(fiy2,ty);
1780 fiz2 = _mm256_add_ps(fiz2,tz);
1782 fjx0 = _mm256_add_ps(fjx0,tx);
1783 fjy0 = _mm256_add_ps(fjy0,ty);
1784 fjz0 = _mm256_add_ps(fjz0,tz);
1786 /**************************
1787 * CALCULATE INTERACTIONS *
1788 **************************/
1790 r21 = _mm256_mul_ps(rsq21,rinv21);
1792 /* EWALD ELECTROSTATICS */
1794 /* Analytical PME correction */
1795 zeta2 = _mm256_mul_ps(beta2,rsq21);
1796 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
1797 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1798 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1799 felec = _mm256_mul_ps(qq21,felec);
1803 /* Calculate temporary vectorial force */
1804 tx = _mm256_mul_ps(fscal,dx21);
1805 ty = _mm256_mul_ps(fscal,dy21);
1806 tz = _mm256_mul_ps(fscal,dz21);
1808 /* Update vectorial force */
1809 fix2 = _mm256_add_ps(fix2,tx);
1810 fiy2 = _mm256_add_ps(fiy2,ty);
1811 fiz2 = _mm256_add_ps(fiz2,tz);
1813 fjx1 = _mm256_add_ps(fjx1,tx);
1814 fjy1 = _mm256_add_ps(fjy1,ty);
1815 fjz1 = _mm256_add_ps(fjz1,tz);
1817 /**************************
1818 * CALCULATE INTERACTIONS *
1819 **************************/
1821 r22 = _mm256_mul_ps(rsq22,rinv22);
1823 /* EWALD ELECTROSTATICS */
1825 /* Analytical PME correction */
1826 zeta2 = _mm256_mul_ps(beta2,rsq22);
1827 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
1828 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1829 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1830 felec = _mm256_mul_ps(qq22,felec);
1834 /* Calculate temporary vectorial force */
1835 tx = _mm256_mul_ps(fscal,dx22);
1836 ty = _mm256_mul_ps(fscal,dy22);
1837 tz = _mm256_mul_ps(fscal,dz22);
1839 /* Update vectorial force */
1840 fix2 = _mm256_add_ps(fix2,tx);
1841 fiy2 = _mm256_add_ps(fiy2,ty);
1842 fiz2 = _mm256_add_ps(fiz2,tz);
1844 fjx2 = _mm256_add_ps(fjx2,tx);
1845 fjy2 = _mm256_add_ps(fjy2,ty);
1846 fjz2 = _mm256_add_ps(fjz2,tz);
1848 fjptrA = f+j_coord_offsetA;
1849 fjptrB = f+j_coord_offsetB;
1850 fjptrC = f+j_coord_offsetC;
1851 fjptrD = f+j_coord_offsetD;
1852 fjptrE = f+j_coord_offsetE;
1853 fjptrF = f+j_coord_offsetF;
1854 fjptrG = f+j_coord_offsetG;
1855 fjptrH = f+j_coord_offsetH;
1857 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1858 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1860 /* Inner loop uses 527 flops */
1863 if(jidx<j_index_end)
1866 /* Get j neighbor index, and coordinate index */
1867 jnrlistA = jjnr[jidx];
1868 jnrlistB = jjnr[jidx+1];
1869 jnrlistC = jjnr[jidx+2];
1870 jnrlistD = jjnr[jidx+3];
1871 jnrlistE = jjnr[jidx+4];
1872 jnrlistF = jjnr[jidx+5];
1873 jnrlistG = jjnr[jidx+6];
1874 jnrlistH = jjnr[jidx+7];
1875 /* Sign of each element will be negative for non-real atoms.
1876 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1877 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1879 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1880 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1882 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1883 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1884 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1885 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1886 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
1887 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
1888 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
1889 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
1890 j_coord_offsetA = DIM*jnrA;
1891 j_coord_offsetB = DIM*jnrB;
1892 j_coord_offsetC = DIM*jnrC;
1893 j_coord_offsetD = DIM*jnrD;
1894 j_coord_offsetE = DIM*jnrE;
1895 j_coord_offsetF = DIM*jnrF;
1896 j_coord_offsetG = DIM*jnrG;
1897 j_coord_offsetH = DIM*jnrH;
1899 /* load j atom coordinates */
1900 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1901 x+j_coord_offsetC,x+j_coord_offsetD,
1902 x+j_coord_offsetE,x+j_coord_offsetF,
1903 x+j_coord_offsetG,x+j_coord_offsetH,
1904 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1906 /* Calculate displacement vector */
1907 dx00 = _mm256_sub_ps(ix0,jx0);
1908 dy00 = _mm256_sub_ps(iy0,jy0);
1909 dz00 = _mm256_sub_ps(iz0,jz0);
1910 dx01 = _mm256_sub_ps(ix0,jx1);
1911 dy01 = _mm256_sub_ps(iy0,jy1);
1912 dz01 = _mm256_sub_ps(iz0,jz1);
1913 dx02 = _mm256_sub_ps(ix0,jx2);
1914 dy02 = _mm256_sub_ps(iy0,jy2);
1915 dz02 = _mm256_sub_ps(iz0,jz2);
1916 dx10 = _mm256_sub_ps(ix1,jx0);
1917 dy10 = _mm256_sub_ps(iy1,jy0);
1918 dz10 = _mm256_sub_ps(iz1,jz0);
1919 dx11 = _mm256_sub_ps(ix1,jx1);
1920 dy11 = _mm256_sub_ps(iy1,jy1);
1921 dz11 = _mm256_sub_ps(iz1,jz1);
1922 dx12 = _mm256_sub_ps(ix1,jx2);
1923 dy12 = _mm256_sub_ps(iy1,jy2);
1924 dz12 = _mm256_sub_ps(iz1,jz2);
1925 dx20 = _mm256_sub_ps(ix2,jx0);
1926 dy20 = _mm256_sub_ps(iy2,jy0);
1927 dz20 = _mm256_sub_ps(iz2,jz0);
1928 dx21 = _mm256_sub_ps(ix2,jx1);
1929 dy21 = _mm256_sub_ps(iy2,jy1);
1930 dz21 = _mm256_sub_ps(iz2,jz1);
1931 dx22 = _mm256_sub_ps(ix2,jx2);
1932 dy22 = _mm256_sub_ps(iy2,jy2);
1933 dz22 = _mm256_sub_ps(iz2,jz2);
1935 /* Calculate squared distance and things based on it */
1936 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1937 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1938 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1939 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1940 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1941 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1942 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1943 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1944 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1946 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1947 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1948 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1949 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1950 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1951 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1952 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1953 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1954 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1956 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1957 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1958 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1959 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1960 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1961 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1962 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1963 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1964 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1966 fjx0 = _mm256_setzero_ps();
1967 fjy0 = _mm256_setzero_ps();
1968 fjz0 = _mm256_setzero_ps();
1969 fjx1 = _mm256_setzero_ps();
1970 fjy1 = _mm256_setzero_ps();
1971 fjz1 = _mm256_setzero_ps();
1972 fjx2 = _mm256_setzero_ps();
1973 fjy2 = _mm256_setzero_ps();
1974 fjz2 = _mm256_setzero_ps();
1976 /**************************
1977 * CALCULATE INTERACTIONS *
1978 **************************/
1980 r00 = _mm256_mul_ps(rsq00,rinv00);
1981 r00 = _mm256_andnot_ps(dummy_mask,r00);
1983 /* EWALD ELECTROSTATICS */
1985 /* Analytical PME correction */
1986 zeta2 = _mm256_mul_ps(beta2,rsq00);
1987 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1988 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1989 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1990 felec = _mm256_mul_ps(qq00,felec);
1992 /* Analytical LJ-PME */
1993 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1994 ewcljrsq = _mm256_mul_ps(ewclj2,rsq00);
1995 ewclj6 = _mm256_mul_ps(ewclj2,_mm256_mul_ps(ewclj2,ewclj2));
1996 exponent = gmx_simd_exp_r(ewcljrsq);
1997 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1998 poly = _mm256_mul_ps(exponent,_mm256_add_ps(_mm256_sub_ps(one,ewcljrsq),_mm256_mul_ps(_mm256_mul_ps(ewcljrsq,ewcljrsq),one_half)));
1999 /* f6A = 6 * C6grid * (1 - poly) */
2000 f6A = _mm256_mul_ps(c6grid_00,_mm256_sub_ps(one,poly));
2001 /* f6B = C6grid * exponent * beta^6 */
2002 f6B = _mm256_mul_ps(_mm256_mul_ps(c6grid_00,one_sixth),_mm256_mul_ps(exponent,ewclj6));
2003 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2004 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);
2006 fscal = _mm256_add_ps(felec,fvdw);
2008 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2010 /* Calculate temporary vectorial force */
2011 tx = _mm256_mul_ps(fscal,dx00);
2012 ty = _mm256_mul_ps(fscal,dy00);
2013 tz = _mm256_mul_ps(fscal,dz00);
2015 /* Update vectorial force */
2016 fix0 = _mm256_add_ps(fix0,tx);
2017 fiy0 = _mm256_add_ps(fiy0,ty);
2018 fiz0 = _mm256_add_ps(fiz0,tz);
2020 fjx0 = _mm256_add_ps(fjx0,tx);
2021 fjy0 = _mm256_add_ps(fjy0,ty);
2022 fjz0 = _mm256_add_ps(fjz0,tz);
2024 /**************************
2025 * CALCULATE INTERACTIONS *
2026 **************************/
2028 r01 = _mm256_mul_ps(rsq01,rinv01);
2029 r01 = _mm256_andnot_ps(dummy_mask,r01);
2031 /* EWALD ELECTROSTATICS */
2033 /* Analytical PME correction */
2034 zeta2 = _mm256_mul_ps(beta2,rsq01);
2035 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
2036 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2037 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2038 felec = _mm256_mul_ps(qq01,felec);
2042 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2044 /* Calculate temporary vectorial force */
2045 tx = _mm256_mul_ps(fscal,dx01);
2046 ty = _mm256_mul_ps(fscal,dy01);
2047 tz = _mm256_mul_ps(fscal,dz01);
2049 /* Update vectorial force */
2050 fix0 = _mm256_add_ps(fix0,tx);
2051 fiy0 = _mm256_add_ps(fiy0,ty);
2052 fiz0 = _mm256_add_ps(fiz0,tz);
2054 fjx1 = _mm256_add_ps(fjx1,tx);
2055 fjy1 = _mm256_add_ps(fjy1,ty);
2056 fjz1 = _mm256_add_ps(fjz1,tz);
2058 /**************************
2059 * CALCULATE INTERACTIONS *
2060 **************************/
2062 r02 = _mm256_mul_ps(rsq02,rinv02);
2063 r02 = _mm256_andnot_ps(dummy_mask,r02);
2065 /* EWALD ELECTROSTATICS */
2067 /* Analytical PME correction */
2068 zeta2 = _mm256_mul_ps(beta2,rsq02);
2069 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
2070 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2071 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2072 felec = _mm256_mul_ps(qq02,felec);
2076 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2078 /* Calculate temporary vectorial force */
2079 tx = _mm256_mul_ps(fscal,dx02);
2080 ty = _mm256_mul_ps(fscal,dy02);
2081 tz = _mm256_mul_ps(fscal,dz02);
2083 /* Update vectorial force */
2084 fix0 = _mm256_add_ps(fix0,tx);
2085 fiy0 = _mm256_add_ps(fiy0,ty);
2086 fiz0 = _mm256_add_ps(fiz0,tz);
2088 fjx2 = _mm256_add_ps(fjx2,tx);
2089 fjy2 = _mm256_add_ps(fjy2,ty);
2090 fjz2 = _mm256_add_ps(fjz2,tz);
2092 /**************************
2093 * CALCULATE INTERACTIONS *
2094 **************************/
2096 r10 = _mm256_mul_ps(rsq10,rinv10);
2097 r10 = _mm256_andnot_ps(dummy_mask,r10);
2099 /* EWALD ELECTROSTATICS */
2101 /* Analytical PME correction */
2102 zeta2 = _mm256_mul_ps(beta2,rsq10);
2103 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2104 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2105 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2106 felec = _mm256_mul_ps(qq10,felec);
2110 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2112 /* Calculate temporary vectorial force */
2113 tx = _mm256_mul_ps(fscal,dx10);
2114 ty = _mm256_mul_ps(fscal,dy10);
2115 tz = _mm256_mul_ps(fscal,dz10);
2117 /* Update vectorial force */
2118 fix1 = _mm256_add_ps(fix1,tx);
2119 fiy1 = _mm256_add_ps(fiy1,ty);
2120 fiz1 = _mm256_add_ps(fiz1,tz);
2122 fjx0 = _mm256_add_ps(fjx0,tx);
2123 fjy0 = _mm256_add_ps(fjy0,ty);
2124 fjz0 = _mm256_add_ps(fjz0,tz);
2126 /**************************
2127 * CALCULATE INTERACTIONS *
2128 **************************/
2130 r11 = _mm256_mul_ps(rsq11,rinv11);
2131 r11 = _mm256_andnot_ps(dummy_mask,r11);
2133 /* EWALD ELECTROSTATICS */
2135 /* Analytical PME correction */
2136 zeta2 = _mm256_mul_ps(beta2,rsq11);
2137 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2138 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2139 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2140 felec = _mm256_mul_ps(qq11,felec);
2144 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2146 /* Calculate temporary vectorial force */
2147 tx = _mm256_mul_ps(fscal,dx11);
2148 ty = _mm256_mul_ps(fscal,dy11);
2149 tz = _mm256_mul_ps(fscal,dz11);
2151 /* Update vectorial force */
2152 fix1 = _mm256_add_ps(fix1,tx);
2153 fiy1 = _mm256_add_ps(fiy1,ty);
2154 fiz1 = _mm256_add_ps(fiz1,tz);
2156 fjx1 = _mm256_add_ps(fjx1,tx);
2157 fjy1 = _mm256_add_ps(fjy1,ty);
2158 fjz1 = _mm256_add_ps(fjz1,tz);
2160 /**************************
2161 * CALCULATE INTERACTIONS *
2162 **************************/
2164 r12 = _mm256_mul_ps(rsq12,rinv12);
2165 r12 = _mm256_andnot_ps(dummy_mask,r12);
2167 /* EWALD ELECTROSTATICS */
2169 /* Analytical PME correction */
2170 zeta2 = _mm256_mul_ps(beta2,rsq12);
2171 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2172 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2173 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2174 felec = _mm256_mul_ps(qq12,felec);
2178 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2180 /* Calculate temporary vectorial force */
2181 tx = _mm256_mul_ps(fscal,dx12);
2182 ty = _mm256_mul_ps(fscal,dy12);
2183 tz = _mm256_mul_ps(fscal,dz12);
2185 /* Update vectorial force */
2186 fix1 = _mm256_add_ps(fix1,tx);
2187 fiy1 = _mm256_add_ps(fiy1,ty);
2188 fiz1 = _mm256_add_ps(fiz1,tz);
2190 fjx2 = _mm256_add_ps(fjx2,tx);
2191 fjy2 = _mm256_add_ps(fjy2,ty);
2192 fjz2 = _mm256_add_ps(fjz2,tz);
2194 /**************************
2195 * CALCULATE INTERACTIONS *
2196 **************************/
2198 r20 = _mm256_mul_ps(rsq20,rinv20);
2199 r20 = _mm256_andnot_ps(dummy_mask,r20);
2201 /* EWALD ELECTROSTATICS */
2203 /* Analytical PME correction */
2204 zeta2 = _mm256_mul_ps(beta2,rsq20);
2205 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
2206 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2207 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2208 felec = _mm256_mul_ps(qq20,felec);
2212 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2214 /* Calculate temporary vectorial force */
2215 tx = _mm256_mul_ps(fscal,dx20);
2216 ty = _mm256_mul_ps(fscal,dy20);
2217 tz = _mm256_mul_ps(fscal,dz20);
2219 /* Update vectorial force */
2220 fix2 = _mm256_add_ps(fix2,tx);
2221 fiy2 = _mm256_add_ps(fiy2,ty);
2222 fiz2 = _mm256_add_ps(fiz2,tz);
2224 fjx0 = _mm256_add_ps(fjx0,tx);
2225 fjy0 = _mm256_add_ps(fjy0,ty);
2226 fjz0 = _mm256_add_ps(fjz0,tz);
2228 /**************************
2229 * CALCULATE INTERACTIONS *
2230 **************************/
2232 r21 = _mm256_mul_ps(rsq21,rinv21);
2233 r21 = _mm256_andnot_ps(dummy_mask,r21);
2235 /* EWALD ELECTROSTATICS */
2237 /* Analytical PME correction */
2238 zeta2 = _mm256_mul_ps(beta2,rsq21);
2239 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2240 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2241 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2242 felec = _mm256_mul_ps(qq21,felec);
2246 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2248 /* Calculate temporary vectorial force */
2249 tx = _mm256_mul_ps(fscal,dx21);
2250 ty = _mm256_mul_ps(fscal,dy21);
2251 tz = _mm256_mul_ps(fscal,dz21);
2253 /* Update vectorial force */
2254 fix2 = _mm256_add_ps(fix2,tx);
2255 fiy2 = _mm256_add_ps(fiy2,ty);
2256 fiz2 = _mm256_add_ps(fiz2,tz);
2258 fjx1 = _mm256_add_ps(fjx1,tx);
2259 fjy1 = _mm256_add_ps(fjy1,ty);
2260 fjz1 = _mm256_add_ps(fjz1,tz);
2262 /**************************
2263 * CALCULATE INTERACTIONS *
2264 **************************/
2266 r22 = _mm256_mul_ps(rsq22,rinv22);
2267 r22 = _mm256_andnot_ps(dummy_mask,r22);
2269 /* EWALD ELECTROSTATICS */
2271 /* Analytical PME correction */
2272 zeta2 = _mm256_mul_ps(beta2,rsq22);
2273 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
2274 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2275 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2276 felec = _mm256_mul_ps(qq22,felec);
2280 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2282 /* Calculate temporary vectorial force */
2283 tx = _mm256_mul_ps(fscal,dx22);
2284 ty = _mm256_mul_ps(fscal,dy22);
2285 tz = _mm256_mul_ps(fscal,dz22);
2287 /* Update vectorial force */
2288 fix2 = _mm256_add_ps(fix2,tx);
2289 fiy2 = _mm256_add_ps(fiy2,ty);
2290 fiz2 = _mm256_add_ps(fiz2,tz);
2292 fjx2 = _mm256_add_ps(fjx2,tx);
2293 fjy2 = _mm256_add_ps(fjy2,ty);
2294 fjz2 = _mm256_add_ps(fjz2,tz);
2296 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2297 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2298 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2299 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2300 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2301 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2302 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2303 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2305 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2306 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2308 /* Inner loop uses 536 flops */
2311 /* End of innermost loop */
2313 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2314 f+i_coord_offset,fshift+i_shift_offset);
2316 /* Increment number of inner iterations */
2317 inneriter += j_index_end - j_index_start;
2319 /* Outer loop uses 18 flops */
2322 /* Increment number of outer iterations */
2325 /* Update outer/inner flops */
2327 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*536);