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 "gromacs/legacyheaders/types/simple.h"
46 #include "gromacs/math/vec.h"
47 #include "gromacs/legacyheaders/nrnb.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_ElecEwSw_VdwNone_GeomW3W3_VF_avx_256_single
54 * Electrostatics interaction: Ewald
55 * VdW interaction: None
56 * Geometry: Water3-Water3
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEwSw_VdwNone_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 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
90 real * vdwioffsetptr1;
91 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
92 real * vdwioffsetptr2;
93 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
94 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
95 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
96 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
97 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
98 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
99 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
100 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
101 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
102 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
103 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
104 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
105 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
106 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
107 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
108 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
109 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
112 __m128i ewitab_lo,ewitab_hi;
113 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
114 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
116 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
117 real rswitch_scalar,d_scalar;
118 __m256 dummy_mask,cutoff_mask;
119 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
120 __m256 one = _mm256_set1_ps(1.0);
121 __m256 two = _mm256_set1_ps(2.0);
127 jindex = nlist->jindex;
129 shiftidx = nlist->shift;
131 shiftvec = fr->shift_vec[0];
132 fshift = fr->fshift[0];
133 facel = _mm256_set1_ps(fr->epsfac);
134 charge = mdatoms->chargeA;
136 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
137 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
138 beta2 = _mm256_mul_ps(beta,beta);
139 beta3 = _mm256_mul_ps(beta,beta2);
141 ewtab = fr->ic->tabq_coul_FDV0;
142 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
143 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
145 /* Setup water-specific parameters */
146 inr = nlist->iinr[0];
147 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
148 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
149 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
151 jq0 = _mm256_set1_ps(charge[inr+0]);
152 jq1 = _mm256_set1_ps(charge[inr+1]);
153 jq2 = _mm256_set1_ps(charge[inr+2]);
154 qq00 = _mm256_mul_ps(iq0,jq0);
155 qq01 = _mm256_mul_ps(iq0,jq1);
156 qq02 = _mm256_mul_ps(iq0,jq2);
157 qq10 = _mm256_mul_ps(iq1,jq0);
158 qq11 = _mm256_mul_ps(iq1,jq1);
159 qq12 = _mm256_mul_ps(iq1,jq2);
160 qq20 = _mm256_mul_ps(iq2,jq0);
161 qq21 = _mm256_mul_ps(iq2,jq1);
162 qq22 = _mm256_mul_ps(iq2,jq2);
164 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
165 rcutoff_scalar = fr->rcoulomb;
166 rcutoff = _mm256_set1_ps(rcutoff_scalar);
167 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
169 rswitch_scalar = fr->rcoulomb_switch;
170 rswitch = _mm256_set1_ps(rswitch_scalar);
171 /* Setup switch parameters */
172 d_scalar = rcutoff_scalar-rswitch_scalar;
173 d = _mm256_set1_ps(d_scalar);
174 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
175 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
176 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
177 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
178 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
179 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
181 /* Avoid stupid compiler warnings */
182 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
195 for(iidx=0;iidx<4*DIM;iidx++)
200 /* Start outer loop over neighborlists */
201 for(iidx=0; iidx<nri; iidx++)
203 /* Load shift vector for this list */
204 i_shift_offset = DIM*shiftidx[iidx];
206 /* Load limits for loop over neighbors */
207 j_index_start = jindex[iidx];
208 j_index_end = jindex[iidx+1];
210 /* Get outer coordinate index */
212 i_coord_offset = DIM*inr;
214 /* Load i particle coords and add shift vector */
215 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
216 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
218 fix0 = _mm256_setzero_ps();
219 fiy0 = _mm256_setzero_ps();
220 fiz0 = _mm256_setzero_ps();
221 fix1 = _mm256_setzero_ps();
222 fiy1 = _mm256_setzero_ps();
223 fiz1 = _mm256_setzero_ps();
224 fix2 = _mm256_setzero_ps();
225 fiy2 = _mm256_setzero_ps();
226 fiz2 = _mm256_setzero_ps();
228 /* Reset potential sums */
229 velecsum = _mm256_setzero_ps();
231 /* Start inner kernel loop */
232 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
235 /* Get j neighbor index, and coordinate index */
244 j_coord_offsetA = DIM*jnrA;
245 j_coord_offsetB = DIM*jnrB;
246 j_coord_offsetC = DIM*jnrC;
247 j_coord_offsetD = DIM*jnrD;
248 j_coord_offsetE = DIM*jnrE;
249 j_coord_offsetF = DIM*jnrF;
250 j_coord_offsetG = DIM*jnrG;
251 j_coord_offsetH = DIM*jnrH;
253 /* load j atom coordinates */
254 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
255 x+j_coord_offsetC,x+j_coord_offsetD,
256 x+j_coord_offsetE,x+j_coord_offsetF,
257 x+j_coord_offsetG,x+j_coord_offsetH,
258 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
260 /* Calculate displacement vector */
261 dx00 = _mm256_sub_ps(ix0,jx0);
262 dy00 = _mm256_sub_ps(iy0,jy0);
263 dz00 = _mm256_sub_ps(iz0,jz0);
264 dx01 = _mm256_sub_ps(ix0,jx1);
265 dy01 = _mm256_sub_ps(iy0,jy1);
266 dz01 = _mm256_sub_ps(iz0,jz1);
267 dx02 = _mm256_sub_ps(ix0,jx2);
268 dy02 = _mm256_sub_ps(iy0,jy2);
269 dz02 = _mm256_sub_ps(iz0,jz2);
270 dx10 = _mm256_sub_ps(ix1,jx0);
271 dy10 = _mm256_sub_ps(iy1,jy0);
272 dz10 = _mm256_sub_ps(iz1,jz0);
273 dx11 = _mm256_sub_ps(ix1,jx1);
274 dy11 = _mm256_sub_ps(iy1,jy1);
275 dz11 = _mm256_sub_ps(iz1,jz1);
276 dx12 = _mm256_sub_ps(ix1,jx2);
277 dy12 = _mm256_sub_ps(iy1,jy2);
278 dz12 = _mm256_sub_ps(iz1,jz2);
279 dx20 = _mm256_sub_ps(ix2,jx0);
280 dy20 = _mm256_sub_ps(iy2,jy0);
281 dz20 = _mm256_sub_ps(iz2,jz0);
282 dx21 = _mm256_sub_ps(ix2,jx1);
283 dy21 = _mm256_sub_ps(iy2,jy1);
284 dz21 = _mm256_sub_ps(iz2,jz1);
285 dx22 = _mm256_sub_ps(ix2,jx2);
286 dy22 = _mm256_sub_ps(iy2,jy2);
287 dz22 = _mm256_sub_ps(iz2,jz2);
289 /* Calculate squared distance and things based on it */
290 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
291 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
292 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
293 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
294 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
295 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
296 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
297 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
298 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
300 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
301 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
302 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
303 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
304 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
305 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
306 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
307 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
308 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
310 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
311 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
312 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
313 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
314 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
315 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
316 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
317 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
318 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
320 fjx0 = _mm256_setzero_ps();
321 fjy0 = _mm256_setzero_ps();
322 fjz0 = _mm256_setzero_ps();
323 fjx1 = _mm256_setzero_ps();
324 fjy1 = _mm256_setzero_ps();
325 fjz1 = _mm256_setzero_ps();
326 fjx2 = _mm256_setzero_ps();
327 fjy2 = _mm256_setzero_ps();
328 fjz2 = _mm256_setzero_ps();
330 /**************************
331 * CALCULATE INTERACTIONS *
332 **************************/
334 if (gmx_mm256_any_lt(rsq00,rcutoff2))
337 r00 = _mm256_mul_ps(rsq00,rinv00);
339 /* EWALD ELECTROSTATICS */
341 /* Analytical PME correction */
342 zeta2 = _mm256_mul_ps(beta2,rsq00);
343 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
344 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
345 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
346 felec = _mm256_mul_ps(qq00,felec);
347 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
348 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
349 velec = _mm256_sub_ps(rinv00,pmecorrV);
350 velec = _mm256_mul_ps(qq00,velec);
352 d = _mm256_sub_ps(r00,rswitch);
353 d = _mm256_max_ps(d,_mm256_setzero_ps());
354 d2 = _mm256_mul_ps(d,d);
355 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
357 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
359 /* Evaluate switch function */
360 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
361 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
362 velec = _mm256_mul_ps(velec,sw);
363 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
365 /* Update potential sum for this i atom from the interaction with this j atom. */
366 velec = _mm256_and_ps(velec,cutoff_mask);
367 velecsum = _mm256_add_ps(velecsum,velec);
371 fscal = _mm256_and_ps(fscal,cutoff_mask);
373 /* Calculate temporary vectorial force */
374 tx = _mm256_mul_ps(fscal,dx00);
375 ty = _mm256_mul_ps(fscal,dy00);
376 tz = _mm256_mul_ps(fscal,dz00);
378 /* Update vectorial force */
379 fix0 = _mm256_add_ps(fix0,tx);
380 fiy0 = _mm256_add_ps(fiy0,ty);
381 fiz0 = _mm256_add_ps(fiz0,tz);
383 fjx0 = _mm256_add_ps(fjx0,tx);
384 fjy0 = _mm256_add_ps(fjy0,ty);
385 fjz0 = _mm256_add_ps(fjz0,tz);
389 /**************************
390 * CALCULATE INTERACTIONS *
391 **************************/
393 if (gmx_mm256_any_lt(rsq01,rcutoff2))
396 r01 = _mm256_mul_ps(rsq01,rinv01);
398 /* EWALD ELECTROSTATICS */
400 /* Analytical PME correction */
401 zeta2 = _mm256_mul_ps(beta2,rsq01);
402 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
403 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
404 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
405 felec = _mm256_mul_ps(qq01,felec);
406 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
407 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
408 velec = _mm256_sub_ps(rinv01,pmecorrV);
409 velec = _mm256_mul_ps(qq01,velec);
411 d = _mm256_sub_ps(r01,rswitch);
412 d = _mm256_max_ps(d,_mm256_setzero_ps());
413 d2 = _mm256_mul_ps(d,d);
414 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
416 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
418 /* Evaluate switch function */
419 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
420 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
421 velec = _mm256_mul_ps(velec,sw);
422 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
424 /* Update potential sum for this i atom from the interaction with this j atom. */
425 velec = _mm256_and_ps(velec,cutoff_mask);
426 velecsum = _mm256_add_ps(velecsum,velec);
430 fscal = _mm256_and_ps(fscal,cutoff_mask);
432 /* Calculate temporary vectorial force */
433 tx = _mm256_mul_ps(fscal,dx01);
434 ty = _mm256_mul_ps(fscal,dy01);
435 tz = _mm256_mul_ps(fscal,dz01);
437 /* Update vectorial force */
438 fix0 = _mm256_add_ps(fix0,tx);
439 fiy0 = _mm256_add_ps(fiy0,ty);
440 fiz0 = _mm256_add_ps(fiz0,tz);
442 fjx1 = _mm256_add_ps(fjx1,tx);
443 fjy1 = _mm256_add_ps(fjy1,ty);
444 fjz1 = _mm256_add_ps(fjz1,tz);
448 /**************************
449 * CALCULATE INTERACTIONS *
450 **************************/
452 if (gmx_mm256_any_lt(rsq02,rcutoff2))
455 r02 = _mm256_mul_ps(rsq02,rinv02);
457 /* EWALD ELECTROSTATICS */
459 /* Analytical PME correction */
460 zeta2 = _mm256_mul_ps(beta2,rsq02);
461 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
462 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
463 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
464 felec = _mm256_mul_ps(qq02,felec);
465 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
466 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
467 velec = _mm256_sub_ps(rinv02,pmecorrV);
468 velec = _mm256_mul_ps(qq02,velec);
470 d = _mm256_sub_ps(r02,rswitch);
471 d = _mm256_max_ps(d,_mm256_setzero_ps());
472 d2 = _mm256_mul_ps(d,d);
473 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
475 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
477 /* Evaluate switch function */
478 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
479 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
480 velec = _mm256_mul_ps(velec,sw);
481 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
483 /* Update potential sum for this i atom from the interaction with this j atom. */
484 velec = _mm256_and_ps(velec,cutoff_mask);
485 velecsum = _mm256_add_ps(velecsum,velec);
489 fscal = _mm256_and_ps(fscal,cutoff_mask);
491 /* Calculate temporary vectorial force */
492 tx = _mm256_mul_ps(fscal,dx02);
493 ty = _mm256_mul_ps(fscal,dy02);
494 tz = _mm256_mul_ps(fscal,dz02);
496 /* Update vectorial force */
497 fix0 = _mm256_add_ps(fix0,tx);
498 fiy0 = _mm256_add_ps(fiy0,ty);
499 fiz0 = _mm256_add_ps(fiz0,tz);
501 fjx2 = _mm256_add_ps(fjx2,tx);
502 fjy2 = _mm256_add_ps(fjy2,ty);
503 fjz2 = _mm256_add_ps(fjz2,tz);
507 /**************************
508 * CALCULATE INTERACTIONS *
509 **************************/
511 if (gmx_mm256_any_lt(rsq10,rcutoff2))
514 r10 = _mm256_mul_ps(rsq10,rinv10);
516 /* EWALD ELECTROSTATICS */
518 /* Analytical PME correction */
519 zeta2 = _mm256_mul_ps(beta2,rsq10);
520 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
521 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
522 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
523 felec = _mm256_mul_ps(qq10,felec);
524 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
525 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
526 velec = _mm256_sub_ps(rinv10,pmecorrV);
527 velec = _mm256_mul_ps(qq10,velec);
529 d = _mm256_sub_ps(r10,rswitch);
530 d = _mm256_max_ps(d,_mm256_setzero_ps());
531 d2 = _mm256_mul_ps(d,d);
532 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
534 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
536 /* Evaluate switch function */
537 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
538 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
539 velec = _mm256_mul_ps(velec,sw);
540 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
542 /* Update potential sum for this i atom from the interaction with this j atom. */
543 velec = _mm256_and_ps(velec,cutoff_mask);
544 velecsum = _mm256_add_ps(velecsum,velec);
548 fscal = _mm256_and_ps(fscal,cutoff_mask);
550 /* Calculate temporary vectorial force */
551 tx = _mm256_mul_ps(fscal,dx10);
552 ty = _mm256_mul_ps(fscal,dy10);
553 tz = _mm256_mul_ps(fscal,dz10);
555 /* Update vectorial force */
556 fix1 = _mm256_add_ps(fix1,tx);
557 fiy1 = _mm256_add_ps(fiy1,ty);
558 fiz1 = _mm256_add_ps(fiz1,tz);
560 fjx0 = _mm256_add_ps(fjx0,tx);
561 fjy0 = _mm256_add_ps(fjy0,ty);
562 fjz0 = _mm256_add_ps(fjz0,tz);
566 /**************************
567 * CALCULATE INTERACTIONS *
568 **************************/
570 if (gmx_mm256_any_lt(rsq11,rcutoff2))
573 r11 = _mm256_mul_ps(rsq11,rinv11);
575 /* EWALD ELECTROSTATICS */
577 /* Analytical PME correction */
578 zeta2 = _mm256_mul_ps(beta2,rsq11);
579 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
580 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
581 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
582 felec = _mm256_mul_ps(qq11,felec);
583 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
584 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
585 velec = _mm256_sub_ps(rinv11,pmecorrV);
586 velec = _mm256_mul_ps(qq11,velec);
588 d = _mm256_sub_ps(r11,rswitch);
589 d = _mm256_max_ps(d,_mm256_setzero_ps());
590 d2 = _mm256_mul_ps(d,d);
591 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
593 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
595 /* Evaluate switch function */
596 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
597 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
598 velec = _mm256_mul_ps(velec,sw);
599 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
601 /* Update potential sum for this i atom from the interaction with this j atom. */
602 velec = _mm256_and_ps(velec,cutoff_mask);
603 velecsum = _mm256_add_ps(velecsum,velec);
607 fscal = _mm256_and_ps(fscal,cutoff_mask);
609 /* Calculate temporary vectorial force */
610 tx = _mm256_mul_ps(fscal,dx11);
611 ty = _mm256_mul_ps(fscal,dy11);
612 tz = _mm256_mul_ps(fscal,dz11);
614 /* Update vectorial force */
615 fix1 = _mm256_add_ps(fix1,tx);
616 fiy1 = _mm256_add_ps(fiy1,ty);
617 fiz1 = _mm256_add_ps(fiz1,tz);
619 fjx1 = _mm256_add_ps(fjx1,tx);
620 fjy1 = _mm256_add_ps(fjy1,ty);
621 fjz1 = _mm256_add_ps(fjz1,tz);
625 /**************************
626 * CALCULATE INTERACTIONS *
627 **************************/
629 if (gmx_mm256_any_lt(rsq12,rcutoff2))
632 r12 = _mm256_mul_ps(rsq12,rinv12);
634 /* EWALD ELECTROSTATICS */
636 /* Analytical PME correction */
637 zeta2 = _mm256_mul_ps(beta2,rsq12);
638 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
639 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
640 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
641 felec = _mm256_mul_ps(qq12,felec);
642 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
643 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
644 velec = _mm256_sub_ps(rinv12,pmecorrV);
645 velec = _mm256_mul_ps(qq12,velec);
647 d = _mm256_sub_ps(r12,rswitch);
648 d = _mm256_max_ps(d,_mm256_setzero_ps());
649 d2 = _mm256_mul_ps(d,d);
650 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
652 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
654 /* Evaluate switch function */
655 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
656 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
657 velec = _mm256_mul_ps(velec,sw);
658 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
660 /* Update potential sum for this i atom from the interaction with this j atom. */
661 velec = _mm256_and_ps(velec,cutoff_mask);
662 velecsum = _mm256_add_ps(velecsum,velec);
666 fscal = _mm256_and_ps(fscal,cutoff_mask);
668 /* Calculate temporary vectorial force */
669 tx = _mm256_mul_ps(fscal,dx12);
670 ty = _mm256_mul_ps(fscal,dy12);
671 tz = _mm256_mul_ps(fscal,dz12);
673 /* Update vectorial force */
674 fix1 = _mm256_add_ps(fix1,tx);
675 fiy1 = _mm256_add_ps(fiy1,ty);
676 fiz1 = _mm256_add_ps(fiz1,tz);
678 fjx2 = _mm256_add_ps(fjx2,tx);
679 fjy2 = _mm256_add_ps(fjy2,ty);
680 fjz2 = _mm256_add_ps(fjz2,tz);
684 /**************************
685 * CALCULATE INTERACTIONS *
686 **************************/
688 if (gmx_mm256_any_lt(rsq20,rcutoff2))
691 r20 = _mm256_mul_ps(rsq20,rinv20);
693 /* EWALD ELECTROSTATICS */
695 /* Analytical PME correction */
696 zeta2 = _mm256_mul_ps(beta2,rsq20);
697 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
698 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
699 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
700 felec = _mm256_mul_ps(qq20,felec);
701 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
702 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
703 velec = _mm256_sub_ps(rinv20,pmecorrV);
704 velec = _mm256_mul_ps(qq20,velec);
706 d = _mm256_sub_ps(r20,rswitch);
707 d = _mm256_max_ps(d,_mm256_setzero_ps());
708 d2 = _mm256_mul_ps(d,d);
709 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
711 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
713 /* Evaluate switch function */
714 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
715 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
716 velec = _mm256_mul_ps(velec,sw);
717 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
719 /* Update potential sum for this i atom from the interaction with this j atom. */
720 velec = _mm256_and_ps(velec,cutoff_mask);
721 velecsum = _mm256_add_ps(velecsum,velec);
725 fscal = _mm256_and_ps(fscal,cutoff_mask);
727 /* Calculate temporary vectorial force */
728 tx = _mm256_mul_ps(fscal,dx20);
729 ty = _mm256_mul_ps(fscal,dy20);
730 tz = _mm256_mul_ps(fscal,dz20);
732 /* Update vectorial force */
733 fix2 = _mm256_add_ps(fix2,tx);
734 fiy2 = _mm256_add_ps(fiy2,ty);
735 fiz2 = _mm256_add_ps(fiz2,tz);
737 fjx0 = _mm256_add_ps(fjx0,tx);
738 fjy0 = _mm256_add_ps(fjy0,ty);
739 fjz0 = _mm256_add_ps(fjz0,tz);
743 /**************************
744 * CALCULATE INTERACTIONS *
745 **************************/
747 if (gmx_mm256_any_lt(rsq21,rcutoff2))
750 r21 = _mm256_mul_ps(rsq21,rinv21);
752 /* EWALD ELECTROSTATICS */
754 /* Analytical PME correction */
755 zeta2 = _mm256_mul_ps(beta2,rsq21);
756 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
757 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
758 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
759 felec = _mm256_mul_ps(qq21,felec);
760 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
761 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
762 velec = _mm256_sub_ps(rinv21,pmecorrV);
763 velec = _mm256_mul_ps(qq21,velec);
765 d = _mm256_sub_ps(r21,rswitch);
766 d = _mm256_max_ps(d,_mm256_setzero_ps());
767 d2 = _mm256_mul_ps(d,d);
768 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
770 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
772 /* Evaluate switch function */
773 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
774 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
775 velec = _mm256_mul_ps(velec,sw);
776 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
778 /* Update potential sum for this i atom from the interaction with this j atom. */
779 velec = _mm256_and_ps(velec,cutoff_mask);
780 velecsum = _mm256_add_ps(velecsum,velec);
784 fscal = _mm256_and_ps(fscal,cutoff_mask);
786 /* Calculate temporary vectorial force */
787 tx = _mm256_mul_ps(fscal,dx21);
788 ty = _mm256_mul_ps(fscal,dy21);
789 tz = _mm256_mul_ps(fscal,dz21);
791 /* Update vectorial force */
792 fix2 = _mm256_add_ps(fix2,tx);
793 fiy2 = _mm256_add_ps(fiy2,ty);
794 fiz2 = _mm256_add_ps(fiz2,tz);
796 fjx1 = _mm256_add_ps(fjx1,tx);
797 fjy1 = _mm256_add_ps(fjy1,ty);
798 fjz1 = _mm256_add_ps(fjz1,tz);
802 /**************************
803 * CALCULATE INTERACTIONS *
804 **************************/
806 if (gmx_mm256_any_lt(rsq22,rcutoff2))
809 r22 = _mm256_mul_ps(rsq22,rinv22);
811 /* EWALD ELECTROSTATICS */
813 /* Analytical PME correction */
814 zeta2 = _mm256_mul_ps(beta2,rsq22);
815 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
816 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
817 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
818 felec = _mm256_mul_ps(qq22,felec);
819 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
820 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
821 velec = _mm256_sub_ps(rinv22,pmecorrV);
822 velec = _mm256_mul_ps(qq22,velec);
824 d = _mm256_sub_ps(r22,rswitch);
825 d = _mm256_max_ps(d,_mm256_setzero_ps());
826 d2 = _mm256_mul_ps(d,d);
827 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
829 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
831 /* Evaluate switch function */
832 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
833 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
834 velec = _mm256_mul_ps(velec,sw);
835 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
837 /* Update potential sum for this i atom from the interaction with this j atom. */
838 velec = _mm256_and_ps(velec,cutoff_mask);
839 velecsum = _mm256_add_ps(velecsum,velec);
843 fscal = _mm256_and_ps(fscal,cutoff_mask);
845 /* Calculate temporary vectorial force */
846 tx = _mm256_mul_ps(fscal,dx22);
847 ty = _mm256_mul_ps(fscal,dy22);
848 tz = _mm256_mul_ps(fscal,dz22);
850 /* Update vectorial force */
851 fix2 = _mm256_add_ps(fix2,tx);
852 fiy2 = _mm256_add_ps(fiy2,ty);
853 fiz2 = _mm256_add_ps(fiz2,tz);
855 fjx2 = _mm256_add_ps(fjx2,tx);
856 fjy2 = _mm256_add_ps(fjy2,ty);
857 fjz2 = _mm256_add_ps(fjz2,tz);
861 fjptrA = f+j_coord_offsetA;
862 fjptrB = f+j_coord_offsetB;
863 fjptrC = f+j_coord_offsetC;
864 fjptrD = f+j_coord_offsetD;
865 fjptrE = f+j_coord_offsetE;
866 fjptrF = f+j_coord_offsetF;
867 fjptrG = f+j_coord_offsetG;
868 fjptrH = f+j_coord_offsetH;
870 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
871 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
873 /* Inner loop uses 972 flops */
879 /* Get j neighbor index, and coordinate index */
880 jnrlistA = jjnr[jidx];
881 jnrlistB = jjnr[jidx+1];
882 jnrlistC = jjnr[jidx+2];
883 jnrlistD = jjnr[jidx+3];
884 jnrlistE = jjnr[jidx+4];
885 jnrlistF = jjnr[jidx+5];
886 jnrlistG = jjnr[jidx+6];
887 jnrlistH = jjnr[jidx+7];
888 /* Sign of each element will be negative for non-real atoms.
889 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
890 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
892 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
893 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
895 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
896 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
897 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
898 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
899 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
900 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
901 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
902 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
903 j_coord_offsetA = DIM*jnrA;
904 j_coord_offsetB = DIM*jnrB;
905 j_coord_offsetC = DIM*jnrC;
906 j_coord_offsetD = DIM*jnrD;
907 j_coord_offsetE = DIM*jnrE;
908 j_coord_offsetF = DIM*jnrF;
909 j_coord_offsetG = DIM*jnrG;
910 j_coord_offsetH = DIM*jnrH;
912 /* load j atom coordinates */
913 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
914 x+j_coord_offsetC,x+j_coord_offsetD,
915 x+j_coord_offsetE,x+j_coord_offsetF,
916 x+j_coord_offsetG,x+j_coord_offsetH,
917 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
919 /* Calculate displacement vector */
920 dx00 = _mm256_sub_ps(ix0,jx0);
921 dy00 = _mm256_sub_ps(iy0,jy0);
922 dz00 = _mm256_sub_ps(iz0,jz0);
923 dx01 = _mm256_sub_ps(ix0,jx1);
924 dy01 = _mm256_sub_ps(iy0,jy1);
925 dz01 = _mm256_sub_ps(iz0,jz1);
926 dx02 = _mm256_sub_ps(ix0,jx2);
927 dy02 = _mm256_sub_ps(iy0,jy2);
928 dz02 = _mm256_sub_ps(iz0,jz2);
929 dx10 = _mm256_sub_ps(ix1,jx0);
930 dy10 = _mm256_sub_ps(iy1,jy0);
931 dz10 = _mm256_sub_ps(iz1,jz0);
932 dx11 = _mm256_sub_ps(ix1,jx1);
933 dy11 = _mm256_sub_ps(iy1,jy1);
934 dz11 = _mm256_sub_ps(iz1,jz1);
935 dx12 = _mm256_sub_ps(ix1,jx2);
936 dy12 = _mm256_sub_ps(iy1,jy2);
937 dz12 = _mm256_sub_ps(iz1,jz2);
938 dx20 = _mm256_sub_ps(ix2,jx0);
939 dy20 = _mm256_sub_ps(iy2,jy0);
940 dz20 = _mm256_sub_ps(iz2,jz0);
941 dx21 = _mm256_sub_ps(ix2,jx1);
942 dy21 = _mm256_sub_ps(iy2,jy1);
943 dz21 = _mm256_sub_ps(iz2,jz1);
944 dx22 = _mm256_sub_ps(ix2,jx2);
945 dy22 = _mm256_sub_ps(iy2,jy2);
946 dz22 = _mm256_sub_ps(iz2,jz2);
948 /* Calculate squared distance and things based on it */
949 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
950 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
951 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
952 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
953 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
954 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
955 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
956 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
957 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
959 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
960 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
961 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
962 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
963 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
964 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
965 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
966 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
967 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
969 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
970 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
971 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
972 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
973 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
974 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
975 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
976 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
977 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
979 fjx0 = _mm256_setzero_ps();
980 fjy0 = _mm256_setzero_ps();
981 fjz0 = _mm256_setzero_ps();
982 fjx1 = _mm256_setzero_ps();
983 fjy1 = _mm256_setzero_ps();
984 fjz1 = _mm256_setzero_ps();
985 fjx2 = _mm256_setzero_ps();
986 fjy2 = _mm256_setzero_ps();
987 fjz2 = _mm256_setzero_ps();
989 /**************************
990 * CALCULATE INTERACTIONS *
991 **************************/
993 if (gmx_mm256_any_lt(rsq00,rcutoff2))
996 r00 = _mm256_mul_ps(rsq00,rinv00);
997 r00 = _mm256_andnot_ps(dummy_mask,r00);
999 /* EWALD ELECTROSTATICS */
1001 /* Analytical PME correction */
1002 zeta2 = _mm256_mul_ps(beta2,rsq00);
1003 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1004 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1005 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1006 felec = _mm256_mul_ps(qq00,felec);
1007 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1008 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1009 velec = _mm256_sub_ps(rinv00,pmecorrV);
1010 velec = _mm256_mul_ps(qq00,velec);
1012 d = _mm256_sub_ps(r00,rswitch);
1013 d = _mm256_max_ps(d,_mm256_setzero_ps());
1014 d2 = _mm256_mul_ps(d,d);
1015 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1017 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1019 /* Evaluate switch function */
1020 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1021 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1022 velec = _mm256_mul_ps(velec,sw);
1023 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1025 /* Update potential sum for this i atom from the interaction with this j atom. */
1026 velec = _mm256_and_ps(velec,cutoff_mask);
1027 velec = _mm256_andnot_ps(dummy_mask,velec);
1028 velecsum = _mm256_add_ps(velecsum,velec);
1032 fscal = _mm256_and_ps(fscal,cutoff_mask);
1034 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1036 /* Calculate temporary vectorial force */
1037 tx = _mm256_mul_ps(fscal,dx00);
1038 ty = _mm256_mul_ps(fscal,dy00);
1039 tz = _mm256_mul_ps(fscal,dz00);
1041 /* Update vectorial force */
1042 fix0 = _mm256_add_ps(fix0,tx);
1043 fiy0 = _mm256_add_ps(fiy0,ty);
1044 fiz0 = _mm256_add_ps(fiz0,tz);
1046 fjx0 = _mm256_add_ps(fjx0,tx);
1047 fjy0 = _mm256_add_ps(fjy0,ty);
1048 fjz0 = _mm256_add_ps(fjz0,tz);
1052 /**************************
1053 * CALCULATE INTERACTIONS *
1054 **************************/
1056 if (gmx_mm256_any_lt(rsq01,rcutoff2))
1059 r01 = _mm256_mul_ps(rsq01,rinv01);
1060 r01 = _mm256_andnot_ps(dummy_mask,r01);
1062 /* EWALD ELECTROSTATICS */
1064 /* Analytical PME correction */
1065 zeta2 = _mm256_mul_ps(beta2,rsq01);
1066 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1067 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1068 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1069 felec = _mm256_mul_ps(qq01,felec);
1070 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1071 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1072 velec = _mm256_sub_ps(rinv01,pmecorrV);
1073 velec = _mm256_mul_ps(qq01,velec);
1075 d = _mm256_sub_ps(r01,rswitch);
1076 d = _mm256_max_ps(d,_mm256_setzero_ps());
1077 d2 = _mm256_mul_ps(d,d);
1078 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1080 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1082 /* Evaluate switch function */
1083 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1084 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
1085 velec = _mm256_mul_ps(velec,sw);
1086 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
1088 /* Update potential sum for this i atom from the interaction with this j atom. */
1089 velec = _mm256_and_ps(velec,cutoff_mask);
1090 velec = _mm256_andnot_ps(dummy_mask,velec);
1091 velecsum = _mm256_add_ps(velecsum,velec);
1095 fscal = _mm256_and_ps(fscal,cutoff_mask);
1097 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1099 /* Calculate temporary vectorial force */
1100 tx = _mm256_mul_ps(fscal,dx01);
1101 ty = _mm256_mul_ps(fscal,dy01);
1102 tz = _mm256_mul_ps(fscal,dz01);
1104 /* Update vectorial force */
1105 fix0 = _mm256_add_ps(fix0,tx);
1106 fiy0 = _mm256_add_ps(fiy0,ty);
1107 fiz0 = _mm256_add_ps(fiz0,tz);
1109 fjx1 = _mm256_add_ps(fjx1,tx);
1110 fjy1 = _mm256_add_ps(fjy1,ty);
1111 fjz1 = _mm256_add_ps(fjz1,tz);
1115 /**************************
1116 * CALCULATE INTERACTIONS *
1117 **************************/
1119 if (gmx_mm256_any_lt(rsq02,rcutoff2))
1122 r02 = _mm256_mul_ps(rsq02,rinv02);
1123 r02 = _mm256_andnot_ps(dummy_mask,r02);
1125 /* EWALD ELECTROSTATICS */
1127 /* Analytical PME correction */
1128 zeta2 = _mm256_mul_ps(beta2,rsq02);
1129 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
1130 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1131 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1132 felec = _mm256_mul_ps(qq02,felec);
1133 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1134 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1135 velec = _mm256_sub_ps(rinv02,pmecorrV);
1136 velec = _mm256_mul_ps(qq02,velec);
1138 d = _mm256_sub_ps(r02,rswitch);
1139 d = _mm256_max_ps(d,_mm256_setzero_ps());
1140 d2 = _mm256_mul_ps(d,d);
1141 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1143 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1145 /* Evaluate switch function */
1146 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1147 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
1148 velec = _mm256_mul_ps(velec,sw);
1149 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
1151 /* Update potential sum for this i atom from the interaction with this j atom. */
1152 velec = _mm256_and_ps(velec,cutoff_mask);
1153 velec = _mm256_andnot_ps(dummy_mask,velec);
1154 velecsum = _mm256_add_ps(velecsum,velec);
1158 fscal = _mm256_and_ps(fscal,cutoff_mask);
1160 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1162 /* Calculate temporary vectorial force */
1163 tx = _mm256_mul_ps(fscal,dx02);
1164 ty = _mm256_mul_ps(fscal,dy02);
1165 tz = _mm256_mul_ps(fscal,dz02);
1167 /* Update vectorial force */
1168 fix0 = _mm256_add_ps(fix0,tx);
1169 fiy0 = _mm256_add_ps(fiy0,ty);
1170 fiz0 = _mm256_add_ps(fiz0,tz);
1172 fjx2 = _mm256_add_ps(fjx2,tx);
1173 fjy2 = _mm256_add_ps(fjy2,ty);
1174 fjz2 = _mm256_add_ps(fjz2,tz);
1178 /**************************
1179 * CALCULATE INTERACTIONS *
1180 **************************/
1182 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1185 r10 = _mm256_mul_ps(rsq10,rinv10);
1186 r10 = _mm256_andnot_ps(dummy_mask,r10);
1188 /* EWALD ELECTROSTATICS */
1190 /* Analytical PME correction */
1191 zeta2 = _mm256_mul_ps(beta2,rsq10);
1192 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
1193 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1194 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1195 felec = _mm256_mul_ps(qq10,felec);
1196 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1197 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1198 velec = _mm256_sub_ps(rinv10,pmecorrV);
1199 velec = _mm256_mul_ps(qq10,velec);
1201 d = _mm256_sub_ps(r10,rswitch);
1202 d = _mm256_max_ps(d,_mm256_setzero_ps());
1203 d2 = _mm256_mul_ps(d,d);
1204 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1206 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1208 /* Evaluate switch function */
1209 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1210 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
1211 velec = _mm256_mul_ps(velec,sw);
1212 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1214 /* Update potential sum for this i atom from the interaction with this j atom. */
1215 velec = _mm256_and_ps(velec,cutoff_mask);
1216 velec = _mm256_andnot_ps(dummy_mask,velec);
1217 velecsum = _mm256_add_ps(velecsum,velec);
1221 fscal = _mm256_and_ps(fscal,cutoff_mask);
1223 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1225 /* Calculate temporary vectorial force */
1226 tx = _mm256_mul_ps(fscal,dx10);
1227 ty = _mm256_mul_ps(fscal,dy10);
1228 tz = _mm256_mul_ps(fscal,dz10);
1230 /* Update vectorial force */
1231 fix1 = _mm256_add_ps(fix1,tx);
1232 fiy1 = _mm256_add_ps(fiy1,ty);
1233 fiz1 = _mm256_add_ps(fiz1,tz);
1235 fjx0 = _mm256_add_ps(fjx0,tx);
1236 fjy0 = _mm256_add_ps(fjy0,ty);
1237 fjz0 = _mm256_add_ps(fjz0,tz);
1241 /**************************
1242 * CALCULATE INTERACTIONS *
1243 **************************/
1245 if (gmx_mm256_any_lt(rsq11,rcutoff2))
1248 r11 = _mm256_mul_ps(rsq11,rinv11);
1249 r11 = _mm256_andnot_ps(dummy_mask,r11);
1251 /* EWALD ELECTROSTATICS */
1253 /* Analytical PME correction */
1254 zeta2 = _mm256_mul_ps(beta2,rsq11);
1255 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
1256 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1257 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1258 felec = _mm256_mul_ps(qq11,felec);
1259 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1260 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1261 velec = _mm256_sub_ps(rinv11,pmecorrV);
1262 velec = _mm256_mul_ps(qq11,velec);
1264 d = _mm256_sub_ps(r11,rswitch);
1265 d = _mm256_max_ps(d,_mm256_setzero_ps());
1266 d2 = _mm256_mul_ps(d,d);
1267 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1269 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1271 /* Evaluate switch function */
1272 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1273 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
1274 velec = _mm256_mul_ps(velec,sw);
1275 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1277 /* Update potential sum for this i atom from the interaction with this j atom. */
1278 velec = _mm256_and_ps(velec,cutoff_mask);
1279 velec = _mm256_andnot_ps(dummy_mask,velec);
1280 velecsum = _mm256_add_ps(velecsum,velec);
1284 fscal = _mm256_and_ps(fscal,cutoff_mask);
1286 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1288 /* Calculate temporary vectorial force */
1289 tx = _mm256_mul_ps(fscal,dx11);
1290 ty = _mm256_mul_ps(fscal,dy11);
1291 tz = _mm256_mul_ps(fscal,dz11);
1293 /* Update vectorial force */
1294 fix1 = _mm256_add_ps(fix1,tx);
1295 fiy1 = _mm256_add_ps(fiy1,ty);
1296 fiz1 = _mm256_add_ps(fiz1,tz);
1298 fjx1 = _mm256_add_ps(fjx1,tx);
1299 fjy1 = _mm256_add_ps(fjy1,ty);
1300 fjz1 = _mm256_add_ps(fjz1,tz);
1304 /**************************
1305 * CALCULATE INTERACTIONS *
1306 **************************/
1308 if (gmx_mm256_any_lt(rsq12,rcutoff2))
1311 r12 = _mm256_mul_ps(rsq12,rinv12);
1312 r12 = _mm256_andnot_ps(dummy_mask,r12);
1314 /* EWALD ELECTROSTATICS */
1316 /* Analytical PME correction */
1317 zeta2 = _mm256_mul_ps(beta2,rsq12);
1318 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
1319 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1320 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1321 felec = _mm256_mul_ps(qq12,felec);
1322 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1323 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1324 velec = _mm256_sub_ps(rinv12,pmecorrV);
1325 velec = _mm256_mul_ps(qq12,velec);
1327 d = _mm256_sub_ps(r12,rswitch);
1328 d = _mm256_max_ps(d,_mm256_setzero_ps());
1329 d2 = _mm256_mul_ps(d,d);
1330 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1332 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1334 /* Evaluate switch function */
1335 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1336 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
1337 velec = _mm256_mul_ps(velec,sw);
1338 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1340 /* Update potential sum for this i atom from the interaction with this j atom. */
1341 velec = _mm256_and_ps(velec,cutoff_mask);
1342 velec = _mm256_andnot_ps(dummy_mask,velec);
1343 velecsum = _mm256_add_ps(velecsum,velec);
1347 fscal = _mm256_and_ps(fscal,cutoff_mask);
1349 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1351 /* Calculate temporary vectorial force */
1352 tx = _mm256_mul_ps(fscal,dx12);
1353 ty = _mm256_mul_ps(fscal,dy12);
1354 tz = _mm256_mul_ps(fscal,dz12);
1356 /* Update vectorial force */
1357 fix1 = _mm256_add_ps(fix1,tx);
1358 fiy1 = _mm256_add_ps(fiy1,ty);
1359 fiz1 = _mm256_add_ps(fiz1,tz);
1361 fjx2 = _mm256_add_ps(fjx2,tx);
1362 fjy2 = _mm256_add_ps(fjy2,ty);
1363 fjz2 = _mm256_add_ps(fjz2,tz);
1367 /**************************
1368 * CALCULATE INTERACTIONS *
1369 **************************/
1371 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1374 r20 = _mm256_mul_ps(rsq20,rinv20);
1375 r20 = _mm256_andnot_ps(dummy_mask,r20);
1377 /* EWALD ELECTROSTATICS */
1379 /* Analytical PME correction */
1380 zeta2 = _mm256_mul_ps(beta2,rsq20);
1381 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1382 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1383 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1384 felec = _mm256_mul_ps(qq20,felec);
1385 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1386 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1387 velec = _mm256_sub_ps(rinv20,pmecorrV);
1388 velec = _mm256_mul_ps(qq20,velec);
1390 d = _mm256_sub_ps(r20,rswitch);
1391 d = _mm256_max_ps(d,_mm256_setzero_ps());
1392 d2 = _mm256_mul_ps(d,d);
1393 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1395 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1397 /* Evaluate switch function */
1398 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1399 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
1400 velec = _mm256_mul_ps(velec,sw);
1401 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1403 /* Update potential sum for this i atom from the interaction with this j atom. */
1404 velec = _mm256_and_ps(velec,cutoff_mask);
1405 velec = _mm256_andnot_ps(dummy_mask,velec);
1406 velecsum = _mm256_add_ps(velecsum,velec);
1410 fscal = _mm256_and_ps(fscal,cutoff_mask);
1412 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1414 /* Calculate temporary vectorial force */
1415 tx = _mm256_mul_ps(fscal,dx20);
1416 ty = _mm256_mul_ps(fscal,dy20);
1417 tz = _mm256_mul_ps(fscal,dz20);
1419 /* Update vectorial force */
1420 fix2 = _mm256_add_ps(fix2,tx);
1421 fiy2 = _mm256_add_ps(fiy2,ty);
1422 fiz2 = _mm256_add_ps(fiz2,tz);
1424 fjx0 = _mm256_add_ps(fjx0,tx);
1425 fjy0 = _mm256_add_ps(fjy0,ty);
1426 fjz0 = _mm256_add_ps(fjz0,tz);
1430 /**************************
1431 * CALCULATE INTERACTIONS *
1432 **************************/
1434 if (gmx_mm256_any_lt(rsq21,rcutoff2))
1437 r21 = _mm256_mul_ps(rsq21,rinv21);
1438 r21 = _mm256_andnot_ps(dummy_mask,r21);
1440 /* EWALD ELECTROSTATICS */
1442 /* Analytical PME correction */
1443 zeta2 = _mm256_mul_ps(beta2,rsq21);
1444 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
1445 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1446 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1447 felec = _mm256_mul_ps(qq21,felec);
1448 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1449 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1450 velec = _mm256_sub_ps(rinv21,pmecorrV);
1451 velec = _mm256_mul_ps(qq21,velec);
1453 d = _mm256_sub_ps(r21,rswitch);
1454 d = _mm256_max_ps(d,_mm256_setzero_ps());
1455 d2 = _mm256_mul_ps(d,d);
1456 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1458 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1460 /* Evaluate switch function */
1461 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1462 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
1463 velec = _mm256_mul_ps(velec,sw);
1464 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1466 /* Update potential sum for this i atom from the interaction with this j atom. */
1467 velec = _mm256_and_ps(velec,cutoff_mask);
1468 velec = _mm256_andnot_ps(dummy_mask,velec);
1469 velecsum = _mm256_add_ps(velecsum,velec);
1473 fscal = _mm256_and_ps(fscal,cutoff_mask);
1475 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1477 /* Calculate temporary vectorial force */
1478 tx = _mm256_mul_ps(fscal,dx21);
1479 ty = _mm256_mul_ps(fscal,dy21);
1480 tz = _mm256_mul_ps(fscal,dz21);
1482 /* Update vectorial force */
1483 fix2 = _mm256_add_ps(fix2,tx);
1484 fiy2 = _mm256_add_ps(fiy2,ty);
1485 fiz2 = _mm256_add_ps(fiz2,tz);
1487 fjx1 = _mm256_add_ps(fjx1,tx);
1488 fjy1 = _mm256_add_ps(fjy1,ty);
1489 fjz1 = _mm256_add_ps(fjz1,tz);
1493 /**************************
1494 * CALCULATE INTERACTIONS *
1495 **************************/
1497 if (gmx_mm256_any_lt(rsq22,rcutoff2))
1500 r22 = _mm256_mul_ps(rsq22,rinv22);
1501 r22 = _mm256_andnot_ps(dummy_mask,r22);
1503 /* EWALD ELECTROSTATICS */
1505 /* Analytical PME correction */
1506 zeta2 = _mm256_mul_ps(beta2,rsq22);
1507 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
1508 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1509 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1510 felec = _mm256_mul_ps(qq22,felec);
1511 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1512 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1513 velec = _mm256_sub_ps(rinv22,pmecorrV);
1514 velec = _mm256_mul_ps(qq22,velec);
1516 d = _mm256_sub_ps(r22,rswitch);
1517 d = _mm256_max_ps(d,_mm256_setzero_ps());
1518 d2 = _mm256_mul_ps(d,d);
1519 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1521 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1523 /* Evaluate switch function */
1524 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1525 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
1526 velec = _mm256_mul_ps(velec,sw);
1527 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1529 /* Update potential sum for this i atom from the interaction with this j atom. */
1530 velec = _mm256_and_ps(velec,cutoff_mask);
1531 velec = _mm256_andnot_ps(dummy_mask,velec);
1532 velecsum = _mm256_add_ps(velecsum,velec);
1536 fscal = _mm256_and_ps(fscal,cutoff_mask);
1538 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1540 /* Calculate temporary vectorial force */
1541 tx = _mm256_mul_ps(fscal,dx22);
1542 ty = _mm256_mul_ps(fscal,dy22);
1543 tz = _mm256_mul_ps(fscal,dz22);
1545 /* Update vectorial force */
1546 fix2 = _mm256_add_ps(fix2,tx);
1547 fiy2 = _mm256_add_ps(fiy2,ty);
1548 fiz2 = _mm256_add_ps(fiz2,tz);
1550 fjx2 = _mm256_add_ps(fjx2,tx);
1551 fjy2 = _mm256_add_ps(fjy2,ty);
1552 fjz2 = _mm256_add_ps(fjz2,tz);
1556 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1557 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1558 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1559 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1560 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1561 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1562 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1563 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1565 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1566 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1568 /* Inner loop uses 981 flops */
1571 /* End of innermost loop */
1573 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1574 f+i_coord_offset,fshift+i_shift_offset);
1577 /* Update potential energies */
1578 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1580 /* Increment number of inner iterations */
1581 inneriter += j_index_end - j_index_start;
1583 /* Outer loop uses 19 flops */
1586 /* Increment number of outer iterations */
1589 /* Update outer/inner flops */
1591 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*981);
1594 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3W3_F_avx_256_single
1595 * Electrostatics interaction: Ewald
1596 * VdW interaction: None
1597 * Geometry: Water3-Water3
1598 * Calculate force/pot: Force
1601 nb_kernel_ElecEwSw_VdwNone_GeomW3W3_F_avx_256_single
1602 (t_nblist * gmx_restrict nlist,
1603 rvec * gmx_restrict xx,
1604 rvec * gmx_restrict ff,
1605 t_forcerec * gmx_restrict fr,
1606 t_mdatoms * gmx_restrict mdatoms,
1607 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1608 t_nrnb * gmx_restrict nrnb)
1610 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1611 * just 0 for non-waters.
1612 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1613 * jnr indices corresponding to data put in the four positions in the SIMD register.
1615 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1616 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1617 int jnrA,jnrB,jnrC,jnrD;
1618 int jnrE,jnrF,jnrG,jnrH;
1619 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1620 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1621 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1622 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1623 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1624 real rcutoff_scalar;
1625 real *shiftvec,*fshift,*x,*f;
1626 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1627 real scratch[4*DIM];
1628 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1629 real * vdwioffsetptr0;
1630 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1631 real * vdwioffsetptr1;
1632 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1633 real * vdwioffsetptr2;
1634 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1635 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1636 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1637 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1638 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1639 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1640 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1641 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1642 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1643 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1644 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1645 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1646 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1647 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1648 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1649 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1650 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
1653 __m128i ewitab_lo,ewitab_hi;
1654 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1655 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1657 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1658 real rswitch_scalar,d_scalar;
1659 __m256 dummy_mask,cutoff_mask;
1660 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1661 __m256 one = _mm256_set1_ps(1.0);
1662 __m256 two = _mm256_set1_ps(2.0);
1668 jindex = nlist->jindex;
1670 shiftidx = nlist->shift;
1672 shiftvec = fr->shift_vec[0];
1673 fshift = fr->fshift[0];
1674 facel = _mm256_set1_ps(fr->epsfac);
1675 charge = mdatoms->chargeA;
1677 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
1678 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1679 beta2 = _mm256_mul_ps(beta,beta);
1680 beta3 = _mm256_mul_ps(beta,beta2);
1682 ewtab = fr->ic->tabq_coul_FDV0;
1683 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
1684 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1686 /* Setup water-specific parameters */
1687 inr = nlist->iinr[0];
1688 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1689 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1690 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1692 jq0 = _mm256_set1_ps(charge[inr+0]);
1693 jq1 = _mm256_set1_ps(charge[inr+1]);
1694 jq2 = _mm256_set1_ps(charge[inr+2]);
1695 qq00 = _mm256_mul_ps(iq0,jq0);
1696 qq01 = _mm256_mul_ps(iq0,jq1);
1697 qq02 = _mm256_mul_ps(iq0,jq2);
1698 qq10 = _mm256_mul_ps(iq1,jq0);
1699 qq11 = _mm256_mul_ps(iq1,jq1);
1700 qq12 = _mm256_mul_ps(iq1,jq2);
1701 qq20 = _mm256_mul_ps(iq2,jq0);
1702 qq21 = _mm256_mul_ps(iq2,jq1);
1703 qq22 = _mm256_mul_ps(iq2,jq2);
1705 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1706 rcutoff_scalar = fr->rcoulomb;
1707 rcutoff = _mm256_set1_ps(rcutoff_scalar);
1708 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
1710 rswitch_scalar = fr->rcoulomb_switch;
1711 rswitch = _mm256_set1_ps(rswitch_scalar);
1712 /* Setup switch parameters */
1713 d_scalar = rcutoff_scalar-rswitch_scalar;
1714 d = _mm256_set1_ps(d_scalar);
1715 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
1716 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1717 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1718 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
1719 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1720 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1722 /* Avoid stupid compiler warnings */
1723 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1724 j_coord_offsetA = 0;
1725 j_coord_offsetB = 0;
1726 j_coord_offsetC = 0;
1727 j_coord_offsetD = 0;
1728 j_coord_offsetE = 0;
1729 j_coord_offsetF = 0;
1730 j_coord_offsetG = 0;
1731 j_coord_offsetH = 0;
1736 for(iidx=0;iidx<4*DIM;iidx++)
1738 scratch[iidx] = 0.0;
1741 /* Start outer loop over neighborlists */
1742 for(iidx=0; iidx<nri; iidx++)
1744 /* Load shift vector for this list */
1745 i_shift_offset = DIM*shiftidx[iidx];
1747 /* Load limits for loop over neighbors */
1748 j_index_start = jindex[iidx];
1749 j_index_end = jindex[iidx+1];
1751 /* Get outer coordinate index */
1753 i_coord_offset = DIM*inr;
1755 /* Load i particle coords and add shift vector */
1756 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1757 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1759 fix0 = _mm256_setzero_ps();
1760 fiy0 = _mm256_setzero_ps();
1761 fiz0 = _mm256_setzero_ps();
1762 fix1 = _mm256_setzero_ps();
1763 fiy1 = _mm256_setzero_ps();
1764 fiz1 = _mm256_setzero_ps();
1765 fix2 = _mm256_setzero_ps();
1766 fiy2 = _mm256_setzero_ps();
1767 fiz2 = _mm256_setzero_ps();
1769 /* Start inner kernel loop */
1770 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1773 /* Get j neighbor index, and coordinate index */
1775 jnrB = jjnr[jidx+1];
1776 jnrC = jjnr[jidx+2];
1777 jnrD = jjnr[jidx+3];
1778 jnrE = jjnr[jidx+4];
1779 jnrF = jjnr[jidx+5];
1780 jnrG = jjnr[jidx+6];
1781 jnrH = jjnr[jidx+7];
1782 j_coord_offsetA = DIM*jnrA;
1783 j_coord_offsetB = DIM*jnrB;
1784 j_coord_offsetC = DIM*jnrC;
1785 j_coord_offsetD = DIM*jnrD;
1786 j_coord_offsetE = DIM*jnrE;
1787 j_coord_offsetF = DIM*jnrF;
1788 j_coord_offsetG = DIM*jnrG;
1789 j_coord_offsetH = DIM*jnrH;
1791 /* load j atom coordinates */
1792 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1793 x+j_coord_offsetC,x+j_coord_offsetD,
1794 x+j_coord_offsetE,x+j_coord_offsetF,
1795 x+j_coord_offsetG,x+j_coord_offsetH,
1796 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1798 /* Calculate displacement vector */
1799 dx00 = _mm256_sub_ps(ix0,jx0);
1800 dy00 = _mm256_sub_ps(iy0,jy0);
1801 dz00 = _mm256_sub_ps(iz0,jz0);
1802 dx01 = _mm256_sub_ps(ix0,jx1);
1803 dy01 = _mm256_sub_ps(iy0,jy1);
1804 dz01 = _mm256_sub_ps(iz0,jz1);
1805 dx02 = _mm256_sub_ps(ix0,jx2);
1806 dy02 = _mm256_sub_ps(iy0,jy2);
1807 dz02 = _mm256_sub_ps(iz0,jz2);
1808 dx10 = _mm256_sub_ps(ix1,jx0);
1809 dy10 = _mm256_sub_ps(iy1,jy0);
1810 dz10 = _mm256_sub_ps(iz1,jz0);
1811 dx11 = _mm256_sub_ps(ix1,jx1);
1812 dy11 = _mm256_sub_ps(iy1,jy1);
1813 dz11 = _mm256_sub_ps(iz1,jz1);
1814 dx12 = _mm256_sub_ps(ix1,jx2);
1815 dy12 = _mm256_sub_ps(iy1,jy2);
1816 dz12 = _mm256_sub_ps(iz1,jz2);
1817 dx20 = _mm256_sub_ps(ix2,jx0);
1818 dy20 = _mm256_sub_ps(iy2,jy0);
1819 dz20 = _mm256_sub_ps(iz2,jz0);
1820 dx21 = _mm256_sub_ps(ix2,jx1);
1821 dy21 = _mm256_sub_ps(iy2,jy1);
1822 dz21 = _mm256_sub_ps(iz2,jz1);
1823 dx22 = _mm256_sub_ps(ix2,jx2);
1824 dy22 = _mm256_sub_ps(iy2,jy2);
1825 dz22 = _mm256_sub_ps(iz2,jz2);
1827 /* Calculate squared distance and things based on it */
1828 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1829 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1830 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1831 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1832 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1833 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1834 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1835 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1836 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1838 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1839 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1840 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1841 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1842 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1843 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1844 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1845 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1846 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1848 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1849 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1850 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1851 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1852 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1853 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1854 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1855 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1856 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1858 fjx0 = _mm256_setzero_ps();
1859 fjy0 = _mm256_setzero_ps();
1860 fjz0 = _mm256_setzero_ps();
1861 fjx1 = _mm256_setzero_ps();
1862 fjy1 = _mm256_setzero_ps();
1863 fjz1 = _mm256_setzero_ps();
1864 fjx2 = _mm256_setzero_ps();
1865 fjy2 = _mm256_setzero_ps();
1866 fjz2 = _mm256_setzero_ps();
1868 /**************************
1869 * CALCULATE INTERACTIONS *
1870 **************************/
1872 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1875 r00 = _mm256_mul_ps(rsq00,rinv00);
1877 /* EWALD ELECTROSTATICS */
1879 /* Analytical PME correction */
1880 zeta2 = _mm256_mul_ps(beta2,rsq00);
1881 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1882 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1883 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1884 felec = _mm256_mul_ps(qq00,felec);
1885 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1886 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1887 velec = _mm256_sub_ps(rinv00,pmecorrV);
1888 velec = _mm256_mul_ps(qq00,velec);
1890 d = _mm256_sub_ps(r00,rswitch);
1891 d = _mm256_max_ps(d,_mm256_setzero_ps());
1892 d2 = _mm256_mul_ps(d,d);
1893 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1895 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1897 /* Evaluate switch function */
1898 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1899 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1900 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1904 fscal = _mm256_and_ps(fscal,cutoff_mask);
1906 /* Calculate temporary vectorial force */
1907 tx = _mm256_mul_ps(fscal,dx00);
1908 ty = _mm256_mul_ps(fscal,dy00);
1909 tz = _mm256_mul_ps(fscal,dz00);
1911 /* Update vectorial force */
1912 fix0 = _mm256_add_ps(fix0,tx);
1913 fiy0 = _mm256_add_ps(fiy0,ty);
1914 fiz0 = _mm256_add_ps(fiz0,tz);
1916 fjx0 = _mm256_add_ps(fjx0,tx);
1917 fjy0 = _mm256_add_ps(fjy0,ty);
1918 fjz0 = _mm256_add_ps(fjz0,tz);
1922 /**************************
1923 * CALCULATE INTERACTIONS *
1924 **************************/
1926 if (gmx_mm256_any_lt(rsq01,rcutoff2))
1929 r01 = _mm256_mul_ps(rsq01,rinv01);
1931 /* EWALD ELECTROSTATICS */
1933 /* Analytical PME correction */
1934 zeta2 = _mm256_mul_ps(beta2,rsq01);
1935 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1936 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1937 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1938 felec = _mm256_mul_ps(qq01,felec);
1939 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1940 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1941 velec = _mm256_sub_ps(rinv01,pmecorrV);
1942 velec = _mm256_mul_ps(qq01,velec);
1944 d = _mm256_sub_ps(r01,rswitch);
1945 d = _mm256_max_ps(d,_mm256_setzero_ps());
1946 d2 = _mm256_mul_ps(d,d);
1947 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1949 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1951 /* Evaluate switch function */
1952 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1953 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
1954 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
1958 fscal = _mm256_and_ps(fscal,cutoff_mask);
1960 /* Calculate temporary vectorial force */
1961 tx = _mm256_mul_ps(fscal,dx01);
1962 ty = _mm256_mul_ps(fscal,dy01);
1963 tz = _mm256_mul_ps(fscal,dz01);
1965 /* Update vectorial force */
1966 fix0 = _mm256_add_ps(fix0,tx);
1967 fiy0 = _mm256_add_ps(fiy0,ty);
1968 fiz0 = _mm256_add_ps(fiz0,tz);
1970 fjx1 = _mm256_add_ps(fjx1,tx);
1971 fjy1 = _mm256_add_ps(fjy1,ty);
1972 fjz1 = _mm256_add_ps(fjz1,tz);
1976 /**************************
1977 * CALCULATE INTERACTIONS *
1978 **************************/
1980 if (gmx_mm256_any_lt(rsq02,rcutoff2))
1983 r02 = _mm256_mul_ps(rsq02,rinv02);
1985 /* EWALD ELECTROSTATICS */
1987 /* Analytical PME correction */
1988 zeta2 = _mm256_mul_ps(beta2,rsq02);
1989 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
1990 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1991 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1992 felec = _mm256_mul_ps(qq02,felec);
1993 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1994 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1995 velec = _mm256_sub_ps(rinv02,pmecorrV);
1996 velec = _mm256_mul_ps(qq02,velec);
1998 d = _mm256_sub_ps(r02,rswitch);
1999 d = _mm256_max_ps(d,_mm256_setzero_ps());
2000 d2 = _mm256_mul_ps(d,d);
2001 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2003 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2005 /* Evaluate switch function */
2006 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2007 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
2008 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
2012 fscal = _mm256_and_ps(fscal,cutoff_mask);
2014 /* Calculate temporary vectorial force */
2015 tx = _mm256_mul_ps(fscal,dx02);
2016 ty = _mm256_mul_ps(fscal,dy02);
2017 tz = _mm256_mul_ps(fscal,dz02);
2019 /* Update vectorial force */
2020 fix0 = _mm256_add_ps(fix0,tx);
2021 fiy0 = _mm256_add_ps(fiy0,ty);
2022 fiz0 = _mm256_add_ps(fiz0,tz);
2024 fjx2 = _mm256_add_ps(fjx2,tx);
2025 fjy2 = _mm256_add_ps(fjy2,ty);
2026 fjz2 = _mm256_add_ps(fjz2,tz);
2030 /**************************
2031 * CALCULATE INTERACTIONS *
2032 **************************/
2034 if (gmx_mm256_any_lt(rsq10,rcutoff2))
2037 r10 = _mm256_mul_ps(rsq10,rinv10);
2039 /* EWALD ELECTROSTATICS */
2041 /* Analytical PME correction */
2042 zeta2 = _mm256_mul_ps(beta2,rsq10);
2043 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2044 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2045 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2046 felec = _mm256_mul_ps(qq10,felec);
2047 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2048 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2049 velec = _mm256_sub_ps(rinv10,pmecorrV);
2050 velec = _mm256_mul_ps(qq10,velec);
2052 d = _mm256_sub_ps(r10,rswitch);
2053 d = _mm256_max_ps(d,_mm256_setzero_ps());
2054 d2 = _mm256_mul_ps(d,d);
2055 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2057 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2059 /* Evaluate switch function */
2060 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2061 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
2062 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
2066 fscal = _mm256_and_ps(fscal,cutoff_mask);
2068 /* Calculate temporary vectorial force */
2069 tx = _mm256_mul_ps(fscal,dx10);
2070 ty = _mm256_mul_ps(fscal,dy10);
2071 tz = _mm256_mul_ps(fscal,dz10);
2073 /* Update vectorial force */
2074 fix1 = _mm256_add_ps(fix1,tx);
2075 fiy1 = _mm256_add_ps(fiy1,ty);
2076 fiz1 = _mm256_add_ps(fiz1,tz);
2078 fjx0 = _mm256_add_ps(fjx0,tx);
2079 fjy0 = _mm256_add_ps(fjy0,ty);
2080 fjz0 = _mm256_add_ps(fjz0,tz);
2084 /**************************
2085 * CALCULATE INTERACTIONS *
2086 **************************/
2088 if (gmx_mm256_any_lt(rsq11,rcutoff2))
2091 r11 = _mm256_mul_ps(rsq11,rinv11);
2093 /* EWALD ELECTROSTATICS */
2095 /* Analytical PME correction */
2096 zeta2 = _mm256_mul_ps(beta2,rsq11);
2097 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2098 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2099 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2100 felec = _mm256_mul_ps(qq11,felec);
2101 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2102 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2103 velec = _mm256_sub_ps(rinv11,pmecorrV);
2104 velec = _mm256_mul_ps(qq11,velec);
2106 d = _mm256_sub_ps(r11,rswitch);
2107 d = _mm256_max_ps(d,_mm256_setzero_ps());
2108 d2 = _mm256_mul_ps(d,d);
2109 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2111 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2113 /* Evaluate switch function */
2114 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2115 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
2116 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
2120 fscal = _mm256_and_ps(fscal,cutoff_mask);
2122 /* Calculate temporary vectorial force */
2123 tx = _mm256_mul_ps(fscal,dx11);
2124 ty = _mm256_mul_ps(fscal,dy11);
2125 tz = _mm256_mul_ps(fscal,dz11);
2127 /* Update vectorial force */
2128 fix1 = _mm256_add_ps(fix1,tx);
2129 fiy1 = _mm256_add_ps(fiy1,ty);
2130 fiz1 = _mm256_add_ps(fiz1,tz);
2132 fjx1 = _mm256_add_ps(fjx1,tx);
2133 fjy1 = _mm256_add_ps(fjy1,ty);
2134 fjz1 = _mm256_add_ps(fjz1,tz);
2138 /**************************
2139 * CALCULATE INTERACTIONS *
2140 **************************/
2142 if (gmx_mm256_any_lt(rsq12,rcutoff2))
2145 r12 = _mm256_mul_ps(rsq12,rinv12);
2147 /* EWALD ELECTROSTATICS */
2149 /* Analytical PME correction */
2150 zeta2 = _mm256_mul_ps(beta2,rsq12);
2151 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2152 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2153 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2154 felec = _mm256_mul_ps(qq12,felec);
2155 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2156 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2157 velec = _mm256_sub_ps(rinv12,pmecorrV);
2158 velec = _mm256_mul_ps(qq12,velec);
2160 d = _mm256_sub_ps(r12,rswitch);
2161 d = _mm256_max_ps(d,_mm256_setzero_ps());
2162 d2 = _mm256_mul_ps(d,d);
2163 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2165 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2167 /* Evaluate switch function */
2168 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2169 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
2170 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
2174 fscal = _mm256_and_ps(fscal,cutoff_mask);
2176 /* Calculate temporary vectorial force */
2177 tx = _mm256_mul_ps(fscal,dx12);
2178 ty = _mm256_mul_ps(fscal,dy12);
2179 tz = _mm256_mul_ps(fscal,dz12);
2181 /* Update vectorial force */
2182 fix1 = _mm256_add_ps(fix1,tx);
2183 fiy1 = _mm256_add_ps(fiy1,ty);
2184 fiz1 = _mm256_add_ps(fiz1,tz);
2186 fjx2 = _mm256_add_ps(fjx2,tx);
2187 fjy2 = _mm256_add_ps(fjy2,ty);
2188 fjz2 = _mm256_add_ps(fjz2,tz);
2192 /**************************
2193 * CALCULATE INTERACTIONS *
2194 **************************/
2196 if (gmx_mm256_any_lt(rsq20,rcutoff2))
2199 r20 = _mm256_mul_ps(rsq20,rinv20);
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);
2209 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2210 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2211 velec = _mm256_sub_ps(rinv20,pmecorrV);
2212 velec = _mm256_mul_ps(qq20,velec);
2214 d = _mm256_sub_ps(r20,rswitch);
2215 d = _mm256_max_ps(d,_mm256_setzero_ps());
2216 d2 = _mm256_mul_ps(d,d);
2217 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2219 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2221 /* Evaluate switch function */
2222 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2223 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
2224 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
2228 fscal = _mm256_and_ps(fscal,cutoff_mask);
2230 /* Calculate temporary vectorial force */
2231 tx = _mm256_mul_ps(fscal,dx20);
2232 ty = _mm256_mul_ps(fscal,dy20);
2233 tz = _mm256_mul_ps(fscal,dz20);
2235 /* Update vectorial force */
2236 fix2 = _mm256_add_ps(fix2,tx);
2237 fiy2 = _mm256_add_ps(fiy2,ty);
2238 fiz2 = _mm256_add_ps(fiz2,tz);
2240 fjx0 = _mm256_add_ps(fjx0,tx);
2241 fjy0 = _mm256_add_ps(fjy0,ty);
2242 fjz0 = _mm256_add_ps(fjz0,tz);
2246 /**************************
2247 * CALCULATE INTERACTIONS *
2248 **************************/
2250 if (gmx_mm256_any_lt(rsq21,rcutoff2))
2253 r21 = _mm256_mul_ps(rsq21,rinv21);
2255 /* EWALD ELECTROSTATICS */
2257 /* Analytical PME correction */
2258 zeta2 = _mm256_mul_ps(beta2,rsq21);
2259 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2260 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2261 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2262 felec = _mm256_mul_ps(qq21,felec);
2263 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2264 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2265 velec = _mm256_sub_ps(rinv21,pmecorrV);
2266 velec = _mm256_mul_ps(qq21,velec);
2268 d = _mm256_sub_ps(r21,rswitch);
2269 d = _mm256_max_ps(d,_mm256_setzero_ps());
2270 d2 = _mm256_mul_ps(d,d);
2271 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2273 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2275 /* Evaluate switch function */
2276 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2277 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
2278 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2282 fscal = _mm256_and_ps(fscal,cutoff_mask);
2284 /* Calculate temporary vectorial force */
2285 tx = _mm256_mul_ps(fscal,dx21);
2286 ty = _mm256_mul_ps(fscal,dy21);
2287 tz = _mm256_mul_ps(fscal,dz21);
2289 /* Update vectorial force */
2290 fix2 = _mm256_add_ps(fix2,tx);
2291 fiy2 = _mm256_add_ps(fiy2,ty);
2292 fiz2 = _mm256_add_ps(fiz2,tz);
2294 fjx1 = _mm256_add_ps(fjx1,tx);
2295 fjy1 = _mm256_add_ps(fjy1,ty);
2296 fjz1 = _mm256_add_ps(fjz1,tz);
2300 /**************************
2301 * CALCULATE INTERACTIONS *
2302 **************************/
2304 if (gmx_mm256_any_lt(rsq22,rcutoff2))
2307 r22 = _mm256_mul_ps(rsq22,rinv22);
2309 /* EWALD ELECTROSTATICS */
2311 /* Analytical PME correction */
2312 zeta2 = _mm256_mul_ps(beta2,rsq22);
2313 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
2314 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2315 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2316 felec = _mm256_mul_ps(qq22,felec);
2317 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2318 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2319 velec = _mm256_sub_ps(rinv22,pmecorrV);
2320 velec = _mm256_mul_ps(qq22,velec);
2322 d = _mm256_sub_ps(r22,rswitch);
2323 d = _mm256_max_ps(d,_mm256_setzero_ps());
2324 d2 = _mm256_mul_ps(d,d);
2325 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2327 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2329 /* Evaluate switch function */
2330 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2331 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
2332 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
2336 fscal = _mm256_and_ps(fscal,cutoff_mask);
2338 /* Calculate temporary vectorial force */
2339 tx = _mm256_mul_ps(fscal,dx22);
2340 ty = _mm256_mul_ps(fscal,dy22);
2341 tz = _mm256_mul_ps(fscal,dz22);
2343 /* Update vectorial force */
2344 fix2 = _mm256_add_ps(fix2,tx);
2345 fiy2 = _mm256_add_ps(fiy2,ty);
2346 fiz2 = _mm256_add_ps(fiz2,tz);
2348 fjx2 = _mm256_add_ps(fjx2,tx);
2349 fjy2 = _mm256_add_ps(fjy2,ty);
2350 fjz2 = _mm256_add_ps(fjz2,tz);
2354 fjptrA = f+j_coord_offsetA;
2355 fjptrB = f+j_coord_offsetB;
2356 fjptrC = f+j_coord_offsetC;
2357 fjptrD = f+j_coord_offsetD;
2358 fjptrE = f+j_coord_offsetE;
2359 fjptrF = f+j_coord_offsetF;
2360 fjptrG = f+j_coord_offsetG;
2361 fjptrH = f+j_coord_offsetH;
2363 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2364 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2366 /* Inner loop uses 945 flops */
2369 if(jidx<j_index_end)
2372 /* Get j neighbor index, and coordinate index */
2373 jnrlistA = jjnr[jidx];
2374 jnrlistB = jjnr[jidx+1];
2375 jnrlistC = jjnr[jidx+2];
2376 jnrlistD = jjnr[jidx+3];
2377 jnrlistE = jjnr[jidx+4];
2378 jnrlistF = jjnr[jidx+5];
2379 jnrlistG = jjnr[jidx+6];
2380 jnrlistH = jjnr[jidx+7];
2381 /* Sign of each element will be negative for non-real atoms.
2382 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2383 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2385 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
2386 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
2388 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
2389 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
2390 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
2391 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
2392 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
2393 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
2394 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
2395 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
2396 j_coord_offsetA = DIM*jnrA;
2397 j_coord_offsetB = DIM*jnrB;
2398 j_coord_offsetC = DIM*jnrC;
2399 j_coord_offsetD = DIM*jnrD;
2400 j_coord_offsetE = DIM*jnrE;
2401 j_coord_offsetF = DIM*jnrF;
2402 j_coord_offsetG = DIM*jnrG;
2403 j_coord_offsetH = DIM*jnrH;
2405 /* load j atom coordinates */
2406 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2407 x+j_coord_offsetC,x+j_coord_offsetD,
2408 x+j_coord_offsetE,x+j_coord_offsetF,
2409 x+j_coord_offsetG,x+j_coord_offsetH,
2410 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2412 /* Calculate displacement vector */
2413 dx00 = _mm256_sub_ps(ix0,jx0);
2414 dy00 = _mm256_sub_ps(iy0,jy0);
2415 dz00 = _mm256_sub_ps(iz0,jz0);
2416 dx01 = _mm256_sub_ps(ix0,jx1);
2417 dy01 = _mm256_sub_ps(iy0,jy1);
2418 dz01 = _mm256_sub_ps(iz0,jz1);
2419 dx02 = _mm256_sub_ps(ix0,jx2);
2420 dy02 = _mm256_sub_ps(iy0,jy2);
2421 dz02 = _mm256_sub_ps(iz0,jz2);
2422 dx10 = _mm256_sub_ps(ix1,jx0);
2423 dy10 = _mm256_sub_ps(iy1,jy0);
2424 dz10 = _mm256_sub_ps(iz1,jz0);
2425 dx11 = _mm256_sub_ps(ix1,jx1);
2426 dy11 = _mm256_sub_ps(iy1,jy1);
2427 dz11 = _mm256_sub_ps(iz1,jz1);
2428 dx12 = _mm256_sub_ps(ix1,jx2);
2429 dy12 = _mm256_sub_ps(iy1,jy2);
2430 dz12 = _mm256_sub_ps(iz1,jz2);
2431 dx20 = _mm256_sub_ps(ix2,jx0);
2432 dy20 = _mm256_sub_ps(iy2,jy0);
2433 dz20 = _mm256_sub_ps(iz2,jz0);
2434 dx21 = _mm256_sub_ps(ix2,jx1);
2435 dy21 = _mm256_sub_ps(iy2,jy1);
2436 dz21 = _mm256_sub_ps(iz2,jz1);
2437 dx22 = _mm256_sub_ps(ix2,jx2);
2438 dy22 = _mm256_sub_ps(iy2,jy2);
2439 dz22 = _mm256_sub_ps(iz2,jz2);
2441 /* Calculate squared distance and things based on it */
2442 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
2443 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
2444 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
2445 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
2446 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
2447 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
2448 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
2449 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
2450 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
2452 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
2453 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
2454 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
2455 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
2456 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
2457 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
2458 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
2459 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
2460 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
2462 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
2463 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
2464 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
2465 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
2466 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
2467 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
2468 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
2469 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
2470 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
2472 fjx0 = _mm256_setzero_ps();
2473 fjy0 = _mm256_setzero_ps();
2474 fjz0 = _mm256_setzero_ps();
2475 fjx1 = _mm256_setzero_ps();
2476 fjy1 = _mm256_setzero_ps();
2477 fjz1 = _mm256_setzero_ps();
2478 fjx2 = _mm256_setzero_ps();
2479 fjy2 = _mm256_setzero_ps();
2480 fjz2 = _mm256_setzero_ps();
2482 /**************************
2483 * CALCULATE INTERACTIONS *
2484 **************************/
2486 if (gmx_mm256_any_lt(rsq00,rcutoff2))
2489 r00 = _mm256_mul_ps(rsq00,rinv00);
2490 r00 = _mm256_andnot_ps(dummy_mask,r00);
2492 /* EWALD ELECTROSTATICS */
2494 /* Analytical PME correction */
2495 zeta2 = _mm256_mul_ps(beta2,rsq00);
2496 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
2497 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2498 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2499 felec = _mm256_mul_ps(qq00,felec);
2500 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2501 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2502 velec = _mm256_sub_ps(rinv00,pmecorrV);
2503 velec = _mm256_mul_ps(qq00,velec);
2505 d = _mm256_sub_ps(r00,rswitch);
2506 d = _mm256_max_ps(d,_mm256_setzero_ps());
2507 d2 = _mm256_mul_ps(d,d);
2508 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2510 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2512 /* Evaluate switch function */
2513 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2514 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
2515 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
2519 fscal = _mm256_and_ps(fscal,cutoff_mask);
2521 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2523 /* Calculate temporary vectorial force */
2524 tx = _mm256_mul_ps(fscal,dx00);
2525 ty = _mm256_mul_ps(fscal,dy00);
2526 tz = _mm256_mul_ps(fscal,dz00);
2528 /* Update vectorial force */
2529 fix0 = _mm256_add_ps(fix0,tx);
2530 fiy0 = _mm256_add_ps(fiy0,ty);
2531 fiz0 = _mm256_add_ps(fiz0,tz);
2533 fjx0 = _mm256_add_ps(fjx0,tx);
2534 fjy0 = _mm256_add_ps(fjy0,ty);
2535 fjz0 = _mm256_add_ps(fjz0,tz);
2539 /**************************
2540 * CALCULATE INTERACTIONS *
2541 **************************/
2543 if (gmx_mm256_any_lt(rsq01,rcutoff2))
2546 r01 = _mm256_mul_ps(rsq01,rinv01);
2547 r01 = _mm256_andnot_ps(dummy_mask,r01);
2549 /* EWALD ELECTROSTATICS */
2551 /* Analytical PME correction */
2552 zeta2 = _mm256_mul_ps(beta2,rsq01);
2553 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
2554 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2555 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2556 felec = _mm256_mul_ps(qq01,felec);
2557 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2558 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2559 velec = _mm256_sub_ps(rinv01,pmecorrV);
2560 velec = _mm256_mul_ps(qq01,velec);
2562 d = _mm256_sub_ps(r01,rswitch);
2563 d = _mm256_max_ps(d,_mm256_setzero_ps());
2564 d2 = _mm256_mul_ps(d,d);
2565 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2567 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2569 /* Evaluate switch function */
2570 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2571 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
2572 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
2576 fscal = _mm256_and_ps(fscal,cutoff_mask);
2578 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2580 /* Calculate temporary vectorial force */
2581 tx = _mm256_mul_ps(fscal,dx01);
2582 ty = _mm256_mul_ps(fscal,dy01);
2583 tz = _mm256_mul_ps(fscal,dz01);
2585 /* Update vectorial force */
2586 fix0 = _mm256_add_ps(fix0,tx);
2587 fiy0 = _mm256_add_ps(fiy0,ty);
2588 fiz0 = _mm256_add_ps(fiz0,tz);
2590 fjx1 = _mm256_add_ps(fjx1,tx);
2591 fjy1 = _mm256_add_ps(fjy1,ty);
2592 fjz1 = _mm256_add_ps(fjz1,tz);
2596 /**************************
2597 * CALCULATE INTERACTIONS *
2598 **************************/
2600 if (gmx_mm256_any_lt(rsq02,rcutoff2))
2603 r02 = _mm256_mul_ps(rsq02,rinv02);
2604 r02 = _mm256_andnot_ps(dummy_mask,r02);
2606 /* EWALD ELECTROSTATICS */
2608 /* Analytical PME correction */
2609 zeta2 = _mm256_mul_ps(beta2,rsq02);
2610 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
2611 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2612 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2613 felec = _mm256_mul_ps(qq02,felec);
2614 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2615 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2616 velec = _mm256_sub_ps(rinv02,pmecorrV);
2617 velec = _mm256_mul_ps(qq02,velec);
2619 d = _mm256_sub_ps(r02,rswitch);
2620 d = _mm256_max_ps(d,_mm256_setzero_ps());
2621 d2 = _mm256_mul_ps(d,d);
2622 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2624 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2626 /* Evaluate switch function */
2627 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2628 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
2629 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
2633 fscal = _mm256_and_ps(fscal,cutoff_mask);
2635 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2637 /* Calculate temporary vectorial force */
2638 tx = _mm256_mul_ps(fscal,dx02);
2639 ty = _mm256_mul_ps(fscal,dy02);
2640 tz = _mm256_mul_ps(fscal,dz02);
2642 /* Update vectorial force */
2643 fix0 = _mm256_add_ps(fix0,tx);
2644 fiy0 = _mm256_add_ps(fiy0,ty);
2645 fiz0 = _mm256_add_ps(fiz0,tz);
2647 fjx2 = _mm256_add_ps(fjx2,tx);
2648 fjy2 = _mm256_add_ps(fjy2,ty);
2649 fjz2 = _mm256_add_ps(fjz2,tz);
2653 /**************************
2654 * CALCULATE INTERACTIONS *
2655 **************************/
2657 if (gmx_mm256_any_lt(rsq10,rcutoff2))
2660 r10 = _mm256_mul_ps(rsq10,rinv10);
2661 r10 = _mm256_andnot_ps(dummy_mask,r10);
2663 /* EWALD ELECTROSTATICS */
2665 /* Analytical PME correction */
2666 zeta2 = _mm256_mul_ps(beta2,rsq10);
2667 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2668 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2669 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2670 felec = _mm256_mul_ps(qq10,felec);
2671 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2672 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2673 velec = _mm256_sub_ps(rinv10,pmecorrV);
2674 velec = _mm256_mul_ps(qq10,velec);
2676 d = _mm256_sub_ps(r10,rswitch);
2677 d = _mm256_max_ps(d,_mm256_setzero_ps());
2678 d2 = _mm256_mul_ps(d,d);
2679 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2681 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2683 /* Evaluate switch function */
2684 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2685 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
2686 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
2690 fscal = _mm256_and_ps(fscal,cutoff_mask);
2692 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2694 /* Calculate temporary vectorial force */
2695 tx = _mm256_mul_ps(fscal,dx10);
2696 ty = _mm256_mul_ps(fscal,dy10);
2697 tz = _mm256_mul_ps(fscal,dz10);
2699 /* Update vectorial force */
2700 fix1 = _mm256_add_ps(fix1,tx);
2701 fiy1 = _mm256_add_ps(fiy1,ty);
2702 fiz1 = _mm256_add_ps(fiz1,tz);
2704 fjx0 = _mm256_add_ps(fjx0,tx);
2705 fjy0 = _mm256_add_ps(fjy0,ty);
2706 fjz0 = _mm256_add_ps(fjz0,tz);
2710 /**************************
2711 * CALCULATE INTERACTIONS *
2712 **************************/
2714 if (gmx_mm256_any_lt(rsq11,rcutoff2))
2717 r11 = _mm256_mul_ps(rsq11,rinv11);
2718 r11 = _mm256_andnot_ps(dummy_mask,r11);
2720 /* EWALD ELECTROSTATICS */
2722 /* Analytical PME correction */
2723 zeta2 = _mm256_mul_ps(beta2,rsq11);
2724 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2725 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2726 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2727 felec = _mm256_mul_ps(qq11,felec);
2728 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2729 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2730 velec = _mm256_sub_ps(rinv11,pmecorrV);
2731 velec = _mm256_mul_ps(qq11,velec);
2733 d = _mm256_sub_ps(r11,rswitch);
2734 d = _mm256_max_ps(d,_mm256_setzero_ps());
2735 d2 = _mm256_mul_ps(d,d);
2736 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2738 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2740 /* Evaluate switch function */
2741 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2742 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
2743 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
2747 fscal = _mm256_and_ps(fscal,cutoff_mask);
2749 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2751 /* Calculate temporary vectorial force */
2752 tx = _mm256_mul_ps(fscal,dx11);
2753 ty = _mm256_mul_ps(fscal,dy11);
2754 tz = _mm256_mul_ps(fscal,dz11);
2756 /* Update vectorial force */
2757 fix1 = _mm256_add_ps(fix1,tx);
2758 fiy1 = _mm256_add_ps(fiy1,ty);
2759 fiz1 = _mm256_add_ps(fiz1,tz);
2761 fjx1 = _mm256_add_ps(fjx1,tx);
2762 fjy1 = _mm256_add_ps(fjy1,ty);
2763 fjz1 = _mm256_add_ps(fjz1,tz);
2767 /**************************
2768 * CALCULATE INTERACTIONS *
2769 **************************/
2771 if (gmx_mm256_any_lt(rsq12,rcutoff2))
2774 r12 = _mm256_mul_ps(rsq12,rinv12);
2775 r12 = _mm256_andnot_ps(dummy_mask,r12);
2777 /* EWALD ELECTROSTATICS */
2779 /* Analytical PME correction */
2780 zeta2 = _mm256_mul_ps(beta2,rsq12);
2781 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2782 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2783 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2784 felec = _mm256_mul_ps(qq12,felec);
2785 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2786 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2787 velec = _mm256_sub_ps(rinv12,pmecorrV);
2788 velec = _mm256_mul_ps(qq12,velec);
2790 d = _mm256_sub_ps(r12,rswitch);
2791 d = _mm256_max_ps(d,_mm256_setzero_ps());
2792 d2 = _mm256_mul_ps(d,d);
2793 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2795 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2797 /* Evaluate switch function */
2798 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2799 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
2800 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
2804 fscal = _mm256_and_ps(fscal,cutoff_mask);
2806 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2808 /* Calculate temporary vectorial force */
2809 tx = _mm256_mul_ps(fscal,dx12);
2810 ty = _mm256_mul_ps(fscal,dy12);
2811 tz = _mm256_mul_ps(fscal,dz12);
2813 /* Update vectorial force */
2814 fix1 = _mm256_add_ps(fix1,tx);
2815 fiy1 = _mm256_add_ps(fiy1,ty);
2816 fiz1 = _mm256_add_ps(fiz1,tz);
2818 fjx2 = _mm256_add_ps(fjx2,tx);
2819 fjy2 = _mm256_add_ps(fjy2,ty);
2820 fjz2 = _mm256_add_ps(fjz2,tz);
2824 /**************************
2825 * CALCULATE INTERACTIONS *
2826 **************************/
2828 if (gmx_mm256_any_lt(rsq20,rcutoff2))
2831 r20 = _mm256_mul_ps(rsq20,rinv20);
2832 r20 = _mm256_andnot_ps(dummy_mask,r20);
2834 /* EWALD ELECTROSTATICS */
2836 /* Analytical PME correction */
2837 zeta2 = _mm256_mul_ps(beta2,rsq20);
2838 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
2839 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2840 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2841 felec = _mm256_mul_ps(qq20,felec);
2842 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2843 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2844 velec = _mm256_sub_ps(rinv20,pmecorrV);
2845 velec = _mm256_mul_ps(qq20,velec);
2847 d = _mm256_sub_ps(r20,rswitch);
2848 d = _mm256_max_ps(d,_mm256_setzero_ps());
2849 d2 = _mm256_mul_ps(d,d);
2850 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2852 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2854 /* Evaluate switch function */
2855 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2856 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
2857 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
2861 fscal = _mm256_and_ps(fscal,cutoff_mask);
2863 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2865 /* Calculate temporary vectorial force */
2866 tx = _mm256_mul_ps(fscal,dx20);
2867 ty = _mm256_mul_ps(fscal,dy20);
2868 tz = _mm256_mul_ps(fscal,dz20);
2870 /* Update vectorial force */
2871 fix2 = _mm256_add_ps(fix2,tx);
2872 fiy2 = _mm256_add_ps(fiy2,ty);
2873 fiz2 = _mm256_add_ps(fiz2,tz);
2875 fjx0 = _mm256_add_ps(fjx0,tx);
2876 fjy0 = _mm256_add_ps(fjy0,ty);
2877 fjz0 = _mm256_add_ps(fjz0,tz);
2881 /**************************
2882 * CALCULATE INTERACTIONS *
2883 **************************/
2885 if (gmx_mm256_any_lt(rsq21,rcutoff2))
2888 r21 = _mm256_mul_ps(rsq21,rinv21);
2889 r21 = _mm256_andnot_ps(dummy_mask,r21);
2891 /* EWALD ELECTROSTATICS */
2893 /* Analytical PME correction */
2894 zeta2 = _mm256_mul_ps(beta2,rsq21);
2895 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2896 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2897 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2898 felec = _mm256_mul_ps(qq21,felec);
2899 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2900 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2901 velec = _mm256_sub_ps(rinv21,pmecorrV);
2902 velec = _mm256_mul_ps(qq21,velec);
2904 d = _mm256_sub_ps(r21,rswitch);
2905 d = _mm256_max_ps(d,_mm256_setzero_ps());
2906 d2 = _mm256_mul_ps(d,d);
2907 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2909 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2911 /* Evaluate switch function */
2912 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2913 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
2914 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2918 fscal = _mm256_and_ps(fscal,cutoff_mask);
2920 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2922 /* Calculate temporary vectorial force */
2923 tx = _mm256_mul_ps(fscal,dx21);
2924 ty = _mm256_mul_ps(fscal,dy21);
2925 tz = _mm256_mul_ps(fscal,dz21);
2927 /* Update vectorial force */
2928 fix2 = _mm256_add_ps(fix2,tx);
2929 fiy2 = _mm256_add_ps(fiy2,ty);
2930 fiz2 = _mm256_add_ps(fiz2,tz);
2932 fjx1 = _mm256_add_ps(fjx1,tx);
2933 fjy1 = _mm256_add_ps(fjy1,ty);
2934 fjz1 = _mm256_add_ps(fjz1,tz);
2938 /**************************
2939 * CALCULATE INTERACTIONS *
2940 **************************/
2942 if (gmx_mm256_any_lt(rsq22,rcutoff2))
2945 r22 = _mm256_mul_ps(rsq22,rinv22);
2946 r22 = _mm256_andnot_ps(dummy_mask,r22);
2948 /* EWALD ELECTROSTATICS */
2950 /* Analytical PME correction */
2951 zeta2 = _mm256_mul_ps(beta2,rsq22);
2952 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
2953 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2954 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2955 felec = _mm256_mul_ps(qq22,felec);
2956 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2957 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2958 velec = _mm256_sub_ps(rinv22,pmecorrV);
2959 velec = _mm256_mul_ps(qq22,velec);
2961 d = _mm256_sub_ps(r22,rswitch);
2962 d = _mm256_max_ps(d,_mm256_setzero_ps());
2963 d2 = _mm256_mul_ps(d,d);
2964 sw = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2966 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2968 /* Evaluate switch function */
2969 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2970 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
2971 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
2975 fscal = _mm256_and_ps(fscal,cutoff_mask);
2977 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2979 /* Calculate temporary vectorial force */
2980 tx = _mm256_mul_ps(fscal,dx22);
2981 ty = _mm256_mul_ps(fscal,dy22);
2982 tz = _mm256_mul_ps(fscal,dz22);
2984 /* Update vectorial force */
2985 fix2 = _mm256_add_ps(fix2,tx);
2986 fiy2 = _mm256_add_ps(fiy2,ty);
2987 fiz2 = _mm256_add_ps(fiz2,tz);
2989 fjx2 = _mm256_add_ps(fjx2,tx);
2990 fjy2 = _mm256_add_ps(fjy2,ty);
2991 fjz2 = _mm256_add_ps(fjz2,tz);
2995 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2996 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2997 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2998 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2999 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
3000 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
3001 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
3002 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
3004 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
3005 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
3007 /* Inner loop uses 954 flops */
3010 /* End of innermost loop */
3012 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
3013 f+i_coord_offset,fshift+i_shift_offset);
3015 /* Increment number of inner iterations */
3016 inneriter += j_index_end - j_index_start;
3018 /* Outer loop uses 18 flops */
3021 /* Increment number of outer iterations */
3024 /* Update outer/inner flops */
3026 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*954);