2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS avx_256_single kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_avx_256_single.h"
48 #include "kernelutil_x86_avx_256_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3W3_VF_avx_256_single
52 * Electrostatics interaction: Ewald
53 * VdW interaction: None
54 * Geometry: Water3-Water3
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSw_VdwNone_GeomW3W3_VF_avx_256_single
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
74 int jnrA,jnrB,jnrC,jnrD;
75 int jnrE,jnrF,jnrG,jnrH;
76 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
77 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
78 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
79 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
80 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
82 real *shiftvec,*fshift,*x,*f;
83 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
85 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 real * vdwioffsetptr0;
87 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
88 real * vdwioffsetptr1;
89 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
90 real * vdwioffsetptr2;
91 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
92 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
93 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
94 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
95 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
96 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
97 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
98 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
99 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
100 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
101 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
102 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
103 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
104 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
105 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
106 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
107 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
110 __m128i ewitab_lo,ewitab_hi;
111 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
112 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
114 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
115 real rswitch_scalar,d_scalar;
116 __m256 dummy_mask,cutoff_mask;
117 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
118 __m256 one = _mm256_set1_ps(1.0);
119 __m256 two = _mm256_set1_ps(2.0);
125 jindex = nlist->jindex;
127 shiftidx = nlist->shift;
129 shiftvec = fr->shift_vec[0];
130 fshift = fr->fshift[0];
131 facel = _mm256_set1_ps(fr->epsfac);
132 charge = mdatoms->chargeA;
134 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
135 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
136 beta2 = _mm256_mul_ps(beta,beta);
137 beta3 = _mm256_mul_ps(beta,beta2);
139 ewtab = fr->ic->tabq_coul_FDV0;
140 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
141 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
143 /* Setup water-specific parameters */
144 inr = nlist->iinr[0];
145 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
146 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
147 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
149 jq0 = _mm256_set1_ps(charge[inr+0]);
150 jq1 = _mm256_set1_ps(charge[inr+1]);
151 jq2 = _mm256_set1_ps(charge[inr+2]);
152 qq00 = _mm256_mul_ps(iq0,jq0);
153 qq01 = _mm256_mul_ps(iq0,jq1);
154 qq02 = _mm256_mul_ps(iq0,jq2);
155 qq10 = _mm256_mul_ps(iq1,jq0);
156 qq11 = _mm256_mul_ps(iq1,jq1);
157 qq12 = _mm256_mul_ps(iq1,jq2);
158 qq20 = _mm256_mul_ps(iq2,jq0);
159 qq21 = _mm256_mul_ps(iq2,jq1);
160 qq22 = _mm256_mul_ps(iq2,jq2);
162 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
163 rcutoff_scalar = fr->rcoulomb;
164 rcutoff = _mm256_set1_ps(rcutoff_scalar);
165 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
167 rswitch_scalar = fr->rcoulomb_switch;
168 rswitch = _mm256_set1_ps(rswitch_scalar);
169 /* Setup switch parameters */
170 d_scalar = rcutoff_scalar-rswitch_scalar;
171 d = _mm256_set1_ps(d_scalar);
172 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
173 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
174 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
175 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
176 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
177 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
179 /* Avoid stupid compiler warnings */
180 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
193 for(iidx=0;iidx<4*DIM;iidx++)
198 /* Start outer loop over neighborlists */
199 for(iidx=0; iidx<nri; iidx++)
201 /* Load shift vector for this list */
202 i_shift_offset = DIM*shiftidx[iidx];
204 /* Load limits for loop over neighbors */
205 j_index_start = jindex[iidx];
206 j_index_end = jindex[iidx+1];
208 /* Get outer coordinate index */
210 i_coord_offset = DIM*inr;
212 /* Load i particle coords and add shift vector */
213 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
214 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
216 fix0 = _mm256_setzero_ps();
217 fiy0 = _mm256_setzero_ps();
218 fiz0 = _mm256_setzero_ps();
219 fix1 = _mm256_setzero_ps();
220 fiy1 = _mm256_setzero_ps();
221 fiz1 = _mm256_setzero_ps();
222 fix2 = _mm256_setzero_ps();
223 fiy2 = _mm256_setzero_ps();
224 fiz2 = _mm256_setzero_ps();
226 /* Reset potential sums */
227 velecsum = _mm256_setzero_ps();
229 /* Start inner kernel loop */
230 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
233 /* Get j neighbor index, and coordinate index */
242 j_coord_offsetA = DIM*jnrA;
243 j_coord_offsetB = DIM*jnrB;
244 j_coord_offsetC = DIM*jnrC;
245 j_coord_offsetD = DIM*jnrD;
246 j_coord_offsetE = DIM*jnrE;
247 j_coord_offsetF = DIM*jnrF;
248 j_coord_offsetG = DIM*jnrG;
249 j_coord_offsetH = DIM*jnrH;
251 /* load j atom coordinates */
252 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
253 x+j_coord_offsetC,x+j_coord_offsetD,
254 x+j_coord_offsetE,x+j_coord_offsetF,
255 x+j_coord_offsetG,x+j_coord_offsetH,
256 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
258 /* Calculate displacement vector */
259 dx00 = _mm256_sub_ps(ix0,jx0);
260 dy00 = _mm256_sub_ps(iy0,jy0);
261 dz00 = _mm256_sub_ps(iz0,jz0);
262 dx01 = _mm256_sub_ps(ix0,jx1);
263 dy01 = _mm256_sub_ps(iy0,jy1);
264 dz01 = _mm256_sub_ps(iz0,jz1);
265 dx02 = _mm256_sub_ps(ix0,jx2);
266 dy02 = _mm256_sub_ps(iy0,jy2);
267 dz02 = _mm256_sub_ps(iz0,jz2);
268 dx10 = _mm256_sub_ps(ix1,jx0);
269 dy10 = _mm256_sub_ps(iy1,jy0);
270 dz10 = _mm256_sub_ps(iz1,jz0);
271 dx11 = _mm256_sub_ps(ix1,jx1);
272 dy11 = _mm256_sub_ps(iy1,jy1);
273 dz11 = _mm256_sub_ps(iz1,jz1);
274 dx12 = _mm256_sub_ps(ix1,jx2);
275 dy12 = _mm256_sub_ps(iy1,jy2);
276 dz12 = _mm256_sub_ps(iz1,jz2);
277 dx20 = _mm256_sub_ps(ix2,jx0);
278 dy20 = _mm256_sub_ps(iy2,jy0);
279 dz20 = _mm256_sub_ps(iz2,jz0);
280 dx21 = _mm256_sub_ps(ix2,jx1);
281 dy21 = _mm256_sub_ps(iy2,jy1);
282 dz21 = _mm256_sub_ps(iz2,jz1);
283 dx22 = _mm256_sub_ps(ix2,jx2);
284 dy22 = _mm256_sub_ps(iy2,jy2);
285 dz22 = _mm256_sub_ps(iz2,jz2);
287 /* Calculate squared distance and things based on it */
288 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
289 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
290 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
291 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
292 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
293 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
294 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
295 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
296 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
298 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
299 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
300 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
301 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
302 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
303 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
304 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
305 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
306 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
308 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
309 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
310 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
311 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
312 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
313 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
314 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
315 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
316 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
318 fjx0 = _mm256_setzero_ps();
319 fjy0 = _mm256_setzero_ps();
320 fjz0 = _mm256_setzero_ps();
321 fjx1 = _mm256_setzero_ps();
322 fjy1 = _mm256_setzero_ps();
323 fjz1 = _mm256_setzero_ps();
324 fjx2 = _mm256_setzero_ps();
325 fjy2 = _mm256_setzero_ps();
326 fjz2 = _mm256_setzero_ps();
328 /**************************
329 * CALCULATE INTERACTIONS *
330 **************************/
332 if (gmx_mm256_any_lt(rsq00,rcutoff2))
335 r00 = _mm256_mul_ps(rsq00,rinv00);
337 /* EWALD ELECTROSTATICS */
339 /* Analytical PME correction */
340 zeta2 = _mm256_mul_ps(beta2,rsq00);
341 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
342 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
343 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
344 felec = _mm256_mul_ps(qq00,felec);
345 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
346 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
347 velec = _mm256_sub_ps(rinv00,pmecorrV);
348 velec = _mm256_mul_ps(qq00,velec);
350 d = _mm256_sub_ps(r00,rswitch);
351 d = _mm256_max_ps(d,_mm256_setzero_ps());
352 d2 = _mm256_mul_ps(d,d);
353 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)))))));
355 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
357 /* Evaluate switch function */
358 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
359 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
360 velec = _mm256_mul_ps(velec,sw);
361 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
363 /* Update potential sum for this i atom from the interaction with this j atom. */
364 velec = _mm256_and_ps(velec,cutoff_mask);
365 velecsum = _mm256_add_ps(velecsum,velec);
369 fscal = _mm256_and_ps(fscal,cutoff_mask);
371 /* Calculate temporary vectorial force */
372 tx = _mm256_mul_ps(fscal,dx00);
373 ty = _mm256_mul_ps(fscal,dy00);
374 tz = _mm256_mul_ps(fscal,dz00);
376 /* Update vectorial force */
377 fix0 = _mm256_add_ps(fix0,tx);
378 fiy0 = _mm256_add_ps(fiy0,ty);
379 fiz0 = _mm256_add_ps(fiz0,tz);
381 fjx0 = _mm256_add_ps(fjx0,tx);
382 fjy0 = _mm256_add_ps(fjy0,ty);
383 fjz0 = _mm256_add_ps(fjz0,tz);
387 /**************************
388 * CALCULATE INTERACTIONS *
389 **************************/
391 if (gmx_mm256_any_lt(rsq01,rcutoff2))
394 r01 = _mm256_mul_ps(rsq01,rinv01);
396 /* EWALD ELECTROSTATICS */
398 /* Analytical PME correction */
399 zeta2 = _mm256_mul_ps(beta2,rsq01);
400 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
401 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
402 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
403 felec = _mm256_mul_ps(qq01,felec);
404 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
405 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
406 velec = _mm256_sub_ps(rinv01,pmecorrV);
407 velec = _mm256_mul_ps(qq01,velec);
409 d = _mm256_sub_ps(r01,rswitch);
410 d = _mm256_max_ps(d,_mm256_setzero_ps());
411 d2 = _mm256_mul_ps(d,d);
412 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)))))));
414 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
416 /* Evaluate switch function */
417 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
418 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
419 velec = _mm256_mul_ps(velec,sw);
420 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
422 /* Update potential sum for this i atom from the interaction with this j atom. */
423 velec = _mm256_and_ps(velec,cutoff_mask);
424 velecsum = _mm256_add_ps(velecsum,velec);
428 fscal = _mm256_and_ps(fscal,cutoff_mask);
430 /* Calculate temporary vectorial force */
431 tx = _mm256_mul_ps(fscal,dx01);
432 ty = _mm256_mul_ps(fscal,dy01);
433 tz = _mm256_mul_ps(fscal,dz01);
435 /* Update vectorial force */
436 fix0 = _mm256_add_ps(fix0,tx);
437 fiy0 = _mm256_add_ps(fiy0,ty);
438 fiz0 = _mm256_add_ps(fiz0,tz);
440 fjx1 = _mm256_add_ps(fjx1,tx);
441 fjy1 = _mm256_add_ps(fjy1,ty);
442 fjz1 = _mm256_add_ps(fjz1,tz);
446 /**************************
447 * CALCULATE INTERACTIONS *
448 **************************/
450 if (gmx_mm256_any_lt(rsq02,rcutoff2))
453 r02 = _mm256_mul_ps(rsq02,rinv02);
455 /* EWALD ELECTROSTATICS */
457 /* Analytical PME correction */
458 zeta2 = _mm256_mul_ps(beta2,rsq02);
459 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
460 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
461 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
462 felec = _mm256_mul_ps(qq02,felec);
463 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
464 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
465 velec = _mm256_sub_ps(rinv02,pmecorrV);
466 velec = _mm256_mul_ps(qq02,velec);
468 d = _mm256_sub_ps(r02,rswitch);
469 d = _mm256_max_ps(d,_mm256_setzero_ps());
470 d2 = _mm256_mul_ps(d,d);
471 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)))))));
473 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
475 /* Evaluate switch function */
476 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
477 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
478 velec = _mm256_mul_ps(velec,sw);
479 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
481 /* Update potential sum for this i atom from the interaction with this j atom. */
482 velec = _mm256_and_ps(velec,cutoff_mask);
483 velecsum = _mm256_add_ps(velecsum,velec);
487 fscal = _mm256_and_ps(fscal,cutoff_mask);
489 /* Calculate temporary vectorial force */
490 tx = _mm256_mul_ps(fscal,dx02);
491 ty = _mm256_mul_ps(fscal,dy02);
492 tz = _mm256_mul_ps(fscal,dz02);
494 /* Update vectorial force */
495 fix0 = _mm256_add_ps(fix0,tx);
496 fiy0 = _mm256_add_ps(fiy0,ty);
497 fiz0 = _mm256_add_ps(fiz0,tz);
499 fjx2 = _mm256_add_ps(fjx2,tx);
500 fjy2 = _mm256_add_ps(fjy2,ty);
501 fjz2 = _mm256_add_ps(fjz2,tz);
505 /**************************
506 * CALCULATE INTERACTIONS *
507 **************************/
509 if (gmx_mm256_any_lt(rsq10,rcutoff2))
512 r10 = _mm256_mul_ps(rsq10,rinv10);
514 /* EWALD ELECTROSTATICS */
516 /* Analytical PME correction */
517 zeta2 = _mm256_mul_ps(beta2,rsq10);
518 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
519 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
520 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
521 felec = _mm256_mul_ps(qq10,felec);
522 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
523 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
524 velec = _mm256_sub_ps(rinv10,pmecorrV);
525 velec = _mm256_mul_ps(qq10,velec);
527 d = _mm256_sub_ps(r10,rswitch);
528 d = _mm256_max_ps(d,_mm256_setzero_ps());
529 d2 = _mm256_mul_ps(d,d);
530 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)))))));
532 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
534 /* Evaluate switch function */
535 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
536 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
537 velec = _mm256_mul_ps(velec,sw);
538 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
540 /* Update potential sum for this i atom from the interaction with this j atom. */
541 velec = _mm256_and_ps(velec,cutoff_mask);
542 velecsum = _mm256_add_ps(velecsum,velec);
546 fscal = _mm256_and_ps(fscal,cutoff_mask);
548 /* Calculate temporary vectorial force */
549 tx = _mm256_mul_ps(fscal,dx10);
550 ty = _mm256_mul_ps(fscal,dy10);
551 tz = _mm256_mul_ps(fscal,dz10);
553 /* Update vectorial force */
554 fix1 = _mm256_add_ps(fix1,tx);
555 fiy1 = _mm256_add_ps(fiy1,ty);
556 fiz1 = _mm256_add_ps(fiz1,tz);
558 fjx0 = _mm256_add_ps(fjx0,tx);
559 fjy0 = _mm256_add_ps(fjy0,ty);
560 fjz0 = _mm256_add_ps(fjz0,tz);
564 /**************************
565 * CALCULATE INTERACTIONS *
566 **************************/
568 if (gmx_mm256_any_lt(rsq11,rcutoff2))
571 r11 = _mm256_mul_ps(rsq11,rinv11);
573 /* EWALD ELECTROSTATICS */
575 /* Analytical PME correction */
576 zeta2 = _mm256_mul_ps(beta2,rsq11);
577 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
578 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
579 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
580 felec = _mm256_mul_ps(qq11,felec);
581 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
582 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
583 velec = _mm256_sub_ps(rinv11,pmecorrV);
584 velec = _mm256_mul_ps(qq11,velec);
586 d = _mm256_sub_ps(r11,rswitch);
587 d = _mm256_max_ps(d,_mm256_setzero_ps());
588 d2 = _mm256_mul_ps(d,d);
589 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)))))));
591 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
593 /* Evaluate switch function */
594 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
595 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
596 velec = _mm256_mul_ps(velec,sw);
597 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
599 /* Update potential sum for this i atom from the interaction with this j atom. */
600 velec = _mm256_and_ps(velec,cutoff_mask);
601 velecsum = _mm256_add_ps(velecsum,velec);
605 fscal = _mm256_and_ps(fscal,cutoff_mask);
607 /* Calculate temporary vectorial force */
608 tx = _mm256_mul_ps(fscal,dx11);
609 ty = _mm256_mul_ps(fscal,dy11);
610 tz = _mm256_mul_ps(fscal,dz11);
612 /* Update vectorial force */
613 fix1 = _mm256_add_ps(fix1,tx);
614 fiy1 = _mm256_add_ps(fiy1,ty);
615 fiz1 = _mm256_add_ps(fiz1,tz);
617 fjx1 = _mm256_add_ps(fjx1,tx);
618 fjy1 = _mm256_add_ps(fjy1,ty);
619 fjz1 = _mm256_add_ps(fjz1,tz);
623 /**************************
624 * CALCULATE INTERACTIONS *
625 **************************/
627 if (gmx_mm256_any_lt(rsq12,rcutoff2))
630 r12 = _mm256_mul_ps(rsq12,rinv12);
632 /* EWALD ELECTROSTATICS */
634 /* Analytical PME correction */
635 zeta2 = _mm256_mul_ps(beta2,rsq12);
636 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
637 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
638 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
639 felec = _mm256_mul_ps(qq12,felec);
640 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
641 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
642 velec = _mm256_sub_ps(rinv12,pmecorrV);
643 velec = _mm256_mul_ps(qq12,velec);
645 d = _mm256_sub_ps(r12,rswitch);
646 d = _mm256_max_ps(d,_mm256_setzero_ps());
647 d2 = _mm256_mul_ps(d,d);
648 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)))))));
650 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
652 /* Evaluate switch function */
653 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
654 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
655 velec = _mm256_mul_ps(velec,sw);
656 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
658 /* Update potential sum for this i atom from the interaction with this j atom. */
659 velec = _mm256_and_ps(velec,cutoff_mask);
660 velecsum = _mm256_add_ps(velecsum,velec);
664 fscal = _mm256_and_ps(fscal,cutoff_mask);
666 /* Calculate temporary vectorial force */
667 tx = _mm256_mul_ps(fscal,dx12);
668 ty = _mm256_mul_ps(fscal,dy12);
669 tz = _mm256_mul_ps(fscal,dz12);
671 /* Update vectorial force */
672 fix1 = _mm256_add_ps(fix1,tx);
673 fiy1 = _mm256_add_ps(fiy1,ty);
674 fiz1 = _mm256_add_ps(fiz1,tz);
676 fjx2 = _mm256_add_ps(fjx2,tx);
677 fjy2 = _mm256_add_ps(fjy2,ty);
678 fjz2 = _mm256_add_ps(fjz2,tz);
682 /**************************
683 * CALCULATE INTERACTIONS *
684 **************************/
686 if (gmx_mm256_any_lt(rsq20,rcutoff2))
689 r20 = _mm256_mul_ps(rsq20,rinv20);
691 /* EWALD ELECTROSTATICS */
693 /* Analytical PME correction */
694 zeta2 = _mm256_mul_ps(beta2,rsq20);
695 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
696 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
697 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
698 felec = _mm256_mul_ps(qq20,felec);
699 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
700 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
701 velec = _mm256_sub_ps(rinv20,pmecorrV);
702 velec = _mm256_mul_ps(qq20,velec);
704 d = _mm256_sub_ps(r20,rswitch);
705 d = _mm256_max_ps(d,_mm256_setzero_ps());
706 d2 = _mm256_mul_ps(d,d);
707 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)))))));
709 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
711 /* Evaluate switch function */
712 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
713 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
714 velec = _mm256_mul_ps(velec,sw);
715 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
717 /* Update potential sum for this i atom from the interaction with this j atom. */
718 velec = _mm256_and_ps(velec,cutoff_mask);
719 velecsum = _mm256_add_ps(velecsum,velec);
723 fscal = _mm256_and_ps(fscal,cutoff_mask);
725 /* Calculate temporary vectorial force */
726 tx = _mm256_mul_ps(fscal,dx20);
727 ty = _mm256_mul_ps(fscal,dy20);
728 tz = _mm256_mul_ps(fscal,dz20);
730 /* Update vectorial force */
731 fix2 = _mm256_add_ps(fix2,tx);
732 fiy2 = _mm256_add_ps(fiy2,ty);
733 fiz2 = _mm256_add_ps(fiz2,tz);
735 fjx0 = _mm256_add_ps(fjx0,tx);
736 fjy0 = _mm256_add_ps(fjy0,ty);
737 fjz0 = _mm256_add_ps(fjz0,tz);
741 /**************************
742 * CALCULATE INTERACTIONS *
743 **************************/
745 if (gmx_mm256_any_lt(rsq21,rcutoff2))
748 r21 = _mm256_mul_ps(rsq21,rinv21);
750 /* EWALD ELECTROSTATICS */
752 /* Analytical PME correction */
753 zeta2 = _mm256_mul_ps(beta2,rsq21);
754 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
755 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
756 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
757 felec = _mm256_mul_ps(qq21,felec);
758 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
759 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
760 velec = _mm256_sub_ps(rinv21,pmecorrV);
761 velec = _mm256_mul_ps(qq21,velec);
763 d = _mm256_sub_ps(r21,rswitch);
764 d = _mm256_max_ps(d,_mm256_setzero_ps());
765 d2 = _mm256_mul_ps(d,d);
766 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)))))));
768 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
770 /* Evaluate switch function */
771 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
772 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
773 velec = _mm256_mul_ps(velec,sw);
774 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
776 /* Update potential sum for this i atom from the interaction with this j atom. */
777 velec = _mm256_and_ps(velec,cutoff_mask);
778 velecsum = _mm256_add_ps(velecsum,velec);
782 fscal = _mm256_and_ps(fscal,cutoff_mask);
784 /* Calculate temporary vectorial force */
785 tx = _mm256_mul_ps(fscal,dx21);
786 ty = _mm256_mul_ps(fscal,dy21);
787 tz = _mm256_mul_ps(fscal,dz21);
789 /* Update vectorial force */
790 fix2 = _mm256_add_ps(fix2,tx);
791 fiy2 = _mm256_add_ps(fiy2,ty);
792 fiz2 = _mm256_add_ps(fiz2,tz);
794 fjx1 = _mm256_add_ps(fjx1,tx);
795 fjy1 = _mm256_add_ps(fjy1,ty);
796 fjz1 = _mm256_add_ps(fjz1,tz);
800 /**************************
801 * CALCULATE INTERACTIONS *
802 **************************/
804 if (gmx_mm256_any_lt(rsq22,rcutoff2))
807 r22 = _mm256_mul_ps(rsq22,rinv22);
809 /* EWALD ELECTROSTATICS */
811 /* Analytical PME correction */
812 zeta2 = _mm256_mul_ps(beta2,rsq22);
813 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
814 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
815 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
816 felec = _mm256_mul_ps(qq22,felec);
817 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
818 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
819 velec = _mm256_sub_ps(rinv22,pmecorrV);
820 velec = _mm256_mul_ps(qq22,velec);
822 d = _mm256_sub_ps(r22,rswitch);
823 d = _mm256_max_ps(d,_mm256_setzero_ps());
824 d2 = _mm256_mul_ps(d,d);
825 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)))))));
827 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
829 /* Evaluate switch function */
830 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
831 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
832 velec = _mm256_mul_ps(velec,sw);
833 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
835 /* Update potential sum for this i atom from the interaction with this j atom. */
836 velec = _mm256_and_ps(velec,cutoff_mask);
837 velecsum = _mm256_add_ps(velecsum,velec);
841 fscal = _mm256_and_ps(fscal,cutoff_mask);
843 /* Calculate temporary vectorial force */
844 tx = _mm256_mul_ps(fscal,dx22);
845 ty = _mm256_mul_ps(fscal,dy22);
846 tz = _mm256_mul_ps(fscal,dz22);
848 /* Update vectorial force */
849 fix2 = _mm256_add_ps(fix2,tx);
850 fiy2 = _mm256_add_ps(fiy2,ty);
851 fiz2 = _mm256_add_ps(fiz2,tz);
853 fjx2 = _mm256_add_ps(fjx2,tx);
854 fjy2 = _mm256_add_ps(fjy2,ty);
855 fjz2 = _mm256_add_ps(fjz2,tz);
859 fjptrA = f+j_coord_offsetA;
860 fjptrB = f+j_coord_offsetB;
861 fjptrC = f+j_coord_offsetC;
862 fjptrD = f+j_coord_offsetD;
863 fjptrE = f+j_coord_offsetE;
864 fjptrF = f+j_coord_offsetF;
865 fjptrG = f+j_coord_offsetG;
866 fjptrH = f+j_coord_offsetH;
868 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
869 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
871 /* Inner loop uses 972 flops */
877 /* Get j neighbor index, and coordinate index */
878 jnrlistA = jjnr[jidx];
879 jnrlistB = jjnr[jidx+1];
880 jnrlistC = jjnr[jidx+2];
881 jnrlistD = jjnr[jidx+3];
882 jnrlistE = jjnr[jidx+4];
883 jnrlistF = jjnr[jidx+5];
884 jnrlistG = jjnr[jidx+6];
885 jnrlistH = jjnr[jidx+7];
886 /* Sign of each element will be negative for non-real atoms.
887 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
888 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
890 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
891 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
893 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
894 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
895 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
896 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
897 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
898 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
899 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
900 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
901 j_coord_offsetA = DIM*jnrA;
902 j_coord_offsetB = DIM*jnrB;
903 j_coord_offsetC = DIM*jnrC;
904 j_coord_offsetD = DIM*jnrD;
905 j_coord_offsetE = DIM*jnrE;
906 j_coord_offsetF = DIM*jnrF;
907 j_coord_offsetG = DIM*jnrG;
908 j_coord_offsetH = DIM*jnrH;
910 /* load j atom coordinates */
911 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
912 x+j_coord_offsetC,x+j_coord_offsetD,
913 x+j_coord_offsetE,x+j_coord_offsetF,
914 x+j_coord_offsetG,x+j_coord_offsetH,
915 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
917 /* Calculate displacement vector */
918 dx00 = _mm256_sub_ps(ix0,jx0);
919 dy00 = _mm256_sub_ps(iy0,jy0);
920 dz00 = _mm256_sub_ps(iz0,jz0);
921 dx01 = _mm256_sub_ps(ix0,jx1);
922 dy01 = _mm256_sub_ps(iy0,jy1);
923 dz01 = _mm256_sub_ps(iz0,jz1);
924 dx02 = _mm256_sub_ps(ix0,jx2);
925 dy02 = _mm256_sub_ps(iy0,jy2);
926 dz02 = _mm256_sub_ps(iz0,jz2);
927 dx10 = _mm256_sub_ps(ix1,jx0);
928 dy10 = _mm256_sub_ps(iy1,jy0);
929 dz10 = _mm256_sub_ps(iz1,jz0);
930 dx11 = _mm256_sub_ps(ix1,jx1);
931 dy11 = _mm256_sub_ps(iy1,jy1);
932 dz11 = _mm256_sub_ps(iz1,jz1);
933 dx12 = _mm256_sub_ps(ix1,jx2);
934 dy12 = _mm256_sub_ps(iy1,jy2);
935 dz12 = _mm256_sub_ps(iz1,jz2);
936 dx20 = _mm256_sub_ps(ix2,jx0);
937 dy20 = _mm256_sub_ps(iy2,jy0);
938 dz20 = _mm256_sub_ps(iz2,jz0);
939 dx21 = _mm256_sub_ps(ix2,jx1);
940 dy21 = _mm256_sub_ps(iy2,jy1);
941 dz21 = _mm256_sub_ps(iz2,jz1);
942 dx22 = _mm256_sub_ps(ix2,jx2);
943 dy22 = _mm256_sub_ps(iy2,jy2);
944 dz22 = _mm256_sub_ps(iz2,jz2);
946 /* Calculate squared distance and things based on it */
947 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
948 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
949 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
950 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
951 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
952 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
953 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
954 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
955 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
957 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
958 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
959 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
960 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
961 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
962 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
963 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
964 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
965 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
967 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
968 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
969 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
970 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
971 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
972 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
973 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
974 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
975 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
977 fjx0 = _mm256_setzero_ps();
978 fjy0 = _mm256_setzero_ps();
979 fjz0 = _mm256_setzero_ps();
980 fjx1 = _mm256_setzero_ps();
981 fjy1 = _mm256_setzero_ps();
982 fjz1 = _mm256_setzero_ps();
983 fjx2 = _mm256_setzero_ps();
984 fjy2 = _mm256_setzero_ps();
985 fjz2 = _mm256_setzero_ps();
987 /**************************
988 * CALCULATE INTERACTIONS *
989 **************************/
991 if (gmx_mm256_any_lt(rsq00,rcutoff2))
994 r00 = _mm256_mul_ps(rsq00,rinv00);
995 r00 = _mm256_andnot_ps(dummy_mask,r00);
997 /* EWALD ELECTROSTATICS */
999 /* Analytical PME correction */
1000 zeta2 = _mm256_mul_ps(beta2,rsq00);
1001 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1002 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1003 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1004 felec = _mm256_mul_ps(qq00,felec);
1005 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1006 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1007 velec = _mm256_sub_ps(rinv00,pmecorrV);
1008 velec = _mm256_mul_ps(qq00,velec);
1010 d = _mm256_sub_ps(r00,rswitch);
1011 d = _mm256_max_ps(d,_mm256_setzero_ps());
1012 d2 = _mm256_mul_ps(d,d);
1013 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)))))));
1015 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1017 /* Evaluate switch function */
1018 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1019 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1020 velec = _mm256_mul_ps(velec,sw);
1021 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1023 /* Update potential sum for this i atom from the interaction with this j atom. */
1024 velec = _mm256_and_ps(velec,cutoff_mask);
1025 velec = _mm256_andnot_ps(dummy_mask,velec);
1026 velecsum = _mm256_add_ps(velecsum,velec);
1030 fscal = _mm256_and_ps(fscal,cutoff_mask);
1032 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1034 /* Calculate temporary vectorial force */
1035 tx = _mm256_mul_ps(fscal,dx00);
1036 ty = _mm256_mul_ps(fscal,dy00);
1037 tz = _mm256_mul_ps(fscal,dz00);
1039 /* Update vectorial force */
1040 fix0 = _mm256_add_ps(fix0,tx);
1041 fiy0 = _mm256_add_ps(fiy0,ty);
1042 fiz0 = _mm256_add_ps(fiz0,tz);
1044 fjx0 = _mm256_add_ps(fjx0,tx);
1045 fjy0 = _mm256_add_ps(fjy0,ty);
1046 fjz0 = _mm256_add_ps(fjz0,tz);
1050 /**************************
1051 * CALCULATE INTERACTIONS *
1052 **************************/
1054 if (gmx_mm256_any_lt(rsq01,rcutoff2))
1057 r01 = _mm256_mul_ps(rsq01,rinv01);
1058 r01 = _mm256_andnot_ps(dummy_mask,r01);
1060 /* EWALD ELECTROSTATICS */
1062 /* Analytical PME correction */
1063 zeta2 = _mm256_mul_ps(beta2,rsq01);
1064 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1065 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1066 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1067 felec = _mm256_mul_ps(qq01,felec);
1068 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1069 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1070 velec = _mm256_sub_ps(rinv01,pmecorrV);
1071 velec = _mm256_mul_ps(qq01,velec);
1073 d = _mm256_sub_ps(r01,rswitch);
1074 d = _mm256_max_ps(d,_mm256_setzero_ps());
1075 d2 = _mm256_mul_ps(d,d);
1076 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)))))));
1078 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1080 /* Evaluate switch function */
1081 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1082 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
1083 velec = _mm256_mul_ps(velec,sw);
1084 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
1086 /* Update potential sum for this i atom from the interaction with this j atom. */
1087 velec = _mm256_and_ps(velec,cutoff_mask);
1088 velec = _mm256_andnot_ps(dummy_mask,velec);
1089 velecsum = _mm256_add_ps(velecsum,velec);
1093 fscal = _mm256_and_ps(fscal,cutoff_mask);
1095 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1097 /* Calculate temporary vectorial force */
1098 tx = _mm256_mul_ps(fscal,dx01);
1099 ty = _mm256_mul_ps(fscal,dy01);
1100 tz = _mm256_mul_ps(fscal,dz01);
1102 /* Update vectorial force */
1103 fix0 = _mm256_add_ps(fix0,tx);
1104 fiy0 = _mm256_add_ps(fiy0,ty);
1105 fiz0 = _mm256_add_ps(fiz0,tz);
1107 fjx1 = _mm256_add_ps(fjx1,tx);
1108 fjy1 = _mm256_add_ps(fjy1,ty);
1109 fjz1 = _mm256_add_ps(fjz1,tz);
1113 /**************************
1114 * CALCULATE INTERACTIONS *
1115 **************************/
1117 if (gmx_mm256_any_lt(rsq02,rcutoff2))
1120 r02 = _mm256_mul_ps(rsq02,rinv02);
1121 r02 = _mm256_andnot_ps(dummy_mask,r02);
1123 /* EWALD ELECTROSTATICS */
1125 /* Analytical PME correction */
1126 zeta2 = _mm256_mul_ps(beta2,rsq02);
1127 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
1128 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1129 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1130 felec = _mm256_mul_ps(qq02,felec);
1131 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1132 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1133 velec = _mm256_sub_ps(rinv02,pmecorrV);
1134 velec = _mm256_mul_ps(qq02,velec);
1136 d = _mm256_sub_ps(r02,rswitch);
1137 d = _mm256_max_ps(d,_mm256_setzero_ps());
1138 d2 = _mm256_mul_ps(d,d);
1139 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)))))));
1141 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1143 /* Evaluate switch function */
1144 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1145 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
1146 velec = _mm256_mul_ps(velec,sw);
1147 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
1149 /* Update potential sum for this i atom from the interaction with this j atom. */
1150 velec = _mm256_and_ps(velec,cutoff_mask);
1151 velec = _mm256_andnot_ps(dummy_mask,velec);
1152 velecsum = _mm256_add_ps(velecsum,velec);
1156 fscal = _mm256_and_ps(fscal,cutoff_mask);
1158 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1160 /* Calculate temporary vectorial force */
1161 tx = _mm256_mul_ps(fscal,dx02);
1162 ty = _mm256_mul_ps(fscal,dy02);
1163 tz = _mm256_mul_ps(fscal,dz02);
1165 /* Update vectorial force */
1166 fix0 = _mm256_add_ps(fix0,tx);
1167 fiy0 = _mm256_add_ps(fiy0,ty);
1168 fiz0 = _mm256_add_ps(fiz0,tz);
1170 fjx2 = _mm256_add_ps(fjx2,tx);
1171 fjy2 = _mm256_add_ps(fjy2,ty);
1172 fjz2 = _mm256_add_ps(fjz2,tz);
1176 /**************************
1177 * CALCULATE INTERACTIONS *
1178 **************************/
1180 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1183 r10 = _mm256_mul_ps(rsq10,rinv10);
1184 r10 = _mm256_andnot_ps(dummy_mask,r10);
1186 /* EWALD ELECTROSTATICS */
1188 /* Analytical PME correction */
1189 zeta2 = _mm256_mul_ps(beta2,rsq10);
1190 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
1191 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1192 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1193 felec = _mm256_mul_ps(qq10,felec);
1194 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1195 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1196 velec = _mm256_sub_ps(rinv10,pmecorrV);
1197 velec = _mm256_mul_ps(qq10,velec);
1199 d = _mm256_sub_ps(r10,rswitch);
1200 d = _mm256_max_ps(d,_mm256_setzero_ps());
1201 d2 = _mm256_mul_ps(d,d);
1202 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)))))));
1204 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1206 /* Evaluate switch function */
1207 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1208 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
1209 velec = _mm256_mul_ps(velec,sw);
1210 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1212 /* Update potential sum for this i atom from the interaction with this j atom. */
1213 velec = _mm256_and_ps(velec,cutoff_mask);
1214 velec = _mm256_andnot_ps(dummy_mask,velec);
1215 velecsum = _mm256_add_ps(velecsum,velec);
1219 fscal = _mm256_and_ps(fscal,cutoff_mask);
1221 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1223 /* Calculate temporary vectorial force */
1224 tx = _mm256_mul_ps(fscal,dx10);
1225 ty = _mm256_mul_ps(fscal,dy10);
1226 tz = _mm256_mul_ps(fscal,dz10);
1228 /* Update vectorial force */
1229 fix1 = _mm256_add_ps(fix1,tx);
1230 fiy1 = _mm256_add_ps(fiy1,ty);
1231 fiz1 = _mm256_add_ps(fiz1,tz);
1233 fjx0 = _mm256_add_ps(fjx0,tx);
1234 fjy0 = _mm256_add_ps(fjy0,ty);
1235 fjz0 = _mm256_add_ps(fjz0,tz);
1239 /**************************
1240 * CALCULATE INTERACTIONS *
1241 **************************/
1243 if (gmx_mm256_any_lt(rsq11,rcutoff2))
1246 r11 = _mm256_mul_ps(rsq11,rinv11);
1247 r11 = _mm256_andnot_ps(dummy_mask,r11);
1249 /* EWALD ELECTROSTATICS */
1251 /* Analytical PME correction */
1252 zeta2 = _mm256_mul_ps(beta2,rsq11);
1253 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
1254 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1255 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1256 felec = _mm256_mul_ps(qq11,felec);
1257 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1258 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1259 velec = _mm256_sub_ps(rinv11,pmecorrV);
1260 velec = _mm256_mul_ps(qq11,velec);
1262 d = _mm256_sub_ps(r11,rswitch);
1263 d = _mm256_max_ps(d,_mm256_setzero_ps());
1264 d2 = _mm256_mul_ps(d,d);
1265 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)))))));
1267 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1269 /* Evaluate switch function */
1270 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1271 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
1272 velec = _mm256_mul_ps(velec,sw);
1273 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1275 /* Update potential sum for this i atom from the interaction with this j atom. */
1276 velec = _mm256_and_ps(velec,cutoff_mask);
1277 velec = _mm256_andnot_ps(dummy_mask,velec);
1278 velecsum = _mm256_add_ps(velecsum,velec);
1282 fscal = _mm256_and_ps(fscal,cutoff_mask);
1284 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1286 /* Calculate temporary vectorial force */
1287 tx = _mm256_mul_ps(fscal,dx11);
1288 ty = _mm256_mul_ps(fscal,dy11);
1289 tz = _mm256_mul_ps(fscal,dz11);
1291 /* Update vectorial force */
1292 fix1 = _mm256_add_ps(fix1,tx);
1293 fiy1 = _mm256_add_ps(fiy1,ty);
1294 fiz1 = _mm256_add_ps(fiz1,tz);
1296 fjx1 = _mm256_add_ps(fjx1,tx);
1297 fjy1 = _mm256_add_ps(fjy1,ty);
1298 fjz1 = _mm256_add_ps(fjz1,tz);
1302 /**************************
1303 * CALCULATE INTERACTIONS *
1304 **************************/
1306 if (gmx_mm256_any_lt(rsq12,rcutoff2))
1309 r12 = _mm256_mul_ps(rsq12,rinv12);
1310 r12 = _mm256_andnot_ps(dummy_mask,r12);
1312 /* EWALD ELECTROSTATICS */
1314 /* Analytical PME correction */
1315 zeta2 = _mm256_mul_ps(beta2,rsq12);
1316 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
1317 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1318 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1319 felec = _mm256_mul_ps(qq12,felec);
1320 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1321 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1322 velec = _mm256_sub_ps(rinv12,pmecorrV);
1323 velec = _mm256_mul_ps(qq12,velec);
1325 d = _mm256_sub_ps(r12,rswitch);
1326 d = _mm256_max_ps(d,_mm256_setzero_ps());
1327 d2 = _mm256_mul_ps(d,d);
1328 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)))))));
1330 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1332 /* Evaluate switch function */
1333 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1334 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
1335 velec = _mm256_mul_ps(velec,sw);
1336 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1338 /* Update potential sum for this i atom from the interaction with this j atom. */
1339 velec = _mm256_and_ps(velec,cutoff_mask);
1340 velec = _mm256_andnot_ps(dummy_mask,velec);
1341 velecsum = _mm256_add_ps(velecsum,velec);
1345 fscal = _mm256_and_ps(fscal,cutoff_mask);
1347 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1349 /* Calculate temporary vectorial force */
1350 tx = _mm256_mul_ps(fscal,dx12);
1351 ty = _mm256_mul_ps(fscal,dy12);
1352 tz = _mm256_mul_ps(fscal,dz12);
1354 /* Update vectorial force */
1355 fix1 = _mm256_add_ps(fix1,tx);
1356 fiy1 = _mm256_add_ps(fiy1,ty);
1357 fiz1 = _mm256_add_ps(fiz1,tz);
1359 fjx2 = _mm256_add_ps(fjx2,tx);
1360 fjy2 = _mm256_add_ps(fjy2,ty);
1361 fjz2 = _mm256_add_ps(fjz2,tz);
1365 /**************************
1366 * CALCULATE INTERACTIONS *
1367 **************************/
1369 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1372 r20 = _mm256_mul_ps(rsq20,rinv20);
1373 r20 = _mm256_andnot_ps(dummy_mask,r20);
1375 /* EWALD ELECTROSTATICS */
1377 /* Analytical PME correction */
1378 zeta2 = _mm256_mul_ps(beta2,rsq20);
1379 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1380 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1381 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1382 felec = _mm256_mul_ps(qq20,felec);
1383 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1384 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1385 velec = _mm256_sub_ps(rinv20,pmecorrV);
1386 velec = _mm256_mul_ps(qq20,velec);
1388 d = _mm256_sub_ps(r20,rswitch);
1389 d = _mm256_max_ps(d,_mm256_setzero_ps());
1390 d2 = _mm256_mul_ps(d,d);
1391 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)))))));
1393 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1395 /* Evaluate switch function */
1396 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1397 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
1398 velec = _mm256_mul_ps(velec,sw);
1399 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1401 /* Update potential sum for this i atom from the interaction with this j atom. */
1402 velec = _mm256_and_ps(velec,cutoff_mask);
1403 velec = _mm256_andnot_ps(dummy_mask,velec);
1404 velecsum = _mm256_add_ps(velecsum,velec);
1408 fscal = _mm256_and_ps(fscal,cutoff_mask);
1410 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1412 /* Calculate temporary vectorial force */
1413 tx = _mm256_mul_ps(fscal,dx20);
1414 ty = _mm256_mul_ps(fscal,dy20);
1415 tz = _mm256_mul_ps(fscal,dz20);
1417 /* Update vectorial force */
1418 fix2 = _mm256_add_ps(fix2,tx);
1419 fiy2 = _mm256_add_ps(fiy2,ty);
1420 fiz2 = _mm256_add_ps(fiz2,tz);
1422 fjx0 = _mm256_add_ps(fjx0,tx);
1423 fjy0 = _mm256_add_ps(fjy0,ty);
1424 fjz0 = _mm256_add_ps(fjz0,tz);
1428 /**************************
1429 * CALCULATE INTERACTIONS *
1430 **************************/
1432 if (gmx_mm256_any_lt(rsq21,rcutoff2))
1435 r21 = _mm256_mul_ps(rsq21,rinv21);
1436 r21 = _mm256_andnot_ps(dummy_mask,r21);
1438 /* EWALD ELECTROSTATICS */
1440 /* Analytical PME correction */
1441 zeta2 = _mm256_mul_ps(beta2,rsq21);
1442 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
1443 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1444 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1445 felec = _mm256_mul_ps(qq21,felec);
1446 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1447 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1448 velec = _mm256_sub_ps(rinv21,pmecorrV);
1449 velec = _mm256_mul_ps(qq21,velec);
1451 d = _mm256_sub_ps(r21,rswitch);
1452 d = _mm256_max_ps(d,_mm256_setzero_ps());
1453 d2 = _mm256_mul_ps(d,d);
1454 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)))))));
1456 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1458 /* Evaluate switch function */
1459 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1460 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
1461 velec = _mm256_mul_ps(velec,sw);
1462 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1464 /* Update potential sum for this i atom from the interaction with this j atom. */
1465 velec = _mm256_and_ps(velec,cutoff_mask);
1466 velec = _mm256_andnot_ps(dummy_mask,velec);
1467 velecsum = _mm256_add_ps(velecsum,velec);
1471 fscal = _mm256_and_ps(fscal,cutoff_mask);
1473 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1475 /* Calculate temporary vectorial force */
1476 tx = _mm256_mul_ps(fscal,dx21);
1477 ty = _mm256_mul_ps(fscal,dy21);
1478 tz = _mm256_mul_ps(fscal,dz21);
1480 /* Update vectorial force */
1481 fix2 = _mm256_add_ps(fix2,tx);
1482 fiy2 = _mm256_add_ps(fiy2,ty);
1483 fiz2 = _mm256_add_ps(fiz2,tz);
1485 fjx1 = _mm256_add_ps(fjx1,tx);
1486 fjy1 = _mm256_add_ps(fjy1,ty);
1487 fjz1 = _mm256_add_ps(fjz1,tz);
1491 /**************************
1492 * CALCULATE INTERACTIONS *
1493 **************************/
1495 if (gmx_mm256_any_lt(rsq22,rcutoff2))
1498 r22 = _mm256_mul_ps(rsq22,rinv22);
1499 r22 = _mm256_andnot_ps(dummy_mask,r22);
1501 /* EWALD ELECTROSTATICS */
1503 /* Analytical PME correction */
1504 zeta2 = _mm256_mul_ps(beta2,rsq22);
1505 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
1506 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1507 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1508 felec = _mm256_mul_ps(qq22,felec);
1509 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1510 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1511 velec = _mm256_sub_ps(rinv22,pmecorrV);
1512 velec = _mm256_mul_ps(qq22,velec);
1514 d = _mm256_sub_ps(r22,rswitch);
1515 d = _mm256_max_ps(d,_mm256_setzero_ps());
1516 d2 = _mm256_mul_ps(d,d);
1517 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)))))));
1519 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1521 /* Evaluate switch function */
1522 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1523 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
1524 velec = _mm256_mul_ps(velec,sw);
1525 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1527 /* Update potential sum for this i atom from the interaction with this j atom. */
1528 velec = _mm256_and_ps(velec,cutoff_mask);
1529 velec = _mm256_andnot_ps(dummy_mask,velec);
1530 velecsum = _mm256_add_ps(velecsum,velec);
1534 fscal = _mm256_and_ps(fscal,cutoff_mask);
1536 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1538 /* Calculate temporary vectorial force */
1539 tx = _mm256_mul_ps(fscal,dx22);
1540 ty = _mm256_mul_ps(fscal,dy22);
1541 tz = _mm256_mul_ps(fscal,dz22);
1543 /* Update vectorial force */
1544 fix2 = _mm256_add_ps(fix2,tx);
1545 fiy2 = _mm256_add_ps(fiy2,ty);
1546 fiz2 = _mm256_add_ps(fiz2,tz);
1548 fjx2 = _mm256_add_ps(fjx2,tx);
1549 fjy2 = _mm256_add_ps(fjy2,ty);
1550 fjz2 = _mm256_add_ps(fjz2,tz);
1554 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1555 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1556 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1557 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1558 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1559 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1560 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1561 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1563 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1564 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1566 /* Inner loop uses 981 flops */
1569 /* End of innermost loop */
1571 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1572 f+i_coord_offset,fshift+i_shift_offset);
1575 /* Update potential energies */
1576 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1578 /* Increment number of inner iterations */
1579 inneriter += j_index_end - j_index_start;
1581 /* Outer loop uses 19 flops */
1584 /* Increment number of outer iterations */
1587 /* Update outer/inner flops */
1589 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*981);
1592 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwNone_GeomW3W3_F_avx_256_single
1593 * Electrostatics interaction: Ewald
1594 * VdW interaction: None
1595 * Geometry: Water3-Water3
1596 * Calculate force/pot: Force
1599 nb_kernel_ElecEwSw_VdwNone_GeomW3W3_F_avx_256_single
1600 (t_nblist * gmx_restrict nlist,
1601 rvec * gmx_restrict xx,
1602 rvec * gmx_restrict ff,
1603 t_forcerec * gmx_restrict fr,
1604 t_mdatoms * gmx_restrict mdatoms,
1605 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1606 t_nrnb * gmx_restrict nrnb)
1608 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1609 * just 0 for non-waters.
1610 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1611 * jnr indices corresponding to data put in the four positions in the SIMD register.
1613 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1614 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1615 int jnrA,jnrB,jnrC,jnrD;
1616 int jnrE,jnrF,jnrG,jnrH;
1617 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1618 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1619 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1620 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1621 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1622 real rcutoff_scalar;
1623 real *shiftvec,*fshift,*x,*f;
1624 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1625 real scratch[4*DIM];
1626 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1627 real * vdwioffsetptr0;
1628 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1629 real * vdwioffsetptr1;
1630 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1631 real * vdwioffsetptr2;
1632 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1633 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1634 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1635 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1636 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1637 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1638 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1639 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1640 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1641 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1642 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1643 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1644 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1645 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1646 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1647 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1648 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
1651 __m128i ewitab_lo,ewitab_hi;
1652 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1653 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1655 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1656 real rswitch_scalar,d_scalar;
1657 __m256 dummy_mask,cutoff_mask;
1658 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1659 __m256 one = _mm256_set1_ps(1.0);
1660 __m256 two = _mm256_set1_ps(2.0);
1666 jindex = nlist->jindex;
1668 shiftidx = nlist->shift;
1670 shiftvec = fr->shift_vec[0];
1671 fshift = fr->fshift[0];
1672 facel = _mm256_set1_ps(fr->epsfac);
1673 charge = mdatoms->chargeA;
1675 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
1676 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1677 beta2 = _mm256_mul_ps(beta,beta);
1678 beta3 = _mm256_mul_ps(beta,beta2);
1680 ewtab = fr->ic->tabq_coul_FDV0;
1681 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
1682 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1684 /* Setup water-specific parameters */
1685 inr = nlist->iinr[0];
1686 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1687 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1688 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1690 jq0 = _mm256_set1_ps(charge[inr+0]);
1691 jq1 = _mm256_set1_ps(charge[inr+1]);
1692 jq2 = _mm256_set1_ps(charge[inr+2]);
1693 qq00 = _mm256_mul_ps(iq0,jq0);
1694 qq01 = _mm256_mul_ps(iq0,jq1);
1695 qq02 = _mm256_mul_ps(iq0,jq2);
1696 qq10 = _mm256_mul_ps(iq1,jq0);
1697 qq11 = _mm256_mul_ps(iq1,jq1);
1698 qq12 = _mm256_mul_ps(iq1,jq2);
1699 qq20 = _mm256_mul_ps(iq2,jq0);
1700 qq21 = _mm256_mul_ps(iq2,jq1);
1701 qq22 = _mm256_mul_ps(iq2,jq2);
1703 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1704 rcutoff_scalar = fr->rcoulomb;
1705 rcutoff = _mm256_set1_ps(rcutoff_scalar);
1706 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
1708 rswitch_scalar = fr->rcoulomb_switch;
1709 rswitch = _mm256_set1_ps(rswitch_scalar);
1710 /* Setup switch parameters */
1711 d_scalar = rcutoff_scalar-rswitch_scalar;
1712 d = _mm256_set1_ps(d_scalar);
1713 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
1714 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1715 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1716 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
1717 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1718 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1720 /* Avoid stupid compiler warnings */
1721 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1722 j_coord_offsetA = 0;
1723 j_coord_offsetB = 0;
1724 j_coord_offsetC = 0;
1725 j_coord_offsetD = 0;
1726 j_coord_offsetE = 0;
1727 j_coord_offsetF = 0;
1728 j_coord_offsetG = 0;
1729 j_coord_offsetH = 0;
1734 for(iidx=0;iidx<4*DIM;iidx++)
1736 scratch[iidx] = 0.0;
1739 /* Start outer loop over neighborlists */
1740 for(iidx=0; iidx<nri; iidx++)
1742 /* Load shift vector for this list */
1743 i_shift_offset = DIM*shiftidx[iidx];
1745 /* Load limits for loop over neighbors */
1746 j_index_start = jindex[iidx];
1747 j_index_end = jindex[iidx+1];
1749 /* Get outer coordinate index */
1751 i_coord_offset = DIM*inr;
1753 /* Load i particle coords and add shift vector */
1754 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1755 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1757 fix0 = _mm256_setzero_ps();
1758 fiy0 = _mm256_setzero_ps();
1759 fiz0 = _mm256_setzero_ps();
1760 fix1 = _mm256_setzero_ps();
1761 fiy1 = _mm256_setzero_ps();
1762 fiz1 = _mm256_setzero_ps();
1763 fix2 = _mm256_setzero_ps();
1764 fiy2 = _mm256_setzero_ps();
1765 fiz2 = _mm256_setzero_ps();
1767 /* Start inner kernel loop */
1768 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1771 /* Get j neighbor index, and coordinate index */
1773 jnrB = jjnr[jidx+1];
1774 jnrC = jjnr[jidx+2];
1775 jnrD = jjnr[jidx+3];
1776 jnrE = jjnr[jidx+4];
1777 jnrF = jjnr[jidx+5];
1778 jnrG = jjnr[jidx+6];
1779 jnrH = jjnr[jidx+7];
1780 j_coord_offsetA = DIM*jnrA;
1781 j_coord_offsetB = DIM*jnrB;
1782 j_coord_offsetC = DIM*jnrC;
1783 j_coord_offsetD = DIM*jnrD;
1784 j_coord_offsetE = DIM*jnrE;
1785 j_coord_offsetF = DIM*jnrF;
1786 j_coord_offsetG = DIM*jnrG;
1787 j_coord_offsetH = DIM*jnrH;
1789 /* load j atom coordinates */
1790 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1791 x+j_coord_offsetC,x+j_coord_offsetD,
1792 x+j_coord_offsetE,x+j_coord_offsetF,
1793 x+j_coord_offsetG,x+j_coord_offsetH,
1794 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1796 /* Calculate displacement vector */
1797 dx00 = _mm256_sub_ps(ix0,jx0);
1798 dy00 = _mm256_sub_ps(iy0,jy0);
1799 dz00 = _mm256_sub_ps(iz0,jz0);
1800 dx01 = _mm256_sub_ps(ix0,jx1);
1801 dy01 = _mm256_sub_ps(iy0,jy1);
1802 dz01 = _mm256_sub_ps(iz0,jz1);
1803 dx02 = _mm256_sub_ps(ix0,jx2);
1804 dy02 = _mm256_sub_ps(iy0,jy2);
1805 dz02 = _mm256_sub_ps(iz0,jz2);
1806 dx10 = _mm256_sub_ps(ix1,jx0);
1807 dy10 = _mm256_sub_ps(iy1,jy0);
1808 dz10 = _mm256_sub_ps(iz1,jz0);
1809 dx11 = _mm256_sub_ps(ix1,jx1);
1810 dy11 = _mm256_sub_ps(iy1,jy1);
1811 dz11 = _mm256_sub_ps(iz1,jz1);
1812 dx12 = _mm256_sub_ps(ix1,jx2);
1813 dy12 = _mm256_sub_ps(iy1,jy2);
1814 dz12 = _mm256_sub_ps(iz1,jz2);
1815 dx20 = _mm256_sub_ps(ix2,jx0);
1816 dy20 = _mm256_sub_ps(iy2,jy0);
1817 dz20 = _mm256_sub_ps(iz2,jz0);
1818 dx21 = _mm256_sub_ps(ix2,jx1);
1819 dy21 = _mm256_sub_ps(iy2,jy1);
1820 dz21 = _mm256_sub_ps(iz2,jz1);
1821 dx22 = _mm256_sub_ps(ix2,jx2);
1822 dy22 = _mm256_sub_ps(iy2,jy2);
1823 dz22 = _mm256_sub_ps(iz2,jz2);
1825 /* Calculate squared distance and things based on it */
1826 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1827 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1828 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1829 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1830 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1831 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1832 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1833 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1834 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1836 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1837 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1838 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1839 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1840 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1841 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1842 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1843 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1844 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1846 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1847 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1848 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1849 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1850 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1851 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1852 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1853 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1854 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1856 fjx0 = _mm256_setzero_ps();
1857 fjy0 = _mm256_setzero_ps();
1858 fjz0 = _mm256_setzero_ps();
1859 fjx1 = _mm256_setzero_ps();
1860 fjy1 = _mm256_setzero_ps();
1861 fjz1 = _mm256_setzero_ps();
1862 fjx2 = _mm256_setzero_ps();
1863 fjy2 = _mm256_setzero_ps();
1864 fjz2 = _mm256_setzero_ps();
1866 /**************************
1867 * CALCULATE INTERACTIONS *
1868 **************************/
1870 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1873 r00 = _mm256_mul_ps(rsq00,rinv00);
1875 /* EWALD ELECTROSTATICS */
1877 /* Analytical PME correction */
1878 zeta2 = _mm256_mul_ps(beta2,rsq00);
1879 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1880 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1881 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1882 felec = _mm256_mul_ps(qq00,felec);
1883 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1884 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1885 velec = _mm256_sub_ps(rinv00,pmecorrV);
1886 velec = _mm256_mul_ps(qq00,velec);
1888 d = _mm256_sub_ps(r00,rswitch);
1889 d = _mm256_max_ps(d,_mm256_setzero_ps());
1890 d2 = _mm256_mul_ps(d,d);
1891 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)))))));
1893 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1895 /* Evaluate switch function */
1896 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1897 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1898 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1902 fscal = _mm256_and_ps(fscal,cutoff_mask);
1904 /* Calculate temporary vectorial force */
1905 tx = _mm256_mul_ps(fscal,dx00);
1906 ty = _mm256_mul_ps(fscal,dy00);
1907 tz = _mm256_mul_ps(fscal,dz00);
1909 /* Update vectorial force */
1910 fix0 = _mm256_add_ps(fix0,tx);
1911 fiy0 = _mm256_add_ps(fiy0,ty);
1912 fiz0 = _mm256_add_ps(fiz0,tz);
1914 fjx0 = _mm256_add_ps(fjx0,tx);
1915 fjy0 = _mm256_add_ps(fjy0,ty);
1916 fjz0 = _mm256_add_ps(fjz0,tz);
1920 /**************************
1921 * CALCULATE INTERACTIONS *
1922 **************************/
1924 if (gmx_mm256_any_lt(rsq01,rcutoff2))
1927 r01 = _mm256_mul_ps(rsq01,rinv01);
1929 /* EWALD ELECTROSTATICS */
1931 /* Analytical PME correction */
1932 zeta2 = _mm256_mul_ps(beta2,rsq01);
1933 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1934 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1935 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1936 felec = _mm256_mul_ps(qq01,felec);
1937 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1938 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1939 velec = _mm256_sub_ps(rinv01,pmecorrV);
1940 velec = _mm256_mul_ps(qq01,velec);
1942 d = _mm256_sub_ps(r01,rswitch);
1943 d = _mm256_max_ps(d,_mm256_setzero_ps());
1944 d2 = _mm256_mul_ps(d,d);
1945 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)))))));
1947 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1949 /* Evaluate switch function */
1950 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1951 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
1952 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
1956 fscal = _mm256_and_ps(fscal,cutoff_mask);
1958 /* Calculate temporary vectorial force */
1959 tx = _mm256_mul_ps(fscal,dx01);
1960 ty = _mm256_mul_ps(fscal,dy01);
1961 tz = _mm256_mul_ps(fscal,dz01);
1963 /* Update vectorial force */
1964 fix0 = _mm256_add_ps(fix0,tx);
1965 fiy0 = _mm256_add_ps(fiy0,ty);
1966 fiz0 = _mm256_add_ps(fiz0,tz);
1968 fjx1 = _mm256_add_ps(fjx1,tx);
1969 fjy1 = _mm256_add_ps(fjy1,ty);
1970 fjz1 = _mm256_add_ps(fjz1,tz);
1974 /**************************
1975 * CALCULATE INTERACTIONS *
1976 **************************/
1978 if (gmx_mm256_any_lt(rsq02,rcutoff2))
1981 r02 = _mm256_mul_ps(rsq02,rinv02);
1983 /* EWALD ELECTROSTATICS */
1985 /* Analytical PME correction */
1986 zeta2 = _mm256_mul_ps(beta2,rsq02);
1987 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
1988 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1989 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1990 felec = _mm256_mul_ps(qq02,felec);
1991 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1992 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1993 velec = _mm256_sub_ps(rinv02,pmecorrV);
1994 velec = _mm256_mul_ps(qq02,velec);
1996 d = _mm256_sub_ps(r02,rswitch);
1997 d = _mm256_max_ps(d,_mm256_setzero_ps());
1998 d2 = _mm256_mul_ps(d,d);
1999 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)))))));
2001 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2003 /* Evaluate switch function */
2004 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2005 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
2006 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
2010 fscal = _mm256_and_ps(fscal,cutoff_mask);
2012 /* Calculate temporary vectorial force */
2013 tx = _mm256_mul_ps(fscal,dx02);
2014 ty = _mm256_mul_ps(fscal,dy02);
2015 tz = _mm256_mul_ps(fscal,dz02);
2017 /* Update vectorial force */
2018 fix0 = _mm256_add_ps(fix0,tx);
2019 fiy0 = _mm256_add_ps(fiy0,ty);
2020 fiz0 = _mm256_add_ps(fiz0,tz);
2022 fjx2 = _mm256_add_ps(fjx2,tx);
2023 fjy2 = _mm256_add_ps(fjy2,ty);
2024 fjz2 = _mm256_add_ps(fjz2,tz);
2028 /**************************
2029 * CALCULATE INTERACTIONS *
2030 **************************/
2032 if (gmx_mm256_any_lt(rsq10,rcutoff2))
2035 r10 = _mm256_mul_ps(rsq10,rinv10);
2037 /* EWALD ELECTROSTATICS */
2039 /* Analytical PME correction */
2040 zeta2 = _mm256_mul_ps(beta2,rsq10);
2041 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2042 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2043 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2044 felec = _mm256_mul_ps(qq10,felec);
2045 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2046 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2047 velec = _mm256_sub_ps(rinv10,pmecorrV);
2048 velec = _mm256_mul_ps(qq10,velec);
2050 d = _mm256_sub_ps(r10,rswitch);
2051 d = _mm256_max_ps(d,_mm256_setzero_ps());
2052 d2 = _mm256_mul_ps(d,d);
2053 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)))))));
2055 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2057 /* Evaluate switch function */
2058 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2059 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
2060 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
2064 fscal = _mm256_and_ps(fscal,cutoff_mask);
2066 /* Calculate temporary vectorial force */
2067 tx = _mm256_mul_ps(fscal,dx10);
2068 ty = _mm256_mul_ps(fscal,dy10);
2069 tz = _mm256_mul_ps(fscal,dz10);
2071 /* Update vectorial force */
2072 fix1 = _mm256_add_ps(fix1,tx);
2073 fiy1 = _mm256_add_ps(fiy1,ty);
2074 fiz1 = _mm256_add_ps(fiz1,tz);
2076 fjx0 = _mm256_add_ps(fjx0,tx);
2077 fjy0 = _mm256_add_ps(fjy0,ty);
2078 fjz0 = _mm256_add_ps(fjz0,tz);
2082 /**************************
2083 * CALCULATE INTERACTIONS *
2084 **************************/
2086 if (gmx_mm256_any_lt(rsq11,rcutoff2))
2089 r11 = _mm256_mul_ps(rsq11,rinv11);
2091 /* EWALD ELECTROSTATICS */
2093 /* Analytical PME correction */
2094 zeta2 = _mm256_mul_ps(beta2,rsq11);
2095 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2096 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2097 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2098 felec = _mm256_mul_ps(qq11,felec);
2099 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2100 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2101 velec = _mm256_sub_ps(rinv11,pmecorrV);
2102 velec = _mm256_mul_ps(qq11,velec);
2104 d = _mm256_sub_ps(r11,rswitch);
2105 d = _mm256_max_ps(d,_mm256_setzero_ps());
2106 d2 = _mm256_mul_ps(d,d);
2107 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)))))));
2109 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2111 /* Evaluate switch function */
2112 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2113 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
2114 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
2118 fscal = _mm256_and_ps(fscal,cutoff_mask);
2120 /* Calculate temporary vectorial force */
2121 tx = _mm256_mul_ps(fscal,dx11);
2122 ty = _mm256_mul_ps(fscal,dy11);
2123 tz = _mm256_mul_ps(fscal,dz11);
2125 /* Update vectorial force */
2126 fix1 = _mm256_add_ps(fix1,tx);
2127 fiy1 = _mm256_add_ps(fiy1,ty);
2128 fiz1 = _mm256_add_ps(fiz1,tz);
2130 fjx1 = _mm256_add_ps(fjx1,tx);
2131 fjy1 = _mm256_add_ps(fjy1,ty);
2132 fjz1 = _mm256_add_ps(fjz1,tz);
2136 /**************************
2137 * CALCULATE INTERACTIONS *
2138 **************************/
2140 if (gmx_mm256_any_lt(rsq12,rcutoff2))
2143 r12 = _mm256_mul_ps(rsq12,rinv12);
2145 /* EWALD ELECTROSTATICS */
2147 /* Analytical PME correction */
2148 zeta2 = _mm256_mul_ps(beta2,rsq12);
2149 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2150 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2151 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2152 felec = _mm256_mul_ps(qq12,felec);
2153 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2154 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2155 velec = _mm256_sub_ps(rinv12,pmecorrV);
2156 velec = _mm256_mul_ps(qq12,velec);
2158 d = _mm256_sub_ps(r12,rswitch);
2159 d = _mm256_max_ps(d,_mm256_setzero_ps());
2160 d2 = _mm256_mul_ps(d,d);
2161 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)))))));
2163 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2165 /* Evaluate switch function */
2166 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2167 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
2168 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
2172 fscal = _mm256_and_ps(fscal,cutoff_mask);
2174 /* Calculate temporary vectorial force */
2175 tx = _mm256_mul_ps(fscal,dx12);
2176 ty = _mm256_mul_ps(fscal,dy12);
2177 tz = _mm256_mul_ps(fscal,dz12);
2179 /* Update vectorial force */
2180 fix1 = _mm256_add_ps(fix1,tx);
2181 fiy1 = _mm256_add_ps(fiy1,ty);
2182 fiz1 = _mm256_add_ps(fiz1,tz);
2184 fjx2 = _mm256_add_ps(fjx2,tx);
2185 fjy2 = _mm256_add_ps(fjy2,ty);
2186 fjz2 = _mm256_add_ps(fjz2,tz);
2190 /**************************
2191 * CALCULATE INTERACTIONS *
2192 **************************/
2194 if (gmx_mm256_any_lt(rsq20,rcutoff2))
2197 r20 = _mm256_mul_ps(rsq20,rinv20);
2199 /* EWALD ELECTROSTATICS */
2201 /* Analytical PME correction */
2202 zeta2 = _mm256_mul_ps(beta2,rsq20);
2203 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
2204 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2205 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2206 felec = _mm256_mul_ps(qq20,felec);
2207 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2208 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2209 velec = _mm256_sub_ps(rinv20,pmecorrV);
2210 velec = _mm256_mul_ps(qq20,velec);
2212 d = _mm256_sub_ps(r20,rswitch);
2213 d = _mm256_max_ps(d,_mm256_setzero_ps());
2214 d2 = _mm256_mul_ps(d,d);
2215 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)))))));
2217 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2219 /* Evaluate switch function */
2220 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2221 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
2222 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
2226 fscal = _mm256_and_ps(fscal,cutoff_mask);
2228 /* Calculate temporary vectorial force */
2229 tx = _mm256_mul_ps(fscal,dx20);
2230 ty = _mm256_mul_ps(fscal,dy20);
2231 tz = _mm256_mul_ps(fscal,dz20);
2233 /* Update vectorial force */
2234 fix2 = _mm256_add_ps(fix2,tx);
2235 fiy2 = _mm256_add_ps(fiy2,ty);
2236 fiz2 = _mm256_add_ps(fiz2,tz);
2238 fjx0 = _mm256_add_ps(fjx0,tx);
2239 fjy0 = _mm256_add_ps(fjy0,ty);
2240 fjz0 = _mm256_add_ps(fjz0,tz);
2244 /**************************
2245 * CALCULATE INTERACTIONS *
2246 **************************/
2248 if (gmx_mm256_any_lt(rsq21,rcutoff2))
2251 r21 = _mm256_mul_ps(rsq21,rinv21);
2253 /* EWALD ELECTROSTATICS */
2255 /* Analytical PME correction */
2256 zeta2 = _mm256_mul_ps(beta2,rsq21);
2257 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2258 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2259 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2260 felec = _mm256_mul_ps(qq21,felec);
2261 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2262 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2263 velec = _mm256_sub_ps(rinv21,pmecorrV);
2264 velec = _mm256_mul_ps(qq21,velec);
2266 d = _mm256_sub_ps(r21,rswitch);
2267 d = _mm256_max_ps(d,_mm256_setzero_ps());
2268 d2 = _mm256_mul_ps(d,d);
2269 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)))))));
2271 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2273 /* Evaluate switch function */
2274 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2275 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
2276 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2280 fscal = _mm256_and_ps(fscal,cutoff_mask);
2282 /* Calculate temporary vectorial force */
2283 tx = _mm256_mul_ps(fscal,dx21);
2284 ty = _mm256_mul_ps(fscal,dy21);
2285 tz = _mm256_mul_ps(fscal,dz21);
2287 /* Update vectorial force */
2288 fix2 = _mm256_add_ps(fix2,tx);
2289 fiy2 = _mm256_add_ps(fiy2,ty);
2290 fiz2 = _mm256_add_ps(fiz2,tz);
2292 fjx1 = _mm256_add_ps(fjx1,tx);
2293 fjy1 = _mm256_add_ps(fjy1,ty);
2294 fjz1 = _mm256_add_ps(fjz1,tz);
2298 /**************************
2299 * CALCULATE INTERACTIONS *
2300 **************************/
2302 if (gmx_mm256_any_lt(rsq22,rcutoff2))
2305 r22 = _mm256_mul_ps(rsq22,rinv22);
2307 /* EWALD ELECTROSTATICS */
2309 /* Analytical PME correction */
2310 zeta2 = _mm256_mul_ps(beta2,rsq22);
2311 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
2312 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2313 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2314 felec = _mm256_mul_ps(qq22,felec);
2315 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2316 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2317 velec = _mm256_sub_ps(rinv22,pmecorrV);
2318 velec = _mm256_mul_ps(qq22,velec);
2320 d = _mm256_sub_ps(r22,rswitch);
2321 d = _mm256_max_ps(d,_mm256_setzero_ps());
2322 d2 = _mm256_mul_ps(d,d);
2323 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)))))));
2325 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2327 /* Evaluate switch function */
2328 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2329 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
2330 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
2334 fscal = _mm256_and_ps(fscal,cutoff_mask);
2336 /* Calculate temporary vectorial force */
2337 tx = _mm256_mul_ps(fscal,dx22);
2338 ty = _mm256_mul_ps(fscal,dy22);
2339 tz = _mm256_mul_ps(fscal,dz22);
2341 /* Update vectorial force */
2342 fix2 = _mm256_add_ps(fix2,tx);
2343 fiy2 = _mm256_add_ps(fiy2,ty);
2344 fiz2 = _mm256_add_ps(fiz2,tz);
2346 fjx2 = _mm256_add_ps(fjx2,tx);
2347 fjy2 = _mm256_add_ps(fjy2,ty);
2348 fjz2 = _mm256_add_ps(fjz2,tz);
2352 fjptrA = f+j_coord_offsetA;
2353 fjptrB = f+j_coord_offsetB;
2354 fjptrC = f+j_coord_offsetC;
2355 fjptrD = f+j_coord_offsetD;
2356 fjptrE = f+j_coord_offsetE;
2357 fjptrF = f+j_coord_offsetF;
2358 fjptrG = f+j_coord_offsetG;
2359 fjptrH = f+j_coord_offsetH;
2361 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2362 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2364 /* Inner loop uses 945 flops */
2367 if(jidx<j_index_end)
2370 /* Get j neighbor index, and coordinate index */
2371 jnrlistA = jjnr[jidx];
2372 jnrlistB = jjnr[jidx+1];
2373 jnrlistC = jjnr[jidx+2];
2374 jnrlistD = jjnr[jidx+3];
2375 jnrlistE = jjnr[jidx+4];
2376 jnrlistF = jjnr[jidx+5];
2377 jnrlistG = jjnr[jidx+6];
2378 jnrlistH = jjnr[jidx+7];
2379 /* Sign of each element will be negative for non-real atoms.
2380 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2381 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2383 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
2384 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
2386 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
2387 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
2388 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
2389 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
2390 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
2391 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
2392 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
2393 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
2394 j_coord_offsetA = DIM*jnrA;
2395 j_coord_offsetB = DIM*jnrB;
2396 j_coord_offsetC = DIM*jnrC;
2397 j_coord_offsetD = DIM*jnrD;
2398 j_coord_offsetE = DIM*jnrE;
2399 j_coord_offsetF = DIM*jnrF;
2400 j_coord_offsetG = DIM*jnrG;
2401 j_coord_offsetH = DIM*jnrH;
2403 /* load j atom coordinates */
2404 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2405 x+j_coord_offsetC,x+j_coord_offsetD,
2406 x+j_coord_offsetE,x+j_coord_offsetF,
2407 x+j_coord_offsetG,x+j_coord_offsetH,
2408 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2410 /* Calculate displacement vector */
2411 dx00 = _mm256_sub_ps(ix0,jx0);
2412 dy00 = _mm256_sub_ps(iy0,jy0);
2413 dz00 = _mm256_sub_ps(iz0,jz0);
2414 dx01 = _mm256_sub_ps(ix0,jx1);
2415 dy01 = _mm256_sub_ps(iy0,jy1);
2416 dz01 = _mm256_sub_ps(iz0,jz1);
2417 dx02 = _mm256_sub_ps(ix0,jx2);
2418 dy02 = _mm256_sub_ps(iy0,jy2);
2419 dz02 = _mm256_sub_ps(iz0,jz2);
2420 dx10 = _mm256_sub_ps(ix1,jx0);
2421 dy10 = _mm256_sub_ps(iy1,jy0);
2422 dz10 = _mm256_sub_ps(iz1,jz0);
2423 dx11 = _mm256_sub_ps(ix1,jx1);
2424 dy11 = _mm256_sub_ps(iy1,jy1);
2425 dz11 = _mm256_sub_ps(iz1,jz1);
2426 dx12 = _mm256_sub_ps(ix1,jx2);
2427 dy12 = _mm256_sub_ps(iy1,jy2);
2428 dz12 = _mm256_sub_ps(iz1,jz2);
2429 dx20 = _mm256_sub_ps(ix2,jx0);
2430 dy20 = _mm256_sub_ps(iy2,jy0);
2431 dz20 = _mm256_sub_ps(iz2,jz0);
2432 dx21 = _mm256_sub_ps(ix2,jx1);
2433 dy21 = _mm256_sub_ps(iy2,jy1);
2434 dz21 = _mm256_sub_ps(iz2,jz1);
2435 dx22 = _mm256_sub_ps(ix2,jx2);
2436 dy22 = _mm256_sub_ps(iy2,jy2);
2437 dz22 = _mm256_sub_ps(iz2,jz2);
2439 /* Calculate squared distance and things based on it */
2440 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
2441 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
2442 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
2443 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
2444 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
2445 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
2446 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
2447 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
2448 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
2450 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
2451 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
2452 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
2453 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
2454 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
2455 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
2456 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
2457 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
2458 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
2460 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
2461 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
2462 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
2463 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
2464 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
2465 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
2466 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
2467 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
2468 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
2470 fjx0 = _mm256_setzero_ps();
2471 fjy0 = _mm256_setzero_ps();
2472 fjz0 = _mm256_setzero_ps();
2473 fjx1 = _mm256_setzero_ps();
2474 fjy1 = _mm256_setzero_ps();
2475 fjz1 = _mm256_setzero_ps();
2476 fjx2 = _mm256_setzero_ps();
2477 fjy2 = _mm256_setzero_ps();
2478 fjz2 = _mm256_setzero_ps();
2480 /**************************
2481 * CALCULATE INTERACTIONS *
2482 **************************/
2484 if (gmx_mm256_any_lt(rsq00,rcutoff2))
2487 r00 = _mm256_mul_ps(rsq00,rinv00);
2488 r00 = _mm256_andnot_ps(dummy_mask,r00);
2490 /* EWALD ELECTROSTATICS */
2492 /* Analytical PME correction */
2493 zeta2 = _mm256_mul_ps(beta2,rsq00);
2494 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
2495 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2496 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2497 felec = _mm256_mul_ps(qq00,felec);
2498 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2499 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2500 velec = _mm256_sub_ps(rinv00,pmecorrV);
2501 velec = _mm256_mul_ps(qq00,velec);
2503 d = _mm256_sub_ps(r00,rswitch);
2504 d = _mm256_max_ps(d,_mm256_setzero_ps());
2505 d2 = _mm256_mul_ps(d,d);
2506 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)))))));
2508 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2510 /* Evaluate switch function */
2511 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2512 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
2513 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
2517 fscal = _mm256_and_ps(fscal,cutoff_mask);
2519 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2521 /* Calculate temporary vectorial force */
2522 tx = _mm256_mul_ps(fscal,dx00);
2523 ty = _mm256_mul_ps(fscal,dy00);
2524 tz = _mm256_mul_ps(fscal,dz00);
2526 /* Update vectorial force */
2527 fix0 = _mm256_add_ps(fix0,tx);
2528 fiy0 = _mm256_add_ps(fiy0,ty);
2529 fiz0 = _mm256_add_ps(fiz0,tz);
2531 fjx0 = _mm256_add_ps(fjx0,tx);
2532 fjy0 = _mm256_add_ps(fjy0,ty);
2533 fjz0 = _mm256_add_ps(fjz0,tz);
2537 /**************************
2538 * CALCULATE INTERACTIONS *
2539 **************************/
2541 if (gmx_mm256_any_lt(rsq01,rcutoff2))
2544 r01 = _mm256_mul_ps(rsq01,rinv01);
2545 r01 = _mm256_andnot_ps(dummy_mask,r01);
2547 /* EWALD ELECTROSTATICS */
2549 /* Analytical PME correction */
2550 zeta2 = _mm256_mul_ps(beta2,rsq01);
2551 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
2552 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2553 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2554 felec = _mm256_mul_ps(qq01,felec);
2555 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2556 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2557 velec = _mm256_sub_ps(rinv01,pmecorrV);
2558 velec = _mm256_mul_ps(qq01,velec);
2560 d = _mm256_sub_ps(r01,rswitch);
2561 d = _mm256_max_ps(d,_mm256_setzero_ps());
2562 d2 = _mm256_mul_ps(d,d);
2563 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)))))));
2565 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2567 /* Evaluate switch function */
2568 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2569 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
2570 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
2574 fscal = _mm256_and_ps(fscal,cutoff_mask);
2576 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2578 /* Calculate temporary vectorial force */
2579 tx = _mm256_mul_ps(fscal,dx01);
2580 ty = _mm256_mul_ps(fscal,dy01);
2581 tz = _mm256_mul_ps(fscal,dz01);
2583 /* Update vectorial force */
2584 fix0 = _mm256_add_ps(fix0,tx);
2585 fiy0 = _mm256_add_ps(fiy0,ty);
2586 fiz0 = _mm256_add_ps(fiz0,tz);
2588 fjx1 = _mm256_add_ps(fjx1,tx);
2589 fjy1 = _mm256_add_ps(fjy1,ty);
2590 fjz1 = _mm256_add_ps(fjz1,tz);
2594 /**************************
2595 * CALCULATE INTERACTIONS *
2596 **************************/
2598 if (gmx_mm256_any_lt(rsq02,rcutoff2))
2601 r02 = _mm256_mul_ps(rsq02,rinv02);
2602 r02 = _mm256_andnot_ps(dummy_mask,r02);
2604 /* EWALD ELECTROSTATICS */
2606 /* Analytical PME correction */
2607 zeta2 = _mm256_mul_ps(beta2,rsq02);
2608 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
2609 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2610 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2611 felec = _mm256_mul_ps(qq02,felec);
2612 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2613 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2614 velec = _mm256_sub_ps(rinv02,pmecorrV);
2615 velec = _mm256_mul_ps(qq02,velec);
2617 d = _mm256_sub_ps(r02,rswitch);
2618 d = _mm256_max_ps(d,_mm256_setzero_ps());
2619 d2 = _mm256_mul_ps(d,d);
2620 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)))))));
2622 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2624 /* Evaluate switch function */
2625 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2626 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
2627 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
2631 fscal = _mm256_and_ps(fscal,cutoff_mask);
2633 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2635 /* Calculate temporary vectorial force */
2636 tx = _mm256_mul_ps(fscal,dx02);
2637 ty = _mm256_mul_ps(fscal,dy02);
2638 tz = _mm256_mul_ps(fscal,dz02);
2640 /* Update vectorial force */
2641 fix0 = _mm256_add_ps(fix0,tx);
2642 fiy0 = _mm256_add_ps(fiy0,ty);
2643 fiz0 = _mm256_add_ps(fiz0,tz);
2645 fjx2 = _mm256_add_ps(fjx2,tx);
2646 fjy2 = _mm256_add_ps(fjy2,ty);
2647 fjz2 = _mm256_add_ps(fjz2,tz);
2651 /**************************
2652 * CALCULATE INTERACTIONS *
2653 **************************/
2655 if (gmx_mm256_any_lt(rsq10,rcutoff2))
2658 r10 = _mm256_mul_ps(rsq10,rinv10);
2659 r10 = _mm256_andnot_ps(dummy_mask,r10);
2661 /* EWALD ELECTROSTATICS */
2663 /* Analytical PME correction */
2664 zeta2 = _mm256_mul_ps(beta2,rsq10);
2665 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2666 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2667 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2668 felec = _mm256_mul_ps(qq10,felec);
2669 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2670 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2671 velec = _mm256_sub_ps(rinv10,pmecorrV);
2672 velec = _mm256_mul_ps(qq10,velec);
2674 d = _mm256_sub_ps(r10,rswitch);
2675 d = _mm256_max_ps(d,_mm256_setzero_ps());
2676 d2 = _mm256_mul_ps(d,d);
2677 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)))))));
2679 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2681 /* Evaluate switch function */
2682 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2683 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
2684 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
2688 fscal = _mm256_and_ps(fscal,cutoff_mask);
2690 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2692 /* Calculate temporary vectorial force */
2693 tx = _mm256_mul_ps(fscal,dx10);
2694 ty = _mm256_mul_ps(fscal,dy10);
2695 tz = _mm256_mul_ps(fscal,dz10);
2697 /* Update vectorial force */
2698 fix1 = _mm256_add_ps(fix1,tx);
2699 fiy1 = _mm256_add_ps(fiy1,ty);
2700 fiz1 = _mm256_add_ps(fiz1,tz);
2702 fjx0 = _mm256_add_ps(fjx0,tx);
2703 fjy0 = _mm256_add_ps(fjy0,ty);
2704 fjz0 = _mm256_add_ps(fjz0,tz);
2708 /**************************
2709 * CALCULATE INTERACTIONS *
2710 **************************/
2712 if (gmx_mm256_any_lt(rsq11,rcutoff2))
2715 r11 = _mm256_mul_ps(rsq11,rinv11);
2716 r11 = _mm256_andnot_ps(dummy_mask,r11);
2718 /* EWALD ELECTROSTATICS */
2720 /* Analytical PME correction */
2721 zeta2 = _mm256_mul_ps(beta2,rsq11);
2722 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2723 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2724 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2725 felec = _mm256_mul_ps(qq11,felec);
2726 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2727 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2728 velec = _mm256_sub_ps(rinv11,pmecorrV);
2729 velec = _mm256_mul_ps(qq11,velec);
2731 d = _mm256_sub_ps(r11,rswitch);
2732 d = _mm256_max_ps(d,_mm256_setzero_ps());
2733 d2 = _mm256_mul_ps(d,d);
2734 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)))))));
2736 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2738 /* Evaluate switch function */
2739 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2740 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
2741 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
2745 fscal = _mm256_and_ps(fscal,cutoff_mask);
2747 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2749 /* Calculate temporary vectorial force */
2750 tx = _mm256_mul_ps(fscal,dx11);
2751 ty = _mm256_mul_ps(fscal,dy11);
2752 tz = _mm256_mul_ps(fscal,dz11);
2754 /* Update vectorial force */
2755 fix1 = _mm256_add_ps(fix1,tx);
2756 fiy1 = _mm256_add_ps(fiy1,ty);
2757 fiz1 = _mm256_add_ps(fiz1,tz);
2759 fjx1 = _mm256_add_ps(fjx1,tx);
2760 fjy1 = _mm256_add_ps(fjy1,ty);
2761 fjz1 = _mm256_add_ps(fjz1,tz);
2765 /**************************
2766 * CALCULATE INTERACTIONS *
2767 **************************/
2769 if (gmx_mm256_any_lt(rsq12,rcutoff2))
2772 r12 = _mm256_mul_ps(rsq12,rinv12);
2773 r12 = _mm256_andnot_ps(dummy_mask,r12);
2775 /* EWALD ELECTROSTATICS */
2777 /* Analytical PME correction */
2778 zeta2 = _mm256_mul_ps(beta2,rsq12);
2779 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2780 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2781 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2782 felec = _mm256_mul_ps(qq12,felec);
2783 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2784 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2785 velec = _mm256_sub_ps(rinv12,pmecorrV);
2786 velec = _mm256_mul_ps(qq12,velec);
2788 d = _mm256_sub_ps(r12,rswitch);
2789 d = _mm256_max_ps(d,_mm256_setzero_ps());
2790 d2 = _mm256_mul_ps(d,d);
2791 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)))))));
2793 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2795 /* Evaluate switch function */
2796 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2797 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
2798 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
2802 fscal = _mm256_and_ps(fscal,cutoff_mask);
2804 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2806 /* Calculate temporary vectorial force */
2807 tx = _mm256_mul_ps(fscal,dx12);
2808 ty = _mm256_mul_ps(fscal,dy12);
2809 tz = _mm256_mul_ps(fscal,dz12);
2811 /* Update vectorial force */
2812 fix1 = _mm256_add_ps(fix1,tx);
2813 fiy1 = _mm256_add_ps(fiy1,ty);
2814 fiz1 = _mm256_add_ps(fiz1,tz);
2816 fjx2 = _mm256_add_ps(fjx2,tx);
2817 fjy2 = _mm256_add_ps(fjy2,ty);
2818 fjz2 = _mm256_add_ps(fjz2,tz);
2822 /**************************
2823 * CALCULATE INTERACTIONS *
2824 **************************/
2826 if (gmx_mm256_any_lt(rsq20,rcutoff2))
2829 r20 = _mm256_mul_ps(rsq20,rinv20);
2830 r20 = _mm256_andnot_ps(dummy_mask,r20);
2832 /* EWALD ELECTROSTATICS */
2834 /* Analytical PME correction */
2835 zeta2 = _mm256_mul_ps(beta2,rsq20);
2836 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
2837 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2838 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2839 felec = _mm256_mul_ps(qq20,felec);
2840 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2841 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2842 velec = _mm256_sub_ps(rinv20,pmecorrV);
2843 velec = _mm256_mul_ps(qq20,velec);
2845 d = _mm256_sub_ps(r20,rswitch);
2846 d = _mm256_max_ps(d,_mm256_setzero_ps());
2847 d2 = _mm256_mul_ps(d,d);
2848 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)))))));
2850 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2852 /* Evaluate switch function */
2853 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2854 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
2855 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
2859 fscal = _mm256_and_ps(fscal,cutoff_mask);
2861 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2863 /* Calculate temporary vectorial force */
2864 tx = _mm256_mul_ps(fscal,dx20);
2865 ty = _mm256_mul_ps(fscal,dy20);
2866 tz = _mm256_mul_ps(fscal,dz20);
2868 /* Update vectorial force */
2869 fix2 = _mm256_add_ps(fix2,tx);
2870 fiy2 = _mm256_add_ps(fiy2,ty);
2871 fiz2 = _mm256_add_ps(fiz2,tz);
2873 fjx0 = _mm256_add_ps(fjx0,tx);
2874 fjy0 = _mm256_add_ps(fjy0,ty);
2875 fjz0 = _mm256_add_ps(fjz0,tz);
2879 /**************************
2880 * CALCULATE INTERACTIONS *
2881 **************************/
2883 if (gmx_mm256_any_lt(rsq21,rcutoff2))
2886 r21 = _mm256_mul_ps(rsq21,rinv21);
2887 r21 = _mm256_andnot_ps(dummy_mask,r21);
2889 /* EWALD ELECTROSTATICS */
2891 /* Analytical PME correction */
2892 zeta2 = _mm256_mul_ps(beta2,rsq21);
2893 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2894 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2895 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2896 felec = _mm256_mul_ps(qq21,felec);
2897 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2898 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2899 velec = _mm256_sub_ps(rinv21,pmecorrV);
2900 velec = _mm256_mul_ps(qq21,velec);
2902 d = _mm256_sub_ps(r21,rswitch);
2903 d = _mm256_max_ps(d,_mm256_setzero_ps());
2904 d2 = _mm256_mul_ps(d,d);
2905 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)))))));
2907 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2909 /* Evaluate switch function */
2910 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2911 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
2912 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2916 fscal = _mm256_and_ps(fscal,cutoff_mask);
2918 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2920 /* Calculate temporary vectorial force */
2921 tx = _mm256_mul_ps(fscal,dx21);
2922 ty = _mm256_mul_ps(fscal,dy21);
2923 tz = _mm256_mul_ps(fscal,dz21);
2925 /* Update vectorial force */
2926 fix2 = _mm256_add_ps(fix2,tx);
2927 fiy2 = _mm256_add_ps(fiy2,ty);
2928 fiz2 = _mm256_add_ps(fiz2,tz);
2930 fjx1 = _mm256_add_ps(fjx1,tx);
2931 fjy1 = _mm256_add_ps(fjy1,ty);
2932 fjz1 = _mm256_add_ps(fjz1,tz);
2936 /**************************
2937 * CALCULATE INTERACTIONS *
2938 **************************/
2940 if (gmx_mm256_any_lt(rsq22,rcutoff2))
2943 r22 = _mm256_mul_ps(rsq22,rinv22);
2944 r22 = _mm256_andnot_ps(dummy_mask,r22);
2946 /* EWALD ELECTROSTATICS */
2948 /* Analytical PME correction */
2949 zeta2 = _mm256_mul_ps(beta2,rsq22);
2950 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
2951 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2952 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2953 felec = _mm256_mul_ps(qq22,felec);
2954 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2955 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2956 velec = _mm256_sub_ps(rinv22,pmecorrV);
2957 velec = _mm256_mul_ps(qq22,velec);
2959 d = _mm256_sub_ps(r22,rswitch);
2960 d = _mm256_max_ps(d,_mm256_setzero_ps());
2961 d2 = _mm256_mul_ps(d,d);
2962 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)))))));
2964 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2966 /* Evaluate switch function */
2967 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2968 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
2969 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
2973 fscal = _mm256_and_ps(fscal,cutoff_mask);
2975 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2977 /* Calculate temporary vectorial force */
2978 tx = _mm256_mul_ps(fscal,dx22);
2979 ty = _mm256_mul_ps(fscal,dy22);
2980 tz = _mm256_mul_ps(fscal,dz22);
2982 /* Update vectorial force */
2983 fix2 = _mm256_add_ps(fix2,tx);
2984 fiy2 = _mm256_add_ps(fiy2,ty);
2985 fiz2 = _mm256_add_ps(fiz2,tz);
2987 fjx2 = _mm256_add_ps(fjx2,tx);
2988 fjy2 = _mm256_add_ps(fjy2,ty);
2989 fjz2 = _mm256_add_ps(fjz2,tz);
2993 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2994 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2995 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2996 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2997 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2998 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2999 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
3000 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
3002 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
3003 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
3005 /* Inner loop uses 954 flops */
3008 /* End of innermost loop */
3010 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
3011 f+i_coord_offset,fshift+i_shift_offset);
3013 /* Increment number of inner iterations */
3014 inneriter += j_index_end - j_index_start;
3016 /* Outer loop uses 18 flops */
3019 /* Increment number of outer iterations */
3022 /* Update outer/inner flops */
3024 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*954);