2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS avx_256_single kernel generator.
42 #include "../nb_kernel.h"
43 #include "gromacs/legacyheaders/types/simple.h"
44 #include "gromacs/math/vec.h"
45 #include "gromacs/legacyheaders/nrnb.h"
47 #include "gromacs/simd/math_x86_avx_256_single.h"
48 #include "kernelutil_x86_avx_256_single.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_VF_avx_256_single
52 * Electrostatics interaction: Ewald
53 * VdW interaction: LennardJones
54 * Geometry: Water3-Water3
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSw_VdwLJSw_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 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
113 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
114 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
116 __m128i ewitab_lo,ewitab_hi;
117 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
118 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
120 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
121 real rswitch_scalar,d_scalar;
122 __m256 dummy_mask,cutoff_mask;
123 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
124 __m256 one = _mm256_set1_ps(1.0);
125 __m256 two = _mm256_set1_ps(2.0);
131 jindex = nlist->jindex;
133 shiftidx = nlist->shift;
135 shiftvec = fr->shift_vec[0];
136 fshift = fr->fshift[0];
137 facel = _mm256_set1_ps(fr->epsfac);
138 charge = mdatoms->chargeA;
139 nvdwtype = fr->ntype;
141 vdwtype = mdatoms->typeA;
143 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
144 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
145 beta2 = _mm256_mul_ps(beta,beta);
146 beta3 = _mm256_mul_ps(beta,beta2);
148 ewtab = fr->ic->tabq_coul_FDV0;
149 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
150 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
152 /* Setup water-specific parameters */
153 inr = nlist->iinr[0];
154 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
155 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
156 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
157 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
159 jq0 = _mm256_set1_ps(charge[inr+0]);
160 jq1 = _mm256_set1_ps(charge[inr+1]);
161 jq2 = _mm256_set1_ps(charge[inr+2]);
162 vdwjidx0A = 2*vdwtype[inr+0];
163 qq00 = _mm256_mul_ps(iq0,jq0);
164 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
165 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
166 qq01 = _mm256_mul_ps(iq0,jq1);
167 qq02 = _mm256_mul_ps(iq0,jq2);
168 qq10 = _mm256_mul_ps(iq1,jq0);
169 qq11 = _mm256_mul_ps(iq1,jq1);
170 qq12 = _mm256_mul_ps(iq1,jq2);
171 qq20 = _mm256_mul_ps(iq2,jq0);
172 qq21 = _mm256_mul_ps(iq2,jq1);
173 qq22 = _mm256_mul_ps(iq2,jq2);
175 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
176 rcutoff_scalar = fr->rcoulomb;
177 rcutoff = _mm256_set1_ps(rcutoff_scalar);
178 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
180 rswitch_scalar = fr->rcoulomb_switch;
181 rswitch = _mm256_set1_ps(rswitch_scalar);
182 /* Setup switch parameters */
183 d_scalar = rcutoff_scalar-rswitch_scalar;
184 d = _mm256_set1_ps(d_scalar);
185 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
186 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
187 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
188 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
189 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
190 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
192 /* Avoid stupid compiler warnings */
193 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
206 for(iidx=0;iidx<4*DIM;iidx++)
211 /* Start outer loop over neighborlists */
212 for(iidx=0; iidx<nri; iidx++)
214 /* Load shift vector for this list */
215 i_shift_offset = DIM*shiftidx[iidx];
217 /* Load limits for loop over neighbors */
218 j_index_start = jindex[iidx];
219 j_index_end = jindex[iidx+1];
221 /* Get outer coordinate index */
223 i_coord_offset = DIM*inr;
225 /* Load i particle coords and add shift vector */
226 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
227 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
229 fix0 = _mm256_setzero_ps();
230 fiy0 = _mm256_setzero_ps();
231 fiz0 = _mm256_setzero_ps();
232 fix1 = _mm256_setzero_ps();
233 fiy1 = _mm256_setzero_ps();
234 fiz1 = _mm256_setzero_ps();
235 fix2 = _mm256_setzero_ps();
236 fiy2 = _mm256_setzero_ps();
237 fiz2 = _mm256_setzero_ps();
239 /* Reset potential sums */
240 velecsum = _mm256_setzero_ps();
241 vvdwsum = _mm256_setzero_ps();
243 /* Start inner kernel loop */
244 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
247 /* Get j neighbor index, and coordinate index */
256 j_coord_offsetA = DIM*jnrA;
257 j_coord_offsetB = DIM*jnrB;
258 j_coord_offsetC = DIM*jnrC;
259 j_coord_offsetD = DIM*jnrD;
260 j_coord_offsetE = DIM*jnrE;
261 j_coord_offsetF = DIM*jnrF;
262 j_coord_offsetG = DIM*jnrG;
263 j_coord_offsetH = DIM*jnrH;
265 /* load j atom coordinates */
266 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
267 x+j_coord_offsetC,x+j_coord_offsetD,
268 x+j_coord_offsetE,x+j_coord_offsetF,
269 x+j_coord_offsetG,x+j_coord_offsetH,
270 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
272 /* Calculate displacement vector */
273 dx00 = _mm256_sub_ps(ix0,jx0);
274 dy00 = _mm256_sub_ps(iy0,jy0);
275 dz00 = _mm256_sub_ps(iz0,jz0);
276 dx01 = _mm256_sub_ps(ix0,jx1);
277 dy01 = _mm256_sub_ps(iy0,jy1);
278 dz01 = _mm256_sub_ps(iz0,jz1);
279 dx02 = _mm256_sub_ps(ix0,jx2);
280 dy02 = _mm256_sub_ps(iy0,jy2);
281 dz02 = _mm256_sub_ps(iz0,jz2);
282 dx10 = _mm256_sub_ps(ix1,jx0);
283 dy10 = _mm256_sub_ps(iy1,jy0);
284 dz10 = _mm256_sub_ps(iz1,jz0);
285 dx11 = _mm256_sub_ps(ix1,jx1);
286 dy11 = _mm256_sub_ps(iy1,jy1);
287 dz11 = _mm256_sub_ps(iz1,jz1);
288 dx12 = _mm256_sub_ps(ix1,jx2);
289 dy12 = _mm256_sub_ps(iy1,jy2);
290 dz12 = _mm256_sub_ps(iz1,jz2);
291 dx20 = _mm256_sub_ps(ix2,jx0);
292 dy20 = _mm256_sub_ps(iy2,jy0);
293 dz20 = _mm256_sub_ps(iz2,jz0);
294 dx21 = _mm256_sub_ps(ix2,jx1);
295 dy21 = _mm256_sub_ps(iy2,jy1);
296 dz21 = _mm256_sub_ps(iz2,jz1);
297 dx22 = _mm256_sub_ps(ix2,jx2);
298 dy22 = _mm256_sub_ps(iy2,jy2);
299 dz22 = _mm256_sub_ps(iz2,jz2);
301 /* Calculate squared distance and things based on it */
302 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
303 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
304 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
305 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
306 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
307 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
308 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
309 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
310 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
312 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
313 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
314 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
315 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
316 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
317 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
318 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
319 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
320 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
322 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
323 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
324 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
325 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
326 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
327 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
328 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
329 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
330 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
332 fjx0 = _mm256_setzero_ps();
333 fjy0 = _mm256_setzero_ps();
334 fjz0 = _mm256_setzero_ps();
335 fjx1 = _mm256_setzero_ps();
336 fjy1 = _mm256_setzero_ps();
337 fjz1 = _mm256_setzero_ps();
338 fjx2 = _mm256_setzero_ps();
339 fjy2 = _mm256_setzero_ps();
340 fjz2 = _mm256_setzero_ps();
342 /**************************
343 * CALCULATE INTERACTIONS *
344 **************************/
346 if (gmx_mm256_any_lt(rsq00,rcutoff2))
349 r00 = _mm256_mul_ps(rsq00,rinv00);
351 /* EWALD ELECTROSTATICS */
353 /* Analytical PME correction */
354 zeta2 = _mm256_mul_ps(beta2,rsq00);
355 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
356 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
357 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
358 felec = _mm256_mul_ps(qq00,felec);
359 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
360 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
361 velec = _mm256_sub_ps(rinv00,pmecorrV);
362 velec = _mm256_mul_ps(qq00,velec);
364 /* LENNARD-JONES DISPERSION/REPULSION */
366 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
367 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
368 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
369 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
370 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
372 d = _mm256_sub_ps(r00,rswitch);
373 d = _mm256_max_ps(d,_mm256_setzero_ps());
374 d2 = _mm256_mul_ps(d,d);
375 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)))))));
377 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
379 /* Evaluate switch function */
380 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
381 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
382 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
383 velec = _mm256_mul_ps(velec,sw);
384 vvdw = _mm256_mul_ps(vvdw,sw);
385 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
387 /* Update potential sum for this i atom from the interaction with this j atom. */
388 velec = _mm256_and_ps(velec,cutoff_mask);
389 velecsum = _mm256_add_ps(velecsum,velec);
390 vvdw = _mm256_and_ps(vvdw,cutoff_mask);
391 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
393 fscal = _mm256_add_ps(felec,fvdw);
395 fscal = _mm256_and_ps(fscal,cutoff_mask);
397 /* Calculate temporary vectorial force */
398 tx = _mm256_mul_ps(fscal,dx00);
399 ty = _mm256_mul_ps(fscal,dy00);
400 tz = _mm256_mul_ps(fscal,dz00);
402 /* Update vectorial force */
403 fix0 = _mm256_add_ps(fix0,tx);
404 fiy0 = _mm256_add_ps(fiy0,ty);
405 fiz0 = _mm256_add_ps(fiz0,tz);
407 fjx0 = _mm256_add_ps(fjx0,tx);
408 fjy0 = _mm256_add_ps(fjy0,ty);
409 fjz0 = _mm256_add_ps(fjz0,tz);
413 /**************************
414 * CALCULATE INTERACTIONS *
415 **************************/
417 if (gmx_mm256_any_lt(rsq01,rcutoff2))
420 r01 = _mm256_mul_ps(rsq01,rinv01);
422 /* EWALD ELECTROSTATICS */
424 /* Analytical PME correction */
425 zeta2 = _mm256_mul_ps(beta2,rsq01);
426 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
427 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
428 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
429 felec = _mm256_mul_ps(qq01,felec);
430 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
431 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
432 velec = _mm256_sub_ps(rinv01,pmecorrV);
433 velec = _mm256_mul_ps(qq01,velec);
435 d = _mm256_sub_ps(r01,rswitch);
436 d = _mm256_max_ps(d,_mm256_setzero_ps());
437 d2 = _mm256_mul_ps(d,d);
438 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)))))));
440 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
442 /* Evaluate switch function */
443 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
444 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
445 velec = _mm256_mul_ps(velec,sw);
446 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
448 /* Update potential sum for this i atom from the interaction with this j atom. */
449 velec = _mm256_and_ps(velec,cutoff_mask);
450 velecsum = _mm256_add_ps(velecsum,velec);
454 fscal = _mm256_and_ps(fscal,cutoff_mask);
456 /* Calculate temporary vectorial force */
457 tx = _mm256_mul_ps(fscal,dx01);
458 ty = _mm256_mul_ps(fscal,dy01);
459 tz = _mm256_mul_ps(fscal,dz01);
461 /* Update vectorial force */
462 fix0 = _mm256_add_ps(fix0,tx);
463 fiy0 = _mm256_add_ps(fiy0,ty);
464 fiz0 = _mm256_add_ps(fiz0,tz);
466 fjx1 = _mm256_add_ps(fjx1,tx);
467 fjy1 = _mm256_add_ps(fjy1,ty);
468 fjz1 = _mm256_add_ps(fjz1,tz);
472 /**************************
473 * CALCULATE INTERACTIONS *
474 **************************/
476 if (gmx_mm256_any_lt(rsq02,rcutoff2))
479 r02 = _mm256_mul_ps(rsq02,rinv02);
481 /* EWALD ELECTROSTATICS */
483 /* Analytical PME correction */
484 zeta2 = _mm256_mul_ps(beta2,rsq02);
485 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
486 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
487 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
488 felec = _mm256_mul_ps(qq02,felec);
489 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
490 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
491 velec = _mm256_sub_ps(rinv02,pmecorrV);
492 velec = _mm256_mul_ps(qq02,velec);
494 d = _mm256_sub_ps(r02,rswitch);
495 d = _mm256_max_ps(d,_mm256_setzero_ps());
496 d2 = _mm256_mul_ps(d,d);
497 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)))))));
499 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
501 /* Evaluate switch function */
502 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
503 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
504 velec = _mm256_mul_ps(velec,sw);
505 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
507 /* Update potential sum for this i atom from the interaction with this j atom. */
508 velec = _mm256_and_ps(velec,cutoff_mask);
509 velecsum = _mm256_add_ps(velecsum,velec);
513 fscal = _mm256_and_ps(fscal,cutoff_mask);
515 /* Calculate temporary vectorial force */
516 tx = _mm256_mul_ps(fscal,dx02);
517 ty = _mm256_mul_ps(fscal,dy02);
518 tz = _mm256_mul_ps(fscal,dz02);
520 /* Update vectorial force */
521 fix0 = _mm256_add_ps(fix0,tx);
522 fiy0 = _mm256_add_ps(fiy0,ty);
523 fiz0 = _mm256_add_ps(fiz0,tz);
525 fjx2 = _mm256_add_ps(fjx2,tx);
526 fjy2 = _mm256_add_ps(fjy2,ty);
527 fjz2 = _mm256_add_ps(fjz2,tz);
531 /**************************
532 * CALCULATE INTERACTIONS *
533 **************************/
535 if (gmx_mm256_any_lt(rsq10,rcutoff2))
538 r10 = _mm256_mul_ps(rsq10,rinv10);
540 /* EWALD ELECTROSTATICS */
542 /* Analytical PME correction */
543 zeta2 = _mm256_mul_ps(beta2,rsq10);
544 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
545 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
546 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
547 felec = _mm256_mul_ps(qq10,felec);
548 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
549 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
550 velec = _mm256_sub_ps(rinv10,pmecorrV);
551 velec = _mm256_mul_ps(qq10,velec);
553 d = _mm256_sub_ps(r10,rswitch);
554 d = _mm256_max_ps(d,_mm256_setzero_ps());
555 d2 = _mm256_mul_ps(d,d);
556 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)))))));
558 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
560 /* Evaluate switch function */
561 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
562 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
563 velec = _mm256_mul_ps(velec,sw);
564 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
566 /* Update potential sum for this i atom from the interaction with this j atom. */
567 velec = _mm256_and_ps(velec,cutoff_mask);
568 velecsum = _mm256_add_ps(velecsum,velec);
572 fscal = _mm256_and_ps(fscal,cutoff_mask);
574 /* Calculate temporary vectorial force */
575 tx = _mm256_mul_ps(fscal,dx10);
576 ty = _mm256_mul_ps(fscal,dy10);
577 tz = _mm256_mul_ps(fscal,dz10);
579 /* Update vectorial force */
580 fix1 = _mm256_add_ps(fix1,tx);
581 fiy1 = _mm256_add_ps(fiy1,ty);
582 fiz1 = _mm256_add_ps(fiz1,tz);
584 fjx0 = _mm256_add_ps(fjx0,tx);
585 fjy0 = _mm256_add_ps(fjy0,ty);
586 fjz0 = _mm256_add_ps(fjz0,tz);
590 /**************************
591 * CALCULATE INTERACTIONS *
592 **************************/
594 if (gmx_mm256_any_lt(rsq11,rcutoff2))
597 r11 = _mm256_mul_ps(rsq11,rinv11);
599 /* EWALD ELECTROSTATICS */
601 /* Analytical PME correction */
602 zeta2 = _mm256_mul_ps(beta2,rsq11);
603 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
604 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
605 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
606 felec = _mm256_mul_ps(qq11,felec);
607 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
608 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
609 velec = _mm256_sub_ps(rinv11,pmecorrV);
610 velec = _mm256_mul_ps(qq11,velec);
612 d = _mm256_sub_ps(r11,rswitch);
613 d = _mm256_max_ps(d,_mm256_setzero_ps());
614 d2 = _mm256_mul_ps(d,d);
615 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)))))));
617 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
619 /* Evaluate switch function */
620 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
621 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
622 velec = _mm256_mul_ps(velec,sw);
623 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
625 /* Update potential sum for this i atom from the interaction with this j atom. */
626 velec = _mm256_and_ps(velec,cutoff_mask);
627 velecsum = _mm256_add_ps(velecsum,velec);
631 fscal = _mm256_and_ps(fscal,cutoff_mask);
633 /* Calculate temporary vectorial force */
634 tx = _mm256_mul_ps(fscal,dx11);
635 ty = _mm256_mul_ps(fscal,dy11);
636 tz = _mm256_mul_ps(fscal,dz11);
638 /* Update vectorial force */
639 fix1 = _mm256_add_ps(fix1,tx);
640 fiy1 = _mm256_add_ps(fiy1,ty);
641 fiz1 = _mm256_add_ps(fiz1,tz);
643 fjx1 = _mm256_add_ps(fjx1,tx);
644 fjy1 = _mm256_add_ps(fjy1,ty);
645 fjz1 = _mm256_add_ps(fjz1,tz);
649 /**************************
650 * CALCULATE INTERACTIONS *
651 **************************/
653 if (gmx_mm256_any_lt(rsq12,rcutoff2))
656 r12 = _mm256_mul_ps(rsq12,rinv12);
658 /* EWALD ELECTROSTATICS */
660 /* Analytical PME correction */
661 zeta2 = _mm256_mul_ps(beta2,rsq12);
662 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
663 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
664 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
665 felec = _mm256_mul_ps(qq12,felec);
666 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
667 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
668 velec = _mm256_sub_ps(rinv12,pmecorrV);
669 velec = _mm256_mul_ps(qq12,velec);
671 d = _mm256_sub_ps(r12,rswitch);
672 d = _mm256_max_ps(d,_mm256_setzero_ps());
673 d2 = _mm256_mul_ps(d,d);
674 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)))))));
676 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
678 /* Evaluate switch function */
679 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
680 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
681 velec = _mm256_mul_ps(velec,sw);
682 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
684 /* Update potential sum for this i atom from the interaction with this j atom. */
685 velec = _mm256_and_ps(velec,cutoff_mask);
686 velecsum = _mm256_add_ps(velecsum,velec);
690 fscal = _mm256_and_ps(fscal,cutoff_mask);
692 /* Calculate temporary vectorial force */
693 tx = _mm256_mul_ps(fscal,dx12);
694 ty = _mm256_mul_ps(fscal,dy12);
695 tz = _mm256_mul_ps(fscal,dz12);
697 /* Update vectorial force */
698 fix1 = _mm256_add_ps(fix1,tx);
699 fiy1 = _mm256_add_ps(fiy1,ty);
700 fiz1 = _mm256_add_ps(fiz1,tz);
702 fjx2 = _mm256_add_ps(fjx2,tx);
703 fjy2 = _mm256_add_ps(fjy2,ty);
704 fjz2 = _mm256_add_ps(fjz2,tz);
708 /**************************
709 * CALCULATE INTERACTIONS *
710 **************************/
712 if (gmx_mm256_any_lt(rsq20,rcutoff2))
715 r20 = _mm256_mul_ps(rsq20,rinv20);
717 /* EWALD ELECTROSTATICS */
719 /* Analytical PME correction */
720 zeta2 = _mm256_mul_ps(beta2,rsq20);
721 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
722 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
723 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
724 felec = _mm256_mul_ps(qq20,felec);
725 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
726 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
727 velec = _mm256_sub_ps(rinv20,pmecorrV);
728 velec = _mm256_mul_ps(qq20,velec);
730 d = _mm256_sub_ps(r20,rswitch);
731 d = _mm256_max_ps(d,_mm256_setzero_ps());
732 d2 = _mm256_mul_ps(d,d);
733 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)))))));
735 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
737 /* Evaluate switch function */
738 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
739 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
740 velec = _mm256_mul_ps(velec,sw);
741 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
743 /* Update potential sum for this i atom from the interaction with this j atom. */
744 velec = _mm256_and_ps(velec,cutoff_mask);
745 velecsum = _mm256_add_ps(velecsum,velec);
749 fscal = _mm256_and_ps(fscal,cutoff_mask);
751 /* Calculate temporary vectorial force */
752 tx = _mm256_mul_ps(fscal,dx20);
753 ty = _mm256_mul_ps(fscal,dy20);
754 tz = _mm256_mul_ps(fscal,dz20);
756 /* Update vectorial force */
757 fix2 = _mm256_add_ps(fix2,tx);
758 fiy2 = _mm256_add_ps(fiy2,ty);
759 fiz2 = _mm256_add_ps(fiz2,tz);
761 fjx0 = _mm256_add_ps(fjx0,tx);
762 fjy0 = _mm256_add_ps(fjy0,ty);
763 fjz0 = _mm256_add_ps(fjz0,tz);
767 /**************************
768 * CALCULATE INTERACTIONS *
769 **************************/
771 if (gmx_mm256_any_lt(rsq21,rcutoff2))
774 r21 = _mm256_mul_ps(rsq21,rinv21);
776 /* EWALD ELECTROSTATICS */
778 /* Analytical PME correction */
779 zeta2 = _mm256_mul_ps(beta2,rsq21);
780 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
781 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
782 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
783 felec = _mm256_mul_ps(qq21,felec);
784 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
785 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
786 velec = _mm256_sub_ps(rinv21,pmecorrV);
787 velec = _mm256_mul_ps(qq21,velec);
789 d = _mm256_sub_ps(r21,rswitch);
790 d = _mm256_max_ps(d,_mm256_setzero_ps());
791 d2 = _mm256_mul_ps(d,d);
792 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)))))));
794 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
796 /* Evaluate switch function */
797 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
798 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
799 velec = _mm256_mul_ps(velec,sw);
800 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
802 /* Update potential sum for this i atom from the interaction with this j atom. */
803 velec = _mm256_and_ps(velec,cutoff_mask);
804 velecsum = _mm256_add_ps(velecsum,velec);
808 fscal = _mm256_and_ps(fscal,cutoff_mask);
810 /* Calculate temporary vectorial force */
811 tx = _mm256_mul_ps(fscal,dx21);
812 ty = _mm256_mul_ps(fscal,dy21);
813 tz = _mm256_mul_ps(fscal,dz21);
815 /* Update vectorial force */
816 fix2 = _mm256_add_ps(fix2,tx);
817 fiy2 = _mm256_add_ps(fiy2,ty);
818 fiz2 = _mm256_add_ps(fiz2,tz);
820 fjx1 = _mm256_add_ps(fjx1,tx);
821 fjy1 = _mm256_add_ps(fjy1,ty);
822 fjz1 = _mm256_add_ps(fjz1,tz);
826 /**************************
827 * CALCULATE INTERACTIONS *
828 **************************/
830 if (gmx_mm256_any_lt(rsq22,rcutoff2))
833 r22 = _mm256_mul_ps(rsq22,rinv22);
835 /* EWALD ELECTROSTATICS */
837 /* Analytical PME correction */
838 zeta2 = _mm256_mul_ps(beta2,rsq22);
839 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
840 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
841 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
842 felec = _mm256_mul_ps(qq22,felec);
843 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
844 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
845 velec = _mm256_sub_ps(rinv22,pmecorrV);
846 velec = _mm256_mul_ps(qq22,velec);
848 d = _mm256_sub_ps(r22,rswitch);
849 d = _mm256_max_ps(d,_mm256_setzero_ps());
850 d2 = _mm256_mul_ps(d,d);
851 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)))))));
853 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
855 /* Evaluate switch function */
856 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
857 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
858 velec = _mm256_mul_ps(velec,sw);
859 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
861 /* Update potential sum for this i atom from the interaction with this j atom. */
862 velec = _mm256_and_ps(velec,cutoff_mask);
863 velecsum = _mm256_add_ps(velecsum,velec);
867 fscal = _mm256_and_ps(fscal,cutoff_mask);
869 /* Calculate temporary vectorial force */
870 tx = _mm256_mul_ps(fscal,dx22);
871 ty = _mm256_mul_ps(fscal,dy22);
872 tz = _mm256_mul_ps(fscal,dz22);
874 /* Update vectorial force */
875 fix2 = _mm256_add_ps(fix2,tx);
876 fiy2 = _mm256_add_ps(fiy2,ty);
877 fiz2 = _mm256_add_ps(fiz2,tz);
879 fjx2 = _mm256_add_ps(fjx2,tx);
880 fjy2 = _mm256_add_ps(fjy2,ty);
881 fjz2 = _mm256_add_ps(fjz2,tz);
885 fjptrA = f+j_coord_offsetA;
886 fjptrB = f+j_coord_offsetB;
887 fjptrC = f+j_coord_offsetC;
888 fjptrD = f+j_coord_offsetD;
889 fjptrE = f+j_coord_offsetE;
890 fjptrF = f+j_coord_offsetF;
891 fjptrG = f+j_coord_offsetG;
892 fjptrH = f+j_coord_offsetH;
894 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
895 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
897 /* Inner loop uses 990 flops */
903 /* Get j neighbor index, and coordinate index */
904 jnrlistA = jjnr[jidx];
905 jnrlistB = jjnr[jidx+1];
906 jnrlistC = jjnr[jidx+2];
907 jnrlistD = jjnr[jidx+3];
908 jnrlistE = jjnr[jidx+4];
909 jnrlistF = jjnr[jidx+5];
910 jnrlistG = jjnr[jidx+6];
911 jnrlistH = jjnr[jidx+7];
912 /* Sign of each element will be negative for non-real atoms.
913 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
914 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
916 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
917 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
919 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
920 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
921 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
922 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
923 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
924 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
925 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
926 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
927 j_coord_offsetA = DIM*jnrA;
928 j_coord_offsetB = DIM*jnrB;
929 j_coord_offsetC = DIM*jnrC;
930 j_coord_offsetD = DIM*jnrD;
931 j_coord_offsetE = DIM*jnrE;
932 j_coord_offsetF = DIM*jnrF;
933 j_coord_offsetG = DIM*jnrG;
934 j_coord_offsetH = DIM*jnrH;
936 /* load j atom coordinates */
937 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
938 x+j_coord_offsetC,x+j_coord_offsetD,
939 x+j_coord_offsetE,x+j_coord_offsetF,
940 x+j_coord_offsetG,x+j_coord_offsetH,
941 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
943 /* Calculate displacement vector */
944 dx00 = _mm256_sub_ps(ix0,jx0);
945 dy00 = _mm256_sub_ps(iy0,jy0);
946 dz00 = _mm256_sub_ps(iz0,jz0);
947 dx01 = _mm256_sub_ps(ix0,jx1);
948 dy01 = _mm256_sub_ps(iy0,jy1);
949 dz01 = _mm256_sub_ps(iz0,jz1);
950 dx02 = _mm256_sub_ps(ix0,jx2);
951 dy02 = _mm256_sub_ps(iy0,jy2);
952 dz02 = _mm256_sub_ps(iz0,jz2);
953 dx10 = _mm256_sub_ps(ix1,jx0);
954 dy10 = _mm256_sub_ps(iy1,jy0);
955 dz10 = _mm256_sub_ps(iz1,jz0);
956 dx11 = _mm256_sub_ps(ix1,jx1);
957 dy11 = _mm256_sub_ps(iy1,jy1);
958 dz11 = _mm256_sub_ps(iz1,jz1);
959 dx12 = _mm256_sub_ps(ix1,jx2);
960 dy12 = _mm256_sub_ps(iy1,jy2);
961 dz12 = _mm256_sub_ps(iz1,jz2);
962 dx20 = _mm256_sub_ps(ix2,jx0);
963 dy20 = _mm256_sub_ps(iy2,jy0);
964 dz20 = _mm256_sub_ps(iz2,jz0);
965 dx21 = _mm256_sub_ps(ix2,jx1);
966 dy21 = _mm256_sub_ps(iy2,jy1);
967 dz21 = _mm256_sub_ps(iz2,jz1);
968 dx22 = _mm256_sub_ps(ix2,jx2);
969 dy22 = _mm256_sub_ps(iy2,jy2);
970 dz22 = _mm256_sub_ps(iz2,jz2);
972 /* Calculate squared distance and things based on it */
973 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
974 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
975 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
976 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
977 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
978 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
979 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
980 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
981 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
983 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
984 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
985 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
986 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
987 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
988 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
989 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
990 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
991 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
993 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
994 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
995 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
996 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
997 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
998 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
999 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1000 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1001 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1003 fjx0 = _mm256_setzero_ps();
1004 fjy0 = _mm256_setzero_ps();
1005 fjz0 = _mm256_setzero_ps();
1006 fjx1 = _mm256_setzero_ps();
1007 fjy1 = _mm256_setzero_ps();
1008 fjz1 = _mm256_setzero_ps();
1009 fjx2 = _mm256_setzero_ps();
1010 fjy2 = _mm256_setzero_ps();
1011 fjz2 = _mm256_setzero_ps();
1013 /**************************
1014 * CALCULATE INTERACTIONS *
1015 **************************/
1017 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1020 r00 = _mm256_mul_ps(rsq00,rinv00);
1021 r00 = _mm256_andnot_ps(dummy_mask,r00);
1023 /* EWALD ELECTROSTATICS */
1025 /* Analytical PME correction */
1026 zeta2 = _mm256_mul_ps(beta2,rsq00);
1027 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1028 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1029 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1030 felec = _mm256_mul_ps(qq00,felec);
1031 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1032 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1033 velec = _mm256_sub_ps(rinv00,pmecorrV);
1034 velec = _mm256_mul_ps(qq00,velec);
1036 /* LENNARD-JONES DISPERSION/REPULSION */
1038 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1039 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
1040 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
1041 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
1042 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
1044 d = _mm256_sub_ps(r00,rswitch);
1045 d = _mm256_max_ps(d,_mm256_setzero_ps());
1046 d2 = _mm256_mul_ps(d,d);
1047 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)))))));
1049 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1051 /* Evaluate switch function */
1052 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1053 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1054 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
1055 velec = _mm256_mul_ps(velec,sw);
1056 vvdw = _mm256_mul_ps(vvdw,sw);
1057 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1059 /* Update potential sum for this i atom from the interaction with this j atom. */
1060 velec = _mm256_and_ps(velec,cutoff_mask);
1061 velec = _mm256_andnot_ps(dummy_mask,velec);
1062 velecsum = _mm256_add_ps(velecsum,velec);
1063 vvdw = _mm256_and_ps(vvdw,cutoff_mask);
1064 vvdw = _mm256_andnot_ps(dummy_mask,vvdw);
1065 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
1067 fscal = _mm256_add_ps(felec,fvdw);
1069 fscal = _mm256_and_ps(fscal,cutoff_mask);
1071 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1073 /* Calculate temporary vectorial force */
1074 tx = _mm256_mul_ps(fscal,dx00);
1075 ty = _mm256_mul_ps(fscal,dy00);
1076 tz = _mm256_mul_ps(fscal,dz00);
1078 /* Update vectorial force */
1079 fix0 = _mm256_add_ps(fix0,tx);
1080 fiy0 = _mm256_add_ps(fiy0,ty);
1081 fiz0 = _mm256_add_ps(fiz0,tz);
1083 fjx0 = _mm256_add_ps(fjx0,tx);
1084 fjy0 = _mm256_add_ps(fjy0,ty);
1085 fjz0 = _mm256_add_ps(fjz0,tz);
1089 /**************************
1090 * CALCULATE INTERACTIONS *
1091 **************************/
1093 if (gmx_mm256_any_lt(rsq01,rcutoff2))
1096 r01 = _mm256_mul_ps(rsq01,rinv01);
1097 r01 = _mm256_andnot_ps(dummy_mask,r01);
1099 /* EWALD ELECTROSTATICS */
1101 /* Analytical PME correction */
1102 zeta2 = _mm256_mul_ps(beta2,rsq01);
1103 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1104 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1105 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1106 felec = _mm256_mul_ps(qq01,felec);
1107 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1108 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1109 velec = _mm256_sub_ps(rinv01,pmecorrV);
1110 velec = _mm256_mul_ps(qq01,velec);
1112 d = _mm256_sub_ps(r01,rswitch);
1113 d = _mm256_max_ps(d,_mm256_setzero_ps());
1114 d2 = _mm256_mul_ps(d,d);
1115 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)))))));
1117 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1119 /* Evaluate switch function */
1120 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1121 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
1122 velec = _mm256_mul_ps(velec,sw);
1123 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
1125 /* Update potential sum for this i atom from the interaction with this j atom. */
1126 velec = _mm256_and_ps(velec,cutoff_mask);
1127 velec = _mm256_andnot_ps(dummy_mask,velec);
1128 velecsum = _mm256_add_ps(velecsum,velec);
1132 fscal = _mm256_and_ps(fscal,cutoff_mask);
1134 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1136 /* Calculate temporary vectorial force */
1137 tx = _mm256_mul_ps(fscal,dx01);
1138 ty = _mm256_mul_ps(fscal,dy01);
1139 tz = _mm256_mul_ps(fscal,dz01);
1141 /* Update vectorial force */
1142 fix0 = _mm256_add_ps(fix0,tx);
1143 fiy0 = _mm256_add_ps(fiy0,ty);
1144 fiz0 = _mm256_add_ps(fiz0,tz);
1146 fjx1 = _mm256_add_ps(fjx1,tx);
1147 fjy1 = _mm256_add_ps(fjy1,ty);
1148 fjz1 = _mm256_add_ps(fjz1,tz);
1152 /**************************
1153 * CALCULATE INTERACTIONS *
1154 **************************/
1156 if (gmx_mm256_any_lt(rsq02,rcutoff2))
1159 r02 = _mm256_mul_ps(rsq02,rinv02);
1160 r02 = _mm256_andnot_ps(dummy_mask,r02);
1162 /* EWALD ELECTROSTATICS */
1164 /* Analytical PME correction */
1165 zeta2 = _mm256_mul_ps(beta2,rsq02);
1166 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
1167 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1168 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1169 felec = _mm256_mul_ps(qq02,felec);
1170 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1171 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1172 velec = _mm256_sub_ps(rinv02,pmecorrV);
1173 velec = _mm256_mul_ps(qq02,velec);
1175 d = _mm256_sub_ps(r02,rswitch);
1176 d = _mm256_max_ps(d,_mm256_setzero_ps());
1177 d2 = _mm256_mul_ps(d,d);
1178 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)))))));
1180 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1182 /* Evaluate switch function */
1183 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1184 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
1185 velec = _mm256_mul_ps(velec,sw);
1186 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
1188 /* Update potential sum for this i atom from the interaction with this j atom. */
1189 velec = _mm256_and_ps(velec,cutoff_mask);
1190 velec = _mm256_andnot_ps(dummy_mask,velec);
1191 velecsum = _mm256_add_ps(velecsum,velec);
1195 fscal = _mm256_and_ps(fscal,cutoff_mask);
1197 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1199 /* Calculate temporary vectorial force */
1200 tx = _mm256_mul_ps(fscal,dx02);
1201 ty = _mm256_mul_ps(fscal,dy02);
1202 tz = _mm256_mul_ps(fscal,dz02);
1204 /* Update vectorial force */
1205 fix0 = _mm256_add_ps(fix0,tx);
1206 fiy0 = _mm256_add_ps(fiy0,ty);
1207 fiz0 = _mm256_add_ps(fiz0,tz);
1209 fjx2 = _mm256_add_ps(fjx2,tx);
1210 fjy2 = _mm256_add_ps(fjy2,ty);
1211 fjz2 = _mm256_add_ps(fjz2,tz);
1215 /**************************
1216 * CALCULATE INTERACTIONS *
1217 **************************/
1219 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1222 r10 = _mm256_mul_ps(rsq10,rinv10);
1223 r10 = _mm256_andnot_ps(dummy_mask,r10);
1225 /* EWALD ELECTROSTATICS */
1227 /* Analytical PME correction */
1228 zeta2 = _mm256_mul_ps(beta2,rsq10);
1229 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
1230 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1231 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1232 felec = _mm256_mul_ps(qq10,felec);
1233 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1234 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1235 velec = _mm256_sub_ps(rinv10,pmecorrV);
1236 velec = _mm256_mul_ps(qq10,velec);
1238 d = _mm256_sub_ps(r10,rswitch);
1239 d = _mm256_max_ps(d,_mm256_setzero_ps());
1240 d2 = _mm256_mul_ps(d,d);
1241 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)))))));
1243 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1245 /* Evaluate switch function */
1246 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1247 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
1248 velec = _mm256_mul_ps(velec,sw);
1249 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1251 /* Update potential sum for this i atom from the interaction with this j atom. */
1252 velec = _mm256_and_ps(velec,cutoff_mask);
1253 velec = _mm256_andnot_ps(dummy_mask,velec);
1254 velecsum = _mm256_add_ps(velecsum,velec);
1258 fscal = _mm256_and_ps(fscal,cutoff_mask);
1260 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1262 /* Calculate temporary vectorial force */
1263 tx = _mm256_mul_ps(fscal,dx10);
1264 ty = _mm256_mul_ps(fscal,dy10);
1265 tz = _mm256_mul_ps(fscal,dz10);
1267 /* Update vectorial force */
1268 fix1 = _mm256_add_ps(fix1,tx);
1269 fiy1 = _mm256_add_ps(fiy1,ty);
1270 fiz1 = _mm256_add_ps(fiz1,tz);
1272 fjx0 = _mm256_add_ps(fjx0,tx);
1273 fjy0 = _mm256_add_ps(fjy0,ty);
1274 fjz0 = _mm256_add_ps(fjz0,tz);
1278 /**************************
1279 * CALCULATE INTERACTIONS *
1280 **************************/
1282 if (gmx_mm256_any_lt(rsq11,rcutoff2))
1285 r11 = _mm256_mul_ps(rsq11,rinv11);
1286 r11 = _mm256_andnot_ps(dummy_mask,r11);
1288 /* EWALD ELECTROSTATICS */
1290 /* Analytical PME correction */
1291 zeta2 = _mm256_mul_ps(beta2,rsq11);
1292 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
1293 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1294 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1295 felec = _mm256_mul_ps(qq11,felec);
1296 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1297 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1298 velec = _mm256_sub_ps(rinv11,pmecorrV);
1299 velec = _mm256_mul_ps(qq11,velec);
1301 d = _mm256_sub_ps(r11,rswitch);
1302 d = _mm256_max_ps(d,_mm256_setzero_ps());
1303 d2 = _mm256_mul_ps(d,d);
1304 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)))))));
1306 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1308 /* Evaluate switch function */
1309 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1310 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
1311 velec = _mm256_mul_ps(velec,sw);
1312 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1314 /* Update potential sum for this i atom from the interaction with this j atom. */
1315 velec = _mm256_and_ps(velec,cutoff_mask);
1316 velec = _mm256_andnot_ps(dummy_mask,velec);
1317 velecsum = _mm256_add_ps(velecsum,velec);
1321 fscal = _mm256_and_ps(fscal,cutoff_mask);
1323 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1325 /* Calculate temporary vectorial force */
1326 tx = _mm256_mul_ps(fscal,dx11);
1327 ty = _mm256_mul_ps(fscal,dy11);
1328 tz = _mm256_mul_ps(fscal,dz11);
1330 /* Update vectorial force */
1331 fix1 = _mm256_add_ps(fix1,tx);
1332 fiy1 = _mm256_add_ps(fiy1,ty);
1333 fiz1 = _mm256_add_ps(fiz1,tz);
1335 fjx1 = _mm256_add_ps(fjx1,tx);
1336 fjy1 = _mm256_add_ps(fjy1,ty);
1337 fjz1 = _mm256_add_ps(fjz1,tz);
1341 /**************************
1342 * CALCULATE INTERACTIONS *
1343 **************************/
1345 if (gmx_mm256_any_lt(rsq12,rcutoff2))
1348 r12 = _mm256_mul_ps(rsq12,rinv12);
1349 r12 = _mm256_andnot_ps(dummy_mask,r12);
1351 /* EWALD ELECTROSTATICS */
1353 /* Analytical PME correction */
1354 zeta2 = _mm256_mul_ps(beta2,rsq12);
1355 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
1356 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1357 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1358 felec = _mm256_mul_ps(qq12,felec);
1359 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1360 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1361 velec = _mm256_sub_ps(rinv12,pmecorrV);
1362 velec = _mm256_mul_ps(qq12,velec);
1364 d = _mm256_sub_ps(r12,rswitch);
1365 d = _mm256_max_ps(d,_mm256_setzero_ps());
1366 d2 = _mm256_mul_ps(d,d);
1367 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)))))));
1369 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1371 /* Evaluate switch function */
1372 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1373 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
1374 velec = _mm256_mul_ps(velec,sw);
1375 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1377 /* Update potential sum for this i atom from the interaction with this j atom. */
1378 velec = _mm256_and_ps(velec,cutoff_mask);
1379 velec = _mm256_andnot_ps(dummy_mask,velec);
1380 velecsum = _mm256_add_ps(velecsum,velec);
1384 fscal = _mm256_and_ps(fscal,cutoff_mask);
1386 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1388 /* Calculate temporary vectorial force */
1389 tx = _mm256_mul_ps(fscal,dx12);
1390 ty = _mm256_mul_ps(fscal,dy12);
1391 tz = _mm256_mul_ps(fscal,dz12);
1393 /* Update vectorial force */
1394 fix1 = _mm256_add_ps(fix1,tx);
1395 fiy1 = _mm256_add_ps(fiy1,ty);
1396 fiz1 = _mm256_add_ps(fiz1,tz);
1398 fjx2 = _mm256_add_ps(fjx2,tx);
1399 fjy2 = _mm256_add_ps(fjy2,ty);
1400 fjz2 = _mm256_add_ps(fjz2,tz);
1404 /**************************
1405 * CALCULATE INTERACTIONS *
1406 **************************/
1408 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1411 r20 = _mm256_mul_ps(rsq20,rinv20);
1412 r20 = _mm256_andnot_ps(dummy_mask,r20);
1414 /* EWALD ELECTROSTATICS */
1416 /* Analytical PME correction */
1417 zeta2 = _mm256_mul_ps(beta2,rsq20);
1418 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1419 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1420 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1421 felec = _mm256_mul_ps(qq20,felec);
1422 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1423 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1424 velec = _mm256_sub_ps(rinv20,pmecorrV);
1425 velec = _mm256_mul_ps(qq20,velec);
1427 d = _mm256_sub_ps(r20,rswitch);
1428 d = _mm256_max_ps(d,_mm256_setzero_ps());
1429 d2 = _mm256_mul_ps(d,d);
1430 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)))))));
1432 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1434 /* Evaluate switch function */
1435 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1436 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
1437 velec = _mm256_mul_ps(velec,sw);
1438 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1440 /* Update potential sum for this i atom from the interaction with this j atom. */
1441 velec = _mm256_and_ps(velec,cutoff_mask);
1442 velec = _mm256_andnot_ps(dummy_mask,velec);
1443 velecsum = _mm256_add_ps(velecsum,velec);
1447 fscal = _mm256_and_ps(fscal,cutoff_mask);
1449 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1451 /* Calculate temporary vectorial force */
1452 tx = _mm256_mul_ps(fscal,dx20);
1453 ty = _mm256_mul_ps(fscal,dy20);
1454 tz = _mm256_mul_ps(fscal,dz20);
1456 /* Update vectorial force */
1457 fix2 = _mm256_add_ps(fix2,tx);
1458 fiy2 = _mm256_add_ps(fiy2,ty);
1459 fiz2 = _mm256_add_ps(fiz2,tz);
1461 fjx0 = _mm256_add_ps(fjx0,tx);
1462 fjy0 = _mm256_add_ps(fjy0,ty);
1463 fjz0 = _mm256_add_ps(fjz0,tz);
1467 /**************************
1468 * CALCULATE INTERACTIONS *
1469 **************************/
1471 if (gmx_mm256_any_lt(rsq21,rcutoff2))
1474 r21 = _mm256_mul_ps(rsq21,rinv21);
1475 r21 = _mm256_andnot_ps(dummy_mask,r21);
1477 /* EWALD ELECTROSTATICS */
1479 /* Analytical PME correction */
1480 zeta2 = _mm256_mul_ps(beta2,rsq21);
1481 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
1482 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1483 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1484 felec = _mm256_mul_ps(qq21,felec);
1485 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1486 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1487 velec = _mm256_sub_ps(rinv21,pmecorrV);
1488 velec = _mm256_mul_ps(qq21,velec);
1490 d = _mm256_sub_ps(r21,rswitch);
1491 d = _mm256_max_ps(d,_mm256_setzero_ps());
1492 d2 = _mm256_mul_ps(d,d);
1493 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)))))));
1495 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1497 /* Evaluate switch function */
1498 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1499 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
1500 velec = _mm256_mul_ps(velec,sw);
1501 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1503 /* Update potential sum for this i atom from the interaction with this j atom. */
1504 velec = _mm256_and_ps(velec,cutoff_mask);
1505 velec = _mm256_andnot_ps(dummy_mask,velec);
1506 velecsum = _mm256_add_ps(velecsum,velec);
1510 fscal = _mm256_and_ps(fscal,cutoff_mask);
1512 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1514 /* Calculate temporary vectorial force */
1515 tx = _mm256_mul_ps(fscal,dx21);
1516 ty = _mm256_mul_ps(fscal,dy21);
1517 tz = _mm256_mul_ps(fscal,dz21);
1519 /* Update vectorial force */
1520 fix2 = _mm256_add_ps(fix2,tx);
1521 fiy2 = _mm256_add_ps(fiy2,ty);
1522 fiz2 = _mm256_add_ps(fiz2,tz);
1524 fjx1 = _mm256_add_ps(fjx1,tx);
1525 fjy1 = _mm256_add_ps(fjy1,ty);
1526 fjz1 = _mm256_add_ps(fjz1,tz);
1530 /**************************
1531 * CALCULATE INTERACTIONS *
1532 **************************/
1534 if (gmx_mm256_any_lt(rsq22,rcutoff2))
1537 r22 = _mm256_mul_ps(rsq22,rinv22);
1538 r22 = _mm256_andnot_ps(dummy_mask,r22);
1540 /* EWALD ELECTROSTATICS */
1542 /* Analytical PME correction */
1543 zeta2 = _mm256_mul_ps(beta2,rsq22);
1544 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
1545 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1546 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1547 felec = _mm256_mul_ps(qq22,felec);
1548 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1549 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1550 velec = _mm256_sub_ps(rinv22,pmecorrV);
1551 velec = _mm256_mul_ps(qq22,velec);
1553 d = _mm256_sub_ps(r22,rswitch);
1554 d = _mm256_max_ps(d,_mm256_setzero_ps());
1555 d2 = _mm256_mul_ps(d,d);
1556 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)))))));
1558 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1560 /* Evaluate switch function */
1561 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1562 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
1563 velec = _mm256_mul_ps(velec,sw);
1564 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1566 /* Update potential sum for this i atom from the interaction with this j atom. */
1567 velec = _mm256_and_ps(velec,cutoff_mask);
1568 velec = _mm256_andnot_ps(dummy_mask,velec);
1569 velecsum = _mm256_add_ps(velecsum,velec);
1573 fscal = _mm256_and_ps(fscal,cutoff_mask);
1575 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1577 /* Calculate temporary vectorial force */
1578 tx = _mm256_mul_ps(fscal,dx22);
1579 ty = _mm256_mul_ps(fscal,dy22);
1580 tz = _mm256_mul_ps(fscal,dz22);
1582 /* Update vectorial force */
1583 fix2 = _mm256_add_ps(fix2,tx);
1584 fiy2 = _mm256_add_ps(fiy2,ty);
1585 fiz2 = _mm256_add_ps(fiz2,tz);
1587 fjx2 = _mm256_add_ps(fjx2,tx);
1588 fjy2 = _mm256_add_ps(fjy2,ty);
1589 fjz2 = _mm256_add_ps(fjz2,tz);
1593 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1594 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1595 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1596 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1597 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1598 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1599 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1600 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1602 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1603 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1605 /* Inner loop uses 999 flops */
1608 /* End of innermost loop */
1610 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1611 f+i_coord_offset,fshift+i_shift_offset);
1614 /* Update potential energies */
1615 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1616 gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1618 /* Increment number of inner iterations */
1619 inneriter += j_index_end - j_index_start;
1621 /* Outer loop uses 20 flops */
1624 /* Increment number of outer iterations */
1627 /* Update outer/inner flops */
1629 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*999);
1632 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_F_avx_256_single
1633 * Electrostatics interaction: Ewald
1634 * VdW interaction: LennardJones
1635 * Geometry: Water3-Water3
1636 * Calculate force/pot: Force
1639 nb_kernel_ElecEwSw_VdwLJSw_GeomW3W3_F_avx_256_single
1640 (t_nblist * gmx_restrict nlist,
1641 rvec * gmx_restrict xx,
1642 rvec * gmx_restrict ff,
1643 t_forcerec * gmx_restrict fr,
1644 t_mdatoms * gmx_restrict mdatoms,
1645 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1646 t_nrnb * gmx_restrict nrnb)
1648 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1649 * just 0 for non-waters.
1650 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1651 * jnr indices corresponding to data put in the four positions in the SIMD register.
1653 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1654 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1655 int jnrA,jnrB,jnrC,jnrD;
1656 int jnrE,jnrF,jnrG,jnrH;
1657 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1658 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1659 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1660 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1661 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1662 real rcutoff_scalar;
1663 real *shiftvec,*fshift,*x,*f;
1664 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1665 real scratch[4*DIM];
1666 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1667 real * vdwioffsetptr0;
1668 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1669 real * vdwioffsetptr1;
1670 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1671 real * vdwioffsetptr2;
1672 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1673 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1674 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1675 int vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1676 __m256 jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1677 int vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1678 __m256 jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1679 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1680 __m256 dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1681 __m256 dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1682 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1683 __m256 dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1684 __m256 dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1685 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1686 __m256 dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1687 __m256 dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1688 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
1691 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1694 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
1695 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
1697 __m128i ewitab_lo,ewitab_hi;
1698 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1699 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1701 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1702 real rswitch_scalar,d_scalar;
1703 __m256 dummy_mask,cutoff_mask;
1704 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1705 __m256 one = _mm256_set1_ps(1.0);
1706 __m256 two = _mm256_set1_ps(2.0);
1712 jindex = nlist->jindex;
1714 shiftidx = nlist->shift;
1716 shiftvec = fr->shift_vec[0];
1717 fshift = fr->fshift[0];
1718 facel = _mm256_set1_ps(fr->epsfac);
1719 charge = mdatoms->chargeA;
1720 nvdwtype = fr->ntype;
1721 vdwparam = fr->nbfp;
1722 vdwtype = mdatoms->typeA;
1724 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
1725 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
1726 beta2 = _mm256_mul_ps(beta,beta);
1727 beta3 = _mm256_mul_ps(beta,beta2);
1729 ewtab = fr->ic->tabq_coul_FDV0;
1730 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
1731 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1733 /* Setup water-specific parameters */
1734 inr = nlist->iinr[0];
1735 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1736 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1737 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1738 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
1740 jq0 = _mm256_set1_ps(charge[inr+0]);
1741 jq1 = _mm256_set1_ps(charge[inr+1]);
1742 jq2 = _mm256_set1_ps(charge[inr+2]);
1743 vdwjidx0A = 2*vdwtype[inr+0];
1744 qq00 = _mm256_mul_ps(iq0,jq0);
1745 c6_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1746 c12_00 = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1747 qq01 = _mm256_mul_ps(iq0,jq1);
1748 qq02 = _mm256_mul_ps(iq0,jq2);
1749 qq10 = _mm256_mul_ps(iq1,jq0);
1750 qq11 = _mm256_mul_ps(iq1,jq1);
1751 qq12 = _mm256_mul_ps(iq1,jq2);
1752 qq20 = _mm256_mul_ps(iq2,jq0);
1753 qq21 = _mm256_mul_ps(iq2,jq1);
1754 qq22 = _mm256_mul_ps(iq2,jq2);
1756 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1757 rcutoff_scalar = fr->rcoulomb;
1758 rcutoff = _mm256_set1_ps(rcutoff_scalar);
1759 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
1761 rswitch_scalar = fr->rcoulomb_switch;
1762 rswitch = _mm256_set1_ps(rswitch_scalar);
1763 /* Setup switch parameters */
1764 d_scalar = rcutoff_scalar-rswitch_scalar;
1765 d = _mm256_set1_ps(d_scalar);
1766 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
1767 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1768 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1769 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
1770 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1771 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1773 /* Avoid stupid compiler warnings */
1774 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1775 j_coord_offsetA = 0;
1776 j_coord_offsetB = 0;
1777 j_coord_offsetC = 0;
1778 j_coord_offsetD = 0;
1779 j_coord_offsetE = 0;
1780 j_coord_offsetF = 0;
1781 j_coord_offsetG = 0;
1782 j_coord_offsetH = 0;
1787 for(iidx=0;iidx<4*DIM;iidx++)
1789 scratch[iidx] = 0.0;
1792 /* Start outer loop over neighborlists */
1793 for(iidx=0; iidx<nri; iidx++)
1795 /* Load shift vector for this list */
1796 i_shift_offset = DIM*shiftidx[iidx];
1798 /* Load limits for loop over neighbors */
1799 j_index_start = jindex[iidx];
1800 j_index_end = jindex[iidx+1];
1802 /* Get outer coordinate index */
1804 i_coord_offset = DIM*inr;
1806 /* Load i particle coords and add shift vector */
1807 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1808 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1810 fix0 = _mm256_setzero_ps();
1811 fiy0 = _mm256_setzero_ps();
1812 fiz0 = _mm256_setzero_ps();
1813 fix1 = _mm256_setzero_ps();
1814 fiy1 = _mm256_setzero_ps();
1815 fiz1 = _mm256_setzero_ps();
1816 fix2 = _mm256_setzero_ps();
1817 fiy2 = _mm256_setzero_ps();
1818 fiz2 = _mm256_setzero_ps();
1820 /* Start inner kernel loop */
1821 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1824 /* Get j neighbor index, and coordinate index */
1826 jnrB = jjnr[jidx+1];
1827 jnrC = jjnr[jidx+2];
1828 jnrD = jjnr[jidx+3];
1829 jnrE = jjnr[jidx+4];
1830 jnrF = jjnr[jidx+5];
1831 jnrG = jjnr[jidx+6];
1832 jnrH = jjnr[jidx+7];
1833 j_coord_offsetA = DIM*jnrA;
1834 j_coord_offsetB = DIM*jnrB;
1835 j_coord_offsetC = DIM*jnrC;
1836 j_coord_offsetD = DIM*jnrD;
1837 j_coord_offsetE = DIM*jnrE;
1838 j_coord_offsetF = DIM*jnrF;
1839 j_coord_offsetG = DIM*jnrG;
1840 j_coord_offsetH = DIM*jnrH;
1842 /* load j atom coordinates */
1843 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1844 x+j_coord_offsetC,x+j_coord_offsetD,
1845 x+j_coord_offsetE,x+j_coord_offsetF,
1846 x+j_coord_offsetG,x+j_coord_offsetH,
1847 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1849 /* Calculate displacement vector */
1850 dx00 = _mm256_sub_ps(ix0,jx0);
1851 dy00 = _mm256_sub_ps(iy0,jy0);
1852 dz00 = _mm256_sub_ps(iz0,jz0);
1853 dx01 = _mm256_sub_ps(ix0,jx1);
1854 dy01 = _mm256_sub_ps(iy0,jy1);
1855 dz01 = _mm256_sub_ps(iz0,jz1);
1856 dx02 = _mm256_sub_ps(ix0,jx2);
1857 dy02 = _mm256_sub_ps(iy0,jy2);
1858 dz02 = _mm256_sub_ps(iz0,jz2);
1859 dx10 = _mm256_sub_ps(ix1,jx0);
1860 dy10 = _mm256_sub_ps(iy1,jy0);
1861 dz10 = _mm256_sub_ps(iz1,jz0);
1862 dx11 = _mm256_sub_ps(ix1,jx1);
1863 dy11 = _mm256_sub_ps(iy1,jy1);
1864 dz11 = _mm256_sub_ps(iz1,jz1);
1865 dx12 = _mm256_sub_ps(ix1,jx2);
1866 dy12 = _mm256_sub_ps(iy1,jy2);
1867 dz12 = _mm256_sub_ps(iz1,jz2);
1868 dx20 = _mm256_sub_ps(ix2,jx0);
1869 dy20 = _mm256_sub_ps(iy2,jy0);
1870 dz20 = _mm256_sub_ps(iz2,jz0);
1871 dx21 = _mm256_sub_ps(ix2,jx1);
1872 dy21 = _mm256_sub_ps(iy2,jy1);
1873 dz21 = _mm256_sub_ps(iz2,jz1);
1874 dx22 = _mm256_sub_ps(ix2,jx2);
1875 dy22 = _mm256_sub_ps(iy2,jy2);
1876 dz22 = _mm256_sub_ps(iz2,jz2);
1878 /* Calculate squared distance and things based on it */
1879 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1880 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1881 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1882 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1883 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1884 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1885 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1886 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1887 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1889 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1890 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
1891 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
1892 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1893 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
1894 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
1895 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1896 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
1897 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
1899 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1900 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
1901 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
1902 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1903 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
1904 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
1905 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1906 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
1907 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
1909 fjx0 = _mm256_setzero_ps();
1910 fjy0 = _mm256_setzero_ps();
1911 fjz0 = _mm256_setzero_ps();
1912 fjx1 = _mm256_setzero_ps();
1913 fjy1 = _mm256_setzero_ps();
1914 fjz1 = _mm256_setzero_ps();
1915 fjx2 = _mm256_setzero_ps();
1916 fjy2 = _mm256_setzero_ps();
1917 fjz2 = _mm256_setzero_ps();
1919 /**************************
1920 * CALCULATE INTERACTIONS *
1921 **************************/
1923 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1926 r00 = _mm256_mul_ps(rsq00,rinv00);
1928 /* EWALD ELECTROSTATICS */
1930 /* Analytical PME correction */
1931 zeta2 = _mm256_mul_ps(beta2,rsq00);
1932 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1933 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1934 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1935 felec = _mm256_mul_ps(qq00,felec);
1936 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1937 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1938 velec = _mm256_sub_ps(rinv00,pmecorrV);
1939 velec = _mm256_mul_ps(qq00,velec);
1941 /* LENNARD-JONES DISPERSION/REPULSION */
1943 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1944 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
1945 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
1946 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
1947 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
1949 d = _mm256_sub_ps(r00,rswitch);
1950 d = _mm256_max_ps(d,_mm256_setzero_ps());
1951 d2 = _mm256_mul_ps(d,d);
1952 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)))))));
1954 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1956 /* Evaluate switch function */
1957 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1958 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1959 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
1960 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1962 fscal = _mm256_add_ps(felec,fvdw);
1964 fscal = _mm256_and_ps(fscal,cutoff_mask);
1966 /* Calculate temporary vectorial force */
1967 tx = _mm256_mul_ps(fscal,dx00);
1968 ty = _mm256_mul_ps(fscal,dy00);
1969 tz = _mm256_mul_ps(fscal,dz00);
1971 /* Update vectorial force */
1972 fix0 = _mm256_add_ps(fix0,tx);
1973 fiy0 = _mm256_add_ps(fiy0,ty);
1974 fiz0 = _mm256_add_ps(fiz0,tz);
1976 fjx0 = _mm256_add_ps(fjx0,tx);
1977 fjy0 = _mm256_add_ps(fjy0,ty);
1978 fjz0 = _mm256_add_ps(fjz0,tz);
1982 /**************************
1983 * CALCULATE INTERACTIONS *
1984 **************************/
1986 if (gmx_mm256_any_lt(rsq01,rcutoff2))
1989 r01 = _mm256_mul_ps(rsq01,rinv01);
1991 /* EWALD ELECTROSTATICS */
1993 /* Analytical PME correction */
1994 zeta2 = _mm256_mul_ps(beta2,rsq01);
1995 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
1996 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1997 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1998 felec = _mm256_mul_ps(qq01,felec);
1999 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2000 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2001 velec = _mm256_sub_ps(rinv01,pmecorrV);
2002 velec = _mm256_mul_ps(qq01,velec);
2004 d = _mm256_sub_ps(r01,rswitch);
2005 d = _mm256_max_ps(d,_mm256_setzero_ps());
2006 d2 = _mm256_mul_ps(d,d);
2007 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)))))));
2009 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2011 /* Evaluate switch function */
2012 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2013 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
2014 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
2018 fscal = _mm256_and_ps(fscal,cutoff_mask);
2020 /* Calculate temporary vectorial force */
2021 tx = _mm256_mul_ps(fscal,dx01);
2022 ty = _mm256_mul_ps(fscal,dy01);
2023 tz = _mm256_mul_ps(fscal,dz01);
2025 /* Update vectorial force */
2026 fix0 = _mm256_add_ps(fix0,tx);
2027 fiy0 = _mm256_add_ps(fiy0,ty);
2028 fiz0 = _mm256_add_ps(fiz0,tz);
2030 fjx1 = _mm256_add_ps(fjx1,tx);
2031 fjy1 = _mm256_add_ps(fjy1,ty);
2032 fjz1 = _mm256_add_ps(fjz1,tz);
2036 /**************************
2037 * CALCULATE INTERACTIONS *
2038 **************************/
2040 if (gmx_mm256_any_lt(rsq02,rcutoff2))
2043 r02 = _mm256_mul_ps(rsq02,rinv02);
2045 /* EWALD ELECTROSTATICS */
2047 /* Analytical PME correction */
2048 zeta2 = _mm256_mul_ps(beta2,rsq02);
2049 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
2050 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2051 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2052 felec = _mm256_mul_ps(qq02,felec);
2053 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2054 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2055 velec = _mm256_sub_ps(rinv02,pmecorrV);
2056 velec = _mm256_mul_ps(qq02,velec);
2058 d = _mm256_sub_ps(r02,rswitch);
2059 d = _mm256_max_ps(d,_mm256_setzero_ps());
2060 d2 = _mm256_mul_ps(d,d);
2061 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)))))));
2063 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2065 /* Evaluate switch function */
2066 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2067 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
2068 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
2072 fscal = _mm256_and_ps(fscal,cutoff_mask);
2074 /* Calculate temporary vectorial force */
2075 tx = _mm256_mul_ps(fscal,dx02);
2076 ty = _mm256_mul_ps(fscal,dy02);
2077 tz = _mm256_mul_ps(fscal,dz02);
2079 /* Update vectorial force */
2080 fix0 = _mm256_add_ps(fix0,tx);
2081 fiy0 = _mm256_add_ps(fiy0,ty);
2082 fiz0 = _mm256_add_ps(fiz0,tz);
2084 fjx2 = _mm256_add_ps(fjx2,tx);
2085 fjy2 = _mm256_add_ps(fjy2,ty);
2086 fjz2 = _mm256_add_ps(fjz2,tz);
2090 /**************************
2091 * CALCULATE INTERACTIONS *
2092 **************************/
2094 if (gmx_mm256_any_lt(rsq10,rcutoff2))
2097 r10 = _mm256_mul_ps(rsq10,rinv10);
2099 /* EWALD ELECTROSTATICS */
2101 /* Analytical PME correction */
2102 zeta2 = _mm256_mul_ps(beta2,rsq10);
2103 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2104 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2105 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2106 felec = _mm256_mul_ps(qq10,felec);
2107 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2108 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2109 velec = _mm256_sub_ps(rinv10,pmecorrV);
2110 velec = _mm256_mul_ps(qq10,velec);
2112 d = _mm256_sub_ps(r10,rswitch);
2113 d = _mm256_max_ps(d,_mm256_setzero_ps());
2114 d2 = _mm256_mul_ps(d,d);
2115 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)))))));
2117 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2119 /* Evaluate switch function */
2120 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2121 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
2122 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
2126 fscal = _mm256_and_ps(fscal,cutoff_mask);
2128 /* Calculate temporary vectorial force */
2129 tx = _mm256_mul_ps(fscal,dx10);
2130 ty = _mm256_mul_ps(fscal,dy10);
2131 tz = _mm256_mul_ps(fscal,dz10);
2133 /* Update vectorial force */
2134 fix1 = _mm256_add_ps(fix1,tx);
2135 fiy1 = _mm256_add_ps(fiy1,ty);
2136 fiz1 = _mm256_add_ps(fiz1,tz);
2138 fjx0 = _mm256_add_ps(fjx0,tx);
2139 fjy0 = _mm256_add_ps(fjy0,ty);
2140 fjz0 = _mm256_add_ps(fjz0,tz);
2144 /**************************
2145 * CALCULATE INTERACTIONS *
2146 **************************/
2148 if (gmx_mm256_any_lt(rsq11,rcutoff2))
2151 r11 = _mm256_mul_ps(rsq11,rinv11);
2153 /* EWALD ELECTROSTATICS */
2155 /* Analytical PME correction */
2156 zeta2 = _mm256_mul_ps(beta2,rsq11);
2157 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2158 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2159 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2160 felec = _mm256_mul_ps(qq11,felec);
2161 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2162 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2163 velec = _mm256_sub_ps(rinv11,pmecorrV);
2164 velec = _mm256_mul_ps(qq11,velec);
2166 d = _mm256_sub_ps(r11,rswitch);
2167 d = _mm256_max_ps(d,_mm256_setzero_ps());
2168 d2 = _mm256_mul_ps(d,d);
2169 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)))))));
2171 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2173 /* Evaluate switch function */
2174 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2175 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
2176 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
2180 fscal = _mm256_and_ps(fscal,cutoff_mask);
2182 /* Calculate temporary vectorial force */
2183 tx = _mm256_mul_ps(fscal,dx11);
2184 ty = _mm256_mul_ps(fscal,dy11);
2185 tz = _mm256_mul_ps(fscal,dz11);
2187 /* Update vectorial force */
2188 fix1 = _mm256_add_ps(fix1,tx);
2189 fiy1 = _mm256_add_ps(fiy1,ty);
2190 fiz1 = _mm256_add_ps(fiz1,tz);
2192 fjx1 = _mm256_add_ps(fjx1,tx);
2193 fjy1 = _mm256_add_ps(fjy1,ty);
2194 fjz1 = _mm256_add_ps(fjz1,tz);
2198 /**************************
2199 * CALCULATE INTERACTIONS *
2200 **************************/
2202 if (gmx_mm256_any_lt(rsq12,rcutoff2))
2205 r12 = _mm256_mul_ps(rsq12,rinv12);
2207 /* EWALD ELECTROSTATICS */
2209 /* Analytical PME correction */
2210 zeta2 = _mm256_mul_ps(beta2,rsq12);
2211 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2212 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2213 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2214 felec = _mm256_mul_ps(qq12,felec);
2215 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2216 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2217 velec = _mm256_sub_ps(rinv12,pmecorrV);
2218 velec = _mm256_mul_ps(qq12,velec);
2220 d = _mm256_sub_ps(r12,rswitch);
2221 d = _mm256_max_ps(d,_mm256_setzero_ps());
2222 d2 = _mm256_mul_ps(d,d);
2223 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)))))));
2225 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2227 /* Evaluate switch function */
2228 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2229 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
2230 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
2234 fscal = _mm256_and_ps(fscal,cutoff_mask);
2236 /* Calculate temporary vectorial force */
2237 tx = _mm256_mul_ps(fscal,dx12);
2238 ty = _mm256_mul_ps(fscal,dy12);
2239 tz = _mm256_mul_ps(fscal,dz12);
2241 /* Update vectorial force */
2242 fix1 = _mm256_add_ps(fix1,tx);
2243 fiy1 = _mm256_add_ps(fiy1,ty);
2244 fiz1 = _mm256_add_ps(fiz1,tz);
2246 fjx2 = _mm256_add_ps(fjx2,tx);
2247 fjy2 = _mm256_add_ps(fjy2,ty);
2248 fjz2 = _mm256_add_ps(fjz2,tz);
2252 /**************************
2253 * CALCULATE INTERACTIONS *
2254 **************************/
2256 if (gmx_mm256_any_lt(rsq20,rcutoff2))
2259 r20 = _mm256_mul_ps(rsq20,rinv20);
2261 /* EWALD ELECTROSTATICS */
2263 /* Analytical PME correction */
2264 zeta2 = _mm256_mul_ps(beta2,rsq20);
2265 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
2266 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2267 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2268 felec = _mm256_mul_ps(qq20,felec);
2269 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2270 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2271 velec = _mm256_sub_ps(rinv20,pmecorrV);
2272 velec = _mm256_mul_ps(qq20,velec);
2274 d = _mm256_sub_ps(r20,rswitch);
2275 d = _mm256_max_ps(d,_mm256_setzero_ps());
2276 d2 = _mm256_mul_ps(d,d);
2277 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)))))));
2279 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2281 /* Evaluate switch function */
2282 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2283 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
2284 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
2288 fscal = _mm256_and_ps(fscal,cutoff_mask);
2290 /* Calculate temporary vectorial force */
2291 tx = _mm256_mul_ps(fscal,dx20);
2292 ty = _mm256_mul_ps(fscal,dy20);
2293 tz = _mm256_mul_ps(fscal,dz20);
2295 /* Update vectorial force */
2296 fix2 = _mm256_add_ps(fix2,tx);
2297 fiy2 = _mm256_add_ps(fiy2,ty);
2298 fiz2 = _mm256_add_ps(fiz2,tz);
2300 fjx0 = _mm256_add_ps(fjx0,tx);
2301 fjy0 = _mm256_add_ps(fjy0,ty);
2302 fjz0 = _mm256_add_ps(fjz0,tz);
2306 /**************************
2307 * CALCULATE INTERACTIONS *
2308 **************************/
2310 if (gmx_mm256_any_lt(rsq21,rcutoff2))
2313 r21 = _mm256_mul_ps(rsq21,rinv21);
2315 /* EWALD ELECTROSTATICS */
2317 /* Analytical PME correction */
2318 zeta2 = _mm256_mul_ps(beta2,rsq21);
2319 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2320 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2321 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2322 felec = _mm256_mul_ps(qq21,felec);
2323 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2324 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2325 velec = _mm256_sub_ps(rinv21,pmecorrV);
2326 velec = _mm256_mul_ps(qq21,velec);
2328 d = _mm256_sub_ps(r21,rswitch);
2329 d = _mm256_max_ps(d,_mm256_setzero_ps());
2330 d2 = _mm256_mul_ps(d,d);
2331 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)))))));
2333 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2335 /* Evaluate switch function */
2336 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2337 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
2338 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2342 fscal = _mm256_and_ps(fscal,cutoff_mask);
2344 /* Calculate temporary vectorial force */
2345 tx = _mm256_mul_ps(fscal,dx21);
2346 ty = _mm256_mul_ps(fscal,dy21);
2347 tz = _mm256_mul_ps(fscal,dz21);
2349 /* Update vectorial force */
2350 fix2 = _mm256_add_ps(fix2,tx);
2351 fiy2 = _mm256_add_ps(fiy2,ty);
2352 fiz2 = _mm256_add_ps(fiz2,tz);
2354 fjx1 = _mm256_add_ps(fjx1,tx);
2355 fjy1 = _mm256_add_ps(fjy1,ty);
2356 fjz1 = _mm256_add_ps(fjz1,tz);
2360 /**************************
2361 * CALCULATE INTERACTIONS *
2362 **************************/
2364 if (gmx_mm256_any_lt(rsq22,rcutoff2))
2367 r22 = _mm256_mul_ps(rsq22,rinv22);
2369 /* EWALD ELECTROSTATICS */
2371 /* Analytical PME correction */
2372 zeta2 = _mm256_mul_ps(beta2,rsq22);
2373 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
2374 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2375 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2376 felec = _mm256_mul_ps(qq22,felec);
2377 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2378 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2379 velec = _mm256_sub_ps(rinv22,pmecorrV);
2380 velec = _mm256_mul_ps(qq22,velec);
2382 d = _mm256_sub_ps(r22,rswitch);
2383 d = _mm256_max_ps(d,_mm256_setzero_ps());
2384 d2 = _mm256_mul_ps(d,d);
2385 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)))))));
2387 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2389 /* Evaluate switch function */
2390 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2391 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
2392 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
2396 fscal = _mm256_and_ps(fscal,cutoff_mask);
2398 /* Calculate temporary vectorial force */
2399 tx = _mm256_mul_ps(fscal,dx22);
2400 ty = _mm256_mul_ps(fscal,dy22);
2401 tz = _mm256_mul_ps(fscal,dz22);
2403 /* Update vectorial force */
2404 fix2 = _mm256_add_ps(fix2,tx);
2405 fiy2 = _mm256_add_ps(fiy2,ty);
2406 fiz2 = _mm256_add_ps(fiz2,tz);
2408 fjx2 = _mm256_add_ps(fjx2,tx);
2409 fjy2 = _mm256_add_ps(fjy2,ty);
2410 fjz2 = _mm256_add_ps(fjz2,tz);
2414 fjptrA = f+j_coord_offsetA;
2415 fjptrB = f+j_coord_offsetB;
2416 fjptrC = f+j_coord_offsetC;
2417 fjptrD = f+j_coord_offsetD;
2418 fjptrE = f+j_coord_offsetE;
2419 fjptrF = f+j_coord_offsetF;
2420 fjptrG = f+j_coord_offsetG;
2421 fjptrH = f+j_coord_offsetH;
2423 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2424 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2426 /* Inner loop uses 960 flops */
2429 if(jidx<j_index_end)
2432 /* Get j neighbor index, and coordinate index */
2433 jnrlistA = jjnr[jidx];
2434 jnrlistB = jjnr[jidx+1];
2435 jnrlistC = jjnr[jidx+2];
2436 jnrlistD = jjnr[jidx+3];
2437 jnrlistE = jjnr[jidx+4];
2438 jnrlistF = jjnr[jidx+5];
2439 jnrlistG = jjnr[jidx+6];
2440 jnrlistH = jjnr[jidx+7];
2441 /* Sign of each element will be negative for non-real atoms.
2442 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2443 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2445 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
2446 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
2448 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
2449 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
2450 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
2451 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
2452 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
2453 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
2454 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
2455 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
2456 j_coord_offsetA = DIM*jnrA;
2457 j_coord_offsetB = DIM*jnrB;
2458 j_coord_offsetC = DIM*jnrC;
2459 j_coord_offsetD = DIM*jnrD;
2460 j_coord_offsetE = DIM*jnrE;
2461 j_coord_offsetF = DIM*jnrF;
2462 j_coord_offsetG = DIM*jnrG;
2463 j_coord_offsetH = DIM*jnrH;
2465 /* load j atom coordinates */
2466 gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2467 x+j_coord_offsetC,x+j_coord_offsetD,
2468 x+j_coord_offsetE,x+j_coord_offsetF,
2469 x+j_coord_offsetG,x+j_coord_offsetH,
2470 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
2472 /* Calculate displacement vector */
2473 dx00 = _mm256_sub_ps(ix0,jx0);
2474 dy00 = _mm256_sub_ps(iy0,jy0);
2475 dz00 = _mm256_sub_ps(iz0,jz0);
2476 dx01 = _mm256_sub_ps(ix0,jx1);
2477 dy01 = _mm256_sub_ps(iy0,jy1);
2478 dz01 = _mm256_sub_ps(iz0,jz1);
2479 dx02 = _mm256_sub_ps(ix0,jx2);
2480 dy02 = _mm256_sub_ps(iy0,jy2);
2481 dz02 = _mm256_sub_ps(iz0,jz2);
2482 dx10 = _mm256_sub_ps(ix1,jx0);
2483 dy10 = _mm256_sub_ps(iy1,jy0);
2484 dz10 = _mm256_sub_ps(iz1,jz0);
2485 dx11 = _mm256_sub_ps(ix1,jx1);
2486 dy11 = _mm256_sub_ps(iy1,jy1);
2487 dz11 = _mm256_sub_ps(iz1,jz1);
2488 dx12 = _mm256_sub_ps(ix1,jx2);
2489 dy12 = _mm256_sub_ps(iy1,jy2);
2490 dz12 = _mm256_sub_ps(iz1,jz2);
2491 dx20 = _mm256_sub_ps(ix2,jx0);
2492 dy20 = _mm256_sub_ps(iy2,jy0);
2493 dz20 = _mm256_sub_ps(iz2,jz0);
2494 dx21 = _mm256_sub_ps(ix2,jx1);
2495 dy21 = _mm256_sub_ps(iy2,jy1);
2496 dz21 = _mm256_sub_ps(iz2,jz1);
2497 dx22 = _mm256_sub_ps(ix2,jx2);
2498 dy22 = _mm256_sub_ps(iy2,jy2);
2499 dz22 = _mm256_sub_ps(iz2,jz2);
2501 /* Calculate squared distance and things based on it */
2502 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
2503 rsq01 = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
2504 rsq02 = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
2505 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
2506 rsq11 = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
2507 rsq12 = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
2508 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
2509 rsq21 = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
2510 rsq22 = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
2512 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
2513 rinv01 = gmx_mm256_invsqrt_ps(rsq01);
2514 rinv02 = gmx_mm256_invsqrt_ps(rsq02);
2515 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
2516 rinv11 = gmx_mm256_invsqrt_ps(rsq11);
2517 rinv12 = gmx_mm256_invsqrt_ps(rsq12);
2518 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
2519 rinv21 = gmx_mm256_invsqrt_ps(rsq21);
2520 rinv22 = gmx_mm256_invsqrt_ps(rsq22);
2522 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
2523 rinvsq01 = _mm256_mul_ps(rinv01,rinv01);
2524 rinvsq02 = _mm256_mul_ps(rinv02,rinv02);
2525 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
2526 rinvsq11 = _mm256_mul_ps(rinv11,rinv11);
2527 rinvsq12 = _mm256_mul_ps(rinv12,rinv12);
2528 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
2529 rinvsq21 = _mm256_mul_ps(rinv21,rinv21);
2530 rinvsq22 = _mm256_mul_ps(rinv22,rinv22);
2532 fjx0 = _mm256_setzero_ps();
2533 fjy0 = _mm256_setzero_ps();
2534 fjz0 = _mm256_setzero_ps();
2535 fjx1 = _mm256_setzero_ps();
2536 fjy1 = _mm256_setzero_ps();
2537 fjz1 = _mm256_setzero_ps();
2538 fjx2 = _mm256_setzero_ps();
2539 fjy2 = _mm256_setzero_ps();
2540 fjz2 = _mm256_setzero_ps();
2542 /**************************
2543 * CALCULATE INTERACTIONS *
2544 **************************/
2546 if (gmx_mm256_any_lt(rsq00,rcutoff2))
2549 r00 = _mm256_mul_ps(rsq00,rinv00);
2550 r00 = _mm256_andnot_ps(dummy_mask,r00);
2552 /* EWALD ELECTROSTATICS */
2554 /* Analytical PME correction */
2555 zeta2 = _mm256_mul_ps(beta2,rsq00);
2556 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
2557 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2558 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2559 felec = _mm256_mul_ps(qq00,felec);
2560 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2561 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2562 velec = _mm256_sub_ps(rinv00,pmecorrV);
2563 velec = _mm256_mul_ps(qq00,velec);
2565 /* LENNARD-JONES DISPERSION/REPULSION */
2567 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2568 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
2569 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
2570 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
2571 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
2573 d = _mm256_sub_ps(r00,rswitch);
2574 d = _mm256_max_ps(d,_mm256_setzero_ps());
2575 d2 = _mm256_mul_ps(d,d);
2576 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)))))));
2578 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2580 /* Evaluate switch function */
2581 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2582 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
2583 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
2584 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
2586 fscal = _mm256_add_ps(felec,fvdw);
2588 fscal = _mm256_and_ps(fscal,cutoff_mask);
2590 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2592 /* Calculate temporary vectorial force */
2593 tx = _mm256_mul_ps(fscal,dx00);
2594 ty = _mm256_mul_ps(fscal,dy00);
2595 tz = _mm256_mul_ps(fscal,dz00);
2597 /* Update vectorial force */
2598 fix0 = _mm256_add_ps(fix0,tx);
2599 fiy0 = _mm256_add_ps(fiy0,ty);
2600 fiz0 = _mm256_add_ps(fiz0,tz);
2602 fjx0 = _mm256_add_ps(fjx0,tx);
2603 fjy0 = _mm256_add_ps(fjy0,ty);
2604 fjz0 = _mm256_add_ps(fjz0,tz);
2608 /**************************
2609 * CALCULATE INTERACTIONS *
2610 **************************/
2612 if (gmx_mm256_any_lt(rsq01,rcutoff2))
2615 r01 = _mm256_mul_ps(rsq01,rinv01);
2616 r01 = _mm256_andnot_ps(dummy_mask,r01);
2618 /* EWALD ELECTROSTATICS */
2620 /* Analytical PME correction */
2621 zeta2 = _mm256_mul_ps(beta2,rsq01);
2622 rinv3 = _mm256_mul_ps(rinvsq01,rinv01);
2623 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2624 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2625 felec = _mm256_mul_ps(qq01,felec);
2626 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2627 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2628 velec = _mm256_sub_ps(rinv01,pmecorrV);
2629 velec = _mm256_mul_ps(qq01,velec);
2631 d = _mm256_sub_ps(r01,rswitch);
2632 d = _mm256_max_ps(d,_mm256_setzero_ps());
2633 d2 = _mm256_mul_ps(d,d);
2634 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)))))));
2636 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2638 /* Evaluate switch function */
2639 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2640 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv01,_mm256_mul_ps(velec,dsw)) );
2641 cutoff_mask = _mm256_cmp_ps(rsq01,rcutoff2,_CMP_LT_OQ);
2645 fscal = _mm256_and_ps(fscal,cutoff_mask);
2647 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2649 /* Calculate temporary vectorial force */
2650 tx = _mm256_mul_ps(fscal,dx01);
2651 ty = _mm256_mul_ps(fscal,dy01);
2652 tz = _mm256_mul_ps(fscal,dz01);
2654 /* Update vectorial force */
2655 fix0 = _mm256_add_ps(fix0,tx);
2656 fiy0 = _mm256_add_ps(fiy0,ty);
2657 fiz0 = _mm256_add_ps(fiz0,tz);
2659 fjx1 = _mm256_add_ps(fjx1,tx);
2660 fjy1 = _mm256_add_ps(fjy1,ty);
2661 fjz1 = _mm256_add_ps(fjz1,tz);
2665 /**************************
2666 * CALCULATE INTERACTIONS *
2667 **************************/
2669 if (gmx_mm256_any_lt(rsq02,rcutoff2))
2672 r02 = _mm256_mul_ps(rsq02,rinv02);
2673 r02 = _mm256_andnot_ps(dummy_mask,r02);
2675 /* EWALD ELECTROSTATICS */
2677 /* Analytical PME correction */
2678 zeta2 = _mm256_mul_ps(beta2,rsq02);
2679 rinv3 = _mm256_mul_ps(rinvsq02,rinv02);
2680 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2681 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2682 felec = _mm256_mul_ps(qq02,felec);
2683 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2684 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2685 velec = _mm256_sub_ps(rinv02,pmecorrV);
2686 velec = _mm256_mul_ps(qq02,velec);
2688 d = _mm256_sub_ps(r02,rswitch);
2689 d = _mm256_max_ps(d,_mm256_setzero_ps());
2690 d2 = _mm256_mul_ps(d,d);
2691 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)))))));
2693 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2695 /* Evaluate switch function */
2696 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2697 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv02,_mm256_mul_ps(velec,dsw)) );
2698 cutoff_mask = _mm256_cmp_ps(rsq02,rcutoff2,_CMP_LT_OQ);
2702 fscal = _mm256_and_ps(fscal,cutoff_mask);
2704 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2706 /* Calculate temporary vectorial force */
2707 tx = _mm256_mul_ps(fscal,dx02);
2708 ty = _mm256_mul_ps(fscal,dy02);
2709 tz = _mm256_mul_ps(fscal,dz02);
2711 /* Update vectorial force */
2712 fix0 = _mm256_add_ps(fix0,tx);
2713 fiy0 = _mm256_add_ps(fiy0,ty);
2714 fiz0 = _mm256_add_ps(fiz0,tz);
2716 fjx2 = _mm256_add_ps(fjx2,tx);
2717 fjy2 = _mm256_add_ps(fjy2,ty);
2718 fjz2 = _mm256_add_ps(fjz2,tz);
2722 /**************************
2723 * CALCULATE INTERACTIONS *
2724 **************************/
2726 if (gmx_mm256_any_lt(rsq10,rcutoff2))
2729 r10 = _mm256_mul_ps(rsq10,rinv10);
2730 r10 = _mm256_andnot_ps(dummy_mask,r10);
2732 /* EWALD ELECTROSTATICS */
2734 /* Analytical PME correction */
2735 zeta2 = _mm256_mul_ps(beta2,rsq10);
2736 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
2737 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2738 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2739 felec = _mm256_mul_ps(qq10,felec);
2740 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2741 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2742 velec = _mm256_sub_ps(rinv10,pmecorrV);
2743 velec = _mm256_mul_ps(qq10,velec);
2745 d = _mm256_sub_ps(r10,rswitch);
2746 d = _mm256_max_ps(d,_mm256_setzero_ps());
2747 d2 = _mm256_mul_ps(d,d);
2748 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)))))));
2750 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2752 /* Evaluate switch function */
2753 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2754 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
2755 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
2759 fscal = _mm256_and_ps(fscal,cutoff_mask);
2761 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2763 /* Calculate temporary vectorial force */
2764 tx = _mm256_mul_ps(fscal,dx10);
2765 ty = _mm256_mul_ps(fscal,dy10);
2766 tz = _mm256_mul_ps(fscal,dz10);
2768 /* Update vectorial force */
2769 fix1 = _mm256_add_ps(fix1,tx);
2770 fiy1 = _mm256_add_ps(fiy1,ty);
2771 fiz1 = _mm256_add_ps(fiz1,tz);
2773 fjx0 = _mm256_add_ps(fjx0,tx);
2774 fjy0 = _mm256_add_ps(fjy0,ty);
2775 fjz0 = _mm256_add_ps(fjz0,tz);
2779 /**************************
2780 * CALCULATE INTERACTIONS *
2781 **************************/
2783 if (gmx_mm256_any_lt(rsq11,rcutoff2))
2786 r11 = _mm256_mul_ps(rsq11,rinv11);
2787 r11 = _mm256_andnot_ps(dummy_mask,r11);
2789 /* EWALD ELECTROSTATICS */
2791 /* Analytical PME correction */
2792 zeta2 = _mm256_mul_ps(beta2,rsq11);
2793 rinv3 = _mm256_mul_ps(rinvsq11,rinv11);
2794 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2795 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2796 felec = _mm256_mul_ps(qq11,felec);
2797 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2798 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2799 velec = _mm256_sub_ps(rinv11,pmecorrV);
2800 velec = _mm256_mul_ps(qq11,velec);
2802 d = _mm256_sub_ps(r11,rswitch);
2803 d = _mm256_max_ps(d,_mm256_setzero_ps());
2804 d2 = _mm256_mul_ps(d,d);
2805 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)))))));
2807 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2809 /* Evaluate switch function */
2810 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2811 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv11,_mm256_mul_ps(velec,dsw)) );
2812 cutoff_mask = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
2816 fscal = _mm256_and_ps(fscal,cutoff_mask);
2818 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2820 /* Calculate temporary vectorial force */
2821 tx = _mm256_mul_ps(fscal,dx11);
2822 ty = _mm256_mul_ps(fscal,dy11);
2823 tz = _mm256_mul_ps(fscal,dz11);
2825 /* Update vectorial force */
2826 fix1 = _mm256_add_ps(fix1,tx);
2827 fiy1 = _mm256_add_ps(fiy1,ty);
2828 fiz1 = _mm256_add_ps(fiz1,tz);
2830 fjx1 = _mm256_add_ps(fjx1,tx);
2831 fjy1 = _mm256_add_ps(fjy1,ty);
2832 fjz1 = _mm256_add_ps(fjz1,tz);
2836 /**************************
2837 * CALCULATE INTERACTIONS *
2838 **************************/
2840 if (gmx_mm256_any_lt(rsq12,rcutoff2))
2843 r12 = _mm256_mul_ps(rsq12,rinv12);
2844 r12 = _mm256_andnot_ps(dummy_mask,r12);
2846 /* EWALD ELECTROSTATICS */
2848 /* Analytical PME correction */
2849 zeta2 = _mm256_mul_ps(beta2,rsq12);
2850 rinv3 = _mm256_mul_ps(rinvsq12,rinv12);
2851 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2852 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2853 felec = _mm256_mul_ps(qq12,felec);
2854 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2855 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2856 velec = _mm256_sub_ps(rinv12,pmecorrV);
2857 velec = _mm256_mul_ps(qq12,velec);
2859 d = _mm256_sub_ps(r12,rswitch);
2860 d = _mm256_max_ps(d,_mm256_setzero_ps());
2861 d2 = _mm256_mul_ps(d,d);
2862 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)))))));
2864 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2866 /* Evaluate switch function */
2867 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2868 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv12,_mm256_mul_ps(velec,dsw)) );
2869 cutoff_mask = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
2873 fscal = _mm256_and_ps(fscal,cutoff_mask);
2875 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2877 /* Calculate temporary vectorial force */
2878 tx = _mm256_mul_ps(fscal,dx12);
2879 ty = _mm256_mul_ps(fscal,dy12);
2880 tz = _mm256_mul_ps(fscal,dz12);
2882 /* Update vectorial force */
2883 fix1 = _mm256_add_ps(fix1,tx);
2884 fiy1 = _mm256_add_ps(fiy1,ty);
2885 fiz1 = _mm256_add_ps(fiz1,tz);
2887 fjx2 = _mm256_add_ps(fjx2,tx);
2888 fjy2 = _mm256_add_ps(fjy2,ty);
2889 fjz2 = _mm256_add_ps(fjz2,tz);
2893 /**************************
2894 * CALCULATE INTERACTIONS *
2895 **************************/
2897 if (gmx_mm256_any_lt(rsq20,rcutoff2))
2900 r20 = _mm256_mul_ps(rsq20,rinv20);
2901 r20 = _mm256_andnot_ps(dummy_mask,r20);
2903 /* EWALD ELECTROSTATICS */
2905 /* Analytical PME correction */
2906 zeta2 = _mm256_mul_ps(beta2,rsq20);
2907 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
2908 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2909 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2910 felec = _mm256_mul_ps(qq20,felec);
2911 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2912 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2913 velec = _mm256_sub_ps(rinv20,pmecorrV);
2914 velec = _mm256_mul_ps(qq20,velec);
2916 d = _mm256_sub_ps(r20,rswitch);
2917 d = _mm256_max_ps(d,_mm256_setzero_ps());
2918 d2 = _mm256_mul_ps(d,d);
2919 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)))))));
2921 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2923 /* Evaluate switch function */
2924 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2925 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
2926 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
2930 fscal = _mm256_and_ps(fscal,cutoff_mask);
2932 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2934 /* Calculate temporary vectorial force */
2935 tx = _mm256_mul_ps(fscal,dx20);
2936 ty = _mm256_mul_ps(fscal,dy20);
2937 tz = _mm256_mul_ps(fscal,dz20);
2939 /* Update vectorial force */
2940 fix2 = _mm256_add_ps(fix2,tx);
2941 fiy2 = _mm256_add_ps(fiy2,ty);
2942 fiz2 = _mm256_add_ps(fiz2,tz);
2944 fjx0 = _mm256_add_ps(fjx0,tx);
2945 fjy0 = _mm256_add_ps(fjy0,ty);
2946 fjz0 = _mm256_add_ps(fjz0,tz);
2950 /**************************
2951 * CALCULATE INTERACTIONS *
2952 **************************/
2954 if (gmx_mm256_any_lt(rsq21,rcutoff2))
2957 r21 = _mm256_mul_ps(rsq21,rinv21);
2958 r21 = _mm256_andnot_ps(dummy_mask,r21);
2960 /* EWALD ELECTROSTATICS */
2962 /* Analytical PME correction */
2963 zeta2 = _mm256_mul_ps(beta2,rsq21);
2964 rinv3 = _mm256_mul_ps(rinvsq21,rinv21);
2965 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
2966 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2967 felec = _mm256_mul_ps(qq21,felec);
2968 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
2969 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
2970 velec = _mm256_sub_ps(rinv21,pmecorrV);
2971 velec = _mm256_mul_ps(qq21,velec);
2973 d = _mm256_sub_ps(r21,rswitch);
2974 d = _mm256_max_ps(d,_mm256_setzero_ps());
2975 d2 = _mm256_mul_ps(d,d);
2976 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)))))));
2978 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2980 /* Evaluate switch function */
2981 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2982 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv21,_mm256_mul_ps(velec,dsw)) );
2983 cutoff_mask = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2987 fscal = _mm256_and_ps(fscal,cutoff_mask);
2989 fscal = _mm256_andnot_ps(dummy_mask,fscal);
2991 /* Calculate temporary vectorial force */
2992 tx = _mm256_mul_ps(fscal,dx21);
2993 ty = _mm256_mul_ps(fscal,dy21);
2994 tz = _mm256_mul_ps(fscal,dz21);
2996 /* Update vectorial force */
2997 fix2 = _mm256_add_ps(fix2,tx);
2998 fiy2 = _mm256_add_ps(fiy2,ty);
2999 fiz2 = _mm256_add_ps(fiz2,tz);
3001 fjx1 = _mm256_add_ps(fjx1,tx);
3002 fjy1 = _mm256_add_ps(fjy1,ty);
3003 fjz1 = _mm256_add_ps(fjz1,tz);
3007 /**************************
3008 * CALCULATE INTERACTIONS *
3009 **************************/
3011 if (gmx_mm256_any_lt(rsq22,rcutoff2))
3014 r22 = _mm256_mul_ps(rsq22,rinv22);
3015 r22 = _mm256_andnot_ps(dummy_mask,r22);
3017 /* EWALD ELECTROSTATICS */
3019 /* Analytical PME correction */
3020 zeta2 = _mm256_mul_ps(beta2,rsq22);
3021 rinv3 = _mm256_mul_ps(rinvsq22,rinv22);
3022 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
3023 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
3024 felec = _mm256_mul_ps(qq22,felec);
3025 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
3026 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
3027 velec = _mm256_sub_ps(rinv22,pmecorrV);
3028 velec = _mm256_mul_ps(qq22,velec);
3030 d = _mm256_sub_ps(r22,rswitch);
3031 d = _mm256_max_ps(d,_mm256_setzero_ps());
3032 d2 = _mm256_mul_ps(d,d);
3033 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)))))));
3035 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
3037 /* Evaluate switch function */
3038 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
3039 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv22,_mm256_mul_ps(velec,dsw)) );
3040 cutoff_mask = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
3044 fscal = _mm256_and_ps(fscal,cutoff_mask);
3046 fscal = _mm256_andnot_ps(dummy_mask,fscal);
3048 /* Calculate temporary vectorial force */
3049 tx = _mm256_mul_ps(fscal,dx22);
3050 ty = _mm256_mul_ps(fscal,dy22);
3051 tz = _mm256_mul_ps(fscal,dz22);
3053 /* Update vectorial force */
3054 fix2 = _mm256_add_ps(fix2,tx);
3055 fiy2 = _mm256_add_ps(fiy2,ty);
3056 fiz2 = _mm256_add_ps(fiz2,tz);
3058 fjx2 = _mm256_add_ps(fjx2,tx);
3059 fjy2 = _mm256_add_ps(fjy2,ty);
3060 fjz2 = _mm256_add_ps(fjz2,tz);
3064 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
3065 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
3066 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
3067 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
3068 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
3069 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
3070 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
3071 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
3073 gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
3074 fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
3076 /* Inner loop uses 969 flops */
3079 /* End of innermost loop */
3081 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
3082 f+i_coord_offset,fshift+i_shift_offset);
3084 /* Increment number of inner iterations */
3085 inneriter += j_index_end - j_index_start;
3087 /* Outer loop uses 18 flops */
3090 /* Increment number of outer iterations */
3093 /* Update outer/inner flops */
3095 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*969);