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_GeomW3P1_VF_avx_256_single
52 * Electrostatics interaction: Ewald
53 * VdW interaction: LennardJones
54 * Geometry: Water3-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_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 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
95 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
96 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
97 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
100 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
103 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
104 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
106 __m128i ewitab_lo,ewitab_hi;
107 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
108 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
110 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
111 real rswitch_scalar,d_scalar;
112 __m256 dummy_mask,cutoff_mask;
113 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
114 __m256 one = _mm256_set1_ps(1.0);
115 __m256 two = _mm256_set1_ps(2.0);
121 jindex = nlist->jindex;
123 shiftidx = nlist->shift;
125 shiftvec = fr->shift_vec[0];
126 fshift = fr->fshift[0];
127 facel = _mm256_set1_ps(fr->epsfac);
128 charge = mdatoms->chargeA;
129 nvdwtype = fr->ntype;
131 vdwtype = mdatoms->typeA;
133 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
134 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
135 beta2 = _mm256_mul_ps(beta,beta);
136 beta3 = _mm256_mul_ps(beta,beta2);
138 ewtab = fr->ic->tabq_coul_FDV0;
139 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
140 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
142 /* Setup water-specific parameters */
143 inr = nlist->iinr[0];
144 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
145 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
146 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
147 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
149 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
150 rcutoff_scalar = fr->rcoulomb;
151 rcutoff = _mm256_set1_ps(rcutoff_scalar);
152 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
154 rswitch_scalar = fr->rcoulomb_switch;
155 rswitch = _mm256_set1_ps(rswitch_scalar);
156 /* Setup switch parameters */
157 d_scalar = rcutoff_scalar-rswitch_scalar;
158 d = _mm256_set1_ps(d_scalar);
159 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
160 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
161 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
162 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
163 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
164 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
166 /* Avoid stupid compiler warnings */
167 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
180 for(iidx=0;iidx<4*DIM;iidx++)
185 /* Start outer loop over neighborlists */
186 for(iidx=0; iidx<nri; iidx++)
188 /* Load shift vector for this list */
189 i_shift_offset = DIM*shiftidx[iidx];
191 /* Load limits for loop over neighbors */
192 j_index_start = jindex[iidx];
193 j_index_end = jindex[iidx+1];
195 /* Get outer coordinate index */
197 i_coord_offset = DIM*inr;
199 /* Load i particle coords and add shift vector */
200 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
201 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
203 fix0 = _mm256_setzero_ps();
204 fiy0 = _mm256_setzero_ps();
205 fiz0 = _mm256_setzero_ps();
206 fix1 = _mm256_setzero_ps();
207 fiy1 = _mm256_setzero_ps();
208 fiz1 = _mm256_setzero_ps();
209 fix2 = _mm256_setzero_ps();
210 fiy2 = _mm256_setzero_ps();
211 fiz2 = _mm256_setzero_ps();
213 /* Reset potential sums */
214 velecsum = _mm256_setzero_ps();
215 vvdwsum = _mm256_setzero_ps();
217 /* Start inner kernel loop */
218 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
221 /* Get j neighbor index, and coordinate index */
230 j_coord_offsetA = DIM*jnrA;
231 j_coord_offsetB = DIM*jnrB;
232 j_coord_offsetC = DIM*jnrC;
233 j_coord_offsetD = DIM*jnrD;
234 j_coord_offsetE = DIM*jnrE;
235 j_coord_offsetF = DIM*jnrF;
236 j_coord_offsetG = DIM*jnrG;
237 j_coord_offsetH = DIM*jnrH;
239 /* load j atom coordinates */
240 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
241 x+j_coord_offsetC,x+j_coord_offsetD,
242 x+j_coord_offsetE,x+j_coord_offsetF,
243 x+j_coord_offsetG,x+j_coord_offsetH,
246 /* Calculate displacement vector */
247 dx00 = _mm256_sub_ps(ix0,jx0);
248 dy00 = _mm256_sub_ps(iy0,jy0);
249 dz00 = _mm256_sub_ps(iz0,jz0);
250 dx10 = _mm256_sub_ps(ix1,jx0);
251 dy10 = _mm256_sub_ps(iy1,jy0);
252 dz10 = _mm256_sub_ps(iz1,jz0);
253 dx20 = _mm256_sub_ps(ix2,jx0);
254 dy20 = _mm256_sub_ps(iy2,jy0);
255 dz20 = _mm256_sub_ps(iz2,jz0);
257 /* Calculate squared distance and things based on it */
258 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
259 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
260 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
262 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
263 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
264 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
266 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
267 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
268 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
270 /* Load parameters for j particles */
271 jq0 = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
272 charge+jnrC+0,charge+jnrD+0,
273 charge+jnrE+0,charge+jnrF+0,
274 charge+jnrG+0,charge+jnrH+0);
275 vdwjidx0A = 2*vdwtype[jnrA+0];
276 vdwjidx0B = 2*vdwtype[jnrB+0];
277 vdwjidx0C = 2*vdwtype[jnrC+0];
278 vdwjidx0D = 2*vdwtype[jnrD+0];
279 vdwjidx0E = 2*vdwtype[jnrE+0];
280 vdwjidx0F = 2*vdwtype[jnrF+0];
281 vdwjidx0G = 2*vdwtype[jnrG+0];
282 vdwjidx0H = 2*vdwtype[jnrH+0];
284 fjx0 = _mm256_setzero_ps();
285 fjy0 = _mm256_setzero_ps();
286 fjz0 = _mm256_setzero_ps();
288 /**************************
289 * CALCULATE INTERACTIONS *
290 **************************/
292 if (gmx_mm256_any_lt(rsq00,rcutoff2))
295 r00 = _mm256_mul_ps(rsq00,rinv00);
297 /* Compute parameters for interactions between i and j atoms */
298 qq00 = _mm256_mul_ps(iq0,jq0);
299 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
300 vdwioffsetptr0+vdwjidx0B,
301 vdwioffsetptr0+vdwjidx0C,
302 vdwioffsetptr0+vdwjidx0D,
303 vdwioffsetptr0+vdwjidx0E,
304 vdwioffsetptr0+vdwjidx0F,
305 vdwioffsetptr0+vdwjidx0G,
306 vdwioffsetptr0+vdwjidx0H,
309 /* EWALD ELECTROSTATICS */
311 /* Analytical PME correction */
312 zeta2 = _mm256_mul_ps(beta2,rsq00);
313 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
314 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
315 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
316 felec = _mm256_mul_ps(qq00,felec);
317 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
318 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
319 velec = _mm256_sub_ps(rinv00,pmecorrV);
320 velec = _mm256_mul_ps(qq00,velec);
322 /* LENNARD-JONES DISPERSION/REPULSION */
324 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
325 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
326 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
327 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
328 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
330 d = _mm256_sub_ps(r00,rswitch);
331 d = _mm256_max_ps(d,_mm256_setzero_ps());
332 d2 = _mm256_mul_ps(d,d);
333 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)))))));
335 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
337 /* Evaluate switch function */
338 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
339 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
340 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
341 velec = _mm256_mul_ps(velec,sw);
342 vvdw = _mm256_mul_ps(vvdw,sw);
343 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
345 /* Update potential sum for this i atom from the interaction with this j atom. */
346 velec = _mm256_and_ps(velec,cutoff_mask);
347 velecsum = _mm256_add_ps(velecsum,velec);
348 vvdw = _mm256_and_ps(vvdw,cutoff_mask);
349 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
351 fscal = _mm256_add_ps(felec,fvdw);
353 fscal = _mm256_and_ps(fscal,cutoff_mask);
355 /* Calculate temporary vectorial force */
356 tx = _mm256_mul_ps(fscal,dx00);
357 ty = _mm256_mul_ps(fscal,dy00);
358 tz = _mm256_mul_ps(fscal,dz00);
360 /* Update vectorial force */
361 fix0 = _mm256_add_ps(fix0,tx);
362 fiy0 = _mm256_add_ps(fiy0,ty);
363 fiz0 = _mm256_add_ps(fiz0,tz);
365 fjx0 = _mm256_add_ps(fjx0,tx);
366 fjy0 = _mm256_add_ps(fjy0,ty);
367 fjz0 = _mm256_add_ps(fjz0,tz);
371 /**************************
372 * CALCULATE INTERACTIONS *
373 **************************/
375 if (gmx_mm256_any_lt(rsq10,rcutoff2))
378 r10 = _mm256_mul_ps(rsq10,rinv10);
380 /* Compute parameters for interactions between i and j atoms */
381 qq10 = _mm256_mul_ps(iq1,jq0);
383 /* EWALD ELECTROSTATICS */
385 /* Analytical PME correction */
386 zeta2 = _mm256_mul_ps(beta2,rsq10);
387 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
388 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
389 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
390 felec = _mm256_mul_ps(qq10,felec);
391 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
392 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
393 velec = _mm256_sub_ps(rinv10,pmecorrV);
394 velec = _mm256_mul_ps(qq10,velec);
396 d = _mm256_sub_ps(r10,rswitch);
397 d = _mm256_max_ps(d,_mm256_setzero_ps());
398 d2 = _mm256_mul_ps(d,d);
399 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)))))));
401 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
403 /* Evaluate switch function */
404 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
405 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
406 velec = _mm256_mul_ps(velec,sw);
407 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
409 /* Update potential sum for this i atom from the interaction with this j atom. */
410 velec = _mm256_and_ps(velec,cutoff_mask);
411 velecsum = _mm256_add_ps(velecsum,velec);
415 fscal = _mm256_and_ps(fscal,cutoff_mask);
417 /* Calculate temporary vectorial force */
418 tx = _mm256_mul_ps(fscal,dx10);
419 ty = _mm256_mul_ps(fscal,dy10);
420 tz = _mm256_mul_ps(fscal,dz10);
422 /* Update vectorial force */
423 fix1 = _mm256_add_ps(fix1,tx);
424 fiy1 = _mm256_add_ps(fiy1,ty);
425 fiz1 = _mm256_add_ps(fiz1,tz);
427 fjx0 = _mm256_add_ps(fjx0,tx);
428 fjy0 = _mm256_add_ps(fjy0,ty);
429 fjz0 = _mm256_add_ps(fjz0,tz);
433 /**************************
434 * CALCULATE INTERACTIONS *
435 **************************/
437 if (gmx_mm256_any_lt(rsq20,rcutoff2))
440 r20 = _mm256_mul_ps(rsq20,rinv20);
442 /* Compute parameters for interactions between i and j atoms */
443 qq20 = _mm256_mul_ps(iq2,jq0);
445 /* EWALD ELECTROSTATICS */
447 /* Analytical PME correction */
448 zeta2 = _mm256_mul_ps(beta2,rsq20);
449 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
450 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
451 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
452 felec = _mm256_mul_ps(qq20,felec);
453 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
454 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
455 velec = _mm256_sub_ps(rinv20,pmecorrV);
456 velec = _mm256_mul_ps(qq20,velec);
458 d = _mm256_sub_ps(r20,rswitch);
459 d = _mm256_max_ps(d,_mm256_setzero_ps());
460 d2 = _mm256_mul_ps(d,d);
461 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)))))));
463 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
465 /* Evaluate switch function */
466 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
467 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
468 velec = _mm256_mul_ps(velec,sw);
469 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
471 /* Update potential sum for this i atom from the interaction with this j atom. */
472 velec = _mm256_and_ps(velec,cutoff_mask);
473 velecsum = _mm256_add_ps(velecsum,velec);
477 fscal = _mm256_and_ps(fscal,cutoff_mask);
479 /* Calculate temporary vectorial force */
480 tx = _mm256_mul_ps(fscal,dx20);
481 ty = _mm256_mul_ps(fscal,dy20);
482 tz = _mm256_mul_ps(fscal,dz20);
484 /* Update vectorial force */
485 fix2 = _mm256_add_ps(fix2,tx);
486 fiy2 = _mm256_add_ps(fiy2,ty);
487 fiz2 = _mm256_add_ps(fiz2,tz);
489 fjx0 = _mm256_add_ps(fjx0,tx);
490 fjy0 = _mm256_add_ps(fjy0,ty);
491 fjz0 = _mm256_add_ps(fjz0,tz);
495 fjptrA = f+j_coord_offsetA;
496 fjptrB = f+j_coord_offsetB;
497 fjptrC = f+j_coord_offsetC;
498 fjptrD = f+j_coord_offsetD;
499 fjptrE = f+j_coord_offsetE;
500 fjptrF = f+j_coord_offsetF;
501 fjptrG = f+j_coord_offsetG;
502 fjptrH = f+j_coord_offsetH;
504 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
506 /* Inner loop uses 345 flops */
512 /* Get j neighbor index, and coordinate index */
513 jnrlistA = jjnr[jidx];
514 jnrlistB = jjnr[jidx+1];
515 jnrlistC = jjnr[jidx+2];
516 jnrlistD = jjnr[jidx+3];
517 jnrlistE = jjnr[jidx+4];
518 jnrlistF = jjnr[jidx+5];
519 jnrlistG = jjnr[jidx+6];
520 jnrlistH = jjnr[jidx+7];
521 /* Sign of each element will be negative for non-real atoms.
522 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
523 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
525 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
526 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
528 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
529 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
530 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
531 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
532 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
533 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
534 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
535 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
536 j_coord_offsetA = DIM*jnrA;
537 j_coord_offsetB = DIM*jnrB;
538 j_coord_offsetC = DIM*jnrC;
539 j_coord_offsetD = DIM*jnrD;
540 j_coord_offsetE = DIM*jnrE;
541 j_coord_offsetF = DIM*jnrF;
542 j_coord_offsetG = DIM*jnrG;
543 j_coord_offsetH = DIM*jnrH;
545 /* load j atom coordinates */
546 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
547 x+j_coord_offsetC,x+j_coord_offsetD,
548 x+j_coord_offsetE,x+j_coord_offsetF,
549 x+j_coord_offsetG,x+j_coord_offsetH,
552 /* Calculate displacement vector */
553 dx00 = _mm256_sub_ps(ix0,jx0);
554 dy00 = _mm256_sub_ps(iy0,jy0);
555 dz00 = _mm256_sub_ps(iz0,jz0);
556 dx10 = _mm256_sub_ps(ix1,jx0);
557 dy10 = _mm256_sub_ps(iy1,jy0);
558 dz10 = _mm256_sub_ps(iz1,jz0);
559 dx20 = _mm256_sub_ps(ix2,jx0);
560 dy20 = _mm256_sub_ps(iy2,jy0);
561 dz20 = _mm256_sub_ps(iz2,jz0);
563 /* Calculate squared distance and things based on it */
564 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
565 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
566 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
568 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
569 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
570 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
572 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
573 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
574 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
576 /* Load parameters for j particles */
577 jq0 = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
578 charge+jnrC+0,charge+jnrD+0,
579 charge+jnrE+0,charge+jnrF+0,
580 charge+jnrG+0,charge+jnrH+0);
581 vdwjidx0A = 2*vdwtype[jnrA+0];
582 vdwjidx0B = 2*vdwtype[jnrB+0];
583 vdwjidx0C = 2*vdwtype[jnrC+0];
584 vdwjidx0D = 2*vdwtype[jnrD+0];
585 vdwjidx0E = 2*vdwtype[jnrE+0];
586 vdwjidx0F = 2*vdwtype[jnrF+0];
587 vdwjidx0G = 2*vdwtype[jnrG+0];
588 vdwjidx0H = 2*vdwtype[jnrH+0];
590 fjx0 = _mm256_setzero_ps();
591 fjy0 = _mm256_setzero_ps();
592 fjz0 = _mm256_setzero_ps();
594 /**************************
595 * CALCULATE INTERACTIONS *
596 **************************/
598 if (gmx_mm256_any_lt(rsq00,rcutoff2))
601 r00 = _mm256_mul_ps(rsq00,rinv00);
602 r00 = _mm256_andnot_ps(dummy_mask,r00);
604 /* Compute parameters for interactions between i and j atoms */
605 qq00 = _mm256_mul_ps(iq0,jq0);
606 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
607 vdwioffsetptr0+vdwjidx0B,
608 vdwioffsetptr0+vdwjidx0C,
609 vdwioffsetptr0+vdwjidx0D,
610 vdwioffsetptr0+vdwjidx0E,
611 vdwioffsetptr0+vdwjidx0F,
612 vdwioffsetptr0+vdwjidx0G,
613 vdwioffsetptr0+vdwjidx0H,
616 /* EWALD ELECTROSTATICS */
618 /* Analytical PME correction */
619 zeta2 = _mm256_mul_ps(beta2,rsq00);
620 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
621 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
622 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
623 felec = _mm256_mul_ps(qq00,felec);
624 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
625 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
626 velec = _mm256_sub_ps(rinv00,pmecorrV);
627 velec = _mm256_mul_ps(qq00,velec);
629 /* LENNARD-JONES DISPERSION/REPULSION */
631 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
632 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
633 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
634 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
635 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
637 d = _mm256_sub_ps(r00,rswitch);
638 d = _mm256_max_ps(d,_mm256_setzero_ps());
639 d2 = _mm256_mul_ps(d,d);
640 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)))))));
642 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
644 /* Evaluate switch function */
645 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
646 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
647 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
648 velec = _mm256_mul_ps(velec,sw);
649 vvdw = _mm256_mul_ps(vvdw,sw);
650 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
652 /* Update potential sum for this i atom from the interaction with this j atom. */
653 velec = _mm256_and_ps(velec,cutoff_mask);
654 velec = _mm256_andnot_ps(dummy_mask,velec);
655 velecsum = _mm256_add_ps(velecsum,velec);
656 vvdw = _mm256_and_ps(vvdw,cutoff_mask);
657 vvdw = _mm256_andnot_ps(dummy_mask,vvdw);
658 vvdwsum = _mm256_add_ps(vvdwsum,vvdw);
660 fscal = _mm256_add_ps(felec,fvdw);
662 fscal = _mm256_and_ps(fscal,cutoff_mask);
664 fscal = _mm256_andnot_ps(dummy_mask,fscal);
666 /* Calculate temporary vectorial force */
667 tx = _mm256_mul_ps(fscal,dx00);
668 ty = _mm256_mul_ps(fscal,dy00);
669 tz = _mm256_mul_ps(fscal,dz00);
671 /* Update vectorial force */
672 fix0 = _mm256_add_ps(fix0,tx);
673 fiy0 = _mm256_add_ps(fiy0,ty);
674 fiz0 = _mm256_add_ps(fiz0,tz);
676 fjx0 = _mm256_add_ps(fjx0,tx);
677 fjy0 = _mm256_add_ps(fjy0,ty);
678 fjz0 = _mm256_add_ps(fjz0,tz);
682 /**************************
683 * CALCULATE INTERACTIONS *
684 **************************/
686 if (gmx_mm256_any_lt(rsq10,rcutoff2))
689 r10 = _mm256_mul_ps(rsq10,rinv10);
690 r10 = _mm256_andnot_ps(dummy_mask,r10);
692 /* Compute parameters for interactions between i and j atoms */
693 qq10 = _mm256_mul_ps(iq1,jq0);
695 /* EWALD ELECTROSTATICS */
697 /* Analytical PME correction */
698 zeta2 = _mm256_mul_ps(beta2,rsq10);
699 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
700 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
701 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
702 felec = _mm256_mul_ps(qq10,felec);
703 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
704 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
705 velec = _mm256_sub_ps(rinv10,pmecorrV);
706 velec = _mm256_mul_ps(qq10,velec);
708 d = _mm256_sub_ps(r10,rswitch);
709 d = _mm256_max_ps(d,_mm256_setzero_ps());
710 d2 = _mm256_mul_ps(d,d);
711 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)))))));
713 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
715 /* Evaluate switch function */
716 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
717 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
718 velec = _mm256_mul_ps(velec,sw);
719 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
721 /* Update potential sum for this i atom from the interaction with this j atom. */
722 velec = _mm256_and_ps(velec,cutoff_mask);
723 velec = _mm256_andnot_ps(dummy_mask,velec);
724 velecsum = _mm256_add_ps(velecsum,velec);
728 fscal = _mm256_and_ps(fscal,cutoff_mask);
730 fscal = _mm256_andnot_ps(dummy_mask,fscal);
732 /* Calculate temporary vectorial force */
733 tx = _mm256_mul_ps(fscal,dx10);
734 ty = _mm256_mul_ps(fscal,dy10);
735 tz = _mm256_mul_ps(fscal,dz10);
737 /* Update vectorial force */
738 fix1 = _mm256_add_ps(fix1,tx);
739 fiy1 = _mm256_add_ps(fiy1,ty);
740 fiz1 = _mm256_add_ps(fiz1,tz);
742 fjx0 = _mm256_add_ps(fjx0,tx);
743 fjy0 = _mm256_add_ps(fjy0,ty);
744 fjz0 = _mm256_add_ps(fjz0,tz);
748 /**************************
749 * CALCULATE INTERACTIONS *
750 **************************/
752 if (gmx_mm256_any_lt(rsq20,rcutoff2))
755 r20 = _mm256_mul_ps(rsq20,rinv20);
756 r20 = _mm256_andnot_ps(dummy_mask,r20);
758 /* Compute parameters for interactions between i and j atoms */
759 qq20 = _mm256_mul_ps(iq2,jq0);
761 /* EWALD ELECTROSTATICS */
763 /* Analytical PME correction */
764 zeta2 = _mm256_mul_ps(beta2,rsq20);
765 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
766 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
767 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
768 felec = _mm256_mul_ps(qq20,felec);
769 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
770 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
771 velec = _mm256_sub_ps(rinv20,pmecorrV);
772 velec = _mm256_mul_ps(qq20,velec);
774 d = _mm256_sub_ps(r20,rswitch);
775 d = _mm256_max_ps(d,_mm256_setzero_ps());
776 d2 = _mm256_mul_ps(d,d);
777 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)))))));
779 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
781 /* Evaluate switch function */
782 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
783 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
784 velec = _mm256_mul_ps(velec,sw);
785 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
787 /* Update potential sum for this i atom from the interaction with this j atom. */
788 velec = _mm256_and_ps(velec,cutoff_mask);
789 velec = _mm256_andnot_ps(dummy_mask,velec);
790 velecsum = _mm256_add_ps(velecsum,velec);
794 fscal = _mm256_and_ps(fscal,cutoff_mask);
796 fscal = _mm256_andnot_ps(dummy_mask,fscal);
798 /* Calculate temporary vectorial force */
799 tx = _mm256_mul_ps(fscal,dx20);
800 ty = _mm256_mul_ps(fscal,dy20);
801 tz = _mm256_mul_ps(fscal,dz20);
803 /* Update vectorial force */
804 fix2 = _mm256_add_ps(fix2,tx);
805 fiy2 = _mm256_add_ps(fiy2,ty);
806 fiz2 = _mm256_add_ps(fiz2,tz);
808 fjx0 = _mm256_add_ps(fjx0,tx);
809 fjy0 = _mm256_add_ps(fjy0,ty);
810 fjz0 = _mm256_add_ps(fjz0,tz);
814 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
815 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
816 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
817 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
818 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
819 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
820 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
821 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
823 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
825 /* Inner loop uses 348 flops */
828 /* End of innermost loop */
830 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
831 f+i_coord_offset,fshift+i_shift_offset);
834 /* Update potential energies */
835 gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
836 gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
838 /* Increment number of inner iterations */
839 inneriter += j_index_end - j_index_start;
841 /* Outer loop uses 20 flops */
844 /* Increment number of outer iterations */
847 /* Update outer/inner flops */
849 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*348);
852 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_avx_256_single
853 * Electrostatics interaction: Ewald
854 * VdW interaction: LennardJones
855 * Geometry: Water3-Particle
856 * Calculate force/pot: Force
859 nb_kernel_ElecEwSw_VdwLJSw_GeomW3P1_F_avx_256_single
860 (t_nblist * gmx_restrict nlist,
861 rvec * gmx_restrict xx,
862 rvec * gmx_restrict ff,
863 t_forcerec * gmx_restrict fr,
864 t_mdatoms * gmx_restrict mdatoms,
865 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
866 t_nrnb * gmx_restrict nrnb)
868 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
869 * just 0 for non-waters.
870 * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
871 * jnr indices corresponding to data put in the four positions in the SIMD register.
873 int i_shift_offset,i_coord_offset,outeriter,inneriter;
874 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
875 int jnrA,jnrB,jnrC,jnrD;
876 int jnrE,jnrF,jnrG,jnrH;
877 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
878 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
879 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
880 int j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
881 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
883 real *shiftvec,*fshift,*x,*f;
884 real *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
886 __m256 tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
887 real * vdwioffsetptr0;
888 __m256 ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
889 real * vdwioffsetptr1;
890 __m256 ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
891 real * vdwioffsetptr2;
892 __m256 ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
893 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
894 __m256 jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
895 __m256 dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
896 __m256 dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
897 __m256 dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
898 __m256 velec,felec,velecsum,facel,crf,krf,krf2;
901 __m256 rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
904 __m256 one_sixth = _mm256_set1_ps(1.0/6.0);
905 __m256 one_twelfth = _mm256_set1_ps(1.0/12.0);
907 __m128i ewitab_lo,ewitab_hi;
908 __m256 ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
909 __m256 beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
911 __m256 rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
912 real rswitch_scalar,d_scalar;
913 __m256 dummy_mask,cutoff_mask;
914 __m256 signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
915 __m256 one = _mm256_set1_ps(1.0);
916 __m256 two = _mm256_set1_ps(2.0);
922 jindex = nlist->jindex;
924 shiftidx = nlist->shift;
926 shiftvec = fr->shift_vec[0];
927 fshift = fr->fshift[0];
928 facel = _mm256_set1_ps(fr->epsfac);
929 charge = mdatoms->chargeA;
930 nvdwtype = fr->ntype;
932 vdwtype = mdatoms->typeA;
934 sh_ewald = _mm256_set1_ps(fr->ic->sh_ewald);
935 beta = _mm256_set1_ps(fr->ic->ewaldcoeff_q);
936 beta2 = _mm256_mul_ps(beta,beta);
937 beta3 = _mm256_mul_ps(beta,beta2);
939 ewtab = fr->ic->tabq_coul_FDV0;
940 ewtabscale = _mm256_set1_ps(fr->ic->tabq_scale);
941 ewtabhalfspace = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
943 /* Setup water-specific parameters */
944 inr = nlist->iinr[0];
945 iq0 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
946 iq1 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
947 iq2 = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
948 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
950 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
951 rcutoff_scalar = fr->rcoulomb;
952 rcutoff = _mm256_set1_ps(rcutoff_scalar);
953 rcutoff2 = _mm256_mul_ps(rcutoff,rcutoff);
955 rswitch_scalar = fr->rcoulomb_switch;
956 rswitch = _mm256_set1_ps(rswitch_scalar);
957 /* Setup switch parameters */
958 d_scalar = rcutoff_scalar-rswitch_scalar;
959 d = _mm256_set1_ps(d_scalar);
960 swV3 = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
961 swV4 = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
962 swV5 = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
963 swF2 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
964 swF3 = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
965 swF4 = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
967 /* Avoid stupid compiler warnings */
968 jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
981 for(iidx=0;iidx<4*DIM;iidx++)
986 /* Start outer loop over neighborlists */
987 for(iidx=0; iidx<nri; iidx++)
989 /* Load shift vector for this list */
990 i_shift_offset = DIM*shiftidx[iidx];
992 /* Load limits for loop over neighbors */
993 j_index_start = jindex[iidx];
994 j_index_end = jindex[iidx+1];
996 /* Get outer coordinate index */
998 i_coord_offset = DIM*inr;
1000 /* Load i particle coords and add shift vector */
1001 gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1002 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1004 fix0 = _mm256_setzero_ps();
1005 fiy0 = _mm256_setzero_ps();
1006 fiz0 = _mm256_setzero_ps();
1007 fix1 = _mm256_setzero_ps();
1008 fiy1 = _mm256_setzero_ps();
1009 fiz1 = _mm256_setzero_ps();
1010 fix2 = _mm256_setzero_ps();
1011 fiy2 = _mm256_setzero_ps();
1012 fiz2 = _mm256_setzero_ps();
1014 /* Start inner kernel loop */
1015 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1018 /* Get j neighbor index, and coordinate index */
1020 jnrB = jjnr[jidx+1];
1021 jnrC = jjnr[jidx+2];
1022 jnrD = jjnr[jidx+3];
1023 jnrE = jjnr[jidx+4];
1024 jnrF = jjnr[jidx+5];
1025 jnrG = jjnr[jidx+6];
1026 jnrH = jjnr[jidx+7];
1027 j_coord_offsetA = DIM*jnrA;
1028 j_coord_offsetB = DIM*jnrB;
1029 j_coord_offsetC = DIM*jnrC;
1030 j_coord_offsetD = DIM*jnrD;
1031 j_coord_offsetE = DIM*jnrE;
1032 j_coord_offsetF = DIM*jnrF;
1033 j_coord_offsetG = DIM*jnrG;
1034 j_coord_offsetH = DIM*jnrH;
1036 /* load j atom coordinates */
1037 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1038 x+j_coord_offsetC,x+j_coord_offsetD,
1039 x+j_coord_offsetE,x+j_coord_offsetF,
1040 x+j_coord_offsetG,x+j_coord_offsetH,
1043 /* Calculate displacement vector */
1044 dx00 = _mm256_sub_ps(ix0,jx0);
1045 dy00 = _mm256_sub_ps(iy0,jy0);
1046 dz00 = _mm256_sub_ps(iz0,jz0);
1047 dx10 = _mm256_sub_ps(ix1,jx0);
1048 dy10 = _mm256_sub_ps(iy1,jy0);
1049 dz10 = _mm256_sub_ps(iz1,jz0);
1050 dx20 = _mm256_sub_ps(ix2,jx0);
1051 dy20 = _mm256_sub_ps(iy2,jy0);
1052 dz20 = _mm256_sub_ps(iz2,jz0);
1054 /* Calculate squared distance and things based on it */
1055 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1056 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1057 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1059 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1060 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1061 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1063 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1064 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1065 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1067 /* Load parameters for j particles */
1068 jq0 = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1069 charge+jnrC+0,charge+jnrD+0,
1070 charge+jnrE+0,charge+jnrF+0,
1071 charge+jnrG+0,charge+jnrH+0);
1072 vdwjidx0A = 2*vdwtype[jnrA+0];
1073 vdwjidx0B = 2*vdwtype[jnrB+0];
1074 vdwjidx0C = 2*vdwtype[jnrC+0];
1075 vdwjidx0D = 2*vdwtype[jnrD+0];
1076 vdwjidx0E = 2*vdwtype[jnrE+0];
1077 vdwjidx0F = 2*vdwtype[jnrF+0];
1078 vdwjidx0G = 2*vdwtype[jnrG+0];
1079 vdwjidx0H = 2*vdwtype[jnrH+0];
1081 fjx0 = _mm256_setzero_ps();
1082 fjy0 = _mm256_setzero_ps();
1083 fjz0 = _mm256_setzero_ps();
1085 /**************************
1086 * CALCULATE INTERACTIONS *
1087 **************************/
1089 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1092 r00 = _mm256_mul_ps(rsq00,rinv00);
1094 /* Compute parameters for interactions between i and j atoms */
1095 qq00 = _mm256_mul_ps(iq0,jq0);
1096 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1097 vdwioffsetptr0+vdwjidx0B,
1098 vdwioffsetptr0+vdwjidx0C,
1099 vdwioffsetptr0+vdwjidx0D,
1100 vdwioffsetptr0+vdwjidx0E,
1101 vdwioffsetptr0+vdwjidx0F,
1102 vdwioffsetptr0+vdwjidx0G,
1103 vdwioffsetptr0+vdwjidx0H,
1106 /* EWALD ELECTROSTATICS */
1108 /* Analytical PME correction */
1109 zeta2 = _mm256_mul_ps(beta2,rsq00);
1110 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1111 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1112 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1113 felec = _mm256_mul_ps(qq00,felec);
1114 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1115 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1116 velec = _mm256_sub_ps(rinv00,pmecorrV);
1117 velec = _mm256_mul_ps(qq00,velec);
1119 /* LENNARD-JONES DISPERSION/REPULSION */
1121 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1122 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
1123 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
1124 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
1125 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
1127 d = _mm256_sub_ps(r00,rswitch);
1128 d = _mm256_max_ps(d,_mm256_setzero_ps());
1129 d2 = _mm256_mul_ps(d,d);
1130 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)))))));
1132 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1134 /* Evaluate switch function */
1135 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1136 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1137 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
1138 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1140 fscal = _mm256_add_ps(felec,fvdw);
1142 fscal = _mm256_and_ps(fscal,cutoff_mask);
1144 /* Calculate temporary vectorial force */
1145 tx = _mm256_mul_ps(fscal,dx00);
1146 ty = _mm256_mul_ps(fscal,dy00);
1147 tz = _mm256_mul_ps(fscal,dz00);
1149 /* Update vectorial force */
1150 fix0 = _mm256_add_ps(fix0,tx);
1151 fiy0 = _mm256_add_ps(fiy0,ty);
1152 fiz0 = _mm256_add_ps(fiz0,tz);
1154 fjx0 = _mm256_add_ps(fjx0,tx);
1155 fjy0 = _mm256_add_ps(fjy0,ty);
1156 fjz0 = _mm256_add_ps(fjz0,tz);
1160 /**************************
1161 * CALCULATE INTERACTIONS *
1162 **************************/
1164 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1167 r10 = _mm256_mul_ps(rsq10,rinv10);
1169 /* Compute parameters for interactions between i and j atoms */
1170 qq10 = _mm256_mul_ps(iq1,jq0);
1172 /* EWALD ELECTROSTATICS */
1174 /* Analytical PME correction */
1175 zeta2 = _mm256_mul_ps(beta2,rsq10);
1176 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
1177 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1178 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1179 felec = _mm256_mul_ps(qq10,felec);
1180 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1181 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1182 velec = _mm256_sub_ps(rinv10,pmecorrV);
1183 velec = _mm256_mul_ps(qq10,velec);
1185 d = _mm256_sub_ps(r10,rswitch);
1186 d = _mm256_max_ps(d,_mm256_setzero_ps());
1187 d2 = _mm256_mul_ps(d,d);
1188 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)))))));
1190 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1192 /* Evaluate switch function */
1193 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1194 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
1195 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1199 fscal = _mm256_and_ps(fscal,cutoff_mask);
1201 /* Calculate temporary vectorial force */
1202 tx = _mm256_mul_ps(fscal,dx10);
1203 ty = _mm256_mul_ps(fscal,dy10);
1204 tz = _mm256_mul_ps(fscal,dz10);
1206 /* Update vectorial force */
1207 fix1 = _mm256_add_ps(fix1,tx);
1208 fiy1 = _mm256_add_ps(fiy1,ty);
1209 fiz1 = _mm256_add_ps(fiz1,tz);
1211 fjx0 = _mm256_add_ps(fjx0,tx);
1212 fjy0 = _mm256_add_ps(fjy0,ty);
1213 fjz0 = _mm256_add_ps(fjz0,tz);
1217 /**************************
1218 * CALCULATE INTERACTIONS *
1219 **************************/
1221 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1224 r20 = _mm256_mul_ps(rsq20,rinv20);
1226 /* Compute parameters for interactions between i and j atoms */
1227 qq20 = _mm256_mul_ps(iq2,jq0);
1229 /* EWALD ELECTROSTATICS */
1231 /* Analytical PME correction */
1232 zeta2 = _mm256_mul_ps(beta2,rsq20);
1233 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1234 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1235 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1236 felec = _mm256_mul_ps(qq20,felec);
1237 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1238 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1239 velec = _mm256_sub_ps(rinv20,pmecorrV);
1240 velec = _mm256_mul_ps(qq20,velec);
1242 d = _mm256_sub_ps(r20,rswitch);
1243 d = _mm256_max_ps(d,_mm256_setzero_ps());
1244 d2 = _mm256_mul_ps(d,d);
1245 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)))))));
1247 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1249 /* Evaluate switch function */
1250 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1251 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
1252 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1256 fscal = _mm256_and_ps(fscal,cutoff_mask);
1258 /* Calculate temporary vectorial force */
1259 tx = _mm256_mul_ps(fscal,dx20);
1260 ty = _mm256_mul_ps(fscal,dy20);
1261 tz = _mm256_mul_ps(fscal,dz20);
1263 /* Update vectorial force */
1264 fix2 = _mm256_add_ps(fix2,tx);
1265 fiy2 = _mm256_add_ps(fiy2,ty);
1266 fiz2 = _mm256_add_ps(fiz2,tz);
1268 fjx0 = _mm256_add_ps(fjx0,tx);
1269 fjy0 = _mm256_add_ps(fjy0,ty);
1270 fjz0 = _mm256_add_ps(fjz0,tz);
1274 fjptrA = f+j_coord_offsetA;
1275 fjptrB = f+j_coord_offsetB;
1276 fjptrC = f+j_coord_offsetC;
1277 fjptrD = f+j_coord_offsetD;
1278 fjptrE = f+j_coord_offsetE;
1279 fjptrF = f+j_coord_offsetF;
1280 fjptrG = f+j_coord_offsetG;
1281 fjptrH = f+j_coord_offsetH;
1283 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1285 /* Inner loop uses 333 flops */
1288 if(jidx<j_index_end)
1291 /* Get j neighbor index, and coordinate index */
1292 jnrlistA = jjnr[jidx];
1293 jnrlistB = jjnr[jidx+1];
1294 jnrlistC = jjnr[jidx+2];
1295 jnrlistD = jjnr[jidx+3];
1296 jnrlistE = jjnr[jidx+4];
1297 jnrlistF = jjnr[jidx+5];
1298 jnrlistG = jjnr[jidx+6];
1299 jnrlistH = jjnr[jidx+7];
1300 /* Sign of each element will be negative for non-real atoms.
1301 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1302 * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1304 dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1305 gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1307 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1308 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1309 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1310 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1311 jnrE = (jnrlistE>=0) ? jnrlistE : 0;
1312 jnrF = (jnrlistF>=0) ? jnrlistF : 0;
1313 jnrG = (jnrlistG>=0) ? jnrlistG : 0;
1314 jnrH = (jnrlistH>=0) ? jnrlistH : 0;
1315 j_coord_offsetA = DIM*jnrA;
1316 j_coord_offsetB = DIM*jnrB;
1317 j_coord_offsetC = DIM*jnrC;
1318 j_coord_offsetD = DIM*jnrD;
1319 j_coord_offsetE = DIM*jnrE;
1320 j_coord_offsetF = DIM*jnrF;
1321 j_coord_offsetG = DIM*jnrG;
1322 j_coord_offsetH = DIM*jnrH;
1324 /* load j atom coordinates */
1325 gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1326 x+j_coord_offsetC,x+j_coord_offsetD,
1327 x+j_coord_offsetE,x+j_coord_offsetF,
1328 x+j_coord_offsetG,x+j_coord_offsetH,
1331 /* Calculate displacement vector */
1332 dx00 = _mm256_sub_ps(ix0,jx0);
1333 dy00 = _mm256_sub_ps(iy0,jy0);
1334 dz00 = _mm256_sub_ps(iz0,jz0);
1335 dx10 = _mm256_sub_ps(ix1,jx0);
1336 dy10 = _mm256_sub_ps(iy1,jy0);
1337 dz10 = _mm256_sub_ps(iz1,jz0);
1338 dx20 = _mm256_sub_ps(ix2,jx0);
1339 dy20 = _mm256_sub_ps(iy2,jy0);
1340 dz20 = _mm256_sub_ps(iz2,jz0);
1342 /* Calculate squared distance and things based on it */
1343 rsq00 = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1344 rsq10 = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1345 rsq20 = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1347 rinv00 = gmx_mm256_invsqrt_ps(rsq00);
1348 rinv10 = gmx_mm256_invsqrt_ps(rsq10);
1349 rinv20 = gmx_mm256_invsqrt_ps(rsq20);
1351 rinvsq00 = _mm256_mul_ps(rinv00,rinv00);
1352 rinvsq10 = _mm256_mul_ps(rinv10,rinv10);
1353 rinvsq20 = _mm256_mul_ps(rinv20,rinv20);
1355 /* Load parameters for j particles */
1356 jq0 = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1357 charge+jnrC+0,charge+jnrD+0,
1358 charge+jnrE+0,charge+jnrF+0,
1359 charge+jnrG+0,charge+jnrH+0);
1360 vdwjidx0A = 2*vdwtype[jnrA+0];
1361 vdwjidx0B = 2*vdwtype[jnrB+0];
1362 vdwjidx0C = 2*vdwtype[jnrC+0];
1363 vdwjidx0D = 2*vdwtype[jnrD+0];
1364 vdwjidx0E = 2*vdwtype[jnrE+0];
1365 vdwjidx0F = 2*vdwtype[jnrF+0];
1366 vdwjidx0G = 2*vdwtype[jnrG+0];
1367 vdwjidx0H = 2*vdwtype[jnrH+0];
1369 fjx0 = _mm256_setzero_ps();
1370 fjy0 = _mm256_setzero_ps();
1371 fjz0 = _mm256_setzero_ps();
1373 /**************************
1374 * CALCULATE INTERACTIONS *
1375 **************************/
1377 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1380 r00 = _mm256_mul_ps(rsq00,rinv00);
1381 r00 = _mm256_andnot_ps(dummy_mask,r00);
1383 /* Compute parameters for interactions between i and j atoms */
1384 qq00 = _mm256_mul_ps(iq0,jq0);
1385 gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1386 vdwioffsetptr0+vdwjidx0B,
1387 vdwioffsetptr0+vdwjidx0C,
1388 vdwioffsetptr0+vdwjidx0D,
1389 vdwioffsetptr0+vdwjidx0E,
1390 vdwioffsetptr0+vdwjidx0F,
1391 vdwioffsetptr0+vdwjidx0G,
1392 vdwioffsetptr0+vdwjidx0H,
1395 /* EWALD ELECTROSTATICS */
1397 /* Analytical PME correction */
1398 zeta2 = _mm256_mul_ps(beta2,rsq00);
1399 rinv3 = _mm256_mul_ps(rinvsq00,rinv00);
1400 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1401 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1402 felec = _mm256_mul_ps(qq00,felec);
1403 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1404 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1405 velec = _mm256_sub_ps(rinv00,pmecorrV);
1406 velec = _mm256_mul_ps(qq00,velec);
1408 /* LENNARD-JONES DISPERSION/REPULSION */
1410 rinvsix = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1411 vvdw6 = _mm256_mul_ps(c6_00,rinvsix);
1412 vvdw12 = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
1413 vvdw = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
1414 fvdw = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
1416 d = _mm256_sub_ps(r00,rswitch);
1417 d = _mm256_max_ps(d,_mm256_setzero_ps());
1418 d2 = _mm256_mul_ps(d,d);
1419 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)))))));
1421 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1423 /* Evaluate switch function */
1424 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1425 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(velec,dsw)) );
1426 fvdw = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
1427 cutoff_mask = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1429 fscal = _mm256_add_ps(felec,fvdw);
1431 fscal = _mm256_and_ps(fscal,cutoff_mask);
1433 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1435 /* Calculate temporary vectorial force */
1436 tx = _mm256_mul_ps(fscal,dx00);
1437 ty = _mm256_mul_ps(fscal,dy00);
1438 tz = _mm256_mul_ps(fscal,dz00);
1440 /* Update vectorial force */
1441 fix0 = _mm256_add_ps(fix0,tx);
1442 fiy0 = _mm256_add_ps(fiy0,ty);
1443 fiz0 = _mm256_add_ps(fiz0,tz);
1445 fjx0 = _mm256_add_ps(fjx0,tx);
1446 fjy0 = _mm256_add_ps(fjy0,ty);
1447 fjz0 = _mm256_add_ps(fjz0,tz);
1451 /**************************
1452 * CALCULATE INTERACTIONS *
1453 **************************/
1455 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1458 r10 = _mm256_mul_ps(rsq10,rinv10);
1459 r10 = _mm256_andnot_ps(dummy_mask,r10);
1461 /* Compute parameters for interactions between i and j atoms */
1462 qq10 = _mm256_mul_ps(iq1,jq0);
1464 /* EWALD ELECTROSTATICS */
1466 /* Analytical PME correction */
1467 zeta2 = _mm256_mul_ps(beta2,rsq10);
1468 rinv3 = _mm256_mul_ps(rinvsq10,rinv10);
1469 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1470 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1471 felec = _mm256_mul_ps(qq10,felec);
1472 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1473 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1474 velec = _mm256_sub_ps(rinv10,pmecorrV);
1475 velec = _mm256_mul_ps(qq10,velec);
1477 d = _mm256_sub_ps(r10,rswitch);
1478 d = _mm256_max_ps(d,_mm256_setzero_ps());
1479 d2 = _mm256_mul_ps(d,d);
1480 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)))))));
1482 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1484 /* Evaluate switch function */
1485 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1486 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv10,_mm256_mul_ps(velec,dsw)) );
1487 cutoff_mask = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1491 fscal = _mm256_and_ps(fscal,cutoff_mask);
1493 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1495 /* Calculate temporary vectorial force */
1496 tx = _mm256_mul_ps(fscal,dx10);
1497 ty = _mm256_mul_ps(fscal,dy10);
1498 tz = _mm256_mul_ps(fscal,dz10);
1500 /* Update vectorial force */
1501 fix1 = _mm256_add_ps(fix1,tx);
1502 fiy1 = _mm256_add_ps(fiy1,ty);
1503 fiz1 = _mm256_add_ps(fiz1,tz);
1505 fjx0 = _mm256_add_ps(fjx0,tx);
1506 fjy0 = _mm256_add_ps(fjy0,ty);
1507 fjz0 = _mm256_add_ps(fjz0,tz);
1511 /**************************
1512 * CALCULATE INTERACTIONS *
1513 **************************/
1515 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1518 r20 = _mm256_mul_ps(rsq20,rinv20);
1519 r20 = _mm256_andnot_ps(dummy_mask,r20);
1521 /* Compute parameters for interactions between i and j atoms */
1522 qq20 = _mm256_mul_ps(iq2,jq0);
1524 /* EWALD ELECTROSTATICS */
1526 /* Analytical PME correction */
1527 zeta2 = _mm256_mul_ps(beta2,rsq20);
1528 rinv3 = _mm256_mul_ps(rinvsq20,rinv20);
1529 pmecorrF = gmx_mm256_pmecorrF_ps(zeta2);
1530 felec = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1531 felec = _mm256_mul_ps(qq20,felec);
1532 pmecorrV = gmx_mm256_pmecorrV_ps(zeta2);
1533 pmecorrV = _mm256_mul_ps(pmecorrV,beta);
1534 velec = _mm256_sub_ps(rinv20,pmecorrV);
1535 velec = _mm256_mul_ps(qq20,velec);
1537 d = _mm256_sub_ps(r20,rswitch);
1538 d = _mm256_max_ps(d,_mm256_setzero_ps());
1539 d2 = _mm256_mul_ps(d,d);
1540 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)))))));
1542 dsw = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1544 /* Evaluate switch function */
1545 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1546 felec = _mm256_sub_ps( _mm256_mul_ps(felec,sw) , _mm256_mul_ps(rinv20,_mm256_mul_ps(velec,dsw)) );
1547 cutoff_mask = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1551 fscal = _mm256_and_ps(fscal,cutoff_mask);
1553 fscal = _mm256_andnot_ps(dummy_mask,fscal);
1555 /* Calculate temporary vectorial force */
1556 tx = _mm256_mul_ps(fscal,dx20);
1557 ty = _mm256_mul_ps(fscal,dy20);
1558 tz = _mm256_mul_ps(fscal,dz20);
1560 /* Update vectorial force */
1561 fix2 = _mm256_add_ps(fix2,tx);
1562 fiy2 = _mm256_add_ps(fiy2,ty);
1563 fiz2 = _mm256_add_ps(fiz2,tz);
1565 fjx0 = _mm256_add_ps(fjx0,tx);
1566 fjy0 = _mm256_add_ps(fjy0,ty);
1567 fjz0 = _mm256_add_ps(fjz0,tz);
1571 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1572 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1573 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1574 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1575 fjptrE = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1576 fjptrF = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1577 fjptrG = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1578 fjptrH = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1580 gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1582 /* Inner loop uses 336 flops */
1585 /* End of innermost loop */
1587 gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1588 f+i_coord_offset,fshift+i_shift_offset);
1590 /* Increment number of inner iterations */
1591 inneriter += j_index_end - j_index_start;
1593 /* Outer loop uses 18 flops */
1596 /* Increment number of outer iterations */
1599 /* Update outer/inner flops */
1601 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*336);