2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013, 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_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "types/simple.h"
49 #include "gmx_math_x86_avx_256_double.h"
50 #include "kernelutil_x86_avx_256_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_256_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LennardJones
56 * Geometry: Water4-Particle
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_256_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
76 int jnrA,jnrB,jnrC,jnrD;
77 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
78 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
79 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
80 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
82 real *shiftvec,*fshift,*x,*f;
83 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
85 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
86 real * vdwioffsetptr0;
87 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
88 real * vdwioffsetptr1;
89 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
90 real * vdwioffsetptr2;
91 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
92 real * vdwioffsetptr3;
93 __m256d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
94 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
95 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
96 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
97 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
98 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
99 __m256d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
100 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
103 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
106 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
107 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
109 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
110 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
112 __m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
113 real rswitch_scalar,d_scalar;
114 __m256d dummy_mask,cutoff_mask;
115 __m128 tmpmask0,tmpmask1;
116 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
117 __m256d one = _mm256_set1_pd(1.0);
118 __m256d two = _mm256_set1_pd(2.0);
124 jindex = nlist->jindex;
126 shiftidx = nlist->shift;
128 shiftvec = fr->shift_vec[0];
129 fshift = fr->fshift[0];
130 facel = _mm256_set1_pd(fr->epsfac);
131 charge = mdatoms->chargeA;
132 nvdwtype = fr->ntype;
134 vdwtype = mdatoms->typeA;
136 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
137 beta = _mm256_set1_pd(fr->ic->ewaldcoeff);
138 beta2 = _mm256_mul_pd(beta,beta);
139 beta3 = _mm256_mul_pd(beta,beta2);
141 ewtab = fr->ic->tabq_coul_FDV0;
142 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
143 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
145 /* Setup water-specific parameters */
146 inr = nlist->iinr[0];
147 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
148 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
149 iq3 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
150 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
152 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
153 rcutoff_scalar = fr->rcoulomb;
154 rcutoff = _mm256_set1_pd(rcutoff_scalar);
155 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
157 rswitch_scalar = fr->rcoulomb_switch;
158 rswitch = _mm256_set1_pd(rswitch_scalar);
159 /* Setup switch parameters */
160 d_scalar = rcutoff_scalar-rswitch_scalar;
161 d = _mm256_set1_pd(d_scalar);
162 swV3 = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
163 swV4 = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
164 swV5 = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
165 swF2 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
166 swF3 = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
167 swF4 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
169 /* Avoid stupid compiler warnings */
170 jnrA = jnrB = jnrC = jnrD = 0;
179 for(iidx=0;iidx<4*DIM;iidx++)
184 /* Start outer loop over neighborlists */
185 for(iidx=0; iidx<nri; iidx++)
187 /* Load shift vector for this list */
188 i_shift_offset = DIM*shiftidx[iidx];
190 /* Load limits for loop over neighbors */
191 j_index_start = jindex[iidx];
192 j_index_end = jindex[iidx+1];
194 /* Get outer coordinate index */
196 i_coord_offset = DIM*inr;
198 /* Load i particle coords and add shift vector */
199 gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
200 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
202 fix0 = _mm256_setzero_pd();
203 fiy0 = _mm256_setzero_pd();
204 fiz0 = _mm256_setzero_pd();
205 fix1 = _mm256_setzero_pd();
206 fiy1 = _mm256_setzero_pd();
207 fiz1 = _mm256_setzero_pd();
208 fix2 = _mm256_setzero_pd();
209 fiy2 = _mm256_setzero_pd();
210 fiz2 = _mm256_setzero_pd();
211 fix3 = _mm256_setzero_pd();
212 fiy3 = _mm256_setzero_pd();
213 fiz3 = _mm256_setzero_pd();
215 /* Reset potential sums */
216 velecsum = _mm256_setzero_pd();
217 vvdwsum = _mm256_setzero_pd();
219 /* Start inner kernel loop */
220 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
223 /* Get j neighbor index, and coordinate index */
228 j_coord_offsetA = DIM*jnrA;
229 j_coord_offsetB = DIM*jnrB;
230 j_coord_offsetC = DIM*jnrC;
231 j_coord_offsetD = DIM*jnrD;
233 /* load j atom coordinates */
234 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
235 x+j_coord_offsetC,x+j_coord_offsetD,
238 /* Calculate displacement vector */
239 dx00 = _mm256_sub_pd(ix0,jx0);
240 dy00 = _mm256_sub_pd(iy0,jy0);
241 dz00 = _mm256_sub_pd(iz0,jz0);
242 dx10 = _mm256_sub_pd(ix1,jx0);
243 dy10 = _mm256_sub_pd(iy1,jy0);
244 dz10 = _mm256_sub_pd(iz1,jz0);
245 dx20 = _mm256_sub_pd(ix2,jx0);
246 dy20 = _mm256_sub_pd(iy2,jy0);
247 dz20 = _mm256_sub_pd(iz2,jz0);
248 dx30 = _mm256_sub_pd(ix3,jx0);
249 dy30 = _mm256_sub_pd(iy3,jy0);
250 dz30 = _mm256_sub_pd(iz3,jz0);
252 /* Calculate squared distance and things based on it */
253 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
254 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
255 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
256 rsq30 = gmx_mm256_calc_rsq_pd(dx30,dy30,dz30);
258 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
259 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
260 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
261 rinv30 = gmx_mm256_invsqrt_pd(rsq30);
263 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
264 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
265 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
266 rinvsq30 = _mm256_mul_pd(rinv30,rinv30);
268 /* Load parameters for j particles */
269 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
270 charge+jnrC+0,charge+jnrD+0);
271 vdwjidx0A = 2*vdwtype[jnrA+0];
272 vdwjidx0B = 2*vdwtype[jnrB+0];
273 vdwjidx0C = 2*vdwtype[jnrC+0];
274 vdwjidx0D = 2*vdwtype[jnrD+0];
276 fjx0 = _mm256_setzero_pd();
277 fjy0 = _mm256_setzero_pd();
278 fjz0 = _mm256_setzero_pd();
280 /**************************
281 * CALCULATE INTERACTIONS *
282 **************************/
284 if (gmx_mm256_any_lt(rsq00,rcutoff2))
287 r00 = _mm256_mul_pd(rsq00,rinv00);
289 /* Compute parameters for interactions between i and j atoms */
290 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
291 vdwioffsetptr0+vdwjidx0B,
292 vdwioffsetptr0+vdwjidx0C,
293 vdwioffsetptr0+vdwjidx0D,
296 /* LENNARD-JONES DISPERSION/REPULSION */
298 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
299 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
300 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
301 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
302 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
304 d = _mm256_sub_pd(r00,rswitch);
305 d = _mm256_max_pd(d,_mm256_setzero_pd());
306 d2 = _mm256_mul_pd(d,d);
307 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
309 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
311 /* Evaluate switch function */
312 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
313 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
314 vvdw = _mm256_mul_pd(vvdw,sw);
315 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
317 /* Update potential sum for this i atom from the interaction with this j atom. */
318 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
319 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
323 fscal = _mm256_and_pd(fscal,cutoff_mask);
325 /* Calculate temporary vectorial force */
326 tx = _mm256_mul_pd(fscal,dx00);
327 ty = _mm256_mul_pd(fscal,dy00);
328 tz = _mm256_mul_pd(fscal,dz00);
330 /* Update vectorial force */
331 fix0 = _mm256_add_pd(fix0,tx);
332 fiy0 = _mm256_add_pd(fiy0,ty);
333 fiz0 = _mm256_add_pd(fiz0,tz);
335 fjx0 = _mm256_add_pd(fjx0,tx);
336 fjy0 = _mm256_add_pd(fjy0,ty);
337 fjz0 = _mm256_add_pd(fjz0,tz);
341 /**************************
342 * CALCULATE INTERACTIONS *
343 **************************/
345 if (gmx_mm256_any_lt(rsq10,rcutoff2))
348 r10 = _mm256_mul_pd(rsq10,rinv10);
350 /* Compute parameters for interactions between i and j atoms */
351 qq10 = _mm256_mul_pd(iq1,jq0);
353 /* EWALD ELECTROSTATICS */
355 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
356 ewrt = _mm256_mul_pd(r10,ewtabscale);
357 ewitab = _mm256_cvttpd_epi32(ewrt);
358 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
359 ewitab = _mm_slli_epi32(ewitab,2);
360 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
361 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
362 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
363 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
364 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
365 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
366 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
367 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
368 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
370 d = _mm256_sub_pd(r10,rswitch);
371 d = _mm256_max_pd(d,_mm256_setzero_pd());
372 d2 = _mm256_mul_pd(d,d);
373 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
375 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
377 /* Evaluate switch function */
378 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
379 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
380 velec = _mm256_mul_pd(velec,sw);
381 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
383 /* Update potential sum for this i atom from the interaction with this j atom. */
384 velec = _mm256_and_pd(velec,cutoff_mask);
385 velecsum = _mm256_add_pd(velecsum,velec);
389 fscal = _mm256_and_pd(fscal,cutoff_mask);
391 /* Calculate temporary vectorial force */
392 tx = _mm256_mul_pd(fscal,dx10);
393 ty = _mm256_mul_pd(fscal,dy10);
394 tz = _mm256_mul_pd(fscal,dz10);
396 /* Update vectorial force */
397 fix1 = _mm256_add_pd(fix1,tx);
398 fiy1 = _mm256_add_pd(fiy1,ty);
399 fiz1 = _mm256_add_pd(fiz1,tz);
401 fjx0 = _mm256_add_pd(fjx0,tx);
402 fjy0 = _mm256_add_pd(fjy0,ty);
403 fjz0 = _mm256_add_pd(fjz0,tz);
407 /**************************
408 * CALCULATE INTERACTIONS *
409 **************************/
411 if (gmx_mm256_any_lt(rsq20,rcutoff2))
414 r20 = _mm256_mul_pd(rsq20,rinv20);
416 /* Compute parameters for interactions between i and j atoms */
417 qq20 = _mm256_mul_pd(iq2,jq0);
419 /* EWALD ELECTROSTATICS */
421 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
422 ewrt = _mm256_mul_pd(r20,ewtabscale);
423 ewitab = _mm256_cvttpd_epi32(ewrt);
424 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
425 ewitab = _mm_slli_epi32(ewitab,2);
426 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
427 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
428 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
429 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
430 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
431 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
432 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
433 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
434 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
436 d = _mm256_sub_pd(r20,rswitch);
437 d = _mm256_max_pd(d,_mm256_setzero_pd());
438 d2 = _mm256_mul_pd(d,d);
439 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
441 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
443 /* Evaluate switch function */
444 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
445 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
446 velec = _mm256_mul_pd(velec,sw);
447 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
449 /* Update potential sum for this i atom from the interaction with this j atom. */
450 velec = _mm256_and_pd(velec,cutoff_mask);
451 velecsum = _mm256_add_pd(velecsum,velec);
455 fscal = _mm256_and_pd(fscal,cutoff_mask);
457 /* Calculate temporary vectorial force */
458 tx = _mm256_mul_pd(fscal,dx20);
459 ty = _mm256_mul_pd(fscal,dy20);
460 tz = _mm256_mul_pd(fscal,dz20);
462 /* Update vectorial force */
463 fix2 = _mm256_add_pd(fix2,tx);
464 fiy2 = _mm256_add_pd(fiy2,ty);
465 fiz2 = _mm256_add_pd(fiz2,tz);
467 fjx0 = _mm256_add_pd(fjx0,tx);
468 fjy0 = _mm256_add_pd(fjy0,ty);
469 fjz0 = _mm256_add_pd(fjz0,tz);
473 /**************************
474 * CALCULATE INTERACTIONS *
475 **************************/
477 if (gmx_mm256_any_lt(rsq30,rcutoff2))
480 r30 = _mm256_mul_pd(rsq30,rinv30);
482 /* Compute parameters for interactions between i and j atoms */
483 qq30 = _mm256_mul_pd(iq3,jq0);
485 /* EWALD ELECTROSTATICS */
487 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
488 ewrt = _mm256_mul_pd(r30,ewtabscale);
489 ewitab = _mm256_cvttpd_epi32(ewrt);
490 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
491 ewitab = _mm_slli_epi32(ewitab,2);
492 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
493 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
494 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
495 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
496 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
497 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
498 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
499 velec = _mm256_mul_pd(qq30,_mm256_sub_pd(rinv30,velec));
500 felec = _mm256_mul_pd(_mm256_mul_pd(qq30,rinv30),_mm256_sub_pd(rinvsq30,felec));
502 d = _mm256_sub_pd(r30,rswitch);
503 d = _mm256_max_pd(d,_mm256_setzero_pd());
504 d2 = _mm256_mul_pd(d,d);
505 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
507 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
509 /* Evaluate switch function */
510 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
511 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv30,_mm256_mul_pd(velec,dsw)) );
512 velec = _mm256_mul_pd(velec,sw);
513 cutoff_mask = _mm256_cmp_pd(rsq30,rcutoff2,_CMP_LT_OQ);
515 /* Update potential sum for this i atom from the interaction with this j atom. */
516 velec = _mm256_and_pd(velec,cutoff_mask);
517 velecsum = _mm256_add_pd(velecsum,velec);
521 fscal = _mm256_and_pd(fscal,cutoff_mask);
523 /* Calculate temporary vectorial force */
524 tx = _mm256_mul_pd(fscal,dx30);
525 ty = _mm256_mul_pd(fscal,dy30);
526 tz = _mm256_mul_pd(fscal,dz30);
528 /* Update vectorial force */
529 fix3 = _mm256_add_pd(fix3,tx);
530 fiy3 = _mm256_add_pd(fiy3,ty);
531 fiz3 = _mm256_add_pd(fiz3,tz);
533 fjx0 = _mm256_add_pd(fjx0,tx);
534 fjy0 = _mm256_add_pd(fjy0,ty);
535 fjz0 = _mm256_add_pd(fjz0,tz);
539 fjptrA = f+j_coord_offsetA;
540 fjptrB = f+j_coord_offsetB;
541 fjptrC = f+j_coord_offsetC;
542 fjptrD = f+j_coord_offsetD;
544 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
546 /* Inner loop uses 257 flops */
552 /* Get j neighbor index, and coordinate index */
553 jnrlistA = jjnr[jidx];
554 jnrlistB = jjnr[jidx+1];
555 jnrlistC = jjnr[jidx+2];
556 jnrlistD = jjnr[jidx+3];
557 /* Sign of each element will be negative for non-real atoms.
558 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
559 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
561 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
563 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
564 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
565 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
567 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
568 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
569 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
570 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
571 j_coord_offsetA = DIM*jnrA;
572 j_coord_offsetB = DIM*jnrB;
573 j_coord_offsetC = DIM*jnrC;
574 j_coord_offsetD = DIM*jnrD;
576 /* load j atom coordinates */
577 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
578 x+j_coord_offsetC,x+j_coord_offsetD,
581 /* Calculate displacement vector */
582 dx00 = _mm256_sub_pd(ix0,jx0);
583 dy00 = _mm256_sub_pd(iy0,jy0);
584 dz00 = _mm256_sub_pd(iz0,jz0);
585 dx10 = _mm256_sub_pd(ix1,jx0);
586 dy10 = _mm256_sub_pd(iy1,jy0);
587 dz10 = _mm256_sub_pd(iz1,jz0);
588 dx20 = _mm256_sub_pd(ix2,jx0);
589 dy20 = _mm256_sub_pd(iy2,jy0);
590 dz20 = _mm256_sub_pd(iz2,jz0);
591 dx30 = _mm256_sub_pd(ix3,jx0);
592 dy30 = _mm256_sub_pd(iy3,jy0);
593 dz30 = _mm256_sub_pd(iz3,jz0);
595 /* Calculate squared distance and things based on it */
596 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
597 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
598 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
599 rsq30 = gmx_mm256_calc_rsq_pd(dx30,dy30,dz30);
601 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
602 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
603 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
604 rinv30 = gmx_mm256_invsqrt_pd(rsq30);
606 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
607 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
608 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
609 rinvsq30 = _mm256_mul_pd(rinv30,rinv30);
611 /* Load parameters for j particles */
612 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
613 charge+jnrC+0,charge+jnrD+0);
614 vdwjidx0A = 2*vdwtype[jnrA+0];
615 vdwjidx0B = 2*vdwtype[jnrB+0];
616 vdwjidx0C = 2*vdwtype[jnrC+0];
617 vdwjidx0D = 2*vdwtype[jnrD+0];
619 fjx0 = _mm256_setzero_pd();
620 fjy0 = _mm256_setzero_pd();
621 fjz0 = _mm256_setzero_pd();
623 /**************************
624 * CALCULATE INTERACTIONS *
625 **************************/
627 if (gmx_mm256_any_lt(rsq00,rcutoff2))
630 r00 = _mm256_mul_pd(rsq00,rinv00);
631 r00 = _mm256_andnot_pd(dummy_mask,r00);
633 /* Compute parameters for interactions between i and j atoms */
634 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
635 vdwioffsetptr0+vdwjidx0B,
636 vdwioffsetptr0+vdwjidx0C,
637 vdwioffsetptr0+vdwjidx0D,
640 /* LENNARD-JONES DISPERSION/REPULSION */
642 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
643 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
644 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
645 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
646 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
648 d = _mm256_sub_pd(r00,rswitch);
649 d = _mm256_max_pd(d,_mm256_setzero_pd());
650 d2 = _mm256_mul_pd(d,d);
651 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
653 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
655 /* Evaluate switch function */
656 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
657 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
658 vvdw = _mm256_mul_pd(vvdw,sw);
659 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
661 /* Update potential sum for this i atom from the interaction with this j atom. */
662 vvdw = _mm256_and_pd(vvdw,cutoff_mask);
663 vvdw = _mm256_andnot_pd(dummy_mask,vvdw);
664 vvdwsum = _mm256_add_pd(vvdwsum,vvdw);
668 fscal = _mm256_and_pd(fscal,cutoff_mask);
670 fscal = _mm256_andnot_pd(dummy_mask,fscal);
672 /* Calculate temporary vectorial force */
673 tx = _mm256_mul_pd(fscal,dx00);
674 ty = _mm256_mul_pd(fscal,dy00);
675 tz = _mm256_mul_pd(fscal,dz00);
677 /* Update vectorial force */
678 fix0 = _mm256_add_pd(fix0,tx);
679 fiy0 = _mm256_add_pd(fiy0,ty);
680 fiz0 = _mm256_add_pd(fiz0,tz);
682 fjx0 = _mm256_add_pd(fjx0,tx);
683 fjy0 = _mm256_add_pd(fjy0,ty);
684 fjz0 = _mm256_add_pd(fjz0,tz);
688 /**************************
689 * CALCULATE INTERACTIONS *
690 **************************/
692 if (gmx_mm256_any_lt(rsq10,rcutoff2))
695 r10 = _mm256_mul_pd(rsq10,rinv10);
696 r10 = _mm256_andnot_pd(dummy_mask,r10);
698 /* Compute parameters for interactions between i and j atoms */
699 qq10 = _mm256_mul_pd(iq1,jq0);
701 /* EWALD ELECTROSTATICS */
703 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
704 ewrt = _mm256_mul_pd(r10,ewtabscale);
705 ewitab = _mm256_cvttpd_epi32(ewrt);
706 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
707 ewitab = _mm_slli_epi32(ewitab,2);
708 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
709 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
710 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
711 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
712 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
713 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
714 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
715 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
716 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
718 d = _mm256_sub_pd(r10,rswitch);
719 d = _mm256_max_pd(d,_mm256_setzero_pd());
720 d2 = _mm256_mul_pd(d,d);
721 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
723 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
725 /* Evaluate switch function */
726 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
727 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
728 velec = _mm256_mul_pd(velec,sw);
729 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
731 /* Update potential sum for this i atom from the interaction with this j atom. */
732 velec = _mm256_and_pd(velec,cutoff_mask);
733 velec = _mm256_andnot_pd(dummy_mask,velec);
734 velecsum = _mm256_add_pd(velecsum,velec);
738 fscal = _mm256_and_pd(fscal,cutoff_mask);
740 fscal = _mm256_andnot_pd(dummy_mask,fscal);
742 /* Calculate temporary vectorial force */
743 tx = _mm256_mul_pd(fscal,dx10);
744 ty = _mm256_mul_pd(fscal,dy10);
745 tz = _mm256_mul_pd(fscal,dz10);
747 /* Update vectorial force */
748 fix1 = _mm256_add_pd(fix1,tx);
749 fiy1 = _mm256_add_pd(fiy1,ty);
750 fiz1 = _mm256_add_pd(fiz1,tz);
752 fjx0 = _mm256_add_pd(fjx0,tx);
753 fjy0 = _mm256_add_pd(fjy0,ty);
754 fjz0 = _mm256_add_pd(fjz0,tz);
758 /**************************
759 * CALCULATE INTERACTIONS *
760 **************************/
762 if (gmx_mm256_any_lt(rsq20,rcutoff2))
765 r20 = _mm256_mul_pd(rsq20,rinv20);
766 r20 = _mm256_andnot_pd(dummy_mask,r20);
768 /* Compute parameters for interactions between i and j atoms */
769 qq20 = _mm256_mul_pd(iq2,jq0);
771 /* EWALD ELECTROSTATICS */
773 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
774 ewrt = _mm256_mul_pd(r20,ewtabscale);
775 ewitab = _mm256_cvttpd_epi32(ewrt);
776 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
777 ewitab = _mm_slli_epi32(ewitab,2);
778 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
779 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
780 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
781 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
782 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
783 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
784 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
785 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
786 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
788 d = _mm256_sub_pd(r20,rswitch);
789 d = _mm256_max_pd(d,_mm256_setzero_pd());
790 d2 = _mm256_mul_pd(d,d);
791 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
793 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
795 /* Evaluate switch function */
796 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
797 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
798 velec = _mm256_mul_pd(velec,sw);
799 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
801 /* Update potential sum for this i atom from the interaction with this j atom. */
802 velec = _mm256_and_pd(velec,cutoff_mask);
803 velec = _mm256_andnot_pd(dummy_mask,velec);
804 velecsum = _mm256_add_pd(velecsum,velec);
808 fscal = _mm256_and_pd(fscal,cutoff_mask);
810 fscal = _mm256_andnot_pd(dummy_mask,fscal);
812 /* Calculate temporary vectorial force */
813 tx = _mm256_mul_pd(fscal,dx20);
814 ty = _mm256_mul_pd(fscal,dy20);
815 tz = _mm256_mul_pd(fscal,dz20);
817 /* Update vectorial force */
818 fix2 = _mm256_add_pd(fix2,tx);
819 fiy2 = _mm256_add_pd(fiy2,ty);
820 fiz2 = _mm256_add_pd(fiz2,tz);
822 fjx0 = _mm256_add_pd(fjx0,tx);
823 fjy0 = _mm256_add_pd(fjy0,ty);
824 fjz0 = _mm256_add_pd(fjz0,tz);
828 /**************************
829 * CALCULATE INTERACTIONS *
830 **************************/
832 if (gmx_mm256_any_lt(rsq30,rcutoff2))
835 r30 = _mm256_mul_pd(rsq30,rinv30);
836 r30 = _mm256_andnot_pd(dummy_mask,r30);
838 /* Compute parameters for interactions between i and j atoms */
839 qq30 = _mm256_mul_pd(iq3,jq0);
841 /* EWALD ELECTROSTATICS */
843 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
844 ewrt = _mm256_mul_pd(r30,ewtabscale);
845 ewitab = _mm256_cvttpd_epi32(ewrt);
846 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
847 ewitab = _mm_slli_epi32(ewitab,2);
848 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
849 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
850 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
851 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
852 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
853 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
854 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
855 velec = _mm256_mul_pd(qq30,_mm256_sub_pd(rinv30,velec));
856 felec = _mm256_mul_pd(_mm256_mul_pd(qq30,rinv30),_mm256_sub_pd(rinvsq30,felec));
858 d = _mm256_sub_pd(r30,rswitch);
859 d = _mm256_max_pd(d,_mm256_setzero_pd());
860 d2 = _mm256_mul_pd(d,d);
861 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
863 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
865 /* Evaluate switch function */
866 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
867 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv30,_mm256_mul_pd(velec,dsw)) );
868 velec = _mm256_mul_pd(velec,sw);
869 cutoff_mask = _mm256_cmp_pd(rsq30,rcutoff2,_CMP_LT_OQ);
871 /* Update potential sum for this i atom from the interaction with this j atom. */
872 velec = _mm256_and_pd(velec,cutoff_mask);
873 velec = _mm256_andnot_pd(dummy_mask,velec);
874 velecsum = _mm256_add_pd(velecsum,velec);
878 fscal = _mm256_and_pd(fscal,cutoff_mask);
880 fscal = _mm256_andnot_pd(dummy_mask,fscal);
882 /* Calculate temporary vectorial force */
883 tx = _mm256_mul_pd(fscal,dx30);
884 ty = _mm256_mul_pd(fscal,dy30);
885 tz = _mm256_mul_pd(fscal,dz30);
887 /* Update vectorial force */
888 fix3 = _mm256_add_pd(fix3,tx);
889 fiy3 = _mm256_add_pd(fiy3,ty);
890 fiz3 = _mm256_add_pd(fiz3,tz);
892 fjx0 = _mm256_add_pd(fjx0,tx);
893 fjy0 = _mm256_add_pd(fjy0,ty);
894 fjz0 = _mm256_add_pd(fjz0,tz);
898 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
899 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
900 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
901 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
903 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
905 /* Inner loop uses 261 flops */
908 /* End of innermost loop */
910 gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
911 f+i_coord_offset,fshift+i_shift_offset);
914 /* Update potential energies */
915 gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
916 gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
918 /* Increment number of inner iterations */
919 inneriter += j_index_end - j_index_start;
921 /* Outer loop uses 26 flops */
924 /* Increment number of outer iterations */
927 /* Update outer/inner flops */
929 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*261);
932 * Gromacs nonbonded kernel: nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_256_double
933 * Electrostatics interaction: Ewald
934 * VdW interaction: LennardJones
935 * Geometry: Water4-Particle
936 * Calculate force/pot: Force
939 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_256_double
940 (t_nblist * gmx_restrict nlist,
941 rvec * gmx_restrict xx,
942 rvec * gmx_restrict ff,
943 t_forcerec * gmx_restrict fr,
944 t_mdatoms * gmx_restrict mdatoms,
945 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
946 t_nrnb * gmx_restrict nrnb)
948 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
949 * just 0 for non-waters.
950 * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
951 * jnr indices corresponding to data put in the four positions in the SIMD register.
953 int i_shift_offset,i_coord_offset,outeriter,inneriter;
954 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
955 int jnrA,jnrB,jnrC,jnrD;
956 int jnrlistA,jnrlistB,jnrlistC,jnrlistD;
957 int jnrlistE,jnrlistF,jnrlistG,jnrlistH;
958 int j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
959 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
961 real *shiftvec,*fshift,*x,*f;
962 real *fjptrA,*fjptrB,*fjptrC,*fjptrD;
964 __m256d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
965 real * vdwioffsetptr0;
966 __m256d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
967 real * vdwioffsetptr1;
968 __m256d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
969 real * vdwioffsetptr2;
970 __m256d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
971 real * vdwioffsetptr3;
972 __m256d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
973 int vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
974 __m256d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
975 __m256d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
976 __m256d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
977 __m256d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
978 __m256d dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
979 __m256d velec,felec,velecsum,facel,crf,krf,krf2;
982 __m256d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
985 __m256d one_sixth = _mm256_set1_pd(1.0/6.0);
986 __m256d one_twelfth = _mm256_set1_pd(1.0/12.0);
988 __m256d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
989 __m256d beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
991 __m256d rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
992 real rswitch_scalar,d_scalar;
993 __m256d dummy_mask,cutoff_mask;
994 __m128 tmpmask0,tmpmask1;
995 __m256d signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
996 __m256d one = _mm256_set1_pd(1.0);
997 __m256d two = _mm256_set1_pd(2.0);
1003 jindex = nlist->jindex;
1005 shiftidx = nlist->shift;
1007 shiftvec = fr->shift_vec[0];
1008 fshift = fr->fshift[0];
1009 facel = _mm256_set1_pd(fr->epsfac);
1010 charge = mdatoms->chargeA;
1011 nvdwtype = fr->ntype;
1012 vdwparam = fr->nbfp;
1013 vdwtype = mdatoms->typeA;
1015 sh_ewald = _mm256_set1_pd(fr->ic->sh_ewald);
1016 beta = _mm256_set1_pd(fr->ic->ewaldcoeff);
1017 beta2 = _mm256_mul_pd(beta,beta);
1018 beta3 = _mm256_mul_pd(beta,beta2);
1020 ewtab = fr->ic->tabq_coul_FDV0;
1021 ewtabscale = _mm256_set1_pd(fr->ic->tabq_scale);
1022 ewtabhalfspace = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
1024 /* Setup water-specific parameters */
1025 inr = nlist->iinr[0];
1026 iq1 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1027 iq2 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1028 iq3 = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1029 vdwioffsetptr0 = vdwparam+2*nvdwtype*vdwtype[inr+0];
1031 /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1032 rcutoff_scalar = fr->rcoulomb;
1033 rcutoff = _mm256_set1_pd(rcutoff_scalar);
1034 rcutoff2 = _mm256_mul_pd(rcutoff,rcutoff);
1036 rswitch_scalar = fr->rcoulomb_switch;
1037 rswitch = _mm256_set1_pd(rswitch_scalar);
1038 /* Setup switch parameters */
1039 d_scalar = rcutoff_scalar-rswitch_scalar;
1040 d = _mm256_set1_pd(d_scalar);
1041 swV3 = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
1042 swV4 = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1043 swV5 = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1044 swF2 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
1045 swF3 = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1046 swF4 = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1048 /* Avoid stupid compiler warnings */
1049 jnrA = jnrB = jnrC = jnrD = 0;
1050 j_coord_offsetA = 0;
1051 j_coord_offsetB = 0;
1052 j_coord_offsetC = 0;
1053 j_coord_offsetD = 0;
1058 for(iidx=0;iidx<4*DIM;iidx++)
1060 scratch[iidx] = 0.0;
1063 /* Start outer loop over neighborlists */
1064 for(iidx=0; iidx<nri; iidx++)
1066 /* Load shift vector for this list */
1067 i_shift_offset = DIM*shiftidx[iidx];
1069 /* Load limits for loop over neighbors */
1070 j_index_start = jindex[iidx];
1071 j_index_end = jindex[iidx+1];
1073 /* Get outer coordinate index */
1075 i_coord_offset = DIM*inr;
1077 /* Load i particle coords and add shift vector */
1078 gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1079 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1081 fix0 = _mm256_setzero_pd();
1082 fiy0 = _mm256_setzero_pd();
1083 fiz0 = _mm256_setzero_pd();
1084 fix1 = _mm256_setzero_pd();
1085 fiy1 = _mm256_setzero_pd();
1086 fiz1 = _mm256_setzero_pd();
1087 fix2 = _mm256_setzero_pd();
1088 fiy2 = _mm256_setzero_pd();
1089 fiz2 = _mm256_setzero_pd();
1090 fix3 = _mm256_setzero_pd();
1091 fiy3 = _mm256_setzero_pd();
1092 fiz3 = _mm256_setzero_pd();
1094 /* Start inner kernel loop */
1095 for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1098 /* Get j neighbor index, and coordinate index */
1100 jnrB = jjnr[jidx+1];
1101 jnrC = jjnr[jidx+2];
1102 jnrD = jjnr[jidx+3];
1103 j_coord_offsetA = DIM*jnrA;
1104 j_coord_offsetB = DIM*jnrB;
1105 j_coord_offsetC = DIM*jnrC;
1106 j_coord_offsetD = DIM*jnrD;
1108 /* load j atom coordinates */
1109 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1110 x+j_coord_offsetC,x+j_coord_offsetD,
1113 /* Calculate displacement vector */
1114 dx00 = _mm256_sub_pd(ix0,jx0);
1115 dy00 = _mm256_sub_pd(iy0,jy0);
1116 dz00 = _mm256_sub_pd(iz0,jz0);
1117 dx10 = _mm256_sub_pd(ix1,jx0);
1118 dy10 = _mm256_sub_pd(iy1,jy0);
1119 dz10 = _mm256_sub_pd(iz1,jz0);
1120 dx20 = _mm256_sub_pd(ix2,jx0);
1121 dy20 = _mm256_sub_pd(iy2,jy0);
1122 dz20 = _mm256_sub_pd(iz2,jz0);
1123 dx30 = _mm256_sub_pd(ix3,jx0);
1124 dy30 = _mm256_sub_pd(iy3,jy0);
1125 dz30 = _mm256_sub_pd(iz3,jz0);
1127 /* Calculate squared distance and things based on it */
1128 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1129 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1130 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1131 rsq30 = gmx_mm256_calc_rsq_pd(dx30,dy30,dz30);
1133 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1134 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1135 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1136 rinv30 = gmx_mm256_invsqrt_pd(rsq30);
1138 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1139 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1140 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1141 rinvsq30 = _mm256_mul_pd(rinv30,rinv30);
1143 /* Load parameters for j particles */
1144 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1145 charge+jnrC+0,charge+jnrD+0);
1146 vdwjidx0A = 2*vdwtype[jnrA+0];
1147 vdwjidx0B = 2*vdwtype[jnrB+0];
1148 vdwjidx0C = 2*vdwtype[jnrC+0];
1149 vdwjidx0D = 2*vdwtype[jnrD+0];
1151 fjx0 = _mm256_setzero_pd();
1152 fjy0 = _mm256_setzero_pd();
1153 fjz0 = _mm256_setzero_pd();
1155 /**************************
1156 * CALCULATE INTERACTIONS *
1157 **************************/
1159 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1162 r00 = _mm256_mul_pd(rsq00,rinv00);
1164 /* Compute parameters for interactions between i and j atoms */
1165 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
1166 vdwioffsetptr0+vdwjidx0B,
1167 vdwioffsetptr0+vdwjidx0C,
1168 vdwioffsetptr0+vdwjidx0D,
1171 /* LENNARD-JONES DISPERSION/REPULSION */
1173 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1174 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
1175 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
1176 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
1177 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
1179 d = _mm256_sub_pd(r00,rswitch);
1180 d = _mm256_max_pd(d,_mm256_setzero_pd());
1181 d2 = _mm256_mul_pd(d,d);
1182 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1184 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1186 /* Evaluate switch function */
1187 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1188 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
1189 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1193 fscal = _mm256_and_pd(fscal,cutoff_mask);
1195 /* Calculate temporary vectorial force */
1196 tx = _mm256_mul_pd(fscal,dx00);
1197 ty = _mm256_mul_pd(fscal,dy00);
1198 tz = _mm256_mul_pd(fscal,dz00);
1200 /* Update vectorial force */
1201 fix0 = _mm256_add_pd(fix0,tx);
1202 fiy0 = _mm256_add_pd(fiy0,ty);
1203 fiz0 = _mm256_add_pd(fiz0,tz);
1205 fjx0 = _mm256_add_pd(fjx0,tx);
1206 fjy0 = _mm256_add_pd(fjy0,ty);
1207 fjz0 = _mm256_add_pd(fjz0,tz);
1211 /**************************
1212 * CALCULATE INTERACTIONS *
1213 **************************/
1215 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1218 r10 = _mm256_mul_pd(rsq10,rinv10);
1220 /* Compute parameters for interactions between i and j atoms */
1221 qq10 = _mm256_mul_pd(iq1,jq0);
1223 /* EWALD ELECTROSTATICS */
1225 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1226 ewrt = _mm256_mul_pd(r10,ewtabscale);
1227 ewitab = _mm256_cvttpd_epi32(ewrt);
1228 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1229 ewitab = _mm_slli_epi32(ewitab,2);
1230 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1231 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1232 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1233 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1234 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1235 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1236 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1237 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1238 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1240 d = _mm256_sub_pd(r10,rswitch);
1241 d = _mm256_max_pd(d,_mm256_setzero_pd());
1242 d2 = _mm256_mul_pd(d,d);
1243 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1245 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1247 /* Evaluate switch function */
1248 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1249 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1250 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1254 fscal = _mm256_and_pd(fscal,cutoff_mask);
1256 /* Calculate temporary vectorial force */
1257 tx = _mm256_mul_pd(fscal,dx10);
1258 ty = _mm256_mul_pd(fscal,dy10);
1259 tz = _mm256_mul_pd(fscal,dz10);
1261 /* Update vectorial force */
1262 fix1 = _mm256_add_pd(fix1,tx);
1263 fiy1 = _mm256_add_pd(fiy1,ty);
1264 fiz1 = _mm256_add_pd(fiz1,tz);
1266 fjx0 = _mm256_add_pd(fjx0,tx);
1267 fjy0 = _mm256_add_pd(fjy0,ty);
1268 fjz0 = _mm256_add_pd(fjz0,tz);
1272 /**************************
1273 * CALCULATE INTERACTIONS *
1274 **************************/
1276 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1279 r20 = _mm256_mul_pd(rsq20,rinv20);
1281 /* Compute parameters for interactions between i and j atoms */
1282 qq20 = _mm256_mul_pd(iq2,jq0);
1284 /* EWALD ELECTROSTATICS */
1286 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1287 ewrt = _mm256_mul_pd(r20,ewtabscale);
1288 ewitab = _mm256_cvttpd_epi32(ewrt);
1289 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1290 ewitab = _mm_slli_epi32(ewitab,2);
1291 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1292 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1293 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1294 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1295 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1296 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1297 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1298 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1299 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1301 d = _mm256_sub_pd(r20,rswitch);
1302 d = _mm256_max_pd(d,_mm256_setzero_pd());
1303 d2 = _mm256_mul_pd(d,d);
1304 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1306 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(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_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1311 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1315 fscal = _mm256_and_pd(fscal,cutoff_mask);
1317 /* Calculate temporary vectorial force */
1318 tx = _mm256_mul_pd(fscal,dx20);
1319 ty = _mm256_mul_pd(fscal,dy20);
1320 tz = _mm256_mul_pd(fscal,dz20);
1322 /* Update vectorial force */
1323 fix2 = _mm256_add_pd(fix2,tx);
1324 fiy2 = _mm256_add_pd(fiy2,ty);
1325 fiz2 = _mm256_add_pd(fiz2,tz);
1327 fjx0 = _mm256_add_pd(fjx0,tx);
1328 fjy0 = _mm256_add_pd(fjy0,ty);
1329 fjz0 = _mm256_add_pd(fjz0,tz);
1333 /**************************
1334 * CALCULATE INTERACTIONS *
1335 **************************/
1337 if (gmx_mm256_any_lt(rsq30,rcutoff2))
1340 r30 = _mm256_mul_pd(rsq30,rinv30);
1342 /* Compute parameters for interactions between i and j atoms */
1343 qq30 = _mm256_mul_pd(iq3,jq0);
1345 /* EWALD ELECTROSTATICS */
1347 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1348 ewrt = _mm256_mul_pd(r30,ewtabscale);
1349 ewitab = _mm256_cvttpd_epi32(ewrt);
1350 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1351 ewitab = _mm_slli_epi32(ewitab,2);
1352 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1353 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1354 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1355 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1356 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1357 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1358 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1359 velec = _mm256_mul_pd(qq30,_mm256_sub_pd(rinv30,velec));
1360 felec = _mm256_mul_pd(_mm256_mul_pd(qq30,rinv30),_mm256_sub_pd(rinvsq30,felec));
1362 d = _mm256_sub_pd(r30,rswitch);
1363 d = _mm256_max_pd(d,_mm256_setzero_pd());
1364 d2 = _mm256_mul_pd(d,d);
1365 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1367 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1369 /* Evaluate switch function */
1370 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1371 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv30,_mm256_mul_pd(velec,dsw)) );
1372 cutoff_mask = _mm256_cmp_pd(rsq30,rcutoff2,_CMP_LT_OQ);
1376 fscal = _mm256_and_pd(fscal,cutoff_mask);
1378 /* Calculate temporary vectorial force */
1379 tx = _mm256_mul_pd(fscal,dx30);
1380 ty = _mm256_mul_pd(fscal,dy30);
1381 tz = _mm256_mul_pd(fscal,dz30);
1383 /* Update vectorial force */
1384 fix3 = _mm256_add_pd(fix3,tx);
1385 fiy3 = _mm256_add_pd(fiy3,ty);
1386 fiz3 = _mm256_add_pd(fiz3,tz);
1388 fjx0 = _mm256_add_pd(fjx0,tx);
1389 fjy0 = _mm256_add_pd(fjy0,ty);
1390 fjz0 = _mm256_add_pd(fjz0,tz);
1394 fjptrA = f+j_coord_offsetA;
1395 fjptrB = f+j_coord_offsetB;
1396 fjptrC = f+j_coord_offsetC;
1397 fjptrD = f+j_coord_offsetD;
1399 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1401 /* Inner loop uses 245 flops */
1404 if(jidx<j_index_end)
1407 /* Get j neighbor index, and coordinate index */
1408 jnrlistA = jjnr[jidx];
1409 jnrlistB = jjnr[jidx+1];
1410 jnrlistC = jjnr[jidx+2];
1411 jnrlistD = jjnr[jidx+3];
1412 /* Sign of each element will be negative for non-real atoms.
1413 * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1414 * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1416 tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1418 tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1419 tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1420 dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1422 jnrA = (jnrlistA>=0) ? jnrlistA : 0;
1423 jnrB = (jnrlistB>=0) ? jnrlistB : 0;
1424 jnrC = (jnrlistC>=0) ? jnrlistC : 0;
1425 jnrD = (jnrlistD>=0) ? jnrlistD : 0;
1426 j_coord_offsetA = DIM*jnrA;
1427 j_coord_offsetB = DIM*jnrB;
1428 j_coord_offsetC = DIM*jnrC;
1429 j_coord_offsetD = DIM*jnrD;
1431 /* load j atom coordinates */
1432 gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1433 x+j_coord_offsetC,x+j_coord_offsetD,
1436 /* Calculate displacement vector */
1437 dx00 = _mm256_sub_pd(ix0,jx0);
1438 dy00 = _mm256_sub_pd(iy0,jy0);
1439 dz00 = _mm256_sub_pd(iz0,jz0);
1440 dx10 = _mm256_sub_pd(ix1,jx0);
1441 dy10 = _mm256_sub_pd(iy1,jy0);
1442 dz10 = _mm256_sub_pd(iz1,jz0);
1443 dx20 = _mm256_sub_pd(ix2,jx0);
1444 dy20 = _mm256_sub_pd(iy2,jy0);
1445 dz20 = _mm256_sub_pd(iz2,jz0);
1446 dx30 = _mm256_sub_pd(ix3,jx0);
1447 dy30 = _mm256_sub_pd(iy3,jy0);
1448 dz30 = _mm256_sub_pd(iz3,jz0);
1450 /* Calculate squared distance and things based on it */
1451 rsq00 = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1452 rsq10 = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1453 rsq20 = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1454 rsq30 = gmx_mm256_calc_rsq_pd(dx30,dy30,dz30);
1456 rinv00 = gmx_mm256_invsqrt_pd(rsq00);
1457 rinv10 = gmx_mm256_invsqrt_pd(rsq10);
1458 rinv20 = gmx_mm256_invsqrt_pd(rsq20);
1459 rinv30 = gmx_mm256_invsqrt_pd(rsq30);
1461 rinvsq00 = _mm256_mul_pd(rinv00,rinv00);
1462 rinvsq10 = _mm256_mul_pd(rinv10,rinv10);
1463 rinvsq20 = _mm256_mul_pd(rinv20,rinv20);
1464 rinvsq30 = _mm256_mul_pd(rinv30,rinv30);
1466 /* Load parameters for j particles */
1467 jq0 = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1468 charge+jnrC+0,charge+jnrD+0);
1469 vdwjidx0A = 2*vdwtype[jnrA+0];
1470 vdwjidx0B = 2*vdwtype[jnrB+0];
1471 vdwjidx0C = 2*vdwtype[jnrC+0];
1472 vdwjidx0D = 2*vdwtype[jnrD+0];
1474 fjx0 = _mm256_setzero_pd();
1475 fjy0 = _mm256_setzero_pd();
1476 fjz0 = _mm256_setzero_pd();
1478 /**************************
1479 * CALCULATE INTERACTIONS *
1480 **************************/
1482 if (gmx_mm256_any_lt(rsq00,rcutoff2))
1485 r00 = _mm256_mul_pd(rsq00,rinv00);
1486 r00 = _mm256_andnot_pd(dummy_mask,r00);
1488 /* Compute parameters for interactions between i and j atoms */
1489 gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
1490 vdwioffsetptr0+vdwjidx0B,
1491 vdwioffsetptr0+vdwjidx0C,
1492 vdwioffsetptr0+vdwjidx0D,
1495 /* LENNARD-JONES DISPERSION/REPULSION */
1497 rinvsix = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1498 vvdw6 = _mm256_mul_pd(c6_00,rinvsix);
1499 vvdw12 = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
1500 vvdw = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
1501 fvdw = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
1503 d = _mm256_sub_pd(r00,rswitch);
1504 d = _mm256_max_pd(d,_mm256_setzero_pd());
1505 d2 = _mm256_mul_pd(d,d);
1506 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1508 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1510 /* Evaluate switch function */
1511 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1512 fvdw = _mm256_sub_pd( _mm256_mul_pd(fvdw,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(vvdw,dsw)) );
1513 cutoff_mask = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1517 fscal = _mm256_and_pd(fscal,cutoff_mask);
1519 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1521 /* Calculate temporary vectorial force */
1522 tx = _mm256_mul_pd(fscal,dx00);
1523 ty = _mm256_mul_pd(fscal,dy00);
1524 tz = _mm256_mul_pd(fscal,dz00);
1526 /* Update vectorial force */
1527 fix0 = _mm256_add_pd(fix0,tx);
1528 fiy0 = _mm256_add_pd(fiy0,ty);
1529 fiz0 = _mm256_add_pd(fiz0,tz);
1531 fjx0 = _mm256_add_pd(fjx0,tx);
1532 fjy0 = _mm256_add_pd(fjy0,ty);
1533 fjz0 = _mm256_add_pd(fjz0,tz);
1537 /**************************
1538 * CALCULATE INTERACTIONS *
1539 **************************/
1541 if (gmx_mm256_any_lt(rsq10,rcutoff2))
1544 r10 = _mm256_mul_pd(rsq10,rinv10);
1545 r10 = _mm256_andnot_pd(dummy_mask,r10);
1547 /* Compute parameters for interactions between i and j atoms */
1548 qq10 = _mm256_mul_pd(iq1,jq0);
1550 /* EWALD ELECTROSTATICS */
1552 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1553 ewrt = _mm256_mul_pd(r10,ewtabscale);
1554 ewitab = _mm256_cvttpd_epi32(ewrt);
1555 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1556 ewitab = _mm_slli_epi32(ewitab,2);
1557 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1558 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1559 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1560 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1561 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1562 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1563 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1564 velec = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1565 felec = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1567 d = _mm256_sub_pd(r10,rswitch);
1568 d = _mm256_max_pd(d,_mm256_setzero_pd());
1569 d2 = _mm256_mul_pd(d,d);
1570 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1572 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1574 /* Evaluate switch function */
1575 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1576 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1577 cutoff_mask = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1581 fscal = _mm256_and_pd(fscal,cutoff_mask);
1583 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1585 /* Calculate temporary vectorial force */
1586 tx = _mm256_mul_pd(fscal,dx10);
1587 ty = _mm256_mul_pd(fscal,dy10);
1588 tz = _mm256_mul_pd(fscal,dz10);
1590 /* Update vectorial force */
1591 fix1 = _mm256_add_pd(fix1,tx);
1592 fiy1 = _mm256_add_pd(fiy1,ty);
1593 fiz1 = _mm256_add_pd(fiz1,tz);
1595 fjx0 = _mm256_add_pd(fjx0,tx);
1596 fjy0 = _mm256_add_pd(fjy0,ty);
1597 fjz0 = _mm256_add_pd(fjz0,tz);
1601 /**************************
1602 * CALCULATE INTERACTIONS *
1603 **************************/
1605 if (gmx_mm256_any_lt(rsq20,rcutoff2))
1608 r20 = _mm256_mul_pd(rsq20,rinv20);
1609 r20 = _mm256_andnot_pd(dummy_mask,r20);
1611 /* Compute parameters for interactions between i and j atoms */
1612 qq20 = _mm256_mul_pd(iq2,jq0);
1614 /* EWALD ELECTROSTATICS */
1616 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1617 ewrt = _mm256_mul_pd(r20,ewtabscale);
1618 ewitab = _mm256_cvttpd_epi32(ewrt);
1619 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1620 ewitab = _mm_slli_epi32(ewitab,2);
1621 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1622 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1623 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1624 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1625 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1626 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1627 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1628 velec = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1629 felec = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1631 d = _mm256_sub_pd(r20,rswitch);
1632 d = _mm256_max_pd(d,_mm256_setzero_pd());
1633 d2 = _mm256_mul_pd(d,d);
1634 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1636 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1638 /* Evaluate switch function */
1639 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1640 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1641 cutoff_mask = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1645 fscal = _mm256_and_pd(fscal,cutoff_mask);
1647 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1649 /* Calculate temporary vectorial force */
1650 tx = _mm256_mul_pd(fscal,dx20);
1651 ty = _mm256_mul_pd(fscal,dy20);
1652 tz = _mm256_mul_pd(fscal,dz20);
1654 /* Update vectorial force */
1655 fix2 = _mm256_add_pd(fix2,tx);
1656 fiy2 = _mm256_add_pd(fiy2,ty);
1657 fiz2 = _mm256_add_pd(fiz2,tz);
1659 fjx0 = _mm256_add_pd(fjx0,tx);
1660 fjy0 = _mm256_add_pd(fjy0,ty);
1661 fjz0 = _mm256_add_pd(fjz0,tz);
1665 /**************************
1666 * CALCULATE INTERACTIONS *
1667 **************************/
1669 if (gmx_mm256_any_lt(rsq30,rcutoff2))
1672 r30 = _mm256_mul_pd(rsq30,rinv30);
1673 r30 = _mm256_andnot_pd(dummy_mask,r30);
1675 /* Compute parameters for interactions between i and j atoms */
1676 qq30 = _mm256_mul_pd(iq3,jq0);
1678 /* EWALD ELECTROSTATICS */
1680 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1681 ewrt = _mm256_mul_pd(r30,ewtabscale);
1682 ewitab = _mm256_cvttpd_epi32(ewrt);
1683 eweps = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1684 ewitab = _mm_slli_epi32(ewitab,2);
1685 ewtabF = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1686 ewtabD = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1687 ewtabV = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1688 ewtabFn = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1689 GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1690 felec = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1691 velec = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1692 velec = _mm256_mul_pd(qq30,_mm256_sub_pd(rinv30,velec));
1693 felec = _mm256_mul_pd(_mm256_mul_pd(qq30,rinv30),_mm256_sub_pd(rinvsq30,felec));
1695 d = _mm256_sub_pd(r30,rswitch);
1696 d = _mm256_max_pd(d,_mm256_setzero_pd());
1697 d2 = _mm256_mul_pd(d,d);
1698 sw = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1700 dsw = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1702 /* Evaluate switch function */
1703 /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1704 felec = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv30,_mm256_mul_pd(velec,dsw)) );
1705 cutoff_mask = _mm256_cmp_pd(rsq30,rcutoff2,_CMP_LT_OQ);
1709 fscal = _mm256_and_pd(fscal,cutoff_mask);
1711 fscal = _mm256_andnot_pd(dummy_mask,fscal);
1713 /* Calculate temporary vectorial force */
1714 tx = _mm256_mul_pd(fscal,dx30);
1715 ty = _mm256_mul_pd(fscal,dy30);
1716 tz = _mm256_mul_pd(fscal,dz30);
1718 /* Update vectorial force */
1719 fix3 = _mm256_add_pd(fix3,tx);
1720 fiy3 = _mm256_add_pd(fiy3,ty);
1721 fiz3 = _mm256_add_pd(fiz3,tz);
1723 fjx0 = _mm256_add_pd(fjx0,tx);
1724 fjy0 = _mm256_add_pd(fjy0,ty);
1725 fjz0 = _mm256_add_pd(fjz0,tz);
1729 fjptrA = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1730 fjptrB = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1731 fjptrC = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1732 fjptrD = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1734 gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1736 /* Inner loop uses 249 flops */
1739 /* End of innermost loop */
1741 gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1742 f+i_coord_offset,fshift+i_shift_offset);
1744 /* Increment number of inner iterations */
1745 inneriter += j_index_end - j_index_start;
1747 /* Outer loop uses 24 flops */
1750 /* Increment number of outer iterations */
1753 /* Update outer/inner flops */
1755 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*249);