src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwLJ_GeomW3W3_avx_256_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_single kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_256_single.h"
  34 #include "kernelutil_x86_avx_256_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3W3_VF_avx_256_single
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water3-Water3
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwLJ_GeomW3W3_VF_avx_256_single
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrE,jnrF,jnrG,jnrH;
  62     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  63     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  64     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  65     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  66     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  67     real             rcutoff_scalar;
  68     real             *shiftvec,*fshift,*x,*f;
  69     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  70     real             scratch[4*DIM];
  71     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  72     real *           vdwioffsetptr0;
  73     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  74     real *           vdwioffsetptr1;
  75     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  76     real *           vdwioffsetptr2;
  77     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  78     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  79     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  80     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  81     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  82     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  83     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  84     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  85     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  86     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  87     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  88     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  89     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  90     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  91     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  92     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  93     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 100     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 101     __m256           dummy_mask,cutoff_mask;
 102     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 103     __m256           one     = _mm256_set1_ps(1.0);
 104     __m256           two     = _mm256_set1_ps(2.0);
 105     x                = xx[0];
 106     f                = ff[0];
 107
 108     nri              = nlist->nri;
 109     iinr             = nlist->iinr;
 110     jindex           = nlist->jindex;
 111     jjnr             = nlist->jjnr;
 112     shiftidx         = nlist->shift;
 113     gid              = nlist->gid;
 114     shiftvec         = fr->shift_vec[0];
 115     fshift           = fr->fshift[0];
 116     facel            = _mm256_set1_ps(fr->epsfac);
 117     charge           = mdatoms->chargeA;
 118     nvdwtype         = fr->ntype;
 119     vdwparam         = fr->nbfp;
 120     vdwtype          = mdatoms->typeA;
 121
 122     /* Setup water-specific parameters */
 123     inr              = nlist->iinr[0];
 124     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 125     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 126     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 127     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 128
 129     jq0              = _mm256_set1_ps(charge[inr+0]);
 130     jq1              = _mm256_set1_ps(charge[inr+1]);
 131     jq2              = _mm256_set1_ps(charge[inr+2]);
 132     vdwjidx0A        = 2*vdwtype[inr+0];
 133     qq00             = _mm256_mul_ps(iq0,jq0);
 134     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 135     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 136     qq01             = _mm256_mul_ps(iq0,jq1);
 137     qq02             = _mm256_mul_ps(iq0,jq2);
 138     qq10             = _mm256_mul_ps(iq1,jq0);
 139     qq11             = _mm256_mul_ps(iq1,jq1);
 140     qq12             = _mm256_mul_ps(iq1,jq2);
 141     qq20             = _mm256_mul_ps(iq2,jq0);
 142     qq21             = _mm256_mul_ps(iq2,jq1);
 143     qq22             = _mm256_mul_ps(iq2,jq2);
 144
 145     /* Avoid stupid compiler warnings */
 146     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 147     j_coord_offsetA = 0;
 148     j_coord_offsetB = 0;
 149     j_coord_offsetC = 0;
 150     j_coord_offsetD = 0;
 151     j_coord_offsetE = 0;
 152     j_coord_offsetF = 0;
 153     j_coord_offsetG = 0;
 154     j_coord_offsetH = 0;
 155
 156     outeriter        = 0;
 157     inneriter        = 0;
 158
 159     for(iidx=0;iidx<4*DIM;iidx++)
 160     {
 161         scratch[iidx] = 0.0;
 162     }
 163
 164     /* Start outer loop over neighborlists */
 165     for(iidx=0; iidx<nri; iidx++)
 166     {
 167         /* Load shift vector for this list */
 168         i_shift_offset   = DIM*shiftidx[iidx];
 169
 170         /* Load limits for loop over neighbors */
 171         j_index_start    = jindex[iidx];
 172         j_index_end      = jindex[iidx+1];
 173
 174         /* Get outer coordinate index */
 175         inr              = iinr[iidx];
 176         i_coord_offset   = DIM*inr;
 177
 178         /* Load i particle coords and add shift vector */
 179         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 180                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 181
 182         fix0             = _mm256_setzero_ps();
 183         fiy0             = _mm256_setzero_ps();
 184         fiz0             = _mm256_setzero_ps();
 185         fix1             = _mm256_setzero_ps();
 186         fiy1             = _mm256_setzero_ps();
 187         fiz1             = _mm256_setzero_ps();
 188         fix2             = _mm256_setzero_ps();
 189         fiy2             = _mm256_setzero_ps();
 190         fiz2             = _mm256_setzero_ps();
 191
 192         /* Reset potential sums */
 193         velecsum         = _mm256_setzero_ps();
 194         vvdwsum          = _mm256_setzero_ps();
 195
 196         /* Start inner kernel loop */
 197         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 198         {
 199
 200             /* Get j neighbor index, and coordinate index */
 201             jnrA             = jjnr[jidx];
 202             jnrB             = jjnr[jidx+1];
 203             jnrC             = jjnr[jidx+2];
 204             jnrD             = jjnr[jidx+3];
 205             jnrE             = jjnr[jidx+4];
 206             jnrF             = jjnr[jidx+5];
 207             jnrG             = jjnr[jidx+6];
 208             jnrH             = jjnr[jidx+7];
 209             j_coord_offsetA  = DIM*jnrA;
 210             j_coord_offsetB  = DIM*jnrB;
 211             j_coord_offsetC  = DIM*jnrC;
 212             j_coord_offsetD  = DIM*jnrD;
 213             j_coord_offsetE  = DIM*jnrE;
 214             j_coord_offsetF  = DIM*jnrF;
 215             j_coord_offsetG  = DIM*jnrG;
 216             j_coord_offsetH  = DIM*jnrH;
 217
 218             /* load j atom coordinates */
 219             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 220                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 221                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 222                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 223                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 224
 225             /* Calculate displacement vector */
 226             dx00             = _mm256_sub_ps(ix0,jx0);
 227             dy00             = _mm256_sub_ps(iy0,jy0);
 228             dz00             = _mm256_sub_ps(iz0,jz0);
 229             dx01             = _mm256_sub_ps(ix0,jx1);
 230             dy01             = _mm256_sub_ps(iy0,jy1);
 231             dz01             = _mm256_sub_ps(iz0,jz1);
 232             dx02             = _mm256_sub_ps(ix0,jx2);
 233             dy02             = _mm256_sub_ps(iy0,jy2);
 234             dz02             = _mm256_sub_ps(iz0,jz2);
 235             dx10             = _mm256_sub_ps(ix1,jx0);
 236             dy10             = _mm256_sub_ps(iy1,jy0);
 237             dz10             = _mm256_sub_ps(iz1,jz0);
 238             dx11             = _mm256_sub_ps(ix1,jx1);
 239             dy11             = _mm256_sub_ps(iy1,jy1);
 240             dz11             = _mm256_sub_ps(iz1,jz1);
 241             dx12             = _mm256_sub_ps(ix1,jx2);
 242             dy12             = _mm256_sub_ps(iy1,jy2);
 243             dz12             = _mm256_sub_ps(iz1,jz2);
 244             dx20             = _mm256_sub_ps(ix2,jx0);
 245             dy20             = _mm256_sub_ps(iy2,jy0);
 246             dz20             = _mm256_sub_ps(iz2,jz0);
 247             dx21             = _mm256_sub_ps(ix2,jx1);
 248             dy21             = _mm256_sub_ps(iy2,jy1);
 249             dz21             = _mm256_sub_ps(iz2,jz1);
 250             dx22             = _mm256_sub_ps(ix2,jx2);
 251             dy22             = _mm256_sub_ps(iy2,jy2);
 252             dz22             = _mm256_sub_ps(iz2,jz2);
 253
 254             /* Calculate squared distance and things based on it */
 255             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 256             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 257             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 258             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 259             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 260             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 261             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 262             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 263             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 264
 265             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 266             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 267             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 268             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 269             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 270             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 271             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 272             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 273             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 274
 275             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 276             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 277             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 278             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 279             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 280             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 281             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 282             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 283             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 284
 285             fjx0             = _mm256_setzero_ps();
 286             fjy0             = _mm256_setzero_ps();
 287             fjz0             = _mm256_setzero_ps();
 288             fjx1             = _mm256_setzero_ps();
 289             fjy1             = _mm256_setzero_ps();
 290             fjz1             = _mm256_setzero_ps();
 291             fjx2             = _mm256_setzero_ps();
 292             fjy2             = _mm256_setzero_ps();
 293             fjz2             = _mm256_setzero_ps();
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             /* COULOMB ELECTROSTATICS */
 300             velec            = _mm256_mul_ps(qq00,rinv00);
 301             felec            = _mm256_mul_ps(velec,rinvsq00);
 302
 303             /* LENNARD-JONES DISPERSION/REPULSION */
 304
 305             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 306             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 307             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 308             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 309             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 310
 311             /* Update potential sum for this i atom from the interaction with this j atom. */
 312             velecsum         = _mm256_add_ps(velecsum,velec);
 313             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 314
 315             fscal            = _mm256_add_ps(felec,fvdw);
 316
 317             /* Calculate temporary vectorial force */
 318             tx               = _mm256_mul_ps(fscal,dx00);
 319             ty               = _mm256_mul_ps(fscal,dy00);
 320             tz               = _mm256_mul_ps(fscal,dz00);
 321
 322             /* Update vectorial force */
 323             fix0             = _mm256_add_ps(fix0,tx);
 324             fiy0             = _mm256_add_ps(fiy0,ty);
 325             fiz0             = _mm256_add_ps(fiz0,tz);
 326
 327             fjx0             = _mm256_add_ps(fjx0,tx);
 328             fjy0             = _mm256_add_ps(fjy0,ty);
 329             fjz0             = _mm256_add_ps(fjz0,tz);
 330
 331             /**************************
 332              * CALCULATE INTERACTIONS *
 333              **************************/
 334
 335             /* COULOMB ELECTROSTATICS */
 336             velec            = _mm256_mul_ps(qq01,rinv01);
 337             felec            = _mm256_mul_ps(velec,rinvsq01);
 338
 339             /* Update potential sum for this i atom from the interaction with this j atom. */
 340             velecsum         = _mm256_add_ps(velecsum,velec);
 341
 342             fscal            = felec;
 343
 344             /* Calculate temporary vectorial force */
 345             tx               = _mm256_mul_ps(fscal,dx01);
 346             ty               = _mm256_mul_ps(fscal,dy01);
 347             tz               = _mm256_mul_ps(fscal,dz01);
 348
 349             /* Update vectorial force */
 350             fix0             = _mm256_add_ps(fix0,tx);
 351             fiy0             = _mm256_add_ps(fiy0,ty);
 352             fiz0             = _mm256_add_ps(fiz0,tz);
 353
 354             fjx1             = _mm256_add_ps(fjx1,tx);
 355             fjy1             = _mm256_add_ps(fjy1,ty);
 356             fjz1             = _mm256_add_ps(fjz1,tz);
 357
 358             /**************************
 359              * CALCULATE INTERACTIONS *
 360              **************************/
 361
 362             /* COULOMB ELECTROSTATICS */
 363             velec            = _mm256_mul_ps(qq02,rinv02);
 364             felec            = _mm256_mul_ps(velec,rinvsq02);
 365
 366             /* Update potential sum for this i atom from the interaction with this j atom. */
 367             velecsum         = _mm256_add_ps(velecsum,velec);
 368
 369             fscal            = felec;
 370
 371             /* Calculate temporary vectorial force */
 372             tx               = _mm256_mul_ps(fscal,dx02);
 373             ty               = _mm256_mul_ps(fscal,dy02);
 374             tz               = _mm256_mul_ps(fscal,dz02);
 375
 376             /* Update vectorial force */
 377             fix0             = _mm256_add_ps(fix0,tx);
 378             fiy0             = _mm256_add_ps(fiy0,ty);
 379             fiz0             = _mm256_add_ps(fiz0,tz);
 380
 381             fjx2             = _mm256_add_ps(fjx2,tx);
 382             fjy2             = _mm256_add_ps(fjy2,ty);
 383             fjz2             = _mm256_add_ps(fjz2,tz);
 384
 385             /**************************
 386              * CALCULATE INTERACTIONS *
 387              **************************/
 388
 389             /* COULOMB ELECTROSTATICS */
 390             velec            = _mm256_mul_ps(qq10,rinv10);
 391             felec            = _mm256_mul_ps(velec,rinvsq10);
 392
 393             /* Update potential sum for this i atom from the interaction with this j atom. */
 394             velecsum         = _mm256_add_ps(velecsum,velec);
 395
 396             fscal            = felec;
 397
 398             /* Calculate temporary vectorial force */
 399             tx               = _mm256_mul_ps(fscal,dx10);
 400             ty               = _mm256_mul_ps(fscal,dy10);
 401             tz               = _mm256_mul_ps(fscal,dz10);
 402
 403             /* Update vectorial force */
 404             fix1             = _mm256_add_ps(fix1,tx);
 405             fiy1             = _mm256_add_ps(fiy1,ty);
 406             fiz1             = _mm256_add_ps(fiz1,tz);
 407
 408             fjx0             = _mm256_add_ps(fjx0,tx);
 409             fjy0             = _mm256_add_ps(fjy0,ty);
 410             fjz0             = _mm256_add_ps(fjz0,tz);
 411
 412             /**************************
 413              * CALCULATE INTERACTIONS *
 414              **************************/
 415
 416             /* COULOMB ELECTROSTATICS */
 417             velec            = _mm256_mul_ps(qq11,rinv11);
 418             felec            = _mm256_mul_ps(velec,rinvsq11);
 419
 420             /* Update potential sum for this i atom from the interaction with this j atom. */
 421             velecsum         = _mm256_add_ps(velecsum,velec);
 422
 423             fscal            = felec;
 424
 425             /* Calculate temporary vectorial force */
 426             tx               = _mm256_mul_ps(fscal,dx11);
 427             ty               = _mm256_mul_ps(fscal,dy11);
 428             tz               = _mm256_mul_ps(fscal,dz11);
 429
 430             /* Update vectorial force */
 431             fix1             = _mm256_add_ps(fix1,tx);
 432             fiy1             = _mm256_add_ps(fiy1,ty);
 433             fiz1             = _mm256_add_ps(fiz1,tz);
 434
 435             fjx1             = _mm256_add_ps(fjx1,tx);
 436             fjy1             = _mm256_add_ps(fjy1,ty);
 437             fjz1             = _mm256_add_ps(fjz1,tz);
 438
 439             /**************************
 440              * CALCULATE INTERACTIONS *
 441              **************************/
 442
 443             /* COULOMB ELECTROSTATICS */
 444             velec            = _mm256_mul_ps(qq12,rinv12);
 445             felec            = _mm256_mul_ps(velec,rinvsq12);
 446
 447             /* Update potential sum for this i atom from the interaction with this j atom. */
 448             velecsum         = _mm256_add_ps(velecsum,velec);
 449
 450             fscal            = felec;
 451
 452             /* Calculate temporary vectorial force */
 453             tx               = _mm256_mul_ps(fscal,dx12);
 454             ty               = _mm256_mul_ps(fscal,dy12);
 455             tz               = _mm256_mul_ps(fscal,dz12);
 456
 457             /* Update vectorial force */
 458             fix1             = _mm256_add_ps(fix1,tx);
 459             fiy1             = _mm256_add_ps(fiy1,ty);
 460             fiz1             = _mm256_add_ps(fiz1,tz);
 461
 462             fjx2             = _mm256_add_ps(fjx2,tx);
 463             fjy2             = _mm256_add_ps(fjy2,ty);
 464             fjz2             = _mm256_add_ps(fjz2,tz);
 465
 466             /**************************
 467              * CALCULATE INTERACTIONS *
 468              **************************/
 469
 470             /* COULOMB ELECTROSTATICS */
 471             velec            = _mm256_mul_ps(qq20,rinv20);
 472             felec            = _mm256_mul_ps(velec,rinvsq20);
 473
 474             /* Update potential sum for this i atom from the interaction with this j atom. */
 475             velecsum         = _mm256_add_ps(velecsum,velec);
 476
 477             fscal            = felec;
 478
 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);
 483
 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);
 488
 489             fjx0             = _mm256_add_ps(fjx0,tx);
 490             fjy0             = _mm256_add_ps(fjy0,ty);
 491             fjz0             = _mm256_add_ps(fjz0,tz);
 492
 493             /**************************
 494              * CALCULATE INTERACTIONS *
 495              **************************/
 496
 497             /* COULOMB ELECTROSTATICS */
 498             velec            = _mm256_mul_ps(qq21,rinv21);
 499             felec            = _mm256_mul_ps(velec,rinvsq21);
 500
 501             /* Update potential sum for this i atom from the interaction with this j atom. */
 502             velecsum         = _mm256_add_ps(velecsum,velec);
 503
 504             fscal            = felec;
 505
 506             /* Calculate temporary vectorial force */
 507             tx               = _mm256_mul_ps(fscal,dx21);
 508             ty               = _mm256_mul_ps(fscal,dy21);
 509             tz               = _mm256_mul_ps(fscal,dz21);
 510
 511             /* Update vectorial force */
 512             fix2             = _mm256_add_ps(fix2,tx);
 513             fiy2             = _mm256_add_ps(fiy2,ty);
 514             fiz2             = _mm256_add_ps(fiz2,tz);
 515
 516             fjx1             = _mm256_add_ps(fjx1,tx);
 517             fjy1             = _mm256_add_ps(fjy1,ty);
 518             fjz1             = _mm256_add_ps(fjz1,tz);
 519
 520             /**************************
 521              * CALCULATE INTERACTIONS *
 522              **************************/
 523
 524             /* COULOMB ELECTROSTATICS */
 525             velec            = _mm256_mul_ps(qq22,rinv22);
 526             felec            = _mm256_mul_ps(velec,rinvsq22);
 527
 528             /* Update potential sum for this i atom from the interaction with this j atom. */
 529             velecsum         = _mm256_add_ps(velecsum,velec);
 530
 531             fscal            = felec;
 532
 533             /* Calculate temporary vectorial force */
 534             tx               = _mm256_mul_ps(fscal,dx22);
 535             ty               = _mm256_mul_ps(fscal,dy22);
 536             tz               = _mm256_mul_ps(fscal,dz22);
 537
 538             /* Update vectorial force */
 539             fix2             = _mm256_add_ps(fix2,tx);
 540             fiy2             = _mm256_add_ps(fiy2,ty);
 541             fiz2             = _mm256_add_ps(fiz2,tz);
 542
 543             fjx2             = _mm256_add_ps(fjx2,tx);
 544             fjy2             = _mm256_add_ps(fjy2,ty);
 545             fjz2             = _mm256_add_ps(fjz2,tz);
 546
 547             fjptrA             = f+j_coord_offsetA;
 548             fjptrB             = f+j_coord_offsetB;
 549             fjptrC             = f+j_coord_offsetC;
 550             fjptrD             = f+j_coord_offsetD;
 551             fjptrE             = f+j_coord_offsetE;
 552             fjptrF             = f+j_coord_offsetF;
 553             fjptrG             = f+j_coord_offsetG;
 554             fjptrH             = f+j_coord_offsetH;
 555
 556             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 557                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 558
 559             /* Inner loop uses 255 flops */
 560         }
 561
 562         if(jidx<j_index_end)
 563         {
 564
 565             /* Get j neighbor index, and coordinate index */
 566             jnrlistA         = jjnr[jidx];
 567             jnrlistB         = jjnr[jidx+1];
 568             jnrlistC         = jjnr[jidx+2];
 569             jnrlistD         = jjnr[jidx+3];
 570             jnrlistE         = jjnr[jidx+4];
 571             jnrlistF         = jjnr[jidx+5];
 572             jnrlistG         = jjnr[jidx+6];
 573             jnrlistH         = jjnr[jidx+7];
 574             /* Sign of each element will be negative for non-real atoms.
 575              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 576              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 577              */
 578             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 579                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 580
 581             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 582             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 583             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 584             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 585             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 586             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 587             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 588             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 589             j_coord_offsetA  = DIM*jnrA;
 590             j_coord_offsetB  = DIM*jnrB;
 591             j_coord_offsetC  = DIM*jnrC;
 592             j_coord_offsetD  = DIM*jnrD;
 593             j_coord_offsetE  = DIM*jnrE;
 594             j_coord_offsetF  = DIM*jnrF;
 595             j_coord_offsetG  = DIM*jnrG;
 596             j_coord_offsetH  = DIM*jnrH;
 597
 598             /* load j atom coordinates */
 599             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 600                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 601                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 602                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 603                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 604
 605             /* Calculate displacement vector */
 606             dx00             = _mm256_sub_ps(ix0,jx0);
 607             dy00             = _mm256_sub_ps(iy0,jy0);
 608             dz00             = _mm256_sub_ps(iz0,jz0);
 609             dx01             = _mm256_sub_ps(ix0,jx1);
 610             dy01             = _mm256_sub_ps(iy0,jy1);
 611             dz01             = _mm256_sub_ps(iz0,jz1);
 612             dx02             = _mm256_sub_ps(ix0,jx2);
 613             dy02             = _mm256_sub_ps(iy0,jy2);
 614             dz02             = _mm256_sub_ps(iz0,jz2);
 615             dx10             = _mm256_sub_ps(ix1,jx0);
 616             dy10             = _mm256_sub_ps(iy1,jy0);
 617             dz10             = _mm256_sub_ps(iz1,jz0);
 618             dx11             = _mm256_sub_ps(ix1,jx1);
 619             dy11             = _mm256_sub_ps(iy1,jy1);
 620             dz11             = _mm256_sub_ps(iz1,jz1);
 621             dx12             = _mm256_sub_ps(ix1,jx2);
 622             dy12             = _mm256_sub_ps(iy1,jy2);
 623             dz12             = _mm256_sub_ps(iz1,jz2);
 624             dx20             = _mm256_sub_ps(ix2,jx0);
 625             dy20             = _mm256_sub_ps(iy2,jy0);
 626             dz20             = _mm256_sub_ps(iz2,jz0);
 627             dx21             = _mm256_sub_ps(ix2,jx1);
 628             dy21             = _mm256_sub_ps(iy2,jy1);
 629             dz21             = _mm256_sub_ps(iz2,jz1);
 630             dx22             = _mm256_sub_ps(ix2,jx2);
 631             dy22             = _mm256_sub_ps(iy2,jy2);
 632             dz22             = _mm256_sub_ps(iz2,jz2);
 633
 634             /* Calculate squared distance and things based on it */
 635             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 636             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 637             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 638             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 639             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 640             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 641             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 642             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 643             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 644
 645             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 646             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 647             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 648             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 649             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 650             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 651             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 652             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 653             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 654
 655             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 656             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 657             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 658             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 659             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 660             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 661             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 662             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 663             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 664
 665             fjx0             = _mm256_setzero_ps();
 666             fjy0             = _mm256_setzero_ps();
 667             fjz0             = _mm256_setzero_ps();
 668             fjx1             = _mm256_setzero_ps();
 669             fjy1             = _mm256_setzero_ps();
 670             fjz1             = _mm256_setzero_ps();
 671             fjx2             = _mm256_setzero_ps();
 672             fjy2             = _mm256_setzero_ps();
 673             fjz2             = _mm256_setzero_ps();
 674
 675             /**************************
 676              * CALCULATE INTERACTIONS *
 677              **************************/
 678
 679             /* COULOMB ELECTROSTATICS */
 680             velec            = _mm256_mul_ps(qq00,rinv00);
 681             felec            = _mm256_mul_ps(velec,rinvsq00);
 682
 683             /* LENNARD-JONES DISPERSION/REPULSION */
 684
 685             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 686             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 687             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 688             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 689             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 690
 691             /* Update potential sum for this i atom from the interaction with this j atom. */
 692             velec            = _mm256_andnot_ps(dummy_mask,velec);
 693             velecsum         = _mm256_add_ps(velecsum,velec);
 694             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 695             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 696
 697             fscal            = _mm256_add_ps(felec,fvdw);
 698
 699             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 700
 701             /* Calculate temporary vectorial force */
 702             tx               = _mm256_mul_ps(fscal,dx00);
 703             ty               = _mm256_mul_ps(fscal,dy00);
 704             tz               = _mm256_mul_ps(fscal,dz00);
 705
 706             /* Update vectorial force */
 707             fix0             = _mm256_add_ps(fix0,tx);
 708             fiy0             = _mm256_add_ps(fiy0,ty);
 709             fiz0             = _mm256_add_ps(fiz0,tz);
 710
 711             fjx0             = _mm256_add_ps(fjx0,tx);
 712             fjy0             = _mm256_add_ps(fjy0,ty);
 713             fjz0             = _mm256_add_ps(fjz0,tz);
 714
 715             /**************************
 716              * CALCULATE INTERACTIONS *
 717              **************************/
 718
 719             /* COULOMB ELECTROSTATICS */
 720             velec            = _mm256_mul_ps(qq01,rinv01);
 721             felec            = _mm256_mul_ps(velec,rinvsq01);
 722
 723             /* Update potential sum for this i atom from the interaction with this j atom. */
 724             velec            = _mm256_andnot_ps(dummy_mask,velec);
 725             velecsum         = _mm256_add_ps(velecsum,velec);
 726
 727             fscal            = felec;
 728
 729             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 730
 731             /* Calculate temporary vectorial force */
 732             tx               = _mm256_mul_ps(fscal,dx01);
 733             ty               = _mm256_mul_ps(fscal,dy01);
 734             tz               = _mm256_mul_ps(fscal,dz01);
 735
 736             /* Update vectorial force */
 737             fix0             = _mm256_add_ps(fix0,tx);
 738             fiy0             = _mm256_add_ps(fiy0,ty);
 739             fiz0             = _mm256_add_ps(fiz0,tz);
 740
 741             fjx1             = _mm256_add_ps(fjx1,tx);
 742             fjy1             = _mm256_add_ps(fjy1,ty);
 743             fjz1             = _mm256_add_ps(fjz1,tz);
 744
 745             /**************************
 746              * CALCULATE INTERACTIONS *
 747              **************************/
 748
 749             /* COULOMB ELECTROSTATICS */
 750             velec            = _mm256_mul_ps(qq02,rinv02);
 751             felec            = _mm256_mul_ps(velec,rinvsq02);
 752
 753             /* Update potential sum for this i atom from the interaction with this j atom. */
 754             velec            = _mm256_andnot_ps(dummy_mask,velec);
 755             velecsum         = _mm256_add_ps(velecsum,velec);
 756
 757             fscal            = felec;
 758
 759             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 760
 761             /* Calculate temporary vectorial force */
 762             tx               = _mm256_mul_ps(fscal,dx02);
 763             ty               = _mm256_mul_ps(fscal,dy02);
 764             tz               = _mm256_mul_ps(fscal,dz02);
 765
 766             /* Update vectorial force */
 767             fix0             = _mm256_add_ps(fix0,tx);
 768             fiy0             = _mm256_add_ps(fiy0,ty);
 769             fiz0             = _mm256_add_ps(fiz0,tz);
 770
 771             fjx2             = _mm256_add_ps(fjx2,tx);
 772             fjy2             = _mm256_add_ps(fjy2,ty);
 773             fjz2             = _mm256_add_ps(fjz2,tz);
 774
 775             /**************************
 776              * CALCULATE INTERACTIONS *
 777              **************************/
 778
 779             /* COULOMB ELECTROSTATICS */
 780             velec            = _mm256_mul_ps(qq10,rinv10);
 781             felec            = _mm256_mul_ps(velec,rinvsq10);
 782
 783             /* Update potential sum for this i atom from the interaction with this j atom. */
 784             velec            = _mm256_andnot_ps(dummy_mask,velec);
 785             velecsum         = _mm256_add_ps(velecsum,velec);
 786
 787             fscal            = felec;
 788
 789             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 790
 791             /* Calculate temporary vectorial force */
 792             tx               = _mm256_mul_ps(fscal,dx10);
 793             ty               = _mm256_mul_ps(fscal,dy10);
 794             tz               = _mm256_mul_ps(fscal,dz10);
 795
 796             /* Update vectorial force */
 797             fix1             = _mm256_add_ps(fix1,tx);
 798             fiy1             = _mm256_add_ps(fiy1,ty);
 799             fiz1             = _mm256_add_ps(fiz1,tz);
 800
 801             fjx0             = _mm256_add_ps(fjx0,tx);
 802             fjy0             = _mm256_add_ps(fjy0,ty);
 803             fjz0             = _mm256_add_ps(fjz0,tz);
 804
 805             /**************************
 806              * CALCULATE INTERACTIONS *
 807              **************************/
 808
 809             /* COULOMB ELECTROSTATICS */
 810             velec            = _mm256_mul_ps(qq11,rinv11);
 811             felec            = _mm256_mul_ps(velec,rinvsq11);
 812
 813             /* Update potential sum for this i atom from the interaction with this j atom. */
 814             velec            = _mm256_andnot_ps(dummy_mask,velec);
 815             velecsum         = _mm256_add_ps(velecsum,velec);
 816
 817             fscal            = felec;
 818
 819             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 820
 821             /* Calculate temporary vectorial force */
 822             tx               = _mm256_mul_ps(fscal,dx11);
 823             ty               = _mm256_mul_ps(fscal,dy11);
 824             tz               = _mm256_mul_ps(fscal,dz11);
 825
 826             /* Update vectorial force */
 827             fix1             = _mm256_add_ps(fix1,tx);
 828             fiy1             = _mm256_add_ps(fiy1,ty);
 829             fiz1             = _mm256_add_ps(fiz1,tz);
 830
 831             fjx1             = _mm256_add_ps(fjx1,tx);
 832             fjy1             = _mm256_add_ps(fjy1,ty);
 833             fjz1             = _mm256_add_ps(fjz1,tz);
 834
 835             /**************************
 836              * CALCULATE INTERACTIONS *
 837              **************************/
 838
 839             /* COULOMB ELECTROSTATICS */
 840             velec            = _mm256_mul_ps(qq12,rinv12);
 841             felec            = _mm256_mul_ps(velec,rinvsq12);
 842
 843             /* Update potential sum for this i atom from the interaction with this j atom. */
 844             velec            = _mm256_andnot_ps(dummy_mask,velec);
 845             velecsum         = _mm256_add_ps(velecsum,velec);
 846
 847             fscal            = felec;
 848
 849             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 850
 851             /* Calculate temporary vectorial force */
 852             tx               = _mm256_mul_ps(fscal,dx12);
 853             ty               = _mm256_mul_ps(fscal,dy12);
 854             tz               = _mm256_mul_ps(fscal,dz12);
 855
 856             /* Update vectorial force */
 857             fix1             = _mm256_add_ps(fix1,tx);
 858             fiy1             = _mm256_add_ps(fiy1,ty);
 859             fiz1             = _mm256_add_ps(fiz1,tz);
 860
 861             fjx2             = _mm256_add_ps(fjx2,tx);
 862             fjy2             = _mm256_add_ps(fjy2,ty);
 863             fjz2             = _mm256_add_ps(fjz2,tz);
 864
 865             /**************************
 866              * CALCULATE INTERACTIONS *
 867              **************************/
 868
 869             /* COULOMB ELECTROSTATICS */
 870             velec            = _mm256_mul_ps(qq20,rinv20);
 871             felec            = _mm256_mul_ps(velec,rinvsq20);
 872
 873             /* Update potential sum for this i atom from the interaction with this j atom. */
 874             velec            = _mm256_andnot_ps(dummy_mask,velec);
 875             velecsum         = _mm256_add_ps(velecsum,velec);
 876
 877             fscal            = felec;
 878
 879             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 880
 881             /* Calculate temporary vectorial force */
 882             tx               = _mm256_mul_ps(fscal,dx20);
 883             ty               = _mm256_mul_ps(fscal,dy20);
 884             tz               = _mm256_mul_ps(fscal,dz20);
 885
 886             /* Update vectorial force */
 887             fix2             = _mm256_add_ps(fix2,tx);
 888             fiy2             = _mm256_add_ps(fiy2,ty);
 889             fiz2             = _mm256_add_ps(fiz2,tz);
 890
 891             fjx0             = _mm256_add_ps(fjx0,tx);
 892             fjy0             = _mm256_add_ps(fjy0,ty);
 893             fjz0             = _mm256_add_ps(fjz0,tz);
 894
 895             /**************************
 896              * CALCULATE INTERACTIONS *
 897              **************************/
 898
 899             /* COULOMB ELECTROSTATICS */
 900             velec            = _mm256_mul_ps(qq21,rinv21);
 901             felec            = _mm256_mul_ps(velec,rinvsq21);
 902
 903             /* Update potential sum for this i atom from the interaction with this j atom. */
 904             velec            = _mm256_andnot_ps(dummy_mask,velec);
 905             velecsum         = _mm256_add_ps(velecsum,velec);
 906
 907             fscal            = felec;
 908
 909             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 910
 911             /* Calculate temporary vectorial force */
 912             tx               = _mm256_mul_ps(fscal,dx21);
 913             ty               = _mm256_mul_ps(fscal,dy21);
 914             tz               = _mm256_mul_ps(fscal,dz21);
 915
 916             /* Update vectorial force */
 917             fix2             = _mm256_add_ps(fix2,tx);
 918             fiy2             = _mm256_add_ps(fiy2,ty);
 919             fiz2             = _mm256_add_ps(fiz2,tz);
 920
 921             fjx1             = _mm256_add_ps(fjx1,tx);
 922             fjy1             = _mm256_add_ps(fjy1,ty);
 923             fjz1             = _mm256_add_ps(fjz1,tz);
 924
 925             /**************************
 926              * CALCULATE INTERACTIONS *
 927              **************************/
 928
 929             /* COULOMB ELECTROSTATICS */
 930             velec            = _mm256_mul_ps(qq22,rinv22);
 931             felec            = _mm256_mul_ps(velec,rinvsq22);
 932
 933             /* Update potential sum for this i atom from the interaction with this j atom. */
 934             velec            = _mm256_andnot_ps(dummy_mask,velec);
 935             velecsum         = _mm256_add_ps(velecsum,velec);
 936
 937             fscal            = felec;
 938
 939             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 940
 941             /* Calculate temporary vectorial force */
 942             tx               = _mm256_mul_ps(fscal,dx22);
 943             ty               = _mm256_mul_ps(fscal,dy22);
 944             tz               = _mm256_mul_ps(fscal,dz22);
 945
 946             /* Update vectorial force */
 947             fix2             = _mm256_add_ps(fix2,tx);
 948             fiy2             = _mm256_add_ps(fiy2,ty);
 949             fiz2             = _mm256_add_ps(fiz2,tz);
 950
 951             fjx2             = _mm256_add_ps(fjx2,tx);
 952             fjy2             = _mm256_add_ps(fjy2,ty);
 953             fjz2             = _mm256_add_ps(fjz2,tz);
 954
 955             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 956             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 957             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 958             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 959             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 960             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 961             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 962             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 963
 964             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 965                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 966
 967             /* Inner loop uses 255 flops */
 968         }
 969
 970         /* End of innermost loop */
 971
 972         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 973                                                  f+i_coord_offset,fshift+i_shift_offset);
 974
 975         ggid                        = gid[iidx];
 976         /* Update potential energies */
 977         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 978         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 979
 980         /* Increment number of inner iterations */
 981         inneriter                  += j_index_end - j_index_start;
 982
 983         /* Outer loop uses 20 flops */
 984     }
 985
 986     /* Increment number of outer iterations */
 987     outeriter        += nri;
 988
 989     /* Update outer/inner flops */
 990
 991     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*255);
 992 }
 993 /*
 994  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW3W3_F_avx_256_single
 995  * Electrostatics interaction: Coulomb
 996  * VdW interaction:            LennardJones
 997  * Geometry:                   Water3-Water3
 998  * Calculate force/pot:        Force
 999  */
1000 void
1001 nb_kernel_ElecCoul_VdwLJ_GeomW3W3_F_avx_256_single
1002                     (t_nblist * gmx_restrict                nlist,
1003                      rvec * gmx_restrict                    xx,
1004                      rvec * gmx_restrict                    ff,
1005                      t_forcerec * gmx_restrict              fr,
1006                      t_mdatoms * gmx_restrict               mdatoms,
1007                      nb_kernel_data_t * gmx_restrict        kernel_data,
1008                      t_nrnb * gmx_restrict                  nrnb)
1009 {
1010     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1011      * just 0 for non-waters.
1012      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1013      * jnr indices corresponding to data put in the four positions in the SIMD register.
1014      */
1015     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1016     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1017     int              jnrA,jnrB,jnrC,jnrD;
1018     int              jnrE,jnrF,jnrG,jnrH;
1019     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1020     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1021     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1022     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1023     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1024     real             rcutoff_scalar;
1025     real             *shiftvec,*fshift,*x,*f;
1026     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1027     real             scratch[4*DIM];
1028     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1029     real *           vdwioffsetptr0;
1030     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1031     real *           vdwioffsetptr1;
1032     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1033     real *           vdwioffsetptr2;
1034     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1035     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1036     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1037     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1038     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1039     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1040     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1041     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1042     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1043     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1044     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1045     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1046     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1047     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1048     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1049     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1050     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1051     real             *charge;
1052     int              nvdwtype;
1053     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1054     int              *vdwtype;
1055     real             *vdwparam;
1056     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1057     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1058     __m256           dummy_mask,cutoff_mask;
1059     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1060     __m256           one     = _mm256_set1_ps(1.0);
1061     __m256           two     = _mm256_set1_ps(2.0);
1062     x                = xx[0];
1063     f                = ff[0];
1064
1065     nri              = nlist->nri;
1066     iinr             = nlist->iinr;
1067     jindex           = nlist->jindex;
1068     jjnr             = nlist->jjnr;
1069     shiftidx         = nlist->shift;
1070     gid              = nlist->gid;
1071     shiftvec         = fr->shift_vec[0];
1072     fshift           = fr->fshift[0];
1073     facel            = _mm256_set1_ps(fr->epsfac);
1074     charge           = mdatoms->chargeA;
1075     nvdwtype         = fr->ntype;
1076     vdwparam         = fr->nbfp;
1077     vdwtype          = mdatoms->typeA;
1078
1079     /* Setup water-specific parameters */
1080     inr              = nlist->iinr[0];
1081     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1082     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1083     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1084     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1085
1086     jq0              = _mm256_set1_ps(charge[inr+0]);
1087     jq1              = _mm256_set1_ps(charge[inr+1]);
1088     jq2              = _mm256_set1_ps(charge[inr+2]);
1089     vdwjidx0A        = 2*vdwtype[inr+0];
1090     qq00             = _mm256_mul_ps(iq0,jq0);
1091     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1092     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1093     qq01             = _mm256_mul_ps(iq0,jq1);
1094     qq02             = _mm256_mul_ps(iq0,jq2);
1095     qq10             = _mm256_mul_ps(iq1,jq0);
1096     qq11             = _mm256_mul_ps(iq1,jq1);
1097     qq12             = _mm256_mul_ps(iq1,jq2);
1098     qq20             = _mm256_mul_ps(iq2,jq0);
1099     qq21             = _mm256_mul_ps(iq2,jq1);
1100     qq22             = _mm256_mul_ps(iq2,jq2);
1101
1102     /* Avoid stupid compiler warnings */
1103     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1104     j_coord_offsetA = 0;
1105     j_coord_offsetB = 0;
1106     j_coord_offsetC = 0;
1107     j_coord_offsetD = 0;
1108     j_coord_offsetE = 0;
1109     j_coord_offsetF = 0;
1110     j_coord_offsetG = 0;
1111     j_coord_offsetH = 0;
1112
1113     outeriter        = 0;
1114     inneriter        = 0;
1115
1116     for(iidx=0;iidx<4*DIM;iidx++)
1117     {
1118         scratch[iidx] = 0.0;
1119     }
1120
1121     /* Start outer loop over neighborlists */
1122     for(iidx=0; iidx<nri; iidx++)
1123     {
1124         /* Load shift vector for this list */
1125         i_shift_offset   = DIM*shiftidx[iidx];
1126
1127         /* Load limits for loop over neighbors */
1128         j_index_start    = jindex[iidx];
1129         j_index_end      = jindex[iidx+1];
1130
1131         /* Get outer coordinate index */
1132         inr              = iinr[iidx];
1133         i_coord_offset   = DIM*inr;
1134
1135         /* Load i particle coords and add shift vector */
1136         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1137                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1138
1139         fix0             = _mm256_setzero_ps();
1140         fiy0             = _mm256_setzero_ps();
1141         fiz0             = _mm256_setzero_ps();
1142         fix1             = _mm256_setzero_ps();
1143         fiy1             = _mm256_setzero_ps();
1144         fiz1             = _mm256_setzero_ps();
1145         fix2             = _mm256_setzero_ps();
1146         fiy2             = _mm256_setzero_ps();
1147         fiz2             = _mm256_setzero_ps();
1148
1149         /* Start inner kernel loop */
1150         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1151         {
1152
1153             /* Get j neighbor index, and coordinate index */
1154             jnrA             = jjnr[jidx];
1155             jnrB             = jjnr[jidx+1];
1156             jnrC             = jjnr[jidx+2];
1157             jnrD             = jjnr[jidx+3];
1158             jnrE             = jjnr[jidx+4];
1159             jnrF             = jjnr[jidx+5];
1160             jnrG             = jjnr[jidx+6];
1161             jnrH             = jjnr[jidx+7];
1162             j_coord_offsetA  = DIM*jnrA;
1163             j_coord_offsetB  = DIM*jnrB;
1164             j_coord_offsetC  = DIM*jnrC;
1165             j_coord_offsetD  = DIM*jnrD;
1166             j_coord_offsetE  = DIM*jnrE;
1167             j_coord_offsetF  = DIM*jnrF;
1168             j_coord_offsetG  = DIM*jnrG;
1169             j_coord_offsetH  = DIM*jnrH;
1170
1171             /* load j atom coordinates */
1172             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1173                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1174                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1175                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1176                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1177
1178             /* Calculate displacement vector */
1179             dx00             = _mm256_sub_ps(ix0,jx0);
1180             dy00             = _mm256_sub_ps(iy0,jy0);
1181             dz00             = _mm256_sub_ps(iz0,jz0);
1182             dx01             = _mm256_sub_ps(ix0,jx1);
1183             dy01             = _mm256_sub_ps(iy0,jy1);
1184             dz01             = _mm256_sub_ps(iz0,jz1);
1185             dx02             = _mm256_sub_ps(ix0,jx2);
1186             dy02             = _mm256_sub_ps(iy0,jy2);
1187             dz02             = _mm256_sub_ps(iz0,jz2);
1188             dx10             = _mm256_sub_ps(ix1,jx0);
1189             dy10             = _mm256_sub_ps(iy1,jy0);
1190             dz10             = _mm256_sub_ps(iz1,jz0);
1191             dx11             = _mm256_sub_ps(ix1,jx1);
1192             dy11             = _mm256_sub_ps(iy1,jy1);
1193             dz11             = _mm256_sub_ps(iz1,jz1);
1194             dx12             = _mm256_sub_ps(ix1,jx2);
1195             dy12             = _mm256_sub_ps(iy1,jy2);
1196             dz12             = _mm256_sub_ps(iz1,jz2);
1197             dx20             = _mm256_sub_ps(ix2,jx0);
1198             dy20             = _mm256_sub_ps(iy2,jy0);
1199             dz20             = _mm256_sub_ps(iz2,jz0);
1200             dx21             = _mm256_sub_ps(ix2,jx1);
1201             dy21             = _mm256_sub_ps(iy2,jy1);
1202             dz21             = _mm256_sub_ps(iz2,jz1);
1203             dx22             = _mm256_sub_ps(ix2,jx2);
1204             dy22             = _mm256_sub_ps(iy2,jy2);
1205             dz22             = _mm256_sub_ps(iz2,jz2);
1206
1207             /* Calculate squared distance and things based on it */
1208             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1209             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1210             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1211             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1212             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1213             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1214             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1215             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1216             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1217
1218             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1219             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1220             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1221             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1222             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1223             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1224             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1225             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1226             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1227
1228             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1229             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1230             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1231             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1232             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1233             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1234             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1235             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1236             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1237
1238             fjx0             = _mm256_setzero_ps();
1239             fjy0             = _mm256_setzero_ps();
1240             fjz0             = _mm256_setzero_ps();
1241             fjx1             = _mm256_setzero_ps();
1242             fjy1             = _mm256_setzero_ps();
1243             fjz1             = _mm256_setzero_ps();
1244             fjx2             = _mm256_setzero_ps();
1245             fjy2             = _mm256_setzero_ps();
1246             fjz2             = _mm256_setzero_ps();
1247
1248             /**************************
1249              * CALCULATE INTERACTIONS *
1250              **************************/
1251
1252             /* COULOMB ELECTROSTATICS */
1253             velec            = _mm256_mul_ps(qq00,rinv00);
1254             felec            = _mm256_mul_ps(velec,rinvsq00);
1255
1256             /* LENNARD-JONES DISPERSION/REPULSION */
1257
1258             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1259             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1260
1261             fscal            = _mm256_add_ps(felec,fvdw);
1262
1263             /* Calculate temporary vectorial force */
1264             tx               = _mm256_mul_ps(fscal,dx00);
1265             ty               = _mm256_mul_ps(fscal,dy00);
1266             tz               = _mm256_mul_ps(fscal,dz00);
1267
1268             /* Update vectorial force */
1269             fix0             = _mm256_add_ps(fix0,tx);
1270             fiy0             = _mm256_add_ps(fiy0,ty);
1271             fiz0             = _mm256_add_ps(fiz0,tz);
1272
1273             fjx0             = _mm256_add_ps(fjx0,tx);
1274             fjy0             = _mm256_add_ps(fjy0,ty);
1275             fjz0             = _mm256_add_ps(fjz0,tz);
1276
1277             /**************************
1278              * CALCULATE INTERACTIONS *
1279              **************************/
1280
1281             /* COULOMB ELECTROSTATICS */
1282             velec            = _mm256_mul_ps(qq01,rinv01);
1283             felec            = _mm256_mul_ps(velec,rinvsq01);
1284
1285             fscal            = felec;
1286
1287             /* Calculate temporary vectorial force */
1288             tx               = _mm256_mul_ps(fscal,dx01);
1289             ty               = _mm256_mul_ps(fscal,dy01);
1290             tz               = _mm256_mul_ps(fscal,dz01);
1291
1292             /* Update vectorial force */
1293             fix0             = _mm256_add_ps(fix0,tx);
1294             fiy0             = _mm256_add_ps(fiy0,ty);
1295             fiz0             = _mm256_add_ps(fiz0,tz);
1296
1297             fjx1             = _mm256_add_ps(fjx1,tx);
1298             fjy1             = _mm256_add_ps(fjy1,ty);
1299             fjz1             = _mm256_add_ps(fjz1,tz);
1300
1301             /**************************
1302              * CALCULATE INTERACTIONS *
1303              **************************/
1304
1305             /* COULOMB ELECTROSTATICS */
1306             velec            = _mm256_mul_ps(qq02,rinv02);
1307             felec            = _mm256_mul_ps(velec,rinvsq02);
1308
1309             fscal            = felec;
1310
1311             /* Calculate temporary vectorial force */
1312             tx               = _mm256_mul_ps(fscal,dx02);
1313             ty               = _mm256_mul_ps(fscal,dy02);
1314             tz               = _mm256_mul_ps(fscal,dz02);
1315
1316             /* Update vectorial force */
1317             fix0             = _mm256_add_ps(fix0,tx);
1318             fiy0             = _mm256_add_ps(fiy0,ty);
1319             fiz0             = _mm256_add_ps(fiz0,tz);
1320
1321             fjx2             = _mm256_add_ps(fjx2,tx);
1322             fjy2             = _mm256_add_ps(fjy2,ty);
1323             fjz2             = _mm256_add_ps(fjz2,tz);
1324
1325             /**************************
1326              * CALCULATE INTERACTIONS *
1327              **************************/
1328
1329             /* COULOMB ELECTROSTATICS */
1330             velec            = _mm256_mul_ps(qq10,rinv10);
1331             felec            = _mm256_mul_ps(velec,rinvsq10);
1332
1333             fscal            = felec;
1334
1335             /* Calculate temporary vectorial force */
1336             tx               = _mm256_mul_ps(fscal,dx10);
1337             ty               = _mm256_mul_ps(fscal,dy10);
1338             tz               = _mm256_mul_ps(fscal,dz10);
1339
1340             /* Update vectorial force */
1341             fix1             = _mm256_add_ps(fix1,tx);
1342             fiy1             = _mm256_add_ps(fiy1,ty);
1343             fiz1             = _mm256_add_ps(fiz1,tz);
1344
1345             fjx0             = _mm256_add_ps(fjx0,tx);
1346             fjy0             = _mm256_add_ps(fjy0,ty);
1347             fjz0             = _mm256_add_ps(fjz0,tz);
1348
1349             /**************************
1350              * CALCULATE INTERACTIONS *
1351              **************************/
1352
1353             /* COULOMB ELECTROSTATICS */
1354             velec            = _mm256_mul_ps(qq11,rinv11);
1355             felec            = _mm256_mul_ps(velec,rinvsq11);
1356
1357             fscal            = felec;
1358
1359             /* Calculate temporary vectorial force */
1360             tx               = _mm256_mul_ps(fscal,dx11);
1361             ty               = _mm256_mul_ps(fscal,dy11);
1362             tz               = _mm256_mul_ps(fscal,dz11);
1363
1364             /* Update vectorial force */
1365             fix1             = _mm256_add_ps(fix1,tx);
1366             fiy1             = _mm256_add_ps(fiy1,ty);
1367             fiz1             = _mm256_add_ps(fiz1,tz);
1368
1369             fjx1             = _mm256_add_ps(fjx1,tx);
1370             fjy1             = _mm256_add_ps(fjy1,ty);
1371             fjz1             = _mm256_add_ps(fjz1,tz);
1372
1373             /**************************
1374              * CALCULATE INTERACTIONS *
1375              **************************/
1376
1377             /* COULOMB ELECTROSTATICS */
1378             velec            = _mm256_mul_ps(qq12,rinv12);
1379             felec            = _mm256_mul_ps(velec,rinvsq12);
1380
1381             fscal            = felec;
1382
1383             /* Calculate temporary vectorial force */
1384             tx               = _mm256_mul_ps(fscal,dx12);
1385             ty               = _mm256_mul_ps(fscal,dy12);
1386             tz               = _mm256_mul_ps(fscal,dz12);
1387
1388             /* Update vectorial force */
1389             fix1             = _mm256_add_ps(fix1,tx);
1390             fiy1             = _mm256_add_ps(fiy1,ty);
1391             fiz1             = _mm256_add_ps(fiz1,tz);
1392
1393             fjx2             = _mm256_add_ps(fjx2,tx);
1394             fjy2             = _mm256_add_ps(fjy2,ty);
1395             fjz2             = _mm256_add_ps(fjz2,tz);
1396
1397             /**************************
1398              * CALCULATE INTERACTIONS *
1399              **************************/
1400
1401             /* COULOMB ELECTROSTATICS */
1402             velec            = _mm256_mul_ps(qq20,rinv20);
1403             felec            = _mm256_mul_ps(velec,rinvsq20);
1404
1405             fscal            = felec;
1406
1407             /* Calculate temporary vectorial force */
1408             tx               = _mm256_mul_ps(fscal,dx20);
1409             ty               = _mm256_mul_ps(fscal,dy20);
1410             tz               = _mm256_mul_ps(fscal,dz20);
1411
1412             /* Update vectorial force */
1413             fix2             = _mm256_add_ps(fix2,tx);
1414             fiy2             = _mm256_add_ps(fiy2,ty);
1415             fiz2             = _mm256_add_ps(fiz2,tz);
1416
1417             fjx0             = _mm256_add_ps(fjx0,tx);
1418             fjy0             = _mm256_add_ps(fjy0,ty);
1419             fjz0             = _mm256_add_ps(fjz0,tz);
1420
1421             /**************************
1422              * CALCULATE INTERACTIONS *
1423              **************************/
1424
1425             /* COULOMB ELECTROSTATICS */
1426             velec            = _mm256_mul_ps(qq21,rinv21);
1427             felec            = _mm256_mul_ps(velec,rinvsq21);
1428
1429             fscal            = felec;
1430
1431             /* Calculate temporary vectorial force */
1432             tx               = _mm256_mul_ps(fscal,dx21);
1433             ty               = _mm256_mul_ps(fscal,dy21);
1434             tz               = _mm256_mul_ps(fscal,dz21);
1435
1436             /* Update vectorial force */
1437             fix2             = _mm256_add_ps(fix2,tx);
1438             fiy2             = _mm256_add_ps(fiy2,ty);
1439             fiz2             = _mm256_add_ps(fiz2,tz);
1440
1441             fjx1             = _mm256_add_ps(fjx1,tx);
1442             fjy1             = _mm256_add_ps(fjy1,ty);
1443             fjz1             = _mm256_add_ps(fjz1,tz);
1444
1445             /**************************
1446              * CALCULATE INTERACTIONS *
1447              **************************/
1448
1449             /* COULOMB ELECTROSTATICS */
1450             velec            = _mm256_mul_ps(qq22,rinv22);
1451             felec            = _mm256_mul_ps(velec,rinvsq22);
1452
1453             fscal            = felec;
1454
1455             /* Calculate temporary vectorial force */
1456             tx               = _mm256_mul_ps(fscal,dx22);
1457             ty               = _mm256_mul_ps(fscal,dy22);
1458             tz               = _mm256_mul_ps(fscal,dz22);
1459
1460             /* Update vectorial force */
1461             fix2             = _mm256_add_ps(fix2,tx);
1462             fiy2             = _mm256_add_ps(fiy2,ty);
1463             fiz2             = _mm256_add_ps(fiz2,tz);
1464
1465             fjx2             = _mm256_add_ps(fjx2,tx);
1466             fjy2             = _mm256_add_ps(fjy2,ty);
1467             fjz2             = _mm256_add_ps(fjz2,tz);
1468
1469             fjptrA             = f+j_coord_offsetA;
1470             fjptrB             = f+j_coord_offsetB;
1471             fjptrC             = f+j_coord_offsetC;
1472             fjptrD             = f+j_coord_offsetD;
1473             fjptrE             = f+j_coord_offsetE;
1474             fjptrF             = f+j_coord_offsetF;
1475             fjptrG             = f+j_coord_offsetG;
1476             fjptrH             = f+j_coord_offsetH;
1477
1478             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1479                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1480
1481             /* Inner loop uses 241 flops */
1482         }
1483
1484         if(jidx<j_index_end)
1485         {
1486
1487             /* Get j neighbor index, and coordinate index */
1488             jnrlistA         = jjnr[jidx];
1489             jnrlistB         = jjnr[jidx+1];
1490             jnrlistC         = jjnr[jidx+2];
1491             jnrlistD         = jjnr[jidx+3];
1492             jnrlistE         = jjnr[jidx+4];
1493             jnrlistF         = jjnr[jidx+5];
1494             jnrlistG         = jjnr[jidx+6];
1495             jnrlistH         = jjnr[jidx+7];
1496             /* Sign of each element will be negative for non-real atoms.
1497              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1498              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1499              */
1500             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1501                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1502
1503             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1504             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1505             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1506             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1507             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1508             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1509             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1510             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1511             j_coord_offsetA  = DIM*jnrA;
1512             j_coord_offsetB  = DIM*jnrB;
1513             j_coord_offsetC  = DIM*jnrC;
1514             j_coord_offsetD  = DIM*jnrD;
1515             j_coord_offsetE  = DIM*jnrE;
1516             j_coord_offsetF  = DIM*jnrF;
1517             j_coord_offsetG  = DIM*jnrG;
1518             j_coord_offsetH  = DIM*jnrH;
1519
1520             /* load j atom coordinates */
1521             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1522                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1523                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1524                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1525                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1526
1527             /* Calculate displacement vector */
1528             dx00             = _mm256_sub_ps(ix0,jx0);
1529             dy00             = _mm256_sub_ps(iy0,jy0);
1530             dz00             = _mm256_sub_ps(iz0,jz0);
1531             dx01             = _mm256_sub_ps(ix0,jx1);
1532             dy01             = _mm256_sub_ps(iy0,jy1);
1533             dz01             = _mm256_sub_ps(iz0,jz1);
1534             dx02             = _mm256_sub_ps(ix0,jx2);
1535             dy02             = _mm256_sub_ps(iy0,jy2);
1536             dz02             = _mm256_sub_ps(iz0,jz2);
1537             dx10             = _mm256_sub_ps(ix1,jx0);
1538             dy10             = _mm256_sub_ps(iy1,jy0);
1539             dz10             = _mm256_sub_ps(iz1,jz0);
1540             dx11             = _mm256_sub_ps(ix1,jx1);
1541             dy11             = _mm256_sub_ps(iy1,jy1);
1542             dz11             = _mm256_sub_ps(iz1,jz1);
1543             dx12             = _mm256_sub_ps(ix1,jx2);
1544             dy12             = _mm256_sub_ps(iy1,jy2);
1545             dz12             = _mm256_sub_ps(iz1,jz2);
1546             dx20             = _mm256_sub_ps(ix2,jx0);
1547             dy20             = _mm256_sub_ps(iy2,jy0);
1548             dz20             = _mm256_sub_ps(iz2,jz0);
1549             dx21             = _mm256_sub_ps(ix2,jx1);
1550             dy21             = _mm256_sub_ps(iy2,jy1);
1551             dz21             = _mm256_sub_ps(iz2,jz1);
1552             dx22             = _mm256_sub_ps(ix2,jx2);
1553             dy22             = _mm256_sub_ps(iy2,jy2);
1554             dz22             = _mm256_sub_ps(iz2,jz2);
1555
1556             /* Calculate squared distance and things based on it */
1557             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1558             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1559             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1560             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1561             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1562             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1563             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1564             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1565             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1566
1567             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1568             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1569             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1570             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1571             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1572             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1573             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1574             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1575             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1576
1577             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1578             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1579             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1580             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1581             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1582             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1583             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1584             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1585             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1586
1587             fjx0             = _mm256_setzero_ps();
1588             fjy0             = _mm256_setzero_ps();
1589             fjz0             = _mm256_setzero_ps();
1590             fjx1             = _mm256_setzero_ps();
1591             fjy1             = _mm256_setzero_ps();
1592             fjz1             = _mm256_setzero_ps();
1593             fjx2             = _mm256_setzero_ps();
1594             fjy2             = _mm256_setzero_ps();
1595             fjz2             = _mm256_setzero_ps();
1596
1597             /**************************
1598              * CALCULATE INTERACTIONS *
1599              **************************/
1600
1601             /* COULOMB ELECTROSTATICS */
1602             velec            = _mm256_mul_ps(qq00,rinv00);
1603             felec            = _mm256_mul_ps(velec,rinvsq00);
1604
1605             /* LENNARD-JONES DISPERSION/REPULSION */
1606
1607             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1608             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1609
1610             fscal            = _mm256_add_ps(felec,fvdw);
1611
1612             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1613
1614             /* Calculate temporary vectorial force */
1615             tx               = _mm256_mul_ps(fscal,dx00);
1616             ty               = _mm256_mul_ps(fscal,dy00);
1617             tz               = _mm256_mul_ps(fscal,dz00);
1618
1619             /* Update vectorial force */
1620             fix0             = _mm256_add_ps(fix0,tx);
1621             fiy0             = _mm256_add_ps(fiy0,ty);
1622             fiz0             = _mm256_add_ps(fiz0,tz);
1623
1624             fjx0             = _mm256_add_ps(fjx0,tx);
1625             fjy0             = _mm256_add_ps(fjy0,ty);
1626             fjz0             = _mm256_add_ps(fjz0,tz);
1627
1628             /**************************
1629              * CALCULATE INTERACTIONS *
1630              **************************/
1631
1632             /* COULOMB ELECTROSTATICS */
1633             velec            = _mm256_mul_ps(qq01,rinv01);
1634             felec            = _mm256_mul_ps(velec,rinvsq01);
1635
1636             fscal            = felec;
1637
1638             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1639
1640             /* Calculate temporary vectorial force */
1641             tx               = _mm256_mul_ps(fscal,dx01);
1642             ty               = _mm256_mul_ps(fscal,dy01);
1643             tz               = _mm256_mul_ps(fscal,dz01);
1644
1645             /* Update vectorial force */
1646             fix0             = _mm256_add_ps(fix0,tx);
1647             fiy0             = _mm256_add_ps(fiy0,ty);
1648             fiz0             = _mm256_add_ps(fiz0,tz);
1649
1650             fjx1             = _mm256_add_ps(fjx1,tx);
1651             fjy1             = _mm256_add_ps(fjy1,ty);
1652             fjz1             = _mm256_add_ps(fjz1,tz);
1653
1654             /**************************
1655              * CALCULATE INTERACTIONS *
1656              **************************/
1657
1658             /* COULOMB ELECTROSTATICS */
1659             velec            = _mm256_mul_ps(qq02,rinv02);
1660             felec            = _mm256_mul_ps(velec,rinvsq02);
1661
1662             fscal            = felec;
1663
1664             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1665
1666             /* Calculate temporary vectorial force */
1667             tx               = _mm256_mul_ps(fscal,dx02);
1668             ty               = _mm256_mul_ps(fscal,dy02);
1669             tz               = _mm256_mul_ps(fscal,dz02);
1670
1671             /* Update vectorial force */
1672             fix0             = _mm256_add_ps(fix0,tx);
1673             fiy0             = _mm256_add_ps(fiy0,ty);
1674             fiz0             = _mm256_add_ps(fiz0,tz);
1675
1676             fjx2             = _mm256_add_ps(fjx2,tx);
1677             fjy2             = _mm256_add_ps(fjy2,ty);
1678             fjz2             = _mm256_add_ps(fjz2,tz);
1679
1680             /**************************
1681              * CALCULATE INTERACTIONS *
1682              **************************/
1683
1684             /* COULOMB ELECTROSTATICS */
1685             velec            = _mm256_mul_ps(qq10,rinv10);
1686             felec            = _mm256_mul_ps(velec,rinvsq10);
1687
1688             fscal            = felec;
1689
1690             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1691
1692             /* Calculate temporary vectorial force */
1693             tx               = _mm256_mul_ps(fscal,dx10);
1694             ty               = _mm256_mul_ps(fscal,dy10);
1695             tz               = _mm256_mul_ps(fscal,dz10);
1696
1697             /* Update vectorial force */
1698             fix1             = _mm256_add_ps(fix1,tx);
1699             fiy1             = _mm256_add_ps(fiy1,ty);
1700             fiz1             = _mm256_add_ps(fiz1,tz);
1701
1702             fjx0             = _mm256_add_ps(fjx0,tx);
1703             fjy0             = _mm256_add_ps(fjy0,ty);
1704             fjz0             = _mm256_add_ps(fjz0,tz);
1705
1706             /**************************
1707              * CALCULATE INTERACTIONS *
1708              **************************/
1709
1710             /* COULOMB ELECTROSTATICS */
1711             velec            = _mm256_mul_ps(qq11,rinv11);
1712             felec            = _mm256_mul_ps(velec,rinvsq11);
1713
1714             fscal            = felec;
1715
1716             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1717
1718             /* Calculate temporary vectorial force */
1719             tx               = _mm256_mul_ps(fscal,dx11);
1720             ty               = _mm256_mul_ps(fscal,dy11);
1721             tz               = _mm256_mul_ps(fscal,dz11);
1722
1723             /* Update vectorial force */
1724             fix1             = _mm256_add_ps(fix1,tx);
1725             fiy1             = _mm256_add_ps(fiy1,ty);
1726             fiz1             = _mm256_add_ps(fiz1,tz);
1727
1728             fjx1             = _mm256_add_ps(fjx1,tx);
1729             fjy1             = _mm256_add_ps(fjy1,ty);
1730             fjz1             = _mm256_add_ps(fjz1,tz);
1731
1732             /**************************
1733              * CALCULATE INTERACTIONS *
1734              **************************/
1735
1736             /* COULOMB ELECTROSTATICS */
1737             velec            = _mm256_mul_ps(qq12,rinv12);
1738             felec            = _mm256_mul_ps(velec,rinvsq12);
1739
1740             fscal            = felec;
1741
1742             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1743
1744             /* Calculate temporary vectorial force */
1745             tx               = _mm256_mul_ps(fscal,dx12);
1746             ty               = _mm256_mul_ps(fscal,dy12);
1747             tz               = _mm256_mul_ps(fscal,dz12);
1748
1749             /* Update vectorial force */
1750             fix1             = _mm256_add_ps(fix1,tx);
1751             fiy1             = _mm256_add_ps(fiy1,ty);
1752             fiz1             = _mm256_add_ps(fiz1,tz);
1753
1754             fjx2             = _mm256_add_ps(fjx2,tx);
1755             fjy2             = _mm256_add_ps(fjy2,ty);
1756             fjz2             = _mm256_add_ps(fjz2,tz);
1757
1758             /**************************
1759              * CALCULATE INTERACTIONS *
1760              **************************/
1761
1762             /* COULOMB ELECTROSTATICS */
1763             velec            = _mm256_mul_ps(qq20,rinv20);
1764             felec            = _mm256_mul_ps(velec,rinvsq20);
1765
1766             fscal            = felec;
1767
1768             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1769
1770             /* Calculate temporary vectorial force */
1771             tx               = _mm256_mul_ps(fscal,dx20);
1772             ty               = _mm256_mul_ps(fscal,dy20);
1773             tz               = _mm256_mul_ps(fscal,dz20);
1774
1775             /* Update vectorial force */
1776             fix2             = _mm256_add_ps(fix2,tx);
1777             fiy2             = _mm256_add_ps(fiy2,ty);
1778             fiz2             = _mm256_add_ps(fiz2,tz);
1779
1780             fjx0             = _mm256_add_ps(fjx0,tx);
1781             fjy0             = _mm256_add_ps(fjy0,ty);
1782             fjz0             = _mm256_add_ps(fjz0,tz);
1783
1784             /**************************
1785              * CALCULATE INTERACTIONS *
1786              **************************/
1787
1788             /* COULOMB ELECTROSTATICS */
1789             velec            = _mm256_mul_ps(qq21,rinv21);
1790             felec            = _mm256_mul_ps(velec,rinvsq21);
1791
1792             fscal            = felec;
1793
1794             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1795
1796             /* Calculate temporary vectorial force */
1797             tx               = _mm256_mul_ps(fscal,dx21);
1798             ty               = _mm256_mul_ps(fscal,dy21);
1799             tz               = _mm256_mul_ps(fscal,dz21);
1800
1801             /* Update vectorial force */
1802             fix2             = _mm256_add_ps(fix2,tx);
1803             fiy2             = _mm256_add_ps(fiy2,ty);
1804             fiz2             = _mm256_add_ps(fiz2,tz);
1805
1806             fjx1             = _mm256_add_ps(fjx1,tx);
1807             fjy1             = _mm256_add_ps(fjy1,ty);
1808             fjz1             = _mm256_add_ps(fjz1,tz);
1809
1810             /**************************
1811              * CALCULATE INTERACTIONS *
1812              **************************/
1813
1814             /* COULOMB ELECTROSTATICS */
1815             velec            = _mm256_mul_ps(qq22,rinv22);
1816             felec            = _mm256_mul_ps(velec,rinvsq22);
1817
1818             fscal            = felec;
1819
1820             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1821
1822             /* Calculate temporary vectorial force */
1823             tx               = _mm256_mul_ps(fscal,dx22);
1824             ty               = _mm256_mul_ps(fscal,dy22);
1825             tz               = _mm256_mul_ps(fscal,dz22);
1826
1827             /* Update vectorial force */
1828             fix2             = _mm256_add_ps(fix2,tx);
1829             fiy2             = _mm256_add_ps(fiy2,ty);
1830             fiz2             = _mm256_add_ps(fiz2,tz);
1831
1832             fjx2             = _mm256_add_ps(fjx2,tx);
1833             fjy2             = _mm256_add_ps(fjy2,ty);
1834             fjz2             = _mm256_add_ps(fjz2,tz);
1835
1836             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1837             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1838             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1839             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1840             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1841             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1842             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1843             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1844
1845             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1846                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1847
1848             /* Inner loop uses 241 flops */
1849         }
1850
1851         /* End of innermost loop */
1852
1853         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1854                                                  f+i_coord_offset,fshift+i_shift_offset);
1855
1856         /* Increment number of inner iterations */
1857         inneriter                  += j_index_end - j_index_start;
1858
1859         /* Outer loop uses 18 flops */
1860     }
1861
1862     /* Increment number of outer iterations */
1863     outeriter        += nri;
1864
1865     /* Update outer/inner flops */
1866
1867     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*241);
1868 }