src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRF_VdwNone_GeomW4W4_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_ElecRF_VdwNone_GeomW4W4_VF_avx_256_single
  38  * Electrostatics interaction: ReactionField
  39  * VdW interaction:            None
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecRF_VdwNone_GeomW4W4_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 *           vdwioffsetptr1;
  73     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  74     real *           vdwioffsetptr2;
  75     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  76     real *           vdwioffsetptr3;
  77     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  78     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  79     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  80     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  81     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  82     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  83     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  84     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  85     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  86     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  87     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  88     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  89     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  90     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  91     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  92     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  93     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     __m256           dummy_mask,cutoff_mask;
  96     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
  97     __m256           one     = _mm256_set1_ps(1.0);
  98     __m256           two     = _mm256_set1_ps(2.0);
  99     x                = xx[0];
 100     f                = ff[0];
 101
 102     nri              = nlist->nri;
 103     iinr             = nlist->iinr;
 104     jindex           = nlist->jindex;
 105     jjnr             = nlist->jjnr;
 106     shiftidx         = nlist->shift;
 107     gid              = nlist->gid;
 108     shiftvec         = fr->shift_vec[0];
 109     fshift           = fr->fshift[0];
 110     facel            = _mm256_set1_ps(fr->epsfac);
 111     charge           = mdatoms->chargeA;
 112     krf              = _mm256_set1_ps(fr->ic->k_rf);
 113     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 114     crf              = _mm256_set1_ps(fr->ic->c_rf);
 115
 116     /* Setup water-specific parameters */
 117     inr              = nlist->iinr[0];
 118     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 119     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 120     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 121
 122     jq1              = _mm256_set1_ps(charge[inr+1]);
 123     jq2              = _mm256_set1_ps(charge[inr+2]);
 124     jq3              = _mm256_set1_ps(charge[inr+3]);
 125     qq11             = _mm256_mul_ps(iq1,jq1);
 126     qq12             = _mm256_mul_ps(iq1,jq2);
 127     qq13             = _mm256_mul_ps(iq1,jq3);
 128     qq21             = _mm256_mul_ps(iq2,jq1);
 129     qq22             = _mm256_mul_ps(iq2,jq2);
 130     qq23             = _mm256_mul_ps(iq2,jq3);
 131     qq31             = _mm256_mul_ps(iq3,jq1);
 132     qq32             = _mm256_mul_ps(iq3,jq2);
 133     qq33             = _mm256_mul_ps(iq3,jq3);
 134
 135     /* Avoid stupid compiler warnings */
 136     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 137     j_coord_offsetA = 0;
 138     j_coord_offsetB = 0;
 139     j_coord_offsetC = 0;
 140     j_coord_offsetD = 0;
 141     j_coord_offsetE = 0;
 142     j_coord_offsetF = 0;
 143     j_coord_offsetG = 0;
 144     j_coord_offsetH = 0;
 145
 146     outeriter        = 0;
 147     inneriter        = 0;
 148
 149     for(iidx=0;iidx<4*DIM;iidx++)
 150     {
 151         scratch[iidx] = 0.0;
 152     }
 153
 154     /* Start outer loop over neighborlists */
 155     for(iidx=0; iidx<nri; iidx++)
 156     {
 157         /* Load shift vector for this list */
 158         i_shift_offset   = DIM*shiftidx[iidx];
 159
 160         /* Load limits for loop over neighbors */
 161         j_index_start    = jindex[iidx];
 162         j_index_end      = jindex[iidx+1];
 163
 164         /* Get outer coordinate index */
 165         inr              = iinr[iidx];
 166         i_coord_offset   = DIM*inr;
 167
 168         /* Load i particle coords and add shift vector */
 169         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 170                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 171
 172         fix1             = _mm256_setzero_ps();
 173         fiy1             = _mm256_setzero_ps();
 174         fiz1             = _mm256_setzero_ps();
 175         fix2             = _mm256_setzero_ps();
 176         fiy2             = _mm256_setzero_ps();
 177         fiz2             = _mm256_setzero_ps();
 178         fix3             = _mm256_setzero_ps();
 179         fiy3             = _mm256_setzero_ps();
 180         fiz3             = _mm256_setzero_ps();
 181
 182         /* Reset potential sums */
 183         velecsum         = _mm256_setzero_ps();
 184
 185         /* Start inner kernel loop */
 186         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 187         {
 188
 189             /* Get j neighbor index, and coordinate index */
 190             jnrA             = jjnr[jidx];
 191             jnrB             = jjnr[jidx+1];
 192             jnrC             = jjnr[jidx+2];
 193             jnrD             = jjnr[jidx+3];
 194             jnrE             = jjnr[jidx+4];
 195             jnrF             = jjnr[jidx+5];
 196             jnrG             = jjnr[jidx+6];
 197             jnrH             = jjnr[jidx+7];
 198             j_coord_offsetA  = DIM*jnrA;
 199             j_coord_offsetB  = DIM*jnrB;
 200             j_coord_offsetC  = DIM*jnrC;
 201             j_coord_offsetD  = DIM*jnrD;
 202             j_coord_offsetE  = DIM*jnrE;
 203             j_coord_offsetF  = DIM*jnrF;
 204             j_coord_offsetG  = DIM*jnrG;
 205             j_coord_offsetH  = DIM*jnrH;
 206
 207             /* load j atom coordinates */
 208             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 209                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 210                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 211                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 212                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 213
 214             /* Calculate displacement vector */
 215             dx11             = _mm256_sub_ps(ix1,jx1);
 216             dy11             = _mm256_sub_ps(iy1,jy1);
 217             dz11             = _mm256_sub_ps(iz1,jz1);
 218             dx12             = _mm256_sub_ps(ix1,jx2);
 219             dy12             = _mm256_sub_ps(iy1,jy2);
 220             dz12             = _mm256_sub_ps(iz1,jz2);
 221             dx13             = _mm256_sub_ps(ix1,jx3);
 222             dy13             = _mm256_sub_ps(iy1,jy3);
 223             dz13             = _mm256_sub_ps(iz1,jz3);
 224             dx21             = _mm256_sub_ps(ix2,jx1);
 225             dy21             = _mm256_sub_ps(iy2,jy1);
 226             dz21             = _mm256_sub_ps(iz2,jz1);
 227             dx22             = _mm256_sub_ps(ix2,jx2);
 228             dy22             = _mm256_sub_ps(iy2,jy2);
 229             dz22             = _mm256_sub_ps(iz2,jz2);
 230             dx23             = _mm256_sub_ps(ix2,jx3);
 231             dy23             = _mm256_sub_ps(iy2,jy3);
 232             dz23             = _mm256_sub_ps(iz2,jz3);
 233             dx31             = _mm256_sub_ps(ix3,jx1);
 234             dy31             = _mm256_sub_ps(iy3,jy1);
 235             dz31             = _mm256_sub_ps(iz3,jz1);
 236             dx32             = _mm256_sub_ps(ix3,jx2);
 237             dy32             = _mm256_sub_ps(iy3,jy2);
 238             dz32             = _mm256_sub_ps(iz3,jz2);
 239             dx33             = _mm256_sub_ps(ix3,jx3);
 240             dy33             = _mm256_sub_ps(iy3,jy3);
 241             dz33             = _mm256_sub_ps(iz3,jz3);
 242
 243             /* Calculate squared distance and things based on it */
 244             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 245             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 246             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 247             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 248             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 249             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 250             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 251             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 252             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 253
 254             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 255             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 256             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 257             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 258             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 259             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 260             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 261             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 262             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 263
 264             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 265             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 266             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 267             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 268             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 269             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 270             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 271             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 272             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 273
 274             fjx1             = _mm256_setzero_ps();
 275             fjy1             = _mm256_setzero_ps();
 276             fjz1             = _mm256_setzero_ps();
 277             fjx2             = _mm256_setzero_ps();
 278             fjy2             = _mm256_setzero_ps();
 279             fjz2             = _mm256_setzero_ps();
 280             fjx3             = _mm256_setzero_ps();
 281             fjy3             = _mm256_setzero_ps();
 282             fjz3             = _mm256_setzero_ps();
 283
 284             /**************************
 285              * CALCULATE INTERACTIONS *
 286              **************************/
 287
 288             /* REACTION-FIELD ELECTROSTATICS */
 289             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 290             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 291
 292             /* Update potential sum for this i atom from the interaction with this j atom. */
 293             velecsum         = _mm256_add_ps(velecsum,velec);
 294
 295             fscal            = felec;
 296
 297             /* Calculate temporary vectorial force */
 298             tx               = _mm256_mul_ps(fscal,dx11);
 299             ty               = _mm256_mul_ps(fscal,dy11);
 300             tz               = _mm256_mul_ps(fscal,dz11);
 301
 302             /* Update vectorial force */
 303             fix1             = _mm256_add_ps(fix1,tx);
 304             fiy1             = _mm256_add_ps(fiy1,ty);
 305             fiz1             = _mm256_add_ps(fiz1,tz);
 306
 307             fjx1             = _mm256_add_ps(fjx1,tx);
 308             fjy1             = _mm256_add_ps(fjy1,ty);
 309             fjz1             = _mm256_add_ps(fjz1,tz);
 310
 311             /**************************
 312              * CALCULATE INTERACTIONS *
 313              **************************/
 314
 315             /* REACTION-FIELD ELECTROSTATICS */
 316             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 317             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 318
 319             /* Update potential sum for this i atom from the interaction with this j atom. */
 320             velecsum         = _mm256_add_ps(velecsum,velec);
 321
 322             fscal            = felec;
 323
 324             /* Calculate temporary vectorial force */
 325             tx               = _mm256_mul_ps(fscal,dx12);
 326             ty               = _mm256_mul_ps(fscal,dy12);
 327             tz               = _mm256_mul_ps(fscal,dz12);
 328
 329             /* Update vectorial force */
 330             fix1             = _mm256_add_ps(fix1,tx);
 331             fiy1             = _mm256_add_ps(fiy1,ty);
 332             fiz1             = _mm256_add_ps(fiz1,tz);
 333
 334             fjx2             = _mm256_add_ps(fjx2,tx);
 335             fjy2             = _mm256_add_ps(fjy2,ty);
 336             fjz2             = _mm256_add_ps(fjz2,tz);
 337
 338             /**************************
 339              * CALCULATE INTERACTIONS *
 340              **************************/
 341
 342             /* REACTION-FIELD ELECTROSTATICS */
 343             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 344             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 345
 346             /* Update potential sum for this i atom from the interaction with this j atom. */
 347             velecsum         = _mm256_add_ps(velecsum,velec);
 348
 349             fscal            = felec;
 350
 351             /* Calculate temporary vectorial force */
 352             tx               = _mm256_mul_ps(fscal,dx13);
 353             ty               = _mm256_mul_ps(fscal,dy13);
 354             tz               = _mm256_mul_ps(fscal,dz13);
 355
 356             /* Update vectorial force */
 357             fix1             = _mm256_add_ps(fix1,tx);
 358             fiy1             = _mm256_add_ps(fiy1,ty);
 359             fiz1             = _mm256_add_ps(fiz1,tz);
 360
 361             fjx3             = _mm256_add_ps(fjx3,tx);
 362             fjy3             = _mm256_add_ps(fjy3,ty);
 363             fjz3             = _mm256_add_ps(fjz3,tz);
 364
 365             /**************************
 366              * CALCULATE INTERACTIONS *
 367              **************************/
 368
 369             /* REACTION-FIELD ELECTROSTATICS */
 370             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 371             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 372
 373             /* Update potential sum for this i atom from the interaction with this j atom. */
 374             velecsum         = _mm256_add_ps(velecsum,velec);
 375
 376             fscal            = felec;
 377
 378             /* Calculate temporary vectorial force */
 379             tx               = _mm256_mul_ps(fscal,dx21);
 380             ty               = _mm256_mul_ps(fscal,dy21);
 381             tz               = _mm256_mul_ps(fscal,dz21);
 382
 383             /* Update vectorial force */
 384             fix2             = _mm256_add_ps(fix2,tx);
 385             fiy2             = _mm256_add_ps(fiy2,ty);
 386             fiz2             = _mm256_add_ps(fiz2,tz);
 387
 388             fjx1             = _mm256_add_ps(fjx1,tx);
 389             fjy1             = _mm256_add_ps(fjy1,ty);
 390             fjz1             = _mm256_add_ps(fjz1,tz);
 391
 392             /**************************
 393              * CALCULATE INTERACTIONS *
 394              **************************/
 395
 396             /* REACTION-FIELD ELECTROSTATICS */
 397             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 398             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 399
 400             /* Update potential sum for this i atom from the interaction with this j atom. */
 401             velecsum         = _mm256_add_ps(velecsum,velec);
 402
 403             fscal            = felec;
 404
 405             /* Calculate temporary vectorial force */
 406             tx               = _mm256_mul_ps(fscal,dx22);
 407             ty               = _mm256_mul_ps(fscal,dy22);
 408             tz               = _mm256_mul_ps(fscal,dz22);
 409
 410             /* Update vectorial force */
 411             fix2             = _mm256_add_ps(fix2,tx);
 412             fiy2             = _mm256_add_ps(fiy2,ty);
 413             fiz2             = _mm256_add_ps(fiz2,tz);
 414
 415             fjx2             = _mm256_add_ps(fjx2,tx);
 416             fjy2             = _mm256_add_ps(fjy2,ty);
 417             fjz2             = _mm256_add_ps(fjz2,tz);
 418
 419             /**************************
 420              * CALCULATE INTERACTIONS *
 421              **************************/
 422
 423             /* REACTION-FIELD ELECTROSTATICS */
 424             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 425             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 426
 427             /* Update potential sum for this i atom from the interaction with this j atom. */
 428             velecsum         = _mm256_add_ps(velecsum,velec);
 429
 430             fscal            = felec;
 431
 432             /* Calculate temporary vectorial force */
 433             tx               = _mm256_mul_ps(fscal,dx23);
 434             ty               = _mm256_mul_ps(fscal,dy23);
 435             tz               = _mm256_mul_ps(fscal,dz23);
 436
 437             /* Update vectorial force */
 438             fix2             = _mm256_add_ps(fix2,tx);
 439             fiy2             = _mm256_add_ps(fiy2,ty);
 440             fiz2             = _mm256_add_ps(fiz2,tz);
 441
 442             fjx3             = _mm256_add_ps(fjx3,tx);
 443             fjy3             = _mm256_add_ps(fjy3,ty);
 444             fjz3             = _mm256_add_ps(fjz3,tz);
 445
 446             /**************************
 447              * CALCULATE INTERACTIONS *
 448              **************************/
 449
 450             /* REACTION-FIELD ELECTROSTATICS */
 451             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 452             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 453
 454             /* Update potential sum for this i atom from the interaction with this j atom. */
 455             velecsum         = _mm256_add_ps(velecsum,velec);
 456
 457             fscal            = felec;
 458
 459             /* Calculate temporary vectorial force */
 460             tx               = _mm256_mul_ps(fscal,dx31);
 461             ty               = _mm256_mul_ps(fscal,dy31);
 462             tz               = _mm256_mul_ps(fscal,dz31);
 463
 464             /* Update vectorial force */
 465             fix3             = _mm256_add_ps(fix3,tx);
 466             fiy3             = _mm256_add_ps(fiy3,ty);
 467             fiz3             = _mm256_add_ps(fiz3,tz);
 468
 469             fjx1             = _mm256_add_ps(fjx1,tx);
 470             fjy1             = _mm256_add_ps(fjy1,ty);
 471             fjz1             = _mm256_add_ps(fjz1,tz);
 472
 473             /**************************
 474              * CALCULATE INTERACTIONS *
 475              **************************/
 476
 477             /* REACTION-FIELD ELECTROSTATICS */
 478             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 479             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 480
 481             /* Update potential sum for this i atom from the interaction with this j atom. */
 482             velecsum         = _mm256_add_ps(velecsum,velec);
 483
 484             fscal            = felec;
 485
 486             /* Calculate temporary vectorial force */
 487             tx               = _mm256_mul_ps(fscal,dx32);
 488             ty               = _mm256_mul_ps(fscal,dy32);
 489             tz               = _mm256_mul_ps(fscal,dz32);
 490
 491             /* Update vectorial force */
 492             fix3             = _mm256_add_ps(fix3,tx);
 493             fiy3             = _mm256_add_ps(fiy3,ty);
 494             fiz3             = _mm256_add_ps(fiz3,tz);
 495
 496             fjx2             = _mm256_add_ps(fjx2,tx);
 497             fjy2             = _mm256_add_ps(fjy2,ty);
 498             fjz2             = _mm256_add_ps(fjz2,tz);
 499
 500             /**************************
 501              * CALCULATE INTERACTIONS *
 502              **************************/
 503
 504             /* REACTION-FIELD ELECTROSTATICS */
 505             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 506             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 507
 508             /* Update potential sum for this i atom from the interaction with this j atom. */
 509             velecsum         = _mm256_add_ps(velecsum,velec);
 510
 511             fscal            = felec;
 512
 513             /* Calculate temporary vectorial force */
 514             tx               = _mm256_mul_ps(fscal,dx33);
 515             ty               = _mm256_mul_ps(fscal,dy33);
 516             tz               = _mm256_mul_ps(fscal,dz33);
 517
 518             /* Update vectorial force */
 519             fix3             = _mm256_add_ps(fix3,tx);
 520             fiy3             = _mm256_add_ps(fiy3,ty);
 521             fiz3             = _mm256_add_ps(fiz3,tz);
 522
 523             fjx3             = _mm256_add_ps(fjx3,tx);
 524             fjy3             = _mm256_add_ps(fjy3,ty);
 525             fjz3             = _mm256_add_ps(fjz3,tz);
 526
 527             fjptrA             = f+j_coord_offsetA;
 528             fjptrB             = f+j_coord_offsetB;
 529             fjptrC             = f+j_coord_offsetC;
 530             fjptrD             = f+j_coord_offsetD;
 531             fjptrE             = f+j_coord_offsetE;
 532             fjptrF             = f+j_coord_offsetF;
 533             fjptrG             = f+j_coord_offsetG;
 534             fjptrH             = f+j_coord_offsetH;
 535
 536             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 537                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 538                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 539
 540             /* Inner loop uses 288 flops */
 541         }
 542
 543         if(jidx<j_index_end)
 544         {
 545
 546             /* Get j neighbor index, and coordinate index */
 547             jnrlistA         = jjnr[jidx];
 548             jnrlistB         = jjnr[jidx+1];
 549             jnrlistC         = jjnr[jidx+2];
 550             jnrlistD         = jjnr[jidx+3];
 551             jnrlistE         = jjnr[jidx+4];
 552             jnrlistF         = jjnr[jidx+5];
 553             jnrlistG         = jjnr[jidx+6];
 554             jnrlistH         = jjnr[jidx+7];
 555             /* Sign of each element will be negative for non-real atoms.
 556              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 557              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 558              */
 559             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 560                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 561
 562             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 563             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 564             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 565             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 566             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 567             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 568             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 569             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 570             j_coord_offsetA  = DIM*jnrA;
 571             j_coord_offsetB  = DIM*jnrB;
 572             j_coord_offsetC  = DIM*jnrC;
 573             j_coord_offsetD  = DIM*jnrD;
 574             j_coord_offsetE  = DIM*jnrE;
 575             j_coord_offsetF  = DIM*jnrF;
 576             j_coord_offsetG  = DIM*jnrG;
 577             j_coord_offsetH  = DIM*jnrH;
 578
 579             /* load j atom coordinates */
 580             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 581                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 582                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 583                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 584                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 585
 586             /* Calculate displacement vector */
 587             dx11             = _mm256_sub_ps(ix1,jx1);
 588             dy11             = _mm256_sub_ps(iy1,jy1);
 589             dz11             = _mm256_sub_ps(iz1,jz1);
 590             dx12             = _mm256_sub_ps(ix1,jx2);
 591             dy12             = _mm256_sub_ps(iy1,jy2);
 592             dz12             = _mm256_sub_ps(iz1,jz2);
 593             dx13             = _mm256_sub_ps(ix1,jx3);
 594             dy13             = _mm256_sub_ps(iy1,jy3);
 595             dz13             = _mm256_sub_ps(iz1,jz3);
 596             dx21             = _mm256_sub_ps(ix2,jx1);
 597             dy21             = _mm256_sub_ps(iy2,jy1);
 598             dz21             = _mm256_sub_ps(iz2,jz1);
 599             dx22             = _mm256_sub_ps(ix2,jx2);
 600             dy22             = _mm256_sub_ps(iy2,jy2);
 601             dz22             = _mm256_sub_ps(iz2,jz2);
 602             dx23             = _mm256_sub_ps(ix2,jx3);
 603             dy23             = _mm256_sub_ps(iy2,jy3);
 604             dz23             = _mm256_sub_ps(iz2,jz3);
 605             dx31             = _mm256_sub_ps(ix3,jx1);
 606             dy31             = _mm256_sub_ps(iy3,jy1);
 607             dz31             = _mm256_sub_ps(iz3,jz1);
 608             dx32             = _mm256_sub_ps(ix3,jx2);
 609             dy32             = _mm256_sub_ps(iy3,jy2);
 610             dz32             = _mm256_sub_ps(iz3,jz2);
 611             dx33             = _mm256_sub_ps(ix3,jx3);
 612             dy33             = _mm256_sub_ps(iy3,jy3);
 613             dz33             = _mm256_sub_ps(iz3,jz3);
 614
 615             /* Calculate squared distance and things based on it */
 616             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 617             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 618             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 619             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 620             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 621             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 622             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 623             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 624             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 625
 626             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 627             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 628             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 629             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 630             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 631             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 632             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 633             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 634             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 635
 636             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 637             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 638             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 639             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 640             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 641             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 642             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 643             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 644             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 645
 646             fjx1             = _mm256_setzero_ps();
 647             fjy1             = _mm256_setzero_ps();
 648             fjz1             = _mm256_setzero_ps();
 649             fjx2             = _mm256_setzero_ps();
 650             fjy2             = _mm256_setzero_ps();
 651             fjz2             = _mm256_setzero_ps();
 652             fjx3             = _mm256_setzero_ps();
 653             fjy3             = _mm256_setzero_ps();
 654             fjz3             = _mm256_setzero_ps();
 655
 656             /**************************
 657              * CALCULATE INTERACTIONS *
 658              **************************/
 659
 660             /* REACTION-FIELD ELECTROSTATICS */
 661             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 662             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 663
 664             /* Update potential sum for this i atom from the interaction with this j atom. */
 665             velec            = _mm256_andnot_ps(dummy_mask,velec);
 666             velecsum         = _mm256_add_ps(velecsum,velec);
 667
 668             fscal            = felec;
 669
 670             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 671
 672             /* Calculate temporary vectorial force */
 673             tx               = _mm256_mul_ps(fscal,dx11);
 674             ty               = _mm256_mul_ps(fscal,dy11);
 675             tz               = _mm256_mul_ps(fscal,dz11);
 676
 677             /* Update vectorial force */
 678             fix1             = _mm256_add_ps(fix1,tx);
 679             fiy1             = _mm256_add_ps(fiy1,ty);
 680             fiz1             = _mm256_add_ps(fiz1,tz);
 681
 682             fjx1             = _mm256_add_ps(fjx1,tx);
 683             fjy1             = _mm256_add_ps(fjy1,ty);
 684             fjz1             = _mm256_add_ps(fjz1,tz);
 685
 686             /**************************
 687              * CALCULATE INTERACTIONS *
 688              **************************/
 689
 690             /* REACTION-FIELD ELECTROSTATICS */
 691             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 692             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 693
 694             /* Update potential sum for this i atom from the interaction with this j atom. */
 695             velec            = _mm256_andnot_ps(dummy_mask,velec);
 696             velecsum         = _mm256_add_ps(velecsum,velec);
 697
 698             fscal            = felec;
 699
 700             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 701
 702             /* Calculate temporary vectorial force */
 703             tx               = _mm256_mul_ps(fscal,dx12);
 704             ty               = _mm256_mul_ps(fscal,dy12);
 705             tz               = _mm256_mul_ps(fscal,dz12);
 706
 707             /* Update vectorial force */
 708             fix1             = _mm256_add_ps(fix1,tx);
 709             fiy1             = _mm256_add_ps(fiy1,ty);
 710             fiz1             = _mm256_add_ps(fiz1,tz);
 711
 712             fjx2             = _mm256_add_ps(fjx2,tx);
 713             fjy2             = _mm256_add_ps(fjy2,ty);
 714             fjz2             = _mm256_add_ps(fjz2,tz);
 715
 716             /**************************
 717              * CALCULATE INTERACTIONS *
 718              **************************/
 719
 720             /* REACTION-FIELD ELECTROSTATICS */
 721             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 722             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 723
 724             /* Update potential sum for this i atom from the interaction with this j atom. */
 725             velec            = _mm256_andnot_ps(dummy_mask,velec);
 726             velecsum         = _mm256_add_ps(velecsum,velec);
 727
 728             fscal            = felec;
 729
 730             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 731
 732             /* Calculate temporary vectorial force */
 733             tx               = _mm256_mul_ps(fscal,dx13);
 734             ty               = _mm256_mul_ps(fscal,dy13);
 735             tz               = _mm256_mul_ps(fscal,dz13);
 736
 737             /* Update vectorial force */
 738             fix1             = _mm256_add_ps(fix1,tx);
 739             fiy1             = _mm256_add_ps(fiy1,ty);
 740             fiz1             = _mm256_add_ps(fiz1,tz);
 741
 742             fjx3             = _mm256_add_ps(fjx3,tx);
 743             fjy3             = _mm256_add_ps(fjy3,ty);
 744             fjz3             = _mm256_add_ps(fjz3,tz);
 745
 746             /**************************
 747              * CALCULATE INTERACTIONS *
 748              **************************/
 749
 750             /* REACTION-FIELD ELECTROSTATICS */
 751             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 752             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 753
 754             /* Update potential sum for this i atom from the interaction with this j atom. */
 755             velec            = _mm256_andnot_ps(dummy_mask,velec);
 756             velecsum         = _mm256_add_ps(velecsum,velec);
 757
 758             fscal            = felec;
 759
 760             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 761
 762             /* Calculate temporary vectorial force */
 763             tx               = _mm256_mul_ps(fscal,dx21);
 764             ty               = _mm256_mul_ps(fscal,dy21);
 765             tz               = _mm256_mul_ps(fscal,dz21);
 766
 767             /* Update vectorial force */
 768             fix2             = _mm256_add_ps(fix2,tx);
 769             fiy2             = _mm256_add_ps(fiy2,ty);
 770             fiz2             = _mm256_add_ps(fiz2,tz);
 771
 772             fjx1             = _mm256_add_ps(fjx1,tx);
 773             fjy1             = _mm256_add_ps(fjy1,ty);
 774             fjz1             = _mm256_add_ps(fjz1,tz);
 775
 776             /**************************
 777              * CALCULATE INTERACTIONS *
 778              **************************/
 779
 780             /* REACTION-FIELD ELECTROSTATICS */
 781             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 782             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 783
 784             /* Update potential sum for this i atom from the interaction with this j atom. */
 785             velec            = _mm256_andnot_ps(dummy_mask,velec);
 786             velecsum         = _mm256_add_ps(velecsum,velec);
 787
 788             fscal            = felec;
 789
 790             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 791
 792             /* Calculate temporary vectorial force */
 793             tx               = _mm256_mul_ps(fscal,dx22);
 794             ty               = _mm256_mul_ps(fscal,dy22);
 795             tz               = _mm256_mul_ps(fscal,dz22);
 796
 797             /* Update vectorial force */
 798             fix2             = _mm256_add_ps(fix2,tx);
 799             fiy2             = _mm256_add_ps(fiy2,ty);
 800             fiz2             = _mm256_add_ps(fiz2,tz);
 801
 802             fjx2             = _mm256_add_ps(fjx2,tx);
 803             fjy2             = _mm256_add_ps(fjy2,ty);
 804             fjz2             = _mm256_add_ps(fjz2,tz);
 805
 806             /**************************
 807              * CALCULATE INTERACTIONS *
 808              **************************/
 809
 810             /* REACTION-FIELD ELECTROSTATICS */
 811             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 812             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 813
 814             /* Update potential sum for this i atom from the interaction with this j atom. */
 815             velec            = _mm256_andnot_ps(dummy_mask,velec);
 816             velecsum         = _mm256_add_ps(velecsum,velec);
 817
 818             fscal            = felec;
 819
 820             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 821
 822             /* Calculate temporary vectorial force */
 823             tx               = _mm256_mul_ps(fscal,dx23);
 824             ty               = _mm256_mul_ps(fscal,dy23);
 825             tz               = _mm256_mul_ps(fscal,dz23);
 826
 827             /* Update vectorial force */
 828             fix2             = _mm256_add_ps(fix2,tx);
 829             fiy2             = _mm256_add_ps(fiy2,ty);
 830             fiz2             = _mm256_add_ps(fiz2,tz);
 831
 832             fjx3             = _mm256_add_ps(fjx3,tx);
 833             fjy3             = _mm256_add_ps(fjy3,ty);
 834             fjz3             = _mm256_add_ps(fjz3,tz);
 835
 836             /**************************
 837              * CALCULATE INTERACTIONS *
 838              **************************/
 839
 840             /* REACTION-FIELD ELECTROSTATICS */
 841             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 842             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 843
 844             /* Update potential sum for this i atom from the interaction with this j atom. */
 845             velec            = _mm256_andnot_ps(dummy_mask,velec);
 846             velecsum         = _mm256_add_ps(velecsum,velec);
 847
 848             fscal            = felec;
 849
 850             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 851
 852             /* Calculate temporary vectorial force */
 853             tx               = _mm256_mul_ps(fscal,dx31);
 854             ty               = _mm256_mul_ps(fscal,dy31);
 855             tz               = _mm256_mul_ps(fscal,dz31);
 856
 857             /* Update vectorial force */
 858             fix3             = _mm256_add_ps(fix3,tx);
 859             fiy3             = _mm256_add_ps(fiy3,ty);
 860             fiz3             = _mm256_add_ps(fiz3,tz);
 861
 862             fjx1             = _mm256_add_ps(fjx1,tx);
 863             fjy1             = _mm256_add_ps(fjy1,ty);
 864             fjz1             = _mm256_add_ps(fjz1,tz);
 865
 866             /**************************
 867              * CALCULATE INTERACTIONS *
 868              **************************/
 869
 870             /* REACTION-FIELD ELECTROSTATICS */
 871             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 872             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 873
 874             /* Update potential sum for this i atom from the interaction with this j atom. */
 875             velec            = _mm256_andnot_ps(dummy_mask,velec);
 876             velecsum         = _mm256_add_ps(velecsum,velec);
 877
 878             fscal            = felec;
 879
 880             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 881
 882             /* Calculate temporary vectorial force */
 883             tx               = _mm256_mul_ps(fscal,dx32);
 884             ty               = _mm256_mul_ps(fscal,dy32);
 885             tz               = _mm256_mul_ps(fscal,dz32);
 886
 887             /* Update vectorial force */
 888             fix3             = _mm256_add_ps(fix3,tx);
 889             fiy3             = _mm256_add_ps(fiy3,ty);
 890             fiz3             = _mm256_add_ps(fiz3,tz);
 891
 892             fjx2             = _mm256_add_ps(fjx2,tx);
 893             fjy2             = _mm256_add_ps(fjy2,ty);
 894             fjz2             = _mm256_add_ps(fjz2,tz);
 895
 896             /**************************
 897              * CALCULATE INTERACTIONS *
 898              **************************/
 899
 900             /* REACTION-FIELD ELECTROSTATICS */
 901             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 902             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 903
 904             /* Update potential sum for this i atom from the interaction with this j atom. */
 905             velec            = _mm256_andnot_ps(dummy_mask,velec);
 906             velecsum         = _mm256_add_ps(velecsum,velec);
 907
 908             fscal            = felec;
 909
 910             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 911
 912             /* Calculate temporary vectorial force */
 913             tx               = _mm256_mul_ps(fscal,dx33);
 914             ty               = _mm256_mul_ps(fscal,dy33);
 915             tz               = _mm256_mul_ps(fscal,dz33);
 916
 917             /* Update vectorial force */
 918             fix3             = _mm256_add_ps(fix3,tx);
 919             fiy3             = _mm256_add_ps(fiy3,ty);
 920             fiz3             = _mm256_add_ps(fiz3,tz);
 921
 922             fjx3             = _mm256_add_ps(fjx3,tx);
 923             fjy3             = _mm256_add_ps(fjy3,ty);
 924             fjz3             = _mm256_add_ps(fjz3,tz);
 925
 926             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 927             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 928             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 929             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 930             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 931             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 932             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 933             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 934
 935             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 936                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 937                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 938
 939             /* Inner loop uses 288 flops */
 940         }
 941
 942         /* End of innermost loop */
 943
 944         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 945                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 946
 947         ggid                        = gid[iidx];
 948         /* Update potential energies */
 949         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 950
 951         /* Increment number of inner iterations */
 952         inneriter                  += j_index_end - j_index_start;
 953
 954         /* Outer loop uses 19 flops */
 955     }
 956
 957     /* Increment number of outer iterations */
 958     outeriter        += nri;
 959
 960     /* Update outer/inner flops */
 961
 962     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*288);
 963 }
 964 /*
 965  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW4W4_F_avx_256_single
 966  * Electrostatics interaction: ReactionField
 967  * VdW interaction:            None
 968  * Geometry:                   Water4-Water4
 969  * Calculate force/pot:        Force
 970  */
 971 void
 972 nb_kernel_ElecRF_VdwNone_GeomW4W4_F_avx_256_single
 973                     (t_nblist * gmx_restrict                nlist,
 974                      rvec * gmx_restrict                    xx,
 975                      rvec * gmx_restrict                    ff,
 976                      t_forcerec * gmx_restrict              fr,
 977                      t_mdatoms * gmx_restrict               mdatoms,
 978                      nb_kernel_data_t * gmx_restrict        kernel_data,
 979                      t_nrnb * gmx_restrict                  nrnb)
 980 {
 981     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 982      * just 0 for non-waters.
 983      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 984      * jnr indices corresponding to data put in the four positions in the SIMD register.
 985      */
 986     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 987     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 988     int              jnrA,jnrB,jnrC,jnrD;
 989     int              jnrE,jnrF,jnrG,jnrH;
 990     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 991     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 992     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 993     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 994     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 995     real             rcutoff_scalar;
 996     real             *shiftvec,*fshift,*x,*f;
 997     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 998     real             scratch[4*DIM];
 999     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1000     real *           vdwioffsetptr1;
1001     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1002     real *           vdwioffsetptr2;
1003     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1004     real *           vdwioffsetptr3;
1005     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1006     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1007     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1008     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1009     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1010     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1011     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1012     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1013     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1014     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1015     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1016     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1017     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1018     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1019     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1020     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1021     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1022     real             *charge;
1023     __m256           dummy_mask,cutoff_mask;
1024     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1025     __m256           one     = _mm256_set1_ps(1.0);
1026     __m256           two     = _mm256_set1_ps(2.0);
1027     x                = xx[0];
1028     f                = ff[0];
1029
1030     nri              = nlist->nri;
1031     iinr             = nlist->iinr;
1032     jindex           = nlist->jindex;
1033     jjnr             = nlist->jjnr;
1034     shiftidx         = nlist->shift;
1035     gid              = nlist->gid;
1036     shiftvec         = fr->shift_vec[0];
1037     fshift           = fr->fshift[0];
1038     facel            = _mm256_set1_ps(fr->epsfac);
1039     charge           = mdatoms->chargeA;
1040     krf              = _mm256_set1_ps(fr->ic->k_rf);
1041     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1042     crf              = _mm256_set1_ps(fr->ic->c_rf);
1043
1044     /* Setup water-specific parameters */
1045     inr              = nlist->iinr[0];
1046     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1047     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1048     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1049
1050     jq1              = _mm256_set1_ps(charge[inr+1]);
1051     jq2              = _mm256_set1_ps(charge[inr+2]);
1052     jq3              = _mm256_set1_ps(charge[inr+3]);
1053     qq11             = _mm256_mul_ps(iq1,jq1);
1054     qq12             = _mm256_mul_ps(iq1,jq2);
1055     qq13             = _mm256_mul_ps(iq1,jq3);
1056     qq21             = _mm256_mul_ps(iq2,jq1);
1057     qq22             = _mm256_mul_ps(iq2,jq2);
1058     qq23             = _mm256_mul_ps(iq2,jq3);
1059     qq31             = _mm256_mul_ps(iq3,jq1);
1060     qq32             = _mm256_mul_ps(iq3,jq2);
1061     qq33             = _mm256_mul_ps(iq3,jq3);
1062
1063     /* Avoid stupid compiler warnings */
1064     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1065     j_coord_offsetA = 0;
1066     j_coord_offsetB = 0;
1067     j_coord_offsetC = 0;
1068     j_coord_offsetD = 0;
1069     j_coord_offsetE = 0;
1070     j_coord_offsetF = 0;
1071     j_coord_offsetG = 0;
1072     j_coord_offsetH = 0;
1073
1074     outeriter        = 0;
1075     inneriter        = 0;
1076
1077     for(iidx=0;iidx<4*DIM;iidx++)
1078     {
1079         scratch[iidx] = 0.0;
1080     }
1081
1082     /* Start outer loop over neighborlists */
1083     for(iidx=0; iidx<nri; iidx++)
1084     {
1085         /* Load shift vector for this list */
1086         i_shift_offset   = DIM*shiftidx[iidx];
1087
1088         /* Load limits for loop over neighbors */
1089         j_index_start    = jindex[iidx];
1090         j_index_end      = jindex[iidx+1];
1091
1092         /* Get outer coordinate index */
1093         inr              = iinr[iidx];
1094         i_coord_offset   = DIM*inr;
1095
1096         /* Load i particle coords and add shift vector */
1097         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1098                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1099
1100         fix1             = _mm256_setzero_ps();
1101         fiy1             = _mm256_setzero_ps();
1102         fiz1             = _mm256_setzero_ps();
1103         fix2             = _mm256_setzero_ps();
1104         fiy2             = _mm256_setzero_ps();
1105         fiz2             = _mm256_setzero_ps();
1106         fix3             = _mm256_setzero_ps();
1107         fiy3             = _mm256_setzero_ps();
1108         fiz3             = _mm256_setzero_ps();
1109
1110         /* Start inner kernel loop */
1111         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1112         {
1113
1114             /* Get j neighbor index, and coordinate index */
1115             jnrA             = jjnr[jidx];
1116             jnrB             = jjnr[jidx+1];
1117             jnrC             = jjnr[jidx+2];
1118             jnrD             = jjnr[jidx+3];
1119             jnrE             = jjnr[jidx+4];
1120             jnrF             = jjnr[jidx+5];
1121             jnrG             = jjnr[jidx+6];
1122             jnrH             = jjnr[jidx+7];
1123             j_coord_offsetA  = DIM*jnrA;
1124             j_coord_offsetB  = DIM*jnrB;
1125             j_coord_offsetC  = DIM*jnrC;
1126             j_coord_offsetD  = DIM*jnrD;
1127             j_coord_offsetE  = DIM*jnrE;
1128             j_coord_offsetF  = DIM*jnrF;
1129             j_coord_offsetG  = DIM*jnrG;
1130             j_coord_offsetH  = DIM*jnrH;
1131
1132             /* load j atom coordinates */
1133             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1134                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1135                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1136                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1137                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1138
1139             /* Calculate displacement vector */
1140             dx11             = _mm256_sub_ps(ix1,jx1);
1141             dy11             = _mm256_sub_ps(iy1,jy1);
1142             dz11             = _mm256_sub_ps(iz1,jz1);
1143             dx12             = _mm256_sub_ps(ix1,jx2);
1144             dy12             = _mm256_sub_ps(iy1,jy2);
1145             dz12             = _mm256_sub_ps(iz1,jz2);
1146             dx13             = _mm256_sub_ps(ix1,jx3);
1147             dy13             = _mm256_sub_ps(iy1,jy3);
1148             dz13             = _mm256_sub_ps(iz1,jz3);
1149             dx21             = _mm256_sub_ps(ix2,jx1);
1150             dy21             = _mm256_sub_ps(iy2,jy1);
1151             dz21             = _mm256_sub_ps(iz2,jz1);
1152             dx22             = _mm256_sub_ps(ix2,jx2);
1153             dy22             = _mm256_sub_ps(iy2,jy2);
1154             dz22             = _mm256_sub_ps(iz2,jz2);
1155             dx23             = _mm256_sub_ps(ix2,jx3);
1156             dy23             = _mm256_sub_ps(iy2,jy3);
1157             dz23             = _mm256_sub_ps(iz2,jz3);
1158             dx31             = _mm256_sub_ps(ix3,jx1);
1159             dy31             = _mm256_sub_ps(iy3,jy1);
1160             dz31             = _mm256_sub_ps(iz3,jz1);
1161             dx32             = _mm256_sub_ps(ix3,jx2);
1162             dy32             = _mm256_sub_ps(iy3,jy2);
1163             dz32             = _mm256_sub_ps(iz3,jz2);
1164             dx33             = _mm256_sub_ps(ix3,jx3);
1165             dy33             = _mm256_sub_ps(iy3,jy3);
1166             dz33             = _mm256_sub_ps(iz3,jz3);
1167
1168             /* Calculate squared distance and things based on it */
1169             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1170             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1171             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1172             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1173             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1174             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1175             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1176             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1177             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1178
1179             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1180             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1181             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1182             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1183             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1184             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1185             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1186             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1187             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1188
1189             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1190             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1191             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1192             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1193             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1194             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1195             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1196             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1197             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1198
1199             fjx1             = _mm256_setzero_ps();
1200             fjy1             = _mm256_setzero_ps();
1201             fjz1             = _mm256_setzero_ps();
1202             fjx2             = _mm256_setzero_ps();
1203             fjy2             = _mm256_setzero_ps();
1204             fjz2             = _mm256_setzero_ps();
1205             fjx3             = _mm256_setzero_ps();
1206             fjy3             = _mm256_setzero_ps();
1207             fjz3             = _mm256_setzero_ps();
1208
1209             /**************************
1210              * CALCULATE INTERACTIONS *
1211              **************************/
1212
1213             /* REACTION-FIELD ELECTROSTATICS */
1214             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1215
1216             fscal            = felec;
1217
1218             /* Calculate temporary vectorial force */
1219             tx               = _mm256_mul_ps(fscal,dx11);
1220             ty               = _mm256_mul_ps(fscal,dy11);
1221             tz               = _mm256_mul_ps(fscal,dz11);
1222
1223             /* Update vectorial force */
1224             fix1             = _mm256_add_ps(fix1,tx);
1225             fiy1             = _mm256_add_ps(fiy1,ty);
1226             fiz1             = _mm256_add_ps(fiz1,tz);
1227
1228             fjx1             = _mm256_add_ps(fjx1,tx);
1229             fjy1             = _mm256_add_ps(fjy1,ty);
1230             fjz1             = _mm256_add_ps(fjz1,tz);
1231
1232             /**************************
1233              * CALCULATE INTERACTIONS *
1234              **************************/
1235
1236             /* REACTION-FIELD ELECTROSTATICS */
1237             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1238
1239             fscal            = felec;
1240
1241             /* Calculate temporary vectorial force */
1242             tx               = _mm256_mul_ps(fscal,dx12);
1243             ty               = _mm256_mul_ps(fscal,dy12);
1244             tz               = _mm256_mul_ps(fscal,dz12);
1245
1246             /* Update vectorial force */
1247             fix1             = _mm256_add_ps(fix1,tx);
1248             fiy1             = _mm256_add_ps(fiy1,ty);
1249             fiz1             = _mm256_add_ps(fiz1,tz);
1250
1251             fjx2             = _mm256_add_ps(fjx2,tx);
1252             fjy2             = _mm256_add_ps(fjy2,ty);
1253             fjz2             = _mm256_add_ps(fjz2,tz);
1254
1255             /**************************
1256              * CALCULATE INTERACTIONS *
1257              **************************/
1258
1259             /* REACTION-FIELD ELECTROSTATICS */
1260             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1261
1262             fscal            = felec;
1263
1264             /* Calculate temporary vectorial force */
1265             tx               = _mm256_mul_ps(fscal,dx13);
1266             ty               = _mm256_mul_ps(fscal,dy13);
1267             tz               = _mm256_mul_ps(fscal,dz13);
1268
1269             /* Update vectorial force */
1270             fix1             = _mm256_add_ps(fix1,tx);
1271             fiy1             = _mm256_add_ps(fiy1,ty);
1272             fiz1             = _mm256_add_ps(fiz1,tz);
1273
1274             fjx3             = _mm256_add_ps(fjx3,tx);
1275             fjy3             = _mm256_add_ps(fjy3,ty);
1276             fjz3             = _mm256_add_ps(fjz3,tz);
1277
1278             /**************************
1279              * CALCULATE INTERACTIONS *
1280              **************************/
1281
1282             /* REACTION-FIELD ELECTROSTATICS */
1283             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1284
1285             fscal            = felec;
1286
1287             /* Calculate temporary vectorial force */
1288             tx               = _mm256_mul_ps(fscal,dx21);
1289             ty               = _mm256_mul_ps(fscal,dy21);
1290             tz               = _mm256_mul_ps(fscal,dz21);
1291
1292             /* Update vectorial force */
1293             fix2             = _mm256_add_ps(fix2,tx);
1294             fiy2             = _mm256_add_ps(fiy2,ty);
1295             fiz2             = _mm256_add_ps(fiz2,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             /* REACTION-FIELD ELECTROSTATICS */
1306             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1307
1308             fscal            = felec;
1309
1310             /* Calculate temporary vectorial force */
1311             tx               = _mm256_mul_ps(fscal,dx22);
1312             ty               = _mm256_mul_ps(fscal,dy22);
1313             tz               = _mm256_mul_ps(fscal,dz22);
1314
1315             /* Update vectorial force */
1316             fix2             = _mm256_add_ps(fix2,tx);
1317             fiy2             = _mm256_add_ps(fiy2,ty);
1318             fiz2             = _mm256_add_ps(fiz2,tz);
1319
1320             fjx2             = _mm256_add_ps(fjx2,tx);
1321             fjy2             = _mm256_add_ps(fjy2,ty);
1322             fjz2             = _mm256_add_ps(fjz2,tz);
1323
1324             /**************************
1325              * CALCULATE INTERACTIONS *
1326              **************************/
1327
1328             /* REACTION-FIELD ELECTROSTATICS */
1329             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1330
1331             fscal            = felec;
1332
1333             /* Calculate temporary vectorial force */
1334             tx               = _mm256_mul_ps(fscal,dx23);
1335             ty               = _mm256_mul_ps(fscal,dy23);
1336             tz               = _mm256_mul_ps(fscal,dz23);
1337
1338             /* Update vectorial force */
1339             fix2             = _mm256_add_ps(fix2,tx);
1340             fiy2             = _mm256_add_ps(fiy2,ty);
1341             fiz2             = _mm256_add_ps(fiz2,tz);
1342
1343             fjx3             = _mm256_add_ps(fjx3,tx);
1344             fjy3             = _mm256_add_ps(fjy3,ty);
1345             fjz3             = _mm256_add_ps(fjz3,tz);
1346
1347             /**************************
1348              * CALCULATE INTERACTIONS *
1349              **************************/
1350
1351             /* REACTION-FIELD ELECTROSTATICS */
1352             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1353
1354             fscal            = felec;
1355
1356             /* Calculate temporary vectorial force */
1357             tx               = _mm256_mul_ps(fscal,dx31);
1358             ty               = _mm256_mul_ps(fscal,dy31);
1359             tz               = _mm256_mul_ps(fscal,dz31);
1360
1361             /* Update vectorial force */
1362             fix3             = _mm256_add_ps(fix3,tx);
1363             fiy3             = _mm256_add_ps(fiy3,ty);
1364             fiz3             = _mm256_add_ps(fiz3,tz);
1365
1366             fjx1             = _mm256_add_ps(fjx1,tx);
1367             fjy1             = _mm256_add_ps(fjy1,ty);
1368             fjz1             = _mm256_add_ps(fjz1,tz);
1369
1370             /**************************
1371              * CALCULATE INTERACTIONS *
1372              **************************/
1373
1374             /* REACTION-FIELD ELECTROSTATICS */
1375             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1376
1377             fscal            = felec;
1378
1379             /* Calculate temporary vectorial force */
1380             tx               = _mm256_mul_ps(fscal,dx32);
1381             ty               = _mm256_mul_ps(fscal,dy32);
1382             tz               = _mm256_mul_ps(fscal,dz32);
1383
1384             /* Update vectorial force */
1385             fix3             = _mm256_add_ps(fix3,tx);
1386             fiy3             = _mm256_add_ps(fiy3,ty);
1387             fiz3             = _mm256_add_ps(fiz3,tz);
1388
1389             fjx2             = _mm256_add_ps(fjx2,tx);
1390             fjy2             = _mm256_add_ps(fjy2,ty);
1391             fjz2             = _mm256_add_ps(fjz2,tz);
1392
1393             /**************************
1394              * CALCULATE INTERACTIONS *
1395              **************************/
1396
1397             /* REACTION-FIELD ELECTROSTATICS */
1398             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1399
1400             fscal            = felec;
1401
1402             /* Calculate temporary vectorial force */
1403             tx               = _mm256_mul_ps(fscal,dx33);
1404             ty               = _mm256_mul_ps(fscal,dy33);
1405             tz               = _mm256_mul_ps(fscal,dz33);
1406
1407             /* Update vectorial force */
1408             fix3             = _mm256_add_ps(fix3,tx);
1409             fiy3             = _mm256_add_ps(fiy3,ty);
1410             fiz3             = _mm256_add_ps(fiz3,tz);
1411
1412             fjx3             = _mm256_add_ps(fjx3,tx);
1413             fjy3             = _mm256_add_ps(fjy3,ty);
1414             fjz3             = _mm256_add_ps(fjz3,tz);
1415
1416             fjptrA             = f+j_coord_offsetA;
1417             fjptrB             = f+j_coord_offsetB;
1418             fjptrC             = f+j_coord_offsetC;
1419             fjptrD             = f+j_coord_offsetD;
1420             fjptrE             = f+j_coord_offsetE;
1421             fjptrF             = f+j_coord_offsetF;
1422             fjptrG             = f+j_coord_offsetG;
1423             fjptrH             = f+j_coord_offsetH;
1424
1425             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1426                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1427                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1428
1429             /* Inner loop uses 243 flops */
1430         }
1431
1432         if(jidx<j_index_end)
1433         {
1434
1435             /* Get j neighbor index, and coordinate index */
1436             jnrlistA         = jjnr[jidx];
1437             jnrlistB         = jjnr[jidx+1];
1438             jnrlistC         = jjnr[jidx+2];
1439             jnrlistD         = jjnr[jidx+3];
1440             jnrlistE         = jjnr[jidx+4];
1441             jnrlistF         = jjnr[jidx+5];
1442             jnrlistG         = jjnr[jidx+6];
1443             jnrlistH         = jjnr[jidx+7];
1444             /* Sign of each element will be negative for non-real atoms.
1445              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1446              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1447              */
1448             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1449                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1450
1451             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1452             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1453             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1454             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1455             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1456             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1457             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1458             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1459             j_coord_offsetA  = DIM*jnrA;
1460             j_coord_offsetB  = DIM*jnrB;
1461             j_coord_offsetC  = DIM*jnrC;
1462             j_coord_offsetD  = DIM*jnrD;
1463             j_coord_offsetE  = DIM*jnrE;
1464             j_coord_offsetF  = DIM*jnrF;
1465             j_coord_offsetG  = DIM*jnrG;
1466             j_coord_offsetH  = DIM*jnrH;
1467
1468             /* load j atom coordinates */
1469             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1470                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1471                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1472                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1473                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1474
1475             /* Calculate displacement vector */
1476             dx11             = _mm256_sub_ps(ix1,jx1);
1477             dy11             = _mm256_sub_ps(iy1,jy1);
1478             dz11             = _mm256_sub_ps(iz1,jz1);
1479             dx12             = _mm256_sub_ps(ix1,jx2);
1480             dy12             = _mm256_sub_ps(iy1,jy2);
1481             dz12             = _mm256_sub_ps(iz1,jz2);
1482             dx13             = _mm256_sub_ps(ix1,jx3);
1483             dy13             = _mm256_sub_ps(iy1,jy3);
1484             dz13             = _mm256_sub_ps(iz1,jz3);
1485             dx21             = _mm256_sub_ps(ix2,jx1);
1486             dy21             = _mm256_sub_ps(iy2,jy1);
1487             dz21             = _mm256_sub_ps(iz2,jz1);
1488             dx22             = _mm256_sub_ps(ix2,jx2);
1489             dy22             = _mm256_sub_ps(iy2,jy2);
1490             dz22             = _mm256_sub_ps(iz2,jz2);
1491             dx23             = _mm256_sub_ps(ix2,jx3);
1492             dy23             = _mm256_sub_ps(iy2,jy3);
1493             dz23             = _mm256_sub_ps(iz2,jz3);
1494             dx31             = _mm256_sub_ps(ix3,jx1);
1495             dy31             = _mm256_sub_ps(iy3,jy1);
1496             dz31             = _mm256_sub_ps(iz3,jz1);
1497             dx32             = _mm256_sub_ps(ix3,jx2);
1498             dy32             = _mm256_sub_ps(iy3,jy2);
1499             dz32             = _mm256_sub_ps(iz3,jz2);
1500             dx33             = _mm256_sub_ps(ix3,jx3);
1501             dy33             = _mm256_sub_ps(iy3,jy3);
1502             dz33             = _mm256_sub_ps(iz3,jz3);
1503
1504             /* Calculate squared distance and things based on it */
1505             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1506             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1507             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1508             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1509             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1510             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1511             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1512             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1513             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1514
1515             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1516             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1517             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1518             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1519             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1520             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1521             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1522             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1523             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1524
1525             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1526             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1527             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1528             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1529             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1530             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1531             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1532             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1533             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1534
1535             fjx1             = _mm256_setzero_ps();
1536             fjy1             = _mm256_setzero_ps();
1537             fjz1             = _mm256_setzero_ps();
1538             fjx2             = _mm256_setzero_ps();
1539             fjy2             = _mm256_setzero_ps();
1540             fjz2             = _mm256_setzero_ps();
1541             fjx3             = _mm256_setzero_ps();
1542             fjy3             = _mm256_setzero_ps();
1543             fjz3             = _mm256_setzero_ps();
1544
1545             /**************************
1546              * CALCULATE INTERACTIONS *
1547              **************************/
1548
1549             /* REACTION-FIELD ELECTROSTATICS */
1550             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1551
1552             fscal            = felec;
1553
1554             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1555
1556             /* Calculate temporary vectorial force */
1557             tx               = _mm256_mul_ps(fscal,dx11);
1558             ty               = _mm256_mul_ps(fscal,dy11);
1559             tz               = _mm256_mul_ps(fscal,dz11);
1560
1561             /* Update vectorial force */
1562             fix1             = _mm256_add_ps(fix1,tx);
1563             fiy1             = _mm256_add_ps(fiy1,ty);
1564             fiz1             = _mm256_add_ps(fiz1,tz);
1565
1566             fjx1             = _mm256_add_ps(fjx1,tx);
1567             fjy1             = _mm256_add_ps(fjy1,ty);
1568             fjz1             = _mm256_add_ps(fjz1,tz);
1569
1570             /**************************
1571              * CALCULATE INTERACTIONS *
1572              **************************/
1573
1574             /* REACTION-FIELD ELECTROSTATICS */
1575             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1576
1577             fscal            = felec;
1578
1579             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1580
1581             /* Calculate temporary vectorial force */
1582             tx               = _mm256_mul_ps(fscal,dx12);
1583             ty               = _mm256_mul_ps(fscal,dy12);
1584             tz               = _mm256_mul_ps(fscal,dz12);
1585
1586             /* Update vectorial force */
1587             fix1             = _mm256_add_ps(fix1,tx);
1588             fiy1             = _mm256_add_ps(fiy1,ty);
1589             fiz1             = _mm256_add_ps(fiz1,tz);
1590
1591             fjx2             = _mm256_add_ps(fjx2,tx);
1592             fjy2             = _mm256_add_ps(fjy2,ty);
1593             fjz2             = _mm256_add_ps(fjz2,tz);
1594
1595             /**************************
1596              * CALCULATE INTERACTIONS *
1597              **************************/
1598
1599             /* REACTION-FIELD ELECTROSTATICS */
1600             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1601
1602             fscal            = felec;
1603
1604             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1605
1606             /* Calculate temporary vectorial force */
1607             tx               = _mm256_mul_ps(fscal,dx13);
1608             ty               = _mm256_mul_ps(fscal,dy13);
1609             tz               = _mm256_mul_ps(fscal,dz13);
1610
1611             /* Update vectorial force */
1612             fix1             = _mm256_add_ps(fix1,tx);
1613             fiy1             = _mm256_add_ps(fiy1,ty);
1614             fiz1             = _mm256_add_ps(fiz1,tz);
1615
1616             fjx3             = _mm256_add_ps(fjx3,tx);
1617             fjy3             = _mm256_add_ps(fjy3,ty);
1618             fjz3             = _mm256_add_ps(fjz3,tz);
1619
1620             /**************************
1621              * CALCULATE INTERACTIONS *
1622              **************************/
1623
1624             /* REACTION-FIELD ELECTROSTATICS */
1625             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1626
1627             fscal            = felec;
1628
1629             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1630
1631             /* Calculate temporary vectorial force */
1632             tx               = _mm256_mul_ps(fscal,dx21);
1633             ty               = _mm256_mul_ps(fscal,dy21);
1634             tz               = _mm256_mul_ps(fscal,dz21);
1635
1636             /* Update vectorial force */
1637             fix2             = _mm256_add_ps(fix2,tx);
1638             fiy2             = _mm256_add_ps(fiy2,ty);
1639             fiz2             = _mm256_add_ps(fiz2,tz);
1640
1641             fjx1             = _mm256_add_ps(fjx1,tx);
1642             fjy1             = _mm256_add_ps(fjy1,ty);
1643             fjz1             = _mm256_add_ps(fjz1,tz);
1644
1645             /**************************
1646              * CALCULATE INTERACTIONS *
1647              **************************/
1648
1649             /* REACTION-FIELD ELECTROSTATICS */
1650             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1651
1652             fscal            = felec;
1653
1654             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1655
1656             /* Calculate temporary vectorial force */
1657             tx               = _mm256_mul_ps(fscal,dx22);
1658             ty               = _mm256_mul_ps(fscal,dy22);
1659             tz               = _mm256_mul_ps(fscal,dz22);
1660
1661             /* Update vectorial force */
1662             fix2             = _mm256_add_ps(fix2,tx);
1663             fiy2             = _mm256_add_ps(fiy2,ty);
1664             fiz2             = _mm256_add_ps(fiz2,tz);
1665
1666             fjx2             = _mm256_add_ps(fjx2,tx);
1667             fjy2             = _mm256_add_ps(fjy2,ty);
1668             fjz2             = _mm256_add_ps(fjz2,tz);
1669
1670             /**************************
1671              * CALCULATE INTERACTIONS *
1672              **************************/
1673
1674             /* REACTION-FIELD ELECTROSTATICS */
1675             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1676
1677             fscal            = felec;
1678
1679             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1680
1681             /* Calculate temporary vectorial force */
1682             tx               = _mm256_mul_ps(fscal,dx23);
1683             ty               = _mm256_mul_ps(fscal,dy23);
1684             tz               = _mm256_mul_ps(fscal,dz23);
1685
1686             /* Update vectorial force */
1687             fix2             = _mm256_add_ps(fix2,tx);
1688             fiy2             = _mm256_add_ps(fiy2,ty);
1689             fiz2             = _mm256_add_ps(fiz2,tz);
1690
1691             fjx3             = _mm256_add_ps(fjx3,tx);
1692             fjy3             = _mm256_add_ps(fjy3,ty);
1693             fjz3             = _mm256_add_ps(fjz3,tz);
1694
1695             /**************************
1696              * CALCULATE INTERACTIONS *
1697              **************************/
1698
1699             /* REACTION-FIELD ELECTROSTATICS */
1700             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1701
1702             fscal            = felec;
1703
1704             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1705
1706             /* Calculate temporary vectorial force */
1707             tx               = _mm256_mul_ps(fscal,dx31);
1708             ty               = _mm256_mul_ps(fscal,dy31);
1709             tz               = _mm256_mul_ps(fscal,dz31);
1710
1711             /* Update vectorial force */
1712             fix3             = _mm256_add_ps(fix3,tx);
1713             fiy3             = _mm256_add_ps(fiy3,ty);
1714             fiz3             = _mm256_add_ps(fiz3,tz);
1715
1716             fjx1             = _mm256_add_ps(fjx1,tx);
1717             fjy1             = _mm256_add_ps(fjy1,ty);
1718             fjz1             = _mm256_add_ps(fjz1,tz);
1719
1720             /**************************
1721              * CALCULATE INTERACTIONS *
1722              **************************/
1723
1724             /* REACTION-FIELD ELECTROSTATICS */
1725             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1726
1727             fscal            = felec;
1728
1729             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1730
1731             /* Calculate temporary vectorial force */
1732             tx               = _mm256_mul_ps(fscal,dx32);
1733             ty               = _mm256_mul_ps(fscal,dy32);
1734             tz               = _mm256_mul_ps(fscal,dz32);
1735
1736             /* Update vectorial force */
1737             fix3             = _mm256_add_ps(fix3,tx);
1738             fiy3             = _mm256_add_ps(fiy3,ty);
1739             fiz3             = _mm256_add_ps(fiz3,tz);
1740
1741             fjx2             = _mm256_add_ps(fjx2,tx);
1742             fjy2             = _mm256_add_ps(fjy2,ty);
1743             fjz2             = _mm256_add_ps(fjz2,tz);
1744
1745             /**************************
1746              * CALCULATE INTERACTIONS *
1747              **************************/
1748
1749             /* REACTION-FIELD ELECTROSTATICS */
1750             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1751
1752             fscal            = felec;
1753
1754             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1755
1756             /* Calculate temporary vectorial force */
1757             tx               = _mm256_mul_ps(fscal,dx33);
1758             ty               = _mm256_mul_ps(fscal,dy33);
1759             tz               = _mm256_mul_ps(fscal,dz33);
1760
1761             /* Update vectorial force */
1762             fix3             = _mm256_add_ps(fix3,tx);
1763             fiy3             = _mm256_add_ps(fiy3,ty);
1764             fiz3             = _mm256_add_ps(fiz3,tz);
1765
1766             fjx3             = _mm256_add_ps(fjx3,tx);
1767             fjy3             = _mm256_add_ps(fjy3,ty);
1768             fjz3             = _mm256_add_ps(fjz3,tz);
1769
1770             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1771             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1772             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1773             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1774             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1775             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1776             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1777             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1778
1779             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1780                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1781                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1782
1783             /* Inner loop uses 243 flops */
1784         }
1785
1786         /* End of innermost loop */
1787
1788         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1789                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1790
1791         /* Increment number of inner iterations */
1792         inneriter                  += j_index_end - j_index_start;
1793
1794         /* Outer loop uses 18 flops */
1795     }
1796
1797     /* Increment number of outer iterations */
1798     outeriter        += nri;
1799
1800     /* Update outer/inner flops */
1801
1802     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*243);
1803 }