src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRF_VdwNone_GeomW4W4_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_single kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_avx_256_single.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW4W4_VF_avx_256_single
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            None
  53  * Geometry:                   Water4-Water4
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRF_VdwNone_GeomW4W4_VF_avx_256_single
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrE,jnrF,jnrG,jnrH;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  79     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  80     real             rcutoff_scalar;
  81     real             *shiftvec,*fshift,*x,*f;
  82     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  83     real             scratch[4*DIM];
  84     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  85     real *           vdwioffsetptr1;
  86     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  87     real *           vdwioffsetptr2;
  88     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  89     real *           vdwioffsetptr3;
  90     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  91     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  92     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  93     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  94     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  95     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  96     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  97     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  98     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  99     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 100     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 101     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 102     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 103     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 104     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 105     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 106     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     __m256           dummy_mask,cutoff_mask;
 109     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 110     __m256           one     = _mm256_set1_ps(1.0);
 111     __m256           two     = _mm256_set1_ps(2.0);
 112     x                = xx[0];
 113     f                = ff[0];
 114
 115     nri              = nlist->nri;
 116     iinr             = nlist->iinr;
 117     jindex           = nlist->jindex;
 118     jjnr             = nlist->jjnr;
 119     shiftidx         = nlist->shift;
 120     gid              = nlist->gid;
 121     shiftvec         = fr->shift_vec[0];
 122     fshift           = fr->fshift[0];
 123     facel            = _mm256_set1_ps(fr->ic->epsfac);
 124     charge           = mdatoms->chargeA;
 125     krf              = _mm256_set1_ps(fr->ic->k_rf);
 126     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 127     crf              = _mm256_set1_ps(fr->ic->c_rf);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 132     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 133     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 134
 135     jq1              = _mm256_set1_ps(charge[inr+1]);
 136     jq2              = _mm256_set1_ps(charge[inr+2]);
 137     jq3              = _mm256_set1_ps(charge[inr+3]);
 138     qq11             = _mm256_mul_ps(iq1,jq1);
 139     qq12             = _mm256_mul_ps(iq1,jq2);
 140     qq13             = _mm256_mul_ps(iq1,jq3);
 141     qq21             = _mm256_mul_ps(iq2,jq1);
 142     qq22             = _mm256_mul_ps(iq2,jq2);
 143     qq23             = _mm256_mul_ps(iq2,jq3);
 144     qq31             = _mm256_mul_ps(iq3,jq1);
 145     qq32             = _mm256_mul_ps(iq3,jq2);
 146     qq33             = _mm256_mul_ps(iq3,jq3);
 147
 148     /* Avoid stupid compiler warnings */
 149     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 150     j_coord_offsetA = 0;
 151     j_coord_offsetB = 0;
 152     j_coord_offsetC = 0;
 153     j_coord_offsetD = 0;
 154     j_coord_offsetE = 0;
 155     j_coord_offsetF = 0;
 156     j_coord_offsetG = 0;
 157     j_coord_offsetH = 0;
 158
 159     outeriter        = 0;
 160     inneriter        = 0;
 161
 162     for(iidx=0;iidx<4*DIM;iidx++)
 163     {
 164         scratch[iidx] = 0.0;
 165     }
 166
 167     /* Start outer loop over neighborlists */
 168     for(iidx=0; iidx<nri; iidx++)
 169     {
 170         /* Load shift vector for this list */
 171         i_shift_offset   = DIM*shiftidx[iidx];
 172
 173         /* Load limits for loop over neighbors */
 174         j_index_start    = jindex[iidx];
 175         j_index_end      = jindex[iidx+1];
 176
 177         /* Get outer coordinate index */
 178         inr              = iinr[iidx];
 179         i_coord_offset   = DIM*inr;
 180
 181         /* Load i particle coords and add shift vector */
 182         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 183                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 184
 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         fix3             = _mm256_setzero_ps();
 192         fiy3             = _mm256_setzero_ps();
 193         fiz3             = _mm256_setzero_ps();
 194
 195         /* Reset potential sums */
 196         velecsum         = _mm256_setzero_ps();
 197
 198         /* Start inner kernel loop */
 199         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 200         {
 201
 202             /* Get j neighbor index, and coordinate index */
 203             jnrA             = jjnr[jidx];
 204             jnrB             = jjnr[jidx+1];
 205             jnrC             = jjnr[jidx+2];
 206             jnrD             = jjnr[jidx+3];
 207             jnrE             = jjnr[jidx+4];
 208             jnrF             = jjnr[jidx+5];
 209             jnrG             = jjnr[jidx+6];
 210             jnrH             = jjnr[jidx+7];
 211             j_coord_offsetA  = DIM*jnrA;
 212             j_coord_offsetB  = DIM*jnrB;
 213             j_coord_offsetC  = DIM*jnrC;
 214             j_coord_offsetD  = DIM*jnrD;
 215             j_coord_offsetE  = DIM*jnrE;
 216             j_coord_offsetF  = DIM*jnrF;
 217             j_coord_offsetG  = DIM*jnrG;
 218             j_coord_offsetH  = DIM*jnrH;
 219
 220             /* load j atom coordinates */
 221             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 222                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 223                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 224                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 225                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 226
 227             /* Calculate displacement vector */
 228             dx11             = _mm256_sub_ps(ix1,jx1);
 229             dy11             = _mm256_sub_ps(iy1,jy1);
 230             dz11             = _mm256_sub_ps(iz1,jz1);
 231             dx12             = _mm256_sub_ps(ix1,jx2);
 232             dy12             = _mm256_sub_ps(iy1,jy2);
 233             dz12             = _mm256_sub_ps(iz1,jz2);
 234             dx13             = _mm256_sub_ps(ix1,jx3);
 235             dy13             = _mm256_sub_ps(iy1,jy3);
 236             dz13             = _mm256_sub_ps(iz1,jz3);
 237             dx21             = _mm256_sub_ps(ix2,jx1);
 238             dy21             = _mm256_sub_ps(iy2,jy1);
 239             dz21             = _mm256_sub_ps(iz2,jz1);
 240             dx22             = _mm256_sub_ps(ix2,jx2);
 241             dy22             = _mm256_sub_ps(iy2,jy2);
 242             dz22             = _mm256_sub_ps(iz2,jz2);
 243             dx23             = _mm256_sub_ps(ix2,jx3);
 244             dy23             = _mm256_sub_ps(iy2,jy3);
 245             dz23             = _mm256_sub_ps(iz2,jz3);
 246             dx31             = _mm256_sub_ps(ix3,jx1);
 247             dy31             = _mm256_sub_ps(iy3,jy1);
 248             dz31             = _mm256_sub_ps(iz3,jz1);
 249             dx32             = _mm256_sub_ps(ix3,jx2);
 250             dy32             = _mm256_sub_ps(iy3,jy2);
 251             dz32             = _mm256_sub_ps(iz3,jz2);
 252             dx33             = _mm256_sub_ps(ix3,jx3);
 253             dy33             = _mm256_sub_ps(iy3,jy3);
 254             dz33             = _mm256_sub_ps(iz3,jz3);
 255
 256             /* Calculate squared distance and things based on it */
 257             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 258             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 259             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 260             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 261             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 262             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 263             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 264             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 265             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 266
 267             rinv11           = avx256_invsqrt_f(rsq11);
 268             rinv12           = avx256_invsqrt_f(rsq12);
 269             rinv13           = avx256_invsqrt_f(rsq13);
 270             rinv21           = avx256_invsqrt_f(rsq21);
 271             rinv22           = avx256_invsqrt_f(rsq22);
 272             rinv23           = avx256_invsqrt_f(rsq23);
 273             rinv31           = avx256_invsqrt_f(rsq31);
 274             rinv32           = avx256_invsqrt_f(rsq32);
 275             rinv33           = avx256_invsqrt_f(rsq33);
 276
 277             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 278             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 279             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 280             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 281             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 282             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 283             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 284             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 285             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 286
 287             fjx1             = _mm256_setzero_ps();
 288             fjy1             = _mm256_setzero_ps();
 289             fjz1             = _mm256_setzero_ps();
 290             fjx2             = _mm256_setzero_ps();
 291             fjy2             = _mm256_setzero_ps();
 292             fjz2             = _mm256_setzero_ps();
 293             fjx3             = _mm256_setzero_ps();
 294             fjy3             = _mm256_setzero_ps();
 295             fjz3             = _mm256_setzero_ps();
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             /* REACTION-FIELD ELECTROSTATICS */
 302             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 303             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 304
 305             /* Update potential sum for this i atom from the interaction with this j atom. */
 306             velecsum         = _mm256_add_ps(velecsum,velec);
 307
 308             fscal            = felec;
 309
 310             /* Calculate temporary vectorial force */
 311             tx               = _mm256_mul_ps(fscal,dx11);
 312             ty               = _mm256_mul_ps(fscal,dy11);
 313             tz               = _mm256_mul_ps(fscal,dz11);
 314
 315             /* Update vectorial force */
 316             fix1             = _mm256_add_ps(fix1,tx);
 317             fiy1             = _mm256_add_ps(fiy1,ty);
 318             fiz1             = _mm256_add_ps(fiz1,tz);
 319
 320             fjx1             = _mm256_add_ps(fjx1,tx);
 321             fjy1             = _mm256_add_ps(fjy1,ty);
 322             fjz1             = _mm256_add_ps(fjz1,tz);
 323
 324             /**************************
 325              * CALCULATE INTERACTIONS *
 326              **************************/
 327
 328             /* REACTION-FIELD ELECTROSTATICS */
 329             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 330             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 331
 332             /* Update potential sum for this i atom from the interaction with this j atom. */
 333             velecsum         = _mm256_add_ps(velecsum,velec);
 334
 335             fscal            = felec;
 336
 337             /* Calculate temporary vectorial force */
 338             tx               = _mm256_mul_ps(fscal,dx12);
 339             ty               = _mm256_mul_ps(fscal,dy12);
 340             tz               = _mm256_mul_ps(fscal,dz12);
 341
 342             /* Update vectorial force */
 343             fix1             = _mm256_add_ps(fix1,tx);
 344             fiy1             = _mm256_add_ps(fiy1,ty);
 345             fiz1             = _mm256_add_ps(fiz1,tz);
 346
 347             fjx2             = _mm256_add_ps(fjx2,tx);
 348             fjy2             = _mm256_add_ps(fjy2,ty);
 349             fjz2             = _mm256_add_ps(fjz2,tz);
 350
 351             /**************************
 352              * CALCULATE INTERACTIONS *
 353              **************************/
 354
 355             /* REACTION-FIELD ELECTROSTATICS */
 356             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 357             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 358
 359             /* Update potential sum for this i atom from the interaction with this j atom. */
 360             velecsum         = _mm256_add_ps(velecsum,velec);
 361
 362             fscal            = felec;
 363
 364             /* Calculate temporary vectorial force */
 365             tx               = _mm256_mul_ps(fscal,dx13);
 366             ty               = _mm256_mul_ps(fscal,dy13);
 367             tz               = _mm256_mul_ps(fscal,dz13);
 368
 369             /* Update vectorial force */
 370             fix1             = _mm256_add_ps(fix1,tx);
 371             fiy1             = _mm256_add_ps(fiy1,ty);
 372             fiz1             = _mm256_add_ps(fiz1,tz);
 373
 374             fjx3             = _mm256_add_ps(fjx3,tx);
 375             fjy3             = _mm256_add_ps(fjy3,ty);
 376             fjz3             = _mm256_add_ps(fjz3,tz);
 377
 378             /**************************
 379              * CALCULATE INTERACTIONS *
 380              **************************/
 381
 382             /* REACTION-FIELD ELECTROSTATICS */
 383             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 384             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 385
 386             /* Update potential sum for this i atom from the interaction with this j atom. */
 387             velecsum         = _mm256_add_ps(velecsum,velec);
 388
 389             fscal            = felec;
 390
 391             /* Calculate temporary vectorial force */
 392             tx               = _mm256_mul_ps(fscal,dx21);
 393             ty               = _mm256_mul_ps(fscal,dy21);
 394             tz               = _mm256_mul_ps(fscal,dz21);
 395
 396             /* Update vectorial force */
 397             fix2             = _mm256_add_ps(fix2,tx);
 398             fiy2             = _mm256_add_ps(fiy2,ty);
 399             fiz2             = _mm256_add_ps(fiz2,tz);
 400
 401             fjx1             = _mm256_add_ps(fjx1,tx);
 402             fjy1             = _mm256_add_ps(fjy1,ty);
 403             fjz1             = _mm256_add_ps(fjz1,tz);
 404
 405             /**************************
 406              * CALCULATE INTERACTIONS *
 407              **************************/
 408
 409             /* REACTION-FIELD ELECTROSTATICS */
 410             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 411             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 412
 413             /* Update potential sum for this i atom from the interaction with this j atom. */
 414             velecsum         = _mm256_add_ps(velecsum,velec);
 415
 416             fscal            = felec;
 417
 418             /* Calculate temporary vectorial force */
 419             tx               = _mm256_mul_ps(fscal,dx22);
 420             ty               = _mm256_mul_ps(fscal,dy22);
 421             tz               = _mm256_mul_ps(fscal,dz22);
 422
 423             /* Update vectorial force */
 424             fix2             = _mm256_add_ps(fix2,tx);
 425             fiy2             = _mm256_add_ps(fiy2,ty);
 426             fiz2             = _mm256_add_ps(fiz2,tz);
 427
 428             fjx2             = _mm256_add_ps(fjx2,tx);
 429             fjy2             = _mm256_add_ps(fjy2,ty);
 430             fjz2             = _mm256_add_ps(fjz2,tz);
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             /* REACTION-FIELD ELECTROSTATICS */
 437             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 438             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 439
 440             /* Update potential sum for this i atom from the interaction with this j atom. */
 441             velecsum         = _mm256_add_ps(velecsum,velec);
 442
 443             fscal            = felec;
 444
 445             /* Calculate temporary vectorial force */
 446             tx               = _mm256_mul_ps(fscal,dx23);
 447             ty               = _mm256_mul_ps(fscal,dy23);
 448             tz               = _mm256_mul_ps(fscal,dz23);
 449
 450             /* Update vectorial force */
 451             fix2             = _mm256_add_ps(fix2,tx);
 452             fiy2             = _mm256_add_ps(fiy2,ty);
 453             fiz2             = _mm256_add_ps(fiz2,tz);
 454
 455             fjx3             = _mm256_add_ps(fjx3,tx);
 456             fjy3             = _mm256_add_ps(fjy3,ty);
 457             fjz3             = _mm256_add_ps(fjz3,tz);
 458
 459             /**************************
 460              * CALCULATE INTERACTIONS *
 461              **************************/
 462
 463             /* REACTION-FIELD ELECTROSTATICS */
 464             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 465             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 466
 467             /* Update potential sum for this i atom from the interaction with this j atom. */
 468             velecsum         = _mm256_add_ps(velecsum,velec);
 469
 470             fscal            = felec;
 471
 472             /* Calculate temporary vectorial force */
 473             tx               = _mm256_mul_ps(fscal,dx31);
 474             ty               = _mm256_mul_ps(fscal,dy31);
 475             tz               = _mm256_mul_ps(fscal,dz31);
 476
 477             /* Update vectorial force */
 478             fix3             = _mm256_add_ps(fix3,tx);
 479             fiy3             = _mm256_add_ps(fiy3,ty);
 480             fiz3             = _mm256_add_ps(fiz3,tz);
 481
 482             fjx1             = _mm256_add_ps(fjx1,tx);
 483             fjy1             = _mm256_add_ps(fjy1,ty);
 484             fjz1             = _mm256_add_ps(fjz1,tz);
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             /* REACTION-FIELD ELECTROSTATICS */
 491             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 492             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 493
 494             /* Update potential sum for this i atom from the interaction with this j atom. */
 495             velecsum         = _mm256_add_ps(velecsum,velec);
 496
 497             fscal            = felec;
 498
 499             /* Calculate temporary vectorial force */
 500             tx               = _mm256_mul_ps(fscal,dx32);
 501             ty               = _mm256_mul_ps(fscal,dy32);
 502             tz               = _mm256_mul_ps(fscal,dz32);
 503
 504             /* Update vectorial force */
 505             fix3             = _mm256_add_ps(fix3,tx);
 506             fiy3             = _mm256_add_ps(fiy3,ty);
 507             fiz3             = _mm256_add_ps(fiz3,tz);
 508
 509             fjx2             = _mm256_add_ps(fjx2,tx);
 510             fjy2             = _mm256_add_ps(fjy2,ty);
 511             fjz2             = _mm256_add_ps(fjz2,tz);
 512
 513             /**************************
 514              * CALCULATE INTERACTIONS *
 515              **************************/
 516
 517             /* REACTION-FIELD ELECTROSTATICS */
 518             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 519             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 520
 521             /* Update potential sum for this i atom from the interaction with this j atom. */
 522             velecsum         = _mm256_add_ps(velecsum,velec);
 523
 524             fscal            = felec;
 525
 526             /* Calculate temporary vectorial force */
 527             tx               = _mm256_mul_ps(fscal,dx33);
 528             ty               = _mm256_mul_ps(fscal,dy33);
 529             tz               = _mm256_mul_ps(fscal,dz33);
 530
 531             /* Update vectorial force */
 532             fix3             = _mm256_add_ps(fix3,tx);
 533             fiy3             = _mm256_add_ps(fiy3,ty);
 534             fiz3             = _mm256_add_ps(fiz3,tz);
 535
 536             fjx3             = _mm256_add_ps(fjx3,tx);
 537             fjy3             = _mm256_add_ps(fjy3,ty);
 538             fjz3             = _mm256_add_ps(fjz3,tz);
 539
 540             fjptrA             = f+j_coord_offsetA;
 541             fjptrB             = f+j_coord_offsetB;
 542             fjptrC             = f+j_coord_offsetC;
 543             fjptrD             = f+j_coord_offsetD;
 544             fjptrE             = f+j_coord_offsetE;
 545             fjptrF             = f+j_coord_offsetF;
 546             fjptrG             = f+j_coord_offsetG;
 547             fjptrH             = f+j_coord_offsetH;
 548
 549             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 550                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 551                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 552
 553             /* Inner loop uses 288 flops */
 554         }
 555
 556         if(jidx<j_index_end)
 557         {
 558
 559             /* Get j neighbor index, and coordinate index */
 560             jnrlistA         = jjnr[jidx];
 561             jnrlistB         = jjnr[jidx+1];
 562             jnrlistC         = jjnr[jidx+2];
 563             jnrlistD         = jjnr[jidx+3];
 564             jnrlistE         = jjnr[jidx+4];
 565             jnrlistF         = jjnr[jidx+5];
 566             jnrlistG         = jjnr[jidx+6];
 567             jnrlistH         = jjnr[jidx+7];
 568             /* Sign of each element will be negative for non-real atoms.
 569              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 570              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 571              */
 572             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 573                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 574
 575             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 576             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 577             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 578             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 579             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 580             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 581             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 582             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 583             j_coord_offsetA  = DIM*jnrA;
 584             j_coord_offsetB  = DIM*jnrB;
 585             j_coord_offsetC  = DIM*jnrC;
 586             j_coord_offsetD  = DIM*jnrD;
 587             j_coord_offsetE  = DIM*jnrE;
 588             j_coord_offsetF  = DIM*jnrF;
 589             j_coord_offsetG  = DIM*jnrG;
 590             j_coord_offsetH  = DIM*jnrH;
 591
 592             /* load j atom coordinates */
 593             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 594                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 595                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 596                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 597                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 598
 599             /* Calculate displacement vector */
 600             dx11             = _mm256_sub_ps(ix1,jx1);
 601             dy11             = _mm256_sub_ps(iy1,jy1);
 602             dz11             = _mm256_sub_ps(iz1,jz1);
 603             dx12             = _mm256_sub_ps(ix1,jx2);
 604             dy12             = _mm256_sub_ps(iy1,jy2);
 605             dz12             = _mm256_sub_ps(iz1,jz2);
 606             dx13             = _mm256_sub_ps(ix1,jx3);
 607             dy13             = _mm256_sub_ps(iy1,jy3);
 608             dz13             = _mm256_sub_ps(iz1,jz3);
 609             dx21             = _mm256_sub_ps(ix2,jx1);
 610             dy21             = _mm256_sub_ps(iy2,jy1);
 611             dz21             = _mm256_sub_ps(iz2,jz1);
 612             dx22             = _mm256_sub_ps(ix2,jx2);
 613             dy22             = _mm256_sub_ps(iy2,jy2);
 614             dz22             = _mm256_sub_ps(iz2,jz2);
 615             dx23             = _mm256_sub_ps(ix2,jx3);
 616             dy23             = _mm256_sub_ps(iy2,jy3);
 617             dz23             = _mm256_sub_ps(iz2,jz3);
 618             dx31             = _mm256_sub_ps(ix3,jx1);
 619             dy31             = _mm256_sub_ps(iy3,jy1);
 620             dz31             = _mm256_sub_ps(iz3,jz1);
 621             dx32             = _mm256_sub_ps(ix3,jx2);
 622             dy32             = _mm256_sub_ps(iy3,jy2);
 623             dz32             = _mm256_sub_ps(iz3,jz2);
 624             dx33             = _mm256_sub_ps(ix3,jx3);
 625             dy33             = _mm256_sub_ps(iy3,jy3);
 626             dz33             = _mm256_sub_ps(iz3,jz3);
 627
 628             /* Calculate squared distance and things based on it */
 629             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 630             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 631             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 632             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 633             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 634             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 635             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 636             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 637             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 638
 639             rinv11           = avx256_invsqrt_f(rsq11);
 640             rinv12           = avx256_invsqrt_f(rsq12);
 641             rinv13           = avx256_invsqrt_f(rsq13);
 642             rinv21           = avx256_invsqrt_f(rsq21);
 643             rinv22           = avx256_invsqrt_f(rsq22);
 644             rinv23           = avx256_invsqrt_f(rsq23);
 645             rinv31           = avx256_invsqrt_f(rsq31);
 646             rinv32           = avx256_invsqrt_f(rsq32);
 647             rinv33           = avx256_invsqrt_f(rsq33);
 648
 649             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 650             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 651             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 652             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 653             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 654             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 655             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 656             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 657             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 658
 659             fjx1             = _mm256_setzero_ps();
 660             fjy1             = _mm256_setzero_ps();
 661             fjz1             = _mm256_setzero_ps();
 662             fjx2             = _mm256_setzero_ps();
 663             fjy2             = _mm256_setzero_ps();
 664             fjz2             = _mm256_setzero_ps();
 665             fjx3             = _mm256_setzero_ps();
 666             fjy3             = _mm256_setzero_ps();
 667             fjz3             = _mm256_setzero_ps();
 668
 669             /**************************
 670              * CALCULATE INTERACTIONS *
 671              **************************/
 672
 673             /* REACTION-FIELD ELECTROSTATICS */
 674             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 675             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 676
 677             /* Update potential sum for this i atom from the interaction with this j atom. */
 678             velec            = _mm256_andnot_ps(dummy_mask,velec);
 679             velecsum         = _mm256_add_ps(velecsum,velec);
 680
 681             fscal            = felec;
 682
 683             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 684
 685             /* Calculate temporary vectorial force */
 686             tx               = _mm256_mul_ps(fscal,dx11);
 687             ty               = _mm256_mul_ps(fscal,dy11);
 688             tz               = _mm256_mul_ps(fscal,dz11);
 689
 690             /* Update vectorial force */
 691             fix1             = _mm256_add_ps(fix1,tx);
 692             fiy1             = _mm256_add_ps(fiy1,ty);
 693             fiz1             = _mm256_add_ps(fiz1,tz);
 694
 695             fjx1             = _mm256_add_ps(fjx1,tx);
 696             fjy1             = _mm256_add_ps(fjy1,ty);
 697             fjz1             = _mm256_add_ps(fjz1,tz);
 698
 699             /**************************
 700              * CALCULATE INTERACTIONS *
 701              **************************/
 702
 703             /* REACTION-FIELD ELECTROSTATICS */
 704             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 705             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 706
 707             /* Update potential sum for this i atom from the interaction with this j atom. */
 708             velec            = _mm256_andnot_ps(dummy_mask,velec);
 709             velecsum         = _mm256_add_ps(velecsum,velec);
 710
 711             fscal            = felec;
 712
 713             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 714
 715             /* Calculate temporary vectorial force */
 716             tx               = _mm256_mul_ps(fscal,dx12);
 717             ty               = _mm256_mul_ps(fscal,dy12);
 718             tz               = _mm256_mul_ps(fscal,dz12);
 719
 720             /* Update vectorial force */
 721             fix1             = _mm256_add_ps(fix1,tx);
 722             fiy1             = _mm256_add_ps(fiy1,ty);
 723             fiz1             = _mm256_add_ps(fiz1,tz);
 724
 725             fjx2             = _mm256_add_ps(fjx2,tx);
 726             fjy2             = _mm256_add_ps(fjy2,ty);
 727             fjz2             = _mm256_add_ps(fjz2,tz);
 728
 729             /**************************
 730              * CALCULATE INTERACTIONS *
 731              **************************/
 732
 733             /* REACTION-FIELD ELECTROSTATICS */
 734             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 735             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 736
 737             /* Update potential sum for this i atom from the interaction with this j atom. */
 738             velec            = _mm256_andnot_ps(dummy_mask,velec);
 739             velecsum         = _mm256_add_ps(velecsum,velec);
 740
 741             fscal            = felec;
 742
 743             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 744
 745             /* Calculate temporary vectorial force */
 746             tx               = _mm256_mul_ps(fscal,dx13);
 747             ty               = _mm256_mul_ps(fscal,dy13);
 748             tz               = _mm256_mul_ps(fscal,dz13);
 749
 750             /* Update vectorial force */
 751             fix1             = _mm256_add_ps(fix1,tx);
 752             fiy1             = _mm256_add_ps(fiy1,ty);
 753             fiz1             = _mm256_add_ps(fiz1,tz);
 754
 755             fjx3             = _mm256_add_ps(fjx3,tx);
 756             fjy3             = _mm256_add_ps(fjy3,ty);
 757             fjz3             = _mm256_add_ps(fjz3,tz);
 758
 759             /**************************
 760              * CALCULATE INTERACTIONS *
 761              **************************/
 762
 763             /* REACTION-FIELD ELECTROSTATICS */
 764             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 765             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 766
 767             /* Update potential sum for this i atom from the interaction with this j atom. */
 768             velec            = _mm256_andnot_ps(dummy_mask,velec);
 769             velecsum         = _mm256_add_ps(velecsum,velec);
 770
 771             fscal            = felec;
 772
 773             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 774
 775             /* Calculate temporary vectorial force */
 776             tx               = _mm256_mul_ps(fscal,dx21);
 777             ty               = _mm256_mul_ps(fscal,dy21);
 778             tz               = _mm256_mul_ps(fscal,dz21);
 779
 780             /* Update vectorial force */
 781             fix2             = _mm256_add_ps(fix2,tx);
 782             fiy2             = _mm256_add_ps(fiy2,ty);
 783             fiz2             = _mm256_add_ps(fiz2,tz);
 784
 785             fjx1             = _mm256_add_ps(fjx1,tx);
 786             fjy1             = _mm256_add_ps(fjy1,ty);
 787             fjz1             = _mm256_add_ps(fjz1,tz);
 788
 789             /**************************
 790              * CALCULATE INTERACTIONS *
 791              **************************/
 792
 793             /* REACTION-FIELD ELECTROSTATICS */
 794             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 795             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 796
 797             /* Update potential sum for this i atom from the interaction with this j atom. */
 798             velec            = _mm256_andnot_ps(dummy_mask,velec);
 799             velecsum         = _mm256_add_ps(velecsum,velec);
 800
 801             fscal            = felec;
 802
 803             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 804
 805             /* Calculate temporary vectorial force */
 806             tx               = _mm256_mul_ps(fscal,dx22);
 807             ty               = _mm256_mul_ps(fscal,dy22);
 808             tz               = _mm256_mul_ps(fscal,dz22);
 809
 810             /* Update vectorial force */
 811             fix2             = _mm256_add_ps(fix2,tx);
 812             fiy2             = _mm256_add_ps(fiy2,ty);
 813             fiz2             = _mm256_add_ps(fiz2,tz);
 814
 815             fjx2             = _mm256_add_ps(fjx2,tx);
 816             fjy2             = _mm256_add_ps(fjy2,ty);
 817             fjz2             = _mm256_add_ps(fjz2,tz);
 818
 819             /**************************
 820              * CALCULATE INTERACTIONS *
 821              **************************/
 822
 823             /* REACTION-FIELD ELECTROSTATICS */
 824             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 825             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 826
 827             /* Update potential sum for this i atom from the interaction with this j atom. */
 828             velec            = _mm256_andnot_ps(dummy_mask,velec);
 829             velecsum         = _mm256_add_ps(velecsum,velec);
 830
 831             fscal            = felec;
 832
 833             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 834
 835             /* Calculate temporary vectorial force */
 836             tx               = _mm256_mul_ps(fscal,dx23);
 837             ty               = _mm256_mul_ps(fscal,dy23);
 838             tz               = _mm256_mul_ps(fscal,dz23);
 839
 840             /* Update vectorial force */
 841             fix2             = _mm256_add_ps(fix2,tx);
 842             fiy2             = _mm256_add_ps(fiy2,ty);
 843             fiz2             = _mm256_add_ps(fiz2,tz);
 844
 845             fjx3             = _mm256_add_ps(fjx3,tx);
 846             fjy3             = _mm256_add_ps(fjy3,ty);
 847             fjz3             = _mm256_add_ps(fjz3,tz);
 848
 849             /**************************
 850              * CALCULATE INTERACTIONS *
 851              **************************/
 852
 853             /* REACTION-FIELD ELECTROSTATICS */
 854             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 855             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 856
 857             /* Update potential sum for this i atom from the interaction with this j atom. */
 858             velec            = _mm256_andnot_ps(dummy_mask,velec);
 859             velecsum         = _mm256_add_ps(velecsum,velec);
 860
 861             fscal            = felec;
 862
 863             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 864
 865             /* Calculate temporary vectorial force */
 866             tx               = _mm256_mul_ps(fscal,dx31);
 867             ty               = _mm256_mul_ps(fscal,dy31);
 868             tz               = _mm256_mul_ps(fscal,dz31);
 869
 870             /* Update vectorial force */
 871             fix3             = _mm256_add_ps(fix3,tx);
 872             fiy3             = _mm256_add_ps(fiy3,ty);
 873             fiz3             = _mm256_add_ps(fiz3,tz);
 874
 875             fjx1             = _mm256_add_ps(fjx1,tx);
 876             fjy1             = _mm256_add_ps(fjy1,ty);
 877             fjz1             = _mm256_add_ps(fjz1,tz);
 878
 879             /**************************
 880              * CALCULATE INTERACTIONS *
 881              **************************/
 882
 883             /* REACTION-FIELD ELECTROSTATICS */
 884             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 885             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 886
 887             /* Update potential sum for this i atom from the interaction with this j atom. */
 888             velec            = _mm256_andnot_ps(dummy_mask,velec);
 889             velecsum         = _mm256_add_ps(velecsum,velec);
 890
 891             fscal            = felec;
 892
 893             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 894
 895             /* Calculate temporary vectorial force */
 896             tx               = _mm256_mul_ps(fscal,dx32);
 897             ty               = _mm256_mul_ps(fscal,dy32);
 898             tz               = _mm256_mul_ps(fscal,dz32);
 899
 900             /* Update vectorial force */
 901             fix3             = _mm256_add_ps(fix3,tx);
 902             fiy3             = _mm256_add_ps(fiy3,ty);
 903             fiz3             = _mm256_add_ps(fiz3,tz);
 904
 905             fjx2             = _mm256_add_ps(fjx2,tx);
 906             fjy2             = _mm256_add_ps(fjy2,ty);
 907             fjz2             = _mm256_add_ps(fjz2,tz);
 908
 909             /**************************
 910              * CALCULATE INTERACTIONS *
 911              **************************/
 912
 913             /* REACTION-FIELD ELECTROSTATICS */
 914             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 915             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 916
 917             /* Update potential sum for this i atom from the interaction with this j atom. */
 918             velec            = _mm256_andnot_ps(dummy_mask,velec);
 919             velecsum         = _mm256_add_ps(velecsum,velec);
 920
 921             fscal            = felec;
 922
 923             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 924
 925             /* Calculate temporary vectorial force */
 926             tx               = _mm256_mul_ps(fscal,dx33);
 927             ty               = _mm256_mul_ps(fscal,dy33);
 928             tz               = _mm256_mul_ps(fscal,dz33);
 929
 930             /* Update vectorial force */
 931             fix3             = _mm256_add_ps(fix3,tx);
 932             fiy3             = _mm256_add_ps(fiy3,ty);
 933             fiz3             = _mm256_add_ps(fiz3,tz);
 934
 935             fjx3             = _mm256_add_ps(fjx3,tx);
 936             fjy3             = _mm256_add_ps(fjy3,ty);
 937             fjz3             = _mm256_add_ps(fjz3,tz);
 938
 939             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 940             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 941             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 942             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 943             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 944             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 945             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 946             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 947
 948             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 949                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 950                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 951
 952             /* Inner loop uses 288 flops */
 953         }
 954
 955         /* End of innermost loop */
 956
 957         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 958                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 959
 960         ggid                        = gid[iidx];
 961         /* Update potential energies */
 962         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 963
 964         /* Increment number of inner iterations */
 965         inneriter                  += j_index_end - j_index_start;
 966
 967         /* Outer loop uses 19 flops */
 968     }
 969
 970     /* Increment number of outer iterations */
 971     outeriter        += nri;
 972
 973     /* Update outer/inner flops */
 974
 975     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*288);
 976 }
 977 /*
 978  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW4W4_F_avx_256_single
 979  * Electrostatics interaction: ReactionField
 980  * VdW interaction:            None
 981  * Geometry:                   Water4-Water4
 982  * Calculate force/pot:        Force
 983  */
 984 void
 985 nb_kernel_ElecRF_VdwNone_GeomW4W4_F_avx_256_single
 986                     (t_nblist                    * gmx_restrict       nlist,
 987                      rvec                        * gmx_restrict          xx,
 988                      rvec                        * gmx_restrict          ff,
 989                      struct t_forcerec           * gmx_restrict          fr,
 990                      t_mdatoms                   * gmx_restrict     mdatoms,
 991                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 992                      t_nrnb                      * gmx_restrict        nrnb)
 993 {
 994     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 995      * just 0 for non-waters.
 996      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 997      * jnr indices corresponding to data put in the four positions in the SIMD register.
 998      */
 999     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1000     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1001     int              jnrA,jnrB,jnrC,jnrD;
1002     int              jnrE,jnrF,jnrG,jnrH;
1003     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1004     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1005     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1006     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1007     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1008     real             rcutoff_scalar;
1009     real             *shiftvec,*fshift,*x,*f;
1010     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1011     real             scratch[4*DIM];
1012     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1013     real *           vdwioffsetptr1;
1014     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1015     real *           vdwioffsetptr2;
1016     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1017     real *           vdwioffsetptr3;
1018     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1019     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1020     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1021     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1022     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1023     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1024     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1025     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1026     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1027     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1028     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1029     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1030     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1031     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1032     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1033     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1034     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1035     real             *charge;
1036     __m256           dummy_mask,cutoff_mask;
1037     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1038     __m256           one     = _mm256_set1_ps(1.0);
1039     __m256           two     = _mm256_set1_ps(2.0);
1040     x                = xx[0];
1041     f                = ff[0];
1042
1043     nri              = nlist->nri;
1044     iinr             = nlist->iinr;
1045     jindex           = nlist->jindex;
1046     jjnr             = nlist->jjnr;
1047     shiftidx         = nlist->shift;
1048     gid              = nlist->gid;
1049     shiftvec         = fr->shift_vec[0];
1050     fshift           = fr->fshift[0];
1051     facel            = _mm256_set1_ps(fr->ic->epsfac);
1052     charge           = mdatoms->chargeA;
1053     krf              = _mm256_set1_ps(fr->ic->k_rf);
1054     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1055     crf              = _mm256_set1_ps(fr->ic->c_rf);
1056
1057     /* Setup water-specific parameters */
1058     inr              = nlist->iinr[0];
1059     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1060     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1061     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1062
1063     jq1              = _mm256_set1_ps(charge[inr+1]);
1064     jq2              = _mm256_set1_ps(charge[inr+2]);
1065     jq3              = _mm256_set1_ps(charge[inr+3]);
1066     qq11             = _mm256_mul_ps(iq1,jq1);
1067     qq12             = _mm256_mul_ps(iq1,jq2);
1068     qq13             = _mm256_mul_ps(iq1,jq3);
1069     qq21             = _mm256_mul_ps(iq2,jq1);
1070     qq22             = _mm256_mul_ps(iq2,jq2);
1071     qq23             = _mm256_mul_ps(iq2,jq3);
1072     qq31             = _mm256_mul_ps(iq3,jq1);
1073     qq32             = _mm256_mul_ps(iq3,jq2);
1074     qq33             = _mm256_mul_ps(iq3,jq3);
1075
1076     /* Avoid stupid compiler warnings */
1077     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1078     j_coord_offsetA = 0;
1079     j_coord_offsetB = 0;
1080     j_coord_offsetC = 0;
1081     j_coord_offsetD = 0;
1082     j_coord_offsetE = 0;
1083     j_coord_offsetF = 0;
1084     j_coord_offsetG = 0;
1085     j_coord_offsetH = 0;
1086
1087     outeriter        = 0;
1088     inneriter        = 0;
1089
1090     for(iidx=0;iidx<4*DIM;iidx++)
1091     {
1092         scratch[iidx] = 0.0;
1093     }
1094
1095     /* Start outer loop over neighborlists */
1096     for(iidx=0; iidx<nri; iidx++)
1097     {
1098         /* Load shift vector for this list */
1099         i_shift_offset   = DIM*shiftidx[iidx];
1100
1101         /* Load limits for loop over neighbors */
1102         j_index_start    = jindex[iidx];
1103         j_index_end      = jindex[iidx+1];
1104
1105         /* Get outer coordinate index */
1106         inr              = iinr[iidx];
1107         i_coord_offset   = DIM*inr;
1108
1109         /* Load i particle coords and add shift vector */
1110         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1111                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1112
1113         fix1             = _mm256_setzero_ps();
1114         fiy1             = _mm256_setzero_ps();
1115         fiz1             = _mm256_setzero_ps();
1116         fix2             = _mm256_setzero_ps();
1117         fiy2             = _mm256_setzero_ps();
1118         fiz2             = _mm256_setzero_ps();
1119         fix3             = _mm256_setzero_ps();
1120         fiy3             = _mm256_setzero_ps();
1121         fiz3             = _mm256_setzero_ps();
1122
1123         /* Start inner kernel loop */
1124         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1125         {
1126
1127             /* Get j neighbor index, and coordinate index */
1128             jnrA             = jjnr[jidx];
1129             jnrB             = jjnr[jidx+1];
1130             jnrC             = jjnr[jidx+2];
1131             jnrD             = jjnr[jidx+3];
1132             jnrE             = jjnr[jidx+4];
1133             jnrF             = jjnr[jidx+5];
1134             jnrG             = jjnr[jidx+6];
1135             jnrH             = jjnr[jidx+7];
1136             j_coord_offsetA  = DIM*jnrA;
1137             j_coord_offsetB  = DIM*jnrB;
1138             j_coord_offsetC  = DIM*jnrC;
1139             j_coord_offsetD  = DIM*jnrD;
1140             j_coord_offsetE  = DIM*jnrE;
1141             j_coord_offsetF  = DIM*jnrF;
1142             j_coord_offsetG  = DIM*jnrG;
1143             j_coord_offsetH  = DIM*jnrH;
1144
1145             /* load j atom coordinates */
1146             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1147                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1148                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1149                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1150                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1151
1152             /* Calculate displacement vector */
1153             dx11             = _mm256_sub_ps(ix1,jx1);
1154             dy11             = _mm256_sub_ps(iy1,jy1);
1155             dz11             = _mm256_sub_ps(iz1,jz1);
1156             dx12             = _mm256_sub_ps(ix1,jx2);
1157             dy12             = _mm256_sub_ps(iy1,jy2);
1158             dz12             = _mm256_sub_ps(iz1,jz2);
1159             dx13             = _mm256_sub_ps(ix1,jx3);
1160             dy13             = _mm256_sub_ps(iy1,jy3);
1161             dz13             = _mm256_sub_ps(iz1,jz3);
1162             dx21             = _mm256_sub_ps(ix2,jx1);
1163             dy21             = _mm256_sub_ps(iy2,jy1);
1164             dz21             = _mm256_sub_ps(iz2,jz1);
1165             dx22             = _mm256_sub_ps(ix2,jx2);
1166             dy22             = _mm256_sub_ps(iy2,jy2);
1167             dz22             = _mm256_sub_ps(iz2,jz2);
1168             dx23             = _mm256_sub_ps(ix2,jx3);
1169             dy23             = _mm256_sub_ps(iy2,jy3);
1170             dz23             = _mm256_sub_ps(iz2,jz3);
1171             dx31             = _mm256_sub_ps(ix3,jx1);
1172             dy31             = _mm256_sub_ps(iy3,jy1);
1173             dz31             = _mm256_sub_ps(iz3,jz1);
1174             dx32             = _mm256_sub_ps(ix3,jx2);
1175             dy32             = _mm256_sub_ps(iy3,jy2);
1176             dz32             = _mm256_sub_ps(iz3,jz2);
1177             dx33             = _mm256_sub_ps(ix3,jx3);
1178             dy33             = _mm256_sub_ps(iy3,jy3);
1179             dz33             = _mm256_sub_ps(iz3,jz3);
1180
1181             /* Calculate squared distance and things based on it */
1182             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1183             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1184             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1185             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1186             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1187             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1188             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1189             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1190             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1191
1192             rinv11           = avx256_invsqrt_f(rsq11);
1193             rinv12           = avx256_invsqrt_f(rsq12);
1194             rinv13           = avx256_invsqrt_f(rsq13);
1195             rinv21           = avx256_invsqrt_f(rsq21);
1196             rinv22           = avx256_invsqrt_f(rsq22);
1197             rinv23           = avx256_invsqrt_f(rsq23);
1198             rinv31           = avx256_invsqrt_f(rsq31);
1199             rinv32           = avx256_invsqrt_f(rsq32);
1200             rinv33           = avx256_invsqrt_f(rsq33);
1201
1202             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1203             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1204             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1205             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1206             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1207             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1208             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1209             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1210             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1211
1212             fjx1             = _mm256_setzero_ps();
1213             fjy1             = _mm256_setzero_ps();
1214             fjz1             = _mm256_setzero_ps();
1215             fjx2             = _mm256_setzero_ps();
1216             fjy2             = _mm256_setzero_ps();
1217             fjz2             = _mm256_setzero_ps();
1218             fjx3             = _mm256_setzero_ps();
1219             fjy3             = _mm256_setzero_ps();
1220             fjz3             = _mm256_setzero_ps();
1221
1222             /**************************
1223              * CALCULATE INTERACTIONS *
1224              **************************/
1225
1226             /* REACTION-FIELD ELECTROSTATICS */
1227             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1228
1229             fscal            = felec;
1230
1231             /* Calculate temporary vectorial force */
1232             tx               = _mm256_mul_ps(fscal,dx11);
1233             ty               = _mm256_mul_ps(fscal,dy11);
1234             tz               = _mm256_mul_ps(fscal,dz11);
1235
1236             /* Update vectorial force */
1237             fix1             = _mm256_add_ps(fix1,tx);
1238             fiy1             = _mm256_add_ps(fiy1,ty);
1239             fiz1             = _mm256_add_ps(fiz1,tz);
1240
1241             fjx1             = _mm256_add_ps(fjx1,tx);
1242             fjy1             = _mm256_add_ps(fjy1,ty);
1243             fjz1             = _mm256_add_ps(fjz1,tz);
1244
1245             /**************************
1246              * CALCULATE INTERACTIONS *
1247              **************************/
1248
1249             /* REACTION-FIELD ELECTROSTATICS */
1250             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1251
1252             fscal            = felec;
1253
1254             /* Calculate temporary vectorial force */
1255             tx               = _mm256_mul_ps(fscal,dx12);
1256             ty               = _mm256_mul_ps(fscal,dy12);
1257             tz               = _mm256_mul_ps(fscal,dz12);
1258
1259             /* Update vectorial force */
1260             fix1             = _mm256_add_ps(fix1,tx);
1261             fiy1             = _mm256_add_ps(fiy1,ty);
1262             fiz1             = _mm256_add_ps(fiz1,tz);
1263
1264             fjx2             = _mm256_add_ps(fjx2,tx);
1265             fjy2             = _mm256_add_ps(fjy2,ty);
1266             fjz2             = _mm256_add_ps(fjz2,tz);
1267
1268             /**************************
1269              * CALCULATE INTERACTIONS *
1270              **************************/
1271
1272             /* REACTION-FIELD ELECTROSTATICS */
1273             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1274
1275             fscal            = felec;
1276
1277             /* Calculate temporary vectorial force */
1278             tx               = _mm256_mul_ps(fscal,dx13);
1279             ty               = _mm256_mul_ps(fscal,dy13);
1280             tz               = _mm256_mul_ps(fscal,dz13);
1281
1282             /* Update vectorial force */
1283             fix1             = _mm256_add_ps(fix1,tx);
1284             fiy1             = _mm256_add_ps(fiy1,ty);
1285             fiz1             = _mm256_add_ps(fiz1,tz);
1286
1287             fjx3             = _mm256_add_ps(fjx3,tx);
1288             fjy3             = _mm256_add_ps(fjy3,ty);
1289             fjz3             = _mm256_add_ps(fjz3,tz);
1290
1291             /**************************
1292              * CALCULATE INTERACTIONS *
1293              **************************/
1294
1295             /* REACTION-FIELD ELECTROSTATICS */
1296             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1297
1298             fscal            = felec;
1299
1300             /* Calculate temporary vectorial force */
1301             tx               = _mm256_mul_ps(fscal,dx21);
1302             ty               = _mm256_mul_ps(fscal,dy21);
1303             tz               = _mm256_mul_ps(fscal,dz21);
1304
1305             /* Update vectorial force */
1306             fix2             = _mm256_add_ps(fix2,tx);
1307             fiy2             = _mm256_add_ps(fiy2,ty);
1308             fiz2             = _mm256_add_ps(fiz2,tz);
1309
1310             fjx1             = _mm256_add_ps(fjx1,tx);
1311             fjy1             = _mm256_add_ps(fjy1,ty);
1312             fjz1             = _mm256_add_ps(fjz1,tz);
1313
1314             /**************************
1315              * CALCULATE INTERACTIONS *
1316              **************************/
1317
1318             /* REACTION-FIELD ELECTROSTATICS */
1319             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1320
1321             fscal            = felec;
1322
1323             /* Calculate temporary vectorial force */
1324             tx               = _mm256_mul_ps(fscal,dx22);
1325             ty               = _mm256_mul_ps(fscal,dy22);
1326             tz               = _mm256_mul_ps(fscal,dz22);
1327
1328             /* Update vectorial force */
1329             fix2             = _mm256_add_ps(fix2,tx);
1330             fiy2             = _mm256_add_ps(fiy2,ty);
1331             fiz2             = _mm256_add_ps(fiz2,tz);
1332
1333             fjx2             = _mm256_add_ps(fjx2,tx);
1334             fjy2             = _mm256_add_ps(fjy2,ty);
1335             fjz2             = _mm256_add_ps(fjz2,tz);
1336
1337             /**************************
1338              * CALCULATE INTERACTIONS *
1339              **************************/
1340
1341             /* REACTION-FIELD ELECTROSTATICS */
1342             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1343
1344             fscal            = felec;
1345
1346             /* Calculate temporary vectorial force */
1347             tx               = _mm256_mul_ps(fscal,dx23);
1348             ty               = _mm256_mul_ps(fscal,dy23);
1349             tz               = _mm256_mul_ps(fscal,dz23);
1350
1351             /* Update vectorial force */
1352             fix2             = _mm256_add_ps(fix2,tx);
1353             fiy2             = _mm256_add_ps(fiy2,ty);
1354             fiz2             = _mm256_add_ps(fiz2,tz);
1355
1356             fjx3             = _mm256_add_ps(fjx3,tx);
1357             fjy3             = _mm256_add_ps(fjy3,ty);
1358             fjz3             = _mm256_add_ps(fjz3,tz);
1359
1360             /**************************
1361              * CALCULATE INTERACTIONS *
1362              **************************/
1363
1364             /* REACTION-FIELD ELECTROSTATICS */
1365             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1366
1367             fscal            = felec;
1368
1369             /* Calculate temporary vectorial force */
1370             tx               = _mm256_mul_ps(fscal,dx31);
1371             ty               = _mm256_mul_ps(fscal,dy31);
1372             tz               = _mm256_mul_ps(fscal,dz31);
1373
1374             /* Update vectorial force */
1375             fix3             = _mm256_add_ps(fix3,tx);
1376             fiy3             = _mm256_add_ps(fiy3,ty);
1377             fiz3             = _mm256_add_ps(fiz3,tz);
1378
1379             fjx1             = _mm256_add_ps(fjx1,tx);
1380             fjy1             = _mm256_add_ps(fjy1,ty);
1381             fjz1             = _mm256_add_ps(fjz1,tz);
1382
1383             /**************************
1384              * CALCULATE INTERACTIONS *
1385              **************************/
1386
1387             /* REACTION-FIELD ELECTROSTATICS */
1388             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1389
1390             fscal            = felec;
1391
1392             /* Calculate temporary vectorial force */
1393             tx               = _mm256_mul_ps(fscal,dx32);
1394             ty               = _mm256_mul_ps(fscal,dy32);
1395             tz               = _mm256_mul_ps(fscal,dz32);
1396
1397             /* Update vectorial force */
1398             fix3             = _mm256_add_ps(fix3,tx);
1399             fiy3             = _mm256_add_ps(fiy3,ty);
1400             fiz3             = _mm256_add_ps(fiz3,tz);
1401
1402             fjx2             = _mm256_add_ps(fjx2,tx);
1403             fjy2             = _mm256_add_ps(fjy2,ty);
1404             fjz2             = _mm256_add_ps(fjz2,tz);
1405
1406             /**************************
1407              * CALCULATE INTERACTIONS *
1408              **************************/
1409
1410             /* REACTION-FIELD ELECTROSTATICS */
1411             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1412
1413             fscal            = felec;
1414
1415             /* Calculate temporary vectorial force */
1416             tx               = _mm256_mul_ps(fscal,dx33);
1417             ty               = _mm256_mul_ps(fscal,dy33);
1418             tz               = _mm256_mul_ps(fscal,dz33);
1419
1420             /* Update vectorial force */
1421             fix3             = _mm256_add_ps(fix3,tx);
1422             fiy3             = _mm256_add_ps(fiy3,ty);
1423             fiz3             = _mm256_add_ps(fiz3,tz);
1424
1425             fjx3             = _mm256_add_ps(fjx3,tx);
1426             fjy3             = _mm256_add_ps(fjy3,ty);
1427             fjz3             = _mm256_add_ps(fjz3,tz);
1428
1429             fjptrA             = f+j_coord_offsetA;
1430             fjptrB             = f+j_coord_offsetB;
1431             fjptrC             = f+j_coord_offsetC;
1432             fjptrD             = f+j_coord_offsetD;
1433             fjptrE             = f+j_coord_offsetE;
1434             fjptrF             = f+j_coord_offsetF;
1435             fjptrG             = f+j_coord_offsetG;
1436             fjptrH             = f+j_coord_offsetH;
1437
1438             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1439                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1440                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1441
1442             /* Inner loop uses 243 flops */
1443         }
1444
1445         if(jidx<j_index_end)
1446         {
1447
1448             /* Get j neighbor index, and coordinate index */
1449             jnrlistA         = jjnr[jidx];
1450             jnrlistB         = jjnr[jidx+1];
1451             jnrlistC         = jjnr[jidx+2];
1452             jnrlistD         = jjnr[jidx+3];
1453             jnrlistE         = jjnr[jidx+4];
1454             jnrlistF         = jjnr[jidx+5];
1455             jnrlistG         = jjnr[jidx+6];
1456             jnrlistH         = jjnr[jidx+7];
1457             /* Sign of each element will be negative for non-real atoms.
1458              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1459              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1460              */
1461             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1462                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1463
1464             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1465             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1466             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1467             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1468             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1469             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1470             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1471             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1472             j_coord_offsetA  = DIM*jnrA;
1473             j_coord_offsetB  = DIM*jnrB;
1474             j_coord_offsetC  = DIM*jnrC;
1475             j_coord_offsetD  = DIM*jnrD;
1476             j_coord_offsetE  = DIM*jnrE;
1477             j_coord_offsetF  = DIM*jnrF;
1478             j_coord_offsetG  = DIM*jnrG;
1479             j_coord_offsetH  = DIM*jnrH;
1480
1481             /* load j atom coordinates */
1482             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1483                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1484                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1485                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1486                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1487
1488             /* Calculate displacement vector */
1489             dx11             = _mm256_sub_ps(ix1,jx1);
1490             dy11             = _mm256_sub_ps(iy1,jy1);
1491             dz11             = _mm256_sub_ps(iz1,jz1);
1492             dx12             = _mm256_sub_ps(ix1,jx2);
1493             dy12             = _mm256_sub_ps(iy1,jy2);
1494             dz12             = _mm256_sub_ps(iz1,jz2);
1495             dx13             = _mm256_sub_ps(ix1,jx3);
1496             dy13             = _mm256_sub_ps(iy1,jy3);
1497             dz13             = _mm256_sub_ps(iz1,jz3);
1498             dx21             = _mm256_sub_ps(ix2,jx1);
1499             dy21             = _mm256_sub_ps(iy2,jy1);
1500             dz21             = _mm256_sub_ps(iz2,jz1);
1501             dx22             = _mm256_sub_ps(ix2,jx2);
1502             dy22             = _mm256_sub_ps(iy2,jy2);
1503             dz22             = _mm256_sub_ps(iz2,jz2);
1504             dx23             = _mm256_sub_ps(ix2,jx3);
1505             dy23             = _mm256_sub_ps(iy2,jy3);
1506             dz23             = _mm256_sub_ps(iz2,jz3);
1507             dx31             = _mm256_sub_ps(ix3,jx1);
1508             dy31             = _mm256_sub_ps(iy3,jy1);
1509             dz31             = _mm256_sub_ps(iz3,jz1);
1510             dx32             = _mm256_sub_ps(ix3,jx2);
1511             dy32             = _mm256_sub_ps(iy3,jy2);
1512             dz32             = _mm256_sub_ps(iz3,jz2);
1513             dx33             = _mm256_sub_ps(ix3,jx3);
1514             dy33             = _mm256_sub_ps(iy3,jy3);
1515             dz33             = _mm256_sub_ps(iz3,jz3);
1516
1517             /* Calculate squared distance and things based on it */
1518             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1519             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1520             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1521             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1522             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1523             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1524             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1525             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1526             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1527
1528             rinv11           = avx256_invsqrt_f(rsq11);
1529             rinv12           = avx256_invsqrt_f(rsq12);
1530             rinv13           = avx256_invsqrt_f(rsq13);
1531             rinv21           = avx256_invsqrt_f(rsq21);
1532             rinv22           = avx256_invsqrt_f(rsq22);
1533             rinv23           = avx256_invsqrt_f(rsq23);
1534             rinv31           = avx256_invsqrt_f(rsq31);
1535             rinv32           = avx256_invsqrt_f(rsq32);
1536             rinv33           = avx256_invsqrt_f(rsq33);
1537
1538             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1539             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1540             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1541             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1542             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1543             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1544             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1545             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1546             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1547
1548             fjx1             = _mm256_setzero_ps();
1549             fjy1             = _mm256_setzero_ps();
1550             fjz1             = _mm256_setzero_ps();
1551             fjx2             = _mm256_setzero_ps();
1552             fjy2             = _mm256_setzero_ps();
1553             fjz2             = _mm256_setzero_ps();
1554             fjx3             = _mm256_setzero_ps();
1555             fjy3             = _mm256_setzero_ps();
1556             fjz3             = _mm256_setzero_ps();
1557
1558             /**************************
1559              * CALCULATE INTERACTIONS *
1560              **************************/
1561
1562             /* REACTION-FIELD ELECTROSTATICS */
1563             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1564
1565             fscal            = felec;
1566
1567             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1568
1569             /* Calculate temporary vectorial force */
1570             tx               = _mm256_mul_ps(fscal,dx11);
1571             ty               = _mm256_mul_ps(fscal,dy11);
1572             tz               = _mm256_mul_ps(fscal,dz11);
1573
1574             /* Update vectorial force */
1575             fix1             = _mm256_add_ps(fix1,tx);
1576             fiy1             = _mm256_add_ps(fiy1,ty);
1577             fiz1             = _mm256_add_ps(fiz1,tz);
1578
1579             fjx1             = _mm256_add_ps(fjx1,tx);
1580             fjy1             = _mm256_add_ps(fjy1,ty);
1581             fjz1             = _mm256_add_ps(fjz1,tz);
1582
1583             /**************************
1584              * CALCULATE INTERACTIONS *
1585              **************************/
1586
1587             /* REACTION-FIELD ELECTROSTATICS */
1588             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1589
1590             fscal            = felec;
1591
1592             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1593
1594             /* Calculate temporary vectorial force */
1595             tx               = _mm256_mul_ps(fscal,dx12);
1596             ty               = _mm256_mul_ps(fscal,dy12);
1597             tz               = _mm256_mul_ps(fscal,dz12);
1598
1599             /* Update vectorial force */
1600             fix1             = _mm256_add_ps(fix1,tx);
1601             fiy1             = _mm256_add_ps(fiy1,ty);
1602             fiz1             = _mm256_add_ps(fiz1,tz);
1603
1604             fjx2             = _mm256_add_ps(fjx2,tx);
1605             fjy2             = _mm256_add_ps(fjy2,ty);
1606             fjz2             = _mm256_add_ps(fjz2,tz);
1607
1608             /**************************
1609              * CALCULATE INTERACTIONS *
1610              **************************/
1611
1612             /* REACTION-FIELD ELECTROSTATICS */
1613             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1614
1615             fscal            = felec;
1616
1617             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1618
1619             /* Calculate temporary vectorial force */
1620             tx               = _mm256_mul_ps(fscal,dx13);
1621             ty               = _mm256_mul_ps(fscal,dy13);
1622             tz               = _mm256_mul_ps(fscal,dz13);
1623
1624             /* Update vectorial force */
1625             fix1             = _mm256_add_ps(fix1,tx);
1626             fiy1             = _mm256_add_ps(fiy1,ty);
1627             fiz1             = _mm256_add_ps(fiz1,tz);
1628
1629             fjx3             = _mm256_add_ps(fjx3,tx);
1630             fjy3             = _mm256_add_ps(fjy3,ty);
1631             fjz3             = _mm256_add_ps(fjz3,tz);
1632
1633             /**************************
1634              * CALCULATE INTERACTIONS *
1635              **************************/
1636
1637             /* REACTION-FIELD ELECTROSTATICS */
1638             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1639
1640             fscal            = felec;
1641
1642             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1643
1644             /* Calculate temporary vectorial force */
1645             tx               = _mm256_mul_ps(fscal,dx21);
1646             ty               = _mm256_mul_ps(fscal,dy21);
1647             tz               = _mm256_mul_ps(fscal,dz21);
1648
1649             /* Update vectorial force */
1650             fix2             = _mm256_add_ps(fix2,tx);
1651             fiy2             = _mm256_add_ps(fiy2,ty);
1652             fiz2             = _mm256_add_ps(fiz2,tz);
1653
1654             fjx1             = _mm256_add_ps(fjx1,tx);
1655             fjy1             = _mm256_add_ps(fjy1,ty);
1656             fjz1             = _mm256_add_ps(fjz1,tz);
1657
1658             /**************************
1659              * CALCULATE INTERACTIONS *
1660              **************************/
1661
1662             /* REACTION-FIELD ELECTROSTATICS */
1663             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1664
1665             fscal            = felec;
1666
1667             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1668
1669             /* Calculate temporary vectorial force */
1670             tx               = _mm256_mul_ps(fscal,dx22);
1671             ty               = _mm256_mul_ps(fscal,dy22);
1672             tz               = _mm256_mul_ps(fscal,dz22);
1673
1674             /* Update vectorial force */
1675             fix2             = _mm256_add_ps(fix2,tx);
1676             fiy2             = _mm256_add_ps(fiy2,ty);
1677             fiz2             = _mm256_add_ps(fiz2,tz);
1678
1679             fjx2             = _mm256_add_ps(fjx2,tx);
1680             fjy2             = _mm256_add_ps(fjy2,ty);
1681             fjz2             = _mm256_add_ps(fjz2,tz);
1682
1683             /**************************
1684              * CALCULATE INTERACTIONS *
1685              **************************/
1686
1687             /* REACTION-FIELD ELECTROSTATICS */
1688             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1689
1690             fscal            = felec;
1691
1692             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1693
1694             /* Calculate temporary vectorial force */
1695             tx               = _mm256_mul_ps(fscal,dx23);
1696             ty               = _mm256_mul_ps(fscal,dy23);
1697             tz               = _mm256_mul_ps(fscal,dz23);
1698
1699             /* Update vectorial force */
1700             fix2             = _mm256_add_ps(fix2,tx);
1701             fiy2             = _mm256_add_ps(fiy2,ty);
1702             fiz2             = _mm256_add_ps(fiz2,tz);
1703
1704             fjx3             = _mm256_add_ps(fjx3,tx);
1705             fjy3             = _mm256_add_ps(fjy3,ty);
1706             fjz3             = _mm256_add_ps(fjz3,tz);
1707
1708             /**************************
1709              * CALCULATE INTERACTIONS *
1710              **************************/
1711
1712             /* REACTION-FIELD ELECTROSTATICS */
1713             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1714
1715             fscal            = felec;
1716
1717             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1718
1719             /* Calculate temporary vectorial force */
1720             tx               = _mm256_mul_ps(fscal,dx31);
1721             ty               = _mm256_mul_ps(fscal,dy31);
1722             tz               = _mm256_mul_ps(fscal,dz31);
1723
1724             /* Update vectorial force */
1725             fix3             = _mm256_add_ps(fix3,tx);
1726             fiy3             = _mm256_add_ps(fiy3,ty);
1727             fiz3             = _mm256_add_ps(fiz3,tz);
1728
1729             fjx1             = _mm256_add_ps(fjx1,tx);
1730             fjy1             = _mm256_add_ps(fjy1,ty);
1731             fjz1             = _mm256_add_ps(fjz1,tz);
1732
1733             /**************************
1734              * CALCULATE INTERACTIONS *
1735              **************************/
1736
1737             /* REACTION-FIELD ELECTROSTATICS */
1738             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
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,dx32);
1746             ty               = _mm256_mul_ps(fscal,dy32);
1747             tz               = _mm256_mul_ps(fscal,dz32);
1748
1749             /* Update vectorial force */
1750             fix3             = _mm256_add_ps(fix3,tx);
1751             fiy3             = _mm256_add_ps(fiy3,ty);
1752             fiz3             = _mm256_add_ps(fiz3,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             /* REACTION-FIELD ELECTROSTATICS */
1763             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1764
1765             fscal            = felec;
1766
1767             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1768
1769             /* Calculate temporary vectorial force */
1770             tx               = _mm256_mul_ps(fscal,dx33);
1771             ty               = _mm256_mul_ps(fscal,dy33);
1772             tz               = _mm256_mul_ps(fscal,dz33);
1773
1774             /* Update vectorial force */
1775             fix3             = _mm256_add_ps(fix3,tx);
1776             fiy3             = _mm256_add_ps(fiy3,ty);
1777             fiz3             = _mm256_add_ps(fiz3,tz);
1778
1779             fjx3             = _mm256_add_ps(fjx3,tx);
1780             fjy3             = _mm256_add_ps(fjy3,ty);
1781             fjz3             = _mm256_add_ps(fjz3,tz);
1782
1783             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1784             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1785             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1786             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1787             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1788             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1789             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1790             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1791
1792             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1793                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1794                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1795
1796             /* Inner loop uses 243 flops */
1797         }
1798
1799         /* End of innermost loop */
1800
1801         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1802                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1803
1804         /* Increment number of inner iterations */
1805         inneriter                  += j_index_end - j_index_start;
1806
1807         /* Outer loop uses 18 flops */
1808     }
1809
1810     /* Increment number of outer iterations */
1811     outeriter        += nri;
1812
1813     /* Update outer/inner flops */
1814
1815     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*243);
1816 }