src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRF_VdwNone_GeomW3W3_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   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/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_single.h"
  50 #include "kernelutil_x86_avx_256_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_avx_256_single
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            None
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRF_VdwNone_GeomW3W3_VF_avx_256_single
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrE,jnrF,jnrG,jnrH;
  78     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  79     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  80     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  81     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  82     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  83     real             rcutoff_scalar;
  84     real             *shiftvec,*fshift,*x,*f;
  85     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  86     real             scratch[4*DIM];
  87     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  88     real *           vdwioffsetptr0;
  89     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  90     real *           vdwioffsetptr1;
  91     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  92     real *           vdwioffsetptr2;
  93     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  97     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  99     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 101     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 102     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 103     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 104     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 105     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 106     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 107     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 108     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 109     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 110     real             *charge;
 111     __m256           dummy_mask,cutoff_mask;
 112     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 113     __m256           one     = _mm256_set1_ps(1.0);
 114     __m256           two     = _mm256_set1_ps(2.0);
 115     x                = xx[0];
 116     f                = ff[0];
 117
 118     nri              = nlist->nri;
 119     iinr             = nlist->iinr;
 120     jindex           = nlist->jindex;
 121     jjnr             = nlist->jjnr;
 122     shiftidx         = nlist->shift;
 123     gid              = nlist->gid;
 124     shiftvec         = fr->shift_vec[0];
 125     fshift           = fr->fshift[0];
 126     facel            = _mm256_set1_ps(fr->epsfac);
 127     charge           = mdatoms->chargeA;
 128     krf              = _mm256_set1_ps(fr->ic->k_rf);
 129     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 130     crf              = _mm256_set1_ps(fr->ic->c_rf);
 131
 132     /* Setup water-specific parameters */
 133     inr              = nlist->iinr[0];
 134     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 135     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 136     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 137
 138     jq0              = _mm256_set1_ps(charge[inr+0]);
 139     jq1              = _mm256_set1_ps(charge[inr+1]);
 140     jq2              = _mm256_set1_ps(charge[inr+2]);
 141     qq00             = _mm256_mul_ps(iq0,jq0);
 142     qq01             = _mm256_mul_ps(iq0,jq1);
 143     qq02             = _mm256_mul_ps(iq0,jq2);
 144     qq10             = _mm256_mul_ps(iq1,jq0);
 145     qq11             = _mm256_mul_ps(iq1,jq1);
 146     qq12             = _mm256_mul_ps(iq1,jq2);
 147     qq20             = _mm256_mul_ps(iq2,jq0);
 148     qq21             = _mm256_mul_ps(iq2,jq1);
 149     qq22             = _mm256_mul_ps(iq2,jq2);
 150
 151     /* Avoid stupid compiler warnings */
 152     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 153     j_coord_offsetA = 0;
 154     j_coord_offsetB = 0;
 155     j_coord_offsetC = 0;
 156     j_coord_offsetD = 0;
 157     j_coord_offsetE = 0;
 158     j_coord_offsetF = 0;
 159     j_coord_offsetG = 0;
 160     j_coord_offsetH = 0;
 161
 162     outeriter        = 0;
 163     inneriter        = 0;
 164
 165     for(iidx=0;iidx<4*DIM;iidx++)
 166     {
 167         scratch[iidx] = 0.0;
 168     }
 169
 170     /* Start outer loop over neighborlists */
 171     for(iidx=0; iidx<nri; iidx++)
 172     {
 173         /* Load shift vector for this list */
 174         i_shift_offset   = DIM*shiftidx[iidx];
 175
 176         /* Load limits for loop over neighbors */
 177         j_index_start    = jindex[iidx];
 178         j_index_end      = jindex[iidx+1];
 179
 180         /* Get outer coordinate index */
 181         inr              = iinr[iidx];
 182         i_coord_offset   = DIM*inr;
 183
 184         /* Load i particle coords and add shift vector */
 185         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 186                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 187
 188         fix0             = _mm256_setzero_ps();
 189         fiy0             = _mm256_setzero_ps();
 190         fiz0             = _mm256_setzero_ps();
 191         fix1             = _mm256_setzero_ps();
 192         fiy1             = _mm256_setzero_ps();
 193         fiz1             = _mm256_setzero_ps();
 194         fix2             = _mm256_setzero_ps();
 195         fiy2             = _mm256_setzero_ps();
 196         fiz2             = _mm256_setzero_ps();
 197
 198         /* Reset potential sums */
 199         velecsum         = _mm256_setzero_ps();
 200
 201         /* Start inner kernel loop */
 202         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 203         {
 204
 205             /* Get j neighbor index, and coordinate index */
 206             jnrA             = jjnr[jidx];
 207             jnrB             = jjnr[jidx+1];
 208             jnrC             = jjnr[jidx+2];
 209             jnrD             = jjnr[jidx+3];
 210             jnrE             = jjnr[jidx+4];
 211             jnrF             = jjnr[jidx+5];
 212             jnrG             = jjnr[jidx+6];
 213             jnrH             = jjnr[jidx+7];
 214             j_coord_offsetA  = DIM*jnrA;
 215             j_coord_offsetB  = DIM*jnrB;
 216             j_coord_offsetC  = DIM*jnrC;
 217             j_coord_offsetD  = DIM*jnrD;
 218             j_coord_offsetE  = DIM*jnrE;
 219             j_coord_offsetF  = DIM*jnrF;
 220             j_coord_offsetG  = DIM*jnrG;
 221             j_coord_offsetH  = DIM*jnrH;
 222
 223             /* load j atom coordinates */
 224             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 225                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 226                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 227                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 228                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 229
 230             /* Calculate displacement vector */
 231             dx00             = _mm256_sub_ps(ix0,jx0);
 232             dy00             = _mm256_sub_ps(iy0,jy0);
 233             dz00             = _mm256_sub_ps(iz0,jz0);
 234             dx01             = _mm256_sub_ps(ix0,jx1);
 235             dy01             = _mm256_sub_ps(iy0,jy1);
 236             dz01             = _mm256_sub_ps(iz0,jz1);
 237             dx02             = _mm256_sub_ps(ix0,jx2);
 238             dy02             = _mm256_sub_ps(iy0,jy2);
 239             dz02             = _mm256_sub_ps(iz0,jz2);
 240             dx10             = _mm256_sub_ps(ix1,jx0);
 241             dy10             = _mm256_sub_ps(iy1,jy0);
 242             dz10             = _mm256_sub_ps(iz1,jz0);
 243             dx11             = _mm256_sub_ps(ix1,jx1);
 244             dy11             = _mm256_sub_ps(iy1,jy1);
 245             dz11             = _mm256_sub_ps(iz1,jz1);
 246             dx12             = _mm256_sub_ps(ix1,jx2);
 247             dy12             = _mm256_sub_ps(iy1,jy2);
 248             dz12             = _mm256_sub_ps(iz1,jz2);
 249             dx20             = _mm256_sub_ps(ix2,jx0);
 250             dy20             = _mm256_sub_ps(iy2,jy0);
 251             dz20             = _mm256_sub_ps(iz2,jz0);
 252             dx21             = _mm256_sub_ps(ix2,jx1);
 253             dy21             = _mm256_sub_ps(iy2,jy1);
 254             dz21             = _mm256_sub_ps(iz2,jz1);
 255             dx22             = _mm256_sub_ps(ix2,jx2);
 256             dy22             = _mm256_sub_ps(iy2,jy2);
 257             dz22             = _mm256_sub_ps(iz2,jz2);
 258
 259             /* Calculate squared distance and things based on it */
 260             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 261             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 262             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 263             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 264             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 265             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 266             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 267             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 268             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 269
 270             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 271             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 272             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 273             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 274             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 275             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 276             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 277             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 278             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 279
 280             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 281             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 282             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 283             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 284             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 285             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 286             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 287             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 288             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 289
 290             fjx0             = _mm256_setzero_ps();
 291             fjy0             = _mm256_setzero_ps();
 292             fjz0             = _mm256_setzero_ps();
 293             fjx1             = _mm256_setzero_ps();
 294             fjy1             = _mm256_setzero_ps();
 295             fjz1             = _mm256_setzero_ps();
 296             fjx2             = _mm256_setzero_ps();
 297             fjy2             = _mm256_setzero_ps();
 298             fjz2             = _mm256_setzero_ps();
 299
 300             /**************************
 301              * CALCULATE INTERACTIONS *
 302              **************************/
 303
 304             /* REACTION-FIELD ELECTROSTATICS */
 305             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 306             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 307
 308             /* Update potential sum for this i atom from the interaction with this j atom. */
 309             velecsum         = _mm256_add_ps(velecsum,velec);
 310
 311             fscal            = felec;
 312
 313             /* Calculate temporary vectorial force */
 314             tx               = _mm256_mul_ps(fscal,dx00);
 315             ty               = _mm256_mul_ps(fscal,dy00);
 316             tz               = _mm256_mul_ps(fscal,dz00);
 317
 318             /* Update vectorial force */
 319             fix0             = _mm256_add_ps(fix0,tx);
 320             fiy0             = _mm256_add_ps(fiy0,ty);
 321             fiz0             = _mm256_add_ps(fiz0,tz);
 322
 323             fjx0             = _mm256_add_ps(fjx0,tx);
 324             fjy0             = _mm256_add_ps(fjy0,ty);
 325             fjz0             = _mm256_add_ps(fjz0,tz);
 326
 327             /**************************
 328              * CALCULATE INTERACTIONS *
 329              **************************/
 330
 331             /* REACTION-FIELD ELECTROSTATICS */
 332             velec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
 333             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
 334
 335             /* Update potential sum for this i atom from the interaction with this j atom. */
 336             velecsum         = _mm256_add_ps(velecsum,velec);
 337
 338             fscal            = felec;
 339
 340             /* Calculate temporary vectorial force */
 341             tx               = _mm256_mul_ps(fscal,dx01);
 342             ty               = _mm256_mul_ps(fscal,dy01);
 343             tz               = _mm256_mul_ps(fscal,dz01);
 344
 345             /* Update vectorial force */
 346             fix0             = _mm256_add_ps(fix0,tx);
 347             fiy0             = _mm256_add_ps(fiy0,ty);
 348             fiz0             = _mm256_add_ps(fiz0,tz);
 349
 350             fjx1             = _mm256_add_ps(fjx1,tx);
 351             fjy1             = _mm256_add_ps(fjy1,ty);
 352             fjz1             = _mm256_add_ps(fjz1,tz);
 353
 354             /**************************
 355              * CALCULATE INTERACTIONS *
 356              **************************/
 357
 358             /* REACTION-FIELD ELECTROSTATICS */
 359             velec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
 360             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
 361
 362             /* Update potential sum for this i atom from the interaction with this j atom. */
 363             velecsum         = _mm256_add_ps(velecsum,velec);
 364
 365             fscal            = felec;
 366
 367             /* Calculate temporary vectorial force */
 368             tx               = _mm256_mul_ps(fscal,dx02);
 369             ty               = _mm256_mul_ps(fscal,dy02);
 370             tz               = _mm256_mul_ps(fscal,dz02);
 371
 372             /* Update vectorial force */
 373             fix0             = _mm256_add_ps(fix0,tx);
 374             fiy0             = _mm256_add_ps(fiy0,ty);
 375             fiz0             = _mm256_add_ps(fiz0,tz);
 376
 377             fjx2             = _mm256_add_ps(fjx2,tx);
 378             fjy2             = _mm256_add_ps(fjy2,ty);
 379             fjz2             = _mm256_add_ps(fjz2,tz);
 380
 381             /**************************
 382              * CALCULATE INTERACTIONS *
 383              **************************/
 384
 385             /* REACTION-FIELD ELECTROSTATICS */
 386             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 387             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 388
 389             /* Update potential sum for this i atom from the interaction with this j atom. */
 390             velecsum         = _mm256_add_ps(velecsum,velec);
 391
 392             fscal            = felec;
 393
 394             /* Calculate temporary vectorial force */
 395             tx               = _mm256_mul_ps(fscal,dx10);
 396             ty               = _mm256_mul_ps(fscal,dy10);
 397             tz               = _mm256_mul_ps(fscal,dz10);
 398
 399             /* Update vectorial force */
 400             fix1             = _mm256_add_ps(fix1,tx);
 401             fiy1             = _mm256_add_ps(fiy1,ty);
 402             fiz1             = _mm256_add_ps(fiz1,tz);
 403
 404             fjx0             = _mm256_add_ps(fjx0,tx);
 405             fjy0             = _mm256_add_ps(fjy0,ty);
 406             fjz0             = _mm256_add_ps(fjz0,tz);
 407
 408             /**************************
 409              * CALCULATE INTERACTIONS *
 410              **************************/
 411
 412             /* REACTION-FIELD ELECTROSTATICS */
 413             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 414             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 415
 416             /* Update potential sum for this i atom from the interaction with this j atom. */
 417             velecsum         = _mm256_add_ps(velecsum,velec);
 418
 419             fscal            = felec;
 420
 421             /* Calculate temporary vectorial force */
 422             tx               = _mm256_mul_ps(fscal,dx11);
 423             ty               = _mm256_mul_ps(fscal,dy11);
 424             tz               = _mm256_mul_ps(fscal,dz11);
 425
 426             /* Update vectorial force */
 427             fix1             = _mm256_add_ps(fix1,tx);
 428             fiy1             = _mm256_add_ps(fiy1,ty);
 429             fiz1             = _mm256_add_ps(fiz1,tz);
 430
 431             fjx1             = _mm256_add_ps(fjx1,tx);
 432             fjy1             = _mm256_add_ps(fjy1,ty);
 433             fjz1             = _mm256_add_ps(fjz1,tz);
 434
 435             /**************************
 436              * CALCULATE INTERACTIONS *
 437              **************************/
 438
 439             /* REACTION-FIELD ELECTROSTATICS */
 440             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 441             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 442
 443             /* Update potential sum for this i atom from the interaction with this j atom. */
 444             velecsum         = _mm256_add_ps(velecsum,velec);
 445
 446             fscal            = felec;
 447
 448             /* Calculate temporary vectorial force */
 449             tx               = _mm256_mul_ps(fscal,dx12);
 450             ty               = _mm256_mul_ps(fscal,dy12);
 451             tz               = _mm256_mul_ps(fscal,dz12);
 452
 453             /* Update vectorial force */
 454             fix1             = _mm256_add_ps(fix1,tx);
 455             fiy1             = _mm256_add_ps(fiy1,ty);
 456             fiz1             = _mm256_add_ps(fiz1,tz);
 457
 458             fjx2             = _mm256_add_ps(fjx2,tx);
 459             fjy2             = _mm256_add_ps(fjy2,ty);
 460             fjz2             = _mm256_add_ps(fjz2,tz);
 461
 462             /**************************
 463              * CALCULATE INTERACTIONS *
 464              **************************/
 465
 466             /* REACTION-FIELD ELECTROSTATICS */
 467             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 468             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 469
 470             /* Update potential sum for this i atom from the interaction with this j atom. */
 471             velecsum         = _mm256_add_ps(velecsum,velec);
 472
 473             fscal            = felec;
 474
 475             /* Calculate temporary vectorial force */
 476             tx               = _mm256_mul_ps(fscal,dx20);
 477             ty               = _mm256_mul_ps(fscal,dy20);
 478             tz               = _mm256_mul_ps(fscal,dz20);
 479
 480             /* Update vectorial force */
 481             fix2             = _mm256_add_ps(fix2,tx);
 482             fiy2             = _mm256_add_ps(fiy2,ty);
 483             fiz2             = _mm256_add_ps(fiz2,tz);
 484
 485             fjx0             = _mm256_add_ps(fjx0,tx);
 486             fjy0             = _mm256_add_ps(fjy0,ty);
 487             fjz0             = _mm256_add_ps(fjz0,tz);
 488
 489             /**************************
 490              * CALCULATE INTERACTIONS *
 491              **************************/
 492
 493             /* REACTION-FIELD ELECTROSTATICS */
 494             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 495             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 496
 497             /* Update potential sum for this i atom from the interaction with this j atom. */
 498             velecsum         = _mm256_add_ps(velecsum,velec);
 499
 500             fscal            = felec;
 501
 502             /* Calculate temporary vectorial force */
 503             tx               = _mm256_mul_ps(fscal,dx21);
 504             ty               = _mm256_mul_ps(fscal,dy21);
 505             tz               = _mm256_mul_ps(fscal,dz21);
 506
 507             /* Update vectorial force */
 508             fix2             = _mm256_add_ps(fix2,tx);
 509             fiy2             = _mm256_add_ps(fiy2,ty);
 510             fiz2             = _mm256_add_ps(fiz2,tz);
 511
 512             fjx1             = _mm256_add_ps(fjx1,tx);
 513             fjy1             = _mm256_add_ps(fjy1,ty);
 514             fjz1             = _mm256_add_ps(fjz1,tz);
 515
 516             /**************************
 517              * CALCULATE INTERACTIONS *
 518              **************************/
 519
 520             /* REACTION-FIELD ELECTROSTATICS */
 521             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 522             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 523
 524             /* Update potential sum for this i atom from the interaction with this j atom. */
 525             velecsum         = _mm256_add_ps(velecsum,velec);
 526
 527             fscal            = felec;
 528
 529             /* Calculate temporary vectorial force */
 530             tx               = _mm256_mul_ps(fscal,dx22);
 531             ty               = _mm256_mul_ps(fscal,dy22);
 532             tz               = _mm256_mul_ps(fscal,dz22);
 533
 534             /* Update vectorial force */
 535             fix2             = _mm256_add_ps(fix2,tx);
 536             fiy2             = _mm256_add_ps(fiy2,ty);
 537             fiz2             = _mm256_add_ps(fiz2,tz);
 538
 539             fjx2             = _mm256_add_ps(fjx2,tx);
 540             fjy2             = _mm256_add_ps(fjy2,ty);
 541             fjz2             = _mm256_add_ps(fjz2,tz);
 542
 543             fjptrA             = f+j_coord_offsetA;
 544             fjptrB             = f+j_coord_offsetB;
 545             fjptrC             = f+j_coord_offsetC;
 546             fjptrD             = f+j_coord_offsetD;
 547             fjptrE             = f+j_coord_offsetE;
 548             fjptrF             = f+j_coord_offsetF;
 549             fjptrG             = f+j_coord_offsetG;
 550             fjptrH             = f+j_coord_offsetH;
 551
 552             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 553                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 554
 555             /* Inner loop uses 288 flops */
 556         }
 557
 558         if(jidx<j_index_end)
 559         {
 560
 561             /* Get j neighbor index, and coordinate index */
 562             jnrlistA         = jjnr[jidx];
 563             jnrlistB         = jjnr[jidx+1];
 564             jnrlistC         = jjnr[jidx+2];
 565             jnrlistD         = jjnr[jidx+3];
 566             jnrlistE         = jjnr[jidx+4];
 567             jnrlistF         = jjnr[jidx+5];
 568             jnrlistG         = jjnr[jidx+6];
 569             jnrlistH         = jjnr[jidx+7];
 570             /* Sign of each element will be negative for non-real atoms.
 571              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 572              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 573              */
 574             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 575                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 576
 577             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 578             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 579             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 580             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 581             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 582             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 583             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 584             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 585             j_coord_offsetA  = DIM*jnrA;
 586             j_coord_offsetB  = DIM*jnrB;
 587             j_coord_offsetC  = DIM*jnrC;
 588             j_coord_offsetD  = DIM*jnrD;
 589             j_coord_offsetE  = DIM*jnrE;
 590             j_coord_offsetF  = DIM*jnrF;
 591             j_coord_offsetG  = DIM*jnrG;
 592             j_coord_offsetH  = DIM*jnrH;
 593
 594             /* load j atom coordinates */
 595             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 596                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 597                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 598                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 599                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 600
 601             /* Calculate displacement vector */
 602             dx00             = _mm256_sub_ps(ix0,jx0);
 603             dy00             = _mm256_sub_ps(iy0,jy0);
 604             dz00             = _mm256_sub_ps(iz0,jz0);
 605             dx01             = _mm256_sub_ps(ix0,jx1);
 606             dy01             = _mm256_sub_ps(iy0,jy1);
 607             dz01             = _mm256_sub_ps(iz0,jz1);
 608             dx02             = _mm256_sub_ps(ix0,jx2);
 609             dy02             = _mm256_sub_ps(iy0,jy2);
 610             dz02             = _mm256_sub_ps(iz0,jz2);
 611             dx10             = _mm256_sub_ps(ix1,jx0);
 612             dy10             = _mm256_sub_ps(iy1,jy0);
 613             dz10             = _mm256_sub_ps(iz1,jz0);
 614             dx11             = _mm256_sub_ps(ix1,jx1);
 615             dy11             = _mm256_sub_ps(iy1,jy1);
 616             dz11             = _mm256_sub_ps(iz1,jz1);
 617             dx12             = _mm256_sub_ps(ix1,jx2);
 618             dy12             = _mm256_sub_ps(iy1,jy2);
 619             dz12             = _mm256_sub_ps(iz1,jz2);
 620             dx20             = _mm256_sub_ps(ix2,jx0);
 621             dy20             = _mm256_sub_ps(iy2,jy0);
 622             dz20             = _mm256_sub_ps(iz2,jz0);
 623             dx21             = _mm256_sub_ps(ix2,jx1);
 624             dy21             = _mm256_sub_ps(iy2,jy1);
 625             dz21             = _mm256_sub_ps(iz2,jz1);
 626             dx22             = _mm256_sub_ps(ix2,jx2);
 627             dy22             = _mm256_sub_ps(iy2,jy2);
 628             dz22             = _mm256_sub_ps(iz2,jz2);
 629
 630             /* Calculate squared distance and things based on it */
 631             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 632             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 633             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 634             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 635             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 636             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 637             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 638             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 639             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 640
 641             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 642             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 643             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 644             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 645             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 646             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 647             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 648             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 649             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 650
 651             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 652             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 653             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 654             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 655             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 656             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 657             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 658             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 659             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 660
 661             fjx0             = _mm256_setzero_ps();
 662             fjy0             = _mm256_setzero_ps();
 663             fjz0             = _mm256_setzero_ps();
 664             fjx1             = _mm256_setzero_ps();
 665             fjy1             = _mm256_setzero_ps();
 666             fjz1             = _mm256_setzero_ps();
 667             fjx2             = _mm256_setzero_ps();
 668             fjy2             = _mm256_setzero_ps();
 669             fjz2             = _mm256_setzero_ps();
 670
 671             /**************************
 672              * CALCULATE INTERACTIONS *
 673              **************************/
 674
 675             /* REACTION-FIELD ELECTROSTATICS */
 676             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 677             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 678
 679             /* Update potential sum for this i atom from the interaction with this j atom. */
 680             velec            = _mm256_andnot_ps(dummy_mask,velec);
 681             velecsum         = _mm256_add_ps(velecsum,velec);
 682
 683             fscal            = felec;
 684
 685             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 686
 687             /* Calculate temporary vectorial force */
 688             tx               = _mm256_mul_ps(fscal,dx00);
 689             ty               = _mm256_mul_ps(fscal,dy00);
 690             tz               = _mm256_mul_ps(fscal,dz00);
 691
 692             /* Update vectorial force */
 693             fix0             = _mm256_add_ps(fix0,tx);
 694             fiy0             = _mm256_add_ps(fiy0,ty);
 695             fiz0             = _mm256_add_ps(fiz0,tz);
 696
 697             fjx0             = _mm256_add_ps(fjx0,tx);
 698             fjy0             = _mm256_add_ps(fjy0,ty);
 699             fjz0             = _mm256_add_ps(fjz0,tz);
 700
 701             /**************************
 702              * CALCULATE INTERACTIONS *
 703              **************************/
 704
 705             /* REACTION-FIELD ELECTROSTATICS */
 706             velec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
 707             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
 708
 709             /* Update potential sum for this i atom from the interaction with this j atom. */
 710             velec            = _mm256_andnot_ps(dummy_mask,velec);
 711             velecsum         = _mm256_add_ps(velecsum,velec);
 712
 713             fscal            = felec;
 714
 715             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 716
 717             /* Calculate temporary vectorial force */
 718             tx               = _mm256_mul_ps(fscal,dx01);
 719             ty               = _mm256_mul_ps(fscal,dy01);
 720             tz               = _mm256_mul_ps(fscal,dz01);
 721
 722             /* Update vectorial force */
 723             fix0             = _mm256_add_ps(fix0,tx);
 724             fiy0             = _mm256_add_ps(fiy0,ty);
 725             fiz0             = _mm256_add_ps(fiz0,tz);
 726
 727             fjx1             = _mm256_add_ps(fjx1,tx);
 728             fjy1             = _mm256_add_ps(fjy1,ty);
 729             fjz1             = _mm256_add_ps(fjz1,tz);
 730
 731             /**************************
 732              * CALCULATE INTERACTIONS *
 733              **************************/
 734
 735             /* REACTION-FIELD ELECTROSTATICS */
 736             velec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
 737             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
 738
 739             /* Update potential sum for this i atom from the interaction with this j atom. */
 740             velec            = _mm256_andnot_ps(dummy_mask,velec);
 741             velecsum         = _mm256_add_ps(velecsum,velec);
 742
 743             fscal            = felec;
 744
 745             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 746
 747             /* Calculate temporary vectorial force */
 748             tx               = _mm256_mul_ps(fscal,dx02);
 749             ty               = _mm256_mul_ps(fscal,dy02);
 750             tz               = _mm256_mul_ps(fscal,dz02);
 751
 752             /* Update vectorial force */
 753             fix0             = _mm256_add_ps(fix0,tx);
 754             fiy0             = _mm256_add_ps(fiy0,ty);
 755             fiz0             = _mm256_add_ps(fiz0,tz);
 756
 757             fjx2             = _mm256_add_ps(fjx2,tx);
 758             fjy2             = _mm256_add_ps(fjy2,ty);
 759             fjz2             = _mm256_add_ps(fjz2,tz);
 760
 761             /**************************
 762              * CALCULATE INTERACTIONS *
 763              **************************/
 764
 765             /* REACTION-FIELD ELECTROSTATICS */
 766             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 767             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 768
 769             /* Update potential sum for this i atom from the interaction with this j atom. */
 770             velec            = _mm256_andnot_ps(dummy_mask,velec);
 771             velecsum         = _mm256_add_ps(velecsum,velec);
 772
 773             fscal            = felec;
 774
 775             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 776
 777             /* Calculate temporary vectorial force */
 778             tx               = _mm256_mul_ps(fscal,dx10);
 779             ty               = _mm256_mul_ps(fscal,dy10);
 780             tz               = _mm256_mul_ps(fscal,dz10);
 781
 782             /* Update vectorial force */
 783             fix1             = _mm256_add_ps(fix1,tx);
 784             fiy1             = _mm256_add_ps(fiy1,ty);
 785             fiz1             = _mm256_add_ps(fiz1,tz);
 786
 787             fjx0             = _mm256_add_ps(fjx0,tx);
 788             fjy0             = _mm256_add_ps(fjy0,ty);
 789             fjz0             = _mm256_add_ps(fjz0,tz);
 790
 791             /**************************
 792              * CALCULATE INTERACTIONS *
 793              **************************/
 794
 795             /* REACTION-FIELD ELECTROSTATICS */
 796             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 797             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 798
 799             /* Update potential sum for this i atom from the interaction with this j atom. */
 800             velec            = _mm256_andnot_ps(dummy_mask,velec);
 801             velecsum         = _mm256_add_ps(velecsum,velec);
 802
 803             fscal            = felec;
 804
 805             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 806
 807             /* Calculate temporary vectorial force */
 808             tx               = _mm256_mul_ps(fscal,dx11);
 809             ty               = _mm256_mul_ps(fscal,dy11);
 810             tz               = _mm256_mul_ps(fscal,dz11);
 811
 812             /* Update vectorial force */
 813             fix1             = _mm256_add_ps(fix1,tx);
 814             fiy1             = _mm256_add_ps(fiy1,ty);
 815             fiz1             = _mm256_add_ps(fiz1,tz);
 816
 817             fjx1             = _mm256_add_ps(fjx1,tx);
 818             fjy1             = _mm256_add_ps(fjy1,ty);
 819             fjz1             = _mm256_add_ps(fjz1,tz);
 820
 821             /**************************
 822              * CALCULATE INTERACTIONS *
 823              **************************/
 824
 825             /* REACTION-FIELD ELECTROSTATICS */
 826             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 827             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 828
 829             /* Update potential sum for this i atom from the interaction with this j atom. */
 830             velec            = _mm256_andnot_ps(dummy_mask,velec);
 831             velecsum         = _mm256_add_ps(velecsum,velec);
 832
 833             fscal            = felec;
 834
 835             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 836
 837             /* Calculate temporary vectorial force */
 838             tx               = _mm256_mul_ps(fscal,dx12);
 839             ty               = _mm256_mul_ps(fscal,dy12);
 840             tz               = _mm256_mul_ps(fscal,dz12);
 841
 842             /* Update vectorial force */
 843             fix1             = _mm256_add_ps(fix1,tx);
 844             fiy1             = _mm256_add_ps(fiy1,ty);
 845             fiz1             = _mm256_add_ps(fiz1,tz);
 846
 847             fjx2             = _mm256_add_ps(fjx2,tx);
 848             fjy2             = _mm256_add_ps(fjy2,ty);
 849             fjz2             = _mm256_add_ps(fjz2,tz);
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             /* REACTION-FIELD ELECTROSTATICS */
 856             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 857             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 858
 859             /* Update potential sum for this i atom from the interaction with this j atom. */
 860             velec            = _mm256_andnot_ps(dummy_mask,velec);
 861             velecsum         = _mm256_add_ps(velecsum,velec);
 862
 863             fscal            = felec;
 864
 865             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 866
 867             /* Calculate temporary vectorial force */
 868             tx               = _mm256_mul_ps(fscal,dx20);
 869             ty               = _mm256_mul_ps(fscal,dy20);
 870             tz               = _mm256_mul_ps(fscal,dz20);
 871
 872             /* Update vectorial force */
 873             fix2             = _mm256_add_ps(fix2,tx);
 874             fiy2             = _mm256_add_ps(fiy2,ty);
 875             fiz2             = _mm256_add_ps(fiz2,tz);
 876
 877             fjx0             = _mm256_add_ps(fjx0,tx);
 878             fjy0             = _mm256_add_ps(fjy0,ty);
 879             fjz0             = _mm256_add_ps(fjz0,tz);
 880
 881             /**************************
 882              * CALCULATE INTERACTIONS *
 883              **************************/
 884
 885             /* REACTION-FIELD ELECTROSTATICS */
 886             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 887             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 888
 889             /* Update potential sum for this i atom from the interaction with this j atom. */
 890             velec            = _mm256_andnot_ps(dummy_mask,velec);
 891             velecsum         = _mm256_add_ps(velecsum,velec);
 892
 893             fscal            = felec;
 894
 895             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 896
 897             /* Calculate temporary vectorial force */
 898             tx               = _mm256_mul_ps(fscal,dx21);
 899             ty               = _mm256_mul_ps(fscal,dy21);
 900             tz               = _mm256_mul_ps(fscal,dz21);
 901
 902             /* Update vectorial force */
 903             fix2             = _mm256_add_ps(fix2,tx);
 904             fiy2             = _mm256_add_ps(fiy2,ty);
 905             fiz2             = _mm256_add_ps(fiz2,tz);
 906
 907             fjx1             = _mm256_add_ps(fjx1,tx);
 908             fjy1             = _mm256_add_ps(fjy1,ty);
 909             fjz1             = _mm256_add_ps(fjz1,tz);
 910
 911             /**************************
 912              * CALCULATE INTERACTIONS *
 913              **************************/
 914
 915             /* REACTION-FIELD ELECTROSTATICS */
 916             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 917             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 918
 919             /* Update potential sum for this i atom from the interaction with this j atom. */
 920             velec            = _mm256_andnot_ps(dummy_mask,velec);
 921             velecsum         = _mm256_add_ps(velecsum,velec);
 922
 923             fscal            = felec;
 924
 925             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 926
 927             /* Calculate temporary vectorial force */
 928             tx               = _mm256_mul_ps(fscal,dx22);
 929             ty               = _mm256_mul_ps(fscal,dy22);
 930             tz               = _mm256_mul_ps(fscal,dz22);
 931
 932             /* Update vectorial force */
 933             fix2             = _mm256_add_ps(fix2,tx);
 934             fiy2             = _mm256_add_ps(fiy2,ty);
 935             fiz2             = _mm256_add_ps(fiz2,tz);
 936
 937             fjx2             = _mm256_add_ps(fjx2,tx);
 938             fjy2             = _mm256_add_ps(fjy2,ty);
 939             fjz2             = _mm256_add_ps(fjz2,tz);
 940
 941             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 942             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 943             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 944             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 945             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 946             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 947             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 948             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 949
 950             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 951                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 952
 953             /* Inner loop uses 288 flops */
 954         }
 955
 956         /* End of innermost loop */
 957
 958         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 959                                                  f+i_coord_offset,fshift+i_shift_offset);
 960
 961         ggid                        = gid[iidx];
 962         /* Update potential energies */
 963         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 964
 965         /* Increment number of inner iterations */
 966         inneriter                  += j_index_end - j_index_start;
 967
 968         /* Outer loop uses 19 flops */
 969     }
 970
 971     /* Increment number of outer iterations */
 972     outeriter        += nri;
 973
 974     /* Update outer/inner flops */
 975
 976     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*288);
 977 }
 978 /*
 979  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_single
 980  * Electrostatics interaction: ReactionField
 981  * VdW interaction:            None
 982  * Geometry:                   Water3-Water3
 983  * Calculate force/pot:        Force
 984  */
 985 void
 986 nb_kernel_ElecRF_VdwNone_GeomW3W3_F_avx_256_single
 987                     (t_nblist                    * gmx_restrict       nlist,
 988                      rvec                        * gmx_restrict          xx,
 989                      rvec                        * gmx_restrict          ff,
 990                      t_forcerec                  * gmx_restrict          fr,
 991                      t_mdatoms                   * gmx_restrict     mdatoms,
 992                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 993                      t_nrnb                      * gmx_restrict        nrnb)
 994 {
 995     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 996      * just 0 for non-waters.
 997      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 998      * jnr indices corresponding to data put in the four positions in the SIMD register.
 999      */
1000     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1001     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1002     int              jnrA,jnrB,jnrC,jnrD;
1003     int              jnrE,jnrF,jnrG,jnrH;
1004     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1005     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1006     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1007     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1008     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1009     real             rcutoff_scalar;
1010     real             *shiftvec,*fshift,*x,*f;
1011     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1012     real             scratch[4*DIM];
1013     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1014     real *           vdwioffsetptr0;
1015     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1016     real *           vdwioffsetptr1;
1017     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1018     real *           vdwioffsetptr2;
1019     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1020     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1021     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1022     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1023     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1024     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1025     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1026     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1027     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1028     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1029     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1030     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1031     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1032     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1033     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1034     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1035     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1036     real             *charge;
1037     __m256           dummy_mask,cutoff_mask;
1038     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1039     __m256           one     = _mm256_set1_ps(1.0);
1040     __m256           two     = _mm256_set1_ps(2.0);
1041     x                = xx[0];
1042     f                = ff[0];
1043
1044     nri              = nlist->nri;
1045     iinr             = nlist->iinr;
1046     jindex           = nlist->jindex;
1047     jjnr             = nlist->jjnr;
1048     shiftidx         = nlist->shift;
1049     gid              = nlist->gid;
1050     shiftvec         = fr->shift_vec[0];
1051     fshift           = fr->fshift[0];
1052     facel            = _mm256_set1_ps(fr->epsfac);
1053     charge           = mdatoms->chargeA;
1054     krf              = _mm256_set1_ps(fr->ic->k_rf);
1055     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1056     crf              = _mm256_set1_ps(fr->ic->c_rf);
1057
1058     /* Setup water-specific parameters */
1059     inr              = nlist->iinr[0];
1060     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1061     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1062     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1063
1064     jq0              = _mm256_set1_ps(charge[inr+0]);
1065     jq1              = _mm256_set1_ps(charge[inr+1]);
1066     jq2              = _mm256_set1_ps(charge[inr+2]);
1067     qq00             = _mm256_mul_ps(iq0,jq0);
1068     qq01             = _mm256_mul_ps(iq0,jq1);
1069     qq02             = _mm256_mul_ps(iq0,jq2);
1070     qq10             = _mm256_mul_ps(iq1,jq0);
1071     qq11             = _mm256_mul_ps(iq1,jq1);
1072     qq12             = _mm256_mul_ps(iq1,jq2);
1073     qq20             = _mm256_mul_ps(iq2,jq0);
1074     qq21             = _mm256_mul_ps(iq2,jq1);
1075     qq22             = _mm256_mul_ps(iq2,jq2);
1076
1077     /* Avoid stupid compiler warnings */
1078     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1079     j_coord_offsetA = 0;
1080     j_coord_offsetB = 0;
1081     j_coord_offsetC = 0;
1082     j_coord_offsetD = 0;
1083     j_coord_offsetE = 0;
1084     j_coord_offsetF = 0;
1085     j_coord_offsetG = 0;
1086     j_coord_offsetH = 0;
1087
1088     outeriter        = 0;
1089     inneriter        = 0;
1090
1091     for(iidx=0;iidx<4*DIM;iidx++)
1092     {
1093         scratch[iidx] = 0.0;
1094     }
1095
1096     /* Start outer loop over neighborlists */
1097     for(iidx=0; iidx<nri; iidx++)
1098     {
1099         /* Load shift vector for this list */
1100         i_shift_offset   = DIM*shiftidx[iidx];
1101
1102         /* Load limits for loop over neighbors */
1103         j_index_start    = jindex[iidx];
1104         j_index_end      = jindex[iidx+1];
1105
1106         /* Get outer coordinate index */
1107         inr              = iinr[iidx];
1108         i_coord_offset   = DIM*inr;
1109
1110         /* Load i particle coords and add shift vector */
1111         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1112                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1113
1114         fix0             = _mm256_setzero_ps();
1115         fiy0             = _mm256_setzero_ps();
1116         fiz0             = _mm256_setzero_ps();
1117         fix1             = _mm256_setzero_ps();
1118         fiy1             = _mm256_setzero_ps();
1119         fiz1             = _mm256_setzero_ps();
1120         fix2             = _mm256_setzero_ps();
1121         fiy2             = _mm256_setzero_ps();
1122         fiz2             = _mm256_setzero_ps();
1123
1124         /* Start inner kernel loop */
1125         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1126         {
1127
1128             /* Get j neighbor index, and coordinate index */
1129             jnrA             = jjnr[jidx];
1130             jnrB             = jjnr[jidx+1];
1131             jnrC             = jjnr[jidx+2];
1132             jnrD             = jjnr[jidx+3];
1133             jnrE             = jjnr[jidx+4];
1134             jnrF             = jjnr[jidx+5];
1135             jnrG             = jjnr[jidx+6];
1136             jnrH             = jjnr[jidx+7];
1137             j_coord_offsetA  = DIM*jnrA;
1138             j_coord_offsetB  = DIM*jnrB;
1139             j_coord_offsetC  = DIM*jnrC;
1140             j_coord_offsetD  = DIM*jnrD;
1141             j_coord_offsetE  = DIM*jnrE;
1142             j_coord_offsetF  = DIM*jnrF;
1143             j_coord_offsetG  = DIM*jnrG;
1144             j_coord_offsetH  = DIM*jnrH;
1145
1146             /* load j atom coordinates */
1147             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1148                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1149                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1150                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1151                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1152
1153             /* Calculate displacement vector */
1154             dx00             = _mm256_sub_ps(ix0,jx0);
1155             dy00             = _mm256_sub_ps(iy0,jy0);
1156             dz00             = _mm256_sub_ps(iz0,jz0);
1157             dx01             = _mm256_sub_ps(ix0,jx1);
1158             dy01             = _mm256_sub_ps(iy0,jy1);
1159             dz01             = _mm256_sub_ps(iz0,jz1);
1160             dx02             = _mm256_sub_ps(ix0,jx2);
1161             dy02             = _mm256_sub_ps(iy0,jy2);
1162             dz02             = _mm256_sub_ps(iz0,jz2);
1163             dx10             = _mm256_sub_ps(ix1,jx0);
1164             dy10             = _mm256_sub_ps(iy1,jy0);
1165             dz10             = _mm256_sub_ps(iz1,jz0);
1166             dx11             = _mm256_sub_ps(ix1,jx1);
1167             dy11             = _mm256_sub_ps(iy1,jy1);
1168             dz11             = _mm256_sub_ps(iz1,jz1);
1169             dx12             = _mm256_sub_ps(ix1,jx2);
1170             dy12             = _mm256_sub_ps(iy1,jy2);
1171             dz12             = _mm256_sub_ps(iz1,jz2);
1172             dx20             = _mm256_sub_ps(ix2,jx0);
1173             dy20             = _mm256_sub_ps(iy2,jy0);
1174             dz20             = _mm256_sub_ps(iz2,jz0);
1175             dx21             = _mm256_sub_ps(ix2,jx1);
1176             dy21             = _mm256_sub_ps(iy2,jy1);
1177             dz21             = _mm256_sub_ps(iz2,jz1);
1178             dx22             = _mm256_sub_ps(ix2,jx2);
1179             dy22             = _mm256_sub_ps(iy2,jy2);
1180             dz22             = _mm256_sub_ps(iz2,jz2);
1181
1182             /* Calculate squared distance and things based on it */
1183             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1184             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1185             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1186             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1187             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1188             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1189             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1190             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1191             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1192
1193             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1194             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1195             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1196             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1197             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1198             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1199             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1200             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1201             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1202
1203             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1204             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1205             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1206             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1207             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1208             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1209             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1210             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1211             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1212
1213             fjx0             = _mm256_setzero_ps();
1214             fjy0             = _mm256_setzero_ps();
1215             fjz0             = _mm256_setzero_ps();
1216             fjx1             = _mm256_setzero_ps();
1217             fjy1             = _mm256_setzero_ps();
1218             fjz1             = _mm256_setzero_ps();
1219             fjx2             = _mm256_setzero_ps();
1220             fjy2             = _mm256_setzero_ps();
1221             fjz2             = _mm256_setzero_ps();
1222
1223             /**************************
1224              * CALCULATE INTERACTIONS *
1225              **************************/
1226
1227             /* REACTION-FIELD ELECTROSTATICS */
1228             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1229
1230             fscal            = felec;
1231
1232             /* Calculate temporary vectorial force */
1233             tx               = _mm256_mul_ps(fscal,dx00);
1234             ty               = _mm256_mul_ps(fscal,dy00);
1235             tz               = _mm256_mul_ps(fscal,dz00);
1236
1237             /* Update vectorial force */
1238             fix0             = _mm256_add_ps(fix0,tx);
1239             fiy0             = _mm256_add_ps(fiy0,ty);
1240             fiz0             = _mm256_add_ps(fiz0,tz);
1241
1242             fjx0             = _mm256_add_ps(fjx0,tx);
1243             fjy0             = _mm256_add_ps(fjy0,ty);
1244             fjz0             = _mm256_add_ps(fjz0,tz);
1245
1246             /**************************
1247              * CALCULATE INTERACTIONS *
1248              **************************/
1249
1250             /* REACTION-FIELD ELECTROSTATICS */
1251             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
1252
1253             fscal            = felec;
1254
1255             /* Calculate temporary vectorial force */
1256             tx               = _mm256_mul_ps(fscal,dx01);
1257             ty               = _mm256_mul_ps(fscal,dy01);
1258             tz               = _mm256_mul_ps(fscal,dz01);
1259
1260             /* Update vectorial force */
1261             fix0             = _mm256_add_ps(fix0,tx);
1262             fiy0             = _mm256_add_ps(fiy0,ty);
1263             fiz0             = _mm256_add_ps(fiz0,tz);
1264
1265             fjx1             = _mm256_add_ps(fjx1,tx);
1266             fjy1             = _mm256_add_ps(fjy1,ty);
1267             fjz1             = _mm256_add_ps(fjz1,tz);
1268
1269             /**************************
1270              * CALCULATE INTERACTIONS *
1271              **************************/
1272
1273             /* REACTION-FIELD ELECTROSTATICS */
1274             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
1275
1276             fscal            = felec;
1277
1278             /* Calculate temporary vectorial force */
1279             tx               = _mm256_mul_ps(fscal,dx02);
1280             ty               = _mm256_mul_ps(fscal,dy02);
1281             tz               = _mm256_mul_ps(fscal,dz02);
1282
1283             /* Update vectorial force */
1284             fix0             = _mm256_add_ps(fix0,tx);
1285             fiy0             = _mm256_add_ps(fiy0,ty);
1286             fiz0             = _mm256_add_ps(fiz0,tz);
1287
1288             fjx2             = _mm256_add_ps(fjx2,tx);
1289             fjy2             = _mm256_add_ps(fjy2,ty);
1290             fjz2             = _mm256_add_ps(fjz2,tz);
1291
1292             /**************************
1293              * CALCULATE INTERACTIONS *
1294              **************************/
1295
1296             /* REACTION-FIELD ELECTROSTATICS */
1297             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1298
1299             fscal            = felec;
1300
1301             /* Calculate temporary vectorial force */
1302             tx               = _mm256_mul_ps(fscal,dx10);
1303             ty               = _mm256_mul_ps(fscal,dy10);
1304             tz               = _mm256_mul_ps(fscal,dz10);
1305
1306             /* Update vectorial force */
1307             fix1             = _mm256_add_ps(fix1,tx);
1308             fiy1             = _mm256_add_ps(fiy1,ty);
1309             fiz1             = _mm256_add_ps(fiz1,tz);
1310
1311             fjx0             = _mm256_add_ps(fjx0,tx);
1312             fjy0             = _mm256_add_ps(fjy0,ty);
1313             fjz0             = _mm256_add_ps(fjz0,tz);
1314
1315             /**************************
1316              * CALCULATE INTERACTIONS *
1317              **************************/
1318
1319             /* REACTION-FIELD ELECTROSTATICS */
1320             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1321
1322             fscal            = felec;
1323
1324             /* Calculate temporary vectorial force */
1325             tx               = _mm256_mul_ps(fscal,dx11);
1326             ty               = _mm256_mul_ps(fscal,dy11);
1327             tz               = _mm256_mul_ps(fscal,dz11);
1328
1329             /* Update vectorial force */
1330             fix1             = _mm256_add_ps(fix1,tx);
1331             fiy1             = _mm256_add_ps(fiy1,ty);
1332             fiz1             = _mm256_add_ps(fiz1,tz);
1333
1334             fjx1             = _mm256_add_ps(fjx1,tx);
1335             fjy1             = _mm256_add_ps(fjy1,ty);
1336             fjz1             = _mm256_add_ps(fjz1,tz);
1337
1338             /**************************
1339              * CALCULATE INTERACTIONS *
1340              **************************/
1341
1342             /* REACTION-FIELD ELECTROSTATICS */
1343             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1344
1345             fscal            = felec;
1346
1347             /* Calculate temporary vectorial force */
1348             tx               = _mm256_mul_ps(fscal,dx12);
1349             ty               = _mm256_mul_ps(fscal,dy12);
1350             tz               = _mm256_mul_ps(fscal,dz12);
1351
1352             /* Update vectorial force */
1353             fix1             = _mm256_add_ps(fix1,tx);
1354             fiy1             = _mm256_add_ps(fiy1,ty);
1355             fiz1             = _mm256_add_ps(fiz1,tz);
1356
1357             fjx2             = _mm256_add_ps(fjx2,tx);
1358             fjy2             = _mm256_add_ps(fjy2,ty);
1359             fjz2             = _mm256_add_ps(fjz2,tz);
1360
1361             /**************************
1362              * CALCULATE INTERACTIONS *
1363              **************************/
1364
1365             /* REACTION-FIELD ELECTROSTATICS */
1366             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1367
1368             fscal            = felec;
1369
1370             /* Calculate temporary vectorial force */
1371             tx               = _mm256_mul_ps(fscal,dx20);
1372             ty               = _mm256_mul_ps(fscal,dy20);
1373             tz               = _mm256_mul_ps(fscal,dz20);
1374
1375             /* Update vectorial force */
1376             fix2             = _mm256_add_ps(fix2,tx);
1377             fiy2             = _mm256_add_ps(fiy2,ty);
1378             fiz2             = _mm256_add_ps(fiz2,tz);
1379
1380             fjx0             = _mm256_add_ps(fjx0,tx);
1381             fjy0             = _mm256_add_ps(fjy0,ty);
1382             fjz0             = _mm256_add_ps(fjz0,tz);
1383
1384             /**************************
1385              * CALCULATE INTERACTIONS *
1386              **************************/
1387
1388             /* REACTION-FIELD ELECTROSTATICS */
1389             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1390
1391             fscal            = felec;
1392
1393             /* Calculate temporary vectorial force */
1394             tx               = _mm256_mul_ps(fscal,dx21);
1395             ty               = _mm256_mul_ps(fscal,dy21);
1396             tz               = _mm256_mul_ps(fscal,dz21);
1397
1398             /* Update vectorial force */
1399             fix2             = _mm256_add_ps(fix2,tx);
1400             fiy2             = _mm256_add_ps(fiy2,ty);
1401             fiz2             = _mm256_add_ps(fiz2,tz);
1402
1403             fjx1             = _mm256_add_ps(fjx1,tx);
1404             fjy1             = _mm256_add_ps(fjy1,ty);
1405             fjz1             = _mm256_add_ps(fjz1,tz);
1406
1407             /**************************
1408              * CALCULATE INTERACTIONS *
1409              **************************/
1410
1411             /* REACTION-FIELD ELECTROSTATICS */
1412             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1413
1414             fscal            = felec;
1415
1416             /* Calculate temporary vectorial force */
1417             tx               = _mm256_mul_ps(fscal,dx22);
1418             ty               = _mm256_mul_ps(fscal,dy22);
1419             tz               = _mm256_mul_ps(fscal,dz22);
1420
1421             /* Update vectorial force */
1422             fix2             = _mm256_add_ps(fix2,tx);
1423             fiy2             = _mm256_add_ps(fiy2,ty);
1424             fiz2             = _mm256_add_ps(fiz2,tz);
1425
1426             fjx2             = _mm256_add_ps(fjx2,tx);
1427             fjy2             = _mm256_add_ps(fjy2,ty);
1428             fjz2             = _mm256_add_ps(fjz2,tz);
1429
1430             fjptrA             = f+j_coord_offsetA;
1431             fjptrB             = f+j_coord_offsetB;
1432             fjptrC             = f+j_coord_offsetC;
1433             fjptrD             = f+j_coord_offsetD;
1434             fjptrE             = f+j_coord_offsetE;
1435             fjptrF             = f+j_coord_offsetF;
1436             fjptrG             = f+j_coord_offsetG;
1437             fjptrH             = f+j_coord_offsetH;
1438
1439             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1440                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
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,x+j_coord_offsetB,
1483                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1484                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1485                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1486                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1487
1488             /* Calculate displacement vector */
1489             dx00             = _mm256_sub_ps(ix0,jx0);
1490             dy00             = _mm256_sub_ps(iy0,jy0);
1491             dz00             = _mm256_sub_ps(iz0,jz0);
1492             dx01             = _mm256_sub_ps(ix0,jx1);
1493             dy01             = _mm256_sub_ps(iy0,jy1);
1494             dz01             = _mm256_sub_ps(iz0,jz1);
1495             dx02             = _mm256_sub_ps(ix0,jx2);
1496             dy02             = _mm256_sub_ps(iy0,jy2);
1497             dz02             = _mm256_sub_ps(iz0,jz2);
1498             dx10             = _mm256_sub_ps(ix1,jx0);
1499             dy10             = _mm256_sub_ps(iy1,jy0);
1500             dz10             = _mm256_sub_ps(iz1,jz0);
1501             dx11             = _mm256_sub_ps(ix1,jx1);
1502             dy11             = _mm256_sub_ps(iy1,jy1);
1503             dz11             = _mm256_sub_ps(iz1,jz1);
1504             dx12             = _mm256_sub_ps(ix1,jx2);
1505             dy12             = _mm256_sub_ps(iy1,jy2);
1506             dz12             = _mm256_sub_ps(iz1,jz2);
1507             dx20             = _mm256_sub_ps(ix2,jx0);
1508             dy20             = _mm256_sub_ps(iy2,jy0);
1509             dz20             = _mm256_sub_ps(iz2,jz0);
1510             dx21             = _mm256_sub_ps(ix2,jx1);
1511             dy21             = _mm256_sub_ps(iy2,jy1);
1512             dz21             = _mm256_sub_ps(iz2,jz1);
1513             dx22             = _mm256_sub_ps(ix2,jx2);
1514             dy22             = _mm256_sub_ps(iy2,jy2);
1515             dz22             = _mm256_sub_ps(iz2,jz2);
1516
1517             /* Calculate squared distance and things based on it */
1518             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1519             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1520             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1521             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1522             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1523             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1524             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1525             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1526             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1527
1528             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1529             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1530             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1531             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1532             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1533             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1534             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1535             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1536             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1537
1538             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1539             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1540             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1541             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1542             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1543             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1544             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1545             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1546             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1547
1548             fjx0             = _mm256_setzero_ps();
1549             fjy0             = _mm256_setzero_ps();
1550             fjz0             = _mm256_setzero_ps();
1551             fjx1             = _mm256_setzero_ps();
1552             fjy1             = _mm256_setzero_ps();
1553             fjz1             = _mm256_setzero_ps();
1554             fjx2             = _mm256_setzero_ps();
1555             fjy2             = _mm256_setzero_ps();
1556             fjz2             = _mm256_setzero_ps();
1557
1558             /**************************
1559              * CALCULATE INTERACTIONS *
1560              **************************/
1561
1562             /* REACTION-FIELD ELECTROSTATICS */
1563             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),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,dx00);
1571             ty               = _mm256_mul_ps(fscal,dy00);
1572             tz               = _mm256_mul_ps(fscal,dz00);
1573
1574             /* Update vectorial force */
1575             fix0             = _mm256_add_ps(fix0,tx);
1576             fiy0             = _mm256_add_ps(fiy0,ty);
1577             fiz0             = _mm256_add_ps(fiz0,tz);
1578
1579             fjx0             = _mm256_add_ps(fjx0,tx);
1580             fjy0             = _mm256_add_ps(fjy0,ty);
1581             fjz0             = _mm256_add_ps(fjz0,tz);
1582
1583             /**************************
1584              * CALCULATE INTERACTIONS *
1585              **************************/
1586
1587             /* REACTION-FIELD ELECTROSTATICS */
1588             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),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,dx01);
1596             ty               = _mm256_mul_ps(fscal,dy01);
1597             tz               = _mm256_mul_ps(fscal,dz01);
1598
1599             /* Update vectorial force */
1600             fix0             = _mm256_add_ps(fix0,tx);
1601             fiy0             = _mm256_add_ps(fiy0,ty);
1602             fiz0             = _mm256_add_ps(fiz0,tz);
1603
1604             fjx1             = _mm256_add_ps(fjx1,tx);
1605             fjy1             = _mm256_add_ps(fjy1,ty);
1606             fjz1             = _mm256_add_ps(fjz1,tz);
1607
1608             /**************************
1609              * CALCULATE INTERACTIONS *
1610              **************************/
1611
1612             /* REACTION-FIELD ELECTROSTATICS */
1613             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),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,dx02);
1621             ty               = _mm256_mul_ps(fscal,dy02);
1622             tz               = _mm256_mul_ps(fscal,dz02);
1623
1624             /* Update vectorial force */
1625             fix0             = _mm256_add_ps(fix0,tx);
1626             fiy0             = _mm256_add_ps(fiy0,ty);
1627             fiz0             = _mm256_add_ps(fiz0,tz);
1628
1629             fjx2             = _mm256_add_ps(fjx2,tx);
1630             fjy2             = _mm256_add_ps(fjy2,ty);
1631             fjz2             = _mm256_add_ps(fjz2,tz);
1632
1633             /**************************
1634              * CALCULATE INTERACTIONS *
1635              **************************/
1636
1637             /* REACTION-FIELD ELECTROSTATICS */
1638             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),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,dx10);
1646             ty               = _mm256_mul_ps(fscal,dy10);
1647             tz               = _mm256_mul_ps(fscal,dz10);
1648
1649             /* Update vectorial force */
1650             fix1             = _mm256_add_ps(fix1,tx);
1651             fiy1             = _mm256_add_ps(fiy1,ty);
1652             fiz1             = _mm256_add_ps(fiz1,tz);
1653
1654             fjx0             = _mm256_add_ps(fjx0,tx);
1655             fjy0             = _mm256_add_ps(fjy0,ty);
1656             fjz0             = _mm256_add_ps(fjz0,tz);
1657
1658             /**************************
1659              * CALCULATE INTERACTIONS *
1660              **************************/
1661
1662             /* REACTION-FIELD ELECTROSTATICS */
1663             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),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,dx11);
1671             ty               = _mm256_mul_ps(fscal,dy11);
1672             tz               = _mm256_mul_ps(fscal,dz11);
1673
1674             /* Update vectorial force */
1675             fix1             = _mm256_add_ps(fix1,tx);
1676             fiy1             = _mm256_add_ps(fiy1,ty);
1677             fiz1             = _mm256_add_ps(fiz1,tz);
1678
1679             fjx1             = _mm256_add_ps(fjx1,tx);
1680             fjy1             = _mm256_add_ps(fjy1,ty);
1681             fjz1             = _mm256_add_ps(fjz1,tz);
1682
1683             /**************************
1684              * CALCULATE INTERACTIONS *
1685              **************************/
1686
1687             /* REACTION-FIELD ELECTROSTATICS */
1688             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),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,dx12);
1696             ty               = _mm256_mul_ps(fscal,dy12);
1697             tz               = _mm256_mul_ps(fscal,dz12);
1698
1699             /* Update vectorial force */
1700             fix1             = _mm256_add_ps(fix1,tx);
1701             fiy1             = _mm256_add_ps(fiy1,ty);
1702             fiz1             = _mm256_add_ps(fiz1,tz);
1703
1704             fjx2             = _mm256_add_ps(fjx2,tx);
1705             fjy2             = _mm256_add_ps(fjy2,ty);
1706             fjz2             = _mm256_add_ps(fjz2,tz);
1707
1708             /**************************
1709              * CALCULATE INTERACTIONS *
1710              **************************/
1711
1712             /* REACTION-FIELD ELECTROSTATICS */
1713             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),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,dx20);
1721             ty               = _mm256_mul_ps(fscal,dy20);
1722             tz               = _mm256_mul_ps(fscal,dz20);
1723
1724             /* Update vectorial force */
1725             fix2             = _mm256_add_ps(fix2,tx);
1726             fiy2             = _mm256_add_ps(fiy2,ty);
1727             fiz2             = _mm256_add_ps(fiz2,tz);
1728
1729             fjx0             = _mm256_add_ps(fjx0,tx);
1730             fjy0             = _mm256_add_ps(fjy0,ty);
1731             fjz0             = _mm256_add_ps(fjz0,tz);
1732
1733             /**************************
1734              * CALCULATE INTERACTIONS *
1735              **************************/
1736
1737             /* REACTION-FIELD ELECTROSTATICS */
1738             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),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,dx21);
1746             ty               = _mm256_mul_ps(fscal,dy21);
1747             tz               = _mm256_mul_ps(fscal,dz21);
1748
1749             /* Update vectorial force */
1750             fix2             = _mm256_add_ps(fix2,tx);
1751             fiy2             = _mm256_add_ps(fiy2,ty);
1752             fiz2             = _mm256_add_ps(fiz2,tz);
1753
1754             fjx1             = _mm256_add_ps(fjx1,tx);
1755             fjy1             = _mm256_add_ps(fjy1,ty);
1756             fjz1             = _mm256_add_ps(fjz1,tz);
1757
1758             /**************************
1759              * CALCULATE INTERACTIONS *
1760              **************************/
1761
1762             /* REACTION-FIELD ELECTROSTATICS */
1763             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),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,dx22);
1771             ty               = _mm256_mul_ps(fscal,dy22);
1772             tz               = _mm256_mul_ps(fscal,dz22);
1773
1774             /* Update vectorial force */
1775             fix2             = _mm256_add_ps(fix2,tx);
1776             fiy2             = _mm256_add_ps(fiy2,ty);
1777             fiz2             = _mm256_add_ps(fiz2,tz);
1778
1779             fjx2             = _mm256_add_ps(fjx2,tx);
1780             fjy2             = _mm256_add_ps(fjy2,ty);
1781             fjz2             = _mm256_add_ps(fjz2,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,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1793                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1794
1795             /* Inner loop uses 243 flops */
1796         }
1797
1798         /* End of innermost loop */
1799
1800         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1801                                                  f+i_coord_offset,fshift+i_shift_offset);
1802
1803         /* Increment number of inner iterations */
1804         inneriter                  += j_index_end - j_index_start;
1805
1806         /* Outer loop uses 18 flops */
1807     }
1808
1809     /* Increment number of outer iterations */
1810     outeriter        += nri;
1811
1812     /* Update outer/inner flops */
1813
1814     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*243);
1815 }