src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRF_VdwLJ_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_VdwLJ_GeomW3W3_VF_avx_256_single
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRF_VdwLJ_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     int              nvdwtype;
 112     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 113     int              *vdwtype;
 114     real             *vdwparam;
 115     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 116     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 117     __m256           dummy_mask,cutoff_mask;
 118     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 119     __m256           one     = _mm256_set1_ps(1.0);
 120     __m256           two     = _mm256_set1_ps(2.0);
 121     x                = xx[0];
 122     f                = ff[0];
 123
 124     nri              = nlist->nri;
 125     iinr             = nlist->iinr;
 126     jindex           = nlist->jindex;
 127     jjnr             = nlist->jjnr;
 128     shiftidx         = nlist->shift;
 129     gid              = nlist->gid;
 130     shiftvec         = fr->shift_vec[0];
 131     fshift           = fr->fshift[0];
 132     facel            = _mm256_set1_ps(fr->epsfac);
 133     charge           = mdatoms->chargeA;
 134     krf              = _mm256_set1_ps(fr->ic->k_rf);
 135     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 136     crf              = _mm256_set1_ps(fr->ic->c_rf);
 137     nvdwtype         = fr->ntype;
 138     vdwparam         = fr->nbfp;
 139     vdwtype          = mdatoms->typeA;
 140
 141     /* Setup water-specific parameters */
 142     inr              = nlist->iinr[0];
 143     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 144     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 145     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 146     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 147
 148     jq0              = _mm256_set1_ps(charge[inr+0]);
 149     jq1              = _mm256_set1_ps(charge[inr+1]);
 150     jq2              = _mm256_set1_ps(charge[inr+2]);
 151     vdwjidx0A        = 2*vdwtype[inr+0];
 152     qq00             = _mm256_mul_ps(iq0,jq0);
 153     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 154     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 155     qq01             = _mm256_mul_ps(iq0,jq1);
 156     qq02             = _mm256_mul_ps(iq0,jq2);
 157     qq10             = _mm256_mul_ps(iq1,jq0);
 158     qq11             = _mm256_mul_ps(iq1,jq1);
 159     qq12             = _mm256_mul_ps(iq1,jq2);
 160     qq20             = _mm256_mul_ps(iq2,jq0);
 161     qq21             = _mm256_mul_ps(iq2,jq1);
 162     qq22             = _mm256_mul_ps(iq2,jq2);
 163
 164     /* Avoid stupid compiler warnings */
 165     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 166     j_coord_offsetA = 0;
 167     j_coord_offsetB = 0;
 168     j_coord_offsetC = 0;
 169     j_coord_offsetD = 0;
 170     j_coord_offsetE = 0;
 171     j_coord_offsetF = 0;
 172     j_coord_offsetG = 0;
 173     j_coord_offsetH = 0;
 174
 175     outeriter        = 0;
 176     inneriter        = 0;
 177
 178     for(iidx=0;iidx<4*DIM;iidx++)
 179     {
 180         scratch[iidx] = 0.0;
 181     }
 182
 183     /* Start outer loop over neighborlists */
 184     for(iidx=0; iidx<nri; iidx++)
 185     {
 186         /* Load shift vector for this list */
 187         i_shift_offset   = DIM*shiftidx[iidx];
 188
 189         /* Load limits for loop over neighbors */
 190         j_index_start    = jindex[iidx];
 191         j_index_end      = jindex[iidx+1];
 192
 193         /* Get outer coordinate index */
 194         inr              = iinr[iidx];
 195         i_coord_offset   = DIM*inr;
 196
 197         /* Load i particle coords and add shift vector */
 198         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 199                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 200
 201         fix0             = _mm256_setzero_ps();
 202         fiy0             = _mm256_setzero_ps();
 203         fiz0             = _mm256_setzero_ps();
 204         fix1             = _mm256_setzero_ps();
 205         fiy1             = _mm256_setzero_ps();
 206         fiz1             = _mm256_setzero_ps();
 207         fix2             = _mm256_setzero_ps();
 208         fiy2             = _mm256_setzero_ps();
 209         fiz2             = _mm256_setzero_ps();
 210
 211         /* Reset potential sums */
 212         velecsum         = _mm256_setzero_ps();
 213         vvdwsum          = _mm256_setzero_ps();
 214
 215         /* Start inner kernel loop */
 216         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 217         {
 218
 219             /* Get j neighbor index, and coordinate index */
 220             jnrA             = jjnr[jidx];
 221             jnrB             = jjnr[jidx+1];
 222             jnrC             = jjnr[jidx+2];
 223             jnrD             = jjnr[jidx+3];
 224             jnrE             = jjnr[jidx+4];
 225             jnrF             = jjnr[jidx+5];
 226             jnrG             = jjnr[jidx+6];
 227             jnrH             = jjnr[jidx+7];
 228             j_coord_offsetA  = DIM*jnrA;
 229             j_coord_offsetB  = DIM*jnrB;
 230             j_coord_offsetC  = DIM*jnrC;
 231             j_coord_offsetD  = DIM*jnrD;
 232             j_coord_offsetE  = DIM*jnrE;
 233             j_coord_offsetF  = DIM*jnrF;
 234             j_coord_offsetG  = DIM*jnrG;
 235             j_coord_offsetH  = DIM*jnrH;
 236
 237             /* load j atom coordinates */
 238             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 239                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 240                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 241                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 242                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 243
 244             /* Calculate displacement vector */
 245             dx00             = _mm256_sub_ps(ix0,jx0);
 246             dy00             = _mm256_sub_ps(iy0,jy0);
 247             dz00             = _mm256_sub_ps(iz0,jz0);
 248             dx01             = _mm256_sub_ps(ix0,jx1);
 249             dy01             = _mm256_sub_ps(iy0,jy1);
 250             dz01             = _mm256_sub_ps(iz0,jz1);
 251             dx02             = _mm256_sub_ps(ix0,jx2);
 252             dy02             = _mm256_sub_ps(iy0,jy2);
 253             dz02             = _mm256_sub_ps(iz0,jz2);
 254             dx10             = _mm256_sub_ps(ix1,jx0);
 255             dy10             = _mm256_sub_ps(iy1,jy0);
 256             dz10             = _mm256_sub_ps(iz1,jz0);
 257             dx11             = _mm256_sub_ps(ix1,jx1);
 258             dy11             = _mm256_sub_ps(iy1,jy1);
 259             dz11             = _mm256_sub_ps(iz1,jz1);
 260             dx12             = _mm256_sub_ps(ix1,jx2);
 261             dy12             = _mm256_sub_ps(iy1,jy2);
 262             dz12             = _mm256_sub_ps(iz1,jz2);
 263             dx20             = _mm256_sub_ps(ix2,jx0);
 264             dy20             = _mm256_sub_ps(iy2,jy0);
 265             dz20             = _mm256_sub_ps(iz2,jz0);
 266             dx21             = _mm256_sub_ps(ix2,jx1);
 267             dy21             = _mm256_sub_ps(iy2,jy1);
 268             dz21             = _mm256_sub_ps(iz2,jz1);
 269             dx22             = _mm256_sub_ps(ix2,jx2);
 270             dy22             = _mm256_sub_ps(iy2,jy2);
 271             dz22             = _mm256_sub_ps(iz2,jz2);
 272
 273             /* Calculate squared distance and things based on it */
 274             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 275             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 276             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 277             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 278             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 279             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 280             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 281             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 282             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 283
 284             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 285             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 286             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 287             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 288             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 289             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 290             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 291             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 292             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 293
 294             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 295             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 296             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 297             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 298             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 299             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 300             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 301             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 302             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 303
 304             fjx0             = _mm256_setzero_ps();
 305             fjy0             = _mm256_setzero_ps();
 306             fjz0             = _mm256_setzero_ps();
 307             fjx1             = _mm256_setzero_ps();
 308             fjy1             = _mm256_setzero_ps();
 309             fjz1             = _mm256_setzero_ps();
 310             fjx2             = _mm256_setzero_ps();
 311             fjy2             = _mm256_setzero_ps();
 312             fjz2             = _mm256_setzero_ps();
 313
 314             /**************************
 315              * CALCULATE INTERACTIONS *
 316              **************************/
 317
 318             /* REACTION-FIELD ELECTROSTATICS */
 319             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 320             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 321
 322             /* LENNARD-JONES DISPERSION/REPULSION */
 323
 324             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 325             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 326             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 327             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 328             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 329
 330             /* Update potential sum for this i atom from the interaction with this j atom. */
 331             velecsum         = _mm256_add_ps(velecsum,velec);
 332             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 333
 334             fscal            = _mm256_add_ps(felec,fvdw);
 335
 336             /* Calculate temporary vectorial force */
 337             tx               = _mm256_mul_ps(fscal,dx00);
 338             ty               = _mm256_mul_ps(fscal,dy00);
 339             tz               = _mm256_mul_ps(fscal,dz00);
 340
 341             /* Update vectorial force */
 342             fix0             = _mm256_add_ps(fix0,tx);
 343             fiy0             = _mm256_add_ps(fiy0,ty);
 344             fiz0             = _mm256_add_ps(fiz0,tz);
 345
 346             fjx0             = _mm256_add_ps(fjx0,tx);
 347             fjy0             = _mm256_add_ps(fjy0,ty);
 348             fjz0             = _mm256_add_ps(fjz0,tz);
 349
 350             /**************************
 351              * CALCULATE INTERACTIONS *
 352              **************************/
 353
 354             /* REACTION-FIELD ELECTROSTATICS */
 355             velec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
 356             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
 357
 358             /* Update potential sum for this i atom from the interaction with this j atom. */
 359             velecsum         = _mm256_add_ps(velecsum,velec);
 360
 361             fscal            = felec;
 362
 363             /* Calculate temporary vectorial force */
 364             tx               = _mm256_mul_ps(fscal,dx01);
 365             ty               = _mm256_mul_ps(fscal,dy01);
 366             tz               = _mm256_mul_ps(fscal,dz01);
 367
 368             /* Update vectorial force */
 369             fix0             = _mm256_add_ps(fix0,tx);
 370             fiy0             = _mm256_add_ps(fiy0,ty);
 371             fiz0             = _mm256_add_ps(fiz0,tz);
 372
 373             fjx1             = _mm256_add_ps(fjx1,tx);
 374             fjy1             = _mm256_add_ps(fjy1,ty);
 375             fjz1             = _mm256_add_ps(fjz1,tz);
 376
 377             /**************************
 378              * CALCULATE INTERACTIONS *
 379              **************************/
 380
 381             /* REACTION-FIELD ELECTROSTATICS */
 382             velec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
 383             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
 384
 385             /* Update potential sum for this i atom from the interaction with this j atom. */
 386             velecsum         = _mm256_add_ps(velecsum,velec);
 387
 388             fscal            = felec;
 389
 390             /* Calculate temporary vectorial force */
 391             tx               = _mm256_mul_ps(fscal,dx02);
 392             ty               = _mm256_mul_ps(fscal,dy02);
 393             tz               = _mm256_mul_ps(fscal,dz02);
 394
 395             /* Update vectorial force */
 396             fix0             = _mm256_add_ps(fix0,tx);
 397             fiy0             = _mm256_add_ps(fiy0,ty);
 398             fiz0             = _mm256_add_ps(fiz0,tz);
 399
 400             fjx2             = _mm256_add_ps(fjx2,tx);
 401             fjy2             = _mm256_add_ps(fjy2,ty);
 402             fjz2             = _mm256_add_ps(fjz2,tz);
 403
 404             /**************************
 405              * CALCULATE INTERACTIONS *
 406              **************************/
 407
 408             /* REACTION-FIELD ELECTROSTATICS */
 409             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 410             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 411
 412             /* Update potential sum for this i atom from the interaction with this j atom. */
 413             velecsum         = _mm256_add_ps(velecsum,velec);
 414
 415             fscal            = felec;
 416
 417             /* Calculate temporary vectorial force */
 418             tx               = _mm256_mul_ps(fscal,dx10);
 419             ty               = _mm256_mul_ps(fscal,dy10);
 420             tz               = _mm256_mul_ps(fscal,dz10);
 421
 422             /* Update vectorial force */
 423             fix1             = _mm256_add_ps(fix1,tx);
 424             fiy1             = _mm256_add_ps(fiy1,ty);
 425             fiz1             = _mm256_add_ps(fiz1,tz);
 426
 427             fjx0             = _mm256_add_ps(fjx0,tx);
 428             fjy0             = _mm256_add_ps(fjy0,ty);
 429             fjz0             = _mm256_add_ps(fjz0,tz);
 430
 431             /**************************
 432              * CALCULATE INTERACTIONS *
 433              **************************/
 434
 435             /* REACTION-FIELD ELECTROSTATICS */
 436             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 437             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 438
 439             /* Update potential sum for this i atom from the interaction with this j atom. */
 440             velecsum         = _mm256_add_ps(velecsum,velec);
 441
 442             fscal            = felec;
 443
 444             /* Calculate temporary vectorial force */
 445             tx               = _mm256_mul_ps(fscal,dx11);
 446             ty               = _mm256_mul_ps(fscal,dy11);
 447             tz               = _mm256_mul_ps(fscal,dz11);
 448
 449             /* Update vectorial force */
 450             fix1             = _mm256_add_ps(fix1,tx);
 451             fiy1             = _mm256_add_ps(fiy1,ty);
 452             fiz1             = _mm256_add_ps(fiz1,tz);
 453
 454             fjx1             = _mm256_add_ps(fjx1,tx);
 455             fjy1             = _mm256_add_ps(fjy1,ty);
 456             fjz1             = _mm256_add_ps(fjz1,tz);
 457
 458             /**************************
 459              * CALCULATE INTERACTIONS *
 460              **************************/
 461
 462             /* REACTION-FIELD ELECTROSTATICS */
 463             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 464             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 465
 466             /* Update potential sum for this i atom from the interaction with this j atom. */
 467             velecsum         = _mm256_add_ps(velecsum,velec);
 468
 469             fscal            = felec;
 470
 471             /* Calculate temporary vectorial force */
 472             tx               = _mm256_mul_ps(fscal,dx12);
 473             ty               = _mm256_mul_ps(fscal,dy12);
 474             tz               = _mm256_mul_ps(fscal,dz12);
 475
 476             /* Update vectorial force */
 477             fix1             = _mm256_add_ps(fix1,tx);
 478             fiy1             = _mm256_add_ps(fiy1,ty);
 479             fiz1             = _mm256_add_ps(fiz1,tz);
 480
 481             fjx2             = _mm256_add_ps(fjx2,tx);
 482             fjy2             = _mm256_add_ps(fjy2,ty);
 483             fjz2             = _mm256_add_ps(fjz2,tz);
 484
 485             /**************************
 486              * CALCULATE INTERACTIONS *
 487              **************************/
 488
 489             /* REACTION-FIELD ELECTROSTATICS */
 490             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 491             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 492
 493             /* Update potential sum for this i atom from the interaction with this j atom. */
 494             velecsum         = _mm256_add_ps(velecsum,velec);
 495
 496             fscal            = felec;
 497
 498             /* Calculate temporary vectorial force */
 499             tx               = _mm256_mul_ps(fscal,dx20);
 500             ty               = _mm256_mul_ps(fscal,dy20);
 501             tz               = _mm256_mul_ps(fscal,dz20);
 502
 503             /* Update vectorial force */
 504             fix2             = _mm256_add_ps(fix2,tx);
 505             fiy2             = _mm256_add_ps(fiy2,ty);
 506             fiz2             = _mm256_add_ps(fiz2,tz);
 507
 508             fjx0             = _mm256_add_ps(fjx0,tx);
 509             fjy0             = _mm256_add_ps(fjy0,ty);
 510             fjz0             = _mm256_add_ps(fjz0,tz);
 511
 512             /**************************
 513              * CALCULATE INTERACTIONS *
 514              **************************/
 515
 516             /* REACTION-FIELD ELECTROSTATICS */
 517             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 518             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 519
 520             /* Update potential sum for this i atom from the interaction with this j atom. */
 521             velecsum         = _mm256_add_ps(velecsum,velec);
 522
 523             fscal            = felec;
 524
 525             /* Calculate temporary vectorial force */
 526             tx               = _mm256_mul_ps(fscal,dx21);
 527             ty               = _mm256_mul_ps(fscal,dy21);
 528             tz               = _mm256_mul_ps(fscal,dz21);
 529
 530             /* Update vectorial force */
 531             fix2             = _mm256_add_ps(fix2,tx);
 532             fiy2             = _mm256_add_ps(fiy2,ty);
 533             fiz2             = _mm256_add_ps(fiz2,tz);
 534
 535             fjx1             = _mm256_add_ps(fjx1,tx);
 536             fjy1             = _mm256_add_ps(fjy1,ty);
 537             fjz1             = _mm256_add_ps(fjz1,tz);
 538
 539             /**************************
 540              * CALCULATE INTERACTIONS *
 541              **************************/
 542
 543             /* REACTION-FIELD ELECTROSTATICS */
 544             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 545             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 546
 547             /* Update potential sum for this i atom from the interaction with this j atom. */
 548             velecsum         = _mm256_add_ps(velecsum,velec);
 549
 550             fscal            = felec;
 551
 552             /* Calculate temporary vectorial force */
 553             tx               = _mm256_mul_ps(fscal,dx22);
 554             ty               = _mm256_mul_ps(fscal,dy22);
 555             tz               = _mm256_mul_ps(fscal,dz22);
 556
 557             /* Update vectorial force */
 558             fix2             = _mm256_add_ps(fix2,tx);
 559             fiy2             = _mm256_add_ps(fiy2,ty);
 560             fiz2             = _mm256_add_ps(fiz2,tz);
 561
 562             fjx2             = _mm256_add_ps(fjx2,tx);
 563             fjy2             = _mm256_add_ps(fjy2,ty);
 564             fjz2             = _mm256_add_ps(fjz2,tz);
 565
 566             fjptrA             = f+j_coord_offsetA;
 567             fjptrB             = f+j_coord_offsetB;
 568             fjptrC             = f+j_coord_offsetC;
 569             fjptrD             = f+j_coord_offsetD;
 570             fjptrE             = f+j_coord_offsetE;
 571             fjptrF             = f+j_coord_offsetF;
 572             fjptrG             = f+j_coord_offsetG;
 573             fjptrH             = f+j_coord_offsetH;
 574
 575             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 576                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 577
 578             /* Inner loop uses 300 flops */
 579         }
 580
 581         if(jidx<j_index_end)
 582         {
 583
 584             /* Get j neighbor index, and coordinate index */
 585             jnrlistA         = jjnr[jidx];
 586             jnrlistB         = jjnr[jidx+1];
 587             jnrlistC         = jjnr[jidx+2];
 588             jnrlistD         = jjnr[jidx+3];
 589             jnrlistE         = jjnr[jidx+4];
 590             jnrlistF         = jjnr[jidx+5];
 591             jnrlistG         = jjnr[jidx+6];
 592             jnrlistH         = jjnr[jidx+7];
 593             /* Sign of each element will be negative for non-real atoms.
 594              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 595              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 596              */
 597             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 598                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 599
 600             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 601             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 602             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 603             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 604             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 605             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 606             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 607             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 608             j_coord_offsetA  = DIM*jnrA;
 609             j_coord_offsetB  = DIM*jnrB;
 610             j_coord_offsetC  = DIM*jnrC;
 611             j_coord_offsetD  = DIM*jnrD;
 612             j_coord_offsetE  = DIM*jnrE;
 613             j_coord_offsetF  = DIM*jnrF;
 614             j_coord_offsetG  = DIM*jnrG;
 615             j_coord_offsetH  = DIM*jnrH;
 616
 617             /* load j atom coordinates */
 618             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 619                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 620                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 621                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 622                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 623
 624             /* Calculate displacement vector */
 625             dx00             = _mm256_sub_ps(ix0,jx0);
 626             dy00             = _mm256_sub_ps(iy0,jy0);
 627             dz00             = _mm256_sub_ps(iz0,jz0);
 628             dx01             = _mm256_sub_ps(ix0,jx1);
 629             dy01             = _mm256_sub_ps(iy0,jy1);
 630             dz01             = _mm256_sub_ps(iz0,jz1);
 631             dx02             = _mm256_sub_ps(ix0,jx2);
 632             dy02             = _mm256_sub_ps(iy0,jy2);
 633             dz02             = _mm256_sub_ps(iz0,jz2);
 634             dx10             = _mm256_sub_ps(ix1,jx0);
 635             dy10             = _mm256_sub_ps(iy1,jy0);
 636             dz10             = _mm256_sub_ps(iz1,jz0);
 637             dx11             = _mm256_sub_ps(ix1,jx1);
 638             dy11             = _mm256_sub_ps(iy1,jy1);
 639             dz11             = _mm256_sub_ps(iz1,jz1);
 640             dx12             = _mm256_sub_ps(ix1,jx2);
 641             dy12             = _mm256_sub_ps(iy1,jy2);
 642             dz12             = _mm256_sub_ps(iz1,jz2);
 643             dx20             = _mm256_sub_ps(ix2,jx0);
 644             dy20             = _mm256_sub_ps(iy2,jy0);
 645             dz20             = _mm256_sub_ps(iz2,jz0);
 646             dx21             = _mm256_sub_ps(ix2,jx1);
 647             dy21             = _mm256_sub_ps(iy2,jy1);
 648             dz21             = _mm256_sub_ps(iz2,jz1);
 649             dx22             = _mm256_sub_ps(ix2,jx2);
 650             dy22             = _mm256_sub_ps(iy2,jy2);
 651             dz22             = _mm256_sub_ps(iz2,jz2);
 652
 653             /* Calculate squared distance and things based on it */
 654             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 655             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 656             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 657             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 658             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 659             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 660             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 661             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 662             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 663
 664             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 665             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 666             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 667             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 668             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 669             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 670             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 671             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 672             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 673
 674             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 675             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 676             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 677             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 678             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 679             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 680             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 681             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 682             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 683
 684             fjx0             = _mm256_setzero_ps();
 685             fjy0             = _mm256_setzero_ps();
 686             fjz0             = _mm256_setzero_ps();
 687             fjx1             = _mm256_setzero_ps();
 688             fjy1             = _mm256_setzero_ps();
 689             fjz1             = _mm256_setzero_ps();
 690             fjx2             = _mm256_setzero_ps();
 691             fjy2             = _mm256_setzero_ps();
 692             fjz2             = _mm256_setzero_ps();
 693
 694             /**************************
 695              * CALCULATE INTERACTIONS *
 696              **************************/
 697
 698             /* REACTION-FIELD ELECTROSTATICS */
 699             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 700             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 701
 702             /* LENNARD-JONES DISPERSION/REPULSION */
 703
 704             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 705             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 706             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 707             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 708             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 709
 710             /* Update potential sum for this i atom from the interaction with this j atom. */
 711             velec            = _mm256_andnot_ps(dummy_mask,velec);
 712             velecsum         = _mm256_add_ps(velecsum,velec);
 713             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 714             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 715
 716             fscal            = _mm256_add_ps(felec,fvdw);
 717
 718             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 719
 720             /* Calculate temporary vectorial force */
 721             tx               = _mm256_mul_ps(fscal,dx00);
 722             ty               = _mm256_mul_ps(fscal,dy00);
 723             tz               = _mm256_mul_ps(fscal,dz00);
 724
 725             /* Update vectorial force */
 726             fix0             = _mm256_add_ps(fix0,tx);
 727             fiy0             = _mm256_add_ps(fiy0,ty);
 728             fiz0             = _mm256_add_ps(fiz0,tz);
 729
 730             fjx0             = _mm256_add_ps(fjx0,tx);
 731             fjy0             = _mm256_add_ps(fjy0,ty);
 732             fjz0             = _mm256_add_ps(fjz0,tz);
 733
 734             /**************************
 735              * CALCULATE INTERACTIONS *
 736              **************************/
 737
 738             /* REACTION-FIELD ELECTROSTATICS */
 739             velec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_add_ps(rinv01,_mm256_mul_ps(krf,rsq01)),crf));
 740             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
 741
 742             /* Update potential sum for this i atom from the interaction with this j atom. */
 743             velec            = _mm256_andnot_ps(dummy_mask,velec);
 744             velecsum         = _mm256_add_ps(velecsum,velec);
 745
 746             fscal            = felec;
 747
 748             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 749
 750             /* Calculate temporary vectorial force */
 751             tx               = _mm256_mul_ps(fscal,dx01);
 752             ty               = _mm256_mul_ps(fscal,dy01);
 753             tz               = _mm256_mul_ps(fscal,dz01);
 754
 755             /* Update vectorial force */
 756             fix0             = _mm256_add_ps(fix0,tx);
 757             fiy0             = _mm256_add_ps(fiy0,ty);
 758             fiz0             = _mm256_add_ps(fiz0,tz);
 759
 760             fjx1             = _mm256_add_ps(fjx1,tx);
 761             fjy1             = _mm256_add_ps(fjy1,ty);
 762             fjz1             = _mm256_add_ps(fjz1,tz);
 763
 764             /**************************
 765              * CALCULATE INTERACTIONS *
 766              **************************/
 767
 768             /* REACTION-FIELD ELECTROSTATICS */
 769             velec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_add_ps(rinv02,_mm256_mul_ps(krf,rsq02)),crf));
 770             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
 771
 772             /* Update potential sum for this i atom from the interaction with this j atom. */
 773             velec            = _mm256_andnot_ps(dummy_mask,velec);
 774             velecsum         = _mm256_add_ps(velecsum,velec);
 775
 776             fscal            = felec;
 777
 778             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 779
 780             /* Calculate temporary vectorial force */
 781             tx               = _mm256_mul_ps(fscal,dx02);
 782             ty               = _mm256_mul_ps(fscal,dy02);
 783             tz               = _mm256_mul_ps(fscal,dz02);
 784
 785             /* Update vectorial force */
 786             fix0             = _mm256_add_ps(fix0,tx);
 787             fiy0             = _mm256_add_ps(fiy0,ty);
 788             fiz0             = _mm256_add_ps(fiz0,tz);
 789
 790             fjx2             = _mm256_add_ps(fjx2,tx);
 791             fjy2             = _mm256_add_ps(fjy2,ty);
 792             fjz2             = _mm256_add_ps(fjz2,tz);
 793
 794             /**************************
 795              * CALCULATE INTERACTIONS *
 796              **************************/
 797
 798             /* REACTION-FIELD ELECTROSTATICS */
 799             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 800             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 801
 802             /* Update potential sum for this i atom from the interaction with this j atom. */
 803             velec            = _mm256_andnot_ps(dummy_mask,velec);
 804             velecsum         = _mm256_add_ps(velecsum,velec);
 805
 806             fscal            = felec;
 807
 808             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 809
 810             /* Calculate temporary vectorial force */
 811             tx               = _mm256_mul_ps(fscal,dx10);
 812             ty               = _mm256_mul_ps(fscal,dy10);
 813             tz               = _mm256_mul_ps(fscal,dz10);
 814
 815             /* Update vectorial force */
 816             fix1             = _mm256_add_ps(fix1,tx);
 817             fiy1             = _mm256_add_ps(fiy1,ty);
 818             fiz1             = _mm256_add_ps(fiz1,tz);
 819
 820             fjx0             = _mm256_add_ps(fjx0,tx);
 821             fjy0             = _mm256_add_ps(fjy0,ty);
 822             fjz0             = _mm256_add_ps(fjz0,tz);
 823
 824             /**************************
 825              * CALCULATE INTERACTIONS *
 826              **************************/
 827
 828             /* REACTION-FIELD ELECTROSTATICS */
 829             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 830             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 831
 832             /* Update potential sum for this i atom from the interaction with this j atom. */
 833             velec            = _mm256_andnot_ps(dummy_mask,velec);
 834             velecsum         = _mm256_add_ps(velecsum,velec);
 835
 836             fscal            = felec;
 837
 838             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 839
 840             /* Calculate temporary vectorial force */
 841             tx               = _mm256_mul_ps(fscal,dx11);
 842             ty               = _mm256_mul_ps(fscal,dy11);
 843             tz               = _mm256_mul_ps(fscal,dz11);
 844
 845             /* Update vectorial force */
 846             fix1             = _mm256_add_ps(fix1,tx);
 847             fiy1             = _mm256_add_ps(fiy1,ty);
 848             fiz1             = _mm256_add_ps(fiz1,tz);
 849
 850             fjx1             = _mm256_add_ps(fjx1,tx);
 851             fjy1             = _mm256_add_ps(fjy1,ty);
 852             fjz1             = _mm256_add_ps(fjz1,tz);
 853
 854             /**************************
 855              * CALCULATE INTERACTIONS *
 856              **************************/
 857
 858             /* REACTION-FIELD ELECTROSTATICS */
 859             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 860             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 861
 862             /* Update potential sum for this i atom from the interaction with this j atom. */
 863             velec            = _mm256_andnot_ps(dummy_mask,velec);
 864             velecsum         = _mm256_add_ps(velecsum,velec);
 865
 866             fscal            = felec;
 867
 868             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 869
 870             /* Calculate temporary vectorial force */
 871             tx               = _mm256_mul_ps(fscal,dx12);
 872             ty               = _mm256_mul_ps(fscal,dy12);
 873             tz               = _mm256_mul_ps(fscal,dz12);
 874
 875             /* Update vectorial force */
 876             fix1             = _mm256_add_ps(fix1,tx);
 877             fiy1             = _mm256_add_ps(fiy1,ty);
 878             fiz1             = _mm256_add_ps(fiz1,tz);
 879
 880             fjx2             = _mm256_add_ps(fjx2,tx);
 881             fjy2             = _mm256_add_ps(fjy2,ty);
 882             fjz2             = _mm256_add_ps(fjz2,tz);
 883
 884             /**************************
 885              * CALCULATE INTERACTIONS *
 886              **************************/
 887
 888             /* REACTION-FIELD ELECTROSTATICS */
 889             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 890             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 891
 892             /* Update potential sum for this i atom from the interaction with this j atom. */
 893             velec            = _mm256_andnot_ps(dummy_mask,velec);
 894             velecsum         = _mm256_add_ps(velecsum,velec);
 895
 896             fscal            = felec;
 897
 898             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 899
 900             /* Calculate temporary vectorial force */
 901             tx               = _mm256_mul_ps(fscal,dx20);
 902             ty               = _mm256_mul_ps(fscal,dy20);
 903             tz               = _mm256_mul_ps(fscal,dz20);
 904
 905             /* Update vectorial force */
 906             fix2             = _mm256_add_ps(fix2,tx);
 907             fiy2             = _mm256_add_ps(fiy2,ty);
 908             fiz2             = _mm256_add_ps(fiz2,tz);
 909
 910             fjx0             = _mm256_add_ps(fjx0,tx);
 911             fjy0             = _mm256_add_ps(fjy0,ty);
 912             fjz0             = _mm256_add_ps(fjz0,tz);
 913
 914             /**************************
 915              * CALCULATE INTERACTIONS *
 916              **************************/
 917
 918             /* REACTION-FIELD ELECTROSTATICS */
 919             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 920             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 921
 922             /* Update potential sum for this i atom from the interaction with this j atom. */
 923             velec            = _mm256_andnot_ps(dummy_mask,velec);
 924             velecsum         = _mm256_add_ps(velecsum,velec);
 925
 926             fscal            = felec;
 927
 928             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 929
 930             /* Calculate temporary vectorial force */
 931             tx               = _mm256_mul_ps(fscal,dx21);
 932             ty               = _mm256_mul_ps(fscal,dy21);
 933             tz               = _mm256_mul_ps(fscal,dz21);
 934
 935             /* Update vectorial force */
 936             fix2             = _mm256_add_ps(fix2,tx);
 937             fiy2             = _mm256_add_ps(fiy2,ty);
 938             fiz2             = _mm256_add_ps(fiz2,tz);
 939
 940             fjx1             = _mm256_add_ps(fjx1,tx);
 941             fjy1             = _mm256_add_ps(fjy1,ty);
 942             fjz1             = _mm256_add_ps(fjz1,tz);
 943
 944             /**************************
 945              * CALCULATE INTERACTIONS *
 946              **************************/
 947
 948             /* REACTION-FIELD ELECTROSTATICS */
 949             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 950             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 951
 952             /* Update potential sum for this i atom from the interaction with this j atom. */
 953             velec            = _mm256_andnot_ps(dummy_mask,velec);
 954             velecsum         = _mm256_add_ps(velecsum,velec);
 955
 956             fscal            = felec;
 957
 958             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 959
 960             /* Calculate temporary vectorial force */
 961             tx               = _mm256_mul_ps(fscal,dx22);
 962             ty               = _mm256_mul_ps(fscal,dy22);
 963             tz               = _mm256_mul_ps(fscal,dz22);
 964
 965             /* Update vectorial force */
 966             fix2             = _mm256_add_ps(fix2,tx);
 967             fiy2             = _mm256_add_ps(fiy2,ty);
 968             fiz2             = _mm256_add_ps(fiz2,tz);
 969
 970             fjx2             = _mm256_add_ps(fjx2,tx);
 971             fjy2             = _mm256_add_ps(fjy2,ty);
 972             fjz2             = _mm256_add_ps(fjz2,tz);
 973
 974             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 975             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 976             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 977             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 978             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 979             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 980             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 981             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 982
 983             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 984                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 985
 986             /* Inner loop uses 300 flops */
 987         }
 988
 989         /* End of innermost loop */
 990
 991         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 992                                                  f+i_coord_offset,fshift+i_shift_offset);
 993
 994         ggid                        = gid[iidx];
 995         /* Update potential energies */
 996         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 997         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 998
 999         /* Increment number of inner iterations */
1000         inneriter                  += j_index_end - j_index_start;
1001
1002         /* Outer loop uses 20 flops */
1003     }
1004
1005     /* Increment number of outer iterations */
1006     outeriter        += nri;
1007
1008     /* Update outer/inner flops */
1009
1010     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*300);
1011 }
1012 /*
1013  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW3W3_F_avx_256_single
1014  * Electrostatics interaction: ReactionField
1015  * VdW interaction:            LennardJones
1016  * Geometry:                   Water3-Water3
1017  * Calculate force/pot:        Force
1018  */
1019 void
1020 nb_kernel_ElecRF_VdwLJ_GeomW3W3_F_avx_256_single
1021                     (t_nblist                    * gmx_restrict       nlist,
1022                      rvec                        * gmx_restrict          xx,
1023                      rvec                        * gmx_restrict          ff,
1024                      t_forcerec                  * gmx_restrict          fr,
1025                      t_mdatoms                   * gmx_restrict     mdatoms,
1026                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1027                      t_nrnb                      * gmx_restrict        nrnb)
1028 {
1029     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1030      * just 0 for non-waters.
1031      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1032      * jnr indices corresponding to data put in the four positions in the SIMD register.
1033      */
1034     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1035     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1036     int              jnrA,jnrB,jnrC,jnrD;
1037     int              jnrE,jnrF,jnrG,jnrH;
1038     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1039     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1040     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1041     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1042     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1043     real             rcutoff_scalar;
1044     real             *shiftvec,*fshift,*x,*f;
1045     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1046     real             scratch[4*DIM];
1047     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1048     real *           vdwioffsetptr0;
1049     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1050     real *           vdwioffsetptr1;
1051     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1052     real *           vdwioffsetptr2;
1053     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1054     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1055     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1056     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1057     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1058     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1059     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1060     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1061     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1062     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1063     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1064     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1065     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1066     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1067     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1068     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1069     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1070     real             *charge;
1071     int              nvdwtype;
1072     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1073     int              *vdwtype;
1074     real             *vdwparam;
1075     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1076     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1077     __m256           dummy_mask,cutoff_mask;
1078     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1079     __m256           one     = _mm256_set1_ps(1.0);
1080     __m256           two     = _mm256_set1_ps(2.0);
1081     x                = xx[0];
1082     f                = ff[0];
1083
1084     nri              = nlist->nri;
1085     iinr             = nlist->iinr;
1086     jindex           = nlist->jindex;
1087     jjnr             = nlist->jjnr;
1088     shiftidx         = nlist->shift;
1089     gid              = nlist->gid;
1090     shiftvec         = fr->shift_vec[0];
1091     fshift           = fr->fshift[0];
1092     facel            = _mm256_set1_ps(fr->epsfac);
1093     charge           = mdatoms->chargeA;
1094     krf              = _mm256_set1_ps(fr->ic->k_rf);
1095     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1096     crf              = _mm256_set1_ps(fr->ic->c_rf);
1097     nvdwtype         = fr->ntype;
1098     vdwparam         = fr->nbfp;
1099     vdwtype          = mdatoms->typeA;
1100
1101     /* Setup water-specific parameters */
1102     inr              = nlist->iinr[0];
1103     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1104     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1105     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1106     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1107
1108     jq0              = _mm256_set1_ps(charge[inr+0]);
1109     jq1              = _mm256_set1_ps(charge[inr+1]);
1110     jq2              = _mm256_set1_ps(charge[inr+2]);
1111     vdwjidx0A        = 2*vdwtype[inr+0];
1112     qq00             = _mm256_mul_ps(iq0,jq0);
1113     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1114     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1115     qq01             = _mm256_mul_ps(iq0,jq1);
1116     qq02             = _mm256_mul_ps(iq0,jq2);
1117     qq10             = _mm256_mul_ps(iq1,jq0);
1118     qq11             = _mm256_mul_ps(iq1,jq1);
1119     qq12             = _mm256_mul_ps(iq1,jq2);
1120     qq20             = _mm256_mul_ps(iq2,jq0);
1121     qq21             = _mm256_mul_ps(iq2,jq1);
1122     qq22             = _mm256_mul_ps(iq2,jq2);
1123
1124     /* Avoid stupid compiler warnings */
1125     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1126     j_coord_offsetA = 0;
1127     j_coord_offsetB = 0;
1128     j_coord_offsetC = 0;
1129     j_coord_offsetD = 0;
1130     j_coord_offsetE = 0;
1131     j_coord_offsetF = 0;
1132     j_coord_offsetG = 0;
1133     j_coord_offsetH = 0;
1134
1135     outeriter        = 0;
1136     inneriter        = 0;
1137
1138     for(iidx=0;iidx<4*DIM;iidx++)
1139     {
1140         scratch[iidx] = 0.0;
1141     }
1142
1143     /* Start outer loop over neighborlists */
1144     for(iidx=0; iidx<nri; iidx++)
1145     {
1146         /* Load shift vector for this list */
1147         i_shift_offset   = DIM*shiftidx[iidx];
1148
1149         /* Load limits for loop over neighbors */
1150         j_index_start    = jindex[iidx];
1151         j_index_end      = jindex[iidx+1];
1152
1153         /* Get outer coordinate index */
1154         inr              = iinr[iidx];
1155         i_coord_offset   = DIM*inr;
1156
1157         /* Load i particle coords and add shift vector */
1158         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1159                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1160
1161         fix0             = _mm256_setzero_ps();
1162         fiy0             = _mm256_setzero_ps();
1163         fiz0             = _mm256_setzero_ps();
1164         fix1             = _mm256_setzero_ps();
1165         fiy1             = _mm256_setzero_ps();
1166         fiz1             = _mm256_setzero_ps();
1167         fix2             = _mm256_setzero_ps();
1168         fiy2             = _mm256_setzero_ps();
1169         fiz2             = _mm256_setzero_ps();
1170
1171         /* Start inner kernel loop */
1172         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1173         {
1174
1175             /* Get j neighbor index, and coordinate index */
1176             jnrA             = jjnr[jidx];
1177             jnrB             = jjnr[jidx+1];
1178             jnrC             = jjnr[jidx+2];
1179             jnrD             = jjnr[jidx+3];
1180             jnrE             = jjnr[jidx+4];
1181             jnrF             = jjnr[jidx+5];
1182             jnrG             = jjnr[jidx+6];
1183             jnrH             = jjnr[jidx+7];
1184             j_coord_offsetA  = DIM*jnrA;
1185             j_coord_offsetB  = DIM*jnrB;
1186             j_coord_offsetC  = DIM*jnrC;
1187             j_coord_offsetD  = DIM*jnrD;
1188             j_coord_offsetE  = DIM*jnrE;
1189             j_coord_offsetF  = DIM*jnrF;
1190             j_coord_offsetG  = DIM*jnrG;
1191             j_coord_offsetH  = DIM*jnrH;
1192
1193             /* load j atom coordinates */
1194             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1195                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1196                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1197                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1198                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1199
1200             /* Calculate displacement vector */
1201             dx00             = _mm256_sub_ps(ix0,jx0);
1202             dy00             = _mm256_sub_ps(iy0,jy0);
1203             dz00             = _mm256_sub_ps(iz0,jz0);
1204             dx01             = _mm256_sub_ps(ix0,jx1);
1205             dy01             = _mm256_sub_ps(iy0,jy1);
1206             dz01             = _mm256_sub_ps(iz0,jz1);
1207             dx02             = _mm256_sub_ps(ix0,jx2);
1208             dy02             = _mm256_sub_ps(iy0,jy2);
1209             dz02             = _mm256_sub_ps(iz0,jz2);
1210             dx10             = _mm256_sub_ps(ix1,jx0);
1211             dy10             = _mm256_sub_ps(iy1,jy0);
1212             dz10             = _mm256_sub_ps(iz1,jz0);
1213             dx11             = _mm256_sub_ps(ix1,jx1);
1214             dy11             = _mm256_sub_ps(iy1,jy1);
1215             dz11             = _mm256_sub_ps(iz1,jz1);
1216             dx12             = _mm256_sub_ps(ix1,jx2);
1217             dy12             = _mm256_sub_ps(iy1,jy2);
1218             dz12             = _mm256_sub_ps(iz1,jz2);
1219             dx20             = _mm256_sub_ps(ix2,jx0);
1220             dy20             = _mm256_sub_ps(iy2,jy0);
1221             dz20             = _mm256_sub_ps(iz2,jz0);
1222             dx21             = _mm256_sub_ps(ix2,jx1);
1223             dy21             = _mm256_sub_ps(iy2,jy1);
1224             dz21             = _mm256_sub_ps(iz2,jz1);
1225             dx22             = _mm256_sub_ps(ix2,jx2);
1226             dy22             = _mm256_sub_ps(iy2,jy2);
1227             dz22             = _mm256_sub_ps(iz2,jz2);
1228
1229             /* Calculate squared distance and things based on it */
1230             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1231             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1232             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1233             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1234             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1235             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1236             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1237             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1238             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1239
1240             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1241             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1242             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1243             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1244             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1245             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1246             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1247             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1248             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1249
1250             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1251             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1252             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1253             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1254             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1255             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1256             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1257             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1258             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1259
1260             fjx0             = _mm256_setzero_ps();
1261             fjy0             = _mm256_setzero_ps();
1262             fjz0             = _mm256_setzero_ps();
1263             fjx1             = _mm256_setzero_ps();
1264             fjy1             = _mm256_setzero_ps();
1265             fjz1             = _mm256_setzero_ps();
1266             fjx2             = _mm256_setzero_ps();
1267             fjy2             = _mm256_setzero_ps();
1268             fjz2             = _mm256_setzero_ps();
1269
1270             /**************************
1271              * CALCULATE INTERACTIONS *
1272              **************************/
1273
1274             /* REACTION-FIELD ELECTROSTATICS */
1275             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1276
1277             /* LENNARD-JONES DISPERSION/REPULSION */
1278
1279             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1280             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1281
1282             fscal            = _mm256_add_ps(felec,fvdw);
1283
1284             /* Calculate temporary vectorial force */
1285             tx               = _mm256_mul_ps(fscal,dx00);
1286             ty               = _mm256_mul_ps(fscal,dy00);
1287             tz               = _mm256_mul_ps(fscal,dz00);
1288
1289             /* Update vectorial force */
1290             fix0             = _mm256_add_ps(fix0,tx);
1291             fiy0             = _mm256_add_ps(fiy0,ty);
1292             fiz0             = _mm256_add_ps(fiz0,tz);
1293
1294             fjx0             = _mm256_add_ps(fjx0,tx);
1295             fjy0             = _mm256_add_ps(fjy0,ty);
1296             fjz0             = _mm256_add_ps(fjz0,tz);
1297
1298             /**************************
1299              * CALCULATE INTERACTIONS *
1300              **************************/
1301
1302             /* REACTION-FIELD ELECTROSTATICS */
1303             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
1304
1305             fscal            = felec;
1306
1307             /* Calculate temporary vectorial force */
1308             tx               = _mm256_mul_ps(fscal,dx01);
1309             ty               = _mm256_mul_ps(fscal,dy01);
1310             tz               = _mm256_mul_ps(fscal,dz01);
1311
1312             /* Update vectorial force */
1313             fix0             = _mm256_add_ps(fix0,tx);
1314             fiy0             = _mm256_add_ps(fiy0,ty);
1315             fiz0             = _mm256_add_ps(fiz0,tz);
1316
1317             fjx1             = _mm256_add_ps(fjx1,tx);
1318             fjy1             = _mm256_add_ps(fjy1,ty);
1319             fjz1             = _mm256_add_ps(fjz1,tz);
1320
1321             /**************************
1322              * CALCULATE INTERACTIONS *
1323              **************************/
1324
1325             /* REACTION-FIELD ELECTROSTATICS */
1326             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
1327
1328             fscal            = felec;
1329
1330             /* Calculate temporary vectorial force */
1331             tx               = _mm256_mul_ps(fscal,dx02);
1332             ty               = _mm256_mul_ps(fscal,dy02);
1333             tz               = _mm256_mul_ps(fscal,dz02);
1334
1335             /* Update vectorial force */
1336             fix0             = _mm256_add_ps(fix0,tx);
1337             fiy0             = _mm256_add_ps(fiy0,ty);
1338             fiz0             = _mm256_add_ps(fiz0,tz);
1339
1340             fjx2             = _mm256_add_ps(fjx2,tx);
1341             fjy2             = _mm256_add_ps(fjy2,ty);
1342             fjz2             = _mm256_add_ps(fjz2,tz);
1343
1344             /**************************
1345              * CALCULATE INTERACTIONS *
1346              **************************/
1347
1348             /* REACTION-FIELD ELECTROSTATICS */
1349             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1350
1351             fscal            = felec;
1352
1353             /* Calculate temporary vectorial force */
1354             tx               = _mm256_mul_ps(fscal,dx10);
1355             ty               = _mm256_mul_ps(fscal,dy10);
1356             tz               = _mm256_mul_ps(fscal,dz10);
1357
1358             /* Update vectorial force */
1359             fix1             = _mm256_add_ps(fix1,tx);
1360             fiy1             = _mm256_add_ps(fiy1,ty);
1361             fiz1             = _mm256_add_ps(fiz1,tz);
1362
1363             fjx0             = _mm256_add_ps(fjx0,tx);
1364             fjy0             = _mm256_add_ps(fjy0,ty);
1365             fjz0             = _mm256_add_ps(fjz0,tz);
1366
1367             /**************************
1368              * CALCULATE INTERACTIONS *
1369              **************************/
1370
1371             /* REACTION-FIELD ELECTROSTATICS */
1372             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1373
1374             fscal            = felec;
1375
1376             /* Calculate temporary vectorial force */
1377             tx               = _mm256_mul_ps(fscal,dx11);
1378             ty               = _mm256_mul_ps(fscal,dy11);
1379             tz               = _mm256_mul_ps(fscal,dz11);
1380
1381             /* Update vectorial force */
1382             fix1             = _mm256_add_ps(fix1,tx);
1383             fiy1             = _mm256_add_ps(fiy1,ty);
1384             fiz1             = _mm256_add_ps(fiz1,tz);
1385
1386             fjx1             = _mm256_add_ps(fjx1,tx);
1387             fjy1             = _mm256_add_ps(fjy1,ty);
1388             fjz1             = _mm256_add_ps(fjz1,tz);
1389
1390             /**************************
1391              * CALCULATE INTERACTIONS *
1392              **************************/
1393
1394             /* REACTION-FIELD ELECTROSTATICS */
1395             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1396
1397             fscal            = felec;
1398
1399             /* Calculate temporary vectorial force */
1400             tx               = _mm256_mul_ps(fscal,dx12);
1401             ty               = _mm256_mul_ps(fscal,dy12);
1402             tz               = _mm256_mul_ps(fscal,dz12);
1403
1404             /* Update vectorial force */
1405             fix1             = _mm256_add_ps(fix1,tx);
1406             fiy1             = _mm256_add_ps(fiy1,ty);
1407             fiz1             = _mm256_add_ps(fiz1,tz);
1408
1409             fjx2             = _mm256_add_ps(fjx2,tx);
1410             fjy2             = _mm256_add_ps(fjy2,ty);
1411             fjz2             = _mm256_add_ps(fjz2,tz);
1412
1413             /**************************
1414              * CALCULATE INTERACTIONS *
1415              **************************/
1416
1417             /* REACTION-FIELD ELECTROSTATICS */
1418             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1419
1420             fscal            = felec;
1421
1422             /* Calculate temporary vectorial force */
1423             tx               = _mm256_mul_ps(fscal,dx20);
1424             ty               = _mm256_mul_ps(fscal,dy20);
1425             tz               = _mm256_mul_ps(fscal,dz20);
1426
1427             /* Update vectorial force */
1428             fix2             = _mm256_add_ps(fix2,tx);
1429             fiy2             = _mm256_add_ps(fiy2,ty);
1430             fiz2             = _mm256_add_ps(fiz2,tz);
1431
1432             fjx0             = _mm256_add_ps(fjx0,tx);
1433             fjy0             = _mm256_add_ps(fjy0,ty);
1434             fjz0             = _mm256_add_ps(fjz0,tz);
1435
1436             /**************************
1437              * CALCULATE INTERACTIONS *
1438              **************************/
1439
1440             /* REACTION-FIELD ELECTROSTATICS */
1441             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1442
1443             fscal            = felec;
1444
1445             /* Calculate temporary vectorial force */
1446             tx               = _mm256_mul_ps(fscal,dx21);
1447             ty               = _mm256_mul_ps(fscal,dy21);
1448             tz               = _mm256_mul_ps(fscal,dz21);
1449
1450             /* Update vectorial force */
1451             fix2             = _mm256_add_ps(fix2,tx);
1452             fiy2             = _mm256_add_ps(fiy2,ty);
1453             fiz2             = _mm256_add_ps(fiz2,tz);
1454
1455             fjx1             = _mm256_add_ps(fjx1,tx);
1456             fjy1             = _mm256_add_ps(fjy1,ty);
1457             fjz1             = _mm256_add_ps(fjz1,tz);
1458
1459             /**************************
1460              * CALCULATE INTERACTIONS *
1461              **************************/
1462
1463             /* REACTION-FIELD ELECTROSTATICS */
1464             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1465
1466             fscal            = felec;
1467
1468             /* Calculate temporary vectorial force */
1469             tx               = _mm256_mul_ps(fscal,dx22);
1470             ty               = _mm256_mul_ps(fscal,dy22);
1471             tz               = _mm256_mul_ps(fscal,dz22);
1472
1473             /* Update vectorial force */
1474             fix2             = _mm256_add_ps(fix2,tx);
1475             fiy2             = _mm256_add_ps(fiy2,ty);
1476             fiz2             = _mm256_add_ps(fiz2,tz);
1477
1478             fjx2             = _mm256_add_ps(fjx2,tx);
1479             fjy2             = _mm256_add_ps(fjy2,ty);
1480             fjz2             = _mm256_add_ps(fjz2,tz);
1481
1482             fjptrA             = f+j_coord_offsetA;
1483             fjptrB             = f+j_coord_offsetB;
1484             fjptrC             = f+j_coord_offsetC;
1485             fjptrD             = f+j_coord_offsetD;
1486             fjptrE             = f+j_coord_offsetE;
1487             fjptrF             = f+j_coord_offsetF;
1488             fjptrG             = f+j_coord_offsetG;
1489             fjptrH             = f+j_coord_offsetH;
1490
1491             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1492                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1493
1494             /* Inner loop uses 250 flops */
1495         }
1496
1497         if(jidx<j_index_end)
1498         {
1499
1500             /* Get j neighbor index, and coordinate index */
1501             jnrlistA         = jjnr[jidx];
1502             jnrlistB         = jjnr[jidx+1];
1503             jnrlistC         = jjnr[jidx+2];
1504             jnrlistD         = jjnr[jidx+3];
1505             jnrlistE         = jjnr[jidx+4];
1506             jnrlistF         = jjnr[jidx+5];
1507             jnrlistG         = jjnr[jidx+6];
1508             jnrlistH         = jjnr[jidx+7];
1509             /* Sign of each element will be negative for non-real atoms.
1510              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1511              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1512              */
1513             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1514                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1515
1516             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1517             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1518             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1519             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1520             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1521             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1522             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1523             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1524             j_coord_offsetA  = DIM*jnrA;
1525             j_coord_offsetB  = DIM*jnrB;
1526             j_coord_offsetC  = DIM*jnrC;
1527             j_coord_offsetD  = DIM*jnrD;
1528             j_coord_offsetE  = DIM*jnrE;
1529             j_coord_offsetF  = DIM*jnrF;
1530             j_coord_offsetG  = DIM*jnrG;
1531             j_coord_offsetH  = DIM*jnrH;
1532
1533             /* load j atom coordinates */
1534             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1535                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1536                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1537                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1538                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1539
1540             /* Calculate displacement vector */
1541             dx00             = _mm256_sub_ps(ix0,jx0);
1542             dy00             = _mm256_sub_ps(iy0,jy0);
1543             dz00             = _mm256_sub_ps(iz0,jz0);
1544             dx01             = _mm256_sub_ps(ix0,jx1);
1545             dy01             = _mm256_sub_ps(iy0,jy1);
1546             dz01             = _mm256_sub_ps(iz0,jz1);
1547             dx02             = _mm256_sub_ps(ix0,jx2);
1548             dy02             = _mm256_sub_ps(iy0,jy2);
1549             dz02             = _mm256_sub_ps(iz0,jz2);
1550             dx10             = _mm256_sub_ps(ix1,jx0);
1551             dy10             = _mm256_sub_ps(iy1,jy0);
1552             dz10             = _mm256_sub_ps(iz1,jz0);
1553             dx11             = _mm256_sub_ps(ix1,jx1);
1554             dy11             = _mm256_sub_ps(iy1,jy1);
1555             dz11             = _mm256_sub_ps(iz1,jz1);
1556             dx12             = _mm256_sub_ps(ix1,jx2);
1557             dy12             = _mm256_sub_ps(iy1,jy2);
1558             dz12             = _mm256_sub_ps(iz1,jz2);
1559             dx20             = _mm256_sub_ps(ix2,jx0);
1560             dy20             = _mm256_sub_ps(iy2,jy0);
1561             dz20             = _mm256_sub_ps(iz2,jz0);
1562             dx21             = _mm256_sub_ps(ix2,jx1);
1563             dy21             = _mm256_sub_ps(iy2,jy1);
1564             dz21             = _mm256_sub_ps(iz2,jz1);
1565             dx22             = _mm256_sub_ps(ix2,jx2);
1566             dy22             = _mm256_sub_ps(iy2,jy2);
1567             dz22             = _mm256_sub_ps(iz2,jz2);
1568
1569             /* Calculate squared distance and things based on it */
1570             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1571             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1572             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1573             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1574             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1575             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1576             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1577             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1578             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1579
1580             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1581             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1582             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1583             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1584             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1585             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1586             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1587             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1588             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1589
1590             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1591             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1592             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1593             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1594             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1595             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1596             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1597             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1598             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1599
1600             fjx0             = _mm256_setzero_ps();
1601             fjy0             = _mm256_setzero_ps();
1602             fjz0             = _mm256_setzero_ps();
1603             fjx1             = _mm256_setzero_ps();
1604             fjy1             = _mm256_setzero_ps();
1605             fjz1             = _mm256_setzero_ps();
1606             fjx2             = _mm256_setzero_ps();
1607             fjy2             = _mm256_setzero_ps();
1608             fjz2             = _mm256_setzero_ps();
1609
1610             /**************************
1611              * CALCULATE INTERACTIONS *
1612              **************************/
1613
1614             /* REACTION-FIELD ELECTROSTATICS */
1615             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1616
1617             /* LENNARD-JONES DISPERSION/REPULSION */
1618
1619             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1620             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1621
1622             fscal            = _mm256_add_ps(felec,fvdw);
1623
1624             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1625
1626             /* Calculate temporary vectorial force */
1627             tx               = _mm256_mul_ps(fscal,dx00);
1628             ty               = _mm256_mul_ps(fscal,dy00);
1629             tz               = _mm256_mul_ps(fscal,dz00);
1630
1631             /* Update vectorial force */
1632             fix0             = _mm256_add_ps(fix0,tx);
1633             fiy0             = _mm256_add_ps(fiy0,ty);
1634             fiz0             = _mm256_add_ps(fiz0,tz);
1635
1636             fjx0             = _mm256_add_ps(fjx0,tx);
1637             fjy0             = _mm256_add_ps(fjy0,ty);
1638             fjz0             = _mm256_add_ps(fjz0,tz);
1639
1640             /**************************
1641              * CALCULATE INTERACTIONS *
1642              **************************/
1643
1644             /* REACTION-FIELD ELECTROSTATICS */
1645             felec            = _mm256_mul_ps(qq01,_mm256_sub_ps(_mm256_mul_ps(rinv01,rinvsq01),krf2));
1646
1647             fscal            = felec;
1648
1649             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1650
1651             /* Calculate temporary vectorial force */
1652             tx               = _mm256_mul_ps(fscal,dx01);
1653             ty               = _mm256_mul_ps(fscal,dy01);
1654             tz               = _mm256_mul_ps(fscal,dz01);
1655
1656             /* Update vectorial force */
1657             fix0             = _mm256_add_ps(fix0,tx);
1658             fiy0             = _mm256_add_ps(fiy0,ty);
1659             fiz0             = _mm256_add_ps(fiz0,tz);
1660
1661             fjx1             = _mm256_add_ps(fjx1,tx);
1662             fjy1             = _mm256_add_ps(fjy1,ty);
1663             fjz1             = _mm256_add_ps(fjz1,tz);
1664
1665             /**************************
1666              * CALCULATE INTERACTIONS *
1667              **************************/
1668
1669             /* REACTION-FIELD ELECTROSTATICS */
1670             felec            = _mm256_mul_ps(qq02,_mm256_sub_ps(_mm256_mul_ps(rinv02,rinvsq02),krf2));
1671
1672             fscal            = felec;
1673
1674             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1675
1676             /* Calculate temporary vectorial force */
1677             tx               = _mm256_mul_ps(fscal,dx02);
1678             ty               = _mm256_mul_ps(fscal,dy02);
1679             tz               = _mm256_mul_ps(fscal,dz02);
1680
1681             /* Update vectorial force */
1682             fix0             = _mm256_add_ps(fix0,tx);
1683             fiy0             = _mm256_add_ps(fiy0,ty);
1684             fiz0             = _mm256_add_ps(fiz0,tz);
1685
1686             fjx2             = _mm256_add_ps(fjx2,tx);
1687             fjy2             = _mm256_add_ps(fjy2,ty);
1688             fjz2             = _mm256_add_ps(fjz2,tz);
1689
1690             /**************************
1691              * CALCULATE INTERACTIONS *
1692              **************************/
1693
1694             /* REACTION-FIELD ELECTROSTATICS */
1695             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1696
1697             fscal            = felec;
1698
1699             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1700
1701             /* Calculate temporary vectorial force */
1702             tx               = _mm256_mul_ps(fscal,dx10);
1703             ty               = _mm256_mul_ps(fscal,dy10);
1704             tz               = _mm256_mul_ps(fscal,dz10);
1705
1706             /* Update vectorial force */
1707             fix1             = _mm256_add_ps(fix1,tx);
1708             fiy1             = _mm256_add_ps(fiy1,ty);
1709             fiz1             = _mm256_add_ps(fiz1,tz);
1710
1711             fjx0             = _mm256_add_ps(fjx0,tx);
1712             fjy0             = _mm256_add_ps(fjy0,ty);
1713             fjz0             = _mm256_add_ps(fjz0,tz);
1714
1715             /**************************
1716              * CALCULATE INTERACTIONS *
1717              **************************/
1718
1719             /* REACTION-FIELD ELECTROSTATICS */
1720             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1721
1722             fscal            = felec;
1723
1724             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1725
1726             /* Calculate temporary vectorial force */
1727             tx               = _mm256_mul_ps(fscal,dx11);
1728             ty               = _mm256_mul_ps(fscal,dy11);
1729             tz               = _mm256_mul_ps(fscal,dz11);
1730
1731             /* Update vectorial force */
1732             fix1             = _mm256_add_ps(fix1,tx);
1733             fiy1             = _mm256_add_ps(fiy1,ty);
1734             fiz1             = _mm256_add_ps(fiz1,tz);
1735
1736             fjx1             = _mm256_add_ps(fjx1,tx);
1737             fjy1             = _mm256_add_ps(fjy1,ty);
1738             fjz1             = _mm256_add_ps(fjz1,tz);
1739
1740             /**************************
1741              * CALCULATE INTERACTIONS *
1742              **************************/
1743
1744             /* REACTION-FIELD ELECTROSTATICS */
1745             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1746
1747             fscal            = felec;
1748
1749             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1750
1751             /* Calculate temporary vectorial force */
1752             tx               = _mm256_mul_ps(fscal,dx12);
1753             ty               = _mm256_mul_ps(fscal,dy12);
1754             tz               = _mm256_mul_ps(fscal,dz12);
1755
1756             /* Update vectorial force */
1757             fix1             = _mm256_add_ps(fix1,tx);
1758             fiy1             = _mm256_add_ps(fiy1,ty);
1759             fiz1             = _mm256_add_ps(fiz1,tz);
1760
1761             fjx2             = _mm256_add_ps(fjx2,tx);
1762             fjy2             = _mm256_add_ps(fjy2,ty);
1763             fjz2             = _mm256_add_ps(fjz2,tz);
1764
1765             /**************************
1766              * CALCULATE INTERACTIONS *
1767              **************************/
1768
1769             /* REACTION-FIELD ELECTROSTATICS */
1770             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1771
1772             fscal            = felec;
1773
1774             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1775
1776             /* Calculate temporary vectorial force */
1777             tx               = _mm256_mul_ps(fscal,dx20);
1778             ty               = _mm256_mul_ps(fscal,dy20);
1779             tz               = _mm256_mul_ps(fscal,dz20);
1780
1781             /* Update vectorial force */
1782             fix2             = _mm256_add_ps(fix2,tx);
1783             fiy2             = _mm256_add_ps(fiy2,ty);
1784             fiz2             = _mm256_add_ps(fiz2,tz);
1785
1786             fjx0             = _mm256_add_ps(fjx0,tx);
1787             fjy0             = _mm256_add_ps(fjy0,ty);
1788             fjz0             = _mm256_add_ps(fjz0,tz);
1789
1790             /**************************
1791              * CALCULATE INTERACTIONS *
1792              **************************/
1793
1794             /* REACTION-FIELD ELECTROSTATICS */
1795             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1796
1797             fscal            = felec;
1798
1799             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1800
1801             /* Calculate temporary vectorial force */
1802             tx               = _mm256_mul_ps(fscal,dx21);
1803             ty               = _mm256_mul_ps(fscal,dy21);
1804             tz               = _mm256_mul_ps(fscal,dz21);
1805
1806             /* Update vectorial force */
1807             fix2             = _mm256_add_ps(fix2,tx);
1808             fiy2             = _mm256_add_ps(fiy2,ty);
1809             fiz2             = _mm256_add_ps(fiz2,tz);
1810
1811             fjx1             = _mm256_add_ps(fjx1,tx);
1812             fjy1             = _mm256_add_ps(fjy1,ty);
1813             fjz1             = _mm256_add_ps(fjz1,tz);
1814
1815             /**************************
1816              * CALCULATE INTERACTIONS *
1817              **************************/
1818
1819             /* REACTION-FIELD ELECTROSTATICS */
1820             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1821
1822             fscal            = felec;
1823
1824             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1825
1826             /* Calculate temporary vectorial force */
1827             tx               = _mm256_mul_ps(fscal,dx22);
1828             ty               = _mm256_mul_ps(fscal,dy22);
1829             tz               = _mm256_mul_ps(fscal,dz22);
1830
1831             /* Update vectorial force */
1832             fix2             = _mm256_add_ps(fix2,tx);
1833             fiy2             = _mm256_add_ps(fiy2,ty);
1834             fiz2             = _mm256_add_ps(fiz2,tz);
1835
1836             fjx2             = _mm256_add_ps(fjx2,tx);
1837             fjy2             = _mm256_add_ps(fjy2,ty);
1838             fjz2             = _mm256_add_ps(fjz2,tz);
1839
1840             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1841             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1842             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1843             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1844             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1845             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1846             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1847             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1848
1849             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1850                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1851
1852             /* Inner loop uses 250 flops */
1853         }
1854
1855         /* End of innermost loop */
1856
1857         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1858                                                  f+i_coord_offset,fshift+i_shift_offset);
1859
1860         /* Increment number of inner iterations */
1861         inneriter                  += j_index_end - j_index_start;
1862
1863         /* Outer loop uses 18 flops */
1864     }
1865
1866     /* Increment number of outer iterations */
1867     outeriter        += nri;
1868
1869     /* Update outer/inner flops */
1870
1871     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*250);
1872 }