src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecRF_VdwLJ_GeomW4W4_avx_256_double.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_double 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_double.h"
  50 #include "kernelutil_x86_avx_256_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_avx_256_double
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_avx_256_double
  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 refer to j loop unrolling done with AVX, e.g. for the four 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              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  79     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  84     real             scratch[4*DIM];
  85     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  95     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  97     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  99     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
 101     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 102     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 103     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 104     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 105     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 106     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 107     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 108     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 109     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 110     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 111     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 112     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 113     real             *charge;
 114     int              nvdwtype;
 115     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 116     int              *vdwtype;
 117     real             *vdwparam;
 118     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 119     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 120     __m256d          dummy_mask,cutoff_mask;
 121     __m128           tmpmask0,tmpmask1;
 122     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 123     __m256d          one     = _mm256_set1_pd(1.0);
 124     __m256d          two     = _mm256_set1_pd(2.0);
 125     x                = xx[0];
 126     f                = ff[0];
 127
 128     nri              = nlist->nri;
 129     iinr             = nlist->iinr;
 130     jindex           = nlist->jindex;
 131     jjnr             = nlist->jjnr;
 132     shiftidx         = nlist->shift;
 133     gid              = nlist->gid;
 134     shiftvec         = fr->shift_vec[0];
 135     fshift           = fr->fshift[0];
 136     facel            = _mm256_set1_pd(fr->epsfac);
 137     charge           = mdatoms->chargeA;
 138     krf              = _mm256_set1_pd(fr->ic->k_rf);
 139     krf2             = _mm256_set1_pd(fr->ic->k_rf*2.0);
 140     crf              = _mm256_set1_pd(fr->ic->c_rf);
 141     nvdwtype         = fr->ntype;
 142     vdwparam         = fr->nbfp;
 143     vdwtype          = mdatoms->typeA;
 144
 145     /* Setup water-specific parameters */
 146     inr              = nlist->iinr[0];
 147     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 148     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 149     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 150     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 151
 152     jq1              = _mm256_set1_pd(charge[inr+1]);
 153     jq2              = _mm256_set1_pd(charge[inr+2]);
 154     jq3              = _mm256_set1_pd(charge[inr+3]);
 155     vdwjidx0A        = 2*vdwtype[inr+0];
 156     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 157     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 158     qq11             = _mm256_mul_pd(iq1,jq1);
 159     qq12             = _mm256_mul_pd(iq1,jq2);
 160     qq13             = _mm256_mul_pd(iq1,jq3);
 161     qq21             = _mm256_mul_pd(iq2,jq1);
 162     qq22             = _mm256_mul_pd(iq2,jq2);
 163     qq23             = _mm256_mul_pd(iq2,jq3);
 164     qq31             = _mm256_mul_pd(iq3,jq1);
 165     qq32             = _mm256_mul_pd(iq3,jq2);
 166     qq33             = _mm256_mul_pd(iq3,jq3);
 167
 168     /* Avoid stupid compiler warnings */
 169     jnrA = jnrB = jnrC = jnrD = 0;
 170     j_coord_offsetA = 0;
 171     j_coord_offsetB = 0;
 172     j_coord_offsetC = 0;
 173     j_coord_offsetD = 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_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 199                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 200
 201         fix0             = _mm256_setzero_pd();
 202         fiy0             = _mm256_setzero_pd();
 203         fiz0             = _mm256_setzero_pd();
 204         fix1             = _mm256_setzero_pd();
 205         fiy1             = _mm256_setzero_pd();
 206         fiz1             = _mm256_setzero_pd();
 207         fix2             = _mm256_setzero_pd();
 208         fiy2             = _mm256_setzero_pd();
 209         fiz2             = _mm256_setzero_pd();
 210         fix3             = _mm256_setzero_pd();
 211         fiy3             = _mm256_setzero_pd();
 212         fiz3             = _mm256_setzero_pd();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm256_setzero_pd();
 216         vvdwsum          = _mm256_setzero_pd();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 220         {
 221
 222             /* Get j neighbor index, and coordinate index */
 223             jnrA             = jjnr[jidx];
 224             jnrB             = jjnr[jidx+1];
 225             jnrC             = jjnr[jidx+2];
 226             jnrD             = jjnr[jidx+3];
 227             j_coord_offsetA  = DIM*jnrA;
 228             j_coord_offsetB  = DIM*jnrB;
 229             j_coord_offsetC  = DIM*jnrC;
 230             j_coord_offsetD  = DIM*jnrD;
 231
 232             /* load j atom coordinates */
 233             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 234                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 235                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 236                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 237
 238             /* Calculate displacement vector */
 239             dx00             = _mm256_sub_pd(ix0,jx0);
 240             dy00             = _mm256_sub_pd(iy0,jy0);
 241             dz00             = _mm256_sub_pd(iz0,jz0);
 242             dx11             = _mm256_sub_pd(ix1,jx1);
 243             dy11             = _mm256_sub_pd(iy1,jy1);
 244             dz11             = _mm256_sub_pd(iz1,jz1);
 245             dx12             = _mm256_sub_pd(ix1,jx2);
 246             dy12             = _mm256_sub_pd(iy1,jy2);
 247             dz12             = _mm256_sub_pd(iz1,jz2);
 248             dx13             = _mm256_sub_pd(ix1,jx3);
 249             dy13             = _mm256_sub_pd(iy1,jy3);
 250             dz13             = _mm256_sub_pd(iz1,jz3);
 251             dx21             = _mm256_sub_pd(ix2,jx1);
 252             dy21             = _mm256_sub_pd(iy2,jy1);
 253             dz21             = _mm256_sub_pd(iz2,jz1);
 254             dx22             = _mm256_sub_pd(ix2,jx2);
 255             dy22             = _mm256_sub_pd(iy2,jy2);
 256             dz22             = _mm256_sub_pd(iz2,jz2);
 257             dx23             = _mm256_sub_pd(ix2,jx3);
 258             dy23             = _mm256_sub_pd(iy2,jy3);
 259             dz23             = _mm256_sub_pd(iz2,jz3);
 260             dx31             = _mm256_sub_pd(ix3,jx1);
 261             dy31             = _mm256_sub_pd(iy3,jy1);
 262             dz31             = _mm256_sub_pd(iz3,jz1);
 263             dx32             = _mm256_sub_pd(ix3,jx2);
 264             dy32             = _mm256_sub_pd(iy3,jy2);
 265             dz32             = _mm256_sub_pd(iz3,jz2);
 266             dx33             = _mm256_sub_pd(ix3,jx3);
 267             dy33             = _mm256_sub_pd(iy3,jy3);
 268             dz33             = _mm256_sub_pd(iz3,jz3);
 269
 270             /* Calculate squared distance and things based on it */
 271             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 272             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 273             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 274             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 275             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 276             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 277             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 278             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 279             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 280             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 281
 282             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 283             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 284             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 285             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 286             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 287             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 288             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 289             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 290             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 291
 292             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 293             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 294             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 295             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 296             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 297             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 298             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 299             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 300             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 301             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 302
 303             fjx0             = _mm256_setzero_pd();
 304             fjy0             = _mm256_setzero_pd();
 305             fjz0             = _mm256_setzero_pd();
 306             fjx1             = _mm256_setzero_pd();
 307             fjy1             = _mm256_setzero_pd();
 308             fjz1             = _mm256_setzero_pd();
 309             fjx2             = _mm256_setzero_pd();
 310             fjy2             = _mm256_setzero_pd();
 311             fjz2             = _mm256_setzero_pd();
 312             fjx3             = _mm256_setzero_pd();
 313             fjy3             = _mm256_setzero_pd();
 314             fjz3             = _mm256_setzero_pd();
 315
 316             /**************************
 317              * CALCULATE INTERACTIONS *
 318              **************************/
 319
 320             /* LENNARD-JONES DISPERSION/REPULSION */
 321
 322             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 323             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 324             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 325             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 326             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 327
 328             /* Update potential sum for this i atom from the interaction with this j atom. */
 329             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 330
 331             fscal            = fvdw;
 332
 333             /* Calculate temporary vectorial force */
 334             tx               = _mm256_mul_pd(fscal,dx00);
 335             ty               = _mm256_mul_pd(fscal,dy00);
 336             tz               = _mm256_mul_pd(fscal,dz00);
 337
 338             /* Update vectorial force */
 339             fix0             = _mm256_add_pd(fix0,tx);
 340             fiy0             = _mm256_add_pd(fiy0,ty);
 341             fiz0             = _mm256_add_pd(fiz0,tz);
 342
 343             fjx0             = _mm256_add_pd(fjx0,tx);
 344             fjy0             = _mm256_add_pd(fjy0,ty);
 345             fjz0             = _mm256_add_pd(fjz0,tz);
 346
 347             /**************************
 348              * CALCULATE INTERACTIONS *
 349              **************************/
 350
 351             /* REACTION-FIELD ELECTROSTATICS */
 352             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
 353             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
 354
 355             /* Update potential sum for this i atom from the interaction with this j atom. */
 356             velecsum         = _mm256_add_pd(velecsum,velec);
 357
 358             fscal            = felec;
 359
 360             /* Calculate temporary vectorial force */
 361             tx               = _mm256_mul_pd(fscal,dx11);
 362             ty               = _mm256_mul_pd(fscal,dy11);
 363             tz               = _mm256_mul_pd(fscal,dz11);
 364
 365             /* Update vectorial force */
 366             fix1             = _mm256_add_pd(fix1,tx);
 367             fiy1             = _mm256_add_pd(fiy1,ty);
 368             fiz1             = _mm256_add_pd(fiz1,tz);
 369
 370             fjx1             = _mm256_add_pd(fjx1,tx);
 371             fjy1             = _mm256_add_pd(fjy1,ty);
 372             fjz1             = _mm256_add_pd(fjz1,tz);
 373
 374             /**************************
 375              * CALCULATE INTERACTIONS *
 376              **************************/
 377
 378             /* REACTION-FIELD ELECTROSTATICS */
 379             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
 380             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velecsum         = _mm256_add_pd(velecsum,velec);
 384
 385             fscal            = felec;
 386
 387             /* Calculate temporary vectorial force */
 388             tx               = _mm256_mul_pd(fscal,dx12);
 389             ty               = _mm256_mul_pd(fscal,dy12);
 390             tz               = _mm256_mul_pd(fscal,dz12);
 391
 392             /* Update vectorial force */
 393             fix1             = _mm256_add_pd(fix1,tx);
 394             fiy1             = _mm256_add_pd(fiy1,ty);
 395             fiz1             = _mm256_add_pd(fiz1,tz);
 396
 397             fjx2             = _mm256_add_pd(fjx2,tx);
 398             fjy2             = _mm256_add_pd(fjy2,ty);
 399             fjz2             = _mm256_add_pd(fjz2,tz);
 400
 401             /**************************
 402              * CALCULATE INTERACTIONS *
 403              **************************/
 404
 405             /* REACTION-FIELD ELECTROSTATICS */
 406             velec            = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_add_pd(rinv13,_mm256_mul_pd(krf,rsq13)),crf));
 407             felec            = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_mul_pd(rinv13,rinvsq13),krf2));
 408
 409             /* Update potential sum for this i atom from the interaction with this j atom. */
 410             velecsum         = _mm256_add_pd(velecsum,velec);
 411
 412             fscal            = felec;
 413
 414             /* Calculate temporary vectorial force */
 415             tx               = _mm256_mul_pd(fscal,dx13);
 416             ty               = _mm256_mul_pd(fscal,dy13);
 417             tz               = _mm256_mul_pd(fscal,dz13);
 418
 419             /* Update vectorial force */
 420             fix1             = _mm256_add_pd(fix1,tx);
 421             fiy1             = _mm256_add_pd(fiy1,ty);
 422             fiz1             = _mm256_add_pd(fiz1,tz);
 423
 424             fjx3             = _mm256_add_pd(fjx3,tx);
 425             fjy3             = _mm256_add_pd(fjy3,ty);
 426             fjz3             = _mm256_add_pd(fjz3,tz);
 427
 428             /**************************
 429              * CALCULATE INTERACTIONS *
 430              **************************/
 431
 432             /* REACTION-FIELD ELECTROSTATICS */
 433             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
 434             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
 435
 436             /* Update potential sum for this i atom from the interaction with this j atom. */
 437             velecsum         = _mm256_add_pd(velecsum,velec);
 438
 439             fscal            = felec;
 440
 441             /* Calculate temporary vectorial force */
 442             tx               = _mm256_mul_pd(fscal,dx21);
 443             ty               = _mm256_mul_pd(fscal,dy21);
 444             tz               = _mm256_mul_pd(fscal,dz21);
 445
 446             /* Update vectorial force */
 447             fix2             = _mm256_add_pd(fix2,tx);
 448             fiy2             = _mm256_add_pd(fiy2,ty);
 449             fiz2             = _mm256_add_pd(fiz2,tz);
 450
 451             fjx1             = _mm256_add_pd(fjx1,tx);
 452             fjy1             = _mm256_add_pd(fjy1,ty);
 453             fjz1             = _mm256_add_pd(fjz1,tz);
 454
 455             /**************************
 456              * CALCULATE INTERACTIONS *
 457              **************************/
 458
 459             /* REACTION-FIELD ELECTROSTATICS */
 460             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
 461             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
 462
 463             /* Update potential sum for this i atom from the interaction with this j atom. */
 464             velecsum         = _mm256_add_pd(velecsum,velec);
 465
 466             fscal            = felec;
 467
 468             /* Calculate temporary vectorial force */
 469             tx               = _mm256_mul_pd(fscal,dx22);
 470             ty               = _mm256_mul_pd(fscal,dy22);
 471             tz               = _mm256_mul_pd(fscal,dz22);
 472
 473             /* Update vectorial force */
 474             fix2             = _mm256_add_pd(fix2,tx);
 475             fiy2             = _mm256_add_pd(fiy2,ty);
 476             fiz2             = _mm256_add_pd(fiz2,tz);
 477
 478             fjx2             = _mm256_add_pd(fjx2,tx);
 479             fjy2             = _mm256_add_pd(fjy2,ty);
 480             fjz2             = _mm256_add_pd(fjz2,tz);
 481
 482             /**************************
 483              * CALCULATE INTERACTIONS *
 484              **************************/
 485
 486             /* REACTION-FIELD ELECTROSTATICS */
 487             velec            = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_add_pd(rinv23,_mm256_mul_pd(krf,rsq23)),crf));
 488             felec            = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_mul_pd(rinv23,rinvsq23),krf2));
 489
 490             /* Update potential sum for this i atom from the interaction with this j atom. */
 491             velecsum         = _mm256_add_pd(velecsum,velec);
 492
 493             fscal            = felec;
 494
 495             /* Calculate temporary vectorial force */
 496             tx               = _mm256_mul_pd(fscal,dx23);
 497             ty               = _mm256_mul_pd(fscal,dy23);
 498             tz               = _mm256_mul_pd(fscal,dz23);
 499
 500             /* Update vectorial force */
 501             fix2             = _mm256_add_pd(fix2,tx);
 502             fiy2             = _mm256_add_pd(fiy2,ty);
 503             fiz2             = _mm256_add_pd(fiz2,tz);
 504
 505             fjx3             = _mm256_add_pd(fjx3,tx);
 506             fjy3             = _mm256_add_pd(fjy3,ty);
 507             fjz3             = _mm256_add_pd(fjz3,tz);
 508
 509             /**************************
 510              * CALCULATE INTERACTIONS *
 511              **************************/
 512
 513             /* REACTION-FIELD ELECTROSTATICS */
 514             velec            = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_add_pd(rinv31,_mm256_mul_pd(krf,rsq31)),crf));
 515             felec            = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_mul_pd(rinv31,rinvsq31),krf2));
 516
 517             /* Update potential sum for this i atom from the interaction with this j atom. */
 518             velecsum         = _mm256_add_pd(velecsum,velec);
 519
 520             fscal            = felec;
 521
 522             /* Calculate temporary vectorial force */
 523             tx               = _mm256_mul_pd(fscal,dx31);
 524             ty               = _mm256_mul_pd(fscal,dy31);
 525             tz               = _mm256_mul_pd(fscal,dz31);
 526
 527             /* Update vectorial force */
 528             fix3             = _mm256_add_pd(fix3,tx);
 529             fiy3             = _mm256_add_pd(fiy3,ty);
 530             fiz3             = _mm256_add_pd(fiz3,tz);
 531
 532             fjx1             = _mm256_add_pd(fjx1,tx);
 533             fjy1             = _mm256_add_pd(fjy1,ty);
 534             fjz1             = _mm256_add_pd(fjz1,tz);
 535
 536             /**************************
 537              * CALCULATE INTERACTIONS *
 538              **************************/
 539
 540             /* REACTION-FIELD ELECTROSTATICS */
 541             velec            = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_add_pd(rinv32,_mm256_mul_pd(krf,rsq32)),crf));
 542             felec            = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_mul_pd(rinv32,rinvsq32),krf2));
 543
 544             /* Update potential sum for this i atom from the interaction with this j atom. */
 545             velecsum         = _mm256_add_pd(velecsum,velec);
 546
 547             fscal            = felec;
 548
 549             /* Calculate temporary vectorial force */
 550             tx               = _mm256_mul_pd(fscal,dx32);
 551             ty               = _mm256_mul_pd(fscal,dy32);
 552             tz               = _mm256_mul_pd(fscal,dz32);
 553
 554             /* Update vectorial force */
 555             fix3             = _mm256_add_pd(fix3,tx);
 556             fiy3             = _mm256_add_pd(fiy3,ty);
 557             fiz3             = _mm256_add_pd(fiz3,tz);
 558
 559             fjx2             = _mm256_add_pd(fjx2,tx);
 560             fjy2             = _mm256_add_pd(fjy2,ty);
 561             fjz2             = _mm256_add_pd(fjz2,tz);
 562
 563             /**************************
 564              * CALCULATE INTERACTIONS *
 565              **************************/
 566
 567             /* REACTION-FIELD ELECTROSTATICS */
 568             velec            = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_add_pd(rinv33,_mm256_mul_pd(krf,rsq33)),crf));
 569             felec            = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_mul_pd(rinv33,rinvsq33),krf2));
 570
 571             /* Update potential sum for this i atom from the interaction with this j atom. */
 572             velecsum         = _mm256_add_pd(velecsum,velec);
 573
 574             fscal            = felec;
 575
 576             /* Calculate temporary vectorial force */
 577             tx               = _mm256_mul_pd(fscal,dx33);
 578             ty               = _mm256_mul_pd(fscal,dy33);
 579             tz               = _mm256_mul_pd(fscal,dz33);
 580
 581             /* Update vectorial force */
 582             fix3             = _mm256_add_pd(fix3,tx);
 583             fiy3             = _mm256_add_pd(fiy3,ty);
 584             fiz3             = _mm256_add_pd(fiz3,tz);
 585
 586             fjx3             = _mm256_add_pd(fjx3,tx);
 587             fjy3             = _mm256_add_pd(fjy3,ty);
 588             fjz3             = _mm256_add_pd(fjz3,tz);
 589
 590             fjptrA             = f+j_coord_offsetA;
 591             fjptrB             = f+j_coord_offsetB;
 592             fjptrC             = f+j_coord_offsetC;
 593             fjptrD             = f+j_coord_offsetD;
 594
 595             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 596                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 597                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 598
 599             /* Inner loop uses 323 flops */
 600         }
 601
 602         if(jidx<j_index_end)
 603         {
 604
 605             /* Get j neighbor index, and coordinate index */
 606             jnrlistA         = jjnr[jidx];
 607             jnrlistB         = jjnr[jidx+1];
 608             jnrlistC         = jjnr[jidx+2];
 609             jnrlistD         = jjnr[jidx+3];
 610             /* Sign of each element will be negative for non-real atoms.
 611              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 612              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 613              */
 614             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 615
 616             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 617             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 618             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 619
 620             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 621             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 622             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 623             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 624             j_coord_offsetA  = DIM*jnrA;
 625             j_coord_offsetB  = DIM*jnrB;
 626             j_coord_offsetC  = DIM*jnrC;
 627             j_coord_offsetD  = DIM*jnrD;
 628
 629             /* load j atom coordinates */
 630             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 631                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 632                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 633                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 634
 635             /* Calculate displacement vector */
 636             dx00             = _mm256_sub_pd(ix0,jx0);
 637             dy00             = _mm256_sub_pd(iy0,jy0);
 638             dz00             = _mm256_sub_pd(iz0,jz0);
 639             dx11             = _mm256_sub_pd(ix1,jx1);
 640             dy11             = _mm256_sub_pd(iy1,jy1);
 641             dz11             = _mm256_sub_pd(iz1,jz1);
 642             dx12             = _mm256_sub_pd(ix1,jx2);
 643             dy12             = _mm256_sub_pd(iy1,jy2);
 644             dz12             = _mm256_sub_pd(iz1,jz2);
 645             dx13             = _mm256_sub_pd(ix1,jx3);
 646             dy13             = _mm256_sub_pd(iy1,jy3);
 647             dz13             = _mm256_sub_pd(iz1,jz3);
 648             dx21             = _mm256_sub_pd(ix2,jx1);
 649             dy21             = _mm256_sub_pd(iy2,jy1);
 650             dz21             = _mm256_sub_pd(iz2,jz1);
 651             dx22             = _mm256_sub_pd(ix2,jx2);
 652             dy22             = _mm256_sub_pd(iy2,jy2);
 653             dz22             = _mm256_sub_pd(iz2,jz2);
 654             dx23             = _mm256_sub_pd(ix2,jx3);
 655             dy23             = _mm256_sub_pd(iy2,jy3);
 656             dz23             = _mm256_sub_pd(iz2,jz3);
 657             dx31             = _mm256_sub_pd(ix3,jx1);
 658             dy31             = _mm256_sub_pd(iy3,jy1);
 659             dz31             = _mm256_sub_pd(iz3,jz1);
 660             dx32             = _mm256_sub_pd(ix3,jx2);
 661             dy32             = _mm256_sub_pd(iy3,jy2);
 662             dz32             = _mm256_sub_pd(iz3,jz2);
 663             dx33             = _mm256_sub_pd(ix3,jx3);
 664             dy33             = _mm256_sub_pd(iy3,jy3);
 665             dz33             = _mm256_sub_pd(iz3,jz3);
 666
 667             /* Calculate squared distance and things based on it */
 668             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 669             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 670             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 671             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 672             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 673             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 674             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 675             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 676             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 677             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 678
 679             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 680             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 681             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 682             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 683             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 684             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 685             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 686             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 687             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 688
 689             rinvsq00         = gmx_mm256_inv_pd(rsq00);
 690             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 691             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 692             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 693             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 694             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 695             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 696             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 697             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 698             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 699
 700             fjx0             = _mm256_setzero_pd();
 701             fjy0             = _mm256_setzero_pd();
 702             fjz0             = _mm256_setzero_pd();
 703             fjx1             = _mm256_setzero_pd();
 704             fjy1             = _mm256_setzero_pd();
 705             fjz1             = _mm256_setzero_pd();
 706             fjx2             = _mm256_setzero_pd();
 707             fjy2             = _mm256_setzero_pd();
 708             fjz2             = _mm256_setzero_pd();
 709             fjx3             = _mm256_setzero_pd();
 710             fjy3             = _mm256_setzero_pd();
 711             fjz3             = _mm256_setzero_pd();
 712
 713             /**************************
 714              * CALCULATE INTERACTIONS *
 715              **************************/
 716
 717             /* LENNARD-JONES DISPERSION/REPULSION */
 718
 719             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 720             vvdw6            = _mm256_mul_pd(c6_00,rinvsix);
 721             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 722             vvdw             = _mm256_sub_pd( _mm256_mul_pd(vvdw12,one_twelfth) , _mm256_mul_pd(vvdw6,one_sixth) );
 723             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,vvdw6),rinvsq00);
 724
 725             /* Update potential sum for this i atom from the interaction with this j atom. */
 726             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 727             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 728
 729             fscal            = fvdw;
 730
 731             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 732
 733             /* Calculate temporary vectorial force */
 734             tx               = _mm256_mul_pd(fscal,dx00);
 735             ty               = _mm256_mul_pd(fscal,dy00);
 736             tz               = _mm256_mul_pd(fscal,dz00);
 737
 738             /* Update vectorial force */
 739             fix0             = _mm256_add_pd(fix0,tx);
 740             fiy0             = _mm256_add_pd(fiy0,ty);
 741             fiz0             = _mm256_add_pd(fiz0,tz);
 742
 743             fjx0             = _mm256_add_pd(fjx0,tx);
 744             fjy0             = _mm256_add_pd(fjy0,ty);
 745             fjz0             = _mm256_add_pd(fjz0,tz);
 746
 747             /**************************
 748              * CALCULATE INTERACTIONS *
 749              **************************/
 750
 751             /* REACTION-FIELD ELECTROSTATICS */
 752             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
 753             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
 754
 755             /* Update potential sum for this i atom from the interaction with this j atom. */
 756             velec            = _mm256_andnot_pd(dummy_mask,velec);
 757             velecsum         = _mm256_add_pd(velecsum,velec);
 758
 759             fscal            = felec;
 760
 761             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 762
 763             /* Calculate temporary vectorial force */
 764             tx               = _mm256_mul_pd(fscal,dx11);
 765             ty               = _mm256_mul_pd(fscal,dy11);
 766             tz               = _mm256_mul_pd(fscal,dz11);
 767
 768             /* Update vectorial force */
 769             fix1             = _mm256_add_pd(fix1,tx);
 770             fiy1             = _mm256_add_pd(fiy1,ty);
 771             fiz1             = _mm256_add_pd(fiz1,tz);
 772
 773             fjx1             = _mm256_add_pd(fjx1,tx);
 774             fjy1             = _mm256_add_pd(fjy1,ty);
 775             fjz1             = _mm256_add_pd(fjz1,tz);
 776
 777             /**************************
 778              * CALCULATE INTERACTIONS *
 779              **************************/
 780
 781             /* REACTION-FIELD ELECTROSTATICS */
 782             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
 783             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
 784
 785             /* Update potential sum for this i atom from the interaction with this j atom. */
 786             velec            = _mm256_andnot_pd(dummy_mask,velec);
 787             velecsum         = _mm256_add_pd(velecsum,velec);
 788
 789             fscal            = felec;
 790
 791             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 792
 793             /* Calculate temporary vectorial force */
 794             tx               = _mm256_mul_pd(fscal,dx12);
 795             ty               = _mm256_mul_pd(fscal,dy12);
 796             tz               = _mm256_mul_pd(fscal,dz12);
 797
 798             /* Update vectorial force */
 799             fix1             = _mm256_add_pd(fix1,tx);
 800             fiy1             = _mm256_add_pd(fiy1,ty);
 801             fiz1             = _mm256_add_pd(fiz1,tz);
 802
 803             fjx2             = _mm256_add_pd(fjx2,tx);
 804             fjy2             = _mm256_add_pd(fjy2,ty);
 805             fjz2             = _mm256_add_pd(fjz2,tz);
 806
 807             /**************************
 808              * CALCULATE INTERACTIONS *
 809              **************************/
 810
 811             /* REACTION-FIELD ELECTROSTATICS */
 812             velec            = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_add_pd(rinv13,_mm256_mul_pd(krf,rsq13)),crf));
 813             felec            = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_mul_pd(rinv13,rinvsq13),krf2));
 814
 815             /* Update potential sum for this i atom from the interaction with this j atom. */
 816             velec            = _mm256_andnot_pd(dummy_mask,velec);
 817             velecsum         = _mm256_add_pd(velecsum,velec);
 818
 819             fscal            = felec;
 820
 821             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 822
 823             /* Calculate temporary vectorial force */
 824             tx               = _mm256_mul_pd(fscal,dx13);
 825             ty               = _mm256_mul_pd(fscal,dy13);
 826             tz               = _mm256_mul_pd(fscal,dz13);
 827
 828             /* Update vectorial force */
 829             fix1             = _mm256_add_pd(fix1,tx);
 830             fiy1             = _mm256_add_pd(fiy1,ty);
 831             fiz1             = _mm256_add_pd(fiz1,tz);
 832
 833             fjx3             = _mm256_add_pd(fjx3,tx);
 834             fjy3             = _mm256_add_pd(fjy3,ty);
 835             fjz3             = _mm256_add_pd(fjz3,tz);
 836
 837             /**************************
 838              * CALCULATE INTERACTIONS *
 839              **************************/
 840
 841             /* REACTION-FIELD ELECTROSTATICS */
 842             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
 843             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
 844
 845             /* Update potential sum for this i atom from the interaction with this j atom. */
 846             velec            = _mm256_andnot_pd(dummy_mask,velec);
 847             velecsum         = _mm256_add_pd(velecsum,velec);
 848
 849             fscal            = felec;
 850
 851             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 852
 853             /* Calculate temporary vectorial force */
 854             tx               = _mm256_mul_pd(fscal,dx21);
 855             ty               = _mm256_mul_pd(fscal,dy21);
 856             tz               = _mm256_mul_pd(fscal,dz21);
 857
 858             /* Update vectorial force */
 859             fix2             = _mm256_add_pd(fix2,tx);
 860             fiy2             = _mm256_add_pd(fiy2,ty);
 861             fiz2             = _mm256_add_pd(fiz2,tz);
 862
 863             fjx1             = _mm256_add_pd(fjx1,tx);
 864             fjy1             = _mm256_add_pd(fjy1,ty);
 865             fjz1             = _mm256_add_pd(fjz1,tz);
 866
 867             /**************************
 868              * CALCULATE INTERACTIONS *
 869              **************************/
 870
 871             /* REACTION-FIELD ELECTROSTATICS */
 872             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
 873             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
 874
 875             /* Update potential sum for this i atom from the interaction with this j atom. */
 876             velec            = _mm256_andnot_pd(dummy_mask,velec);
 877             velecsum         = _mm256_add_pd(velecsum,velec);
 878
 879             fscal            = felec;
 880
 881             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 882
 883             /* Calculate temporary vectorial force */
 884             tx               = _mm256_mul_pd(fscal,dx22);
 885             ty               = _mm256_mul_pd(fscal,dy22);
 886             tz               = _mm256_mul_pd(fscal,dz22);
 887
 888             /* Update vectorial force */
 889             fix2             = _mm256_add_pd(fix2,tx);
 890             fiy2             = _mm256_add_pd(fiy2,ty);
 891             fiz2             = _mm256_add_pd(fiz2,tz);
 892
 893             fjx2             = _mm256_add_pd(fjx2,tx);
 894             fjy2             = _mm256_add_pd(fjy2,ty);
 895             fjz2             = _mm256_add_pd(fjz2,tz);
 896
 897             /**************************
 898              * CALCULATE INTERACTIONS *
 899              **************************/
 900
 901             /* REACTION-FIELD ELECTROSTATICS */
 902             velec            = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_add_pd(rinv23,_mm256_mul_pd(krf,rsq23)),crf));
 903             felec            = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_mul_pd(rinv23,rinvsq23),krf2));
 904
 905             /* Update potential sum for this i atom from the interaction with this j atom. */
 906             velec            = _mm256_andnot_pd(dummy_mask,velec);
 907             velecsum         = _mm256_add_pd(velecsum,velec);
 908
 909             fscal            = felec;
 910
 911             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 912
 913             /* Calculate temporary vectorial force */
 914             tx               = _mm256_mul_pd(fscal,dx23);
 915             ty               = _mm256_mul_pd(fscal,dy23);
 916             tz               = _mm256_mul_pd(fscal,dz23);
 917
 918             /* Update vectorial force */
 919             fix2             = _mm256_add_pd(fix2,tx);
 920             fiy2             = _mm256_add_pd(fiy2,ty);
 921             fiz2             = _mm256_add_pd(fiz2,tz);
 922
 923             fjx3             = _mm256_add_pd(fjx3,tx);
 924             fjy3             = _mm256_add_pd(fjy3,ty);
 925             fjz3             = _mm256_add_pd(fjz3,tz);
 926
 927             /**************************
 928              * CALCULATE INTERACTIONS *
 929              **************************/
 930
 931             /* REACTION-FIELD ELECTROSTATICS */
 932             velec            = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_add_pd(rinv31,_mm256_mul_pd(krf,rsq31)),crf));
 933             felec            = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_mul_pd(rinv31,rinvsq31),krf2));
 934
 935             /* Update potential sum for this i atom from the interaction with this j atom. */
 936             velec            = _mm256_andnot_pd(dummy_mask,velec);
 937             velecsum         = _mm256_add_pd(velecsum,velec);
 938
 939             fscal            = felec;
 940
 941             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 942
 943             /* Calculate temporary vectorial force */
 944             tx               = _mm256_mul_pd(fscal,dx31);
 945             ty               = _mm256_mul_pd(fscal,dy31);
 946             tz               = _mm256_mul_pd(fscal,dz31);
 947
 948             /* Update vectorial force */
 949             fix3             = _mm256_add_pd(fix3,tx);
 950             fiy3             = _mm256_add_pd(fiy3,ty);
 951             fiz3             = _mm256_add_pd(fiz3,tz);
 952
 953             fjx1             = _mm256_add_pd(fjx1,tx);
 954             fjy1             = _mm256_add_pd(fjy1,ty);
 955             fjz1             = _mm256_add_pd(fjz1,tz);
 956
 957             /**************************
 958              * CALCULATE INTERACTIONS *
 959              **************************/
 960
 961             /* REACTION-FIELD ELECTROSTATICS */
 962             velec            = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_add_pd(rinv32,_mm256_mul_pd(krf,rsq32)),crf));
 963             felec            = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_mul_pd(rinv32,rinvsq32),krf2));
 964
 965             /* Update potential sum for this i atom from the interaction with this j atom. */
 966             velec            = _mm256_andnot_pd(dummy_mask,velec);
 967             velecsum         = _mm256_add_pd(velecsum,velec);
 968
 969             fscal            = felec;
 970
 971             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 972
 973             /* Calculate temporary vectorial force */
 974             tx               = _mm256_mul_pd(fscal,dx32);
 975             ty               = _mm256_mul_pd(fscal,dy32);
 976             tz               = _mm256_mul_pd(fscal,dz32);
 977
 978             /* Update vectorial force */
 979             fix3             = _mm256_add_pd(fix3,tx);
 980             fiy3             = _mm256_add_pd(fiy3,ty);
 981             fiz3             = _mm256_add_pd(fiz3,tz);
 982
 983             fjx2             = _mm256_add_pd(fjx2,tx);
 984             fjy2             = _mm256_add_pd(fjy2,ty);
 985             fjz2             = _mm256_add_pd(fjz2,tz);
 986
 987             /**************************
 988              * CALCULATE INTERACTIONS *
 989              **************************/
 990
 991             /* REACTION-FIELD ELECTROSTATICS */
 992             velec            = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_add_pd(rinv33,_mm256_mul_pd(krf,rsq33)),crf));
 993             felec            = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_mul_pd(rinv33,rinvsq33),krf2));
 994
 995             /* Update potential sum for this i atom from the interaction with this j atom. */
 996             velec            = _mm256_andnot_pd(dummy_mask,velec);
 997             velecsum         = _mm256_add_pd(velecsum,velec);
 998
 999             fscal            = felec;
1000
1001             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1002
1003             /* Calculate temporary vectorial force */
1004             tx               = _mm256_mul_pd(fscal,dx33);
1005             ty               = _mm256_mul_pd(fscal,dy33);
1006             tz               = _mm256_mul_pd(fscal,dz33);
1007
1008             /* Update vectorial force */
1009             fix3             = _mm256_add_pd(fix3,tx);
1010             fiy3             = _mm256_add_pd(fiy3,ty);
1011             fiz3             = _mm256_add_pd(fiz3,tz);
1012
1013             fjx3             = _mm256_add_pd(fjx3,tx);
1014             fjy3             = _mm256_add_pd(fjy3,ty);
1015             fjz3             = _mm256_add_pd(fjz3,tz);
1016
1017             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1018             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1019             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1020             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1021
1022             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1023                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1024                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1025
1026             /* Inner loop uses 323 flops */
1027         }
1028
1029         /* End of innermost loop */
1030
1031         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1032                                                  f+i_coord_offset,fshift+i_shift_offset);
1033
1034         ggid                        = gid[iidx];
1035         /* Update potential energies */
1036         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1037         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1038
1039         /* Increment number of inner iterations */
1040         inneriter                  += j_index_end - j_index_start;
1041
1042         /* Outer loop uses 26 flops */
1043     }
1044
1045     /* Increment number of outer iterations */
1046     outeriter        += nri;
1047
1048     /* Update outer/inner flops */
1049
1050     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*323);
1051 }
1052 /*
1053  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_double
1054  * Electrostatics interaction: ReactionField
1055  * VdW interaction:            LennardJones
1056  * Geometry:                   Water4-Water4
1057  * Calculate force/pot:        Force
1058  */
1059 void
1060 nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_double
1061                     (t_nblist                    * gmx_restrict       nlist,
1062                      rvec                        * gmx_restrict          xx,
1063                      rvec                        * gmx_restrict          ff,
1064                      t_forcerec                  * gmx_restrict          fr,
1065                      t_mdatoms                   * gmx_restrict     mdatoms,
1066                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1067                      t_nrnb                      * gmx_restrict        nrnb)
1068 {
1069     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1070      * just 0 for non-waters.
1071      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1072      * jnr indices corresponding to data put in the four positions in the SIMD register.
1073      */
1074     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1075     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1076     int              jnrA,jnrB,jnrC,jnrD;
1077     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1078     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1079     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1080     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1081     real             rcutoff_scalar;
1082     real             *shiftvec,*fshift,*x,*f;
1083     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1084     real             scratch[4*DIM];
1085     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1086     real *           vdwioffsetptr0;
1087     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1088     real *           vdwioffsetptr1;
1089     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1090     real *           vdwioffsetptr2;
1091     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1092     real *           vdwioffsetptr3;
1093     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1094     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1095     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1096     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1097     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1098     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1099     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1100     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1101     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1102     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1103     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1104     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1105     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1106     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1107     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1108     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1109     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1110     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1111     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1112     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1113     real             *charge;
1114     int              nvdwtype;
1115     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1116     int              *vdwtype;
1117     real             *vdwparam;
1118     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1119     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1120     __m256d          dummy_mask,cutoff_mask;
1121     __m128           tmpmask0,tmpmask1;
1122     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1123     __m256d          one     = _mm256_set1_pd(1.0);
1124     __m256d          two     = _mm256_set1_pd(2.0);
1125     x                = xx[0];
1126     f                = ff[0];
1127
1128     nri              = nlist->nri;
1129     iinr             = nlist->iinr;
1130     jindex           = nlist->jindex;
1131     jjnr             = nlist->jjnr;
1132     shiftidx         = nlist->shift;
1133     gid              = nlist->gid;
1134     shiftvec         = fr->shift_vec[0];
1135     fshift           = fr->fshift[0];
1136     facel            = _mm256_set1_pd(fr->epsfac);
1137     charge           = mdatoms->chargeA;
1138     krf              = _mm256_set1_pd(fr->ic->k_rf);
1139     krf2             = _mm256_set1_pd(fr->ic->k_rf*2.0);
1140     crf              = _mm256_set1_pd(fr->ic->c_rf);
1141     nvdwtype         = fr->ntype;
1142     vdwparam         = fr->nbfp;
1143     vdwtype          = mdatoms->typeA;
1144
1145     /* Setup water-specific parameters */
1146     inr              = nlist->iinr[0];
1147     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1148     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1149     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1150     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1151
1152     jq1              = _mm256_set1_pd(charge[inr+1]);
1153     jq2              = _mm256_set1_pd(charge[inr+2]);
1154     jq3              = _mm256_set1_pd(charge[inr+3]);
1155     vdwjidx0A        = 2*vdwtype[inr+0];
1156     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1157     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1158     qq11             = _mm256_mul_pd(iq1,jq1);
1159     qq12             = _mm256_mul_pd(iq1,jq2);
1160     qq13             = _mm256_mul_pd(iq1,jq3);
1161     qq21             = _mm256_mul_pd(iq2,jq1);
1162     qq22             = _mm256_mul_pd(iq2,jq2);
1163     qq23             = _mm256_mul_pd(iq2,jq3);
1164     qq31             = _mm256_mul_pd(iq3,jq1);
1165     qq32             = _mm256_mul_pd(iq3,jq2);
1166     qq33             = _mm256_mul_pd(iq3,jq3);
1167
1168     /* Avoid stupid compiler warnings */
1169     jnrA = jnrB = jnrC = jnrD = 0;
1170     j_coord_offsetA = 0;
1171     j_coord_offsetB = 0;
1172     j_coord_offsetC = 0;
1173     j_coord_offsetD = 0;
1174
1175     outeriter        = 0;
1176     inneriter        = 0;
1177
1178     for(iidx=0;iidx<4*DIM;iidx++)
1179     {
1180         scratch[iidx] = 0.0;
1181     }
1182
1183     /* Start outer loop over neighborlists */
1184     for(iidx=0; iidx<nri; iidx++)
1185     {
1186         /* Load shift vector for this list */
1187         i_shift_offset   = DIM*shiftidx[iidx];
1188
1189         /* Load limits for loop over neighbors */
1190         j_index_start    = jindex[iidx];
1191         j_index_end      = jindex[iidx+1];
1192
1193         /* Get outer coordinate index */
1194         inr              = iinr[iidx];
1195         i_coord_offset   = DIM*inr;
1196
1197         /* Load i particle coords and add shift vector */
1198         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1199                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1200
1201         fix0             = _mm256_setzero_pd();
1202         fiy0             = _mm256_setzero_pd();
1203         fiz0             = _mm256_setzero_pd();
1204         fix1             = _mm256_setzero_pd();
1205         fiy1             = _mm256_setzero_pd();
1206         fiz1             = _mm256_setzero_pd();
1207         fix2             = _mm256_setzero_pd();
1208         fiy2             = _mm256_setzero_pd();
1209         fiz2             = _mm256_setzero_pd();
1210         fix3             = _mm256_setzero_pd();
1211         fiy3             = _mm256_setzero_pd();
1212         fiz3             = _mm256_setzero_pd();
1213
1214         /* Start inner kernel loop */
1215         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1216         {
1217
1218             /* Get j neighbor index, and coordinate index */
1219             jnrA             = jjnr[jidx];
1220             jnrB             = jjnr[jidx+1];
1221             jnrC             = jjnr[jidx+2];
1222             jnrD             = jjnr[jidx+3];
1223             j_coord_offsetA  = DIM*jnrA;
1224             j_coord_offsetB  = DIM*jnrB;
1225             j_coord_offsetC  = DIM*jnrC;
1226             j_coord_offsetD  = DIM*jnrD;
1227
1228             /* load j atom coordinates */
1229             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1230                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1231                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1232                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1233
1234             /* Calculate displacement vector */
1235             dx00             = _mm256_sub_pd(ix0,jx0);
1236             dy00             = _mm256_sub_pd(iy0,jy0);
1237             dz00             = _mm256_sub_pd(iz0,jz0);
1238             dx11             = _mm256_sub_pd(ix1,jx1);
1239             dy11             = _mm256_sub_pd(iy1,jy1);
1240             dz11             = _mm256_sub_pd(iz1,jz1);
1241             dx12             = _mm256_sub_pd(ix1,jx2);
1242             dy12             = _mm256_sub_pd(iy1,jy2);
1243             dz12             = _mm256_sub_pd(iz1,jz2);
1244             dx13             = _mm256_sub_pd(ix1,jx3);
1245             dy13             = _mm256_sub_pd(iy1,jy3);
1246             dz13             = _mm256_sub_pd(iz1,jz3);
1247             dx21             = _mm256_sub_pd(ix2,jx1);
1248             dy21             = _mm256_sub_pd(iy2,jy1);
1249             dz21             = _mm256_sub_pd(iz2,jz1);
1250             dx22             = _mm256_sub_pd(ix2,jx2);
1251             dy22             = _mm256_sub_pd(iy2,jy2);
1252             dz22             = _mm256_sub_pd(iz2,jz2);
1253             dx23             = _mm256_sub_pd(ix2,jx3);
1254             dy23             = _mm256_sub_pd(iy2,jy3);
1255             dz23             = _mm256_sub_pd(iz2,jz3);
1256             dx31             = _mm256_sub_pd(ix3,jx1);
1257             dy31             = _mm256_sub_pd(iy3,jy1);
1258             dz31             = _mm256_sub_pd(iz3,jz1);
1259             dx32             = _mm256_sub_pd(ix3,jx2);
1260             dy32             = _mm256_sub_pd(iy3,jy2);
1261             dz32             = _mm256_sub_pd(iz3,jz2);
1262             dx33             = _mm256_sub_pd(ix3,jx3);
1263             dy33             = _mm256_sub_pd(iy3,jy3);
1264             dz33             = _mm256_sub_pd(iz3,jz3);
1265
1266             /* Calculate squared distance and things based on it */
1267             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1268             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1269             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1270             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1271             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1272             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1273             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1274             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1275             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1276             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1277
1278             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1279             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1280             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1281             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1282             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1283             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1284             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1285             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1286             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1287
1288             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1289             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1290             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1291             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1292             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1293             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1294             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1295             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1296             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1297             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1298
1299             fjx0             = _mm256_setzero_pd();
1300             fjy0             = _mm256_setzero_pd();
1301             fjz0             = _mm256_setzero_pd();
1302             fjx1             = _mm256_setzero_pd();
1303             fjy1             = _mm256_setzero_pd();
1304             fjz1             = _mm256_setzero_pd();
1305             fjx2             = _mm256_setzero_pd();
1306             fjy2             = _mm256_setzero_pd();
1307             fjz2             = _mm256_setzero_pd();
1308             fjx3             = _mm256_setzero_pd();
1309             fjy3             = _mm256_setzero_pd();
1310             fjz3             = _mm256_setzero_pd();
1311
1312             /**************************
1313              * CALCULATE INTERACTIONS *
1314              **************************/
1315
1316             /* LENNARD-JONES DISPERSION/REPULSION */
1317
1318             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1319             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1320
1321             fscal            = fvdw;
1322
1323             /* Calculate temporary vectorial force */
1324             tx               = _mm256_mul_pd(fscal,dx00);
1325             ty               = _mm256_mul_pd(fscal,dy00);
1326             tz               = _mm256_mul_pd(fscal,dz00);
1327
1328             /* Update vectorial force */
1329             fix0             = _mm256_add_pd(fix0,tx);
1330             fiy0             = _mm256_add_pd(fiy0,ty);
1331             fiz0             = _mm256_add_pd(fiz0,tz);
1332
1333             fjx0             = _mm256_add_pd(fjx0,tx);
1334             fjy0             = _mm256_add_pd(fjy0,ty);
1335             fjz0             = _mm256_add_pd(fjz0,tz);
1336
1337             /**************************
1338              * CALCULATE INTERACTIONS *
1339              **************************/
1340
1341             /* REACTION-FIELD ELECTROSTATICS */
1342             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1343
1344             fscal            = felec;
1345
1346             /* Calculate temporary vectorial force */
1347             tx               = _mm256_mul_pd(fscal,dx11);
1348             ty               = _mm256_mul_pd(fscal,dy11);
1349             tz               = _mm256_mul_pd(fscal,dz11);
1350
1351             /* Update vectorial force */
1352             fix1             = _mm256_add_pd(fix1,tx);
1353             fiy1             = _mm256_add_pd(fiy1,ty);
1354             fiz1             = _mm256_add_pd(fiz1,tz);
1355
1356             fjx1             = _mm256_add_pd(fjx1,tx);
1357             fjy1             = _mm256_add_pd(fjy1,ty);
1358             fjz1             = _mm256_add_pd(fjz1,tz);
1359
1360             /**************************
1361              * CALCULATE INTERACTIONS *
1362              **************************/
1363
1364             /* REACTION-FIELD ELECTROSTATICS */
1365             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1366
1367             fscal            = felec;
1368
1369             /* Calculate temporary vectorial force */
1370             tx               = _mm256_mul_pd(fscal,dx12);
1371             ty               = _mm256_mul_pd(fscal,dy12);
1372             tz               = _mm256_mul_pd(fscal,dz12);
1373
1374             /* Update vectorial force */
1375             fix1             = _mm256_add_pd(fix1,tx);
1376             fiy1             = _mm256_add_pd(fiy1,ty);
1377             fiz1             = _mm256_add_pd(fiz1,tz);
1378
1379             fjx2             = _mm256_add_pd(fjx2,tx);
1380             fjy2             = _mm256_add_pd(fjy2,ty);
1381             fjz2             = _mm256_add_pd(fjz2,tz);
1382
1383             /**************************
1384              * CALCULATE INTERACTIONS *
1385              **************************/
1386
1387             /* REACTION-FIELD ELECTROSTATICS */
1388             felec            = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_mul_pd(rinv13,rinvsq13),krf2));
1389
1390             fscal            = felec;
1391
1392             /* Calculate temporary vectorial force */
1393             tx               = _mm256_mul_pd(fscal,dx13);
1394             ty               = _mm256_mul_pd(fscal,dy13);
1395             tz               = _mm256_mul_pd(fscal,dz13);
1396
1397             /* Update vectorial force */
1398             fix1             = _mm256_add_pd(fix1,tx);
1399             fiy1             = _mm256_add_pd(fiy1,ty);
1400             fiz1             = _mm256_add_pd(fiz1,tz);
1401
1402             fjx3             = _mm256_add_pd(fjx3,tx);
1403             fjy3             = _mm256_add_pd(fjy3,ty);
1404             fjz3             = _mm256_add_pd(fjz3,tz);
1405
1406             /**************************
1407              * CALCULATE INTERACTIONS *
1408              **************************/
1409
1410             /* REACTION-FIELD ELECTROSTATICS */
1411             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1412
1413             fscal            = felec;
1414
1415             /* Calculate temporary vectorial force */
1416             tx               = _mm256_mul_pd(fscal,dx21);
1417             ty               = _mm256_mul_pd(fscal,dy21);
1418             tz               = _mm256_mul_pd(fscal,dz21);
1419
1420             /* Update vectorial force */
1421             fix2             = _mm256_add_pd(fix2,tx);
1422             fiy2             = _mm256_add_pd(fiy2,ty);
1423             fiz2             = _mm256_add_pd(fiz2,tz);
1424
1425             fjx1             = _mm256_add_pd(fjx1,tx);
1426             fjy1             = _mm256_add_pd(fjy1,ty);
1427             fjz1             = _mm256_add_pd(fjz1,tz);
1428
1429             /**************************
1430              * CALCULATE INTERACTIONS *
1431              **************************/
1432
1433             /* REACTION-FIELD ELECTROSTATICS */
1434             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1435
1436             fscal            = felec;
1437
1438             /* Calculate temporary vectorial force */
1439             tx               = _mm256_mul_pd(fscal,dx22);
1440             ty               = _mm256_mul_pd(fscal,dy22);
1441             tz               = _mm256_mul_pd(fscal,dz22);
1442
1443             /* Update vectorial force */
1444             fix2             = _mm256_add_pd(fix2,tx);
1445             fiy2             = _mm256_add_pd(fiy2,ty);
1446             fiz2             = _mm256_add_pd(fiz2,tz);
1447
1448             fjx2             = _mm256_add_pd(fjx2,tx);
1449             fjy2             = _mm256_add_pd(fjy2,ty);
1450             fjz2             = _mm256_add_pd(fjz2,tz);
1451
1452             /**************************
1453              * CALCULATE INTERACTIONS *
1454              **************************/
1455
1456             /* REACTION-FIELD ELECTROSTATICS */
1457             felec            = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_mul_pd(rinv23,rinvsq23),krf2));
1458
1459             fscal            = felec;
1460
1461             /* Calculate temporary vectorial force */
1462             tx               = _mm256_mul_pd(fscal,dx23);
1463             ty               = _mm256_mul_pd(fscal,dy23);
1464             tz               = _mm256_mul_pd(fscal,dz23);
1465
1466             /* Update vectorial force */
1467             fix2             = _mm256_add_pd(fix2,tx);
1468             fiy2             = _mm256_add_pd(fiy2,ty);
1469             fiz2             = _mm256_add_pd(fiz2,tz);
1470
1471             fjx3             = _mm256_add_pd(fjx3,tx);
1472             fjy3             = _mm256_add_pd(fjy3,ty);
1473             fjz3             = _mm256_add_pd(fjz3,tz);
1474
1475             /**************************
1476              * CALCULATE INTERACTIONS *
1477              **************************/
1478
1479             /* REACTION-FIELD ELECTROSTATICS */
1480             felec            = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_mul_pd(rinv31,rinvsq31),krf2));
1481
1482             fscal            = felec;
1483
1484             /* Calculate temporary vectorial force */
1485             tx               = _mm256_mul_pd(fscal,dx31);
1486             ty               = _mm256_mul_pd(fscal,dy31);
1487             tz               = _mm256_mul_pd(fscal,dz31);
1488
1489             /* Update vectorial force */
1490             fix3             = _mm256_add_pd(fix3,tx);
1491             fiy3             = _mm256_add_pd(fiy3,ty);
1492             fiz3             = _mm256_add_pd(fiz3,tz);
1493
1494             fjx1             = _mm256_add_pd(fjx1,tx);
1495             fjy1             = _mm256_add_pd(fjy1,ty);
1496             fjz1             = _mm256_add_pd(fjz1,tz);
1497
1498             /**************************
1499              * CALCULATE INTERACTIONS *
1500              **************************/
1501
1502             /* REACTION-FIELD ELECTROSTATICS */
1503             felec            = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_mul_pd(rinv32,rinvsq32),krf2));
1504
1505             fscal            = felec;
1506
1507             /* Calculate temporary vectorial force */
1508             tx               = _mm256_mul_pd(fscal,dx32);
1509             ty               = _mm256_mul_pd(fscal,dy32);
1510             tz               = _mm256_mul_pd(fscal,dz32);
1511
1512             /* Update vectorial force */
1513             fix3             = _mm256_add_pd(fix3,tx);
1514             fiy3             = _mm256_add_pd(fiy3,ty);
1515             fiz3             = _mm256_add_pd(fiz3,tz);
1516
1517             fjx2             = _mm256_add_pd(fjx2,tx);
1518             fjy2             = _mm256_add_pd(fjy2,ty);
1519             fjz2             = _mm256_add_pd(fjz2,tz);
1520
1521             /**************************
1522              * CALCULATE INTERACTIONS *
1523              **************************/
1524
1525             /* REACTION-FIELD ELECTROSTATICS */
1526             felec            = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_mul_pd(rinv33,rinvsq33),krf2));
1527
1528             fscal            = felec;
1529
1530             /* Calculate temporary vectorial force */
1531             tx               = _mm256_mul_pd(fscal,dx33);
1532             ty               = _mm256_mul_pd(fscal,dy33);
1533             tz               = _mm256_mul_pd(fscal,dz33);
1534
1535             /* Update vectorial force */
1536             fix3             = _mm256_add_pd(fix3,tx);
1537             fiy3             = _mm256_add_pd(fiy3,ty);
1538             fiz3             = _mm256_add_pd(fiz3,tz);
1539
1540             fjx3             = _mm256_add_pd(fjx3,tx);
1541             fjy3             = _mm256_add_pd(fjy3,ty);
1542             fjz3             = _mm256_add_pd(fjz3,tz);
1543
1544             fjptrA             = f+j_coord_offsetA;
1545             fjptrB             = f+j_coord_offsetB;
1546             fjptrC             = f+j_coord_offsetC;
1547             fjptrD             = f+j_coord_offsetD;
1548
1549             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1550                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1551                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1552
1553             /* Inner loop uses 273 flops */
1554         }
1555
1556         if(jidx<j_index_end)
1557         {
1558
1559             /* Get j neighbor index, and coordinate index */
1560             jnrlistA         = jjnr[jidx];
1561             jnrlistB         = jjnr[jidx+1];
1562             jnrlistC         = jjnr[jidx+2];
1563             jnrlistD         = jjnr[jidx+3];
1564             /* Sign of each element will be negative for non-real atoms.
1565              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1566              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1567              */
1568             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1569
1570             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1571             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1572             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1573
1574             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1575             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1576             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1577             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1578             j_coord_offsetA  = DIM*jnrA;
1579             j_coord_offsetB  = DIM*jnrB;
1580             j_coord_offsetC  = DIM*jnrC;
1581             j_coord_offsetD  = DIM*jnrD;
1582
1583             /* load j atom coordinates */
1584             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1585                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1586                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1587                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1588
1589             /* Calculate displacement vector */
1590             dx00             = _mm256_sub_pd(ix0,jx0);
1591             dy00             = _mm256_sub_pd(iy0,jy0);
1592             dz00             = _mm256_sub_pd(iz0,jz0);
1593             dx11             = _mm256_sub_pd(ix1,jx1);
1594             dy11             = _mm256_sub_pd(iy1,jy1);
1595             dz11             = _mm256_sub_pd(iz1,jz1);
1596             dx12             = _mm256_sub_pd(ix1,jx2);
1597             dy12             = _mm256_sub_pd(iy1,jy2);
1598             dz12             = _mm256_sub_pd(iz1,jz2);
1599             dx13             = _mm256_sub_pd(ix1,jx3);
1600             dy13             = _mm256_sub_pd(iy1,jy3);
1601             dz13             = _mm256_sub_pd(iz1,jz3);
1602             dx21             = _mm256_sub_pd(ix2,jx1);
1603             dy21             = _mm256_sub_pd(iy2,jy1);
1604             dz21             = _mm256_sub_pd(iz2,jz1);
1605             dx22             = _mm256_sub_pd(ix2,jx2);
1606             dy22             = _mm256_sub_pd(iy2,jy2);
1607             dz22             = _mm256_sub_pd(iz2,jz2);
1608             dx23             = _mm256_sub_pd(ix2,jx3);
1609             dy23             = _mm256_sub_pd(iy2,jy3);
1610             dz23             = _mm256_sub_pd(iz2,jz3);
1611             dx31             = _mm256_sub_pd(ix3,jx1);
1612             dy31             = _mm256_sub_pd(iy3,jy1);
1613             dz31             = _mm256_sub_pd(iz3,jz1);
1614             dx32             = _mm256_sub_pd(ix3,jx2);
1615             dy32             = _mm256_sub_pd(iy3,jy2);
1616             dz32             = _mm256_sub_pd(iz3,jz2);
1617             dx33             = _mm256_sub_pd(ix3,jx3);
1618             dy33             = _mm256_sub_pd(iy3,jy3);
1619             dz33             = _mm256_sub_pd(iz3,jz3);
1620
1621             /* Calculate squared distance and things based on it */
1622             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1623             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1624             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1625             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1626             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1627             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1628             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1629             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1630             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1631             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1632
1633             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1634             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1635             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1636             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1637             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1638             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1639             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1640             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1641             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1642
1643             rinvsq00         = gmx_mm256_inv_pd(rsq00);
1644             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1645             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1646             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1647             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1648             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1649             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1650             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1651             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1652             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1653
1654             fjx0             = _mm256_setzero_pd();
1655             fjy0             = _mm256_setzero_pd();
1656             fjz0             = _mm256_setzero_pd();
1657             fjx1             = _mm256_setzero_pd();
1658             fjy1             = _mm256_setzero_pd();
1659             fjz1             = _mm256_setzero_pd();
1660             fjx2             = _mm256_setzero_pd();
1661             fjy2             = _mm256_setzero_pd();
1662             fjz2             = _mm256_setzero_pd();
1663             fjx3             = _mm256_setzero_pd();
1664             fjy3             = _mm256_setzero_pd();
1665             fjz3             = _mm256_setzero_pd();
1666
1667             /**************************
1668              * CALCULATE INTERACTIONS *
1669              **************************/
1670
1671             /* LENNARD-JONES DISPERSION/REPULSION */
1672
1673             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1674             fvdw             = _mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),c6_00),_mm256_mul_pd(rinvsix,rinvsq00));
1675
1676             fscal            = fvdw;
1677
1678             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1679
1680             /* Calculate temporary vectorial force */
1681             tx               = _mm256_mul_pd(fscal,dx00);
1682             ty               = _mm256_mul_pd(fscal,dy00);
1683             tz               = _mm256_mul_pd(fscal,dz00);
1684
1685             /* Update vectorial force */
1686             fix0             = _mm256_add_pd(fix0,tx);
1687             fiy0             = _mm256_add_pd(fiy0,ty);
1688             fiz0             = _mm256_add_pd(fiz0,tz);
1689
1690             fjx0             = _mm256_add_pd(fjx0,tx);
1691             fjy0             = _mm256_add_pd(fjy0,ty);
1692             fjz0             = _mm256_add_pd(fjz0,tz);
1693
1694             /**************************
1695              * CALCULATE INTERACTIONS *
1696              **************************/
1697
1698             /* REACTION-FIELD ELECTROSTATICS */
1699             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1700
1701             fscal            = felec;
1702
1703             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1704
1705             /* Calculate temporary vectorial force */
1706             tx               = _mm256_mul_pd(fscal,dx11);
1707             ty               = _mm256_mul_pd(fscal,dy11);
1708             tz               = _mm256_mul_pd(fscal,dz11);
1709
1710             /* Update vectorial force */
1711             fix1             = _mm256_add_pd(fix1,tx);
1712             fiy1             = _mm256_add_pd(fiy1,ty);
1713             fiz1             = _mm256_add_pd(fiz1,tz);
1714
1715             fjx1             = _mm256_add_pd(fjx1,tx);
1716             fjy1             = _mm256_add_pd(fjy1,ty);
1717             fjz1             = _mm256_add_pd(fjz1,tz);
1718
1719             /**************************
1720              * CALCULATE INTERACTIONS *
1721              **************************/
1722
1723             /* REACTION-FIELD ELECTROSTATICS */
1724             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1725
1726             fscal            = felec;
1727
1728             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1729
1730             /* Calculate temporary vectorial force */
1731             tx               = _mm256_mul_pd(fscal,dx12);
1732             ty               = _mm256_mul_pd(fscal,dy12);
1733             tz               = _mm256_mul_pd(fscal,dz12);
1734
1735             /* Update vectorial force */
1736             fix1             = _mm256_add_pd(fix1,tx);
1737             fiy1             = _mm256_add_pd(fiy1,ty);
1738             fiz1             = _mm256_add_pd(fiz1,tz);
1739
1740             fjx2             = _mm256_add_pd(fjx2,tx);
1741             fjy2             = _mm256_add_pd(fjy2,ty);
1742             fjz2             = _mm256_add_pd(fjz2,tz);
1743
1744             /**************************
1745              * CALCULATE INTERACTIONS *
1746              **************************/
1747
1748             /* REACTION-FIELD ELECTROSTATICS */
1749             felec            = _mm256_mul_pd(qq13,_mm256_sub_pd(_mm256_mul_pd(rinv13,rinvsq13),krf2));
1750
1751             fscal            = felec;
1752
1753             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1754
1755             /* Calculate temporary vectorial force */
1756             tx               = _mm256_mul_pd(fscal,dx13);
1757             ty               = _mm256_mul_pd(fscal,dy13);
1758             tz               = _mm256_mul_pd(fscal,dz13);
1759
1760             /* Update vectorial force */
1761             fix1             = _mm256_add_pd(fix1,tx);
1762             fiy1             = _mm256_add_pd(fiy1,ty);
1763             fiz1             = _mm256_add_pd(fiz1,tz);
1764
1765             fjx3             = _mm256_add_pd(fjx3,tx);
1766             fjy3             = _mm256_add_pd(fjy3,ty);
1767             fjz3             = _mm256_add_pd(fjz3,tz);
1768
1769             /**************************
1770              * CALCULATE INTERACTIONS *
1771              **************************/
1772
1773             /* REACTION-FIELD ELECTROSTATICS */
1774             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1775
1776             fscal            = felec;
1777
1778             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1779
1780             /* Calculate temporary vectorial force */
1781             tx               = _mm256_mul_pd(fscal,dx21);
1782             ty               = _mm256_mul_pd(fscal,dy21);
1783             tz               = _mm256_mul_pd(fscal,dz21);
1784
1785             /* Update vectorial force */
1786             fix2             = _mm256_add_pd(fix2,tx);
1787             fiy2             = _mm256_add_pd(fiy2,ty);
1788             fiz2             = _mm256_add_pd(fiz2,tz);
1789
1790             fjx1             = _mm256_add_pd(fjx1,tx);
1791             fjy1             = _mm256_add_pd(fjy1,ty);
1792             fjz1             = _mm256_add_pd(fjz1,tz);
1793
1794             /**************************
1795              * CALCULATE INTERACTIONS *
1796              **************************/
1797
1798             /* REACTION-FIELD ELECTROSTATICS */
1799             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1800
1801             fscal            = felec;
1802
1803             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1804
1805             /* Calculate temporary vectorial force */
1806             tx               = _mm256_mul_pd(fscal,dx22);
1807             ty               = _mm256_mul_pd(fscal,dy22);
1808             tz               = _mm256_mul_pd(fscal,dz22);
1809
1810             /* Update vectorial force */
1811             fix2             = _mm256_add_pd(fix2,tx);
1812             fiy2             = _mm256_add_pd(fiy2,ty);
1813             fiz2             = _mm256_add_pd(fiz2,tz);
1814
1815             fjx2             = _mm256_add_pd(fjx2,tx);
1816             fjy2             = _mm256_add_pd(fjy2,ty);
1817             fjz2             = _mm256_add_pd(fjz2,tz);
1818
1819             /**************************
1820              * CALCULATE INTERACTIONS *
1821              **************************/
1822
1823             /* REACTION-FIELD ELECTROSTATICS */
1824             felec            = _mm256_mul_pd(qq23,_mm256_sub_pd(_mm256_mul_pd(rinv23,rinvsq23),krf2));
1825
1826             fscal            = felec;
1827
1828             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1829
1830             /* Calculate temporary vectorial force */
1831             tx               = _mm256_mul_pd(fscal,dx23);
1832             ty               = _mm256_mul_pd(fscal,dy23);
1833             tz               = _mm256_mul_pd(fscal,dz23);
1834
1835             /* Update vectorial force */
1836             fix2             = _mm256_add_pd(fix2,tx);
1837             fiy2             = _mm256_add_pd(fiy2,ty);
1838             fiz2             = _mm256_add_pd(fiz2,tz);
1839
1840             fjx3             = _mm256_add_pd(fjx3,tx);
1841             fjy3             = _mm256_add_pd(fjy3,ty);
1842             fjz3             = _mm256_add_pd(fjz3,tz);
1843
1844             /**************************
1845              * CALCULATE INTERACTIONS *
1846              **************************/
1847
1848             /* REACTION-FIELD ELECTROSTATICS */
1849             felec            = _mm256_mul_pd(qq31,_mm256_sub_pd(_mm256_mul_pd(rinv31,rinvsq31),krf2));
1850
1851             fscal            = felec;
1852
1853             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1854
1855             /* Calculate temporary vectorial force */
1856             tx               = _mm256_mul_pd(fscal,dx31);
1857             ty               = _mm256_mul_pd(fscal,dy31);
1858             tz               = _mm256_mul_pd(fscal,dz31);
1859
1860             /* Update vectorial force */
1861             fix3             = _mm256_add_pd(fix3,tx);
1862             fiy3             = _mm256_add_pd(fiy3,ty);
1863             fiz3             = _mm256_add_pd(fiz3,tz);
1864
1865             fjx1             = _mm256_add_pd(fjx1,tx);
1866             fjy1             = _mm256_add_pd(fjy1,ty);
1867             fjz1             = _mm256_add_pd(fjz1,tz);
1868
1869             /**************************
1870              * CALCULATE INTERACTIONS *
1871              **************************/
1872
1873             /* REACTION-FIELD ELECTROSTATICS */
1874             felec            = _mm256_mul_pd(qq32,_mm256_sub_pd(_mm256_mul_pd(rinv32,rinvsq32),krf2));
1875
1876             fscal            = felec;
1877
1878             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1879
1880             /* Calculate temporary vectorial force */
1881             tx               = _mm256_mul_pd(fscal,dx32);
1882             ty               = _mm256_mul_pd(fscal,dy32);
1883             tz               = _mm256_mul_pd(fscal,dz32);
1884
1885             /* Update vectorial force */
1886             fix3             = _mm256_add_pd(fix3,tx);
1887             fiy3             = _mm256_add_pd(fiy3,ty);
1888             fiz3             = _mm256_add_pd(fiz3,tz);
1889
1890             fjx2             = _mm256_add_pd(fjx2,tx);
1891             fjy2             = _mm256_add_pd(fjy2,ty);
1892             fjz2             = _mm256_add_pd(fjz2,tz);
1893
1894             /**************************
1895              * CALCULATE INTERACTIONS *
1896              **************************/
1897
1898             /* REACTION-FIELD ELECTROSTATICS */
1899             felec            = _mm256_mul_pd(qq33,_mm256_sub_pd(_mm256_mul_pd(rinv33,rinvsq33),krf2));
1900
1901             fscal            = felec;
1902
1903             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1904
1905             /* Calculate temporary vectorial force */
1906             tx               = _mm256_mul_pd(fscal,dx33);
1907             ty               = _mm256_mul_pd(fscal,dy33);
1908             tz               = _mm256_mul_pd(fscal,dz33);
1909
1910             /* Update vectorial force */
1911             fix3             = _mm256_add_pd(fix3,tx);
1912             fiy3             = _mm256_add_pd(fiy3,ty);
1913             fiz3             = _mm256_add_pd(fiz3,tz);
1914
1915             fjx3             = _mm256_add_pd(fjx3,tx);
1916             fjy3             = _mm256_add_pd(fjy3,ty);
1917             fjz3             = _mm256_add_pd(fjz3,tz);
1918
1919             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1920             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1921             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1922             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1923
1924             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1925                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1926                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1927
1928             /* Inner loop uses 273 flops */
1929         }
1930
1931         /* End of innermost loop */
1932
1933         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1934                                                  f+i_coord_offset,fshift+i_shift_offset);
1935
1936         /* Increment number of inner iterations */
1937         inneriter                  += j_index_end - j_index_start;
1938
1939         /* Outer loop uses 24 flops */
1940     }
1941
1942     /* Increment number of outer iterations */
1943     outeriter        += nri;
1944
1945     /* Update outer/inner flops */
1946
1947     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*273);
1948 }