src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwLJ_GeomW4W4_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_single.h"
  48 #include "kernelutil_x86_avx_256_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_single
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrE,jnrF,jnrG,jnrH;
  76     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  77     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  78     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  79     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  84     real             scratch[4*DIM];
  85     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  97     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  99     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
 101     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 102     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 103     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 104     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 105     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 106     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 107     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 108     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 109     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 110     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 111     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 112     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 113     real             *charge;
 114     int              nvdwtype;
 115     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 116     int              *vdwtype;
 117     real             *vdwparam;
 118     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 119     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 120     __m256           dummy_mask,cutoff_mask;
 121     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 122     __m256           one     = _mm256_set1_ps(1.0);
 123     __m256           two     = _mm256_set1_ps(2.0);
 124     x                = xx[0];
 125     f                = ff[0];
 126
 127     nri              = nlist->nri;
 128     iinr             = nlist->iinr;
 129     jindex           = nlist->jindex;
 130     jjnr             = nlist->jjnr;
 131     shiftidx         = nlist->shift;
 132     gid              = nlist->gid;
 133     shiftvec         = fr->shift_vec[0];
 134     fshift           = fr->fshift[0];
 135     facel            = _mm256_set1_ps(fr->epsfac);
 136     charge           = mdatoms->chargeA;
 137     nvdwtype         = fr->ntype;
 138     vdwparam         = fr->nbfp;
 139     vdwtype          = mdatoms->typeA;
 140
 141     /* Setup water-specific parameters */
 142     inr              = nlist->iinr[0];
 143     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 144     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 145     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 146     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 147
 148     jq1              = _mm256_set1_ps(charge[inr+1]);
 149     jq2              = _mm256_set1_ps(charge[inr+2]);
 150     jq3              = _mm256_set1_ps(charge[inr+3]);
 151     vdwjidx0A        = 2*vdwtype[inr+0];
 152     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 153     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 154     qq11             = _mm256_mul_ps(iq1,jq1);
 155     qq12             = _mm256_mul_ps(iq1,jq2);
 156     qq13             = _mm256_mul_ps(iq1,jq3);
 157     qq21             = _mm256_mul_ps(iq2,jq1);
 158     qq22             = _mm256_mul_ps(iq2,jq2);
 159     qq23             = _mm256_mul_ps(iq2,jq3);
 160     qq31             = _mm256_mul_ps(iq3,jq1);
 161     qq32             = _mm256_mul_ps(iq3,jq2);
 162     qq33             = _mm256_mul_ps(iq3,jq3);
 163
 164     /* Avoid stupid compiler warnings */
 165     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 166     j_coord_offsetA = 0;
 167     j_coord_offsetB = 0;
 168     j_coord_offsetC = 0;
 169     j_coord_offsetD = 0;
 170     j_coord_offsetE = 0;
 171     j_coord_offsetF = 0;
 172     j_coord_offsetG = 0;
 173     j_coord_offsetH = 0;
 174
 175     outeriter        = 0;
 176     inneriter        = 0;
 177
 178     for(iidx=0;iidx<4*DIM;iidx++)
 179     {
 180         scratch[iidx] = 0.0;
 181     }
 182
 183     /* Start outer loop over neighborlists */
 184     for(iidx=0; iidx<nri; iidx++)
 185     {
 186         /* Load shift vector for this list */
 187         i_shift_offset   = DIM*shiftidx[iidx];
 188
 189         /* Load limits for loop over neighbors */
 190         j_index_start    = jindex[iidx];
 191         j_index_end      = jindex[iidx+1];
 192
 193         /* Get outer coordinate index */
 194         inr              = iinr[iidx];
 195         i_coord_offset   = DIM*inr;
 196
 197         /* Load i particle coords and add shift vector */
 198         gmx_mm256_load_shift_and_4rvec_broadcast_ps(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_ps();
 202         fiy0             = _mm256_setzero_ps();
 203         fiz0             = _mm256_setzero_ps();
 204         fix1             = _mm256_setzero_ps();
 205         fiy1             = _mm256_setzero_ps();
 206         fiz1             = _mm256_setzero_ps();
 207         fix2             = _mm256_setzero_ps();
 208         fiy2             = _mm256_setzero_ps();
 209         fiz2             = _mm256_setzero_ps();
 210         fix3             = _mm256_setzero_ps();
 211         fiy3             = _mm256_setzero_ps();
 212         fiz3             = _mm256_setzero_ps();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm256_setzero_ps();
 216         vvdwsum          = _mm256_setzero_ps();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 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             jnrE             = jjnr[jidx+4];
 228             jnrF             = jjnr[jidx+5];
 229             jnrG             = jjnr[jidx+6];
 230             jnrH             = jjnr[jidx+7];
 231             j_coord_offsetA  = DIM*jnrA;
 232             j_coord_offsetB  = DIM*jnrB;
 233             j_coord_offsetC  = DIM*jnrC;
 234             j_coord_offsetD  = DIM*jnrD;
 235             j_coord_offsetE  = DIM*jnrE;
 236             j_coord_offsetF  = DIM*jnrF;
 237             j_coord_offsetG  = DIM*jnrG;
 238             j_coord_offsetH  = DIM*jnrH;
 239
 240             /* load j atom coordinates */
 241             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 242                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 243                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 244                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 245                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 246                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 247
 248             /* Calculate displacement vector */
 249             dx00             = _mm256_sub_ps(ix0,jx0);
 250             dy00             = _mm256_sub_ps(iy0,jy0);
 251             dz00             = _mm256_sub_ps(iz0,jz0);
 252             dx11             = _mm256_sub_ps(ix1,jx1);
 253             dy11             = _mm256_sub_ps(iy1,jy1);
 254             dz11             = _mm256_sub_ps(iz1,jz1);
 255             dx12             = _mm256_sub_ps(ix1,jx2);
 256             dy12             = _mm256_sub_ps(iy1,jy2);
 257             dz12             = _mm256_sub_ps(iz1,jz2);
 258             dx13             = _mm256_sub_ps(ix1,jx3);
 259             dy13             = _mm256_sub_ps(iy1,jy3);
 260             dz13             = _mm256_sub_ps(iz1,jz3);
 261             dx21             = _mm256_sub_ps(ix2,jx1);
 262             dy21             = _mm256_sub_ps(iy2,jy1);
 263             dz21             = _mm256_sub_ps(iz2,jz1);
 264             dx22             = _mm256_sub_ps(ix2,jx2);
 265             dy22             = _mm256_sub_ps(iy2,jy2);
 266             dz22             = _mm256_sub_ps(iz2,jz2);
 267             dx23             = _mm256_sub_ps(ix2,jx3);
 268             dy23             = _mm256_sub_ps(iy2,jy3);
 269             dz23             = _mm256_sub_ps(iz2,jz3);
 270             dx31             = _mm256_sub_ps(ix3,jx1);
 271             dy31             = _mm256_sub_ps(iy3,jy1);
 272             dz31             = _mm256_sub_ps(iz3,jz1);
 273             dx32             = _mm256_sub_ps(ix3,jx2);
 274             dy32             = _mm256_sub_ps(iy3,jy2);
 275             dz32             = _mm256_sub_ps(iz3,jz2);
 276             dx33             = _mm256_sub_ps(ix3,jx3);
 277             dy33             = _mm256_sub_ps(iy3,jy3);
 278             dz33             = _mm256_sub_ps(iz3,jz3);
 279
 280             /* Calculate squared distance and things based on it */
 281             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 282             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 283             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 284             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 285             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 286             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 287             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 288             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 289             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 290             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 291
 292             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 293             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 294             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 295             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 296             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 297             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 298             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 299             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 300             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 301
 302             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 303             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 304             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 305             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 306             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 307             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 308             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 309             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 310             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 311             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 312
 313             fjx0             = _mm256_setzero_ps();
 314             fjy0             = _mm256_setzero_ps();
 315             fjz0             = _mm256_setzero_ps();
 316             fjx1             = _mm256_setzero_ps();
 317             fjy1             = _mm256_setzero_ps();
 318             fjz1             = _mm256_setzero_ps();
 319             fjx2             = _mm256_setzero_ps();
 320             fjy2             = _mm256_setzero_ps();
 321             fjz2             = _mm256_setzero_ps();
 322             fjx3             = _mm256_setzero_ps();
 323             fjy3             = _mm256_setzero_ps();
 324             fjz3             = _mm256_setzero_ps();
 325
 326             /**************************
 327              * CALCULATE INTERACTIONS *
 328              **************************/
 329
 330             /* LENNARD-JONES DISPERSION/REPULSION */
 331
 332             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 333             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 334             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 335             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 336             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 337
 338             /* Update potential sum for this i atom from the interaction with this j atom. */
 339             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 340
 341             fscal            = fvdw;
 342
 343             /* Calculate temporary vectorial force */
 344             tx               = _mm256_mul_ps(fscal,dx00);
 345             ty               = _mm256_mul_ps(fscal,dy00);
 346             tz               = _mm256_mul_ps(fscal,dz00);
 347
 348             /* Update vectorial force */
 349             fix0             = _mm256_add_ps(fix0,tx);
 350             fiy0             = _mm256_add_ps(fiy0,ty);
 351             fiz0             = _mm256_add_ps(fiz0,tz);
 352
 353             fjx0             = _mm256_add_ps(fjx0,tx);
 354             fjy0             = _mm256_add_ps(fjy0,ty);
 355             fjz0             = _mm256_add_ps(fjz0,tz);
 356
 357             /**************************
 358              * CALCULATE INTERACTIONS *
 359              **************************/
 360
 361             /* COULOMB ELECTROSTATICS */
 362             velec            = _mm256_mul_ps(qq11,rinv11);
 363             felec            = _mm256_mul_ps(velec,rinvsq11);
 364
 365             /* Update potential sum for this i atom from the interaction with this j atom. */
 366             velecsum         = _mm256_add_ps(velecsum,velec);
 367
 368             fscal            = felec;
 369
 370             /* Calculate temporary vectorial force */
 371             tx               = _mm256_mul_ps(fscal,dx11);
 372             ty               = _mm256_mul_ps(fscal,dy11);
 373             tz               = _mm256_mul_ps(fscal,dz11);
 374
 375             /* Update vectorial force */
 376             fix1             = _mm256_add_ps(fix1,tx);
 377             fiy1             = _mm256_add_ps(fiy1,ty);
 378             fiz1             = _mm256_add_ps(fiz1,tz);
 379
 380             fjx1             = _mm256_add_ps(fjx1,tx);
 381             fjy1             = _mm256_add_ps(fjy1,ty);
 382             fjz1             = _mm256_add_ps(fjz1,tz);
 383
 384             /**************************
 385              * CALCULATE INTERACTIONS *
 386              **************************/
 387
 388             /* COULOMB ELECTROSTATICS */
 389             velec            = _mm256_mul_ps(qq12,rinv12);
 390             felec            = _mm256_mul_ps(velec,rinvsq12);
 391
 392             /* Update potential sum for this i atom from the interaction with this j atom. */
 393             velecsum         = _mm256_add_ps(velecsum,velec);
 394
 395             fscal            = felec;
 396
 397             /* Calculate temporary vectorial force */
 398             tx               = _mm256_mul_ps(fscal,dx12);
 399             ty               = _mm256_mul_ps(fscal,dy12);
 400             tz               = _mm256_mul_ps(fscal,dz12);
 401
 402             /* Update vectorial force */
 403             fix1             = _mm256_add_ps(fix1,tx);
 404             fiy1             = _mm256_add_ps(fiy1,ty);
 405             fiz1             = _mm256_add_ps(fiz1,tz);
 406
 407             fjx2             = _mm256_add_ps(fjx2,tx);
 408             fjy2             = _mm256_add_ps(fjy2,ty);
 409             fjz2             = _mm256_add_ps(fjz2,tz);
 410
 411             /**************************
 412              * CALCULATE INTERACTIONS *
 413              **************************/
 414
 415             /* COULOMB ELECTROSTATICS */
 416             velec            = _mm256_mul_ps(qq13,rinv13);
 417             felec            = _mm256_mul_ps(velec,rinvsq13);
 418
 419             /* Update potential sum for this i atom from the interaction with this j atom. */
 420             velecsum         = _mm256_add_ps(velecsum,velec);
 421
 422             fscal            = felec;
 423
 424             /* Calculate temporary vectorial force */
 425             tx               = _mm256_mul_ps(fscal,dx13);
 426             ty               = _mm256_mul_ps(fscal,dy13);
 427             tz               = _mm256_mul_ps(fscal,dz13);
 428
 429             /* Update vectorial force */
 430             fix1             = _mm256_add_ps(fix1,tx);
 431             fiy1             = _mm256_add_ps(fiy1,ty);
 432             fiz1             = _mm256_add_ps(fiz1,tz);
 433
 434             fjx3             = _mm256_add_ps(fjx3,tx);
 435             fjy3             = _mm256_add_ps(fjy3,ty);
 436             fjz3             = _mm256_add_ps(fjz3,tz);
 437
 438             /**************************
 439              * CALCULATE INTERACTIONS *
 440              **************************/
 441
 442             /* COULOMB ELECTROSTATICS */
 443             velec            = _mm256_mul_ps(qq21,rinv21);
 444             felec            = _mm256_mul_ps(velec,rinvsq21);
 445
 446             /* Update potential sum for this i atom from the interaction with this j atom. */
 447             velecsum         = _mm256_add_ps(velecsum,velec);
 448
 449             fscal            = felec;
 450
 451             /* Calculate temporary vectorial force */
 452             tx               = _mm256_mul_ps(fscal,dx21);
 453             ty               = _mm256_mul_ps(fscal,dy21);
 454             tz               = _mm256_mul_ps(fscal,dz21);
 455
 456             /* Update vectorial force */
 457             fix2             = _mm256_add_ps(fix2,tx);
 458             fiy2             = _mm256_add_ps(fiy2,ty);
 459             fiz2             = _mm256_add_ps(fiz2,tz);
 460
 461             fjx1             = _mm256_add_ps(fjx1,tx);
 462             fjy1             = _mm256_add_ps(fjy1,ty);
 463             fjz1             = _mm256_add_ps(fjz1,tz);
 464
 465             /**************************
 466              * CALCULATE INTERACTIONS *
 467              **************************/
 468
 469             /* COULOMB ELECTROSTATICS */
 470             velec            = _mm256_mul_ps(qq22,rinv22);
 471             felec            = _mm256_mul_ps(velec,rinvsq22);
 472
 473             /* Update potential sum for this i atom from the interaction with this j atom. */
 474             velecsum         = _mm256_add_ps(velecsum,velec);
 475
 476             fscal            = felec;
 477
 478             /* Calculate temporary vectorial force */
 479             tx               = _mm256_mul_ps(fscal,dx22);
 480             ty               = _mm256_mul_ps(fscal,dy22);
 481             tz               = _mm256_mul_ps(fscal,dz22);
 482
 483             /* Update vectorial force */
 484             fix2             = _mm256_add_ps(fix2,tx);
 485             fiy2             = _mm256_add_ps(fiy2,ty);
 486             fiz2             = _mm256_add_ps(fiz2,tz);
 487
 488             fjx2             = _mm256_add_ps(fjx2,tx);
 489             fjy2             = _mm256_add_ps(fjy2,ty);
 490             fjz2             = _mm256_add_ps(fjz2,tz);
 491
 492             /**************************
 493              * CALCULATE INTERACTIONS *
 494              **************************/
 495
 496             /* COULOMB ELECTROSTATICS */
 497             velec            = _mm256_mul_ps(qq23,rinv23);
 498             felec            = _mm256_mul_ps(velec,rinvsq23);
 499
 500             /* Update potential sum for this i atom from the interaction with this j atom. */
 501             velecsum         = _mm256_add_ps(velecsum,velec);
 502
 503             fscal            = felec;
 504
 505             /* Calculate temporary vectorial force */
 506             tx               = _mm256_mul_ps(fscal,dx23);
 507             ty               = _mm256_mul_ps(fscal,dy23);
 508             tz               = _mm256_mul_ps(fscal,dz23);
 509
 510             /* Update vectorial force */
 511             fix2             = _mm256_add_ps(fix2,tx);
 512             fiy2             = _mm256_add_ps(fiy2,ty);
 513             fiz2             = _mm256_add_ps(fiz2,tz);
 514
 515             fjx3             = _mm256_add_ps(fjx3,tx);
 516             fjy3             = _mm256_add_ps(fjy3,ty);
 517             fjz3             = _mm256_add_ps(fjz3,tz);
 518
 519             /**************************
 520              * CALCULATE INTERACTIONS *
 521              **************************/
 522
 523             /* COULOMB ELECTROSTATICS */
 524             velec            = _mm256_mul_ps(qq31,rinv31);
 525             felec            = _mm256_mul_ps(velec,rinvsq31);
 526
 527             /* Update potential sum for this i atom from the interaction with this j atom. */
 528             velecsum         = _mm256_add_ps(velecsum,velec);
 529
 530             fscal            = felec;
 531
 532             /* Calculate temporary vectorial force */
 533             tx               = _mm256_mul_ps(fscal,dx31);
 534             ty               = _mm256_mul_ps(fscal,dy31);
 535             tz               = _mm256_mul_ps(fscal,dz31);
 536
 537             /* Update vectorial force */
 538             fix3             = _mm256_add_ps(fix3,tx);
 539             fiy3             = _mm256_add_ps(fiy3,ty);
 540             fiz3             = _mm256_add_ps(fiz3,tz);
 541
 542             fjx1             = _mm256_add_ps(fjx1,tx);
 543             fjy1             = _mm256_add_ps(fjy1,ty);
 544             fjz1             = _mm256_add_ps(fjz1,tz);
 545
 546             /**************************
 547              * CALCULATE INTERACTIONS *
 548              **************************/
 549
 550             /* COULOMB ELECTROSTATICS */
 551             velec            = _mm256_mul_ps(qq32,rinv32);
 552             felec            = _mm256_mul_ps(velec,rinvsq32);
 553
 554             /* Update potential sum for this i atom from the interaction with this j atom. */
 555             velecsum         = _mm256_add_ps(velecsum,velec);
 556
 557             fscal            = felec;
 558
 559             /* Calculate temporary vectorial force */
 560             tx               = _mm256_mul_ps(fscal,dx32);
 561             ty               = _mm256_mul_ps(fscal,dy32);
 562             tz               = _mm256_mul_ps(fscal,dz32);
 563
 564             /* Update vectorial force */
 565             fix3             = _mm256_add_ps(fix3,tx);
 566             fiy3             = _mm256_add_ps(fiy3,ty);
 567             fiz3             = _mm256_add_ps(fiz3,tz);
 568
 569             fjx2             = _mm256_add_ps(fjx2,tx);
 570             fjy2             = _mm256_add_ps(fjy2,ty);
 571             fjz2             = _mm256_add_ps(fjz2,tz);
 572
 573             /**************************
 574              * CALCULATE INTERACTIONS *
 575              **************************/
 576
 577             /* COULOMB ELECTROSTATICS */
 578             velec            = _mm256_mul_ps(qq33,rinv33);
 579             felec            = _mm256_mul_ps(velec,rinvsq33);
 580
 581             /* Update potential sum for this i atom from the interaction with this j atom. */
 582             velecsum         = _mm256_add_ps(velecsum,velec);
 583
 584             fscal            = felec;
 585
 586             /* Calculate temporary vectorial force */
 587             tx               = _mm256_mul_ps(fscal,dx33);
 588             ty               = _mm256_mul_ps(fscal,dy33);
 589             tz               = _mm256_mul_ps(fscal,dz33);
 590
 591             /* Update vectorial force */
 592             fix3             = _mm256_add_ps(fix3,tx);
 593             fiy3             = _mm256_add_ps(fiy3,ty);
 594             fiz3             = _mm256_add_ps(fiz3,tz);
 595
 596             fjx3             = _mm256_add_ps(fjx3,tx);
 597             fjy3             = _mm256_add_ps(fjy3,ty);
 598             fjz3             = _mm256_add_ps(fjz3,tz);
 599
 600             fjptrA             = f+j_coord_offsetA;
 601             fjptrB             = f+j_coord_offsetB;
 602             fjptrC             = f+j_coord_offsetC;
 603             fjptrD             = f+j_coord_offsetD;
 604             fjptrE             = f+j_coord_offsetE;
 605             fjptrF             = f+j_coord_offsetF;
 606             fjptrG             = f+j_coord_offsetG;
 607             fjptrH             = f+j_coord_offsetH;
 608
 609             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 610                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 611                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 612
 613             /* Inner loop uses 278 flops */
 614         }
 615
 616         if(jidx<j_index_end)
 617         {
 618
 619             /* Get j neighbor index, and coordinate index */
 620             jnrlistA         = jjnr[jidx];
 621             jnrlistB         = jjnr[jidx+1];
 622             jnrlistC         = jjnr[jidx+2];
 623             jnrlistD         = jjnr[jidx+3];
 624             jnrlistE         = jjnr[jidx+4];
 625             jnrlistF         = jjnr[jidx+5];
 626             jnrlistG         = jjnr[jidx+6];
 627             jnrlistH         = jjnr[jidx+7];
 628             /* Sign of each element will be negative for non-real atoms.
 629              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 630              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 631              */
 632             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 633                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 634
 635             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 636             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 637             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 638             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 639             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 640             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 641             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 642             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 643             j_coord_offsetA  = DIM*jnrA;
 644             j_coord_offsetB  = DIM*jnrB;
 645             j_coord_offsetC  = DIM*jnrC;
 646             j_coord_offsetD  = DIM*jnrD;
 647             j_coord_offsetE  = DIM*jnrE;
 648             j_coord_offsetF  = DIM*jnrF;
 649             j_coord_offsetG  = DIM*jnrG;
 650             j_coord_offsetH  = DIM*jnrH;
 651
 652             /* load j atom coordinates */
 653             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 654                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 655                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 656                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 657                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 658                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 659
 660             /* Calculate displacement vector */
 661             dx00             = _mm256_sub_ps(ix0,jx0);
 662             dy00             = _mm256_sub_ps(iy0,jy0);
 663             dz00             = _mm256_sub_ps(iz0,jz0);
 664             dx11             = _mm256_sub_ps(ix1,jx1);
 665             dy11             = _mm256_sub_ps(iy1,jy1);
 666             dz11             = _mm256_sub_ps(iz1,jz1);
 667             dx12             = _mm256_sub_ps(ix1,jx2);
 668             dy12             = _mm256_sub_ps(iy1,jy2);
 669             dz12             = _mm256_sub_ps(iz1,jz2);
 670             dx13             = _mm256_sub_ps(ix1,jx3);
 671             dy13             = _mm256_sub_ps(iy1,jy3);
 672             dz13             = _mm256_sub_ps(iz1,jz3);
 673             dx21             = _mm256_sub_ps(ix2,jx1);
 674             dy21             = _mm256_sub_ps(iy2,jy1);
 675             dz21             = _mm256_sub_ps(iz2,jz1);
 676             dx22             = _mm256_sub_ps(ix2,jx2);
 677             dy22             = _mm256_sub_ps(iy2,jy2);
 678             dz22             = _mm256_sub_ps(iz2,jz2);
 679             dx23             = _mm256_sub_ps(ix2,jx3);
 680             dy23             = _mm256_sub_ps(iy2,jy3);
 681             dz23             = _mm256_sub_ps(iz2,jz3);
 682             dx31             = _mm256_sub_ps(ix3,jx1);
 683             dy31             = _mm256_sub_ps(iy3,jy1);
 684             dz31             = _mm256_sub_ps(iz3,jz1);
 685             dx32             = _mm256_sub_ps(ix3,jx2);
 686             dy32             = _mm256_sub_ps(iy3,jy2);
 687             dz32             = _mm256_sub_ps(iz3,jz2);
 688             dx33             = _mm256_sub_ps(ix3,jx3);
 689             dy33             = _mm256_sub_ps(iy3,jy3);
 690             dz33             = _mm256_sub_ps(iz3,jz3);
 691
 692             /* Calculate squared distance and things based on it */
 693             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 694             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 695             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 696             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 697             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 698             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 699             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 700             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 701             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 702             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 703
 704             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 705             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 706             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 707             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 708             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 709             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 710             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 711             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 712             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 713
 714             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 715             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 716             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 717             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 718             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 719             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 720             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 721             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 722             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 723             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 724
 725             fjx0             = _mm256_setzero_ps();
 726             fjy0             = _mm256_setzero_ps();
 727             fjz0             = _mm256_setzero_ps();
 728             fjx1             = _mm256_setzero_ps();
 729             fjy1             = _mm256_setzero_ps();
 730             fjz1             = _mm256_setzero_ps();
 731             fjx2             = _mm256_setzero_ps();
 732             fjy2             = _mm256_setzero_ps();
 733             fjz2             = _mm256_setzero_ps();
 734             fjx3             = _mm256_setzero_ps();
 735             fjy3             = _mm256_setzero_ps();
 736             fjz3             = _mm256_setzero_ps();
 737
 738             /**************************
 739              * CALCULATE INTERACTIONS *
 740              **************************/
 741
 742             /* LENNARD-JONES DISPERSION/REPULSION */
 743
 744             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 745             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 746             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 747             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 748             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 749
 750             /* Update potential sum for this i atom from the interaction with this j atom. */
 751             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 752             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 753
 754             fscal            = fvdw;
 755
 756             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 757
 758             /* Calculate temporary vectorial force */
 759             tx               = _mm256_mul_ps(fscal,dx00);
 760             ty               = _mm256_mul_ps(fscal,dy00);
 761             tz               = _mm256_mul_ps(fscal,dz00);
 762
 763             /* Update vectorial force */
 764             fix0             = _mm256_add_ps(fix0,tx);
 765             fiy0             = _mm256_add_ps(fiy0,ty);
 766             fiz0             = _mm256_add_ps(fiz0,tz);
 767
 768             fjx0             = _mm256_add_ps(fjx0,tx);
 769             fjy0             = _mm256_add_ps(fjy0,ty);
 770             fjz0             = _mm256_add_ps(fjz0,tz);
 771
 772             /**************************
 773              * CALCULATE INTERACTIONS *
 774              **************************/
 775
 776             /* COULOMB ELECTROSTATICS */
 777             velec            = _mm256_mul_ps(qq11,rinv11);
 778             felec            = _mm256_mul_ps(velec,rinvsq11);
 779
 780             /* Update potential sum for this i atom from the interaction with this j atom. */
 781             velec            = _mm256_andnot_ps(dummy_mask,velec);
 782             velecsum         = _mm256_add_ps(velecsum,velec);
 783
 784             fscal            = felec;
 785
 786             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 787
 788             /* Calculate temporary vectorial force */
 789             tx               = _mm256_mul_ps(fscal,dx11);
 790             ty               = _mm256_mul_ps(fscal,dy11);
 791             tz               = _mm256_mul_ps(fscal,dz11);
 792
 793             /* Update vectorial force */
 794             fix1             = _mm256_add_ps(fix1,tx);
 795             fiy1             = _mm256_add_ps(fiy1,ty);
 796             fiz1             = _mm256_add_ps(fiz1,tz);
 797
 798             fjx1             = _mm256_add_ps(fjx1,tx);
 799             fjy1             = _mm256_add_ps(fjy1,ty);
 800             fjz1             = _mm256_add_ps(fjz1,tz);
 801
 802             /**************************
 803              * CALCULATE INTERACTIONS *
 804              **************************/
 805
 806             /* COULOMB ELECTROSTATICS */
 807             velec            = _mm256_mul_ps(qq12,rinv12);
 808             felec            = _mm256_mul_ps(velec,rinvsq12);
 809
 810             /* Update potential sum for this i atom from the interaction with this j atom. */
 811             velec            = _mm256_andnot_ps(dummy_mask,velec);
 812             velecsum         = _mm256_add_ps(velecsum,velec);
 813
 814             fscal            = felec;
 815
 816             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 817
 818             /* Calculate temporary vectorial force */
 819             tx               = _mm256_mul_ps(fscal,dx12);
 820             ty               = _mm256_mul_ps(fscal,dy12);
 821             tz               = _mm256_mul_ps(fscal,dz12);
 822
 823             /* Update vectorial force */
 824             fix1             = _mm256_add_ps(fix1,tx);
 825             fiy1             = _mm256_add_ps(fiy1,ty);
 826             fiz1             = _mm256_add_ps(fiz1,tz);
 827
 828             fjx2             = _mm256_add_ps(fjx2,tx);
 829             fjy2             = _mm256_add_ps(fjy2,ty);
 830             fjz2             = _mm256_add_ps(fjz2,tz);
 831
 832             /**************************
 833              * CALCULATE INTERACTIONS *
 834              **************************/
 835
 836             /* COULOMB ELECTROSTATICS */
 837             velec            = _mm256_mul_ps(qq13,rinv13);
 838             felec            = _mm256_mul_ps(velec,rinvsq13);
 839
 840             /* Update potential sum for this i atom from the interaction with this j atom. */
 841             velec            = _mm256_andnot_ps(dummy_mask,velec);
 842             velecsum         = _mm256_add_ps(velecsum,velec);
 843
 844             fscal            = felec;
 845
 846             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 847
 848             /* Calculate temporary vectorial force */
 849             tx               = _mm256_mul_ps(fscal,dx13);
 850             ty               = _mm256_mul_ps(fscal,dy13);
 851             tz               = _mm256_mul_ps(fscal,dz13);
 852
 853             /* Update vectorial force */
 854             fix1             = _mm256_add_ps(fix1,tx);
 855             fiy1             = _mm256_add_ps(fiy1,ty);
 856             fiz1             = _mm256_add_ps(fiz1,tz);
 857
 858             fjx3             = _mm256_add_ps(fjx3,tx);
 859             fjy3             = _mm256_add_ps(fjy3,ty);
 860             fjz3             = _mm256_add_ps(fjz3,tz);
 861
 862             /**************************
 863              * CALCULATE INTERACTIONS *
 864              **************************/
 865
 866             /* COULOMB ELECTROSTATICS */
 867             velec            = _mm256_mul_ps(qq21,rinv21);
 868             felec            = _mm256_mul_ps(velec,rinvsq21);
 869
 870             /* Update potential sum for this i atom from the interaction with this j atom. */
 871             velec            = _mm256_andnot_ps(dummy_mask,velec);
 872             velecsum         = _mm256_add_ps(velecsum,velec);
 873
 874             fscal            = felec;
 875
 876             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 877
 878             /* Calculate temporary vectorial force */
 879             tx               = _mm256_mul_ps(fscal,dx21);
 880             ty               = _mm256_mul_ps(fscal,dy21);
 881             tz               = _mm256_mul_ps(fscal,dz21);
 882
 883             /* Update vectorial force */
 884             fix2             = _mm256_add_ps(fix2,tx);
 885             fiy2             = _mm256_add_ps(fiy2,ty);
 886             fiz2             = _mm256_add_ps(fiz2,tz);
 887
 888             fjx1             = _mm256_add_ps(fjx1,tx);
 889             fjy1             = _mm256_add_ps(fjy1,ty);
 890             fjz1             = _mm256_add_ps(fjz1,tz);
 891
 892             /**************************
 893              * CALCULATE INTERACTIONS *
 894              **************************/
 895
 896             /* COULOMB ELECTROSTATICS */
 897             velec            = _mm256_mul_ps(qq22,rinv22);
 898             felec            = _mm256_mul_ps(velec,rinvsq22);
 899
 900             /* Update potential sum for this i atom from the interaction with this j atom. */
 901             velec            = _mm256_andnot_ps(dummy_mask,velec);
 902             velecsum         = _mm256_add_ps(velecsum,velec);
 903
 904             fscal            = felec;
 905
 906             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 907
 908             /* Calculate temporary vectorial force */
 909             tx               = _mm256_mul_ps(fscal,dx22);
 910             ty               = _mm256_mul_ps(fscal,dy22);
 911             tz               = _mm256_mul_ps(fscal,dz22);
 912
 913             /* Update vectorial force */
 914             fix2             = _mm256_add_ps(fix2,tx);
 915             fiy2             = _mm256_add_ps(fiy2,ty);
 916             fiz2             = _mm256_add_ps(fiz2,tz);
 917
 918             fjx2             = _mm256_add_ps(fjx2,tx);
 919             fjy2             = _mm256_add_ps(fjy2,ty);
 920             fjz2             = _mm256_add_ps(fjz2,tz);
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             /* COULOMB ELECTROSTATICS */
 927             velec            = _mm256_mul_ps(qq23,rinv23);
 928             felec            = _mm256_mul_ps(velec,rinvsq23);
 929
 930             /* Update potential sum for this i atom from the interaction with this j atom. */
 931             velec            = _mm256_andnot_ps(dummy_mask,velec);
 932             velecsum         = _mm256_add_ps(velecsum,velec);
 933
 934             fscal            = felec;
 935
 936             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 937
 938             /* Calculate temporary vectorial force */
 939             tx               = _mm256_mul_ps(fscal,dx23);
 940             ty               = _mm256_mul_ps(fscal,dy23);
 941             tz               = _mm256_mul_ps(fscal,dz23);
 942
 943             /* Update vectorial force */
 944             fix2             = _mm256_add_ps(fix2,tx);
 945             fiy2             = _mm256_add_ps(fiy2,ty);
 946             fiz2             = _mm256_add_ps(fiz2,tz);
 947
 948             fjx3             = _mm256_add_ps(fjx3,tx);
 949             fjy3             = _mm256_add_ps(fjy3,ty);
 950             fjz3             = _mm256_add_ps(fjz3,tz);
 951
 952             /**************************
 953              * CALCULATE INTERACTIONS *
 954              **************************/
 955
 956             /* COULOMB ELECTROSTATICS */
 957             velec            = _mm256_mul_ps(qq31,rinv31);
 958             felec            = _mm256_mul_ps(velec,rinvsq31);
 959
 960             /* Update potential sum for this i atom from the interaction with this j atom. */
 961             velec            = _mm256_andnot_ps(dummy_mask,velec);
 962             velecsum         = _mm256_add_ps(velecsum,velec);
 963
 964             fscal            = felec;
 965
 966             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 967
 968             /* Calculate temporary vectorial force */
 969             tx               = _mm256_mul_ps(fscal,dx31);
 970             ty               = _mm256_mul_ps(fscal,dy31);
 971             tz               = _mm256_mul_ps(fscal,dz31);
 972
 973             /* Update vectorial force */
 974             fix3             = _mm256_add_ps(fix3,tx);
 975             fiy3             = _mm256_add_ps(fiy3,ty);
 976             fiz3             = _mm256_add_ps(fiz3,tz);
 977
 978             fjx1             = _mm256_add_ps(fjx1,tx);
 979             fjy1             = _mm256_add_ps(fjy1,ty);
 980             fjz1             = _mm256_add_ps(fjz1,tz);
 981
 982             /**************************
 983              * CALCULATE INTERACTIONS *
 984              **************************/
 985
 986             /* COULOMB ELECTROSTATICS */
 987             velec            = _mm256_mul_ps(qq32,rinv32);
 988             felec            = _mm256_mul_ps(velec,rinvsq32);
 989
 990             /* Update potential sum for this i atom from the interaction with this j atom. */
 991             velec            = _mm256_andnot_ps(dummy_mask,velec);
 992             velecsum         = _mm256_add_ps(velecsum,velec);
 993
 994             fscal            = felec;
 995
 996             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 997
 998             /* Calculate temporary vectorial force */
 999             tx               = _mm256_mul_ps(fscal,dx32);
1000             ty               = _mm256_mul_ps(fscal,dy32);
1001             tz               = _mm256_mul_ps(fscal,dz32);
1002
1003             /* Update vectorial force */
1004             fix3             = _mm256_add_ps(fix3,tx);
1005             fiy3             = _mm256_add_ps(fiy3,ty);
1006             fiz3             = _mm256_add_ps(fiz3,tz);
1007
1008             fjx2             = _mm256_add_ps(fjx2,tx);
1009             fjy2             = _mm256_add_ps(fjy2,ty);
1010             fjz2             = _mm256_add_ps(fjz2,tz);
1011
1012             /**************************
1013              * CALCULATE INTERACTIONS *
1014              **************************/
1015
1016             /* COULOMB ELECTROSTATICS */
1017             velec            = _mm256_mul_ps(qq33,rinv33);
1018             felec            = _mm256_mul_ps(velec,rinvsq33);
1019
1020             /* Update potential sum for this i atom from the interaction with this j atom. */
1021             velec            = _mm256_andnot_ps(dummy_mask,velec);
1022             velecsum         = _mm256_add_ps(velecsum,velec);
1023
1024             fscal            = felec;
1025
1026             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1027
1028             /* Calculate temporary vectorial force */
1029             tx               = _mm256_mul_ps(fscal,dx33);
1030             ty               = _mm256_mul_ps(fscal,dy33);
1031             tz               = _mm256_mul_ps(fscal,dz33);
1032
1033             /* Update vectorial force */
1034             fix3             = _mm256_add_ps(fix3,tx);
1035             fiy3             = _mm256_add_ps(fiy3,ty);
1036             fiz3             = _mm256_add_ps(fiz3,tz);
1037
1038             fjx3             = _mm256_add_ps(fjx3,tx);
1039             fjy3             = _mm256_add_ps(fjy3,ty);
1040             fjz3             = _mm256_add_ps(fjz3,tz);
1041
1042             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1043             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1044             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1045             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1046             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1047             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1048             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1049             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1050
1051             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1052                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1053                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1054
1055             /* Inner loop uses 278 flops */
1056         }
1057
1058         /* End of innermost loop */
1059
1060         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1061                                                  f+i_coord_offset,fshift+i_shift_offset);
1062
1063         ggid                        = gid[iidx];
1064         /* Update potential energies */
1065         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1066         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1067
1068         /* Increment number of inner iterations */
1069         inneriter                  += j_index_end - j_index_start;
1070
1071         /* Outer loop uses 26 flops */
1072     }
1073
1074     /* Increment number of outer iterations */
1075     outeriter        += nri;
1076
1077     /* Update outer/inner flops */
1078
1079     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*278);
1080 }
1081 /*
1082  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_single
1083  * Electrostatics interaction: Coulomb
1084  * VdW interaction:            LennardJones
1085  * Geometry:                   Water4-Water4
1086  * Calculate force/pot:        Force
1087  */
1088 void
1089 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_single
1090                     (t_nblist                    * gmx_restrict       nlist,
1091                      rvec                        * gmx_restrict          xx,
1092                      rvec                        * gmx_restrict          ff,
1093                      t_forcerec                  * gmx_restrict          fr,
1094                      t_mdatoms                   * gmx_restrict     mdatoms,
1095                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1096                      t_nrnb                      * gmx_restrict        nrnb)
1097 {
1098     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1099      * just 0 for non-waters.
1100      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1101      * jnr indices corresponding to data put in the four positions in the SIMD register.
1102      */
1103     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1104     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1105     int              jnrA,jnrB,jnrC,jnrD;
1106     int              jnrE,jnrF,jnrG,jnrH;
1107     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1108     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1109     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1110     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1111     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1112     real             rcutoff_scalar;
1113     real             *shiftvec,*fshift,*x,*f;
1114     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1115     real             scratch[4*DIM];
1116     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1117     real *           vdwioffsetptr0;
1118     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1119     real *           vdwioffsetptr1;
1120     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1121     real *           vdwioffsetptr2;
1122     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1123     real *           vdwioffsetptr3;
1124     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1125     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1126     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1127     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1128     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1129     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1130     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1131     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1132     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1133     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1134     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1135     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1136     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1137     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1138     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1139     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1140     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1141     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1142     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1143     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1144     real             *charge;
1145     int              nvdwtype;
1146     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1147     int              *vdwtype;
1148     real             *vdwparam;
1149     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1150     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1151     __m256           dummy_mask,cutoff_mask;
1152     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1153     __m256           one     = _mm256_set1_ps(1.0);
1154     __m256           two     = _mm256_set1_ps(2.0);
1155     x                = xx[0];
1156     f                = ff[0];
1157
1158     nri              = nlist->nri;
1159     iinr             = nlist->iinr;
1160     jindex           = nlist->jindex;
1161     jjnr             = nlist->jjnr;
1162     shiftidx         = nlist->shift;
1163     gid              = nlist->gid;
1164     shiftvec         = fr->shift_vec[0];
1165     fshift           = fr->fshift[0];
1166     facel            = _mm256_set1_ps(fr->epsfac);
1167     charge           = mdatoms->chargeA;
1168     nvdwtype         = fr->ntype;
1169     vdwparam         = fr->nbfp;
1170     vdwtype          = mdatoms->typeA;
1171
1172     /* Setup water-specific parameters */
1173     inr              = nlist->iinr[0];
1174     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1175     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1176     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1177     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1178
1179     jq1              = _mm256_set1_ps(charge[inr+1]);
1180     jq2              = _mm256_set1_ps(charge[inr+2]);
1181     jq3              = _mm256_set1_ps(charge[inr+3]);
1182     vdwjidx0A        = 2*vdwtype[inr+0];
1183     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1184     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1185     qq11             = _mm256_mul_ps(iq1,jq1);
1186     qq12             = _mm256_mul_ps(iq1,jq2);
1187     qq13             = _mm256_mul_ps(iq1,jq3);
1188     qq21             = _mm256_mul_ps(iq2,jq1);
1189     qq22             = _mm256_mul_ps(iq2,jq2);
1190     qq23             = _mm256_mul_ps(iq2,jq3);
1191     qq31             = _mm256_mul_ps(iq3,jq1);
1192     qq32             = _mm256_mul_ps(iq3,jq2);
1193     qq33             = _mm256_mul_ps(iq3,jq3);
1194
1195     /* Avoid stupid compiler warnings */
1196     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1197     j_coord_offsetA = 0;
1198     j_coord_offsetB = 0;
1199     j_coord_offsetC = 0;
1200     j_coord_offsetD = 0;
1201     j_coord_offsetE = 0;
1202     j_coord_offsetF = 0;
1203     j_coord_offsetG = 0;
1204     j_coord_offsetH = 0;
1205
1206     outeriter        = 0;
1207     inneriter        = 0;
1208
1209     for(iidx=0;iidx<4*DIM;iidx++)
1210     {
1211         scratch[iidx] = 0.0;
1212     }
1213
1214     /* Start outer loop over neighborlists */
1215     for(iidx=0; iidx<nri; iidx++)
1216     {
1217         /* Load shift vector for this list */
1218         i_shift_offset   = DIM*shiftidx[iidx];
1219
1220         /* Load limits for loop over neighbors */
1221         j_index_start    = jindex[iidx];
1222         j_index_end      = jindex[iidx+1];
1223
1224         /* Get outer coordinate index */
1225         inr              = iinr[iidx];
1226         i_coord_offset   = DIM*inr;
1227
1228         /* Load i particle coords and add shift vector */
1229         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1230                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1231
1232         fix0             = _mm256_setzero_ps();
1233         fiy0             = _mm256_setzero_ps();
1234         fiz0             = _mm256_setzero_ps();
1235         fix1             = _mm256_setzero_ps();
1236         fiy1             = _mm256_setzero_ps();
1237         fiz1             = _mm256_setzero_ps();
1238         fix2             = _mm256_setzero_ps();
1239         fiy2             = _mm256_setzero_ps();
1240         fiz2             = _mm256_setzero_ps();
1241         fix3             = _mm256_setzero_ps();
1242         fiy3             = _mm256_setzero_ps();
1243         fiz3             = _mm256_setzero_ps();
1244
1245         /* Start inner kernel loop */
1246         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1247         {
1248
1249             /* Get j neighbor index, and coordinate index */
1250             jnrA             = jjnr[jidx];
1251             jnrB             = jjnr[jidx+1];
1252             jnrC             = jjnr[jidx+2];
1253             jnrD             = jjnr[jidx+3];
1254             jnrE             = jjnr[jidx+4];
1255             jnrF             = jjnr[jidx+5];
1256             jnrG             = jjnr[jidx+6];
1257             jnrH             = jjnr[jidx+7];
1258             j_coord_offsetA  = DIM*jnrA;
1259             j_coord_offsetB  = DIM*jnrB;
1260             j_coord_offsetC  = DIM*jnrC;
1261             j_coord_offsetD  = DIM*jnrD;
1262             j_coord_offsetE  = DIM*jnrE;
1263             j_coord_offsetF  = DIM*jnrF;
1264             j_coord_offsetG  = DIM*jnrG;
1265             j_coord_offsetH  = DIM*jnrH;
1266
1267             /* load j atom coordinates */
1268             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1269                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1270                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1271                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1272                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1273                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1274
1275             /* Calculate displacement vector */
1276             dx00             = _mm256_sub_ps(ix0,jx0);
1277             dy00             = _mm256_sub_ps(iy0,jy0);
1278             dz00             = _mm256_sub_ps(iz0,jz0);
1279             dx11             = _mm256_sub_ps(ix1,jx1);
1280             dy11             = _mm256_sub_ps(iy1,jy1);
1281             dz11             = _mm256_sub_ps(iz1,jz1);
1282             dx12             = _mm256_sub_ps(ix1,jx2);
1283             dy12             = _mm256_sub_ps(iy1,jy2);
1284             dz12             = _mm256_sub_ps(iz1,jz2);
1285             dx13             = _mm256_sub_ps(ix1,jx3);
1286             dy13             = _mm256_sub_ps(iy1,jy3);
1287             dz13             = _mm256_sub_ps(iz1,jz3);
1288             dx21             = _mm256_sub_ps(ix2,jx1);
1289             dy21             = _mm256_sub_ps(iy2,jy1);
1290             dz21             = _mm256_sub_ps(iz2,jz1);
1291             dx22             = _mm256_sub_ps(ix2,jx2);
1292             dy22             = _mm256_sub_ps(iy2,jy2);
1293             dz22             = _mm256_sub_ps(iz2,jz2);
1294             dx23             = _mm256_sub_ps(ix2,jx3);
1295             dy23             = _mm256_sub_ps(iy2,jy3);
1296             dz23             = _mm256_sub_ps(iz2,jz3);
1297             dx31             = _mm256_sub_ps(ix3,jx1);
1298             dy31             = _mm256_sub_ps(iy3,jy1);
1299             dz31             = _mm256_sub_ps(iz3,jz1);
1300             dx32             = _mm256_sub_ps(ix3,jx2);
1301             dy32             = _mm256_sub_ps(iy3,jy2);
1302             dz32             = _mm256_sub_ps(iz3,jz2);
1303             dx33             = _mm256_sub_ps(ix3,jx3);
1304             dy33             = _mm256_sub_ps(iy3,jy3);
1305             dz33             = _mm256_sub_ps(iz3,jz3);
1306
1307             /* Calculate squared distance and things based on it */
1308             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1309             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1310             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1311             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1312             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1313             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1314             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1315             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1316             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1317             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1318
1319             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1320             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1321             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1322             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1323             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1324             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1325             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1326             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1327             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1328
1329             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1330             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1331             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1332             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1333             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1334             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1335             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1336             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1337             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1338             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1339
1340             fjx0             = _mm256_setzero_ps();
1341             fjy0             = _mm256_setzero_ps();
1342             fjz0             = _mm256_setzero_ps();
1343             fjx1             = _mm256_setzero_ps();
1344             fjy1             = _mm256_setzero_ps();
1345             fjz1             = _mm256_setzero_ps();
1346             fjx2             = _mm256_setzero_ps();
1347             fjy2             = _mm256_setzero_ps();
1348             fjz2             = _mm256_setzero_ps();
1349             fjx3             = _mm256_setzero_ps();
1350             fjy3             = _mm256_setzero_ps();
1351             fjz3             = _mm256_setzero_ps();
1352
1353             /**************************
1354              * CALCULATE INTERACTIONS *
1355              **************************/
1356
1357             /* LENNARD-JONES DISPERSION/REPULSION */
1358
1359             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1360             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1361
1362             fscal            = fvdw;
1363
1364             /* Calculate temporary vectorial force */
1365             tx               = _mm256_mul_ps(fscal,dx00);
1366             ty               = _mm256_mul_ps(fscal,dy00);
1367             tz               = _mm256_mul_ps(fscal,dz00);
1368
1369             /* Update vectorial force */
1370             fix0             = _mm256_add_ps(fix0,tx);
1371             fiy0             = _mm256_add_ps(fiy0,ty);
1372             fiz0             = _mm256_add_ps(fiz0,tz);
1373
1374             fjx0             = _mm256_add_ps(fjx0,tx);
1375             fjy0             = _mm256_add_ps(fjy0,ty);
1376             fjz0             = _mm256_add_ps(fjz0,tz);
1377
1378             /**************************
1379              * CALCULATE INTERACTIONS *
1380              **************************/
1381
1382             /* COULOMB ELECTROSTATICS */
1383             velec            = _mm256_mul_ps(qq11,rinv11);
1384             felec            = _mm256_mul_ps(velec,rinvsq11);
1385
1386             fscal            = felec;
1387
1388             /* Calculate temporary vectorial force */
1389             tx               = _mm256_mul_ps(fscal,dx11);
1390             ty               = _mm256_mul_ps(fscal,dy11);
1391             tz               = _mm256_mul_ps(fscal,dz11);
1392
1393             /* Update vectorial force */
1394             fix1             = _mm256_add_ps(fix1,tx);
1395             fiy1             = _mm256_add_ps(fiy1,ty);
1396             fiz1             = _mm256_add_ps(fiz1,tz);
1397
1398             fjx1             = _mm256_add_ps(fjx1,tx);
1399             fjy1             = _mm256_add_ps(fjy1,ty);
1400             fjz1             = _mm256_add_ps(fjz1,tz);
1401
1402             /**************************
1403              * CALCULATE INTERACTIONS *
1404              **************************/
1405
1406             /* COULOMB ELECTROSTATICS */
1407             velec            = _mm256_mul_ps(qq12,rinv12);
1408             felec            = _mm256_mul_ps(velec,rinvsq12);
1409
1410             fscal            = felec;
1411
1412             /* Calculate temporary vectorial force */
1413             tx               = _mm256_mul_ps(fscal,dx12);
1414             ty               = _mm256_mul_ps(fscal,dy12);
1415             tz               = _mm256_mul_ps(fscal,dz12);
1416
1417             /* Update vectorial force */
1418             fix1             = _mm256_add_ps(fix1,tx);
1419             fiy1             = _mm256_add_ps(fiy1,ty);
1420             fiz1             = _mm256_add_ps(fiz1,tz);
1421
1422             fjx2             = _mm256_add_ps(fjx2,tx);
1423             fjy2             = _mm256_add_ps(fjy2,ty);
1424             fjz2             = _mm256_add_ps(fjz2,tz);
1425
1426             /**************************
1427              * CALCULATE INTERACTIONS *
1428              **************************/
1429
1430             /* COULOMB ELECTROSTATICS */
1431             velec            = _mm256_mul_ps(qq13,rinv13);
1432             felec            = _mm256_mul_ps(velec,rinvsq13);
1433
1434             fscal            = felec;
1435
1436             /* Calculate temporary vectorial force */
1437             tx               = _mm256_mul_ps(fscal,dx13);
1438             ty               = _mm256_mul_ps(fscal,dy13);
1439             tz               = _mm256_mul_ps(fscal,dz13);
1440
1441             /* Update vectorial force */
1442             fix1             = _mm256_add_ps(fix1,tx);
1443             fiy1             = _mm256_add_ps(fiy1,ty);
1444             fiz1             = _mm256_add_ps(fiz1,tz);
1445
1446             fjx3             = _mm256_add_ps(fjx3,tx);
1447             fjy3             = _mm256_add_ps(fjy3,ty);
1448             fjz3             = _mm256_add_ps(fjz3,tz);
1449
1450             /**************************
1451              * CALCULATE INTERACTIONS *
1452              **************************/
1453
1454             /* COULOMB ELECTROSTATICS */
1455             velec            = _mm256_mul_ps(qq21,rinv21);
1456             felec            = _mm256_mul_ps(velec,rinvsq21);
1457
1458             fscal            = felec;
1459
1460             /* Calculate temporary vectorial force */
1461             tx               = _mm256_mul_ps(fscal,dx21);
1462             ty               = _mm256_mul_ps(fscal,dy21);
1463             tz               = _mm256_mul_ps(fscal,dz21);
1464
1465             /* Update vectorial force */
1466             fix2             = _mm256_add_ps(fix2,tx);
1467             fiy2             = _mm256_add_ps(fiy2,ty);
1468             fiz2             = _mm256_add_ps(fiz2,tz);
1469
1470             fjx1             = _mm256_add_ps(fjx1,tx);
1471             fjy1             = _mm256_add_ps(fjy1,ty);
1472             fjz1             = _mm256_add_ps(fjz1,tz);
1473
1474             /**************************
1475              * CALCULATE INTERACTIONS *
1476              **************************/
1477
1478             /* COULOMB ELECTROSTATICS */
1479             velec            = _mm256_mul_ps(qq22,rinv22);
1480             felec            = _mm256_mul_ps(velec,rinvsq22);
1481
1482             fscal            = felec;
1483
1484             /* Calculate temporary vectorial force */
1485             tx               = _mm256_mul_ps(fscal,dx22);
1486             ty               = _mm256_mul_ps(fscal,dy22);
1487             tz               = _mm256_mul_ps(fscal,dz22);
1488
1489             /* Update vectorial force */
1490             fix2             = _mm256_add_ps(fix2,tx);
1491             fiy2             = _mm256_add_ps(fiy2,ty);
1492             fiz2             = _mm256_add_ps(fiz2,tz);
1493
1494             fjx2             = _mm256_add_ps(fjx2,tx);
1495             fjy2             = _mm256_add_ps(fjy2,ty);
1496             fjz2             = _mm256_add_ps(fjz2,tz);
1497
1498             /**************************
1499              * CALCULATE INTERACTIONS *
1500              **************************/
1501
1502             /* COULOMB ELECTROSTATICS */
1503             velec            = _mm256_mul_ps(qq23,rinv23);
1504             felec            = _mm256_mul_ps(velec,rinvsq23);
1505
1506             fscal            = felec;
1507
1508             /* Calculate temporary vectorial force */
1509             tx               = _mm256_mul_ps(fscal,dx23);
1510             ty               = _mm256_mul_ps(fscal,dy23);
1511             tz               = _mm256_mul_ps(fscal,dz23);
1512
1513             /* Update vectorial force */
1514             fix2             = _mm256_add_ps(fix2,tx);
1515             fiy2             = _mm256_add_ps(fiy2,ty);
1516             fiz2             = _mm256_add_ps(fiz2,tz);
1517
1518             fjx3             = _mm256_add_ps(fjx3,tx);
1519             fjy3             = _mm256_add_ps(fjy3,ty);
1520             fjz3             = _mm256_add_ps(fjz3,tz);
1521
1522             /**************************
1523              * CALCULATE INTERACTIONS *
1524              **************************/
1525
1526             /* COULOMB ELECTROSTATICS */
1527             velec            = _mm256_mul_ps(qq31,rinv31);
1528             felec            = _mm256_mul_ps(velec,rinvsq31);
1529
1530             fscal            = felec;
1531
1532             /* Calculate temporary vectorial force */
1533             tx               = _mm256_mul_ps(fscal,dx31);
1534             ty               = _mm256_mul_ps(fscal,dy31);
1535             tz               = _mm256_mul_ps(fscal,dz31);
1536
1537             /* Update vectorial force */
1538             fix3             = _mm256_add_ps(fix3,tx);
1539             fiy3             = _mm256_add_ps(fiy3,ty);
1540             fiz3             = _mm256_add_ps(fiz3,tz);
1541
1542             fjx1             = _mm256_add_ps(fjx1,tx);
1543             fjy1             = _mm256_add_ps(fjy1,ty);
1544             fjz1             = _mm256_add_ps(fjz1,tz);
1545
1546             /**************************
1547              * CALCULATE INTERACTIONS *
1548              **************************/
1549
1550             /* COULOMB ELECTROSTATICS */
1551             velec            = _mm256_mul_ps(qq32,rinv32);
1552             felec            = _mm256_mul_ps(velec,rinvsq32);
1553
1554             fscal            = felec;
1555
1556             /* Calculate temporary vectorial force */
1557             tx               = _mm256_mul_ps(fscal,dx32);
1558             ty               = _mm256_mul_ps(fscal,dy32);
1559             tz               = _mm256_mul_ps(fscal,dz32);
1560
1561             /* Update vectorial force */
1562             fix3             = _mm256_add_ps(fix3,tx);
1563             fiy3             = _mm256_add_ps(fiy3,ty);
1564             fiz3             = _mm256_add_ps(fiz3,tz);
1565
1566             fjx2             = _mm256_add_ps(fjx2,tx);
1567             fjy2             = _mm256_add_ps(fjy2,ty);
1568             fjz2             = _mm256_add_ps(fjz2,tz);
1569
1570             /**************************
1571              * CALCULATE INTERACTIONS *
1572              **************************/
1573
1574             /* COULOMB ELECTROSTATICS */
1575             velec            = _mm256_mul_ps(qq33,rinv33);
1576             felec            = _mm256_mul_ps(velec,rinvsq33);
1577
1578             fscal            = felec;
1579
1580             /* Calculate temporary vectorial force */
1581             tx               = _mm256_mul_ps(fscal,dx33);
1582             ty               = _mm256_mul_ps(fscal,dy33);
1583             tz               = _mm256_mul_ps(fscal,dz33);
1584
1585             /* Update vectorial force */
1586             fix3             = _mm256_add_ps(fix3,tx);
1587             fiy3             = _mm256_add_ps(fiy3,ty);
1588             fiz3             = _mm256_add_ps(fiz3,tz);
1589
1590             fjx3             = _mm256_add_ps(fjx3,tx);
1591             fjy3             = _mm256_add_ps(fjy3,ty);
1592             fjz3             = _mm256_add_ps(fjz3,tz);
1593
1594             fjptrA             = f+j_coord_offsetA;
1595             fjptrB             = f+j_coord_offsetB;
1596             fjptrC             = f+j_coord_offsetC;
1597             fjptrD             = f+j_coord_offsetD;
1598             fjptrE             = f+j_coord_offsetE;
1599             fjptrF             = f+j_coord_offsetF;
1600             fjptrG             = f+j_coord_offsetG;
1601             fjptrH             = f+j_coord_offsetH;
1602
1603             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1604                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1605                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1606
1607             /* Inner loop uses 264 flops */
1608         }
1609
1610         if(jidx<j_index_end)
1611         {
1612
1613             /* Get j neighbor index, and coordinate index */
1614             jnrlistA         = jjnr[jidx];
1615             jnrlistB         = jjnr[jidx+1];
1616             jnrlistC         = jjnr[jidx+2];
1617             jnrlistD         = jjnr[jidx+3];
1618             jnrlistE         = jjnr[jidx+4];
1619             jnrlistF         = jjnr[jidx+5];
1620             jnrlistG         = jjnr[jidx+6];
1621             jnrlistH         = jjnr[jidx+7];
1622             /* Sign of each element will be negative for non-real atoms.
1623              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1624              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1625              */
1626             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1627                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1628
1629             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1630             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1631             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1632             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1633             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1634             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1635             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1636             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1637             j_coord_offsetA  = DIM*jnrA;
1638             j_coord_offsetB  = DIM*jnrB;
1639             j_coord_offsetC  = DIM*jnrC;
1640             j_coord_offsetD  = DIM*jnrD;
1641             j_coord_offsetE  = DIM*jnrE;
1642             j_coord_offsetF  = DIM*jnrF;
1643             j_coord_offsetG  = DIM*jnrG;
1644             j_coord_offsetH  = DIM*jnrH;
1645
1646             /* load j atom coordinates */
1647             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1648                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1649                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1650                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1651                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1652                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1653
1654             /* Calculate displacement vector */
1655             dx00             = _mm256_sub_ps(ix0,jx0);
1656             dy00             = _mm256_sub_ps(iy0,jy0);
1657             dz00             = _mm256_sub_ps(iz0,jz0);
1658             dx11             = _mm256_sub_ps(ix1,jx1);
1659             dy11             = _mm256_sub_ps(iy1,jy1);
1660             dz11             = _mm256_sub_ps(iz1,jz1);
1661             dx12             = _mm256_sub_ps(ix1,jx2);
1662             dy12             = _mm256_sub_ps(iy1,jy2);
1663             dz12             = _mm256_sub_ps(iz1,jz2);
1664             dx13             = _mm256_sub_ps(ix1,jx3);
1665             dy13             = _mm256_sub_ps(iy1,jy3);
1666             dz13             = _mm256_sub_ps(iz1,jz3);
1667             dx21             = _mm256_sub_ps(ix2,jx1);
1668             dy21             = _mm256_sub_ps(iy2,jy1);
1669             dz21             = _mm256_sub_ps(iz2,jz1);
1670             dx22             = _mm256_sub_ps(ix2,jx2);
1671             dy22             = _mm256_sub_ps(iy2,jy2);
1672             dz22             = _mm256_sub_ps(iz2,jz2);
1673             dx23             = _mm256_sub_ps(ix2,jx3);
1674             dy23             = _mm256_sub_ps(iy2,jy3);
1675             dz23             = _mm256_sub_ps(iz2,jz3);
1676             dx31             = _mm256_sub_ps(ix3,jx1);
1677             dy31             = _mm256_sub_ps(iy3,jy1);
1678             dz31             = _mm256_sub_ps(iz3,jz1);
1679             dx32             = _mm256_sub_ps(ix3,jx2);
1680             dy32             = _mm256_sub_ps(iy3,jy2);
1681             dz32             = _mm256_sub_ps(iz3,jz2);
1682             dx33             = _mm256_sub_ps(ix3,jx3);
1683             dy33             = _mm256_sub_ps(iy3,jy3);
1684             dz33             = _mm256_sub_ps(iz3,jz3);
1685
1686             /* Calculate squared distance and things based on it */
1687             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1688             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1689             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1690             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1691             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1692             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1693             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1694             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1695             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1696             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1697
1698             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1699             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1700             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1701             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1702             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1703             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1704             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1705             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1706             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1707
1708             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1709             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1710             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1711             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1712             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1713             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1714             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1715             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1716             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1717             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1718
1719             fjx0             = _mm256_setzero_ps();
1720             fjy0             = _mm256_setzero_ps();
1721             fjz0             = _mm256_setzero_ps();
1722             fjx1             = _mm256_setzero_ps();
1723             fjy1             = _mm256_setzero_ps();
1724             fjz1             = _mm256_setzero_ps();
1725             fjx2             = _mm256_setzero_ps();
1726             fjy2             = _mm256_setzero_ps();
1727             fjz2             = _mm256_setzero_ps();
1728             fjx3             = _mm256_setzero_ps();
1729             fjy3             = _mm256_setzero_ps();
1730             fjz3             = _mm256_setzero_ps();
1731
1732             /**************************
1733              * CALCULATE INTERACTIONS *
1734              **************************/
1735
1736             /* LENNARD-JONES DISPERSION/REPULSION */
1737
1738             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1739             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1740
1741             fscal            = fvdw;
1742
1743             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1744
1745             /* Calculate temporary vectorial force */
1746             tx               = _mm256_mul_ps(fscal,dx00);
1747             ty               = _mm256_mul_ps(fscal,dy00);
1748             tz               = _mm256_mul_ps(fscal,dz00);
1749
1750             /* Update vectorial force */
1751             fix0             = _mm256_add_ps(fix0,tx);
1752             fiy0             = _mm256_add_ps(fiy0,ty);
1753             fiz0             = _mm256_add_ps(fiz0,tz);
1754
1755             fjx0             = _mm256_add_ps(fjx0,tx);
1756             fjy0             = _mm256_add_ps(fjy0,ty);
1757             fjz0             = _mm256_add_ps(fjz0,tz);
1758
1759             /**************************
1760              * CALCULATE INTERACTIONS *
1761              **************************/
1762
1763             /* COULOMB ELECTROSTATICS */
1764             velec            = _mm256_mul_ps(qq11,rinv11);
1765             felec            = _mm256_mul_ps(velec,rinvsq11);
1766
1767             fscal            = felec;
1768
1769             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1770
1771             /* Calculate temporary vectorial force */
1772             tx               = _mm256_mul_ps(fscal,dx11);
1773             ty               = _mm256_mul_ps(fscal,dy11);
1774             tz               = _mm256_mul_ps(fscal,dz11);
1775
1776             /* Update vectorial force */
1777             fix1             = _mm256_add_ps(fix1,tx);
1778             fiy1             = _mm256_add_ps(fiy1,ty);
1779             fiz1             = _mm256_add_ps(fiz1,tz);
1780
1781             fjx1             = _mm256_add_ps(fjx1,tx);
1782             fjy1             = _mm256_add_ps(fjy1,ty);
1783             fjz1             = _mm256_add_ps(fjz1,tz);
1784
1785             /**************************
1786              * CALCULATE INTERACTIONS *
1787              **************************/
1788
1789             /* COULOMB ELECTROSTATICS */
1790             velec            = _mm256_mul_ps(qq12,rinv12);
1791             felec            = _mm256_mul_ps(velec,rinvsq12);
1792
1793             fscal            = felec;
1794
1795             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1796
1797             /* Calculate temporary vectorial force */
1798             tx               = _mm256_mul_ps(fscal,dx12);
1799             ty               = _mm256_mul_ps(fscal,dy12);
1800             tz               = _mm256_mul_ps(fscal,dz12);
1801
1802             /* Update vectorial force */
1803             fix1             = _mm256_add_ps(fix1,tx);
1804             fiy1             = _mm256_add_ps(fiy1,ty);
1805             fiz1             = _mm256_add_ps(fiz1,tz);
1806
1807             fjx2             = _mm256_add_ps(fjx2,tx);
1808             fjy2             = _mm256_add_ps(fjy2,ty);
1809             fjz2             = _mm256_add_ps(fjz2,tz);
1810
1811             /**************************
1812              * CALCULATE INTERACTIONS *
1813              **************************/
1814
1815             /* COULOMB ELECTROSTATICS */
1816             velec            = _mm256_mul_ps(qq13,rinv13);
1817             felec            = _mm256_mul_ps(velec,rinvsq13);
1818
1819             fscal            = felec;
1820
1821             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1822
1823             /* Calculate temporary vectorial force */
1824             tx               = _mm256_mul_ps(fscal,dx13);
1825             ty               = _mm256_mul_ps(fscal,dy13);
1826             tz               = _mm256_mul_ps(fscal,dz13);
1827
1828             /* Update vectorial force */
1829             fix1             = _mm256_add_ps(fix1,tx);
1830             fiy1             = _mm256_add_ps(fiy1,ty);
1831             fiz1             = _mm256_add_ps(fiz1,tz);
1832
1833             fjx3             = _mm256_add_ps(fjx3,tx);
1834             fjy3             = _mm256_add_ps(fjy3,ty);
1835             fjz3             = _mm256_add_ps(fjz3,tz);
1836
1837             /**************************
1838              * CALCULATE INTERACTIONS *
1839              **************************/
1840
1841             /* COULOMB ELECTROSTATICS */
1842             velec            = _mm256_mul_ps(qq21,rinv21);
1843             felec            = _mm256_mul_ps(velec,rinvsq21);
1844
1845             fscal            = felec;
1846
1847             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1848
1849             /* Calculate temporary vectorial force */
1850             tx               = _mm256_mul_ps(fscal,dx21);
1851             ty               = _mm256_mul_ps(fscal,dy21);
1852             tz               = _mm256_mul_ps(fscal,dz21);
1853
1854             /* Update vectorial force */
1855             fix2             = _mm256_add_ps(fix2,tx);
1856             fiy2             = _mm256_add_ps(fiy2,ty);
1857             fiz2             = _mm256_add_ps(fiz2,tz);
1858
1859             fjx1             = _mm256_add_ps(fjx1,tx);
1860             fjy1             = _mm256_add_ps(fjy1,ty);
1861             fjz1             = _mm256_add_ps(fjz1,tz);
1862
1863             /**************************
1864              * CALCULATE INTERACTIONS *
1865              **************************/
1866
1867             /* COULOMB ELECTROSTATICS */
1868             velec            = _mm256_mul_ps(qq22,rinv22);
1869             felec            = _mm256_mul_ps(velec,rinvsq22);
1870
1871             fscal            = felec;
1872
1873             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1874
1875             /* Calculate temporary vectorial force */
1876             tx               = _mm256_mul_ps(fscal,dx22);
1877             ty               = _mm256_mul_ps(fscal,dy22);
1878             tz               = _mm256_mul_ps(fscal,dz22);
1879
1880             /* Update vectorial force */
1881             fix2             = _mm256_add_ps(fix2,tx);
1882             fiy2             = _mm256_add_ps(fiy2,ty);
1883             fiz2             = _mm256_add_ps(fiz2,tz);
1884
1885             fjx2             = _mm256_add_ps(fjx2,tx);
1886             fjy2             = _mm256_add_ps(fjy2,ty);
1887             fjz2             = _mm256_add_ps(fjz2,tz);
1888
1889             /**************************
1890              * CALCULATE INTERACTIONS *
1891              **************************/
1892
1893             /* COULOMB ELECTROSTATICS */
1894             velec            = _mm256_mul_ps(qq23,rinv23);
1895             felec            = _mm256_mul_ps(velec,rinvsq23);
1896
1897             fscal            = felec;
1898
1899             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1900
1901             /* Calculate temporary vectorial force */
1902             tx               = _mm256_mul_ps(fscal,dx23);
1903             ty               = _mm256_mul_ps(fscal,dy23);
1904             tz               = _mm256_mul_ps(fscal,dz23);
1905
1906             /* Update vectorial force */
1907             fix2             = _mm256_add_ps(fix2,tx);
1908             fiy2             = _mm256_add_ps(fiy2,ty);
1909             fiz2             = _mm256_add_ps(fiz2,tz);
1910
1911             fjx3             = _mm256_add_ps(fjx3,tx);
1912             fjy3             = _mm256_add_ps(fjy3,ty);
1913             fjz3             = _mm256_add_ps(fjz3,tz);
1914
1915             /**************************
1916              * CALCULATE INTERACTIONS *
1917              **************************/
1918
1919             /* COULOMB ELECTROSTATICS */
1920             velec            = _mm256_mul_ps(qq31,rinv31);
1921             felec            = _mm256_mul_ps(velec,rinvsq31);
1922
1923             fscal            = felec;
1924
1925             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1926
1927             /* Calculate temporary vectorial force */
1928             tx               = _mm256_mul_ps(fscal,dx31);
1929             ty               = _mm256_mul_ps(fscal,dy31);
1930             tz               = _mm256_mul_ps(fscal,dz31);
1931
1932             /* Update vectorial force */
1933             fix3             = _mm256_add_ps(fix3,tx);
1934             fiy3             = _mm256_add_ps(fiy3,ty);
1935             fiz3             = _mm256_add_ps(fiz3,tz);
1936
1937             fjx1             = _mm256_add_ps(fjx1,tx);
1938             fjy1             = _mm256_add_ps(fjy1,ty);
1939             fjz1             = _mm256_add_ps(fjz1,tz);
1940
1941             /**************************
1942              * CALCULATE INTERACTIONS *
1943              **************************/
1944
1945             /* COULOMB ELECTROSTATICS */
1946             velec            = _mm256_mul_ps(qq32,rinv32);
1947             felec            = _mm256_mul_ps(velec,rinvsq32);
1948
1949             fscal            = felec;
1950
1951             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1952
1953             /* Calculate temporary vectorial force */
1954             tx               = _mm256_mul_ps(fscal,dx32);
1955             ty               = _mm256_mul_ps(fscal,dy32);
1956             tz               = _mm256_mul_ps(fscal,dz32);
1957
1958             /* Update vectorial force */
1959             fix3             = _mm256_add_ps(fix3,tx);
1960             fiy3             = _mm256_add_ps(fiy3,ty);
1961             fiz3             = _mm256_add_ps(fiz3,tz);
1962
1963             fjx2             = _mm256_add_ps(fjx2,tx);
1964             fjy2             = _mm256_add_ps(fjy2,ty);
1965             fjz2             = _mm256_add_ps(fjz2,tz);
1966
1967             /**************************
1968              * CALCULATE INTERACTIONS *
1969              **************************/
1970
1971             /* COULOMB ELECTROSTATICS */
1972             velec            = _mm256_mul_ps(qq33,rinv33);
1973             felec            = _mm256_mul_ps(velec,rinvsq33);
1974
1975             fscal            = felec;
1976
1977             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1978
1979             /* Calculate temporary vectorial force */
1980             tx               = _mm256_mul_ps(fscal,dx33);
1981             ty               = _mm256_mul_ps(fscal,dy33);
1982             tz               = _mm256_mul_ps(fscal,dz33);
1983
1984             /* Update vectorial force */
1985             fix3             = _mm256_add_ps(fix3,tx);
1986             fiy3             = _mm256_add_ps(fiy3,ty);
1987             fiz3             = _mm256_add_ps(fiz3,tz);
1988
1989             fjx3             = _mm256_add_ps(fjx3,tx);
1990             fjy3             = _mm256_add_ps(fjy3,ty);
1991             fjz3             = _mm256_add_ps(fjz3,tz);
1992
1993             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1994             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1995             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1996             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1997             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1998             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1999             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2000             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2001
2002             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2003                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2004                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2005
2006             /* Inner loop uses 264 flops */
2007         }
2008
2009         /* End of innermost loop */
2010
2011         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2012                                                  f+i_coord_offset,fshift+i_shift_offset);
2013
2014         /* Increment number of inner iterations */
2015         inneriter                  += j_index_end - j_index_start;
2016
2017         /* Outer loop uses 24 flops */
2018     }
2019
2020     /* Increment number of outer iterations */
2021     outeriter        += nri;
2022
2023     /* Update outer/inner flops */
2024
2025     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*264);
2026 }