src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwLJ_GeomW4W4_avx_256_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_256_single kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_256_single.h"
  34 #include "kernelutil_x86_avx_256_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_single
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_VF_avx_256_single
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrE,jnrF,jnrG,jnrH;
  62     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  63     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  64     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  65     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  66     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  67     real             rcutoff_scalar;
  68     real             *shiftvec,*fshift,*x,*f;
  69     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  70     real             scratch[4*DIM];
  71     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  72     real *           vdwioffsetptr0;
  73     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  74     real *           vdwioffsetptr1;
  75     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  76     real *           vdwioffsetptr2;
  77     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  78     real *           vdwioffsetptr3;
  79     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  80     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  81     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  82     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  83     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  84     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  85     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  86     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  87     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  88     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  89     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  90     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  91     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  92     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  93     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  94     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  95     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  96     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  97     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  98     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  99     real             *charge;
 100     int              nvdwtype;
 101     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 102     int              *vdwtype;
 103     real             *vdwparam;
 104     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 105     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 106     __m256           dummy_mask,cutoff_mask;
 107     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 108     __m256           one     = _mm256_set1_ps(1.0);
 109     __m256           two     = _mm256_set1_ps(2.0);
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = _mm256_set1_ps(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123     nvdwtype         = fr->ntype;
 124     vdwparam         = fr->nbfp;
 125     vdwtype          = mdatoms->typeA;
 126
 127     /* Setup water-specific parameters */
 128     inr              = nlist->iinr[0];
 129     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 130     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 131     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 132     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 133
 134     jq1              = _mm256_set1_ps(charge[inr+1]);
 135     jq2              = _mm256_set1_ps(charge[inr+2]);
 136     jq3              = _mm256_set1_ps(charge[inr+3]);
 137     vdwjidx0A        = 2*vdwtype[inr+0];
 138     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 139     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 140     qq11             = _mm256_mul_ps(iq1,jq1);
 141     qq12             = _mm256_mul_ps(iq1,jq2);
 142     qq13             = _mm256_mul_ps(iq1,jq3);
 143     qq21             = _mm256_mul_ps(iq2,jq1);
 144     qq22             = _mm256_mul_ps(iq2,jq2);
 145     qq23             = _mm256_mul_ps(iq2,jq3);
 146     qq31             = _mm256_mul_ps(iq3,jq1);
 147     qq32             = _mm256_mul_ps(iq3,jq2);
 148     qq33             = _mm256_mul_ps(iq3,jq3);
 149
 150     /* Avoid stupid compiler warnings */
 151     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 152     j_coord_offsetA = 0;
 153     j_coord_offsetB = 0;
 154     j_coord_offsetC = 0;
 155     j_coord_offsetD = 0;
 156     j_coord_offsetE = 0;
 157     j_coord_offsetF = 0;
 158     j_coord_offsetG = 0;
 159     j_coord_offsetH = 0;
 160
 161     outeriter        = 0;
 162     inneriter        = 0;
 163
 164     for(iidx=0;iidx<4*DIM;iidx++)
 165     {
 166         scratch[iidx] = 0.0;
 167     }
 168
 169     /* Start outer loop over neighborlists */
 170     for(iidx=0; iidx<nri; iidx++)
 171     {
 172         /* Load shift vector for this list */
 173         i_shift_offset   = DIM*shiftidx[iidx];
 174
 175         /* Load limits for loop over neighbors */
 176         j_index_start    = jindex[iidx];
 177         j_index_end      = jindex[iidx+1];
 178
 179         /* Get outer coordinate index */
 180         inr              = iinr[iidx];
 181         i_coord_offset   = DIM*inr;
 182
 183         /* Load i particle coords and add shift vector */
 184         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 185                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 186
 187         fix0             = _mm256_setzero_ps();
 188         fiy0             = _mm256_setzero_ps();
 189         fiz0             = _mm256_setzero_ps();
 190         fix1             = _mm256_setzero_ps();
 191         fiy1             = _mm256_setzero_ps();
 192         fiz1             = _mm256_setzero_ps();
 193         fix2             = _mm256_setzero_ps();
 194         fiy2             = _mm256_setzero_ps();
 195         fiz2             = _mm256_setzero_ps();
 196         fix3             = _mm256_setzero_ps();
 197         fiy3             = _mm256_setzero_ps();
 198         fiz3             = _mm256_setzero_ps();
 199
 200         /* Reset potential sums */
 201         velecsum         = _mm256_setzero_ps();
 202         vvdwsum          = _mm256_setzero_ps();
 203
 204         /* Start inner kernel loop */
 205         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 206         {
 207
 208             /* Get j neighbor index, and coordinate index */
 209             jnrA             = jjnr[jidx];
 210             jnrB             = jjnr[jidx+1];
 211             jnrC             = jjnr[jidx+2];
 212             jnrD             = jjnr[jidx+3];
 213             jnrE             = jjnr[jidx+4];
 214             jnrF             = jjnr[jidx+5];
 215             jnrG             = jjnr[jidx+6];
 216             jnrH             = jjnr[jidx+7];
 217             j_coord_offsetA  = DIM*jnrA;
 218             j_coord_offsetB  = DIM*jnrB;
 219             j_coord_offsetC  = DIM*jnrC;
 220             j_coord_offsetD  = DIM*jnrD;
 221             j_coord_offsetE  = DIM*jnrE;
 222             j_coord_offsetF  = DIM*jnrF;
 223             j_coord_offsetG  = DIM*jnrG;
 224             j_coord_offsetH  = DIM*jnrH;
 225
 226             /* load j atom coordinates */
 227             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 228                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 229                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 230                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 231                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 232                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 233
 234             /* Calculate displacement vector */
 235             dx00             = _mm256_sub_ps(ix0,jx0);
 236             dy00             = _mm256_sub_ps(iy0,jy0);
 237             dz00             = _mm256_sub_ps(iz0,jz0);
 238             dx11             = _mm256_sub_ps(ix1,jx1);
 239             dy11             = _mm256_sub_ps(iy1,jy1);
 240             dz11             = _mm256_sub_ps(iz1,jz1);
 241             dx12             = _mm256_sub_ps(ix1,jx2);
 242             dy12             = _mm256_sub_ps(iy1,jy2);
 243             dz12             = _mm256_sub_ps(iz1,jz2);
 244             dx13             = _mm256_sub_ps(ix1,jx3);
 245             dy13             = _mm256_sub_ps(iy1,jy3);
 246             dz13             = _mm256_sub_ps(iz1,jz3);
 247             dx21             = _mm256_sub_ps(ix2,jx1);
 248             dy21             = _mm256_sub_ps(iy2,jy1);
 249             dz21             = _mm256_sub_ps(iz2,jz1);
 250             dx22             = _mm256_sub_ps(ix2,jx2);
 251             dy22             = _mm256_sub_ps(iy2,jy2);
 252             dz22             = _mm256_sub_ps(iz2,jz2);
 253             dx23             = _mm256_sub_ps(ix2,jx3);
 254             dy23             = _mm256_sub_ps(iy2,jy3);
 255             dz23             = _mm256_sub_ps(iz2,jz3);
 256             dx31             = _mm256_sub_ps(ix3,jx1);
 257             dy31             = _mm256_sub_ps(iy3,jy1);
 258             dz31             = _mm256_sub_ps(iz3,jz1);
 259             dx32             = _mm256_sub_ps(ix3,jx2);
 260             dy32             = _mm256_sub_ps(iy3,jy2);
 261             dz32             = _mm256_sub_ps(iz3,jz2);
 262             dx33             = _mm256_sub_ps(ix3,jx3);
 263             dy33             = _mm256_sub_ps(iy3,jy3);
 264             dz33             = _mm256_sub_ps(iz3,jz3);
 265
 266             /* Calculate squared distance and things based on it */
 267             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 268             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 269             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 270             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 271             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 272             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 273             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 274             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 275             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 276             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 277
 278             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 279             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 280             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 281             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 282             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 283             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 284             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 285             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 286             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 287
 288             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 289             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 290             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 291             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 292             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 293             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 294             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 295             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 296             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 297             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 298
 299             fjx0             = _mm256_setzero_ps();
 300             fjy0             = _mm256_setzero_ps();
 301             fjz0             = _mm256_setzero_ps();
 302             fjx1             = _mm256_setzero_ps();
 303             fjy1             = _mm256_setzero_ps();
 304             fjz1             = _mm256_setzero_ps();
 305             fjx2             = _mm256_setzero_ps();
 306             fjy2             = _mm256_setzero_ps();
 307             fjz2             = _mm256_setzero_ps();
 308             fjx3             = _mm256_setzero_ps();
 309             fjy3             = _mm256_setzero_ps();
 310             fjz3             = _mm256_setzero_ps();
 311
 312             /**************************
 313              * CALCULATE INTERACTIONS *
 314              **************************/
 315
 316             /* LENNARD-JONES DISPERSION/REPULSION */
 317
 318             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 319             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 320             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 321             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 322             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 323
 324             /* Update potential sum for this i atom from the interaction with this j atom. */
 325             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 326
 327             fscal            = fvdw;
 328
 329             /* Calculate temporary vectorial force */
 330             tx               = _mm256_mul_ps(fscal,dx00);
 331             ty               = _mm256_mul_ps(fscal,dy00);
 332             tz               = _mm256_mul_ps(fscal,dz00);
 333
 334             /* Update vectorial force */
 335             fix0             = _mm256_add_ps(fix0,tx);
 336             fiy0             = _mm256_add_ps(fiy0,ty);
 337             fiz0             = _mm256_add_ps(fiz0,tz);
 338
 339             fjx0             = _mm256_add_ps(fjx0,tx);
 340             fjy0             = _mm256_add_ps(fjy0,ty);
 341             fjz0             = _mm256_add_ps(fjz0,tz);
 342
 343             /**************************
 344              * CALCULATE INTERACTIONS *
 345              **************************/
 346
 347             /* COULOMB ELECTROSTATICS */
 348             velec            = _mm256_mul_ps(qq11,rinv11);
 349             felec            = _mm256_mul_ps(velec,rinvsq11);
 350
 351             /* Update potential sum for this i atom from the interaction with this j atom. */
 352             velecsum         = _mm256_add_ps(velecsum,velec);
 353
 354             fscal            = felec;
 355
 356             /* Calculate temporary vectorial force */
 357             tx               = _mm256_mul_ps(fscal,dx11);
 358             ty               = _mm256_mul_ps(fscal,dy11);
 359             tz               = _mm256_mul_ps(fscal,dz11);
 360
 361             /* Update vectorial force */
 362             fix1             = _mm256_add_ps(fix1,tx);
 363             fiy1             = _mm256_add_ps(fiy1,ty);
 364             fiz1             = _mm256_add_ps(fiz1,tz);
 365
 366             fjx1             = _mm256_add_ps(fjx1,tx);
 367             fjy1             = _mm256_add_ps(fjy1,ty);
 368             fjz1             = _mm256_add_ps(fjz1,tz);
 369
 370             /**************************
 371              * CALCULATE INTERACTIONS *
 372              **************************/
 373
 374             /* COULOMB ELECTROSTATICS */
 375             velec            = _mm256_mul_ps(qq12,rinv12);
 376             felec            = _mm256_mul_ps(velec,rinvsq12);
 377
 378             /* Update potential sum for this i atom from the interaction with this j atom. */
 379             velecsum         = _mm256_add_ps(velecsum,velec);
 380
 381             fscal            = felec;
 382
 383             /* Calculate temporary vectorial force */
 384             tx               = _mm256_mul_ps(fscal,dx12);
 385             ty               = _mm256_mul_ps(fscal,dy12);
 386             tz               = _mm256_mul_ps(fscal,dz12);
 387
 388             /* Update vectorial force */
 389             fix1             = _mm256_add_ps(fix1,tx);
 390             fiy1             = _mm256_add_ps(fiy1,ty);
 391             fiz1             = _mm256_add_ps(fiz1,tz);
 392
 393             fjx2             = _mm256_add_ps(fjx2,tx);
 394             fjy2             = _mm256_add_ps(fjy2,ty);
 395             fjz2             = _mm256_add_ps(fjz2,tz);
 396
 397             /**************************
 398              * CALCULATE INTERACTIONS *
 399              **************************/
 400
 401             /* COULOMB ELECTROSTATICS */
 402             velec            = _mm256_mul_ps(qq13,rinv13);
 403             felec            = _mm256_mul_ps(velec,rinvsq13);
 404
 405             /* Update potential sum for this i atom from the interaction with this j atom. */
 406             velecsum         = _mm256_add_ps(velecsum,velec);
 407
 408             fscal            = felec;
 409
 410             /* Calculate temporary vectorial force */
 411             tx               = _mm256_mul_ps(fscal,dx13);
 412             ty               = _mm256_mul_ps(fscal,dy13);
 413             tz               = _mm256_mul_ps(fscal,dz13);
 414
 415             /* Update vectorial force */
 416             fix1             = _mm256_add_ps(fix1,tx);
 417             fiy1             = _mm256_add_ps(fiy1,ty);
 418             fiz1             = _mm256_add_ps(fiz1,tz);
 419
 420             fjx3             = _mm256_add_ps(fjx3,tx);
 421             fjy3             = _mm256_add_ps(fjy3,ty);
 422             fjz3             = _mm256_add_ps(fjz3,tz);
 423
 424             /**************************
 425              * CALCULATE INTERACTIONS *
 426              **************************/
 427
 428             /* COULOMB ELECTROSTATICS */
 429             velec            = _mm256_mul_ps(qq21,rinv21);
 430             felec            = _mm256_mul_ps(velec,rinvsq21);
 431
 432             /* Update potential sum for this i atom from the interaction with this j atom. */
 433             velecsum         = _mm256_add_ps(velecsum,velec);
 434
 435             fscal            = felec;
 436
 437             /* Calculate temporary vectorial force */
 438             tx               = _mm256_mul_ps(fscal,dx21);
 439             ty               = _mm256_mul_ps(fscal,dy21);
 440             tz               = _mm256_mul_ps(fscal,dz21);
 441
 442             /* Update vectorial force */
 443             fix2             = _mm256_add_ps(fix2,tx);
 444             fiy2             = _mm256_add_ps(fiy2,ty);
 445             fiz2             = _mm256_add_ps(fiz2,tz);
 446
 447             fjx1             = _mm256_add_ps(fjx1,tx);
 448             fjy1             = _mm256_add_ps(fjy1,ty);
 449             fjz1             = _mm256_add_ps(fjz1,tz);
 450
 451             /**************************
 452              * CALCULATE INTERACTIONS *
 453              **************************/
 454
 455             /* COULOMB ELECTROSTATICS */
 456             velec            = _mm256_mul_ps(qq22,rinv22);
 457             felec            = _mm256_mul_ps(velec,rinvsq22);
 458
 459             /* Update potential sum for this i atom from the interaction with this j atom. */
 460             velecsum         = _mm256_add_ps(velecsum,velec);
 461
 462             fscal            = felec;
 463
 464             /* Calculate temporary vectorial force */
 465             tx               = _mm256_mul_ps(fscal,dx22);
 466             ty               = _mm256_mul_ps(fscal,dy22);
 467             tz               = _mm256_mul_ps(fscal,dz22);
 468
 469             /* Update vectorial force */
 470             fix2             = _mm256_add_ps(fix2,tx);
 471             fiy2             = _mm256_add_ps(fiy2,ty);
 472             fiz2             = _mm256_add_ps(fiz2,tz);
 473
 474             fjx2             = _mm256_add_ps(fjx2,tx);
 475             fjy2             = _mm256_add_ps(fjy2,ty);
 476             fjz2             = _mm256_add_ps(fjz2,tz);
 477
 478             /**************************
 479              * CALCULATE INTERACTIONS *
 480              **************************/
 481
 482             /* COULOMB ELECTROSTATICS */
 483             velec            = _mm256_mul_ps(qq23,rinv23);
 484             felec            = _mm256_mul_ps(velec,rinvsq23);
 485
 486             /* Update potential sum for this i atom from the interaction with this j atom. */
 487             velecsum         = _mm256_add_ps(velecsum,velec);
 488
 489             fscal            = felec;
 490
 491             /* Calculate temporary vectorial force */
 492             tx               = _mm256_mul_ps(fscal,dx23);
 493             ty               = _mm256_mul_ps(fscal,dy23);
 494             tz               = _mm256_mul_ps(fscal,dz23);
 495
 496             /* Update vectorial force */
 497             fix2             = _mm256_add_ps(fix2,tx);
 498             fiy2             = _mm256_add_ps(fiy2,ty);
 499             fiz2             = _mm256_add_ps(fiz2,tz);
 500
 501             fjx3             = _mm256_add_ps(fjx3,tx);
 502             fjy3             = _mm256_add_ps(fjy3,ty);
 503             fjz3             = _mm256_add_ps(fjz3,tz);
 504
 505             /**************************
 506              * CALCULATE INTERACTIONS *
 507              **************************/
 508
 509             /* COULOMB ELECTROSTATICS */
 510             velec            = _mm256_mul_ps(qq31,rinv31);
 511             felec            = _mm256_mul_ps(velec,rinvsq31);
 512
 513             /* Update potential sum for this i atom from the interaction with this j atom. */
 514             velecsum         = _mm256_add_ps(velecsum,velec);
 515
 516             fscal            = felec;
 517
 518             /* Calculate temporary vectorial force */
 519             tx               = _mm256_mul_ps(fscal,dx31);
 520             ty               = _mm256_mul_ps(fscal,dy31);
 521             tz               = _mm256_mul_ps(fscal,dz31);
 522
 523             /* Update vectorial force */
 524             fix3             = _mm256_add_ps(fix3,tx);
 525             fiy3             = _mm256_add_ps(fiy3,ty);
 526             fiz3             = _mm256_add_ps(fiz3,tz);
 527
 528             fjx1             = _mm256_add_ps(fjx1,tx);
 529             fjy1             = _mm256_add_ps(fjy1,ty);
 530             fjz1             = _mm256_add_ps(fjz1,tz);
 531
 532             /**************************
 533              * CALCULATE INTERACTIONS *
 534              **************************/
 535
 536             /* COULOMB ELECTROSTATICS */
 537             velec            = _mm256_mul_ps(qq32,rinv32);
 538             felec            = _mm256_mul_ps(velec,rinvsq32);
 539
 540             /* Update potential sum for this i atom from the interaction with this j atom. */
 541             velecsum         = _mm256_add_ps(velecsum,velec);
 542
 543             fscal            = felec;
 544
 545             /* Calculate temporary vectorial force */
 546             tx               = _mm256_mul_ps(fscal,dx32);
 547             ty               = _mm256_mul_ps(fscal,dy32);
 548             tz               = _mm256_mul_ps(fscal,dz32);
 549
 550             /* Update vectorial force */
 551             fix3             = _mm256_add_ps(fix3,tx);
 552             fiy3             = _mm256_add_ps(fiy3,ty);
 553             fiz3             = _mm256_add_ps(fiz3,tz);
 554
 555             fjx2             = _mm256_add_ps(fjx2,tx);
 556             fjy2             = _mm256_add_ps(fjy2,ty);
 557             fjz2             = _mm256_add_ps(fjz2,tz);
 558
 559             /**************************
 560              * CALCULATE INTERACTIONS *
 561              **************************/
 562
 563             /* COULOMB ELECTROSTATICS */
 564             velec            = _mm256_mul_ps(qq33,rinv33);
 565             felec            = _mm256_mul_ps(velec,rinvsq33);
 566
 567             /* Update potential sum for this i atom from the interaction with this j atom. */
 568             velecsum         = _mm256_add_ps(velecsum,velec);
 569
 570             fscal            = felec;
 571
 572             /* Calculate temporary vectorial force */
 573             tx               = _mm256_mul_ps(fscal,dx33);
 574             ty               = _mm256_mul_ps(fscal,dy33);
 575             tz               = _mm256_mul_ps(fscal,dz33);
 576
 577             /* Update vectorial force */
 578             fix3             = _mm256_add_ps(fix3,tx);
 579             fiy3             = _mm256_add_ps(fiy3,ty);
 580             fiz3             = _mm256_add_ps(fiz3,tz);
 581
 582             fjx3             = _mm256_add_ps(fjx3,tx);
 583             fjy3             = _mm256_add_ps(fjy3,ty);
 584             fjz3             = _mm256_add_ps(fjz3,tz);
 585
 586             fjptrA             = f+j_coord_offsetA;
 587             fjptrB             = f+j_coord_offsetB;
 588             fjptrC             = f+j_coord_offsetC;
 589             fjptrD             = f+j_coord_offsetD;
 590             fjptrE             = f+j_coord_offsetE;
 591             fjptrF             = f+j_coord_offsetF;
 592             fjptrG             = f+j_coord_offsetG;
 593             fjptrH             = f+j_coord_offsetH;
 594
 595             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 596                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 597                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 598
 599             /* Inner loop uses 278 flops */
 600         }
 601
 602         if(jidx<j_index_end)
 603         {
 604
 605             /* Get j neighbor index, and coordinate index */
 606             jnrlistA         = jjnr[jidx];
 607             jnrlistB         = jjnr[jidx+1];
 608             jnrlistC         = jjnr[jidx+2];
 609             jnrlistD         = jjnr[jidx+3];
 610             jnrlistE         = jjnr[jidx+4];
 611             jnrlistF         = jjnr[jidx+5];
 612             jnrlistG         = jjnr[jidx+6];
 613             jnrlistH         = jjnr[jidx+7];
 614             /* Sign of each element will be negative for non-real atoms.
 615              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 616              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 617              */
 618             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 619                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 620
 621             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 622             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 623             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 624             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 625             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 626             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 627             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 628             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 629             j_coord_offsetA  = DIM*jnrA;
 630             j_coord_offsetB  = DIM*jnrB;
 631             j_coord_offsetC  = DIM*jnrC;
 632             j_coord_offsetD  = DIM*jnrD;
 633             j_coord_offsetE  = DIM*jnrE;
 634             j_coord_offsetF  = DIM*jnrF;
 635             j_coord_offsetG  = DIM*jnrG;
 636             j_coord_offsetH  = DIM*jnrH;
 637
 638             /* load j atom coordinates */
 639             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 640                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 641                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 642                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 643                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 644                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 645
 646             /* Calculate displacement vector */
 647             dx00             = _mm256_sub_ps(ix0,jx0);
 648             dy00             = _mm256_sub_ps(iy0,jy0);
 649             dz00             = _mm256_sub_ps(iz0,jz0);
 650             dx11             = _mm256_sub_ps(ix1,jx1);
 651             dy11             = _mm256_sub_ps(iy1,jy1);
 652             dz11             = _mm256_sub_ps(iz1,jz1);
 653             dx12             = _mm256_sub_ps(ix1,jx2);
 654             dy12             = _mm256_sub_ps(iy1,jy2);
 655             dz12             = _mm256_sub_ps(iz1,jz2);
 656             dx13             = _mm256_sub_ps(ix1,jx3);
 657             dy13             = _mm256_sub_ps(iy1,jy3);
 658             dz13             = _mm256_sub_ps(iz1,jz3);
 659             dx21             = _mm256_sub_ps(ix2,jx1);
 660             dy21             = _mm256_sub_ps(iy2,jy1);
 661             dz21             = _mm256_sub_ps(iz2,jz1);
 662             dx22             = _mm256_sub_ps(ix2,jx2);
 663             dy22             = _mm256_sub_ps(iy2,jy2);
 664             dz22             = _mm256_sub_ps(iz2,jz2);
 665             dx23             = _mm256_sub_ps(ix2,jx3);
 666             dy23             = _mm256_sub_ps(iy2,jy3);
 667             dz23             = _mm256_sub_ps(iz2,jz3);
 668             dx31             = _mm256_sub_ps(ix3,jx1);
 669             dy31             = _mm256_sub_ps(iy3,jy1);
 670             dz31             = _mm256_sub_ps(iz3,jz1);
 671             dx32             = _mm256_sub_ps(ix3,jx2);
 672             dy32             = _mm256_sub_ps(iy3,jy2);
 673             dz32             = _mm256_sub_ps(iz3,jz2);
 674             dx33             = _mm256_sub_ps(ix3,jx3);
 675             dy33             = _mm256_sub_ps(iy3,jy3);
 676             dz33             = _mm256_sub_ps(iz3,jz3);
 677
 678             /* Calculate squared distance and things based on it */
 679             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 680             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 681             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 682             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 683             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 684             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 685             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 686             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 687             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 688             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 689
 690             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 691             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 692             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 693             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 694             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 695             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 696             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 697             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 698             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 699
 700             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 701             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 702             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 703             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 704             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 705             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 706             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 707             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 708             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 709             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 710
 711             fjx0             = _mm256_setzero_ps();
 712             fjy0             = _mm256_setzero_ps();
 713             fjz0             = _mm256_setzero_ps();
 714             fjx1             = _mm256_setzero_ps();
 715             fjy1             = _mm256_setzero_ps();
 716             fjz1             = _mm256_setzero_ps();
 717             fjx2             = _mm256_setzero_ps();
 718             fjy2             = _mm256_setzero_ps();
 719             fjz2             = _mm256_setzero_ps();
 720             fjx3             = _mm256_setzero_ps();
 721             fjy3             = _mm256_setzero_ps();
 722             fjz3             = _mm256_setzero_ps();
 723
 724             /**************************
 725              * CALCULATE INTERACTIONS *
 726              **************************/
 727
 728             /* LENNARD-JONES DISPERSION/REPULSION */
 729
 730             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 731             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 732             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 733             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 734             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 735
 736             /* Update potential sum for this i atom from the interaction with this j atom. */
 737             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 738             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 739
 740             fscal            = fvdw;
 741
 742             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 743
 744             /* Calculate temporary vectorial force */
 745             tx               = _mm256_mul_ps(fscal,dx00);
 746             ty               = _mm256_mul_ps(fscal,dy00);
 747             tz               = _mm256_mul_ps(fscal,dz00);
 748
 749             /* Update vectorial force */
 750             fix0             = _mm256_add_ps(fix0,tx);
 751             fiy0             = _mm256_add_ps(fiy0,ty);
 752             fiz0             = _mm256_add_ps(fiz0,tz);
 753
 754             fjx0             = _mm256_add_ps(fjx0,tx);
 755             fjy0             = _mm256_add_ps(fjy0,ty);
 756             fjz0             = _mm256_add_ps(fjz0,tz);
 757
 758             /**************************
 759              * CALCULATE INTERACTIONS *
 760              **************************/
 761
 762             /* COULOMB ELECTROSTATICS */
 763             velec            = _mm256_mul_ps(qq11,rinv11);
 764             felec            = _mm256_mul_ps(velec,rinvsq11);
 765
 766             /* Update potential sum for this i atom from the interaction with this j atom. */
 767             velec            = _mm256_andnot_ps(dummy_mask,velec);
 768             velecsum         = _mm256_add_ps(velecsum,velec);
 769
 770             fscal            = felec;
 771
 772             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 773
 774             /* Calculate temporary vectorial force */
 775             tx               = _mm256_mul_ps(fscal,dx11);
 776             ty               = _mm256_mul_ps(fscal,dy11);
 777             tz               = _mm256_mul_ps(fscal,dz11);
 778
 779             /* Update vectorial force */
 780             fix1             = _mm256_add_ps(fix1,tx);
 781             fiy1             = _mm256_add_ps(fiy1,ty);
 782             fiz1             = _mm256_add_ps(fiz1,tz);
 783
 784             fjx1             = _mm256_add_ps(fjx1,tx);
 785             fjy1             = _mm256_add_ps(fjy1,ty);
 786             fjz1             = _mm256_add_ps(fjz1,tz);
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             /* COULOMB ELECTROSTATICS */
 793             velec            = _mm256_mul_ps(qq12,rinv12);
 794             felec            = _mm256_mul_ps(velec,rinvsq12);
 795
 796             /* Update potential sum for this i atom from the interaction with this j atom. */
 797             velec            = _mm256_andnot_ps(dummy_mask,velec);
 798             velecsum         = _mm256_add_ps(velecsum,velec);
 799
 800             fscal            = felec;
 801
 802             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 803
 804             /* Calculate temporary vectorial force */
 805             tx               = _mm256_mul_ps(fscal,dx12);
 806             ty               = _mm256_mul_ps(fscal,dy12);
 807             tz               = _mm256_mul_ps(fscal,dz12);
 808
 809             /* Update vectorial force */
 810             fix1             = _mm256_add_ps(fix1,tx);
 811             fiy1             = _mm256_add_ps(fiy1,ty);
 812             fiz1             = _mm256_add_ps(fiz1,tz);
 813
 814             fjx2             = _mm256_add_ps(fjx2,tx);
 815             fjy2             = _mm256_add_ps(fjy2,ty);
 816             fjz2             = _mm256_add_ps(fjz2,tz);
 817
 818             /**************************
 819              * CALCULATE INTERACTIONS *
 820              **************************/
 821
 822             /* COULOMB ELECTROSTATICS */
 823             velec            = _mm256_mul_ps(qq13,rinv13);
 824             felec            = _mm256_mul_ps(velec,rinvsq13);
 825
 826             /* Update potential sum for this i atom from the interaction with this j atom. */
 827             velec            = _mm256_andnot_ps(dummy_mask,velec);
 828             velecsum         = _mm256_add_ps(velecsum,velec);
 829
 830             fscal            = felec;
 831
 832             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 833
 834             /* Calculate temporary vectorial force */
 835             tx               = _mm256_mul_ps(fscal,dx13);
 836             ty               = _mm256_mul_ps(fscal,dy13);
 837             tz               = _mm256_mul_ps(fscal,dz13);
 838
 839             /* Update vectorial force */
 840             fix1             = _mm256_add_ps(fix1,tx);
 841             fiy1             = _mm256_add_ps(fiy1,ty);
 842             fiz1             = _mm256_add_ps(fiz1,tz);
 843
 844             fjx3             = _mm256_add_ps(fjx3,tx);
 845             fjy3             = _mm256_add_ps(fjy3,ty);
 846             fjz3             = _mm256_add_ps(fjz3,tz);
 847
 848             /**************************
 849              * CALCULATE INTERACTIONS *
 850              **************************/
 851
 852             /* COULOMB ELECTROSTATICS */
 853             velec            = _mm256_mul_ps(qq21,rinv21);
 854             felec            = _mm256_mul_ps(velec,rinvsq21);
 855
 856             /* Update potential sum for this i atom from the interaction with this j atom. */
 857             velec            = _mm256_andnot_ps(dummy_mask,velec);
 858             velecsum         = _mm256_add_ps(velecsum,velec);
 859
 860             fscal            = felec;
 861
 862             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 863
 864             /* Calculate temporary vectorial force */
 865             tx               = _mm256_mul_ps(fscal,dx21);
 866             ty               = _mm256_mul_ps(fscal,dy21);
 867             tz               = _mm256_mul_ps(fscal,dz21);
 868
 869             /* Update vectorial force */
 870             fix2             = _mm256_add_ps(fix2,tx);
 871             fiy2             = _mm256_add_ps(fiy2,ty);
 872             fiz2             = _mm256_add_ps(fiz2,tz);
 873
 874             fjx1             = _mm256_add_ps(fjx1,tx);
 875             fjy1             = _mm256_add_ps(fjy1,ty);
 876             fjz1             = _mm256_add_ps(fjz1,tz);
 877
 878             /**************************
 879              * CALCULATE INTERACTIONS *
 880              **************************/
 881
 882             /* COULOMB ELECTROSTATICS */
 883             velec            = _mm256_mul_ps(qq22,rinv22);
 884             felec            = _mm256_mul_ps(velec,rinvsq22);
 885
 886             /* Update potential sum for this i atom from the interaction with this j atom. */
 887             velec            = _mm256_andnot_ps(dummy_mask,velec);
 888             velecsum         = _mm256_add_ps(velecsum,velec);
 889
 890             fscal            = felec;
 891
 892             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 893
 894             /* Calculate temporary vectorial force */
 895             tx               = _mm256_mul_ps(fscal,dx22);
 896             ty               = _mm256_mul_ps(fscal,dy22);
 897             tz               = _mm256_mul_ps(fscal,dz22);
 898
 899             /* Update vectorial force */
 900             fix2             = _mm256_add_ps(fix2,tx);
 901             fiy2             = _mm256_add_ps(fiy2,ty);
 902             fiz2             = _mm256_add_ps(fiz2,tz);
 903
 904             fjx2             = _mm256_add_ps(fjx2,tx);
 905             fjy2             = _mm256_add_ps(fjy2,ty);
 906             fjz2             = _mm256_add_ps(fjz2,tz);
 907
 908             /**************************
 909              * CALCULATE INTERACTIONS *
 910              **************************/
 911
 912             /* COULOMB ELECTROSTATICS */
 913             velec            = _mm256_mul_ps(qq23,rinv23);
 914             felec            = _mm256_mul_ps(velec,rinvsq23);
 915
 916             /* Update potential sum for this i atom from the interaction with this j atom. */
 917             velec            = _mm256_andnot_ps(dummy_mask,velec);
 918             velecsum         = _mm256_add_ps(velecsum,velec);
 919
 920             fscal            = felec;
 921
 922             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 923
 924             /* Calculate temporary vectorial force */
 925             tx               = _mm256_mul_ps(fscal,dx23);
 926             ty               = _mm256_mul_ps(fscal,dy23);
 927             tz               = _mm256_mul_ps(fscal,dz23);
 928
 929             /* Update vectorial force */
 930             fix2             = _mm256_add_ps(fix2,tx);
 931             fiy2             = _mm256_add_ps(fiy2,ty);
 932             fiz2             = _mm256_add_ps(fiz2,tz);
 933
 934             fjx3             = _mm256_add_ps(fjx3,tx);
 935             fjy3             = _mm256_add_ps(fjy3,ty);
 936             fjz3             = _mm256_add_ps(fjz3,tz);
 937
 938             /**************************
 939              * CALCULATE INTERACTIONS *
 940              **************************/
 941
 942             /* COULOMB ELECTROSTATICS */
 943             velec            = _mm256_mul_ps(qq31,rinv31);
 944             felec            = _mm256_mul_ps(velec,rinvsq31);
 945
 946             /* Update potential sum for this i atom from the interaction with this j atom. */
 947             velec            = _mm256_andnot_ps(dummy_mask,velec);
 948             velecsum         = _mm256_add_ps(velecsum,velec);
 949
 950             fscal            = felec;
 951
 952             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 953
 954             /* Calculate temporary vectorial force */
 955             tx               = _mm256_mul_ps(fscal,dx31);
 956             ty               = _mm256_mul_ps(fscal,dy31);
 957             tz               = _mm256_mul_ps(fscal,dz31);
 958
 959             /* Update vectorial force */
 960             fix3             = _mm256_add_ps(fix3,tx);
 961             fiy3             = _mm256_add_ps(fiy3,ty);
 962             fiz3             = _mm256_add_ps(fiz3,tz);
 963
 964             fjx1             = _mm256_add_ps(fjx1,tx);
 965             fjy1             = _mm256_add_ps(fjy1,ty);
 966             fjz1             = _mm256_add_ps(fjz1,tz);
 967
 968             /**************************
 969              * CALCULATE INTERACTIONS *
 970              **************************/
 971
 972             /* COULOMB ELECTROSTATICS */
 973             velec            = _mm256_mul_ps(qq32,rinv32);
 974             felec            = _mm256_mul_ps(velec,rinvsq32);
 975
 976             /* Update potential sum for this i atom from the interaction with this j atom. */
 977             velec            = _mm256_andnot_ps(dummy_mask,velec);
 978             velecsum         = _mm256_add_ps(velecsum,velec);
 979
 980             fscal            = felec;
 981
 982             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 983
 984             /* Calculate temporary vectorial force */
 985             tx               = _mm256_mul_ps(fscal,dx32);
 986             ty               = _mm256_mul_ps(fscal,dy32);
 987             tz               = _mm256_mul_ps(fscal,dz32);
 988
 989             /* Update vectorial force */
 990             fix3             = _mm256_add_ps(fix3,tx);
 991             fiy3             = _mm256_add_ps(fiy3,ty);
 992             fiz3             = _mm256_add_ps(fiz3,tz);
 993
 994             fjx2             = _mm256_add_ps(fjx2,tx);
 995             fjy2             = _mm256_add_ps(fjy2,ty);
 996             fjz2             = _mm256_add_ps(fjz2,tz);
 997
 998             /**************************
 999              * CALCULATE INTERACTIONS *
1000              **************************/
1001
1002             /* COULOMB ELECTROSTATICS */
1003             velec            = _mm256_mul_ps(qq33,rinv33);
1004             felec            = _mm256_mul_ps(velec,rinvsq33);
1005
1006             /* Update potential sum for this i atom from the interaction with this j atom. */
1007             velec            = _mm256_andnot_ps(dummy_mask,velec);
1008             velecsum         = _mm256_add_ps(velecsum,velec);
1009
1010             fscal            = felec;
1011
1012             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1013
1014             /* Calculate temporary vectorial force */
1015             tx               = _mm256_mul_ps(fscal,dx33);
1016             ty               = _mm256_mul_ps(fscal,dy33);
1017             tz               = _mm256_mul_ps(fscal,dz33);
1018
1019             /* Update vectorial force */
1020             fix3             = _mm256_add_ps(fix3,tx);
1021             fiy3             = _mm256_add_ps(fiy3,ty);
1022             fiz3             = _mm256_add_ps(fiz3,tz);
1023
1024             fjx3             = _mm256_add_ps(fjx3,tx);
1025             fjy3             = _mm256_add_ps(fjy3,ty);
1026             fjz3             = _mm256_add_ps(fjz3,tz);
1027
1028             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1029             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1030             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1031             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1032             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1033             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1034             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1035             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1036
1037             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1038                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1039                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1040
1041             /* Inner loop uses 278 flops */
1042         }
1043
1044         /* End of innermost loop */
1045
1046         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1047                                                  f+i_coord_offset,fshift+i_shift_offset);
1048
1049         ggid                        = gid[iidx];
1050         /* Update potential energies */
1051         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1052         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1053
1054         /* Increment number of inner iterations */
1055         inneriter                  += j_index_end - j_index_start;
1056
1057         /* Outer loop uses 26 flops */
1058     }
1059
1060     /* Increment number of outer iterations */
1061     outeriter        += nri;
1062
1063     /* Update outer/inner flops */
1064
1065     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*278);
1066 }
1067 /*
1068  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_single
1069  * Electrostatics interaction: Coulomb
1070  * VdW interaction:            LennardJones
1071  * Geometry:                   Water4-Water4
1072  * Calculate force/pot:        Force
1073  */
1074 void
1075 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_single
1076                     (t_nblist * gmx_restrict                nlist,
1077                      rvec * gmx_restrict                    xx,
1078                      rvec * gmx_restrict                    ff,
1079                      t_forcerec * gmx_restrict              fr,
1080                      t_mdatoms * gmx_restrict               mdatoms,
1081                      nb_kernel_data_t * gmx_restrict        kernel_data,
1082                      t_nrnb * gmx_restrict                  nrnb)
1083 {
1084     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1085      * just 0 for non-waters.
1086      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1087      * jnr indices corresponding to data put in the four positions in the SIMD register.
1088      */
1089     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1090     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1091     int              jnrA,jnrB,jnrC,jnrD;
1092     int              jnrE,jnrF,jnrG,jnrH;
1093     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1094     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1095     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1096     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1097     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1098     real             rcutoff_scalar;
1099     real             *shiftvec,*fshift,*x,*f;
1100     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1101     real             scratch[4*DIM];
1102     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1103     real *           vdwioffsetptr0;
1104     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1105     real *           vdwioffsetptr1;
1106     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1107     real *           vdwioffsetptr2;
1108     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1109     real *           vdwioffsetptr3;
1110     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1111     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1112     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1113     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1114     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1115     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1116     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1117     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1118     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1119     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1120     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1121     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1122     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1123     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1124     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1125     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1126     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1127     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1128     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1129     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1130     real             *charge;
1131     int              nvdwtype;
1132     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1133     int              *vdwtype;
1134     real             *vdwparam;
1135     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1136     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1137     __m256           dummy_mask,cutoff_mask;
1138     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1139     __m256           one     = _mm256_set1_ps(1.0);
1140     __m256           two     = _mm256_set1_ps(2.0);
1141     x                = xx[0];
1142     f                = ff[0];
1143
1144     nri              = nlist->nri;
1145     iinr             = nlist->iinr;
1146     jindex           = nlist->jindex;
1147     jjnr             = nlist->jjnr;
1148     shiftidx         = nlist->shift;
1149     gid              = nlist->gid;
1150     shiftvec         = fr->shift_vec[0];
1151     fshift           = fr->fshift[0];
1152     facel            = _mm256_set1_ps(fr->epsfac);
1153     charge           = mdatoms->chargeA;
1154     nvdwtype         = fr->ntype;
1155     vdwparam         = fr->nbfp;
1156     vdwtype          = mdatoms->typeA;
1157
1158     /* Setup water-specific parameters */
1159     inr              = nlist->iinr[0];
1160     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1161     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1162     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1163     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1164
1165     jq1              = _mm256_set1_ps(charge[inr+1]);
1166     jq2              = _mm256_set1_ps(charge[inr+2]);
1167     jq3              = _mm256_set1_ps(charge[inr+3]);
1168     vdwjidx0A        = 2*vdwtype[inr+0];
1169     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1170     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1171     qq11             = _mm256_mul_ps(iq1,jq1);
1172     qq12             = _mm256_mul_ps(iq1,jq2);
1173     qq13             = _mm256_mul_ps(iq1,jq3);
1174     qq21             = _mm256_mul_ps(iq2,jq1);
1175     qq22             = _mm256_mul_ps(iq2,jq2);
1176     qq23             = _mm256_mul_ps(iq2,jq3);
1177     qq31             = _mm256_mul_ps(iq3,jq1);
1178     qq32             = _mm256_mul_ps(iq3,jq2);
1179     qq33             = _mm256_mul_ps(iq3,jq3);
1180
1181     /* Avoid stupid compiler warnings */
1182     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1183     j_coord_offsetA = 0;
1184     j_coord_offsetB = 0;
1185     j_coord_offsetC = 0;
1186     j_coord_offsetD = 0;
1187     j_coord_offsetE = 0;
1188     j_coord_offsetF = 0;
1189     j_coord_offsetG = 0;
1190     j_coord_offsetH = 0;
1191
1192     outeriter        = 0;
1193     inneriter        = 0;
1194
1195     for(iidx=0;iidx<4*DIM;iidx++)
1196     {
1197         scratch[iidx] = 0.0;
1198     }
1199
1200     /* Start outer loop over neighborlists */
1201     for(iidx=0; iidx<nri; iidx++)
1202     {
1203         /* Load shift vector for this list */
1204         i_shift_offset   = DIM*shiftidx[iidx];
1205
1206         /* Load limits for loop over neighbors */
1207         j_index_start    = jindex[iidx];
1208         j_index_end      = jindex[iidx+1];
1209
1210         /* Get outer coordinate index */
1211         inr              = iinr[iidx];
1212         i_coord_offset   = DIM*inr;
1213
1214         /* Load i particle coords and add shift vector */
1215         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1216                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1217
1218         fix0             = _mm256_setzero_ps();
1219         fiy0             = _mm256_setzero_ps();
1220         fiz0             = _mm256_setzero_ps();
1221         fix1             = _mm256_setzero_ps();
1222         fiy1             = _mm256_setzero_ps();
1223         fiz1             = _mm256_setzero_ps();
1224         fix2             = _mm256_setzero_ps();
1225         fiy2             = _mm256_setzero_ps();
1226         fiz2             = _mm256_setzero_ps();
1227         fix3             = _mm256_setzero_ps();
1228         fiy3             = _mm256_setzero_ps();
1229         fiz3             = _mm256_setzero_ps();
1230
1231         /* Start inner kernel loop */
1232         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1233         {
1234
1235             /* Get j neighbor index, and coordinate index */
1236             jnrA             = jjnr[jidx];
1237             jnrB             = jjnr[jidx+1];
1238             jnrC             = jjnr[jidx+2];
1239             jnrD             = jjnr[jidx+3];
1240             jnrE             = jjnr[jidx+4];
1241             jnrF             = jjnr[jidx+5];
1242             jnrG             = jjnr[jidx+6];
1243             jnrH             = jjnr[jidx+7];
1244             j_coord_offsetA  = DIM*jnrA;
1245             j_coord_offsetB  = DIM*jnrB;
1246             j_coord_offsetC  = DIM*jnrC;
1247             j_coord_offsetD  = DIM*jnrD;
1248             j_coord_offsetE  = DIM*jnrE;
1249             j_coord_offsetF  = DIM*jnrF;
1250             j_coord_offsetG  = DIM*jnrG;
1251             j_coord_offsetH  = DIM*jnrH;
1252
1253             /* load j atom coordinates */
1254             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1255                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1256                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1257                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1258                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1259                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1260
1261             /* Calculate displacement vector */
1262             dx00             = _mm256_sub_ps(ix0,jx0);
1263             dy00             = _mm256_sub_ps(iy0,jy0);
1264             dz00             = _mm256_sub_ps(iz0,jz0);
1265             dx11             = _mm256_sub_ps(ix1,jx1);
1266             dy11             = _mm256_sub_ps(iy1,jy1);
1267             dz11             = _mm256_sub_ps(iz1,jz1);
1268             dx12             = _mm256_sub_ps(ix1,jx2);
1269             dy12             = _mm256_sub_ps(iy1,jy2);
1270             dz12             = _mm256_sub_ps(iz1,jz2);
1271             dx13             = _mm256_sub_ps(ix1,jx3);
1272             dy13             = _mm256_sub_ps(iy1,jy3);
1273             dz13             = _mm256_sub_ps(iz1,jz3);
1274             dx21             = _mm256_sub_ps(ix2,jx1);
1275             dy21             = _mm256_sub_ps(iy2,jy1);
1276             dz21             = _mm256_sub_ps(iz2,jz1);
1277             dx22             = _mm256_sub_ps(ix2,jx2);
1278             dy22             = _mm256_sub_ps(iy2,jy2);
1279             dz22             = _mm256_sub_ps(iz2,jz2);
1280             dx23             = _mm256_sub_ps(ix2,jx3);
1281             dy23             = _mm256_sub_ps(iy2,jy3);
1282             dz23             = _mm256_sub_ps(iz2,jz3);
1283             dx31             = _mm256_sub_ps(ix3,jx1);
1284             dy31             = _mm256_sub_ps(iy3,jy1);
1285             dz31             = _mm256_sub_ps(iz3,jz1);
1286             dx32             = _mm256_sub_ps(ix3,jx2);
1287             dy32             = _mm256_sub_ps(iy3,jy2);
1288             dz32             = _mm256_sub_ps(iz3,jz2);
1289             dx33             = _mm256_sub_ps(ix3,jx3);
1290             dy33             = _mm256_sub_ps(iy3,jy3);
1291             dz33             = _mm256_sub_ps(iz3,jz3);
1292
1293             /* Calculate squared distance and things based on it */
1294             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1295             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1296             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1297             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1298             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1299             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1300             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1301             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1302             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1303             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1304
1305             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1306             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1307             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1308             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1309             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1310             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1311             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1312             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1313             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1314
1315             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1316             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1317             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1318             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1319             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1320             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1321             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1322             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1323             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1324             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1325
1326             fjx0             = _mm256_setzero_ps();
1327             fjy0             = _mm256_setzero_ps();
1328             fjz0             = _mm256_setzero_ps();
1329             fjx1             = _mm256_setzero_ps();
1330             fjy1             = _mm256_setzero_ps();
1331             fjz1             = _mm256_setzero_ps();
1332             fjx2             = _mm256_setzero_ps();
1333             fjy2             = _mm256_setzero_ps();
1334             fjz2             = _mm256_setzero_ps();
1335             fjx3             = _mm256_setzero_ps();
1336             fjy3             = _mm256_setzero_ps();
1337             fjz3             = _mm256_setzero_ps();
1338
1339             /**************************
1340              * CALCULATE INTERACTIONS *
1341              **************************/
1342
1343             /* LENNARD-JONES DISPERSION/REPULSION */
1344
1345             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1346             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1347
1348             fscal            = fvdw;
1349
1350             /* Calculate temporary vectorial force */
1351             tx               = _mm256_mul_ps(fscal,dx00);
1352             ty               = _mm256_mul_ps(fscal,dy00);
1353             tz               = _mm256_mul_ps(fscal,dz00);
1354
1355             /* Update vectorial force */
1356             fix0             = _mm256_add_ps(fix0,tx);
1357             fiy0             = _mm256_add_ps(fiy0,ty);
1358             fiz0             = _mm256_add_ps(fiz0,tz);
1359
1360             fjx0             = _mm256_add_ps(fjx0,tx);
1361             fjy0             = _mm256_add_ps(fjy0,ty);
1362             fjz0             = _mm256_add_ps(fjz0,tz);
1363
1364             /**************************
1365              * CALCULATE INTERACTIONS *
1366              **************************/
1367
1368             /* COULOMB ELECTROSTATICS */
1369             velec            = _mm256_mul_ps(qq11,rinv11);
1370             felec            = _mm256_mul_ps(velec,rinvsq11);
1371
1372             fscal            = felec;
1373
1374             /* Calculate temporary vectorial force */
1375             tx               = _mm256_mul_ps(fscal,dx11);
1376             ty               = _mm256_mul_ps(fscal,dy11);
1377             tz               = _mm256_mul_ps(fscal,dz11);
1378
1379             /* Update vectorial force */
1380             fix1             = _mm256_add_ps(fix1,tx);
1381             fiy1             = _mm256_add_ps(fiy1,ty);
1382             fiz1             = _mm256_add_ps(fiz1,tz);
1383
1384             fjx1             = _mm256_add_ps(fjx1,tx);
1385             fjy1             = _mm256_add_ps(fjy1,ty);
1386             fjz1             = _mm256_add_ps(fjz1,tz);
1387
1388             /**************************
1389              * CALCULATE INTERACTIONS *
1390              **************************/
1391
1392             /* COULOMB ELECTROSTATICS */
1393             velec            = _mm256_mul_ps(qq12,rinv12);
1394             felec            = _mm256_mul_ps(velec,rinvsq12);
1395
1396             fscal            = felec;
1397
1398             /* Calculate temporary vectorial force */
1399             tx               = _mm256_mul_ps(fscal,dx12);
1400             ty               = _mm256_mul_ps(fscal,dy12);
1401             tz               = _mm256_mul_ps(fscal,dz12);
1402
1403             /* Update vectorial force */
1404             fix1             = _mm256_add_ps(fix1,tx);
1405             fiy1             = _mm256_add_ps(fiy1,ty);
1406             fiz1             = _mm256_add_ps(fiz1,tz);
1407
1408             fjx2             = _mm256_add_ps(fjx2,tx);
1409             fjy2             = _mm256_add_ps(fjy2,ty);
1410             fjz2             = _mm256_add_ps(fjz2,tz);
1411
1412             /**************************
1413              * CALCULATE INTERACTIONS *
1414              **************************/
1415
1416             /* COULOMB ELECTROSTATICS */
1417             velec            = _mm256_mul_ps(qq13,rinv13);
1418             felec            = _mm256_mul_ps(velec,rinvsq13);
1419
1420             fscal            = felec;
1421
1422             /* Calculate temporary vectorial force */
1423             tx               = _mm256_mul_ps(fscal,dx13);
1424             ty               = _mm256_mul_ps(fscal,dy13);
1425             tz               = _mm256_mul_ps(fscal,dz13);
1426
1427             /* Update vectorial force */
1428             fix1             = _mm256_add_ps(fix1,tx);
1429             fiy1             = _mm256_add_ps(fiy1,ty);
1430             fiz1             = _mm256_add_ps(fiz1,tz);
1431
1432             fjx3             = _mm256_add_ps(fjx3,tx);
1433             fjy3             = _mm256_add_ps(fjy3,ty);
1434             fjz3             = _mm256_add_ps(fjz3,tz);
1435
1436             /**************************
1437              * CALCULATE INTERACTIONS *
1438              **************************/
1439
1440             /* COULOMB ELECTROSTATICS */
1441             velec            = _mm256_mul_ps(qq21,rinv21);
1442             felec            = _mm256_mul_ps(velec,rinvsq21);
1443
1444             fscal            = felec;
1445
1446             /* Calculate temporary vectorial force */
1447             tx               = _mm256_mul_ps(fscal,dx21);
1448             ty               = _mm256_mul_ps(fscal,dy21);
1449             tz               = _mm256_mul_ps(fscal,dz21);
1450
1451             /* Update vectorial force */
1452             fix2             = _mm256_add_ps(fix2,tx);
1453             fiy2             = _mm256_add_ps(fiy2,ty);
1454             fiz2             = _mm256_add_ps(fiz2,tz);
1455
1456             fjx1             = _mm256_add_ps(fjx1,tx);
1457             fjy1             = _mm256_add_ps(fjy1,ty);
1458             fjz1             = _mm256_add_ps(fjz1,tz);
1459
1460             /**************************
1461              * CALCULATE INTERACTIONS *
1462              **************************/
1463
1464             /* COULOMB ELECTROSTATICS */
1465             velec            = _mm256_mul_ps(qq22,rinv22);
1466             felec            = _mm256_mul_ps(velec,rinvsq22);
1467
1468             fscal            = felec;
1469
1470             /* Calculate temporary vectorial force */
1471             tx               = _mm256_mul_ps(fscal,dx22);
1472             ty               = _mm256_mul_ps(fscal,dy22);
1473             tz               = _mm256_mul_ps(fscal,dz22);
1474
1475             /* Update vectorial force */
1476             fix2             = _mm256_add_ps(fix2,tx);
1477             fiy2             = _mm256_add_ps(fiy2,ty);
1478             fiz2             = _mm256_add_ps(fiz2,tz);
1479
1480             fjx2             = _mm256_add_ps(fjx2,tx);
1481             fjy2             = _mm256_add_ps(fjy2,ty);
1482             fjz2             = _mm256_add_ps(fjz2,tz);
1483
1484             /**************************
1485              * CALCULATE INTERACTIONS *
1486              **************************/
1487
1488             /* COULOMB ELECTROSTATICS */
1489             velec            = _mm256_mul_ps(qq23,rinv23);
1490             felec            = _mm256_mul_ps(velec,rinvsq23);
1491
1492             fscal            = felec;
1493
1494             /* Calculate temporary vectorial force */
1495             tx               = _mm256_mul_ps(fscal,dx23);
1496             ty               = _mm256_mul_ps(fscal,dy23);
1497             tz               = _mm256_mul_ps(fscal,dz23);
1498
1499             /* Update vectorial force */
1500             fix2             = _mm256_add_ps(fix2,tx);
1501             fiy2             = _mm256_add_ps(fiy2,ty);
1502             fiz2             = _mm256_add_ps(fiz2,tz);
1503
1504             fjx3             = _mm256_add_ps(fjx3,tx);
1505             fjy3             = _mm256_add_ps(fjy3,ty);
1506             fjz3             = _mm256_add_ps(fjz3,tz);
1507
1508             /**************************
1509              * CALCULATE INTERACTIONS *
1510              **************************/
1511
1512             /* COULOMB ELECTROSTATICS */
1513             velec            = _mm256_mul_ps(qq31,rinv31);
1514             felec            = _mm256_mul_ps(velec,rinvsq31);
1515
1516             fscal            = felec;
1517
1518             /* Calculate temporary vectorial force */
1519             tx               = _mm256_mul_ps(fscal,dx31);
1520             ty               = _mm256_mul_ps(fscal,dy31);
1521             tz               = _mm256_mul_ps(fscal,dz31);
1522
1523             /* Update vectorial force */
1524             fix3             = _mm256_add_ps(fix3,tx);
1525             fiy3             = _mm256_add_ps(fiy3,ty);
1526             fiz3             = _mm256_add_ps(fiz3,tz);
1527
1528             fjx1             = _mm256_add_ps(fjx1,tx);
1529             fjy1             = _mm256_add_ps(fjy1,ty);
1530             fjz1             = _mm256_add_ps(fjz1,tz);
1531
1532             /**************************
1533              * CALCULATE INTERACTIONS *
1534              **************************/
1535
1536             /* COULOMB ELECTROSTATICS */
1537             velec            = _mm256_mul_ps(qq32,rinv32);
1538             felec            = _mm256_mul_ps(velec,rinvsq32);
1539
1540             fscal            = felec;
1541
1542             /* Calculate temporary vectorial force */
1543             tx               = _mm256_mul_ps(fscal,dx32);
1544             ty               = _mm256_mul_ps(fscal,dy32);
1545             tz               = _mm256_mul_ps(fscal,dz32);
1546
1547             /* Update vectorial force */
1548             fix3             = _mm256_add_ps(fix3,tx);
1549             fiy3             = _mm256_add_ps(fiy3,ty);
1550             fiz3             = _mm256_add_ps(fiz3,tz);
1551
1552             fjx2             = _mm256_add_ps(fjx2,tx);
1553             fjy2             = _mm256_add_ps(fjy2,ty);
1554             fjz2             = _mm256_add_ps(fjz2,tz);
1555
1556             /**************************
1557              * CALCULATE INTERACTIONS *
1558              **************************/
1559
1560             /* COULOMB ELECTROSTATICS */
1561             velec            = _mm256_mul_ps(qq33,rinv33);
1562             felec            = _mm256_mul_ps(velec,rinvsq33);
1563
1564             fscal            = felec;
1565
1566             /* Calculate temporary vectorial force */
1567             tx               = _mm256_mul_ps(fscal,dx33);
1568             ty               = _mm256_mul_ps(fscal,dy33);
1569             tz               = _mm256_mul_ps(fscal,dz33);
1570
1571             /* Update vectorial force */
1572             fix3             = _mm256_add_ps(fix3,tx);
1573             fiy3             = _mm256_add_ps(fiy3,ty);
1574             fiz3             = _mm256_add_ps(fiz3,tz);
1575
1576             fjx3             = _mm256_add_ps(fjx3,tx);
1577             fjy3             = _mm256_add_ps(fjy3,ty);
1578             fjz3             = _mm256_add_ps(fjz3,tz);
1579
1580             fjptrA             = f+j_coord_offsetA;
1581             fjptrB             = f+j_coord_offsetB;
1582             fjptrC             = f+j_coord_offsetC;
1583             fjptrD             = f+j_coord_offsetD;
1584             fjptrE             = f+j_coord_offsetE;
1585             fjptrF             = f+j_coord_offsetF;
1586             fjptrG             = f+j_coord_offsetG;
1587             fjptrH             = f+j_coord_offsetH;
1588
1589             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1590                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1591                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1592
1593             /* Inner loop uses 264 flops */
1594         }
1595
1596         if(jidx<j_index_end)
1597         {
1598
1599             /* Get j neighbor index, and coordinate index */
1600             jnrlistA         = jjnr[jidx];
1601             jnrlistB         = jjnr[jidx+1];
1602             jnrlistC         = jjnr[jidx+2];
1603             jnrlistD         = jjnr[jidx+3];
1604             jnrlistE         = jjnr[jidx+4];
1605             jnrlistF         = jjnr[jidx+5];
1606             jnrlistG         = jjnr[jidx+6];
1607             jnrlistH         = jjnr[jidx+7];
1608             /* Sign of each element will be negative for non-real atoms.
1609              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1610              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1611              */
1612             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1613                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1614
1615             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1616             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1617             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1618             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1619             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1620             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1621             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1622             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1623             j_coord_offsetA  = DIM*jnrA;
1624             j_coord_offsetB  = DIM*jnrB;
1625             j_coord_offsetC  = DIM*jnrC;
1626             j_coord_offsetD  = DIM*jnrD;
1627             j_coord_offsetE  = DIM*jnrE;
1628             j_coord_offsetF  = DIM*jnrF;
1629             j_coord_offsetG  = DIM*jnrG;
1630             j_coord_offsetH  = DIM*jnrH;
1631
1632             /* load j atom coordinates */
1633             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1634                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1635                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1636                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1637                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1638                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1639
1640             /* Calculate displacement vector */
1641             dx00             = _mm256_sub_ps(ix0,jx0);
1642             dy00             = _mm256_sub_ps(iy0,jy0);
1643             dz00             = _mm256_sub_ps(iz0,jz0);
1644             dx11             = _mm256_sub_ps(ix1,jx1);
1645             dy11             = _mm256_sub_ps(iy1,jy1);
1646             dz11             = _mm256_sub_ps(iz1,jz1);
1647             dx12             = _mm256_sub_ps(ix1,jx2);
1648             dy12             = _mm256_sub_ps(iy1,jy2);
1649             dz12             = _mm256_sub_ps(iz1,jz2);
1650             dx13             = _mm256_sub_ps(ix1,jx3);
1651             dy13             = _mm256_sub_ps(iy1,jy3);
1652             dz13             = _mm256_sub_ps(iz1,jz3);
1653             dx21             = _mm256_sub_ps(ix2,jx1);
1654             dy21             = _mm256_sub_ps(iy2,jy1);
1655             dz21             = _mm256_sub_ps(iz2,jz1);
1656             dx22             = _mm256_sub_ps(ix2,jx2);
1657             dy22             = _mm256_sub_ps(iy2,jy2);
1658             dz22             = _mm256_sub_ps(iz2,jz2);
1659             dx23             = _mm256_sub_ps(ix2,jx3);
1660             dy23             = _mm256_sub_ps(iy2,jy3);
1661             dz23             = _mm256_sub_ps(iz2,jz3);
1662             dx31             = _mm256_sub_ps(ix3,jx1);
1663             dy31             = _mm256_sub_ps(iy3,jy1);
1664             dz31             = _mm256_sub_ps(iz3,jz1);
1665             dx32             = _mm256_sub_ps(ix3,jx2);
1666             dy32             = _mm256_sub_ps(iy3,jy2);
1667             dz32             = _mm256_sub_ps(iz3,jz2);
1668             dx33             = _mm256_sub_ps(ix3,jx3);
1669             dy33             = _mm256_sub_ps(iy3,jy3);
1670             dz33             = _mm256_sub_ps(iz3,jz3);
1671
1672             /* Calculate squared distance and things based on it */
1673             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1674             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1675             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1676             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1677             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1678             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1679             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1680             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1681             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1682             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1683
1684             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1685             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1686             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1687             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1688             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1689             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1690             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1691             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1692             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1693
1694             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1695             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1696             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1697             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1698             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1699             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1700             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1701             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1702             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1703             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1704
1705             fjx0             = _mm256_setzero_ps();
1706             fjy0             = _mm256_setzero_ps();
1707             fjz0             = _mm256_setzero_ps();
1708             fjx1             = _mm256_setzero_ps();
1709             fjy1             = _mm256_setzero_ps();
1710             fjz1             = _mm256_setzero_ps();
1711             fjx2             = _mm256_setzero_ps();
1712             fjy2             = _mm256_setzero_ps();
1713             fjz2             = _mm256_setzero_ps();
1714             fjx3             = _mm256_setzero_ps();
1715             fjy3             = _mm256_setzero_ps();
1716             fjz3             = _mm256_setzero_ps();
1717
1718             /**************************
1719              * CALCULATE INTERACTIONS *
1720              **************************/
1721
1722             /* LENNARD-JONES DISPERSION/REPULSION */
1723
1724             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1725             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1726
1727             fscal            = fvdw;
1728
1729             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1730
1731             /* Calculate temporary vectorial force */
1732             tx               = _mm256_mul_ps(fscal,dx00);
1733             ty               = _mm256_mul_ps(fscal,dy00);
1734             tz               = _mm256_mul_ps(fscal,dz00);
1735
1736             /* Update vectorial force */
1737             fix0             = _mm256_add_ps(fix0,tx);
1738             fiy0             = _mm256_add_ps(fiy0,ty);
1739             fiz0             = _mm256_add_ps(fiz0,tz);
1740
1741             fjx0             = _mm256_add_ps(fjx0,tx);
1742             fjy0             = _mm256_add_ps(fjy0,ty);
1743             fjz0             = _mm256_add_ps(fjz0,tz);
1744
1745             /**************************
1746              * CALCULATE INTERACTIONS *
1747              **************************/
1748
1749             /* COULOMB ELECTROSTATICS */
1750             velec            = _mm256_mul_ps(qq11,rinv11);
1751             felec            = _mm256_mul_ps(velec,rinvsq11);
1752
1753             fscal            = felec;
1754
1755             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1756
1757             /* Calculate temporary vectorial force */
1758             tx               = _mm256_mul_ps(fscal,dx11);
1759             ty               = _mm256_mul_ps(fscal,dy11);
1760             tz               = _mm256_mul_ps(fscal,dz11);
1761
1762             /* Update vectorial force */
1763             fix1             = _mm256_add_ps(fix1,tx);
1764             fiy1             = _mm256_add_ps(fiy1,ty);
1765             fiz1             = _mm256_add_ps(fiz1,tz);
1766
1767             fjx1             = _mm256_add_ps(fjx1,tx);
1768             fjy1             = _mm256_add_ps(fjy1,ty);
1769             fjz1             = _mm256_add_ps(fjz1,tz);
1770
1771             /**************************
1772              * CALCULATE INTERACTIONS *
1773              **************************/
1774
1775             /* COULOMB ELECTROSTATICS */
1776             velec            = _mm256_mul_ps(qq12,rinv12);
1777             felec            = _mm256_mul_ps(velec,rinvsq12);
1778
1779             fscal            = felec;
1780
1781             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1782
1783             /* Calculate temporary vectorial force */
1784             tx               = _mm256_mul_ps(fscal,dx12);
1785             ty               = _mm256_mul_ps(fscal,dy12);
1786             tz               = _mm256_mul_ps(fscal,dz12);
1787
1788             /* Update vectorial force */
1789             fix1             = _mm256_add_ps(fix1,tx);
1790             fiy1             = _mm256_add_ps(fiy1,ty);
1791             fiz1             = _mm256_add_ps(fiz1,tz);
1792
1793             fjx2             = _mm256_add_ps(fjx2,tx);
1794             fjy2             = _mm256_add_ps(fjy2,ty);
1795             fjz2             = _mm256_add_ps(fjz2,tz);
1796
1797             /**************************
1798              * CALCULATE INTERACTIONS *
1799              **************************/
1800
1801             /* COULOMB ELECTROSTATICS */
1802             velec            = _mm256_mul_ps(qq13,rinv13);
1803             felec            = _mm256_mul_ps(velec,rinvsq13);
1804
1805             fscal            = felec;
1806
1807             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1808
1809             /* Calculate temporary vectorial force */
1810             tx               = _mm256_mul_ps(fscal,dx13);
1811             ty               = _mm256_mul_ps(fscal,dy13);
1812             tz               = _mm256_mul_ps(fscal,dz13);
1813
1814             /* Update vectorial force */
1815             fix1             = _mm256_add_ps(fix1,tx);
1816             fiy1             = _mm256_add_ps(fiy1,ty);
1817             fiz1             = _mm256_add_ps(fiz1,tz);
1818
1819             fjx3             = _mm256_add_ps(fjx3,tx);
1820             fjy3             = _mm256_add_ps(fjy3,ty);
1821             fjz3             = _mm256_add_ps(fjz3,tz);
1822
1823             /**************************
1824              * CALCULATE INTERACTIONS *
1825              **************************/
1826
1827             /* COULOMB ELECTROSTATICS */
1828             velec            = _mm256_mul_ps(qq21,rinv21);
1829             felec            = _mm256_mul_ps(velec,rinvsq21);
1830
1831             fscal            = felec;
1832
1833             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1834
1835             /* Calculate temporary vectorial force */
1836             tx               = _mm256_mul_ps(fscal,dx21);
1837             ty               = _mm256_mul_ps(fscal,dy21);
1838             tz               = _mm256_mul_ps(fscal,dz21);
1839
1840             /* Update vectorial force */
1841             fix2             = _mm256_add_ps(fix2,tx);
1842             fiy2             = _mm256_add_ps(fiy2,ty);
1843             fiz2             = _mm256_add_ps(fiz2,tz);
1844
1845             fjx1             = _mm256_add_ps(fjx1,tx);
1846             fjy1             = _mm256_add_ps(fjy1,ty);
1847             fjz1             = _mm256_add_ps(fjz1,tz);
1848
1849             /**************************
1850              * CALCULATE INTERACTIONS *
1851              **************************/
1852
1853             /* COULOMB ELECTROSTATICS */
1854             velec            = _mm256_mul_ps(qq22,rinv22);
1855             felec            = _mm256_mul_ps(velec,rinvsq22);
1856
1857             fscal            = felec;
1858
1859             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1860
1861             /* Calculate temporary vectorial force */
1862             tx               = _mm256_mul_ps(fscal,dx22);
1863             ty               = _mm256_mul_ps(fscal,dy22);
1864             tz               = _mm256_mul_ps(fscal,dz22);
1865
1866             /* Update vectorial force */
1867             fix2             = _mm256_add_ps(fix2,tx);
1868             fiy2             = _mm256_add_ps(fiy2,ty);
1869             fiz2             = _mm256_add_ps(fiz2,tz);
1870
1871             fjx2             = _mm256_add_ps(fjx2,tx);
1872             fjy2             = _mm256_add_ps(fjy2,ty);
1873             fjz2             = _mm256_add_ps(fjz2,tz);
1874
1875             /**************************
1876              * CALCULATE INTERACTIONS *
1877              **************************/
1878
1879             /* COULOMB ELECTROSTATICS */
1880             velec            = _mm256_mul_ps(qq23,rinv23);
1881             felec            = _mm256_mul_ps(velec,rinvsq23);
1882
1883             fscal            = felec;
1884
1885             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1886
1887             /* Calculate temporary vectorial force */
1888             tx               = _mm256_mul_ps(fscal,dx23);
1889             ty               = _mm256_mul_ps(fscal,dy23);
1890             tz               = _mm256_mul_ps(fscal,dz23);
1891
1892             /* Update vectorial force */
1893             fix2             = _mm256_add_ps(fix2,tx);
1894             fiy2             = _mm256_add_ps(fiy2,ty);
1895             fiz2             = _mm256_add_ps(fiz2,tz);
1896
1897             fjx3             = _mm256_add_ps(fjx3,tx);
1898             fjy3             = _mm256_add_ps(fjy3,ty);
1899             fjz3             = _mm256_add_ps(fjz3,tz);
1900
1901             /**************************
1902              * CALCULATE INTERACTIONS *
1903              **************************/
1904
1905             /* COULOMB ELECTROSTATICS */
1906             velec            = _mm256_mul_ps(qq31,rinv31);
1907             felec            = _mm256_mul_ps(velec,rinvsq31);
1908
1909             fscal            = felec;
1910
1911             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1912
1913             /* Calculate temporary vectorial force */
1914             tx               = _mm256_mul_ps(fscal,dx31);
1915             ty               = _mm256_mul_ps(fscal,dy31);
1916             tz               = _mm256_mul_ps(fscal,dz31);
1917
1918             /* Update vectorial force */
1919             fix3             = _mm256_add_ps(fix3,tx);
1920             fiy3             = _mm256_add_ps(fiy3,ty);
1921             fiz3             = _mm256_add_ps(fiz3,tz);
1922
1923             fjx1             = _mm256_add_ps(fjx1,tx);
1924             fjy1             = _mm256_add_ps(fjy1,ty);
1925             fjz1             = _mm256_add_ps(fjz1,tz);
1926
1927             /**************************
1928              * CALCULATE INTERACTIONS *
1929              **************************/
1930
1931             /* COULOMB ELECTROSTATICS */
1932             velec            = _mm256_mul_ps(qq32,rinv32);
1933             felec            = _mm256_mul_ps(velec,rinvsq32);
1934
1935             fscal            = felec;
1936
1937             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1938
1939             /* Calculate temporary vectorial force */
1940             tx               = _mm256_mul_ps(fscal,dx32);
1941             ty               = _mm256_mul_ps(fscal,dy32);
1942             tz               = _mm256_mul_ps(fscal,dz32);
1943
1944             /* Update vectorial force */
1945             fix3             = _mm256_add_ps(fix3,tx);
1946             fiy3             = _mm256_add_ps(fiy3,ty);
1947             fiz3             = _mm256_add_ps(fiz3,tz);
1948
1949             fjx2             = _mm256_add_ps(fjx2,tx);
1950             fjy2             = _mm256_add_ps(fjy2,ty);
1951             fjz2             = _mm256_add_ps(fjz2,tz);
1952
1953             /**************************
1954              * CALCULATE INTERACTIONS *
1955              **************************/
1956
1957             /* COULOMB ELECTROSTATICS */
1958             velec            = _mm256_mul_ps(qq33,rinv33);
1959             felec            = _mm256_mul_ps(velec,rinvsq33);
1960
1961             fscal            = felec;
1962
1963             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1964
1965             /* Calculate temporary vectorial force */
1966             tx               = _mm256_mul_ps(fscal,dx33);
1967             ty               = _mm256_mul_ps(fscal,dy33);
1968             tz               = _mm256_mul_ps(fscal,dz33);
1969
1970             /* Update vectorial force */
1971             fix3             = _mm256_add_ps(fix3,tx);
1972             fiy3             = _mm256_add_ps(fiy3,ty);
1973             fiz3             = _mm256_add_ps(fiz3,tz);
1974
1975             fjx3             = _mm256_add_ps(fjx3,tx);
1976             fjy3             = _mm256_add_ps(fjy3,ty);
1977             fjz3             = _mm256_add_ps(fjz3,tz);
1978
1979             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1980             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1981             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1982             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1983             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1984             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1985             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1986             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1987
1988             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1989                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1990                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1991
1992             /* Inner loop uses 264 flops */
1993         }
1994
1995         /* End of innermost loop */
1996
1997         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1998                                                  f+i_coord_offset,fshift+i_shift_offset);
1999
2000         /* Increment number of inner iterations */
2001         inneriter                  += j_index_end - j_index_start;
2002
2003         /* Outer loop uses 24 flops */
2004     }
2005
2006     /* Increment number of outer iterations */
2007     outeriter        += nri;
2008
2009     /* Update outer/inner flops */
2010
2011     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*264);
2012 }