src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRF_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_ElecRF_VdwLJ_GeomW4W4_VF_avx_256_single
  38  * Electrostatics interaction: ReactionField
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecRF_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     krf              = _mm256_set1_ps(fr->ic->k_rf);
 124     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 125     crf              = _mm256_set1_ps(fr->ic->c_rf);
 126     nvdwtype         = fr->ntype;
 127     vdwparam         = fr->nbfp;
 128     vdwtype          = mdatoms->typeA;
 129
 130     /* Setup water-specific parameters */
 131     inr              = nlist->iinr[0];
 132     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 133     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 134     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 135     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 136
 137     jq1              = _mm256_set1_ps(charge[inr+1]);
 138     jq2              = _mm256_set1_ps(charge[inr+2]);
 139     jq3              = _mm256_set1_ps(charge[inr+3]);
 140     vdwjidx0A        = 2*vdwtype[inr+0];
 141     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 142     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 143     qq11             = _mm256_mul_ps(iq1,jq1);
 144     qq12             = _mm256_mul_ps(iq1,jq2);
 145     qq13             = _mm256_mul_ps(iq1,jq3);
 146     qq21             = _mm256_mul_ps(iq2,jq1);
 147     qq22             = _mm256_mul_ps(iq2,jq2);
 148     qq23             = _mm256_mul_ps(iq2,jq3);
 149     qq31             = _mm256_mul_ps(iq3,jq1);
 150     qq32             = _mm256_mul_ps(iq3,jq2);
 151     qq33             = _mm256_mul_ps(iq3,jq3);
 152
 153     /* Avoid stupid compiler warnings */
 154     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 155     j_coord_offsetA = 0;
 156     j_coord_offsetB = 0;
 157     j_coord_offsetC = 0;
 158     j_coord_offsetD = 0;
 159     j_coord_offsetE = 0;
 160     j_coord_offsetF = 0;
 161     j_coord_offsetG = 0;
 162     j_coord_offsetH = 0;
 163
 164     outeriter        = 0;
 165     inneriter        = 0;
 166
 167     for(iidx=0;iidx<4*DIM;iidx++)
 168     {
 169         scratch[iidx] = 0.0;
 170     }
 171
 172     /* Start outer loop over neighborlists */
 173     for(iidx=0; iidx<nri; iidx++)
 174     {
 175         /* Load shift vector for this list */
 176         i_shift_offset   = DIM*shiftidx[iidx];
 177
 178         /* Load limits for loop over neighbors */
 179         j_index_start    = jindex[iidx];
 180         j_index_end      = jindex[iidx+1];
 181
 182         /* Get outer coordinate index */
 183         inr              = iinr[iidx];
 184         i_coord_offset   = DIM*inr;
 185
 186         /* Load i particle coords and add shift vector */
 187         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 188                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 189
 190         fix0             = _mm256_setzero_ps();
 191         fiy0             = _mm256_setzero_ps();
 192         fiz0             = _mm256_setzero_ps();
 193         fix1             = _mm256_setzero_ps();
 194         fiy1             = _mm256_setzero_ps();
 195         fiz1             = _mm256_setzero_ps();
 196         fix2             = _mm256_setzero_ps();
 197         fiy2             = _mm256_setzero_ps();
 198         fiz2             = _mm256_setzero_ps();
 199         fix3             = _mm256_setzero_ps();
 200         fiy3             = _mm256_setzero_ps();
 201         fiz3             = _mm256_setzero_ps();
 202
 203         /* Reset potential sums */
 204         velecsum         = _mm256_setzero_ps();
 205         vvdwsum          = _mm256_setzero_ps();
 206
 207         /* Start inner kernel loop */
 208         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 209         {
 210
 211             /* Get j neighbor index, and coordinate index */
 212             jnrA             = jjnr[jidx];
 213             jnrB             = jjnr[jidx+1];
 214             jnrC             = jjnr[jidx+2];
 215             jnrD             = jjnr[jidx+3];
 216             jnrE             = jjnr[jidx+4];
 217             jnrF             = jjnr[jidx+5];
 218             jnrG             = jjnr[jidx+6];
 219             jnrH             = jjnr[jidx+7];
 220             j_coord_offsetA  = DIM*jnrA;
 221             j_coord_offsetB  = DIM*jnrB;
 222             j_coord_offsetC  = DIM*jnrC;
 223             j_coord_offsetD  = DIM*jnrD;
 224             j_coord_offsetE  = DIM*jnrE;
 225             j_coord_offsetF  = DIM*jnrF;
 226             j_coord_offsetG  = DIM*jnrG;
 227             j_coord_offsetH  = DIM*jnrH;
 228
 229             /* load j atom coordinates */
 230             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 231                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 232                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 233                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 234                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 235                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 236
 237             /* Calculate displacement vector */
 238             dx00             = _mm256_sub_ps(ix0,jx0);
 239             dy00             = _mm256_sub_ps(iy0,jy0);
 240             dz00             = _mm256_sub_ps(iz0,jz0);
 241             dx11             = _mm256_sub_ps(ix1,jx1);
 242             dy11             = _mm256_sub_ps(iy1,jy1);
 243             dz11             = _mm256_sub_ps(iz1,jz1);
 244             dx12             = _mm256_sub_ps(ix1,jx2);
 245             dy12             = _mm256_sub_ps(iy1,jy2);
 246             dz12             = _mm256_sub_ps(iz1,jz2);
 247             dx13             = _mm256_sub_ps(ix1,jx3);
 248             dy13             = _mm256_sub_ps(iy1,jy3);
 249             dz13             = _mm256_sub_ps(iz1,jz3);
 250             dx21             = _mm256_sub_ps(ix2,jx1);
 251             dy21             = _mm256_sub_ps(iy2,jy1);
 252             dz21             = _mm256_sub_ps(iz2,jz1);
 253             dx22             = _mm256_sub_ps(ix2,jx2);
 254             dy22             = _mm256_sub_ps(iy2,jy2);
 255             dz22             = _mm256_sub_ps(iz2,jz2);
 256             dx23             = _mm256_sub_ps(ix2,jx3);
 257             dy23             = _mm256_sub_ps(iy2,jy3);
 258             dz23             = _mm256_sub_ps(iz2,jz3);
 259             dx31             = _mm256_sub_ps(ix3,jx1);
 260             dy31             = _mm256_sub_ps(iy3,jy1);
 261             dz31             = _mm256_sub_ps(iz3,jz1);
 262             dx32             = _mm256_sub_ps(ix3,jx2);
 263             dy32             = _mm256_sub_ps(iy3,jy2);
 264             dz32             = _mm256_sub_ps(iz3,jz2);
 265             dx33             = _mm256_sub_ps(ix3,jx3);
 266             dy33             = _mm256_sub_ps(iy3,jy3);
 267             dz33             = _mm256_sub_ps(iz3,jz3);
 268
 269             /* Calculate squared distance and things based on it */
 270             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 271             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 272             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 273             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 274             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 275             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 276             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 277             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 278             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 279             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 280
 281             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 282             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 283             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 284             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 285             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 286             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 287             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 288             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 289             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 290
 291             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 292             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 293             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 294             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 295             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 296             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 297             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 298             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 299             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 300             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 301
 302             fjx0             = _mm256_setzero_ps();
 303             fjy0             = _mm256_setzero_ps();
 304             fjz0             = _mm256_setzero_ps();
 305             fjx1             = _mm256_setzero_ps();
 306             fjy1             = _mm256_setzero_ps();
 307             fjz1             = _mm256_setzero_ps();
 308             fjx2             = _mm256_setzero_ps();
 309             fjy2             = _mm256_setzero_ps();
 310             fjz2             = _mm256_setzero_ps();
 311             fjx3             = _mm256_setzero_ps();
 312             fjy3             = _mm256_setzero_ps();
 313             fjz3             = _mm256_setzero_ps();
 314
 315             /**************************
 316              * CALCULATE INTERACTIONS *
 317              **************************/
 318
 319             /* LENNARD-JONES DISPERSION/REPULSION */
 320
 321             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 322             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 323             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 324             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 325             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 326
 327             /* Update potential sum for this i atom from the interaction with this j atom. */
 328             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 329
 330             fscal            = fvdw;
 331
 332             /* Calculate temporary vectorial force */
 333             tx               = _mm256_mul_ps(fscal,dx00);
 334             ty               = _mm256_mul_ps(fscal,dy00);
 335             tz               = _mm256_mul_ps(fscal,dz00);
 336
 337             /* Update vectorial force */
 338             fix0             = _mm256_add_ps(fix0,tx);
 339             fiy0             = _mm256_add_ps(fiy0,ty);
 340             fiz0             = _mm256_add_ps(fiz0,tz);
 341
 342             fjx0             = _mm256_add_ps(fjx0,tx);
 343             fjy0             = _mm256_add_ps(fjy0,ty);
 344             fjz0             = _mm256_add_ps(fjz0,tz);
 345
 346             /**************************
 347              * CALCULATE INTERACTIONS *
 348              **************************/
 349
 350             /* REACTION-FIELD ELECTROSTATICS */
 351             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 352             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 353
 354             /* Update potential sum for this i atom from the interaction with this j atom. */
 355             velecsum         = _mm256_add_ps(velecsum,velec);
 356
 357             fscal            = felec;
 358
 359             /* Calculate temporary vectorial force */
 360             tx               = _mm256_mul_ps(fscal,dx11);
 361             ty               = _mm256_mul_ps(fscal,dy11);
 362             tz               = _mm256_mul_ps(fscal,dz11);
 363
 364             /* Update vectorial force */
 365             fix1             = _mm256_add_ps(fix1,tx);
 366             fiy1             = _mm256_add_ps(fiy1,ty);
 367             fiz1             = _mm256_add_ps(fiz1,tz);
 368
 369             fjx1             = _mm256_add_ps(fjx1,tx);
 370             fjy1             = _mm256_add_ps(fjy1,ty);
 371             fjz1             = _mm256_add_ps(fjz1,tz);
 372
 373             /**************************
 374              * CALCULATE INTERACTIONS *
 375              **************************/
 376
 377             /* REACTION-FIELD ELECTROSTATICS */
 378             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 379             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 380
 381             /* Update potential sum for this i atom from the interaction with this j atom. */
 382             velecsum         = _mm256_add_ps(velecsum,velec);
 383
 384             fscal            = felec;
 385
 386             /* Calculate temporary vectorial force */
 387             tx               = _mm256_mul_ps(fscal,dx12);
 388             ty               = _mm256_mul_ps(fscal,dy12);
 389             tz               = _mm256_mul_ps(fscal,dz12);
 390
 391             /* Update vectorial force */
 392             fix1             = _mm256_add_ps(fix1,tx);
 393             fiy1             = _mm256_add_ps(fiy1,ty);
 394             fiz1             = _mm256_add_ps(fiz1,tz);
 395
 396             fjx2             = _mm256_add_ps(fjx2,tx);
 397             fjy2             = _mm256_add_ps(fjy2,ty);
 398             fjz2             = _mm256_add_ps(fjz2,tz);
 399
 400             /**************************
 401              * CALCULATE INTERACTIONS *
 402              **************************/
 403
 404             /* REACTION-FIELD ELECTROSTATICS */
 405             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 406             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 407
 408             /* Update potential sum for this i atom from the interaction with this j atom. */
 409             velecsum         = _mm256_add_ps(velecsum,velec);
 410
 411             fscal            = felec;
 412
 413             /* Calculate temporary vectorial force */
 414             tx               = _mm256_mul_ps(fscal,dx13);
 415             ty               = _mm256_mul_ps(fscal,dy13);
 416             tz               = _mm256_mul_ps(fscal,dz13);
 417
 418             /* Update vectorial force */
 419             fix1             = _mm256_add_ps(fix1,tx);
 420             fiy1             = _mm256_add_ps(fiy1,ty);
 421             fiz1             = _mm256_add_ps(fiz1,tz);
 422
 423             fjx3             = _mm256_add_ps(fjx3,tx);
 424             fjy3             = _mm256_add_ps(fjy3,ty);
 425             fjz3             = _mm256_add_ps(fjz3,tz);
 426
 427             /**************************
 428              * CALCULATE INTERACTIONS *
 429              **************************/
 430
 431             /* REACTION-FIELD ELECTROSTATICS */
 432             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 433             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 434
 435             /* Update potential sum for this i atom from the interaction with this j atom. */
 436             velecsum         = _mm256_add_ps(velecsum,velec);
 437
 438             fscal            = felec;
 439
 440             /* Calculate temporary vectorial force */
 441             tx               = _mm256_mul_ps(fscal,dx21);
 442             ty               = _mm256_mul_ps(fscal,dy21);
 443             tz               = _mm256_mul_ps(fscal,dz21);
 444
 445             /* Update vectorial force */
 446             fix2             = _mm256_add_ps(fix2,tx);
 447             fiy2             = _mm256_add_ps(fiy2,ty);
 448             fiz2             = _mm256_add_ps(fiz2,tz);
 449
 450             fjx1             = _mm256_add_ps(fjx1,tx);
 451             fjy1             = _mm256_add_ps(fjy1,ty);
 452             fjz1             = _mm256_add_ps(fjz1,tz);
 453
 454             /**************************
 455              * CALCULATE INTERACTIONS *
 456              **************************/
 457
 458             /* REACTION-FIELD ELECTROSTATICS */
 459             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 460             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 461
 462             /* Update potential sum for this i atom from the interaction with this j atom. */
 463             velecsum         = _mm256_add_ps(velecsum,velec);
 464
 465             fscal            = felec;
 466
 467             /* Calculate temporary vectorial force */
 468             tx               = _mm256_mul_ps(fscal,dx22);
 469             ty               = _mm256_mul_ps(fscal,dy22);
 470             tz               = _mm256_mul_ps(fscal,dz22);
 471
 472             /* Update vectorial force */
 473             fix2             = _mm256_add_ps(fix2,tx);
 474             fiy2             = _mm256_add_ps(fiy2,ty);
 475             fiz2             = _mm256_add_ps(fiz2,tz);
 476
 477             fjx2             = _mm256_add_ps(fjx2,tx);
 478             fjy2             = _mm256_add_ps(fjy2,ty);
 479             fjz2             = _mm256_add_ps(fjz2,tz);
 480
 481             /**************************
 482              * CALCULATE INTERACTIONS *
 483              **************************/
 484
 485             /* REACTION-FIELD ELECTROSTATICS */
 486             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 487             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 488
 489             /* Update potential sum for this i atom from the interaction with this j atom. */
 490             velecsum         = _mm256_add_ps(velecsum,velec);
 491
 492             fscal            = felec;
 493
 494             /* Calculate temporary vectorial force */
 495             tx               = _mm256_mul_ps(fscal,dx23);
 496             ty               = _mm256_mul_ps(fscal,dy23);
 497             tz               = _mm256_mul_ps(fscal,dz23);
 498
 499             /* Update vectorial force */
 500             fix2             = _mm256_add_ps(fix2,tx);
 501             fiy2             = _mm256_add_ps(fiy2,ty);
 502             fiz2             = _mm256_add_ps(fiz2,tz);
 503
 504             fjx3             = _mm256_add_ps(fjx3,tx);
 505             fjy3             = _mm256_add_ps(fjy3,ty);
 506             fjz3             = _mm256_add_ps(fjz3,tz);
 507
 508             /**************************
 509              * CALCULATE INTERACTIONS *
 510              **************************/
 511
 512             /* REACTION-FIELD ELECTROSTATICS */
 513             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 514             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 515
 516             /* Update potential sum for this i atom from the interaction with this j atom. */
 517             velecsum         = _mm256_add_ps(velecsum,velec);
 518
 519             fscal            = felec;
 520
 521             /* Calculate temporary vectorial force */
 522             tx               = _mm256_mul_ps(fscal,dx31);
 523             ty               = _mm256_mul_ps(fscal,dy31);
 524             tz               = _mm256_mul_ps(fscal,dz31);
 525
 526             /* Update vectorial force */
 527             fix3             = _mm256_add_ps(fix3,tx);
 528             fiy3             = _mm256_add_ps(fiy3,ty);
 529             fiz3             = _mm256_add_ps(fiz3,tz);
 530
 531             fjx1             = _mm256_add_ps(fjx1,tx);
 532             fjy1             = _mm256_add_ps(fjy1,ty);
 533             fjz1             = _mm256_add_ps(fjz1,tz);
 534
 535             /**************************
 536              * CALCULATE INTERACTIONS *
 537              **************************/
 538
 539             /* REACTION-FIELD ELECTROSTATICS */
 540             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 541             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 542
 543             /* Update potential sum for this i atom from the interaction with this j atom. */
 544             velecsum         = _mm256_add_ps(velecsum,velec);
 545
 546             fscal            = felec;
 547
 548             /* Calculate temporary vectorial force */
 549             tx               = _mm256_mul_ps(fscal,dx32);
 550             ty               = _mm256_mul_ps(fscal,dy32);
 551             tz               = _mm256_mul_ps(fscal,dz32);
 552
 553             /* Update vectorial force */
 554             fix3             = _mm256_add_ps(fix3,tx);
 555             fiy3             = _mm256_add_ps(fiy3,ty);
 556             fiz3             = _mm256_add_ps(fiz3,tz);
 557
 558             fjx2             = _mm256_add_ps(fjx2,tx);
 559             fjy2             = _mm256_add_ps(fjy2,ty);
 560             fjz2             = _mm256_add_ps(fjz2,tz);
 561
 562             /**************************
 563              * CALCULATE INTERACTIONS *
 564              **************************/
 565
 566             /* REACTION-FIELD ELECTROSTATICS */
 567             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 568             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 569
 570             /* Update potential sum for this i atom from the interaction with this j atom. */
 571             velecsum         = _mm256_add_ps(velecsum,velec);
 572
 573             fscal            = felec;
 574
 575             /* Calculate temporary vectorial force */
 576             tx               = _mm256_mul_ps(fscal,dx33);
 577             ty               = _mm256_mul_ps(fscal,dy33);
 578             tz               = _mm256_mul_ps(fscal,dz33);
 579
 580             /* Update vectorial force */
 581             fix3             = _mm256_add_ps(fix3,tx);
 582             fiy3             = _mm256_add_ps(fiy3,ty);
 583             fiz3             = _mm256_add_ps(fiz3,tz);
 584
 585             fjx3             = _mm256_add_ps(fjx3,tx);
 586             fjy3             = _mm256_add_ps(fjy3,ty);
 587             fjz3             = _mm256_add_ps(fjz3,tz);
 588
 589             fjptrA             = f+j_coord_offsetA;
 590             fjptrB             = f+j_coord_offsetB;
 591             fjptrC             = f+j_coord_offsetC;
 592             fjptrD             = f+j_coord_offsetD;
 593             fjptrE             = f+j_coord_offsetE;
 594             fjptrF             = f+j_coord_offsetF;
 595             fjptrG             = f+j_coord_offsetG;
 596             fjptrH             = f+j_coord_offsetH;
 597
 598             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 599                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 600                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 601
 602             /* Inner loop uses 323 flops */
 603         }
 604
 605         if(jidx<j_index_end)
 606         {
 607
 608             /* Get j neighbor index, and coordinate index */
 609             jnrlistA         = jjnr[jidx];
 610             jnrlistB         = jjnr[jidx+1];
 611             jnrlistC         = jjnr[jidx+2];
 612             jnrlistD         = jjnr[jidx+3];
 613             jnrlistE         = jjnr[jidx+4];
 614             jnrlistF         = jjnr[jidx+5];
 615             jnrlistG         = jjnr[jidx+6];
 616             jnrlistH         = jjnr[jidx+7];
 617             /* Sign of each element will be negative for non-real atoms.
 618              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 619              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 620              */
 621             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 622                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 623
 624             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 625             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 626             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 627             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 628             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 629             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 630             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 631             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 632             j_coord_offsetA  = DIM*jnrA;
 633             j_coord_offsetB  = DIM*jnrB;
 634             j_coord_offsetC  = DIM*jnrC;
 635             j_coord_offsetD  = DIM*jnrD;
 636             j_coord_offsetE  = DIM*jnrE;
 637             j_coord_offsetF  = DIM*jnrF;
 638             j_coord_offsetG  = DIM*jnrG;
 639             j_coord_offsetH  = DIM*jnrH;
 640
 641             /* load j atom coordinates */
 642             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 643                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 644                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 645                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 646                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 647                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 648
 649             /* Calculate displacement vector */
 650             dx00             = _mm256_sub_ps(ix0,jx0);
 651             dy00             = _mm256_sub_ps(iy0,jy0);
 652             dz00             = _mm256_sub_ps(iz0,jz0);
 653             dx11             = _mm256_sub_ps(ix1,jx1);
 654             dy11             = _mm256_sub_ps(iy1,jy1);
 655             dz11             = _mm256_sub_ps(iz1,jz1);
 656             dx12             = _mm256_sub_ps(ix1,jx2);
 657             dy12             = _mm256_sub_ps(iy1,jy2);
 658             dz12             = _mm256_sub_ps(iz1,jz2);
 659             dx13             = _mm256_sub_ps(ix1,jx3);
 660             dy13             = _mm256_sub_ps(iy1,jy3);
 661             dz13             = _mm256_sub_ps(iz1,jz3);
 662             dx21             = _mm256_sub_ps(ix2,jx1);
 663             dy21             = _mm256_sub_ps(iy2,jy1);
 664             dz21             = _mm256_sub_ps(iz2,jz1);
 665             dx22             = _mm256_sub_ps(ix2,jx2);
 666             dy22             = _mm256_sub_ps(iy2,jy2);
 667             dz22             = _mm256_sub_ps(iz2,jz2);
 668             dx23             = _mm256_sub_ps(ix2,jx3);
 669             dy23             = _mm256_sub_ps(iy2,jy3);
 670             dz23             = _mm256_sub_ps(iz2,jz3);
 671             dx31             = _mm256_sub_ps(ix3,jx1);
 672             dy31             = _mm256_sub_ps(iy3,jy1);
 673             dz31             = _mm256_sub_ps(iz3,jz1);
 674             dx32             = _mm256_sub_ps(ix3,jx2);
 675             dy32             = _mm256_sub_ps(iy3,jy2);
 676             dz32             = _mm256_sub_ps(iz3,jz2);
 677             dx33             = _mm256_sub_ps(ix3,jx3);
 678             dy33             = _mm256_sub_ps(iy3,jy3);
 679             dz33             = _mm256_sub_ps(iz3,jz3);
 680
 681             /* Calculate squared distance and things based on it */
 682             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 683             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 684             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 685             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 686             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 687             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 688             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 689             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 690             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 691             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 692
 693             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 694             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 695             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 696             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 697             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 698             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 699             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 700             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 701             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 702
 703             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 704             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 705             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 706             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 707             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 708             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 709             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 710             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 711             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 712             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 713
 714             fjx0             = _mm256_setzero_ps();
 715             fjy0             = _mm256_setzero_ps();
 716             fjz0             = _mm256_setzero_ps();
 717             fjx1             = _mm256_setzero_ps();
 718             fjy1             = _mm256_setzero_ps();
 719             fjz1             = _mm256_setzero_ps();
 720             fjx2             = _mm256_setzero_ps();
 721             fjy2             = _mm256_setzero_ps();
 722             fjz2             = _mm256_setzero_ps();
 723             fjx3             = _mm256_setzero_ps();
 724             fjy3             = _mm256_setzero_ps();
 725             fjz3             = _mm256_setzero_ps();
 726
 727             /**************************
 728              * CALCULATE INTERACTIONS *
 729              **************************/
 730
 731             /* LENNARD-JONES DISPERSION/REPULSION */
 732
 733             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 734             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 735             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 736             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 737             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 738
 739             /* Update potential sum for this i atom from the interaction with this j atom. */
 740             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 741             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 742
 743             fscal            = fvdw;
 744
 745             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 746
 747             /* Calculate temporary vectorial force */
 748             tx               = _mm256_mul_ps(fscal,dx00);
 749             ty               = _mm256_mul_ps(fscal,dy00);
 750             tz               = _mm256_mul_ps(fscal,dz00);
 751
 752             /* Update vectorial force */
 753             fix0             = _mm256_add_ps(fix0,tx);
 754             fiy0             = _mm256_add_ps(fiy0,ty);
 755             fiz0             = _mm256_add_ps(fiz0,tz);
 756
 757             fjx0             = _mm256_add_ps(fjx0,tx);
 758             fjy0             = _mm256_add_ps(fjy0,ty);
 759             fjz0             = _mm256_add_ps(fjz0,tz);
 760
 761             /**************************
 762              * CALCULATE INTERACTIONS *
 763              **************************/
 764
 765             /* REACTION-FIELD ELECTROSTATICS */
 766             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 767             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 768
 769             /* Update potential sum for this i atom from the interaction with this j atom. */
 770             velec            = _mm256_andnot_ps(dummy_mask,velec);
 771             velecsum         = _mm256_add_ps(velecsum,velec);
 772
 773             fscal            = felec;
 774
 775             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 776
 777             /* Calculate temporary vectorial force */
 778             tx               = _mm256_mul_ps(fscal,dx11);
 779             ty               = _mm256_mul_ps(fscal,dy11);
 780             tz               = _mm256_mul_ps(fscal,dz11);
 781
 782             /* Update vectorial force */
 783             fix1             = _mm256_add_ps(fix1,tx);
 784             fiy1             = _mm256_add_ps(fiy1,ty);
 785             fiz1             = _mm256_add_ps(fiz1,tz);
 786
 787             fjx1             = _mm256_add_ps(fjx1,tx);
 788             fjy1             = _mm256_add_ps(fjy1,ty);
 789             fjz1             = _mm256_add_ps(fjz1,tz);
 790
 791             /**************************
 792              * CALCULATE INTERACTIONS *
 793              **************************/
 794
 795             /* REACTION-FIELD ELECTROSTATICS */
 796             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 797             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 798
 799             /* Update potential sum for this i atom from the interaction with this j atom. */
 800             velec            = _mm256_andnot_ps(dummy_mask,velec);
 801             velecsum         = _mm256_add_ps(velecsum,velec);
 802
 803             fscal            = felec;
 804
 805             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 806
 807             /* Calculate temporary vectorial force */
 808             tx               = _mm256_mul_ps(fscal,dx12);
 809             ty               = _mm256_mul_ps(fscal,dy12);
 810             tz               = _mm256_mul_ps(fscal,dz12);
 811
 812             /* Update vectorial force */
 813             fix1             = _mm256_add_ps(fix1,tx);
 814             fiy1             = _mm256_add_ps(fiy1,ty);
 815             fiz1             = _mm256_add_ps(fiz1,tz);
 816
 817             fjx2             = _mm256_add_ps(fjx2,tx);
 818             fjy2             = _mm256_add_ps(fjy2,ty);
 819             fjz2             = _mm256_add_ps(fjz2,tz);
 820
 821             /**************************
 822              * CALCULATE INTERACTIONS *
 823              **************************/
 824
 825             /* REACTION-FIELD ELECTROSTATICS */
 826             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 827             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 828
 829             /* Update potential sum for this i atom from the interaction with this j atom. */
 830             velec            = _mm256_andnot_ps(dummy_mask,velec);
 831             velecsum         = _mm256_add_ps(velecsum,velec);
 832
 833             fscal            = felec;
 834
 835             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 836
 837             /* Calculate temporary vectorial force */
 838             tx               = _mm256_mul_ps(fscal,dx13);
 839             ty               = _mm256_mul_ps(fscal,dy13);
 840             tz               = _mm256_mul_ps(fscal,dz13);
 841
 842             /* Update vectorial force */
 843             fix1             = _mm256_add_ps(fix1,tx);
 844             fiy1             = _mm256_add_ps(fiy1,ty);
 845             fiz1             = _mm256_add_ps(fiz1,tz);
 846
 847             fjx3             = _mm256_add_ps(fjx3,tx);
 848             fjy3             = _mm256_add_ps(fjy3,ty);
 849             fjz3             = _mm256_add_ps(fjz3,tz);
 850
 851             /**************************
 852              * CALCULATE INTERACTIONS *
 853              **************************/
 854
 855             /* REACTION-FIELD ELECTROSTATICS */
 856             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 857             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 858
 859             /* Update potential sum for this i atom from the interaction with this j atom. */
 860             velec            = _mm256_andnot_ps(dummy_mask,velec);
 861             velecsum         = _mm256_add_ps(velecsum,velec);
 862
 863             fscal            = felec;
 864
 865             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 866
 867             /* Calculate temporary vectorial force */
 868             tx               = _mm256_mul_ps(fscal,dx21);
 869             ty               = _mm256_mul_ps(fscal,dy21);
 870             tz               = _mm256_mul_ps(fscal,dz21);
 871
 872             /* Update vectorial force */
 873             fix2             = _mm256_add_ps(fix2,tx);
 874             fiy2             = _mm256_add_ps(fiy2,ty);
 875             fiz2             = _mm256_add_ps(fiz2,tz);
 876
 877             fjx1             = _mm256_add_ps(fjx1,tx);
 878             fjy1             = _mm256_add_ps(fjy1,ty);
 879             fjz1             = _mm256_add_ps(fjz1,tz);
 880
 881             /**************************
 882              * CALCULATE INTERACTIONS *
 883              **************************/
 884
 885             /* REACTION-FIELD ELECTROSTATICS */
 886             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 887             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 888
 889             /* Update potential sum for this i atom from the interaction with this j atom. */
 890             velec            = _mm256_andnot_ps(dummy_mask,velec);
 891             velecsum         = _mm256_add_ps(velecsum,velec);
 892
 893             fscal            = felec;
 894
 895             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 896
 897             /* Calculate temporary vectorial force */
 898             tx               = _mm256_mul_ps(fscal,dx22);
 899             ty               = _mm256_mul_ps(fscal,dy22);
 900             tz               = _mm256_mul_ps(fscal,dz22);
 901
 902             /* Update vectorial force */
 903             fix2             = _mm256_add_ps(fix2,tx);
 904             fiy2             = _mm256_add_ps(fiy2,ty);
 905             fiz2             = _mm256_add_ps(fiz2,tz);
 906
 907             fjx2             = _mm256_add_ps(fjx2,tx);
 908             fjy2             = _mm256_add_ps(fjy2,ty);
 909             fjz2             = _mm256_add_ps(fjz2,tz);
 910
 911             /**************************
 912              * CALCULATE INTERACTIONS *
 913              **************************/
 914
 915             /* REACTION-FIELD ELECTROSTATICS */
 916             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 917             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 918
 919             /* Update potential sum for this i atom from the interaction with this j atom. */
 920             velec            = _mm256_andnot_ps(dummy_mask,velec);
 921             velecsum         = _mm256_add_ps(velecsum,velec);
 922
 923             fscal            = felec;
 924
 925             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 926
 927             /* Calculate temporary vectorial force */
 928             tx               = _mm256_mul_ps(fscal,dx23);
 929             ty               = _mm256_mul_ps(fscal,dy23);
 930             tz               = _mm256_mul_ps(fscal,dz23);
 931
 932             /* Update vectorial force */
 933             fix2             = _mm256_add_ps(fix2,tx);
 934             fiy2             = _mm256_add_ps(fiy2,ty);
 935             fiz2             = _mm256_add_ps(fiz2,tz);
 936
 937             fjx3             = _mm256_add_ps(fjx3,tx);
 938             fjy3             = _mm256_add_ps(fjy3,ty);
 939             fjz3             = _mm256_add_ps(fjz3,tz);
 940
 941             /**************************
 942              * CALCULATE INTERACTIONS *
 943              **************************/
 944
 945             /* REACTION-FIELD ELECTROSTATICS */
 946             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 947             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 948
 949             /* Update potential sum for this i atom from the interaction with this j atom. */
 950             velec            = _mm256_andnot_ps(dummy_mask,velec);
 951             velecsum         = _mm256_add_ps(velecsum,velec);
 952
 953             fscal            = felec;
 954
 955             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 956
 957             /* Calculate temporary vectorial force */
 958             tx               = _mm256_mul_ps(fscal,dx31);
 959             ty               = _mm256_mul_ps(fscal,dy31);
 960             tz               = _mm256_mul_ps(fscal,dz31);
 961
 962             /* Update vectorial force */
 963             fix3             = _mm256_add_ps(fix3,tx);
 964             fiy3             = _mm256_add_ps(fiy3,ty);
 965             fiz3             = _mm256_add_ps(fiz3,tz);
 966
 967             fjx1             = _mm256_add_ps(fjx1,tx);
 968             fjy1             = _mm256_add_ps(fjy1,ty);
 969             fjz1             = _mm256_add_ps(fjz1,tz);
 970
 971             /**************************
 972              * CALCULATE INTERACTIONS *
 973              **************************/
 974
 975             /* REACTION-FIELD ELECTROSTATICS */
 976             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 977             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 978
 979             /* Update potential sum for this i atom from the interaction with this j atom. */
 980             velec            = _mm256_andnot_ps(dummy_mask,velec);
 981             velecsum         = _mm256_add_ps(velecsum,velec);
 982
 983             fscal            = felec;
 984
 985             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 986
 987             /* Calculate temporary vectorial force */
 988             tx               = _mm256_mul_ps(fscal,dx32);
 989             ty               = _mm256_mul_ps(fscal,dy32);
 990             tz               = _mm256_mul_ps(fscal,dz32);
 991
 992             /* Update vectorial force */
 993             fix3             = _mm256_add_ps(fix3,tx);
 994             fiy3             = _mm256_add_ps(fiy3,ty);
 995             fiz3             = _mm256_add_ps(fiz3,tz);
 996
 997             fjx2             = _mm256_add_ps(fjx2,tx);
 998             fjy2             = _mm256_add_ps(fjy2,ty);
 999             fjz2             = _mm256_add_ps(fjz2,tz);
1000
1001             /**************************
1002              * CALCULATE INTERACTIONS *
1003              **************************/
1004
1005             /* REACTION-FIELD ELECTROSTATICS */
1006             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
1007             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1008
1009             /* Update potential sum for this i atom from the interaction with this j atom. */
1010             velec            = _mm256_andnot_ps(dummy_mask,velec);
1011             velecsum         = _mm256_add_ps(velecsum,velec);
1012
1013             fscal            = felec;
1014
1015             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1016
1017             /* Calculate temporary vectorial force */
1018             tx               = _mm256_mul_ps(fscal,dx33);
1019             ty               = _mm256_mul_ps(fscal,dy33);
1020             tz               = _mm256_mul_ps(fscal,dz33);
1021
1022             /* Update vectorial force */
1023             fix3             = _mm256_add_ps(fix3,tx);
1024             fiy3             = _mm256_add_ps(fiy3,ty);
1025             fiz3             = _mm256_add_ps(fiz3,tz);
1026
1027             fjx3             = _mm256_add_ps(fjx3,tx);
1028             fjy3             = _mm256_add_ps(fjy3,ty);
1029             fjz3             = _mm256_add_ps(fjz3,tz);
1030
1031             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1032             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1033             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1034             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1035             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1036             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1037             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1038             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1039
1040             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1041                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1042                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1043
1044             /* Inner loop uses 323 flops */
1045         }
1046
1047         /* End of innermost loop */
1048
1049         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1050                                                  f+i_coord_offset,fshift+i_shift_offset);
1051
1052         ggid                        = gid[iidx];
1053         /* Update potential energies */
1054         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1055         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1056
1057         /* Increment number of inner iterations */
1058         inneriter                  += j_index_end - j_index_start;
1059
1060         /* Outer loop uses 26 flops */
1061     }
1062
1063     /* Increment number of outer iterations */
1064     outeriter        += nri;
1065
1066     /* Update outer/inner flops */
1067
1068     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*323);
1069 }
1070 /*
1071  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_single
1072  * Electrostatics interaction: ReactionField
1073  * VdW interaction:            LennardJones
1074  * Geometry:                   Water4-Water4
1075  * Calculate force/pot:        Force
1076  */
1077 void
1078 nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_single
1079                     (t_nblist * gmx_restrict                nlist,
1080                      rvec * gmx_restrict                    xx,
1081                      rvec * gmx_restrict                    ff,
1082                      t_forcerec * gmx_restrict              fr,
1083                      t_mdatoms * gmx_restrict               mdatoms,
1084                      nb_kernel_data_t * gmx_restrict        kernel_data,
1085                      t_nrnb * gmx_restrict                  nrnb)
1086 {
1087     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1088      * just 0 for non-waters.
1089      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1090      * jnr indices corresponding to data put in the four positions in the SIMD register.
1091      */
1092     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1093     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1094     int              jnrA,jnrB,jnrC,jnrD;
1095     int              jnrE,jnrF,jnrG,jnrH;
1096     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1097     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1098     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1099     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1100     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1101     real             rcutoff_scalar;
1102     real             *shiftvec,*fshift,*x,*f;
1103     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1104     real             scratch[4*DIM];
1105     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1106     real *           vdwioffsetptr0;
1107     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1108     real *           vdwioffsetptr1;
1109     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1110     real *           vdwioffsetptr2;
1111     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1112     real *           vdwioffsetptr3;
1113     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1114     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1115     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1116     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1117     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1118     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1119     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1120     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1121     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1122     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1123     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1124     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1125     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1126     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1127     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1128     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1129     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1130     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1131     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1132     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1133     real             *charge;
1134     int              nvdwtype;
1135     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1136     int              *vdwtype;
1137     real             *vdwparam;
1138     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1139     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1140     __m256           dummy_mask,cutoff_mask;
1141     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1142     __m256           one     = _mm256_set1_ps(1.0);
1143     __m256           two     = _mm256_set1_ps(2.0);
1144     x                = xx[0];
1145     f                = ff[0];
1146
1147     nri              = nlist->nri;
1148     iinr             = nlist->iinr;
1149     jindex           = nlist->jindex;
1150     jjnr             = nlist->jjnr;
1151     shiftidx         = nlist->shift;
1152     gid              = nlist->gid;
1153     shiftvec         = fr->shift_vec[0];
1154     fshift           = fr->fshift[0];
1155     facel            = _mm256_set1_ps(fr->epsfac);
1156     charge           = mdatoms->chargeA;
1157     krf              = _mm256_set1_ps(fr->ic->k_rf);
1158     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1159     crf              = _mm256_set1_ps(fr->ic->c_rf);
1160     nvdwtype         = fr->ntype;
1161     vdwparam         = fr->nbfp;
1162     vdwtype          = mdatoms->typeA;
1163
1164     /* Setup water-specific parameters */
1165     inr              = nlist->iinr[0];
1166     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1167     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1168     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1169     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1170
1171     jq1              = _mm256_set1_ps(charge[inr+1]);
1172     jq2              = _mm256_set1_ps(charge[inr+2]);
1173     jq3              = _mm256_set1_ps(charge[inr+3]);
1174     vdwjidx0A        = 2*vdwtype[inr+0];
1175     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1176     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1177     qq11             = _mm256_mul_ps(iq1,jq1);
1178     qq12             = _mm256_mul_ps(iq1,jq2);
1179     qq13             = _mm256_mul_ps(iq1,jq3);
1180     qq21             = _mm256_mul_ps(iq2,jq1);
1181     qq22             = _mm256_mul_ps(iq2,jq2);
1182     qq23             = _mm256_mul_ps(iq2,jq3);
1183     qq31             = _mm256_mul_ps(iq3,jq1);
1184     qq32             = _mm256_mul_ps(iq3,jq2);
1185     qq33             = _mm256_mul_ps(iq3,jq3);
1186
1187     /* Avoid stupid compiler warnings */
1188     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1189     j_coord_offsetA = 0;
1190     j_coord_offsetB = 0;
1191     j_coord_offsetC = 0;
1192     j_coord_offsetD = 0;
1193     j_coord_offsetE = 0;
1194     j_coord_offsetF = 0;
1195     j_coord_offsetG = 0;
1196     j_coord_offsetH = 0;
1197
1198     outeriter        = 0;
1199     inneriter        = 0;
1200
1201     for(iidx=0;iidx<4*DIM;iidx++)
1202     {
1203         scratch[iidx] = 0.0;
1204     }
1205
1206     /* Start outer loop over neighborlists */
1207     for(iidx=0; iidx<nri; iidx++)
1208     {
1209         /* Load shift vector for this list */
1210         i_shift_offset   = DIM*shiftidx[iidx];
1211
1212         /* Load limits for loop over neighbors */
1213         j_index_start    = jindex[iidx];
1214         j_index_end      = jindex[iidx+1];
1215
1216         /* Get outer coordinate index */
1217         inr              = iinr[iidx];
1218         i_coord_offset   = DIM*inr;
1219
1220         /* Load i particle coords and add shift vector */
1221         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1222                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1223
1224         fix0             = _mm256_setzero_ps();
1225         fiy0             = _mm256_setzero_ps();
1226         fiz0             = _mm256_setzero_ps();
1227         fix1             = _mm256_setzero_ps();
1228         fiy1             = _mm256_setzero_ps();
1229         fiz1             = _mm256_setzero_ps();
1230         fix2             = _mm256_setzero_ps();
1231         fiy2             = _mm256_setzero_ps();
1232         fiz2             = _mm256_setzero_ps();
1233         fix3             = _mm256_setzero_ps();
1234         fiy3             = _mm256_setzero_ps();
1235         fiz3             = _mm256_setzero_ps();
1236
1237         /* Start inner kernel loop */
1238         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1239         {
1240
1241             /* Get j neighbor index, and coordinate index */
1242             jnrA             = jjnr[jidx];
1243             jnrB             = jjnr[jidx+1];
1244             jnrC             = jjnr[jidx+2];
1245             jnrD             = jjnr[jidx+3];
1246             jnrE             = jjnr[jidx+4];
1247             jnrF             = jjnr[jidx+5];
1248             jnrG             = jjnr[jidx+6];
1249             jnrH             = jjnr[jidx+7];
1250             j_coord_offsetA  = DIM*jnrA;
1251             j_coord_offsetB  = DIM*jnrB;
1252             j_coord_offsetC  = DIM*jnrC;
1253             j_coord_offsetD  = DIM*jnrD;
1254             j_coord_offsetE  = DIM*jnrE;
1255             j_coord_offsetF  = DIM*jnrF;
1256             j_coord_offsetG  = DIM*jnrG;
1257             j_coord_offsetH  = DIM*jnrH;
1258
1259             /* load j atom coordinates */
1260             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1261                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1262                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1263                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1264                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1265                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1266
1267             /* Calculate displacement vector */
1268             dx00             = _mm256_sub_ps(ix0,jx0);
1269             dy00             = _mm256_sub_ps(iy0,jy0);
1270             dz00             = _mm256_sub_ps(iz0,jz0);
1271             dx11             = _mm256_sub_ps(ix1,jx1);
1272             dy11             = _mm256_sub_ps(iy1,jy1);
1273             dz11             = _mm256_sub_ps(iz1,jz1);
1274             dx12             = _mm256_sub_ps(ix1,jx2);
1275             dy12             = _mm256_sub_ps(iy1,jy2);
1276             dz12             = _mm256_sub_ps(iz1,jz2);
1277             dx13             = _mm256_sub_ps(ix1,jx3);
1278             dy13             = _mm256_sub_ps(iy1,jy3);
1279             dz13             = _mm256_sub_ps(iz1,jz3);
1280             dx21             = _mm256_sub_ps(ix2,jx1);
1281             dy21             = _mm256_sub_ps(iy2,jy1);
1282             dz21             = _mm256_sub_ps(iz2,jz1);
1283             dx22             = _mm256_sub_ps(ix2,jx2);
1284             dy22             = _mm256_sub_ps(iy2,jy2);
1285             dz22             = _mm256_sub_ps(iz2,jz2);
1286             dx23             = _mm256_sub_ps(ix2,jx3);
1287             dy23             = _mm256_sub_ps(iy2,jy3);
1288             dz23             = _mm256_sub_ps(iz2,jz3);
1289             dx31             = _mm256_sub_ps(ix3,jx1);
1290             dy31             = _mm256_sub_ps(iy3,jy1);
1291             dz31             = _mm256_sub_ps(iz3,jz1);
1292             dx32             = _mm256_sub_ps(ix3,jx2);
1293             dy32             = _mm256_sub_ps(iy3,jy2);
1294             dz32             = _mm256_sub_ps(iz3,jz2);
1295             dx33             = _mm256_sub_ps(ix3,jx3);
1296             dy33             = _mm256_sub_ps(iy3,jy3);
1297             dz33             = _mm256_sub_ps(iz3,jz3);
1298
1299             /* Calculate squared distance and things based on it */
1300             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1301             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1302             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1303             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1304             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1305             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1306             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1307             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1308             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1309             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1310
1311             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1312             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1313             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1314             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1315             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1316             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1317             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1318             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1319             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1320
1321             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1322             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1323             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1324             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1325             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1326             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1327             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1328             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1329             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1330             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1331
1332             fjx0             = _mm256_setzero_ps();
1333             fjy0             = _mm256_setzero_ps();
1334             fjz0             = _mm256_setzero_ps();
1335             fjx1             = _mm256_setzero_ps();
1336             fjy1             = _mm256_setzero_ps();
1337             fjz1             = _mm256_setzero_ps();
1338             fjx2             = _mm256_setzero_ps();
1339             fjy2             = _mm256_setzero_ps();
1340             fjz2             = _mm256_setzero_ps();
1341             fjx3             = _mm256_setzero_ps();
1342             fjy3             = _mm256_setzero_ps();
1343             fjz3             = _mm256_setzero_ps();
1344
1345             /**************************
1346              * CALCULATE INTERACTIONS *
1347              **************************/
1348
1349             /* LENNARD-JONES DISPERSION/REPULSION */
1350
1351             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1352             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1353
1354             fscal            = fvdw;
1355
1356             /* Calculate temporary vectorial force */
1357             tx               = _mm256_mul_ps(fscal,dx00);
1358             ty               = _mm256_mul_ps(fscal,dy00);
1359             tz               = _mm256_mul_ps(fscal,dz00);
1360
1361             /* Update vectorial force */
1362             fix0             = _mm256_add_ps(fix0,tx);
1363             fiy0             = _mm256_add_ps(fiy0,ty);
1364             fiz0             = _mm256_add_ps(fiz0,tz);
1365
1366             fjx0             = _mm256_add_ps(fjx0,tx);
1367             fjy0             = _mm256_add_ps(fjy0,ty);
1368             fjz0             = _mm256_add_ps(fjz0,tz);
1369
1370             /**************************
1371              * CALCULATE INTERACTIONS *
1372              **************************/
1373
1374             /* REACTION-FIELD ELECTROSTATICS */
1375             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1376
1377             fscal            = felec;
1378
1379             /* Calculate temporary vectorial force */
1380             tx               = _mm256_mul_ps(fscal,dx11);
1381             ty               = _mm256_mul_ps(fscal,dy11);
1382             tz               = _mm256_mul_ps(fscal,dz11);
1383
1384             /* Update vectorial force */
1385             fix1             = _mm256_add_ps(fix1,tx);
1386             fiy1             = _mm256_add_ps(fiy1,ty);
1387             fiz1             = _mm256_add_ps(fiz1,tz);
1388
1389             fjx1             = _mm256_add_ps(fjx1,tx);
1390             fjy1             = _mm256_add_ps(fjy1,ty);
1391             fjz1             = _mm256_add_ps(fjz1,tz);
1392
1393             /**************************
1394              * CALCULATE INTERACTIONS *
1395              **************************/
1396
1397             /* REACTION-FIELD ELECTROSTATICS */
1398             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1399
1400             fscal            = felec;
1401
1402             /* Calculate temporary vectorial force */
1403             tx               = _mm256_mul_ps(fscal,dx12);
1404             ty               = _mm256_mul_ps(fscal,dy12);
1405             tz               = _mm256_mul_ps(fscal,dz12);
1406
1407             /* Update vectorial force */
1408             fix1             = _mm256_add_ps(fix1,tx);
1409             fiy1             = _mm256_add_ps(fiy1,ty);
1410             fiz1             = _mm256_add_ps(fiz1,tz);
1411
1412             fjx2             = _mm256_add_ps(fjx2,tx);
1413             fjy2             = _mm256_add_ps(fjy2,ty);
1414             fjz2             = _mm256_add_ps(fjz2,tz);
1415
1416             /**************************
1417              * CALCULATE INTERACTIONS *
1418              **************************/
1419
1420             /* REACTION-FIELD ELECTROSTATICS */
1421             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1422
1423             fscal            = felec;
1424
1425             /* Calculate temporary vectorial force */
1426             tx               = _mm256_mul_ps(fscal,dx13);
1427             ty               = _mm256_mul_ps(fscal,dy13);
1428             tz               = _mm256_mul_ps(fscal,dz13);
1429
1430             /* Update vectorial force */
1431             fix1             = _mm256_add_ps(fix1,tx);
1432             fiy1             = _mm256_add_ps(fiy1,ty);
1433             fiz1             = _mm256_add_ps(fiz1,tz);
1434
1435             fjx3             = _mm256_add_ps(fjx3,tx);
1436             fjy3             = _mm256_add_ps(fjy3,ty);
1437             fjz3             = _mm256_add_ps(fjz3,tz);
1438
1439             /**************************
1440              * CALCULATE INTERACTIONS *
1441              **************************/
1442
1443             /* REACTION-FIELD ELECTROSTATICS */
1444             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1445
1446             fscal            = felec;
1447
1448             /* Calculate temporary vectorial force */
1449             tx               = _mm256_mul_ps(fscal,dx21);
1450             ty               = _mm256_mul_ps(fscal,dy21);
1451             tz               = _mm256_mul_ps(fscal,dz21);
1452
1453             /* Update vectorial force */
1454             fix2             = _mm256_add_ps(fix2,tx);
1455             fiy2             = _mm256_add_ps(fiy2,ty);
1456             fiz2             = _mm256_add_ps(fiz2,tz);
1457
1458             fjx1             = _mm256_add_ps(fjx1,tx);
1459             fjy1             = _mm256_add_ps(fjy1,ty);
1460             fjz1             = _mm256_add_ps(fjz1,tz);
1461
1462             /**************************
1463              * CALCULATE INTERACTIONS *
1464              **************************/
1465
1466             /* REACTION-FIELD ELECTROSTATICS */
1467             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1468
1469             fscal            = felec;
1470
1471             /* Calculate temporary vectorial force */
1472             tx               = _mm256_mul_ps(fscal,dx22);
1473             ty               = _mm256_mul_ps(fscal,dy22);
1474             tz               = _mm256_mul_ps(fscal,dz22);
1475
1476             /* Update vectorial force */
1477             fix2             = _mm256_add_ps(fix2,tx);
1478             fiy2             = _mm256_add_ps(fiy2,ty);
1479             fiz2             = _mm256_add_ps(fiz2,tz);
1480
1481             fjx2             = _mm256_add_ps(fjx2,tx);
1482             fjy2             = _mm256_add_ps(fjy2,ty);
1483             fjz2             = _mm256_add_ps(fjz2,tz);
1484
1485             /**************************
1486              * CALCULATE INTERACTIONS *
1487              **************************/
1488
1489             /* REACTION-FIELD ELECTROSTATICS */
1490             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
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             /* REACTION-FIELD ELECTROSTATICS */
1513             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1514
1515             fscal            = felec;
1516
1517             /* Calculate temporary vectorial force */
1518             tx               = _mm256_mul_ps(fscal,dx31);
1519             ty               = _mm256_mul_ps(fscal,dy31);
1520             tz               = _mm256_mul_ps(fscal,dz31);
1521
1522             /* Update vectorial force */
1523             fix3             = _mm256_add_ps(fix3,tx);
1524             fiy3             = _mm256_add_ps(fiy3,ty);
1525             fiz3             = _mm256_add_ps(fiz3,tz);
1526
1527             fjx1             = _mm256_add_ps(fjx1,tx);
1528             fjy1             = _mm256_add_ps(fjy1,ty);
1529             fjz1             = _mm256_add_ps(fjz1,tz);
1530
1531             /**************************
1532              * CALCULATE INTERACTIONS *
1533              **************************/
1534
1535             /* REACTION-FIELD ELECTROSTATICS */
1536             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1537
1538             fscal            = felec;
1539
1540             /* Calculate temporary vectorial force */
1541             tx               = _mm256_mul_ps(fscal,dx32);
1542             ty               = _mm256_mul_ps(fscal,dy32);
1543             tz               = _mm256_mul_ps(fscal,dz32);
1544
1545             /* Update vectorial force */
1546             fix3             = _mm256_add_ps(fix3,tx);
1547             fiy3             = _mm256_add_ps(fiy3,ty);
1548             fiz3             = _mm256_add_ps(fiz3,tz);
1549
1550             fjx2             = _mm256_add_ps(fjx2,tx);
1551             fjy2             = _mm256_add_ps(fjy2,ty);
1552             fjz2             = _mm256_add_ps(fjz2,tz);
1553
1554             /**************************
1555              * CALCULATE INTERACTIONS *
1556              **************************/
1557
1558             /* REACTION-FIELD ELECTROSTATICS */
1559             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1560
1561             fscal            = felec;
1562
1563             /* Calculate temporary vectorial force */
1564             tx               = _mm256_mul_ps(fscal,dx33);
1565             ty               = _mm256_mul_ps(fscal,dy33);
1566             tz               = _mm256_mul_ps(fscal,dz33);
1567
1568             /* Update vectorial force */
1569             fix3             = _mm256_add_ps(fix3,tx);
1570             fiy3             = _mm256_add_ps(fiy3,ty);
1571             fiz3             = _mm256_add_ps(fiz3,tz);
1572
1573             fjx3             = _mm256_add_ps(fjx3,tx);
1574             fjy3             = _mm256_add_ps(fjy3,ty);
1575             fjz3             = _mm256_add_ps(fjz3,tz);
1576
1577             fjptrA             = f+j_coord_offsetA;
1578             fjptrB             = f+j_coord_offsetB;
1579             fjptrC             = f+j_coord_offsetC;
1580             fjptrD             = f+j_coord_offsetD;
1581             fjptrE             = f+j_coord_offsetE;
1582             fjptrF             = f+j_coord_offsetF;
1583             fjptrG             = f+j_coord_offsetG;
1584             fjptrH             = f+j_coord_offsetH;
1585
1586             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1587                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1588                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1589
1590             /* Inner loop uses 273 flops */
1591         }
1592
1593         if(jidx<j_index_end)
1594         {
1595
1596             /* Get j neighbor index, and coordinate index */
1597             jnrlistA         = jjnr[jidx];
1598             jnrlistB         = jjnr[jidx+1];
1599             jnrlistC         = jjnr[jidx+2];
1600             jnrlistD         = jjnr[jidx+3];
1601             jnrlistE         = jjnr[jidx+4];
1602             jnrlistF         = jjnr[jidx+5];
1603             jnrlistG         = jjnr[jidx+6];
1604             jnrlistH         = jjnr[jidx+7];
1605             /* Sign of each element will be negative for non-real atoms.
1606              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1607              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1608              */
1609             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1610                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1611
1612             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1613             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1614             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1615             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1616             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1617             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1618             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1619             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1620             j_coord_offsetA  = DIM*jnrA;
1621             j_coord_offsetB  = DIM*jnrB;
1622             j_coord_offsetC  = DIM*jnrC;
1623             j_coord_offsetD  = DIM*jnrD;
1624             j_coord_offsetE  = DIM*jnrE;
1625             j_coord_offsetF  = DIM*jnrF;
1626             j_coord_offsetG  = DIM*jnrG;
1627             j_coord_offsetH  = DIM*jnrH;
1628
1629             /* load j atom coordinates */
1630             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1631                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1632                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1633                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1634                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1635                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1636
1637             /* Calculate displacement vector */
1638             dx00             = _mm256_sub_ps(ix0,jx0);
1639             dy00             = _mm256_sub_ps(iy0,jy0);
1640             dz00             = _mm256_sub_ps(iz0,jz0);
1641             dx11             = _mm256_sub_ps(ix1,jx1);
1642             dy11             = _mm256_sub_ps(iy1,jy1);
1643             dz11             = _mm256_sub_ps(iz1,jz1);
1644             dx12             = _mm256_sub_ps(ix1,jx2);
1645             dy12             = _mm256_sub_ps(iy1,jy2);
1646             dz12             = _mm256_sub_ps(iz1,jz2);
1647             dx13             = _mm256_sub_ps(ix1,jx3);
1648             dy13             = _mm256_sub_ps(iy1,jy3);
1649             dz13             = _mm256_sub_ps(iz1,jz3);
1650             dx21             = _mm256_sub_ps(ix2,jx1);
1651             dy21             = _mm256_sub_ps(iy2,jy1);
1652             dz21             = _mm256_sub_ps(iz2,jz1);
1653             dx22             = _mm256_sub_ps(ix2,jx2);
1654             dy22             = _mm256_sub_ps(iy2,jy2);
1655             dz22             = _mm256_sub_ps(iz2,jz2);
1656             dx23             = _mm256_sub_ps(ix2,jx3);
1657             dy23             = _mm256_sub_ps(iy2,jy3);
1658             dz23             = _mm256_sub_ps(iz2,jz3);
1659             dx31             = _mm256_sub_ps(ix3,jx1);
1660             dy31             = _mm256_sub_ps(iy3,jy1);
1661             dz31             = _mm256_sub_ps(iz3,jz1);
1662             dx32             = _mm256_sub_ps(ix3,jx2);
1663             dy32             = _mm256_sub_ps(iy3,jy2);
1664             dz32             = _mm256_sub_ps(iz3,jz2);
1665             dx33             = _mm256_sub_ps(ix3,jx3);
1666             dy33             = _mm256_sub_ps(iy3,jy3);
1667             dz33             = _mm256_sub_ps(iz3,jz3);
1668
1669             /* Calculate squared distance and things based on it */
1670             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1671             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1672             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1673             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1674             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1675             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1676             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1677             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1678             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1679             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1680
1681             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1682             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1683             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1684             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1685             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1686             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1687             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1688             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1689             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1690
1691             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1692             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1693             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1694             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1695             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1696             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1697             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1698             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1699             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1700             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1701
1702             fjx0             = _mm256_setzero_ps();
1703             fjy0             = _mm256_setzero_ps();
1704             fjz0             = _mm256_setzero_ps();
1705             fjx1             = _mm256_setzero_ps();
1706             fjy1             = _mm256_setzero_ps();
1707             fjz1             = _mm256_setzero_ps();
1708             fjx2             = _mm256_setzero_ps();
1709             fjy2             = _mm256_setzero_ps();
1710             fjz2             = _mm256_setzero_ps();
1711             fjx3             = _mm256_setzero_ps();
1712             fjy3             = _mm256_setzero_ps();
1713             fjz3             = _mm256_setzero_ps();
1714
1715             /**************************
1716              * CALCULATE INTERACTIONS *
1717              **************************/
1718
1719             /* LENNARD-JONES DISPERSION/REPULSION */
1720
1721             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1722             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1723
1724             fscal            = fvdw;
1725
1726             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1727
1728             /* Calculate temporary vectorial force */
1729             tx               = _mm256_mul_ps(fscal,dx00);
1730             ty               = _mm256_mul_ps(fscal,dy00);
1731             tz               = _mm256_mul_ps(fscal,dz00);
1732
1733             /* Update vectorial force */
1734             fix0             = _mm256_add_ps(fix0,tx);
1735             fiy0             = _mm256_add_ps(fiy0,ty);
1736             fiz0             = _mm256_add_ps(fiz0,tz);
1737
1738             fjx0             = _mm256_add_ps(fjx0,tx);
1739             fjy0             = _mm256_add_ps(fjy0,ty);
1740             fjz0             = _mm256_add_ps(fjz0,tz);
1741
1742             /**************************
1743              * CALCULATE INTERACTIONS *
1744              **************************/
1745
1746             /* REACTION-FIELD ELECTROSTATICS */
1747             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1748
1749             fscal            = felec;
1750
1751             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1752
1753             /* Calculate temporary vectorial force */
1754             tx               = _mm256_mul_ps(fscal,dx11);
1755             ty               = _mm256_mul_ps(fscal,dy11);
1756             tz               = _mm256_mul_ps(fscal,dz11);
1757
1758             /* Update vectorial force */
1759             fix1             = _mm256_add_ps(fix1,tx);
1760             fiy1             = _mm256_add_ps(fiy1,ty);
1761             fiz1             = _mm256_add_ps(fiz1,tz);
1762
1763             fjx1             = _mm256_add_ps(fjx1,tx);
1764             fjy1             = _mm256_add_ps(fjy1,ty);
1765             fjz1             = _mm256_add_ps(fjz1,tz);
1766
1767             /**************************
1768              * CALCULATE INTERACTIONS *
1769              **************************/
1770
1771             /* REACTION-FIELD ELECTROSTATICS */
1772             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1773
1774             fscal            = felec;
1775
1776             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1777
1778             /* Calculate temporary vectorial force */
1779             tx               = _mm256_mul_ps(fscal,dx12);
1780             ty               = _mm256_mul_ps(fscal,dy12);
1781             tz               = _mm256_mul_ps(fscal,dz12);
1782
1783             /* Update vectorial force */
1784             fix1             = _mm256_add_ps(fix1,tx);
1785             fiy1             = _mm256_add_ps(fiy1,ty);
1786             fiz1             = _mm256_add_ps(fiz1,tz);
1787
1788             fjx2             = _mm256_add_ps(fjx2,tx);
1789             fjy2             = _mm256_add_ps(fjy2,ty);
1790             fjz2             = _mm256_add_ps(fjz2,tz);
1791
1792             /**************************
1793              * CALCULATE INTERACTIONS *
1794              **************************/
1795
1796             /* REACTION-FIELD ELECTROSTATICS */
1797             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1798
1799             fscal            = felec;
1800
1801             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1802
1803             /* Calculate temporary vectorial force */
1804             tx               = _mm256_mul_ps(fscal,dx13);
1805             ty               = _mm256_mul_ps(fscal,dy13);
1806             tz               = _mm256_mul_ps(fscal,dz13);
1807
1808             /* Update vectorial force */
1809             fix1             = _mm256_add_ps(fix1,tx);
1810             fiy1             = _mm256_add_ps(fiy1,ty);
1811             fiz1             = _mm256_add_ps(fiz1,tz);
1812
1813             fjx3             = _mm256_add_ps(fjx3,tx);
1814             fjy3             = _mm256_add_ps(fjy3,ty);
1815             fjz3             = _mm256_add_ps(fjz3,tz);
1816
1817             /**************************
1818              * CALCULATE INTERACTIONS *
1819              **************************/
1820
1821             /* REACTION-FIELD ELECTROSTATICS */
1822             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1823
1824             fscal            = felec;
1825
1826             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1827
1828             /* Calculate temporary vectorial force */
1829             tx               = _mm256_mul_ps(fscal,dx21);
1830             ty               = _mm256_mul_ps(fscal,dy21);
1831             tz               = _mm256_mul_ps(fscal,dz21);
1832
1833             /* Update vectorial force */
1834             fix2             = _mm256_add_ps(fix2,tx);
1835             fiy2             = _mm256_add_ps(fiy2,ty);
1836             fiz2             = _mm256_add_ps(fiz2,tz);
1837
1838             fjx1             = _mm256_add_ps(fjx1,tx);
1839             fjy1             = _mm256_add_ps(fjy1,ty);
1840             fjz1             = _mm256_add_ps(fjz1,tz);
1841
1842             /**************************
1843              * CALCULATE INTERACTIONS *
1844              **************************/
1845
1846             /* REACTION-FIELD ELECTROSTATICS */
1847             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1848
1849             fscal            = felec;
1850
1851             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1852
1853             /* Calculate temporary vectorial force */
1854             tx               = _mm256_mul_ps(fscal,dx22);
1855             ty               = _mm256_mul_ps(fscal,dy22);
1856             tz               = _mm256_mul_ps(fscal,dz22);
1857
1858             /* Update vectorial force */
1859             fix2             = _mm256_add_ps(fix2,tx);
1860             fiy2             = _mm256_add_ps(fiy2,ty);
1861             fiz2             = _mm256_add_ps(fiz2,tz);
1862
1863             fjx2             = _mm256_add_ps(fjx2,tx);
1864             fjy2             = _mm256_add_ps(fjy2,ty);
1865             fjz2             = _mm256_add_ps(fjz2,tz);
1866
1867             /**************************
1868              * CALCULATE INTERACTIONS *
1869              **************************/
1870
1871             /* REACTION-FIELD ELECTROSTATICS */
1872             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1873
1874             fscal            = felec;
1875
1876             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1877
1878             /* Calculate temporary vectorial force */
1879             tx               = _mm256_mul_ps(fscal,dx23);
1880             ty               = _mm256_mul_ps(fscal,dy23);
1881             tz               = _mm256_mul_ps(fscal,dz23);
1882
1883             /* Update vectorial force */
1884             fix2             = _mm256_add_ps(fix2,tx);
1885             fiy2             = _mm256_add_ps(fiy2,ty);
1886             fiz2             = _mm256_add_ps(fiz2,tz);
1887
1888             fjx3             = _mm256_add_ps(fjx3,tx);
1889             fjy3             = _mm256_add_ps(fjy3,ty);
1890             fjz3             = _mm256_add_ps(fjz3,tz);
1891
1892             /**************************
1893              * CALCULATE INTERACTIONS *
1894              **************************/
1895
1896             /* REACTION-FIELD ELECTROSTATICS */
1897             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1898
1899             fscal            = felec;
1900
1901             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1902
1903             /* Calculate temporary vectorial force */
1904             tx               = _mm256_mul_ps(fscal,dx31);
1905             ty               = _mm256_mul_ps(fscal,dy31);
1906             tz               = _mm256_mul_ps(fscal,dz31);
1907
1908             /* Update vectorial force */
1909             fix3             = _mm256_add_ps(fix3,tx);
1910             fiy3             = _mm256_add_ps(fiy3,ty);
1911             fiz3             = _mm256_add_ps(fiz3,tz);
1912
1913             fjx1             = _mm256_add_ps(fjx1,tx);
1914             fjy1             = _mm256_add_ps(fjy1,ty);
1915             fjz1             = _mm256_add_ps(fjz1,tz);
1916
1917             /**************************
1918              * CALCULATE INTERACTIONS *
1919              **************************/
1920
1921             /* REACTION-FIELD ELECTROSTATICS */
1922             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1923
1924             fscal            = felec;
1925
1926             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1927
1928             /* Calculate temporary vectorial force */
1929             tx               = _mm256_mul_ps(fscal,dx32);
1930             ty               = _mm256_mul_ps(fscal,dy32);
1931             tz               = _mm256_mul_ps(fscal,dz32);
1932
1933             /* Update vectorial force */
1934             fix3             = _mm256_add_ps(fix3,tx);
1935             fiy3             = _mm256_add_ps(fiy3,ty);
1936             fiz3             = _mm256_add_ps(fiz3,tz);
1937
1938             fjx2             = _mm256_add_ps(fjx2,tx);
1939             fjy2             = _mm256_add_ps(fjy2,ty);
1940             fjz2             = _mm256_add_ps(fjz2,tz);
1941
1942             /**************************
1943              * CALCULATE INTERACTIONS *
1944              **************************/
1945
1946             /* REACTION-FIELD ELECTROSTATICS */
1947             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1948
1949             fscal            = felec;
1950
1951             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1952
1953             /* Calculate temporary vectorial force */
1954             tx               = _mm256_mul_ps(fscal,dx33);
1955             ty               = _mm256_mul_ps(fscal,dy33);
1956             tz               = _mm256_mul_ps(fscal,dz33);
1957
1958             /* Update vectorial force */
1959             fix3             = _mm256_add_ps(fix3,tx);
1960             fiy3             = _mm256_add_ps(fiy3,ty);
1961             fiz3             = _mm256_add_ps(fiz3,tz);
1962
1963             fjx3             = _mm256_add_ps(fjx3,tx);
1964             fjy3             = _mm256_add_ps(fjy3,ty);
1965             fjz3             = _mm256_add_ps(fjz3,tz);
1966
1967             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1968             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1969             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1970             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1971             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1972             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1973             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1974             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1975
1976             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1977                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1978                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1979
1980             /* Inner loop uses 273 flops */
1981         }
1982
1983         /* End of innermost loop */
1984
1985         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1986                                                  f+i_coord_offset,fshift+i_shift_offset);
1987
1988         /* Increment number of inner iterations */
1989         inneriter                  += j_index_end - j_index_start;
1990
1991         /* Outer loop uses 24 flops */
1992     }
1993
1994     /* Increment number of outer iterations */
1995     outeriter        += nri;
1996
1997     /* Update outer/inner flops */
1998
1999     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*273);
2000 }