src/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEw_VdwLJ_GeomW3W3_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_ElecEw_VdwLJ_GeomW3W3_VF_avx_256_single
  38  * Electrostatics interaction: Ewald
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water3-Water3
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecEw_VdwLJ_GeomW3W3_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     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  79     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  80     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  81     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  82     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  83     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  84     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  85     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  86     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  87     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  88     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  89     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  90     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  91     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  92     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  93     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 100     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 101     __m256i          ewitab;
 102     __m128i          ewitab_lo,ewitab_hi;
 103     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 104     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 105     real             *ewtab;
 106     __m256           dummy_mask,cutoff_mask;
 107     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 108     __m256           one     = _mm256_set1_ps(1.0);
 109     __m256           two     = _mm256_set1_ps(2.0);
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = _mm256_set1_ps(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123     nvdwtype         = fr->ntype;
 124     vdwparam         = fr->nbfp;
 125     vdwtype          = mdatoms->typeA;
 126
 127     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
 128     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff);
 129     beta2            = _mm256_mul_ps(beta,beta);
 130     beta3            = _mm256_mul_ps(beta,beta2);
 131
 132     ewtab            = fr->ic->tabq_coul_FDV0;
 133     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
 134     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
 135
 136     /* Setup water-specific parameters */
 137     inr              = nlist->iinr[0];
 138     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 139     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 140     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 141     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 142
 143     jq0              = _mm256_set1_ps(charge[inr+0]);
 144     jq1              = _mm256_set1_ps(charge[inr+1]);
 145     jq2              = _mm256_set1_ps(charge[inr+2]);
 146     vdwjidx0A        = 2*vdwtype[inr+0];
 147     qq00             = _mm256_mul_ps(iq0,jq0);
 148     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 149     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 150     qq01             = _mm256_mul_ps(iq0,jq1);
 151     qq02             = _mm256_mul_ps(iq0,jq2);
 152     qq10             = _mm256_mul_ps(iq1,jq0);
 153     qq11             = _mm256_mul_ps(iq1,jq1);
 154     qq12             = _mm256_mul_ps(iq1,jq2);
 155     qq20             = _mm256_mul_ps(iq2,jq0);
 156     qq21             = _mm256_mul_ps(iq2,jq1);
 157     qq22             = _mm256_mul_ps(iq2,jq2);
 158
 159     /* Avoid stupid compiler warnings */
 160     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 161     j_coord_offsetA = 0;
 162     j_coord_offsetB = 0;
 163     j_coord_offsetC = 0;
 164     j_coord_offsetD = 0;
 165     j_coord_offsetE = 0;
 166     j_coord_offsetF = 0;
 167     j_coord_offsetG = 0;
 168     j_coord_offsetH = 0;
 169
 170     outeriter        = 0;
 171     inneriter        = 0;
 172
 173     for(iidx=0;iidx<4*DIM;iidx++)
 174     {
 175         scratch[iidx] = 0.0;
 176     }
 177
 178     /* Start outer loop over neighborlists */
 179     for(iidx=0; iidx<nri; iidx++)
 180     {
 181         /* Load shift vector for this list */
 182         i_shift_offset   = DIM*shiftidx[iidx];
 183
 184         /* Load limits for loop over neighbors */
 185         j_index_start    = jindex[iidx];
 186         j_index_end      = jindex[iidx+1];
 187
 188         /* Get outer coordinate index */
 189         inr              = iinr[iidx];
 190         i_coord_offset   = DIM*inr;
 191
 192         /* Load i particle coords and add shift vector */
 193         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 194                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 195
 196         fix0             = _mm256_setzero_ps();
 197         fiy0             = _mm256_setzero_ps();
 198         fiz0             = _mm256_setzero_ps();
 199         fix1             = _mm256_setzero_ps();
 200         fiy1             = _mm256_setzero_ps();
 201         fiz1             = _mm256_setzero_ps();
 202         fix2             = _mm256_setzero_ps();
 203         fiy2             = _mm256_setzero_ps();
 204         fiz2             = _mm256_setzero_ps();
 205
 206         /* Reset potential sums */
 207         velecsum         = _mm256_setzero_ps();
 208         vvdwsum          = _mm256_setzero_ps();
 209
 210         /* Start inner kernel loop */
 211         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 212         {
 213
 214             /* Get j neighbor index, and coordinate index */
 215             jnrA             = jjnr[jidx];
 216             jnrB             = jjnr[jidx+1];
 217             jnrC             = jjnr[jidx+2];
 218             jnrD             = jjnr[jidx+3];
 219             jnrE             = jjnr[jidx+4];
 220             jnrF             = jjnr[jidx+5];
 221             jnrG             = jjnr[jidx+6];
 222             jnrH             = jjnr[jidx+7];
 223             j_coord_offsetA  = DIM*jnrA;
 224             j_coord_offsetB  = DIM*jnrB;
 225             j_coord_offsetC  = DIM*jnrC;
 226             j_coord_offsetD  = DIM*jnrD;
 227             j_coord_offsetE  = DIM*jnrE;
 228             j_coord_offsetF  = DIM*jnrF;
 229             j_coord_offsetG  = DIM*jnrG;
 230             j_coord_offsetH  = DIM*jnrH;
 231
 232             /* load j atom coordinates */
 233             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 234                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 235                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 236                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 237                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 238
 239             /* Calculate displacement vector */
 240             dx00             = _mm256_sub_ps(ix0,jx0);
 241             dy00             = _mm256_sub_ps(iy0,jy0);
 242             dz00             = _mm256_sub_ps(iz0,jz0);
 243             dx01             = _mm256_sub_ps(ix0,jx1);
 244             dy01             = _mm256_sub_ps(iy0,jy1);
 245             dz01             = _mm256_sub_ps(iz0,jz1);
 246             dx02             = _mm256_sub_ps(ix0,jx2);
 247             dy02             = _mm256_sub_ps(iy0,jy2);
 248             dz02             = _mm256_sub_ps(iz0,jz2);
 249             dx10             = _mm256_sub_ps(ix1,jx0);
 250             dy10             = _mm256_sub_ps(iy1,jy0);
 251             dz10             = _mm256_sub_ps(iz1,jz0);
 252             dx11             = _mm256_sub_ps(ix1,jx1);
 253             dy11             = _mm256_sub_ps(iy1,jy1);
 254             dz11             = _mm256_sub_ps(iz1,jz1);
 255             dx12             = _mm256_sub_ps(ix1,jx2);
 256             dy12             = _mm256_sub_ps(iy1,jy2);
 257             dz12             = _mm256_sub_ps(iz1,jz2);
 258             dx20             = _mm256_sub_ps(ix2,jx0);
 259             dy20             = _mm256_sub_ps(iy2,jy0);
 260             dz20             = _mm256_sub_ps(iz2,jz0);
 261             dx21             = _mm256_sub_ps(ix2,jx1);
 262             dy21             = _mm256_sub_ps(iy2,jy1);
 263             dz21             = _mm256_sub_ps(iz2,jz1);
 264             dx22             = _mm256_sub_ps(ix2,jx2);
 265             dy22             = _mm256_sub_ps(iy2,jy2);
 266             dz22             = _mm256_sub_ps(iz2,jz2);
 267
 268             /* Calculate squared distance and things based on it */
 269             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 270             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 271             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 272             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 273             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 274             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 275             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 276             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 277             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 278
 279             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 280             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 281             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 282             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 283             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 284             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 285             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 286             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 287             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 288
 289             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 290             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 291             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 292             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 293             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 294             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 295             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 296             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 297             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 298
 299             fjx0             = _mm256_setzero_ps();
 300             fjy0             = _mm256_setzero_ps();
 301             fjz0             = _mm256_setzero_ps();
 302             fjx1             = _mm256_setzero_ps();
 303             fjy1             = _mm256_setzero_ps();
 304             fjz1             = _mm256_setzero_ps();
 305             fjx2             = _mm256_setzero_ps();
 306             fjy2             = _mm256_setzero_ps();
 307             fjz2             = _mm256_setzero_ps();
 308
 309             /**************************
 310              * CALCULATE INTERACTIONS *
 311              **************************/
 312
 313             r00              = _mm256_mul_ps(rsq00,rinv00);
 314
 315             /* EWALD ELECTROSTATICS */
 316
 317             /* Analytical PME correction */
 318             zeta2            = _mm256_mul_ps(beta2,rsq00);
 319             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
 320             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 321             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 322             felec            = _mm256_mul_ps(qq00,felec);
 323             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 324             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 325             velec            = _mm256_sub_ps(rinv00,pmecorrV);
 326             velec            = _mm256_mul_ps(qq00,velec);
 327
 328             /* LENNARD-JONES DISPERSION/REPULSION */
 329
 330             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 331             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 332             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 333             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 334             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 335
 336             /* Update potential sum for this i atom from the interaction with this j atom. */
 337             velecsum         = _mm256_add_ps(velecsum,velec);
 338             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 339
 340             fscal            = _mm256_add_ps(felec,fvdw);
 341
 342             /* Calculate temporary vectorial force */
 343             tx               = _mm256_mul_ps(fscal,dx00);
 344             ty               = _mm256_mul_ps(fscal,dy00);
 345             tz               = _mm256_mul_ps(fscal,dz00);
 346
 347             /* Update vectorial force */
 348             fix0             = _mm256_add_ps(fix0,tx);
 349             fiy0             = _mm256_add_ps(fiy0,ty);
 350             fiz0             = _mm256_add_ps(fiz0,tz);
 351
 352             fjx0             = _mm256_add_ps(fjx0,tx);
 353             fjy0             = _mm256_add_ps(fjy0,ty);
 354             fjz0             = _mm256_add_ps(fjz0,tz);
 355
 356             /**************************
 357              * CALCULATE INTERACTIONS *
 358              **************************/
 359
 360             r01              = _mm256_mul_ps(rsq01,rinv01);
 361
 362             /* EWALD ELECTROSTATICS */
 363
 364             /* Analytical PME correction */
 365             zeta2            = _mm256_mul_ps(beta2,rsq01);
 366             rinv3            = _mm256_mul_ps(rinvsq01,rinv01);
 367             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 368             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 369             felec            = _mm256_mul_ps(qq01,felec);
 370             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 371             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 372             velec            = _mm256_sub_ps(rinv01,pmecorrV);
 373             velec            = _mm256_mul_ps(qq01,velec);
 374
 375             /* Update potential sum for this i atom from the interaction with this j atom. */
 376             velecsum         = _mm256_add_ps(velecsum,velec);
 377
 378             fscal            = felec;
 379
 380             /* Calculate temporary vectorial force */
 381             tx               = _mm256_mul_ps(fscal,dx01);
 382             ty               = _mm256_mul_ps(fscal,dy01);
 383             tz               = _mm256_mul_ps(fscal,dz01);
 384
 385             /* Update vectorial force */
 386             fix0             = _mm256_add_ps(fix0,tx);
 387             fiy0             = _mm256_add_ps(fiy0,ty);
 388             fiz0             = _mm256_add_ps(fiz0,tz);
 389
 390             fjx1             = _mm256_add_ps(fjx1,tx);
 391             fjy1             = _mm256_add_ps(fjy1,ty);
 392             fjz1             = _mm256_add_ps(fjz1,tz);
 393
 394             /**************************
 395              * CALCULATE INTERACTIONS *
 396              **************************/
 397
 398             r02              = _mm256_mul_ps(rsq02,rinv02);
 399
 400             /* EWALD ELECTROSTATICS */
 401
 402             /* Analytical PME correction */
 403             zeta2            = _mm256_mul_ps(beta2,rsq02);
 404             rinv3            = _mm256_mul_ps(rinvsq02,rinv02);
 405             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 406             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 407             felec            = _mm256_mul_ps(qq02,felec);
 408             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 409             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 410             velec            = _mm256_sub_ps(rinv02,pmecorrV);
 411             velec            = _mm256_mul_ps(qq02,velec);
 412
 413             /* Update potential sum for this i atom from the interaction with this j atom. */
 414             velecsum         = _mm256_add_ps(velecsum,velec);
 415
 416             fscal            = felec;
 417
 418             /* Calculate temporary vectorial force */
 419             tx               = _mm256_mul_ps(fscal,dx02);
 420             ty               = _mm256_mul_ps(fscal,dy02);
 421             tz               = _mm256_mul_ps(fscal,dz02);
 422
 423             /* Update vectorial force */
 424             fix0             = _mm256_add_ps(fix0,tx);
 425             fiy0             = _mm256_add_ps(fiy0,ty);
 426             fiz0             = _mm256_add_ps(fiz0,tz);
 427
 428             fjx2             = _mm256_add_ps(fjx2,tx);
 429             fjy2             = _mm256_add_ps(fjy2,ty);
 430             fjz2             = _mm256_add_ps(fjz2,tz);
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             r10              = _mm256_mul_ps(rsq10,rinv10);
 437
 438             /* EWALD ELECTROSTATICS */
 439
 440             /* Analytical PME correction */
 441             zeta2            = _mm256_mul_ps(beta2,rsq10);
 442             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
 443             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 444             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 445             felec            = _mm256_mul_ps(qq10,felec);
 446             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 447             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 448             velec            = _mm256_sub_ps(rinv10,pmecorrV);
 449             velec            = _mm256_mul_ps(qq10,velec);
 450
 451             /* Update potential sum for this i atom from the interaction with this j atom. */
 452             velecsum         = _mm256_add_ps(velecsum,velec);
 453
 454             fscal            = felec;
 455
 456             /* Calculate temporary vectorial force */
 457             tx               = _mm256_mul_ps(fscal,dx10);
 458             ty               = _mm256_mul_ps(fscal,dy10);
 459             tz               = _mm256_mul_ps(fscal,dz10);
 460
 461             /* Update vectorial force */
 462             fix1             = _mm256_add_ps(fix1,tx);
 463             fiy1             = _mm256_add_ps(fiy1,ty);
 464             fiz1             = _mm256_add_ps(fiz1,tz);
 465
 466             fjx0             = _mm256_add_ps(fjx0,tx);
 467             fjy0             = _mm256_add_ps(fjy0,ty);
 468             fjz0             = _mm256_add_ps(fjz0,tz);
 469
 470             /**************************
 471              * CALCULATE INTERACTIONS *
 472              **************************/
 473
 474             r11              = _mm256_mul_ps(rsq11,rinv11);
 475
 476             /* EWALD ELECTROSTATICS */
 477
 478             /* Analytical PME correction */
 479             zeta2            = _mm256_mul_ps(beta2,rsq11);
 480             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 481             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 482             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 483             felec            = _mm256_mul_ps(qq11,felec);
 484             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 485             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 486             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 487             velec            = _mm256_mul_ps(qq11,velec);
 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,dx11);
 496             ty               = _mm256_mul_ps(fscal,dy11);
 497             tz               = _mm256_mul_ps(fscal,dz11);
 498
 499             /* Update vectorial force */
 500             fix1             = _mm256_add_ps(fix1,tx);
 501             fiy1             = _mm256_add_ps(fiy1,ty);
 502             fiz1             = _mm256_add_ps(fiz1,tz);
 503
 504             fjx1             = _mm256_add_ps(fjx1,tx);
 505             fjy1             = _mm256_add_ps(fjy1,ty);
 506             fjz1             = _mm256_add_ps(fjz1,tz);
 507
 508             /**************************
 509              * CALCULATE INTERACTIONS *
 510              **************************/
 511
 512             r12              = _mm256_mul_ps(rsq12,rinv12);
 513
 514             /* EWALD ELECTROSTATICS */
 515
 516             /* Analytical PME correction */
 517             zeta2            = _mm256_mul_ps(beta2,rsq12);
 518             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
 519             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 520             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 521             felec            = _mm256_mul_ps(qq12,felec);
 522             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 523             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 524             velec            = _mm256_sub_ps(rinv12,pmecorrV);
 525             velec            = _mm256_mul_ps(qq12,velec);
 526
 527             /* Update potential sum for this i atom from the interaction with this j atom. */
 528             velecsum         = _mm256_add_ps(velecsum,velec);
 529
 530             fscal            = felec;
 531
 532             /* Calculate temporary vectorial force */
 533             tx               = _mm256_mul_ps(fscal,dx12);
 534             ty               = _mm256_mul_ps(fscal,dy12);
 535             tz               = _mm256_mul_ps(fscal,dz12);
 536
 537             /* Update vectorial force */
 538             fix1             = _mm256_add_ps(fix1,tx);
 539             fiy1             = _mm256_add_ps(fiy1,ty);
 540             fiz1             = _mm256_add_ps(fiz1,tz);
 541
 542             fjx2             = _mm256_add_ps(fjx2,tx);
 543             fjy2             = _mm256_add_ps(fjy2,ty);
 544             fjz2             = _mm256_add_ps(fjz2,tz);
 545
 546             /**************************
 547              * CALCULATE INTERACTIONS *
 548              **************************/
 549
 550             r20              = _mm256_mul_ps(rsq20,rinv20);
 551
 552             /* EWALD ELECTROSTATICS */
 553
 554             /* Analytical PME correction */
 555             zeta2            = _mm256_mul_ps(beta2,rsq20);
 556             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
 557             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 558             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 559             felec            = _mm256_mul_ps(qq20,felec);
 560             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 561             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 562             velec            = _mm256_sub_ps(rinv20,pmecorrV);
 563             velec            = _mm256_mul_ps(qq20,velec);
 564
 565             /* Update potential sum for this i atom from the interaction with this j atom. */
 566             velecsum         = _mm256_add_ps(velecsum,velec);
 567
 568             fscal            = felec;
 569
 570             /* Calculate temporary vectorial force */
 571             tx               = _mm256_mul_ps(fscal,dx20);
 572             ty               = _mm256_mul_ps(fscal,dy20);
 573             tz               = _mm256_mul_ps(fscal,dz20);
 574
 575             /* Update vectorial force */
 576             fix2             = _mm256_add_ps(fix2,tx);
 577             fiy2             = _mm256_add_ps(fiy2,ty);
 578             fiz2             = _mm256_add_ps(fiz2,tz);
 579
 580             fjx0             = _mm256_add_ps(fjx0,tx);
 581             fjy0             = _mm256_add_ps(fjy0,ty);
 582             fjz0             = _mm256_add_ps(fjz0,tz);
 583
 584             /**************************
 585              * CALCULATE INTERACTIONS *
 586              **************************/
 587
 588             r21              = _mm256_mul_ps(rsq21,rinv21);
 589
 590             /* EWALD ELECTROSTATICS */
 591
 592             /* Analytical PME correction */
 593             zeta2            = _mm256_mul_ps(beta2,rsq21);
 594             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
 595             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 596             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 597             felec            = _mm256_mul_ps(qq21,felec);
 598             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 599             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 600             velec            = _mm256_sub_ps(rinv21,pmecorrV);
 601             velec            = _mm256_mul_ps(qq21,velec);
 602
 603             /* Update potential sum for this i atom from the interaction with this j atom. */
 604             velecsum         = _mm256_add_ps(velecsum,velec);
 605
 606             fscal            = felec;
 607
 608             /* Calculate temporary vectorial force */
 609             tx               = _mm256_mul_ps(fscal,dx21);
 610             ty               = _mm256_mul_ps(fscal,dy21);
 611             tz               = _mm256_mul_ps(fscal,dz21);
 612
 613             /* Update vectorial force */
 614             fix2             = _mm256_add_ps(fix2,tx);
 615             fiy2             = _mm256_add_ps(fiy2,ty);
 616             fiz2             = _mm256_add_ps(fiz2,tz);
 617
 618             fjx1             = _mm256_add_ps(fjx1,tx);
 619             fjy1             = _mm256_add_ps(fjy1,ty);
 620             fjz1             = _mm256_add_ps(fjz1,tz);
 621
 622             /**************************
 623              * CALCULATE INTERACTIONS *
 624              **************************/
 625
 626             r22              = _mm256_mul_ps(rsq22,rinv22);
 627
 628             /* EWALD ELECTROSTATICS */
 629
 630             /* Analytical PME correction */
 631             zeta2            = _mm256_mul_ps(beta2,rsq22);
 632             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
 633             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 634             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 635             felec            = _mm256_mul_ps(qq22,felec);
 636             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 637             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 638             velec            = _mm256_sub_ps(rinv22,pmecorrV);
 639             velec            = _mm256_mul_ps(qq22,velec);
 640
 641             /* Update potential sum for this i atom from the interaction with this j atom. */
 642             velecsum         = _mm256_add_ps(velecsum,velec);
 643
 644             fscal            = felec;
 645
 646             /* Calculate temporary vectorial force */
 647             tx               = _mm256_mul_ps(fscal,dx22);
 648             ty               = _mm256_mul_ps(fscal,dy22);
 649             tz               = _mm256_mul_ps(fscal,dz22);
 650
 651             /* Update vectorial force */
 652             fix2             = _mm256_add_ps(fix2,tx);
 653             fiy2             = _mm256_add_ps(fiy2,ty);
 654             fiz2             = _mm256_add_ps(fiz2,tz);
 655
 656             fjx2             = _mm256_add_ps(fjx2,tx);
 657             fjy2             = _mm256_add_ps(fjy2,ty);
 658             fjz2             = _mm256_add_ps(fjz2,tz);
 659
 660             fjptrA             = f+j_coord_offsetA;
 661             fjptrB             = f+j_coord_offsetB;
 662             fjptrC             = f+j_coord_offsetC;
 663             fjptrD             = f+j_coord_offsetD;
 664             fjptrE             = f+j_coord_offsetE;
 665             fjptrF             = f+j_coord_offsetF;
 666             fjptrG             = f+j_coord_offsetG;
 667             fjptrH             = f+j_coord_offsetH;
 668
 669             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 670                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 671
 672             /* Inner loop uses 768 flops */
 673         }
 674
 675         if(jidx<j_index_end)
 676         {
 677
 678             /* Get j neighbor index, and coordinate index */
 679             jnrlistA         = jjnr[jidx];
 680             jnrlistB         = jjnr[jidx+1];
 681             jnrlistC         = jjnr[jidx+2];
 682             jnrlistD         = jjnr[jidx+3];
 683             jnrlistE         = jjnr[jidx+4];
 684             jnrlistF         = jjnr[jidx+5];
 685             jnrlistG         = jjnr[jidx+6];
 686             jnrlistH         = jjnr[jidx+7];
 687             /* Sign of each element will be negative for non-real atoms.
 688              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 689              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 690              */
 691             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 692                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 693
 694             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 695             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 696             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 697             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 698             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 699             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 700             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 701             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 702             j_coord_offsetA  = DIM*jnrA;
 703             j_coord_offsetB  = DIM*jnrB;
 704             j_coord_offsetC  = DIM*jnrC;
 705             j_coord_offsetD  = DIM*jnrD;
 706             j_coord_offsetE  = DIM*jnrE;
 707             j_coord_offsetF  = DIM*jnrF;
 708             j_coord_offsetG  = DIM*jnrG;
 709             j_coord_offsetH  = DIM*jnrH;
 710
 711             /* load j atom coordinates */
 712             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 713                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 714                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 715                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 716                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 717
 718             /* Calculate displacement vector */
 719             dx00             = _mm256_sub_ps(ix0,jx0);
 720             dy00             = _mm256_sub_ps(iy0,jy0);
 721             dz00             = _mm256_sub_ps(iz0,jz0);
 722             dx01             = _mm256_sub_ps(ix0,jx1);
 723             dy01             = _mm256_sub_ps(iy0,jy1);
 724             dz01             = _mm256_sub_ps(iz0,jz1);
 725             dx02             = _mm256_sub_ps(ix0,jx2);
 726             dy02             = _mm256_sub_ps(iy0,jy2);
 727             dz02             = _mm256_sub_ps(iz0,jz2);
 728             dx10             = _mm256_sub_ps(ix1,jx0);
 729             dy10             = _mm256_sub_ps(iy1,jy0);
 730             dz10             = _mm256_sub_ps(iz1,jz0);
 731             dx11             = _mm256_sub_ps(ix1,jx1);
 732             dy11             = _mm256_sub_ps(iy1,jy1);
 733             dz11             = _mm256_sub_ps(iz1,jz1);
 734             dx12             = _mm256_sub_ps(ix1,jx2);
 735             dy12             = _mm256_sub_ps(iy1,jy2);
 736             dz12             = _mm256_sub_ps(iz1,jz2);
 737             dx20             = _mm256_sub_ps(ix2,jx0);
 738             dy20             = _mm256_sub_ps(iy2,jy0);
 739             dz20             = _mm256_sub_ps(iz2,jz0);
 740             dx21             = _mm256_sub_ps(ix2,jx1);
 741             dy21             = _mm256_sub_ps(iy2,jy1);
 742             dz21             = _mm256_sub_ps(iz2,jz1);
 743             dx22             = _mm256_sub_ps(ix2,jx2);
 744             dy22             = _mm256_sub_ps(iy2,jy2);
 745             dz22             = _mm256_sub_ps(iz2,jz2);
 746
 747             /* Calculate squared distance and things based on it */
 748             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 749             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 750             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 751             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 752             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 753             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 754             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 755             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 756             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 757
 758             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 759             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 760             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 761             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 762             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 763             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 764             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 765             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 766             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 767
 768             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 769             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 770             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 771             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 772             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 773             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 774             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 775             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 776             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 777
 778             fjx0             = _mm256_setzero_ps();
 779             fjy0             = _mm256_setzero_ps();
 780             fjz0             = _mm256_setzero_ps();
 781             fjx1             = _mm256_setzero_ps();
 782             fjy1             = _mm256_setzero_ps();
 783             fjz1             = _mm256_setzero_ps();
 784             fjx2             = _mm256_setzero_ps();
 785             fjy2             = _mm256_setzero_ps();
 786             fjz2             = _mm256_setzero_ps();
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             r00              = _mm256_mul_ps(rsq00,rinv00);
 793             r00              = _mm256_andnot_ps(dummy_mask,r00);
 794
 795             /* EWALD ELECTROSTATICS */
 796
 797             /* Analytical PME correction */
 798             zeta2            = _mm256_mul_ps(beta2,rsq00);
 799             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
 800             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 801             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 802             felec            = _mm256_mul_ps(qq00,felec);
 803             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 804             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 805             velec            = _mm256_sub_ps(rinv00,pmecorrV);
 806             velec            = _mm256_mul_ps(qq00,velec);
 807
 808             /* LENNARD-JONES DISPERSION/REPULSION */
 809
 810             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 811             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 812             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 813             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 814             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 815
 816             /* Update potential sum for this i atom from the interaction with this j atom. */
 817             velec            = _mm256_andnot_ps(dummy_mask,velec);
 818             velecsum         = _mm256_add_ps(velecsum,velec);
 819             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 820             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 821
 822             fscal            = _mm256_add_ps(felec,fvdw);
 823
 824             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 825
 826             /* Calculate temporary vectorial force */
 827             tx               = _mm256_mul_ps(fscal,dx00);
 828             ty               = _mm256_mul_ps(fscal,dy00);
 829             tz               = _mm256_mul_ps(fscal,dz00);
 830
 831             /* Update vectorial force */
 832             fix0             = _mm256_add_ps(fix0,tx);
 833             fiy0             = _mm256_add_ps(fiy0,ty);
 834             fiz0             = _mm256_add_ps(fiz0,tz);
 835
 836             fjx0             = _mm256_add_ps(fjx0,tx);
 837             fjy0             = _mm256_add_ps(fjy0,ty);
 838             fjz0             = _mm256_add_ps(fjz0,tz);
 839
 840             /**************************
 841              * CALCULATE INTERACTIONS *
 842              **************************/
 843
 844             r01              = _mm256_mul_ps(rsq01,rinv01);
 845             r01              = _mm256_andnot_ps(dummy_mask,r01);
 846
 847             /* EWALD ELECTROSTATICS */
 848
 849             /* Analytical PME correction */
 850             zeta2            = _mm256_mul_ps(beta2,rsq01);
 851             rinv3            = _mm256_mul_ps(rinvsq01,rinv01);
 852             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 853             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 854             felec            = _mm256_mul_ps(qq01,felec);
 855             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 856             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 857             velec            = _mm256_sub_ps(rinv01,pmecorrV);
 858             velec            = _mm256_mul_ps(qq01,velec);
 859
 860             /* Update potential sum for this i atom from the interaction with this j atom. */
 861             velec            = _mm256_andnot_ps(dummy_mask,velec);
 862             velecsum         = _mm256_add_ps(velecsum,velec);
 863
 864             fscal            = felec;
 865
 866             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 867
 868             /* Calculate temporary vectorial force */
 869             tx               = _mm256_mul_ps(fscal,dx01);
 870             ty               = _mm256_mul_ps(fscal,dy01);
 871             tz               = _mm256_mul_ps(fscal,dz01);
 872
 873             /* Update vectorial force */
 874             fix0             = _mm256_add_ps(fix0,tx);
 875             fiy0             = _mm256_add_ps(fiy0,ty);
 876             fiz0             = _mm256_add_ps(fiz0,tz);
 877
 878             fjx1             = _mm256_add_ps(fjx1,tx);
 879             fjy1             = _mm256_add_ps(fjy1,ty);
 880             fjz1             = _mm256_add_ps(fjz1,tz);
 881
 882             /**************************
 883              * CALCULATE INTERACTIONS *
 884              **************************/
 885
 886             r02              = _mm256_mul_ps(rsq02,rinv02);
 887             r02              = _mm256_andnot_ps(dummy_mask,r02);
 888
 889             /* EWALD ELECTROSTATICS */
 890
 891             /* Analytical PME correction */
 892             zeta2            = _mm256_mul_ps(beta2,rsq02);
 893             rinv3            = _mm256_mul_ps(rinvsq02,rinv02);
 894             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 895             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 896             felec            = _mm256_mul_ps(qq02,felec);
 897             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 898             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 899             velec            = _mm256_sub_ps(rinv02,pmecorrV);
 900             velec            = _mm256_mul_ps(qq02,velec);
 901
 902             /* Update potential sum for this i atom from the interaction with this j atom. */
 903             velec            = _mm256_andnot_ps(dummy_mask,velec);
 904             velecsum         = _mm256_add_ps(velecsum,velec);
 905
 906             fscal            = felec;
 907
 908             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 909
 910             /* Calculate temporary vectorial force */
 911             tx               = _mm256_mul_ps(fscal,dx02);
 912             ty               = _mm256_mul_ps(fscal,dy02);
 913             tz               = _mm256_mul_ps(fscal,dz02);
 914
 915             /* Update vectorial force */
 916             fix0             = _mm256_add_ps(fix0,tx);
 917             fiy0             = _mm256_add_ps(fiy0,ty);
 918             fiz0             = _mm256_add_ps(fiz0,tz);
 919
 920             fjx2             = _mm256_add_ps(fjx2,tx);
 921             fjy2             = _mm256_add_ps(fjy2,ty);
 922             fjz2             = _mm256_add_ps(fjz2,tz);
 923
 924             /**************************
 925              * CALCULATE INTERACTIONS *
 926              **************************/
 927
 928             r10              = _mm256_mul_ps(rsq10,rinv10);
 929             r10              = _mm256_andnot_ps(dummy_mask,r10);
 930
 931             /* EWALD ELECTROSTATICS */
 932
 933             /* Analytical PME correction */
 934             zeta2            = _mm256_mul_ps(beta2,rsq10);
 935             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
 936             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 937             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 938             felec            = _mm256_mul_ps(qq10,felec);
 939             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 940             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 941             velec            = _mm256_sub_ps(rinv10,pmecorrV);
 942             velec            = _mm256_mul_ps(qq10,velec);
 943
 944             /* Update potential sum for this i atom from the interaction with this j atom. */
 945             velec            = _mm256_andnot_ps(dummy_mask,velec);
 946             velecsum         = _mm256_add_ps(velecsum,velec);
 947
 948             fscal            = felec;
 949
 950             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 951
 952             /* Calculate temporary vectorial force */
 953             tx               = _mm256_mul_ps(fscal,dx10);
 954             ty               = _mm256_mul_ps(fscal,dy10);
 955             tz               = _mm256_mul_ps(fscal,dz10);
 956
 957             /* Update vectorial force */
 958             fix1             = _mm256_add_ps(fix1,tx);
 959             fiy1             = _mm256_add_ps(fiy1,ty);
 960             fiz1             = _mm256_add_ps(fiz1,tz);
 961
 962             fjx0             = _mm256_add_ps(fjx0,tx);
 963             fjy0             = _mm256_add_ps(fjy0,ty);
 964             fjz0             = _mm256_add_ps(fjz0,tz);
 965
 966             /**************************
 967              * CALCULATE INTERACTIONS *
 968              **************************/
 969
 970             r11              = _mm256_mul_ps(rsq11,rinv11);
 971             r11              = _mm256_andnot_ps(dummy_mask,r11);
 972
 973             /* EWALD ELECTROSTATICS */
 974
 975             /* Analytical PME correction */
 976             zeta2            = _mm256_mul_ps(beta2,rsq11);
 977             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
 978             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
 979             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
 980             felec            = _mm256_mul_ps(qq11,felec);
 981             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
 982             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
 983             velec            = _mm256_sub_ps(rinv11,pmecorrV);
 984             velec            = _mm256_mul_ps(qq11,velec);
 985
 986             /* Update potential sum for this i atom from the interaction with this j atom. */
 987             velec            = _mm256_andnot_ps(dummy_mask,velec);
 988             velecsum         = _mm256_add_ps(velecsum,velec);
 989
 990             fscal            = felec;
 991
 992             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 993
 994             /* Calculate temporary vectorial force */
 995             tx               = _mm256_mul_ps(fscal,dx11);
 996             ty               = _mm256_mul_ps(fscal,dy11);
 997             tz               = _mm256_mul_ps(fscal,dz11);
 998
 999             /* Update vectorial force */
1000             fix1             = _mm256_add_ps(fix1,tx);
1001             fiy1             = _mm256_add_ps(fiy1,ty);
1002             fiz1             = _mm256_add_ps(fiz1,tz);
1003
1004             fjx1             = _mm256_add_ps(fjx1,tx);
1005             fjy1             = _mm256_add_ps(fjy1,ty);
1006             fjz1             = _mm256_add_ps(fjz1,tz);
1007
1008             /**************************
1009              * CALCULATE INTERACTIONS *
1010              **************************/
1011
1012             r12              = _mm256_mul_ps(rsq12,rinv12);
1013             r12              = _mm256_andnot_ps(dummy_mask,r12);
1014
1015             /* EWALD ELECTROSTATICS */
1016
1017             /* Analytical PME correction */
1018             zeta2            = _mm256_mul_ps(beta2,rsq12);
1019             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
1020             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1021             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1022             felec            = _mm256_mul_ps(qq12,felec);
1023             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1024             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1025             velec            = _mm256_sub_ps(rinv12,pmecorrV);
1026             velec            = _mm256_mul_ps(qq12,velec);
1027
1028             /* Update potential sum for this i atom from the interaction with this j atom. */
1029             velec            = _mm256_andnot_ps(dummy_mask,velec);
1030             velecsum         = _mm256_add_ps(velecsum,velec);
1031
1032             fscal            = felec;
1033
1034             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1035
1036             /* Calculate temporary vectorial force */
1037             tx               = _mm256_mul_ps(fscal,dx12);
1038             ty               = _mm256_mul_ps(fscal,dy12);
1039             tz               = _mm256_mul_ps(fscal,dz12);
1040
1041             /* Update vectorial force */
1042             fix1             = _mm256_add_ps(fix1,tx);
1043             fiy1             = _mm256_add_ps(fiy1,ty);
1044             fiz1             = _mm256_add_ps(fiz1,tz);
1045
1046             fjx2             = _mm256_add_ps(fjx2,tx);
1047             fjy2             = _mm256_add_ps(fjy2,ty);
1048             fjz2             = _mm256_add_ps(fjz2,tz);
1049
1050             /**************************
1051              * CALCULATE INTERACTIONS *
1052              **************************/
1053
1054             r20              = _mm256_mul_ps(rsq20,rinv20);
1055             r20              = _mm256_andnot_ps(dummy_mask,r20);
1056
1057             /* EWALD ELECTROSTATICS */
1058
1059             /* Analytical PME correction */
1060             zeta2            = _mm256_mul_ps(beta2,rsq20);
1061             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
1062             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1063             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1064             felec            = _mm256_mul_ps(qq20,felec);
1065             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1066             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1067             velec            = _mm256_sub_ps(rinv20,pmecorrV);
1068             velec            = _mm256_mul_ps(qq20,velec);
1069
1070             /* Update potential sum for this i atom from the interaction with this j atom. */
1071             velec            = _mm256_andnot_ps(dummy_mask,velec);
1072             velecsum         = _mm256_add_ps(velecsum,velec);
1073
1074             fscal            = felec;
1075
1076             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1077
1078             /* Calculate temporary vectorial force */
1079             tx               = _mm256_mul_ps(fscal,dx20);
1080             ty               = _mm256_mul_ps(fscal,dy20);
1081             tz               = _mm256_mul_ps(fscal,dz20);
1082
1083             /* Update vectorial force */
1084             fix2             = _mm256_add_ps(fix2,tx);
1085             fiy2             = _mm256_add_ps(fiy2,ty);
1086             fiz2             = _mm256_add_ps(fiz2,tz);
1087
1088             fjx0             = _mm256_add_ps(fjx0,tx);
1089             fjy0             = _mm256_add_ps(fjy0,ty);
1090             fjz0             = _mm256_add_ps(fjz0,tz);
1091
1092             /**************************
1093              * CALCULATE INTERACTIONS *
1094              **************************/
1095
1096             r21              = _mm256_mul_ps(rsq21,rinv21);
1097             r21              = _mm256_andnot_ps(dummy_mask,r21);
1098
1099             /* EWALD ELECTROSTATICS */
1100
1101             /* Analytical PME correction */
1102             zeta2            = _mm256_mul_ps(beta2,rsq21);
1103             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1104             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1105             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1106             felec            = _mm256_mul_ps(qq21,felec);
1107             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1108             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1109             velec            = _mm256_sub_ps(rinv21,pmecorrV);
1110             velec            = _mm256_mul_ps(qq21,velec);
1111
1112             /* Update potential sum for this i atom from the interaction with this j atom. */
1113             velec            = _mm256_andnot_ps(dummy_mask,velec);
1114             velecsum         = _mm256_add_ps(velecsum,velec);
1115
1116             fscal            = felec;
1117
1118             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1119
1120             /* Calculate temporary vectorial force */
1121             tx               = _mm256_mul_ps(fscal,dx21);
1122             ty               = _mm256_mul_ps(fscal,dy21);
1123             tz               = _mm256_mul_ps(fscal,dz21);
1124
1125             /* Update vectorial force */
1126             fix2             = _mm256_add_ps(fix2,tx);
1127             fiy2             = _mm256_add_ps(fiy2,ty);
1128             fiz2             = _mm256_add_ps(fiz2,tz);
1129
1130             fjx1             = _mm256_add_ps(fjx1,tx);
1131             fjy1             = _mm256_add_ps(fjy1,ty);
1132             fjz1             = _mm256_add_ps(fjz1,tz);
1133
1134             /**************************
1135              * CALCULATE INTERACTIONS *
1136              **************************/
1137
1138             r22              = _mm256_mul_ps(rsq22,rinv22);
1139             r22              = _mm256_andnot_ps(dummy_mask,r22);
1140
1141             /* EWALD ELECTROSTATICS */
1142
1143             /* Analytical PME correction */
1144             zeta2            = _mm256_mul_ps(beta2,rsq22);
1145             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1146             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1147             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1148             felec            = _mm256_mul_ps(qq22,felec);
1149             pmecorrV         = gmx_mm256_pmecorrV_ps(zeta2);
1150             pmecorrV         = _mm256_mul_ps(pmecorrV,beta);
1151             velec            = _mm256_sub_ps(rinv22,pmecorrV);
1152             velec            = _mm256_mul_ps(qq22,velec);
1153
1154             /* Update potential sum for this i atom from the interaction with this j atom. */
1155             velec            = _mm256_andnot_ps(dummy_mask,velec);
1156             velecsum         = _mm256_add_ps(velecsum,velec);
1157
1158             fscal            = felec;
1159
1160             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1161
1162             /* Calculate temporary vectorial force */
1163             tx               = _mm256_mul_ps(fscal,dx22);
1164             ty               = _mm256_mul_ps(fscal,dy22);
1165             tz               = _mm256_mul_ps(fscal,dz22);
1166
1167             /* Update vectorial force */
1168             fix2             = _mm256_add_ps(fix2,tx);
1169             fiy2             = _mm256_add_ps(fiy2,ty);
1170             fiz2             = _mm256_add_ps(fiz2,tz);
1171
1172             fjx2             = _mm256_add_ps(fjx2,tx);
1173             fjy2             = _mm256_add_ps(fjy2,ty);
1174             fjz2             = _mm256_add_ps(fjz2,tz);
1175
1176             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1177             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1178             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1179             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1180             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1181             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1182             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1183             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1184
1185             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1186                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1187
1188             /* Inner loop uses 777 flops */
1189         }
1190
1191         /* End of innermost loop */
1192
1193         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1194                                                  f+i_coord_offset,fshift+i_shift_offset);
1195
1196         ggid                        = gid[iidx];
1197         /* Update potential energies */
1198         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1199         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1200
1201         /* Increment number of inner iterations */
1202         inneriter                  += j_index_end - j_index_start;
1203
1204         /* Outer loop uses 20 flops */
1205     }
1206
1207     /* Increment number of outer iterations */
1208     outeriter        += nri;
1209
1210     /* Update outer/inner flops */
1211
1212     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*777);
1213 }
1214 /*
1215  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW3W3_F_avx_256_single
1216  * Electrostatics interaction: Ewald
1217  * VdW interaction:            LennardJones
1218  * Geometry:                   Water3-Water3
1219  * Calculate force/pot:        Force
1220  */
1221 void
1222 nb_kernel_ElecEw_VdwLJ_GeomW3W3_F_avx_256_single
1223                     (t_nblist * gmx_restrict                nlist,
1224                      rvec * gmx_restrict                    xx,
1225                      rvec * gmx_restrict                    ff,
1226                      t_forcerec * gmx_restrict              fr,
1227                      t_mdatoms * gmx_restrict               mdatoms,
1228                      nb_kernel_data_t * gmx_restrict        kernel_data,
1229                      t_nrnb * gmx_restrict                  nrnb)
1230 {
1231     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1232      * just 0 for non-waters.
1233      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1234      * jnr indices corresponding to data put in the four positions in the SIMD register.
1235      */
1236     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1237     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1238     int              jnrA,jnrB,jnrC,jnrD;
1239     int              jnrE,jnrF,jnrG,jnrH;
1240     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1241     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1242     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1243     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1244     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1245     real             rcutoff_scalar;
1246     real             *shiftvec,*fshift,*x,*f;
1247     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1248     real             scratch[4*DIM];
1249     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1250     real *           vdwioffsetptr0;
1251     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1252     real *           vdwioffsetptr1;
1253     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1254     real *           vdwioffsetptr2;
1255     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1256     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1257     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1258     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1259     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1260     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1261     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1262     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1263     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1264     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1265     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1266     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1267     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1268     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1269     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1270     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1271     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1272     real             *charge;
1273     int              nvdwtype;
1274     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1275     int              *vdwtype;
1276     real             *vdwparam;
1277     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1278     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1279     __m256i          ewitab;
1280     __m128i          ewitab_lo,ewitab_hi;
1281     __m256           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1282     __m256           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
1283     real             *ewtab;
1284     __m256           dummy_mask,cutoff_mask;
1285     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1286     __m256           one     = _mm256_set1_ps(1.0);
1287     __m256           two     = _mm256_set1_ps(2.0);
1288     x                = xx[0];
1289     f                = ff[0];
1290
1291     nri              = nlist->nri;
1292     iinr             = nlist->iinr;
1293     jindex           = nlist->jindex;
1294     jjnr             = nlist->jjnr;
1295     shiftidx         = nlist->shift;
1296     gid              = nlist->gid;
1297     shiftvec         = fr->shift_vec[0];
1298     fshift           = fr->fshift[0];
1299     facel            = _mm256_set1_ps(fr->epsfac);
1300     charge           = mdatoms->chargeA;
1301     nvdwtype         = fr->ntype;
1302     vdwparam         = fr->nbfp;
1303     vdwtype          = mdatoms->typeA;
1304
1305     sh_ewald         = _mm256_set1_ps(fr->ic->sh_ewald);
1306     beta             = _mm256_set1_ps(fr->ic->ewaldcoeff);
1307     beta2            = _mm256_mul_ps(beta,beta);
1308     beta3            = _mm256_mul_ps(beta,beta2);
1309
1310     ewtab            = fr->ic->tabq_coul_F;
1311     ewtabscale       = _mm256_set1_ps(fr->ic->tabq_scale);
1312     ewtabhalfspace   = _mm256_set1_ps(0.5/fr->ic->tabq_scale);
1313
1314     /* Setup water-specific parameters */
1315     inr              = nlist->iinr[0];
1316     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1317     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1318     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1319     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1320
1321     jq0              = _mm256_set1_ps(charge[inr+0]);
1322     jq1              = _mm256_set1_ps(charge[inr+1]);
1323     jq2              = _mm256_set1_ps(charge[inr+2]);
1324     vdwjidx0A        = 2*vdwtype[inr+0];
1325     qq00             = _mm256_mul_ps(iq0,jq0);
1326     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1327     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1328     qq01             = _mm256_mul_ps(iq0,jq1);
1329     qq02             = _mm256_mul_ps(iq0,jq2);
1330     qq10             = _mm256_mul_ps(iq1,jq0);
1331     qq11             = _mm256_mul_ps(iq1,jq1);
1332     qq12             = _mm256_mul_ps(iq1,jq2);
1333     qq20             = _mm256_mul_ps(iq2,jq0);
1334     qq21             = _mm256_mul_ps(iq2,jq1);
1335     qq22             = _mm256_mul_ps(iq2,jq2);
1336
1337     /* Avoid stupid compiler warnings */
1338     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1339     j_coord_offsetA = 0;
1340     j_coord_offsetB = 0;
1341     j_coord_offsetC = 0;
1342     j_coord_offsetD = 0;
1343     j_coord_offsetE = 0;
1344     j_coord_offsetF = 0;
1345     j_coord_offsetG = 0;
1346     j_coord_offsetH = 0;
1347
1348     outeriter        = 0;
1349     inneriter        = 0;
1350
1351     for(iidx=0;iidx<4*DIM;iidx++)
1352     {
1353         scratch[iidx] = 0.0;
1354     }
1355
1356     /* Start outer loop over neighborlists */
1357     for(iidx=0; iidx<nri; iidx++)
1358     {
1359         /* Load shift vector for this list */
1360         i_shift_offset   = DIM*shiftidx[iidx];
1361
1362         /* Load limits for loop over neighbors */
1363         j_index_start    = jindex[iidx];
1364         j_index_end      = jindex[iidx+1];
1365
1366         /* Get outer coordinate index */
1367         inr              = iinr[iidx];
1368         i_coord_offset   = DIM*inr;
1369
1370         /* Load i particle coords and add shift vector */
1371         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1372                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1373
1374         fix0             = _mm256_setzero_ps();
1375         fiy0             = _mm256_setzero_ps();
1376         fiz0             = _mm256_setzero_ps();
1377         fix1             = _mm256_setzero_ps();
1378         fiy1             = _mm256_setzero_ps();
1379         fiz1             = _mm256_setzero_ps();
1380         fix2             = _mm256_setzero_ps();
1381         fiy2             = _mm256_setzero_ps();
1382         fiz2             = _mm256_setzero_ps();
1383
1384         /* Start inner kernel loop */
1385         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1386         {
1387
1388             /* Get j neighbor index, and coordinate index */
1389             jnrA             = jjnr[jidx];
1390             jnrB             = jjnr[jidx+1];
1391             jnrC             = jjnr[jidx+2];
1392             jnrD             = jjnr[jidx+3];
1393             jnrE             = jjnr[jidx+4];
1394             jnrF             = jjnr[jidx+5];
1395             jnrG             = jjnr[jidx+6];
1396             jnrH             = jjnr[jidx+7];
1397             j_coord_offsetA  = DIM*jnrA;
1398             j_coord_offsetB  = DIM*jnrB;
1399             j_coord_offsetC  = DIM*jnrC;
1400             j_coord_offsetD  = DIM*jnrD;
1401             j_coord_offsetE  = DIM*jnrE;
1402             j_coord_offsetF  = DIM*jnrF;
1403             j_coord_offsetG  = DIM*jnrG;
1404             j_coord_offsetH  = DIM*jnrH;
1405
1406             /* load j atom coordinates */
1407             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1408                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1409                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1410                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1411                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1412
1413             /* Calculate displacement vector */
1414             dx00             = _mm256_sub_ps(ix0,jx0);
1415             dy00             = _mm256_sub_ps(iy0,jy0);
1416             dz00             = _mm256_sub_ps(iz0,jz0);
1417             dx01             = _mm256_sub_ps(ix0,jx1);
1418             dy01             = _mm256_sub_ps(iy0,jy1);
1419             dz01             = _mm256_sub_ps(iz0,jz1);
1420             dx02             = _mm256_sub_ps(ix0,jx2);
1421             dy02             = _mm256_sub_ps(iy0,jy2);
1422             dz02             = _mm256_sub_ps(iz0,jz2);
1423             dx10             = _mm256_sub_ps(ix1,jx0);
1424             dy10             = _mm256_sub_ps(iy1,jy0);
1425             dz10             = _mm256_sub_ps(iz1,jz0);
1426             dx11             = _mm256_sub_ps(ix1,jx1);
1427             dy11             = _mm256_sub_ps(iy1,jy1);
1428             dz11             = _mm256_sub_ps(iz1,jz1);
1429             dx12             = _mm256_sub_ps(ix1,jx2);
1430             dy12             = _mm256_sub_ps(iy1,jy2);
1431             dz12             = _mm256_sub_ps(iz1,jz2);
1432             dx20             = _mm256_sub_ps(ix2,jx0);
1433             dy20             = _mm256_sub_ps(iy2,jy0);
1434             dz20             = _mm256_sub_ps(iz2,jz0);
1435             dx21             = _mm256_sub_ps(ix2,jx1);
1436             dy21             = _mm256_sub_ps(iy2,jy1);
1437             dz21             = _mm256_sub_ps(iz2,jz1);
1438             dx22             = _mm256_sub_ps(ix2,jx2);
1439             dy22             = _mm256_sub_ps(iy2,jy2);
1440             dz22             = _mm256_sub_ps(iz2,jz2);
1441
1442             /* Calculate squared distance and things based on it */
1443             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1444             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1445             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1446             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1447             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1448             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1449             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1450             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1451             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1452
1453             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1454             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1455             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1456             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1457             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1458             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1459             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1460             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1461             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1462
1463             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1464             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1465             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1466             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1467             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1468             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1469             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1470             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1471             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1472
1473             fjx0             = _mm256_setzero_ps();
1474             fjy0             = _mm256_setzero_ps();
1475             fjz0             = _mm256_setzero_ps();
1476             fjx1             = _mm256_setzero_ps();
1477             fjy1             = _mm256_setzero_ps();
1478             fjz1             = _mm256_setzero_ps();
1479             fjx2             = _mm256_setzero_ps();
1480             fjy2             = _mm256_setzero_ps();
1481             fjz2             = _mm256_setzero_ps();
1482
1483             /**************************
1484              * CALCULATE INTERACTIONS *
1485              **************************/
1486
1487             r00              = _mm256_mul_ps(rsq00,rinv00);
1488
1489             /* EWALD ELECTROSTATICS */
1490
1491             /* Analytical PME correction */
1492             zeta2            = _mm256_mul_ps(beta2,rsq00);
1493             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
1494             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1495             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1496             felec            = _mm256_mul_ps(qq00,felec);
1497
1498             /* LENNARD-JONES DISPERSION/REPULSION */
1499
1500             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1501             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1502
1503             fscal            = _mm256_add_ps(felec,fvdw);
1504
1505             /* Calculate temporary vectorial force */
1506             tx               = _mm256_mul_ps(fscal,dx00);
1507             ty               = _mm256_mul_ps(fscal,dy00);
1508             tz               = _mm256_mul_ps(fscal,dz00);
1509
1510             /* Update vectorial force */
1511             fix0             = _mm256_add_ps(fix0,tx);
1512             fiy0             = _mm256_add_ps(fiy0,ty);
1513             fiz0             = _mm256_add_ps(fiz0,tz);
1514
1515             fjx0             = _mm256_add_ps(fjx0,tx);
1516             fjy0             = _mm256_add_ps(fjy0,ty);
1517             fjz0             = _mm256_add_ps(fjz0,tz);
1518
1519             /**************************
1520              * CALCULATE INTERACTIONS *
1521              **************************/
1522
1523             r01              = _mm256_mul_ps(rsq01,rinv01);
1524
1525             /* EWALD ELECTROSTATICS */
1526
1527             /* Analytical PME correction */
1528             zeta2            = _mm256_mul_ps(beta2,rsq01);
1529             rinv3            = _mm256_mul_ps(rinvsq01,rinv01);
1530             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1531             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1532             felec            = _mm256_mul_ps(qq01,felec);
1533
1534             fscal            = felec;
1535
1536             /* Calculate temporary vectorial force */
1537             tx               = _mm256_mul_ps(fscal,dx01);
1538             ty               = _mm256_mul_ps(fscal,dy01);
1539             tz               = _mm256_mul_ps(fscal,dz01);
1540
1541             /* Update vectorial force */
1542             fix0             = _mm256_add_ps(fix0,tx);
1543             fiy0             = _mm256_add_ps(fiy0,ty);
1544             fiz0             = _mm256_add_ps(fiz0,tz);
1545
1546             fjx1             = _mm256_add_ps(fjx1,tx);
1547             fjy1             = _mm256_add_ps(fjy1,ty);
1548             fjz1             = _mm256_add_ps(fjz1,tz);
1549
1550             /**************************
1551              * CALCULATE INTERACTIONS *
1552              **************************/
1553
1554             r02              = _mm256_mul_ps(rsq02,rinv02);
1555
1556             /* EWALD ELECTROSTATICS */
1557
1558             /* Analytical PME correction */
1559             zeta2            = _mm256_mul_ps(beta2,rsq02);
1560             rinv3            = _mm256_mul_ps(rinvsq02,rinv02);
1561             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1562             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1563             felec            = _mm256_mul_ps(qq02,felec);
1564
1565             fscal            = felec;
1566
1567             /* Calculate temporary vectorial force */
1568             tx               = _mm256_mul_ps(fscal,dx02);
1569             ty               = _mm256_mul_ps(fscal,dy02);
1570             tz               = _mm256_mul_ps(fscal,dz02);
1571
1572             /* Update vectorial force */
1573             fix0             = _mm256_add_ps(fix0,tx);
1574             fiy0             = _mm256_add_ps(fiy0,ty);
1575             fiz0             = _mm256_add_ps(fiz0,tz);
1576
1577             fjx2             = _mm256_add_ps(fjx2,tx);
1578             fjy2             = _mm256_add_ps(fjy2,ty);
1579             fjz2             = _mm256_add_ps(fjz2,tz);
1580
1581             /**************************
1582              * CALCULATE INTERACTIONS *
1583              **************************/
1584
1585             r10              = _mm256_mul_ps(rsq10,rinv10);
1586
1587             /* EWALD ELECTROSTATICS */
1588
1589             /* Analytical PME correction */
1590             zeta2            = _mm256_mul_ps(beta2,rsq10);
1591             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
1592             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1593             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1594             felec            = _mm256_mul_ps(qq10,felec);
1595
1596             fscal            = felec;
1597
1598             /* Calculate temporary vectorial force */
1599             tx               = _mm256_mul_ps(fscal,dx10);
1600             ty               = _mm256_mul_ps(fscal,dy10);
1601             tz               = _mm256_mul_ps(fscal,dz10);
1602
1603             /* Update vectorial force */
1604             fix1             = _mm256_add_ps(fix1,tx);
1605             fiy1             = _mm256_add_ps(fiy1,ty);
1606             fiz1             = _mm256_add_ps(fiz1,tz);
1607
1608             fjx0             = _mm256_add_ps(fjx0,tx);
1609             fjy0             = _mm256_add_ps(fjy0,ty);
1610             fjz0             = _mm256_add_ps(fjz0,tz);
1611
1612             /**************************
1613              * CALCULATE INTERACTIONS *
1614              **************************/
1615
1616             r11              = _mm256_mul_ps(rsq11,rinv11);
1617
1618             /* EWALD ELECTROSTATICS */
1619
1620             /* Analytical PME correction */
1621             zeta2            = _mm256_mul_ps(beta2,rsq11);
1622             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
1623             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1624             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1625             felec            = _mm256_mul_ps(qq11,felec);
1626
1627             fscal            = felec;
1628
1629             /* Calculate temporary vectorial force */
1630             tx               = _mm256_mul_ps(fscal,dx11);
1631             ty               = _mm256_mul_ps(fscal,dy11);
1632             tz               = _mm256_mul_ps(fscal,dz11);
1633
1634             /* Update vectorial force */
1635             fix1             = _mm256_add_ps(fix1,tx);
1636             fiy1             = _mm256_add_ps(fiy1,ty);
1637             fiz1             = _mm256_add_ps(fiz1,tz);
1638
1639             fjx1             = _mm256_add_ps(fjx1,tx);
1640             fjy1             = _mm256_add_ps(fjy1,ty);
1641             fjz1             = _mm256_add_ps(fjz1,tz);
1642
1643             /**************************
1644              * CALCULATE INTERACTIONS *
1645              **************************/
1646
1647             r12              = _mm256_mul_ps(rsq12,rinv12);
1648
1649             /* EWALD ELECTROSTATICS */
1650
1651             /* Analytical PME correction */
1652             zeta2            = _mm256_mul_ps(beta2,rsq12);
1653             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
1654             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1655             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1656             felec            = _mm256_mul_ps(qq12,felec);
1657
1658             fscal            = felec;
1659
1660             /* Calculate temporary vectorial force */
1661             tx               = _mm256_mul_ps(fscal,dx12);
1662             ty               = _mm256_mul_ps(fscal,dy12);
1663             tz               = _mm256_mul_ps(fscal,dz12);
1664
1665             /* Update vectorial force */
1666             fix1             = _mm256_add_ps(fix1,tx);
1667             fiy1             = _mm256_add_ps(fiy1,ty);
1668             fiz1             = _mm256_add_ps(fiz1,tz);
1669
1670             fjx2             = _mm256_add_ps(fjx2,tx);
1671             fjy2             = _mm256_add_ps(fjy2,ty);
1672             fjz2             = _mm256_add_ps(fjz2,tz);
1673
1674             /**************************
1675              * CALCULATE INTERACTIONS *
1676              **************************/
1677
1678             r20              = _mm256_mul_ps(rsq20,rinv20);
1679
1680             /* EWALD ELECTROSTATICS */
1681
1682             /* Analytical PME correction */
1683             zeta2            = _mm256_mul_ps(beta2,rsq20);
1684             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
1685             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1686             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1687             felec            = _mm256_mul_ps(qq20,felec);
1688
1689             fscal            = felec;
1690
1691             /* Calculate temporary vectorial force */
1692             tx               = _mm256_mul_ps(fscal,dx20);
1693             ty               = _mm256_mul_ps(fscal,dy20);
1694             tz               = _mm256_mul_ps(fscal,dz20);
1695
1696             /* Update vectorial force */
1697             fix2             = _mm256_add_ps(fix2,tx);
1698             fiy2             = _mm256_add_ps(fiy2,ty);
1699             fiz2             = _mm256_add_ps(fiz2,tz);
1700
1701             fjx0             = _mm256_add_ps(fjx0,tx);
1702             fjy0             = _mm256_add_ps(fjy0,ty);
1703             fjz0             = _mm256_add_ps(fjz0,tz);
1704
1705             /**************************
1706              * CALCULATE INTERACTIONS *
1707              **************************/
1708
1709             r21              = _mm256_mul_ps(rsq21,rinv21);
1710
1711             /* EWALD ELECTROSTATICS */
1712
1713             /* Analytical PME correction */
1714             zeta2            = _mm256_mul_ps(beta2,rsq21);
1715             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
1716             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1717             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1718             felec            = _mm256_mul_ps(qq21,felec);
1719
1720             fscal            = felec;
1721
1722             /* Calculate temporary vectorial force */
1723             tx               = _mm256_mul_ps(fscal,dx21);
1724             ty               = _mm256_mul_ps(fscal,dy21);
1725             tz               = _mm256_mul_ps(fscal,dz21);
1726
1727             /* Update vectorial force */
1728             fix2             = _mm256_add_ps(fix2,tx);
1729             fiy2             = _mm256_add_ps(fiy2,ty);
1730             fiz2             = _mm256_add_ps(fiz2,tz);
1731
1732             fjx1             = _mm256_add_ps(fjx1,tx);
1733             fjy1             = _mm256_add_ps(fjy1,ty);
1734             fjz1             = _mm256_add_ps(fjz1,tz);
1735
1736             /**************************
1737              * CALCULATE INTERACTIONS *
1738              **************************/
1739
1740             r22              = _mm256_mul_ps(rsq22,rinv22);
1741
1742             /* EWALD ELECTROSTATICS */
1743
1744             /* Analytical PME correction */
1745             zeta2            = _mm256_mul_ps(beta2,rsq22);
1746             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
1747             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1748             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1749             felec            = _mm256_mul_ps(qq22,felec);
1750
1751             fscal            = felec;
1752
1753             /* Calculate temporary vectorial force */
1754             tx               = _mm256_mul_ps(fscal,dx22);
1755             ty               = _mm256_mul_ps(fscal,dy22);
1756             tz               = _mm256_mul_ps(fscal,dz22);
1757
1758             /* Update vectorial force */
1759             fix2             = _mm256_add_ps(fix2,tx);
1760             fiy2             = _mm256_add_ps(fiy2,ty);
1761             fiz2             = _mm256_add_ps(fiz2,tz);
1762
1763             fjx2             = _mm256_add_ps(fjx2,tx);
1764             fjy2             = _mm256_add_ps(fjy2,ty);
1765             fjz2             = _mm256_add_ps(fjz2,tz);
1766
1767             fjptrA             = f+j_coord_offsetA;
1768             fjptrB             = f+j_coord_offsetB;
1769             fjptrC             = f+j_coord_offsetC;
1770             fjptrD             = f+j_coord_offsetD;
1771             fjptrE             = f+j_coord_offsetE;
1772             fjptrF             = f+j_coord_offsetF;
1773             fjptrG             = f+j_coord_offsetG;
1774             fjptrH             = f+j_coord_offsetH;
1775
1776             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1777                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1778
1779             /* Inner loop uses 511 flops */
1780         }
1781
1782         if(jidx<j_index_end)
1783         {
1784
1785             /* Get j neighbor index, and coordinate index */
1786             jnrlistA         = jjnr[jidx];
1787             jnrlistB         = jjnr[jidx+1];
1788             jnrlistC         = jjnr[jidx+2];
1789             jnrlistD         = jjnr[jidx+3];
1790             jnrlistE         = jjnr[jidx+4];
1791             jnrlistF         = jjnr[jidx+5];
1792             jnrlistG         = jjnr[jidx+6];
1793             jnrlistH         = jjnr[jidx+7];
1794             /* Sign of each element will be negative for non-real atoms.
1795              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1796              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1797              */
1798             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1799                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1800
1801             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1802             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1803             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1804             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1805             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1806             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1807             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1808             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1809             j_coord_offsetA  = DIM*jnrA;
1810             j_coord_offsetB  = DIM*jnrB;
1811             j_coord_offsetC  = DIM*jnrC;
1812             j_coord_offsetD  = DIM*jnrD;
1813             j_coord_offsetE  = DIM*jnrE;
1814             j_coord_offsetF  = DIM*jnrF;
1815             j_coord_offsetG  = DIM*jnrG;
1816             j_coord_offsetH  = DIM*jnrH;
1817
1818             /* load j atom coordinates */
1819             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1820                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1821                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1822                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1823                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1824
1825             /* Calculate displacement vector */
1826             dx00             = _mm256_sub_ps(ix0,jx0);
1827             dy00             = _mm256_sub_ps(iy0,jy0);
1828             dz00             = _mm256_sub_ps(iz0,jz0);
1829             dx01             = _mm256_sub_ps(ix0,jx1);
1830             dy01             = _mm256_sub_ps(iy0,jy1);
1831             dz01             = _mm256_sub_ps(iz0,jz1);
1832             dx02             = _mm256_sub_ps(ix0,jx2);
1833             dy02             = _mm256_sub_ps(iy0,jy2);
1834             dz02             = _mm256_sub_ps(iz0,jz2);
1835             dx10             = _mm256_sub_ps(ix1,jx0);
1836             dy10             = _mm256_sub_ps(iy1,jy0);
1837             dz10             = _mm256_sub_ps(iz1,jz0);
1838             dx11             = _mm256_sub_ps(ix1,jx1);
1839             dy11             = _mm256_sub_ps(iy1,jy1);
1840             dz11             = _mm256_sub_ps(iz1,jz1);
1841             dx12             = _mm256_sub_ps(ix1,jx2);
1842             dy12             = _mm256_sub_ps(iy1,jy2);
1843             dz12             = _mm256_sub_ps(iz1,jz2);
1844             dx20             = _mm256_sub_ps(ix2,jx0);
1845             dy20             = _mm256_sub_ps(iy2,jy0);
1846             dz20             = _mm256_sub_ps(iz2,jz0);
1847             dx21             = _mm256_sub_ps(ix2,jx1);
1848             dy21             = _mm256_sub_ps(iy2,jy1);
1849             dz21             = _mm256_sub_ps(iz2,jz1);
1850             dx22             = _mm256_sub_ps(ix2,jx2);
1851             dy22             = _mm256_sub_ps(iy2,jy2);
1852             dz22             = _mm256_sub_ps(iz2,jz2);
1853
1854             /* Calculate squared distance and things based on it */
1855             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1856             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1857             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1858             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1859             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1860             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1861             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1862             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1863             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1864
1865             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1866             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1867             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1868             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1869             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1870             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1871             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1872             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1873             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1874
1875             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1876             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1877             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1878             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1879             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1880             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1881             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1882             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1883             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1884
1885             fjx0             = _mm256_setzero_ps();
1886             fjy0             = _mm256_setzero_ps();
1887             fjz0             = _mm256_setzero_ps();
1888             fjx1             = _mm256_setzero_ps();
1889             fjy1             = _mm256_setzero_ps();
1890             fjz1             = _mm256_setzero_ps();
1891             fjx2             = _mm256_setzero_ps();
1892             fjy2             = _mm256_setzero_ps();
1893             fjz2             = _mm256_setzero_ps();
1894
1895             /**************************
1896              * CALCULATE INTERACTIONS *
1897              **************************/
1898
1899             r00              = _mm256_mul_ps(rsq00,rinv00);
1900             r00              = _mm256_andnot_ps(dummy_mask,r00);
1901
1902             /* EWALD ELECTROSTATICS */
1903
1904             /* Analytical PME correction */
1905             zeta2            = _mm256_mul_ps(beta2,rsq00);
1906             rinv3            = _mm256_mul_ps(rinvsq00,rinv00);
1907             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1908             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1909             felec            = _mm256_mul_ps(qq00,felec);
1910
1911             /* LENNARD-JONES DISPERSION/REPULSION */
1912
1913             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1914             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1915
1916             fscal            = _mm256_add_ps(felec,fvdw);
1917
1918             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1919
1920             /* Calculate temporary vectorial force */
1921             tx               = _mm256_mul_ps(fscal,dx00);
1922             ty               = _mm256_mul_ps(fscal,dy00);
1923             tz               = _mm256_mul_ps(fscal,dz00);
1924
1925             /* Update vectorial force */
1926             fix0             = _mm256_add_ps(fix0,tx);
1927             fiy0             = _mm256_add_ps(fiy0,ty);
1928             fiz0             = _mm256_add_ps(fiz0,tz);
1929
1930             fjx0             = _mm256_add_ps(fjx0,tx);
1931             fjy0             = _mm256_add_ps(fjy0,ty);
1932             fjz0             = _mm256_add_ps(fjz0,tz);
1933
1934             /**************************
1935              * CALCULATE INTERACTIONS *
1936              **************************/
1937
1938             r01              = _mm256_mul_ps(rsq01,rinv01);
1939             r01              = _mm256_andnot_ps(dummy_mask,r01);
1940
1941             /* EWALD ELECTROSTATICS */
1942
1943             /* Analytical PME correction */
1944             zeta2            = _mm256_mul_ps(beta2,rsq01);
1945             rinv3            = _mm256_mul_ps(rinvsq01,rinv01);
1946             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1947             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1948             felec            = _mm256_mul_ps(qq01,felec);
1949
1950             fscal            = felec;
1951
1952             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1953
1954             /* Calculate temporary vectorial force */
1955             tx               = _mm256_mul_ps(fscal,dx01);
1956             ty               = _mm256_mul_ps(fscal,dy01);
1957             tz               = _mm256_mul_ps(fscal,dz01);
1958
1959             /* Update vectorial force */
1960             fix0             = _mm256_add_ps(fix0,tx);
1961             fiy0             = _mm256_add_ps(fiy0,ty);
1962             fiz0             = _mm256_add_ps(fiz0,tz);
1963
1964             fjx1             = _mm256_add_ps(fjx1,tx);
1965             fjy1             = _mm256_add_ps(fjy1,ty);
1966             fjz1             = _mm256_add_ps(fjz1,tz);
1967
1968             /**************************
1969              * CALCULATE INTERACTIONS *
1970              **************************/
1971
1972             r02              = _mm256_mul_ps(rsq02,rinv02);
1973             r02              = _mm256_andnot_ps(dummy_mask,r02);
1974
1975             /* EWALD ELECTROSTATICS */
1976
1977             /* Analytical PME correction */
1978             zeta2            = _mm256_mul_ps(beta2,rsq02);
1979             rinv3            = _mm256_mul_ps(rinvsq02,rinv02);
1980             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
1981             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
1982             felec            = _mm256_mul_ps(qq02,felec);
1983
1984             fscal            = felec;
1985
1986             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1987
1988             /* Calculate temporary vectorial force */
1989             tx               = _mm256_mul_ps(fscal,dx02);
1990             ty               = _mm256_mul_ps(fscal,dy02);
1991             tz               = _mm256_mul_ps(fscal,dz02);
1992
1993             /* Update vectorial force */
1994             fix0             = _mm256_add_ps(fix0,tx);
1995             fiy0             = _mm256_add_ps(fiy0,ty);
1996             fiz0             = _mm256_add_ps(fiz0,tz);
1997
1998             fjx2             = _mm256_add_ps(fjx2,tx);
1999             fjy2             = _mm256_add_ps(fjy2,ty);
2000             fjz2             = _mm256_add_ps(fjz2,tz);
2001
2002             /**************************
2003              * CALCULATE INTERACTIONS *
2004              **************************/
2005
2006             r10              = _mm256_mul_ps(rsq10,rinv10);
2007             r10              = _mm256_andnot_ps(dummy_mask,r10);
2008
2009             /* EWALD ELECTROSTATICS */
2010
2011             /* Analytical PME correction */
2012             zeta2            = _mm256_mul_ps(beta2,rsq10);
2013             rinv3            = _mm256_mul_ps(rinvsq10,rinv10);
2014             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2015             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2016             felec            = _mm256_mul_ps(qq10,felec);
2017
2018             fscal            = felec;
2019
2020             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2021
2022             /* Calculate temporary vectorial force */
2023             tx               = _mm256_mul_ps(fscal,dx10);
2024             ty               = _mm256_mul_ps(fscal,dy10);
2025             tz               = _mm256_mul_ps(fscal,dz10);
2026
2027             /* Update vectorial force */
2028             fix1             = _mm256_add_ps(fix1,tx);
2029             fiy1             = _mm256_add_ps(fiy1,ty);
2030             fiz1             = _mm256_add_ps(fiz1,tz);
2031
2032             fjx0             = _mm256_add_ps(fjx0,tx);
2033             fjy0             = _mm256_add_ps(fjy0,ty);
2034             fjz0             = _mm256_add_ps(fjz0,tz);
2035
2036             /**************************
2037              * CALCULATE INTERACTIONS *
2038              **************************/
2039
2040             r11              = _mm256_mul_ps(rsq11,rinv11);
2041             r11              = _mm256_andnot_ps(dummy_mask,r11);
2042
2043             /* EWALD ELECTROSTATICS */
2044
2045             /* Analytical PME correction */
2046             zeta2            = _mm256_mul_ps(beta2,rsq11);
2047             rinv3            = _mm256_mul_ps(rinvsq11,rinv11);
2048             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2049             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2050             felec            = _mm256_mul_ps(qq11,felec);
2051
2052             fscal            = felec;
2053
2054             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2055
2056             /* Calculate temporary vectorial force */
2057             tx               = _mm256_mul_ps(fscal,dx11);
2058             ty               = _mm256_mul_ps(fscal,dy11);
2059             tz               = _mm256_mul_ps(fscal,dz11);
2060
2061             /* Update vectorial force */
2062             fix1             = _mm256_add_ps(fix1,tx);
2063             fiy1             = _mm256_add_ps(fiy1,ty);
2064             fiz1             = _mm256_add_ps(fiz1,tz);
2065
2066             fjx1             = _mm256_add_ps(fjx1,tx);
2067             fjy1             = _mm256_add_ps(fjy1,ty);
2068             fjz1             = _mm256_add_ps(fjz1,tz);
2069
2070             /**************************
2071              * CALCULATE INTERACTIONS *
2072              **************************/
2073
2074             r12              = _mm256_mul_ps(rsq12,rinv12);
2075             r12              = _mm256_andnot_ps(dummy_mask,r12);
2076
2077             /* EWALD ELECTROSTATICS */
2078
2079             /* Analytical PME correction */
2080             zeta2            = _mm256_mul_ps(beta2,rsq12);
2081             rinv3            = _mm256_mul_ps(rinvsq12,rinv12);
2082             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2083             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2084             felec            = _mm256_mul_ps(qq12,felec);
2085
2086             fscal            = felec;
2087
2088             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2089
2090             /* Calculate temporary vectorial force */
2091             tx               = _mm256_mul_ps(fscal,dx12);
2092             ty               = _mm256_mul_ps(fscal,dy12);
2093             tz               = _mm256_mul_ps(fscal,dz12);
2094
2095             /* Update vectorial force */
2096             fix1             = _mm256_add_ps(fix1,tx);
2097             fiy1             = _mm256_add_ps(fiy1,ty);
2098             fiz1             = _mm256_add_ps(fiz1,tz);
2099
2100             fjx2             = _mm256_add_ps(fjx2,tx);
2101             fjy2             = _mm256_add_ps(fjy2,ty);
2102             fjz2             = _mm256_add_ps(fjz2,tz);
2103
2104             /**************************
2105              * CALCULATE INTERACTIONS *
2106              **************************/
2107
2108             r20              = _mm256_mul_ps(rsq20,rinv20);
2109             r20              = _mm256_andnot_ps(dummy_mask,r20);
2110
2111             /* EWALD ELECTROSTATICS */
2112
2113             /* Analytical PME correction */
2114             zeta2            = _mm256_mul_ps(beta2,rsq20);
2115             rinv3            = _mm256_mul_ps(rinvsq20,rinv20);
2116             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2117             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2118             felec            = _mm256_mul_ps(qq20,felec);
2119
2120             fscal            = felec;
2121
2122             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2123
2124             /* Calculate temporary vectorial force */
2125             tx               = _mm256_mul_ps(fscal,dx20);
2126             ty               = _mm256_mul_ps(fscal,dy20);
2127             tz               = _mm256_mul_ps(fscal,dz20);
2128
2129             /* Update vectorial force */
2130             fix2             = _mm256_add_ps(fix2,tx);
2131             fiy2             = _mm256_add_ps(fiy2,ty);
2132             fiz2             = _mm256_add_ps(fiz2,tz);
2133
2134             fjx0             = _mm256_add_ps(fjx0,tx);
2135             fjy0             = _mm256_add_ps(fjy0,ty);
2136             fjz0             = _mm256_add_ps(fjz0,tz);
2137
2138             /**************************
2139              * CALCULATE INTERACTIONS *
2140              **************************/
2141
2142             r21              = _mm256_mul_ps(rsq21,rinv21);
2143             r21              = _mm256_andnot_ps(dummy_mask,r21);
2144
2145             /* EWALD ELECTROSTATICS */
2146
2147             /* Analytical PME correction */
2148             zeta2            = _mm256_mul_ps(beta2,rsq21);
2149             rinv3            = _mm256_mul_ps(rinvsq21,rinv21);
2150             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2151             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2152             felec            = _mm256_mul_ps(qq21,felec);
2153
2154             fscal            = felec;
2155
2156             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2157
2158             /* Calculate temporary vectorial force */
2159             tx               = _mm256_mul_ps(fscal,dx21);
2160             ty               = _mm256_mul_ps(fscal,dy21);
2161             tz               = _mm256_mul_ps(fscal,dz21);
2162
2163             /* Update vectorial force */
2164             fix2             = _mm256_add_ps(fix2,tx);
2165             fiy2             = _mm256_add_ps(fiy2,ty);
2166             fiz2             = _mm256_add_ps(fiz2,tz);
2167
2168             fjx1             = _mm256_add_ps(fjx1,tx);
2169             fjy1             = _mm256_add_ps(fjy1,ty);
2170             fjz1             = _mm256_add_ps(fjz1,tz);
2171
2172             /**************************
2173              * CALCULATE INTERACTIONS *
2174              **************************/
2175
2176             r22              = _mm256_mul_ps(rsq22,rinv22);
2177             r22              = _mm256_andnot_ps(dummy_mask,r22);
2178
2179             /* EWALD ELECTROSTATICS */
2180
2181             /* Analytical PME correction */
2182             zeta2            = _mm256_mul_ps(beta2,rsq22);
2183             rinv3            = _mm256_mul_ps(rinvsq22,rinv22);
2184             pmecorrF         = gmx_mm256_pmecorrF_ps(zeta2);
2185             felec            = _mm256_add_ps( _mm256_mul_ps(pmecorrF,beta3), rinv3);
2186             felec            = _mm256_mul_ps(qq22,felec);
2187
2188             fscal            = felec;
2189
2190             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2191
2192             /* Calculate temporary vectorial force */
2193             tx               = _mm256_mul_ps(fscal,dx22);
2194             ty               = _mm256_mul_ps(fscal,dy22);
2195             tz               = _mm256_mul_ps(fscal,dz22);
2196
2197             /* Update vectorial force */
2198             fix2             = _mm256_add_ps(fix2,tx);
2199             fiy2             = _mm256_add_ps(fiy2,ty);
2200             fiz2             = _mm256_add_ps(fiz2,tz);
2201
2202             fjx2             = _mm256_add_ps(fjx2,tx);
2203             fjy2             = _mm256_add_ps(fjy2,ty);
2204             fjz2             = _mm256_add_ps(fjz2,tz);
2205
2206             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2207             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2208             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2209             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2210             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2211             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2212             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2213             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2214
2215             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2216                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2217
2218             /* Inner loop uses 520 flops */
2219         }
2220
2221         /* End of innermost loop */
2222
2223         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2224                                                  f+i_coord_offset,fshift+i_shift_offset);
2225
2226         /* Increment number of inner iterations */
2227         inneriter                  += j_index_end - j_index_start;
2228
2229         /* Outer loop uses 18 flops */
2230     }
2231
2232     /* Increment number of outer iterations */
2233     outeriter        += nri;
2234
2235     /* Update outer/inner flops */
2236
2237     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*520);
2238 }