src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_avx_256_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_double.h"
  50 #include "kernelutil_x86_avx_256_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_VF_avx_256_double
  54  * Electrostatics interaction: Coulomb
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_VF_avx_256_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  79     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  84     real             scratch[4*DIM];
  85     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  95     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  97     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 100     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 101     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 102     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 103     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 104     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 105     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 106     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 107     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 108     real             *charge;
 109     int              nvdwtype;
 110     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 111     int              *vdwtype;
 112     real             *vdwparam;
 113     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 114     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 115     __m128i          vfitab;
 116     __m128i          ifour       = _mm_set1_epi32(4);
 117     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 118     real             *vftab;
 119     __m256d          dummy_mask,cutoff_mask;
 120     __m128           tmpmask0,tmpmask1;
 121     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 122     __m256d          one     = _mm256_set1_pd(1.0);
 123     __m256d          two     = _mm256_set1_pd(2.0);
 124     x                = xx[0];
 125     f                = ff[0];
 126
 127     nri              = nlist->nri;
 128     iinr             = nlist->iinr;
 129     jindex           = nlist->jindex;
 130     jjnr             = nlist->jjnr;
 131     shiftidx         = nlist->shift;
 132     gid              = nlist->gid;
 133     shiftvec         = fr->shift_vec[0];
 134     fshift           = fr->fshift[0];
 135     facel            = _mm256_set1_pd(fr->epsfac);
 136     charge           = mdatoms->chargeA;
 137     nvdwtype         = fr->ntype;
 138     vdwparam         = fr->nbfp;
 139     vdwtype          = mdatoms->typeA;
 140
 141     vftab            = kernel_data->table_vdw->data;
 142     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
 143
 144     /* Setup water-specific parameters */
 145     inr              = nlist->iinr[0];
 146     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 147     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 148     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 149     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 150
 151     jq0              = _mm256_set1_pd(charge[inr+0]);
 152     jq1              = _mm256_set1_pd(charge[inr+1]);
 153     jq2              = _mm256_set1_pd(charge[inr+2]);
 154     vdwjidx0A        = 2*vdwtype[inr+0];
 155     qq00             = _mm256_mul_pd(iq0,jq0);
 156     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 157     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 158     qq01             = _mm256_mul_pd(iq0,jq1);
 159     qq02             = _mm256_mul_pd(iq0,jq2);
 160     qq10             = _mm256_mul_pd(iq1,jq0);
 161     qq11             = _mm256_mul_pd(iq1,jq1);
 162     qq12             = _mm256_mul_pd(iq1,jq2);
 163     qq20             = _mm256_mul_pd(iq2,jq0);
 164     qq21             = _mm256_mul_pd(iq2,jq1);
 165     qq22             = _mm256_mul_pd(iq2,jq2);
 166
 167     /* Avoid stupid compiler warnings */
 168     jnrA = jnrB = jnrC = jnrD = 0;
 169     j_coord_offsetA = 0;
 170     j_coord_offsetB = 0;
 171     j_coord_offsetC = 0;
 172     j_coord_offsetD = 0;
 173
 174     outeriter        = 0;
 175     inneriter        = 0;
 176
 177     for(iidx=0;iidx<4*DIM;iidx++)
 178     {
 179         scratch[iidx] = 0.0;
 180     }
 181
 182     /* Start outer loop over neighborlists */
 183     for(iidx=0; iidx<nri; iidx++)
 184     {
 185         /* Load shift vector for this list */
 186         i_shift_offset   = DIM*shiftidx[iidx];
 187
 188         /* Load limits for loop over neighbors */
 189         j_index_start    = jindex[iidx];
 190         j_index_end      = jindex[iidx+1];
 191
 192         /* Get outer coordinate index */
 193         inr              = iinr[iidx];
 194         i_coord_offset   = DIM*inr;
 195
 196         /* Load i particle coords and add shift vector */
 197         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 198                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 199
 200         fix0             = _mm256_setzero_pd();
 201         fiy0             = _mm256_setzero_pd();
 202         fiz0             = _mm256_setzero_pd();
 203         fix1             = _mm256_setzero_pd();
 204         fiy1             = _mm256_setzero_pd();
 205         fiz1             = _mm256_setzero_pd();
 206         fix2             = _mm256_setzero_pd();
 207         fiy2             = _mm256_setzero_pd();
 208         fiz2             = _mm256_setzero_pd();
 209
 210         /* Reset potential sums */
 211         velecsum         = _mm256_setzero_pd();
 212         vvdwsum          = _mm256_setzero_pd();
 213
 214         /* Start inner kernel loop */
 215         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 216         {
 217
 218             /* Get j neighbor index, and coordinate index */
 219             jnrA             = jjnr[jidx];
 220             jnrB             = jjnr[jidx+1];
 221             jnrC             = jjnr[jidx+2];
 222             jnrD             = jjnr[jidx+3];
 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
 228             /* load j atom coordinates */
 229             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 230                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 231                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 232
 233             /* Calculate displacement vector */
 234             dx00             = _mm256_sub_pd(ix0,jx0);
 235             dy00             = _mm256_sub_pd(iy0,jy0);
 236             dz00             = _mm256_sub_pd(iz0,jz0);
 237             dx01             = _mm256_sub_pd(ix0,jx1);
 238             dy01             = _mm256_sub_pd(iy0,jy1);
 239             dz01             = _mm256_sub_pd(iz0,jz1);
 240             dx02             = _mm256_sub_pd(ix0,jx2);
 241             dy02             = _mm256_sub_pd(iy0,jy2);
 242             dz02             = _mm256_sub_pd(iz0,jz2);
 243             dx10             = _mm256_sub_pd(ix1,jx0);
 244             dy10             = _mm256_sub_pd(iy1,jy0);
 245             dz10             = _mm256_sub_pd(iz1,jz0);
 246             dx11             = _mm256_sub_pd(ix1,jx1);
 247             dy11             = _mm256_sub_pd(iy1,jy1);
 248             dz11             = _mm256_sub_pd(iz1,jz1);
 249             dx12             = _mm256_sub_pd(ix1,jx2);
 250             dy12             = _mm256_sub_pd(iy1,jy2);
 251             dz12             = _mm256_sub_pd(iz1,jz2);
 252             dx20             = _mm256_sub_pd(ix2,jx0);
 253             dy20             = _mm256_sub_pd(iy2,jy0);
 254             dz20             = _mm256_sub_pd(iz2,jz0);
 255             dx21             = _mm256_sub_pd(ix2,jx1);
 256             dy21             = _mm256_sub_pd(iy2,jy1);
 257             dz21             = _mm256_sub_pd(iz2,jz1);
 258             dx22             = _mm256_sub_pd(ix2,jx2);
 259             dy22             = _mm256_sub_pd(iy2,jy2);
 260             dz22             = _mm256_sub_pd(iz2,jz2);
 261
 262             /* Calculate squared distance and things based on it */
 263             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 264             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 265             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 266             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 267             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 268             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 269             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 270             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 271             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 272
 273             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 274             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 275             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 276             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 277             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 278             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 279             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 280             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 281             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 282
 283             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 284             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 285             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 286             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 287             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 288             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 289             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 290             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 291             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 292
 293             fjx0             = _mm256_setzero_pd();
 294             fjy0             = _mm256_setzero_pd();
 295             fjz0             = _mm256_setzero_pd();
 296             fjx1             = _mm256_setzero_pd();
 297             fjy1             = _mm256_setzero_pd();
 298             fjz1             = _mm256_setzero_pd();
 299             fjx2             = _mm256_setzero_pd();
 300             fjy2             = _mm256_setzero_pd();
 301             fjz2             = _mm256_setzero_pd();
 302
 303             /**************************
 304              * CALCULATE INTERACTIONS *
 305              **************************/
 306
 307             r00              = _mm256_mul_pd(rsq00,rinv00);
 308
 309             /* Calculate table index by multiplying r with table scale and truncate to integer */
 310             rt               = _mm256_mul_pd(r00,vftabscale);
 311             vfitab           = _mm256_cvttpd_epi32(rt);
 312             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 313             vfitab           = _mm_slli_epi32(vfitab,3);
 314
 315             /* COULOMB ELECTROSTATICS */
 316             velec            = _mm256_mul_pd(qq00,rinv00);
 317             felec            = _mm256_mul_pd(velec,rinvsq00);
 318
 319             /* CUBIC SPLINE TABLE DISPERSION */
 320             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 321             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 322             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 323             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 324             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 325             Heps             = _mm256_mul_pd(vfeps,H);
 326             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 327             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 328             vvdw6            = _mm256_mul_pd(c6_00,VV);
 329             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 330             fvdw6            = _mm256_mul_pd(c6_00,FF);
 331
 332             /* CUBIC SPLINE TABLE REPULSION */
 333             vfitab           = _mm_add_epi32(vfitab,ifour);
 334             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 335             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 336             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 337             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 338             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 339             Heps             = _mm256_mul_pd(vfeps,H);
 340             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 341             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 342             vvdw12           = _mm256_mul_pd(c12_00,VV);
 343             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 344             fvdw12           = _mm256_mul_pd(c12_00,FF);
 345             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 346             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 347
 348             /* Update potential sum for this i atom from the interaction with this j atom. */
 349             velecsum         = _mm256_add_pd(velecsum,velec);
 350             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 351
 352             fscal            = _mm256_add_pd(felec,fvdw);
 353
 354             /* Calculate temporary vectorial force */
 355             tx               = _mm256_mul_pd(fscal,dx00);
 356             ty               = _mm256_mul_pd(fscal,dy00);
 357             tz               = _mm256_mul_pd(fscal,dz00);
 358
 359             /* Update vectorial force */
 360             fix0             = _mm256_add_pd(fix0,tx);
 361             fiy0             = _mm256_add_pd(fiy0,ty);
 362             fiz0             = _mm256_add_pd(fiz0,tz);
 363
 364             fjx0             = _mm256_add_pd(fjx0,tx);
 365             fjy0             = _mm256_add_pd(fjy0,ty);
 366             fjz0             = _mm256_add_pd(fjz0,tz);
 367
 368             /**************************
 369              * CALCULATE INTERACTIONS *
 370              **************************/
 371
 372             /* COULOMB ELECTROSTATICS */
 373             velec            = _mm256_mul_pd(qq01,rinv01);
 374             felec            = _mm256_mul_pd(velec,rinvsq01);
 375
 376             /* Update potential sum for this i atom from the interaction with this j atom. */
 377             velecsum         = _mm256_add_pd(velecsum,velec);
 378
 379             fscal            = felec;
 380
 381             /* Calculate temporary vectorial force */
 382             tx               = _mm256_mul_pd(fscal,dx01);
 383             ty               = _mm256_mul_pd(fscal,dy01);
 384             tz               = _mm256_mul_pd(fscal,dz01);
 385
 386             /* Update vectorial force */
 387             fix0             = _mm256_add_pd(fix0,tx);
 388             fiy0             = _mm256_add_pd(fiy0,ty);
 389             fiz0             = _mm256_add_pd(fiz0,tz);
 390
 391             fjx1             = _mm256_add_pd(fjx1,tx);
 392             fjy1             = _mm256_add_pd(fjy1,ty);
 393             fjz1             = _mm256_add_pd(fjz1,tz);
 394
 395             /**************************
 396              * CALCULATE INTERACTIONS *
 397              **************************/
 398
 399             /* COULOMB ELECTROSTATICS */
 400             velec            = _mm256_mul_pd(qq02,rinv02);
 401             felec            = _mm256_mul_pd(velec,rinvsq02);
 402
 403             /* Update potential sum for this i atom from the interaction with this j atom. */
 404             velecsum         = _mm256_add_pd(velecsum,velec);
 405
 406             fscal            = felec;
 407
 408             /* Calculate temporary vectorial force */
 409             tx               = _mm256_mul_pd(fscal,dx02);
 410             ty               = _mm256_mul_pd(fscal,dy02);
 411             tz               = _mm256_mul_pd(fscal,dz02);
 412
 413             /* Update vectorial force */
 414             fix0             = _mm256_add_pd(fix0,tx);
 415             fiy0             = _mm256_add_pd(fiy0,ty);
 416             fiz0             = _mm256_add_pd(fiz0,tz);
 417
 418             fjx2             = _mm256_add_pd(fjx2,tx);
 419             fjy2             = _mm256_add_pd(fjy2,ty);
 420             fjz2             = _mm256_add_pd(fjz2,tz);
 421
 422             /**************************
 423              * CALCULATE INTERACTIONS *
 424              **************************/
 425
 426             /* COULOMB ELECTROSTATICS */
 427             velec            = _mm256_mul_pd(qq10,rinv10);
 428             felec            = _mm256_mul_pd(velec,rinvsq10);
 429
 430             /* Update potential sum for this i atom from the interaction with this j atom. */
 431             velecsum         = _mm256_add_pd(velecsum,velec);
 432
 433             fscal            = felec;
 434
 435             /* Calculate temporary vectorial force */
 436             tx               = _mm256_mul_pd(fscal,dx10);
 437             ty               = _mm256_mul_pd(fscal,dy10);
 438             tz               = _mm256_mul_pd(fscal,dz10);
 439
 440             /* Update vectorial force */
 441             fix1             = _mm256_add_pd(fix1,tx);
 442             fiy1             = _mm256_add_pd(fiy1,ty);
 443             fiz1             = _mm256_add_pd(fiz1,tz);
 444
 445             fjx0             = _mm256_add_pd(fjx0,tx);
 446             fjy0             = _mm256_add_pd(fjy0,ty);
 447             fjz0             = _mm256_add_pd(fjz0,tz);
 448
 449             /**************************
 450              * CALCULATE INTERACTIONS *
 451              **************************/
 452
 453             /* COULOMB ELECTROSTATICS */
 454             velec            = _mm256_mul_pd(qq11,rinv11);
 455             felec            = _mm256_mul_pd(velec,rinvsq11);
 456
 457             /* Update potential sum for this i atom from the interaction with this j atom. */
 458             velecsum         = _mm256_add_pd(velecsum,velec);
 459
 460             fscal            = felec;
 461
 462             /* Calculate temporary vectorial force */
 463             tx               = _mm256_mul_pd(fscal,dx11);
 464             ty               = _mm256_mul_pd(fscal,dy11);
 465             tz               = _mm256_mul_pd(fscal,dz11);
 466
 467             /* Update vectorial force */
 468             fix1             = _mm256_add_pd(fix1,tx);
 469             fiy1             = _mm256_add_pd(fiy1,ty);
 470             fiz1             = _mm256_add_pd(fiz1,tz);
 471
 472             fjx1             = _mm256_add_pd(fjx1,tx);
 473             fjy1             = _mm256_add_pd(fjy1,ty);
 474             fjz1             = _mm256_add_pd(fjz1,tz);
 475
 476             /**************************
 477              * CALCULATE INTERACTIONS *
 478              **************************/
 479
 480             /* COULOMB ELECTROSTATICS */
 481             velec            = _mm256_mul_pd(qq12,rinv12);
 482             felec            = _mm256_mul_pd(velec,rinvsq12);
 483
 484             /* Update potential sum for this i atom from the interaction with this j atom. */
 485             velecsum         = _mm256_add_pd(velecsum,velec);
 486
 487             fscal            = felec;
 488
 489             /* Calculate temporary vectorial force */
 490             tx               = _mm256_mul_pd(fscal,dx12);
 491             ty               = _mm256_mul_pd(fscal,dy12);
 492             tz               = _mm256_mul_pd(fscal,dz12);
 493
 494             /* Update vectorial force */
 495             fix1             = _mm256_add_pd(fix1,tx);
 496             fiy1             = _mm256_add_pd(fiy1,ty);
 497             fiz1             = _mm256_add_pd(fiz1,tz);
 498
 499             fjx2             = _mm256_add_pd(fjx2,tx);
 500             fjy2             = _mm256_add_pd(fjy2,ty);
 501             fjz2             = _mm256_add_pd(fjz2,tz);
 502
 503             /**************************
 504              * CALCULATE INTERACTIONS *
 505              **************************/
 506
 507             /* COULOMB ELECTROSTATICS */
 508             velec            = _mm256_mul_pd(qq20,rinv20);
 509             felec            = _mm256_mul_pd(velec,rinvsq20);
 510
 511             /* Update potential sum for this i atom from the interaction with this j atom. */
 512             velecsum         = _mm256_add_pd(velecsum,velec);
 513
 514             fscal            = felec;
 515
 516             /* Calculate temporary vectorial force */
 517             tx               = _mm256_mul_pd(fscal,dx20);
 518             ty               = _mm256_mul_pd(fscal,dy20);
 519             tz               = _mm256_mul_pd(fscal,dz20);
 520
 521             /* Update vectorial force */
 522             fix2             = _mm256_add_pd(fix2,tx);
 523             fiy2             = _mm256_add_pd(fiy2,ty);
 524             fiz2             = _mm256_add_pd(fiz2,tz);
 525
 526             fjx0             = _mm256_add_pd(fjx0,tx);
 527             fjy0             = _mm256_add_pd(fjy0,ty);
 528             fjz0             = _mm256_add_pd(fjz0,tz);
 529
 530             /**************************
 531              * CALCULATE INTERACTIONS *
 532              **************************/
 533
 534             /* COULOMB ELECTROSTATICS */
 535             velec            = _mm256_mul_pd(qq21,rinv21);
 536             felec            = _mm256_mul_pd(velec,rinvsq21);
 537
 538             /* Update potential sum for this i atom from the interaction with this j atom. */
 539             velecsum         = _mm256_add_pd(velecsum,velec);
 540
 541             fscal            = felec;
 542
 543             /* Calculate temporary vectorial force */
 544             tx               = _mm256_mul_pd(fscal,dx21);
 545             ty               = _mm256_mul_pd(fscal,dy21);
 546             tz               = _mm256_mul_pd(fscal,dz21);
 547
 548             /* Update vectorial force */
 549             fix2             = _mm256_add_pd(fix2,tx);
 550             fiy2             = _mm256_add_pd(fiy2,ty);
 551             fiz2             = _mm256_add_pd(fiz2,tz);
 552
 553             fjx1             = _mm256_add_pd(fjx1,tx);
 554             fjy1             = _mm256_add_pd(fjy1,ty);
 555             fjz1             = _mm256_add_pd(fjz1,tz);
 556
 557             /**************************
 558              * CALCULATE INTERACTIONS *
 559              **************************/
 560
 561             /* COULOMB ELECTROSTATICS */
 562             velec            = _mm256_mul_pd(qq22,rinv22);
 563             felec            = _mm256_mul_pd(velec,rinvsq22);
 564
 565             /* Update potential sum for this i atom from the interaction with this j atom. */
 566             velecsum         = _mm256_add_pd(velecsum,velec);
 567
 568             fscal            = felec;
 569
 570             /* Calculate temporary vectorial force */
 571             tx               = _mm256_mul_pd(fscal,dx22);
 572             ty               = _mm256_mul_pd(fscal,dy22);
 573             tz               = _mm256_mul_pd(fscal,dz22);
 574
 575             /* Update vectorial force */
 576             fix2             = _mm256_add_pd(fix2,tx);
 577             fiy2             = _mm256_add_pd(fiy2,ty);
 578             fiz2             = _mm256_add_pd(fiz2,tz);
 579
 580             fjx2             = _mm256_add_pd(fjx2,tx);
 581             fjy2             = _mm256_add_pd(fjy2,ty);
 582             fjz2             = _mm256_add_pd(fjz2,tz);
 583
 584             fjptrA             = f+j_coord_offsetA;
 585             fjptrB             = f+j_coord_offsetB;
 586             fjptrC             = f+j_coord_offsetC;
 587             fjptrD             = f+j_coord_offsetD;
 588
 589             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 590                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 591
 592             /* Inner loop uses 278 flops */
 593         }
 594
 595         if(jidx<j_index_end)
 596         {
 597
 598             /* Get j neighbor index, and coordinate index */
 599             jnrlistA         = jjnr[jidx];
 600             jnrlistB         = jjnr[jidx+1];
 601             jnrlistC         = jjnr[jidx+2];
 602             jnrlistD         = jjnr[jidx+3];
 603             /* Sign of each element will be negative for non-real atoms.
 604              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 605              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 606              */
 607             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 608
 609             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 610             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 611             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 612
 613             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 614             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 615             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 616             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 617             j_coord_offsetA  = DIM*jnrA;
 618             j_coord_offsetB  = DIM*jnrB;
 619             j_coord_offsetC  = DIM*jnrC;
 620             j_coord_offsetD  = DIM*jnrD;
 621
 622             /* load j atom coordinates */
 623             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 624                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 625                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 626
 627             /* Calculate displacement vector */
 628             dx00             = _mm256_sub_pd(ix0,jx0);
 629             dy00             = _mm256_sub_pd(iy0,jy0);
 630             dz00             = _mm256_sub_pd(iz0,jz0);
 631             dx01             = _mm256_sub_pd(ix0,jx1);
 632             dy01             = _mm256_sub_pd(iy0,jy1);
 633             dz01             = _mm256_sub_pd(iz0,jz1);
 634             dx02             = _mm256_sub_pd(ix0,jx2);
 635             dy02             = _mm256_sub_pd(iy0,jy2);
 636             dz02             = _mm256_sub_pd(iz0,jz2);
 637             dx10             = _mm256_sub_pd(ix1,jx0);
 638             dy10             = _mm256_sub_pd(iy1,jy0);
 639             dz10             = _mm256_sub_pd(iz1,jz0);
 640             dx11             = _mm256_sub_pd(ix1,jx1);
 641             dy11             = _mm256_sub_pd(iy1,jy1);
 642             dz11             = _mm256_sub_pd(iz1,jz1);
 643             dx12             = _mm256_sub_pd(ix1,jx2);
 644             dy12             = _mm256_sub_pd(iy1,jy2);
 645             dz12             = _mm256_sub_pd(iz1,jz2);
 646             dx20             = _mm256_sub_pd(ix2,jx0);
 647             dy20             = _mm256_sub_pd(iy2,jy0);
 648             dz20             = _mm256_sub_pd(iz2,jz0);
 649             dx21             = _mm256_sub_pd(ix2,jx1);
 650             dy21             = _mm256_sub_pd(iy2,jy1);
 651             dz21             = _mm256_sub_pd(iz2,jz1);
 652             dx22             = _mm256_sub_pd(ix2,jx2);
 653             dy22             = _mm256_sub_pd(iy2,jy2);
 654             dz22             = _mm256_sub_pd(iz2,jz2);
 655
 656             /* Calculate squared distance and things based on it */
 657             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 658             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 659             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 660             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 661             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 662             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 663             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 664             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 665             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 666
 667             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 668             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 669             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 670             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 671             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 672             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 673             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 674             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 675             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 676
 677             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 678             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 679             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 680             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 681             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 682             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 683             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 684             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 685             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 686
 687             fjx0             = _mm256_setzero_pd();
 688             fjy0             = _mm256_setzero_pd();
 689             fjz0             = _mm256_setzero_pd();
 690             fjx1             = _mm256_setzero_pd();
 691             fjy1             = _mm256_setzero_pd();
 692             fjz1             = _mm256_setzero_pd();
 693             fjx2             = _mm256_setzero_pd();
 694             fjy2             = _mm256_setzero_pd();
 695             fjz2             = _mm256_setzero_pd();
 696
 697             /**************************
 698              * CALCULATE INTERACTIONS *
 699              **************************/
 700
 701             r00              = _mm256_mul_pd(rsq00,rinv00);
 702             r00              = _mm256_andnot_pd(dummy_mask,r00);
 703
 704             /* Calculate table index by multiplying r with table scale and truncate to integer */
 705             rt               = _mm256_mul_pd(r00,vftabscale);
 706             vfitab           = _mm256_cvttpd_epi32(rt);
 707             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 708             vfitab           = _mm_slli_epi32(vfitab,3);
 709
 710             /* COULOMB ELECTROSTATICS */
 711             velec            = _mm256_mul_pd(qq00,rinv00);
 712             felec            = _mm256_mul_pd(velec,rinvsq00);
 713
 714             /* CUBIC SPLINE TABLE DISPERSION */
 715             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 716             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 717             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 718             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 719             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 720             Heps             = _mm256_mul_pd(vfeps,H);
 721             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 722             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 723             vvdw6            = _mm256_mul_pd(c6_00,VV);
 724             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 725             fvdw6            = _mm256_mul_pd(c6_00,FF);
 726
 727             /* CUBIC SPLINE TABLE REPULSION */
 728             vfitab           = _mm_add_epi32(vfitab,ifour);
 729             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 730             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 731             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 732             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 733             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 734             Heps             = _mm256_mul_pd(vfeps,H);
 735             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 736             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 737             vvdw12           = _mm256_mul_pd(c12_00,VV);
 738             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 739             fvdw12           = _mm256_mul_pd(c12_00,FF);
 740             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 741             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 742
 743             /* Update potential sum for this i atom from the interaction with this j atom. */
 744             velec            = _mm256_andnot_pd(dummy_mask,velec);
 745             velecsum         = _mm256_add_pd(velecsum,velec);
 746             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 747             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 748
 749             fscal            = _mm256_add_pd(felec,fvdw);
 750
 751             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 752
 753             /* Calculate temporary vectorial force */
 754             tx               = _mm256_mul_pd(fscal,dx00);
 755             ty               = _mm256_mul_pd(fscal,dy00);
 756             tz               = _mm256_mul_pd(fscal,dz00);
 757
 758             /* Update vectorial force */
 759             fix0             = _mm256_add_pd(fix0,tx);
 760             fiy0             = _mm256_add_pd(fiy0,ty);
 761             fiz0             = _mm256_add_pd(fiz0,tz);
 762
 763             fjx0             = _mm256_add_pd(fjx0,tx);
 764             fjy0             = _mm256_add_pd(fjy0,ty);
 765             fjz0             = _mm256_add_pd(fjz0,tz);
 766
 767             /**************************
 768              * CALCULATE INTERACTIONS *
 769              **************************/
 770
 771             /* COULOMB ELECTROSTATICS */
 772             velec            = _mm256_mul_pd(qq01,rinv01);
 773             felec            = _mm256_mul_pd(velec,rinvsq01);
 774
 775             /* Update potential sum for this i atom from the interaction with this j atom. */
 776             velec            = _mm256_andnot_pd(dummy_mask,velec);
 777             velecsum         = _mm256_add_pd(velecsum,velec);
 778
 779             fscal            = felec;
 780
 781             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 782
 783             /* Calculate temporary vectorial force */
 784             tx               = _mm256_mul_pd(fscal,dx01);
 785             ty               = _mm256_mul_pd(fscal,dy01);
 786             tz               = _mm256_mul_pd(fscal,dz01);
 787
 788             /* Update vectorial force */
 789             fix0             = _mm256_add_pd(fix0,tx);
 790             fiy0             = _mm256_add_pd(fiy0,ty);
 791             fiz0             = _mm256_add_pd(fiz0,tz);
 792
 793             fjx1             = _mm256_add_pd(fjx1,tx);
 794             fjy1             = _mm256_add_pd(fjy1,ty);
 795             fjz1             = _mm256_add_pd(fjz1,tz);
 796
 797             /**************************
 798              * CALCULATE INTERACTIONS *
 799              **************************/
 800
 801             /* COULOMB ELECTROSTATICS */
 802             velec            = _mm256_mul_pd(qq02,rinv02);
 803             felec            = _mm256_mul_pd(velec,rinvsq02);
 804
 805             /* Update potential sum for this i atom from the interaction with this j atom. */
 806             velec            = _mm256_andnot_pd(dummy_mask,velec);
 807             velecsum         = _mm256_add_pd(velecsum,velec);
 808
 809             fscal            = felec;
 810
 811             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 812
 813             /* Calculate temporary vectorial force */
 814             tx               = _mm256_mul_pd(fscal,dx02);
 815             ty               = _mm256_mul_pd(fscal,dy02);
 816             tz               = _mm256_mul_pd(fscal,dz02);
 817
 818             /* Update vectorial force */
 819             fix0             = _mm256_add_pd(fix0,tx);
 820             fiy0             = _mm256_add_pd(fiy0,ty);
 821             fiz0             = _mm256_add_pd(fiz0,tz);
 822
 823             fjx2             = _mm256_add_pd(fjx2,tx);
 824             fjy2             = _mm256_add_pd(fjy2,ty);
 825             fjz2             = _mm256_add_pd(fjz2,tz);
 826
 827             /**************************
 828              * CALCULATE INTERACTIONS *
 829              **************************/
 830
 831             /* COULOMB ELECTROSTATICS */
 832             velec            = _mm256_mul_pd(qq10,rinv10);
 833             felec            = _mm256_mul_pd(velec,rinvsq10);
 834
 835             /* Update potential sum for this i atom from the interaction with this j atom. */
 836             velec            = _mm256_andnot_pd(dummy_mask,velec);
 837             velecsum         = _mm256_add_pd(velecsum,velec);
 838
 839             fscal            = felec;
 840
 841             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 842
 843             /* Calculate temporary vectorial force */
 844             tx               = _mm256_mul_pd(fscal,dx10);
 845             ty               = _mm256_mul_pd(fscal,dy10);
 846             tz               = _mm256_mul_pd(fscal,dz10);
 847
 848             /* Update vectorial force */
 849             fix1             = _mm256_add_pd(fix1,tx);
 850             fiy1             = _mm256_add_pd(fiy1,ty);
 851             fiz1             = _mm256_add_pd(fiz1,tz);
 852
 853             fjx0             = _mm256_add_pd(fjx0,tx);
 854             fjy0             = _mm256_add_pd(fjy0,ty);
 855             fjz0             = _mm256_add_pd(fjz0,tz);
 856
 857             /**************************
 858              * CALCULATE INTERACTIONS *
 859              **************************/
 860
 861             /* COULOMB ELECTROSTATICS */
 862             velec            = _mm256_mul_pd(qq11,rinv11);
 863             felec            = _mm256_mul_pd(velec,rinvsq11);
 864
 865             /* Update potential sum for this i atom from the interaction with this j atom. */
 866             velec            = _mm256_andnot_pd(dummy_mask,velec);
 867             velecsum         = _mm256_add_pd(velecsum,velec);
 868
 869             fscal            = felec;
 870
 871             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 872
 873             /* Calculate temporary vectorial force */
 874             tx               = _mm256_mul_pd(fscal,dx11);
 875             ty               = _mm256_mul_pd(fscal,dy11);
 876             tz               = _mm256_mul_pd(fscal,dz11);
 877
 878             /* Update vectorial force */
 879             fix1             = _mm256_add_pd(fix1,tx);
 880             fiy1             = _mm256_add_pd(fiy1,ty);
 881             fiz1             = _mm256_add_pd(fiz1,tz);
 882
 883             fjx1             = _mm256_add_pd(fjx1,tx);
 884             fjy1             = _mm256_add_pd(fjy1,ty);
 885             fjz1             = _mm256_add_pd(fjz1,tz);
 886
 887             /**************************
 888              * CALCULATE INTERACTIONS *
 889              **************************/
 890
 891             /* COULOMB ELECTROSTATICS */
 892             velec            = _mm256_mul_pd(qq12,rinv12);
 893             felec            = _mm256_mul_pd(velec,rinvsq12);
 894
 895             /* Update potential sum for this i atom from the interaction with this j atom. */
 896             velec            = _mm256_andnot_pd(dummy_mask,velec);
 897             velecsum         = _mm256_add_pd(velecsum,velec);
 898
 899             fscal            = felec;
 900
 901             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 902
 903             /* Calculate temporary vectorial force */
 904             tx               = _mm256_mul_pd(fscal,dx12);
 905             ty               = _mm256_mul_pd(fscal,dy12);
 906             tz               = _mm256_mul_pd(fscal,dz12);
 907
 908             /* Update vectorial force */
 909             fix1             = _mm256_add_pd(fix1,tx);
 910             fiy1             = _mm256_add_pd(fiy1,ty);
 911             fiz1             = _mm256_add_pd(fiz1,tz);
 912
 913             fjx2             = _mm256_add_pd(fjx2,tx);
 914             fjy2             = _mm256_add_pd(fjy2,ty);
 915             fjz2             = _mm256_add_pd(fjz2,tz);
 916
 917             /**************************
 918              * CALCULATE INTERACTIONS *
 919              **************************/
 920
 921             /* COULOMB ELECTROSTATICS */
 922             velec            = _mm256_mul_pd(qq20,rinv20);
 923             felec            = _mm256_mul_pd(velec,rinvsq20);
 924
 925             /* Update potential sum for this i atom from the interaction with this j atom. */
 926             velec            = _mm256_andnot_pd(dummy_mask,velec);
 927             velecsum         = _mm256_add_pd(velecsum,velec);
 928
 929             fscal            = felec;
 930
 931             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 932
 933             /* Calculate temporary vectorial force */
 934             tx               = _mm256_mul_pd(fscal,dx20);
 935             ty               = _mm256_mul_pd(fscal,dy20);
 936             tz               = _mm256_mul_pd(fscal,dz20);
 937
 938             /* Update vectorial force */
 939             fix2             = _mm256_add_pd(fix2,tx);
 940             fiy2             = _mm256_add_pd(fiy2,ty);
 941             fiz2             = _mm256_add_pd(fiz2,tz);
 942
 943             fjx0             = _mm256_add_pd(fjx0,tx);
 944             fjy0             = _mm256_add_pd(fjy0,ty);
 945             fjz0             = _mm256_add_pd(fjz0,tz);
 946
 947             /**************************
 948              * CALCULATE INTERACTIONS *
 949              **************************/
 950
 951             /* COULOMB ELECTROSTATICS */
 952             velec            = _mm256_mul_pd(qq21,rinv21);
 953             felec            = _mm256_mul_pd(velec,rinvsq21);
 954
 955             /* Update potential sum for this i atom from the interaction with this j atom. */
 956             velec            = _mm256_andnot_pd(dummy_mask,velec);
 957             velecsum         = _mm256_add_pd(velecsum,velec);
 958
 959             fscal            = felec;
 960
 961             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 962
 963             /* Calculate temporary vectorial force */
 964             tx               = _mm256_mul_pd(fscal,dx21);
 965             ty               = _mm256_mul_pd(fscal,dy21);
 966             tz               = _mm256_mul_pd(fscal,dz21);
 967
 968             /* Update vectorial force */
 969             fix2             = _mm256_add_pd(fix2,tx);
 970             fiy2             = _mm256_add_pd(fiy2,ty);
 971             fiz2             = _mm256_add_pd(fiz2,tz);
 972
 973             fjx1             = _mm256_add_pd(fjx1,tx);
 974             fjy1             = _mm256_add_pd(fjy1,ty);
 975             fjz1             = _mm256_add_pd(fjz1,tz);
 976
 977             /**************************
 978              * CALCULATE INTERACTIONS *
 979              **************************/
 980
 981             /* COULOMB ELECTROSTATICS */
 982             velec            = _mm256_mul_pd(qq22,rinv22);
 983             felec            = _mm256_mul_pd(velec,rinvsq22);
 984
 985             /* Update potential sum for this i atom from the interaction with this j atom. */
 986             velec            = _mm256_andnot_pd(dummy_mask,velec);
 987             velecsum         = _mm256_add_pd(velecsum,velec);
 988
 989             fscal            = felec;
 990
 991             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 992
 993             /* Calculate temporary vectorial force */
 994             tx               = _mm256_mul_pd(fscal,dx22);
 995             ty               = _mm256_mul_pd(fscal,dy22);
 996             tz               = _mm256_mul_pd(fscal,dz22);
 997
 998             /* Update vectorial force */
 999             fix2             = _mm256_add_pd(fix2,tx);
1000             fiy2             = _mm256_add_pd(fiy2,ty);
1001             fiz2             = _mm256_add_pd(fiz2,tz);
1002
1003             fjx2             = _mm256_add_pd(fjx2,tx);
1004             fjy2             = _mm256_add_pd(fjy2,ty);
1005             fjz2             = _mm256_add_pd(fjz2,tz);
1006
1007             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1008             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1009             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1010             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1011
1012             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1013                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1014
1015             /* Inner loop uses 279 flops */
1016         }
1017
1018         /* End of innermost loop */
1019
1020         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1021                                                  f+i_coord_offset,fshift+i_shift_offset);
1022
1023         ggid                        = gid[iidx];
1024         /* Update potential energies */
1025         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1026         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1027
1028         /* Increment number of inner iterations */
1029         inneriter                  += j_index_end - j_index_start;
1030
1031         /* Outer loop uses 20 flops */
1032     }
1033
1034     /* Increment number of outer iterations */
1035     outeriter        += nri;
1036
1037     /* Update outer/inner flops */
1038
1039     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*279);
1040 }
1041 /*
1042  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_F_avx_256_double
1043  * Electrostatics interaction: Coulomb
1044  * VdW interaction:            CubicSplineTable
1045  * Geometry:                   Water3-Water3
1046  * Calculate force/pot:        Force
1047  */
1048 void
1049 nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_F_avx_256_double
1050                     (t_nblist                    * gmx_restrict       nlist,
1051                      rvec                        * gmx_restrict          xx,
1052                      rvec                        * gmx_restrict          ff,
1053                      t_forcerec                  * gmx_restrict          fr,
1054                      t_mdatoms                   * gmx_restrict     mdatoms,
1055                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1056                      t_nrnb                      * gmx_restrict        nrnb)
1057 {
1058     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1059      * just 0 for non-waters.
1060      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1061      * jnr indices corresponding to data put in the four positions in the SIMD register.
1062      */
1063     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1064     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1065     int              jnrA,jnrB,jnrC,jnrD;
1066     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1067     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1068     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1069     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1070     real             rcutoff_scalar;
1071     real             *shiftvec,*fshift,*x,*f;
1072     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1073     real             scratch[4*DIM];
1074     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1075     real *           vdwioffsetptr0;
1076     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1077     real *           vdwioffsetptr1;
1078     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1079     real *           vdwioffsetptr2;
1080     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1081     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1082     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1083     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1084     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1085     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1086     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1087     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1088     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1089     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1090     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1091     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1092     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1093     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1094     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1095     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1096     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1097     real             *charge;
1098     int              nvdwtype;
1099     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1100     int              *vdwtype;
1101     real             *vdwparam;
1102     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1103     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1104     __m128i          vfitab;
1105     __m128i          ifour       = _mm_set1_epi32(4);
1106     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1107     real             *vftab;
1108     __m256d          dummy_mask,cutoff_mask;
1109     __m128           tmpmask0,tmpmask1;
1110     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1111     __m256d          one     = _mm256_set1_pd(1.0);
1112     __m256d          two     = _mm256_set1_pd(2.0);
1113     x                = xx[0];
1114     f                = ff[0];
1115
1116     nri              = nlist->nri;
1117     iinr             = nlist->iinr;
1118     jindex           = nlist->jindex;
1119     jjnr             = nlist->jjnr;
1120     shiftidx         = nlist->shift;
1121     gid              = nlist->gid;
1122     shiftvec         = fr->shift_vec[0];
1123     fshift           = fr->fshift[0];
1124     facel            = _mm256_set1_pd(fr->epsfac);
1125     charge           = mdatoms->chargeA;
1126     nvdwtype         = fr->ntype;
1127     vdwparam         = fr->nbfp;
1128     vdwtype          = mdatoms->typeA;
1129
1130     vftab            = kernel_data->table_vdw->data;
1131     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
1132
1133     /* Setup water-specific parameters */
1134     inr              = nlist->iinr[0];
1135     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1136     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1137     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1138     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1139
1140     jq0              = _mm256_set1_pd(charge[inr+0]);
1141     jq1              = _mm256_set1_pd(charge[inr+1]);
1142     jq2              = _mm256_set1_pd(charge[inr+2]);
1143     vdwjidx0A        = 2*vdwtype[inr+0];
1144     qq00             = _mm256_mul_pd(iq0,jq0);
1145     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1146     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1147     qq01             = _mm256_mul_pd(iq0,jq1);
1148     qq02             = _mm256_mul_pd(iq0,jq2);
1149     qq10             = _mm256_mul_pd(iq1,jq0);
1150     qq11             = _mm256_mul_pd(iq1,jq1);
1151     qq12             = _mm256_mul_pd(iq1,jq2);
1152     qq20             = _mm256_mul_pd(iq2,jq0);
1153     qq21             = _mm256_mul_pd(iq2,jq1);
1154     qq22             = _mm256_mul_pd(iq2,jq2);
1155
1156     /* Avoid stupid compiler warnings */
1157     jnrA = jnrB = jnrC = jnrD = 0;
1158     j_coord_offsetA = 0;
1159     j_coord_offsetB = 0;
1160     j_coord_offsetC = 0;
1161     j_coord_offsetD = 0;
1162
1163     outeriter        = 0;
1164     inneriter        = 0;
1165
1166     for(iidx=0;iidx<4*DIM;iidx++)
1167     {
1168         scratch[iidx] = 0.0;
1169     }
1170
1171     /* Start outer loop over neighborlists */
1172     for(iidx=0; iidx<nri; iidx++)
1173     {
1174         /* Load shift vector for this list */
1175         i_shift_offset   = DIM*shiftidx[iidx];
1176
1177         /* Load limits for loop over neighbors */
1178         j_index_start    = jindex[iidx];
1179         j_index_end      = jindex[iidx+1];
1180
1181         /* Get outer coordinate index */
1182         inr              = iinr[iidx];
1183         i_coord_offset   = DIM*inr;
1184
1185         /* Load i particle coords and add shift vector */
1186         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1187                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1188
1189         fix0             = _mm256_setzero_pd();
1190         fiy0             = _mm256_setzero_pd();
1191         fiz0             = _mm256_setzero_pd();
1192         fix1             = _mm256_setzero_pd();
1193         fiy1             = _mm256_setzero_pd();
1194         fiz1             = _mm256_setzero_pd();
1195         fix2             = _mm256_setzero_pd();
1196         fiy2             = _mm256_setzero_pd();
1197         fiz2             = _mm256_setzero_pd();
1198
1199         /* Start inner kernel loop */
1200         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1201         {
1202
1203             /* Get j neighbor index, and coordinate index */
1204             jnrA             = jjnr[jidx];
1205             jnrB             = jjnr[jidx+1];
1206             jnrC             = jjnr[jidx+2];
1207             jnrD             = jjnr[jidx+3];
1208             j_coord_offsetA  = DIM*jnrA;
1209             j_coord_offsetB  = DIM*jnrB;
1210             j_coord_offsetC  = DIM*jnrC;
1211             j_coord_offsetD  = DIM*jnrD;
1212
1213             /* load j atom coordinates */
1214             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1215                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1216                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1217
1218             /* Calculate displacement vector */
1219             dx00             = _mm256_sub_pd(ix0,jx0);
1220             dy00             = _mm256_sub_pd(iy0,jy0);
1221             dz00             = _mm256_sub_pd(iz0,jz0);
1222             dx01             = _mm256_sub_pd(ix0,jx1);
1223             dy01             = _mm256_sub_pd(iy0,jy1);
1224             dz01             = _mm256_sub_pd(iz0,jz1);
1225             dx02             = _mm256_sub_pd(ix0,jx2);
1226             dy02             = _mm256_sub_pd(iy0,jy2);
1227             dz02             = _mm256_sub_pd(iz0,jz2);
1228             dx10             = _mm256_sub_pd(ix1,jx0);
1229             dy10             = _mm256_sub_pd(iy1,jy0);
1230             dz10             = _mm256_sub_pd(iz1,jz0);
1231             dx11             = _mm256_sub_pd(ix1,jx1);
1232             dy11             = _mm256_sub_pd(iy1,jy1);
1233             dz11             = _mm256_sub_pd(iz1,jz1);
1234             dx12             = _mm256_sub_pd(ix1,jx2);
1235             dy12             = _mm256_sub_pd(iy1,jy2);
1236             dz12             = _mm256_sub_pd(iz1,jz2);
1237             dx20             = _mm256_sub_pd(ix2,jx0);
1238             dy20             = _mm256_sub_pd(iy2,jy0);
1239             dz20             = _mm256_sub_pd(iz2,jz0);
1240             dx21             = _mm256_sub_pd(ix2,jx1);
1241             dy21             = _mm256_sub_pd(iy2,jy1);
1242             dz21             = _mm256_sub_pd(iz2,jz1);
1243             dx22             = _mm256_sub_pd(ix2,jx2);
1244             dy22             = _mm256_sub_pd(iy2,jy2);
1245             dz22             = _mm256_sub_pd(iz2,jz2);
1246
1247             /* Calculate squared distance and things based on it */
1248             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1249             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1250             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1251             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1252             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1253             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1254             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1255             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1256             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1257
1258             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1259             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1260             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1261             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1262             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1263             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1264             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1265             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1266             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1267
1268             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1269             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1270             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1271             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1272             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1273             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1274             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1275             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1276             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1277
1278             fjx0             = _mm256_setzero_pd();
1279             fjy0             = _mm256_setzero_pd();
1280             fjz0             = _mm256_setzero_pd();
1281             fjx1             = _mm256_setzero_pd();
1282             fjy1             = _mm256_setzero_pd();
1283             fjz1             = _mm256_setzero_pd();
1284             fjx2             = _mm256_setzero_pd();
1285             fjy2             = _mm256_setzero_pd();
1286             fjz2             = _mm256_setzero_pd();
1287
1288             /**************************
1289              * CALCULATE INTERACTIONS *
1290              **************************/
1291
1292             r00              = _mm256_mul_pd(rsq00,rinv00);
1293
1294             /* Calculate table index by multiplying r with table scale and truncate to integer */
1295             rt               = _mm256_mul_pd(r00,vftabscale);
1296             vfitab           = _mm256_cvttpd_epi32(rt);
1297             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1298             vfitab           = _mm_slli_epi32(vfitab,3);
1299
1300             /* COULOMB ELECTROSTATICS */
1301             velec            = _mm256_mul_pd(qq00,rinv00);
1302             felec            = _mm256_mul_pd(velec,rinvsq00);
1303
1304             /* CUBIC SPLINE TABLE DISPERSION */
1305             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1306             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1307             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1308             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1309             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1310             Heps             = _mm256_mul_pd(vfeps,H);
1311             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1312             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1313             fvdw6            = _mm256_mul_pd(c6_00,FF);
1314
1315             /* CUBIC SPLINE TABLE REPULSION */
1316             vfitab           = _mm_add_epi32(vfitab,ifour);
1317             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1318             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1319             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1320             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1321             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1322             Heps             = _mm256_mul_pd(vfeps,H);
1323             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1324             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1325             fvdw12           = _mm256_mul_pd(c12_00,FF);
1326             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1327
1328             fscal            = _mm256_add_pd(felec,fvdw);
1329
1330             /* Calculate temporary vectorial force */
1331             tx               = _mm256_mul_pd(fscal,dx00);
1332             ty               = _mm256_mul_pd(fscal,dy00);
1333             tz               = _mm256_mul_pd(fscal,dz00);
1334
1335             /* Update vectorial force */
1336             fix0             = _mm256_add_pd(fix0,tx);
1337             fiy0             = _mm256_add_pd(fiy0,ty);
1338             fiz0             = _mm256_add_pd(fiz0,tz);
1339
1340             fjx0             = _mm256_add_pd(fjx0,tx);
1341             fjy0             = _mm256_add_pd(fjy0,ty);
1342             fjz0             = _mm256_add_pd(fjz0,tz);
1343
1344             /**************************
1345              * CALCULATE INTERACTIONS *
1346              **************************/
1347
1348             /* COULOMB ELECTROSTATICS */
1349             velec            = _mm256_mul_pd(qq01,rinv01);
1350             felec            = _mm256_mul_pd(velec,rinvsq01);
1351
1352             fscal            = felec;
1353
1354             /* Calculate temporary vectorial force */
1355             tx               = _mm256_mul_pd(fscal,dx01);
1356             ty               = _mm256_mul_pd(fscal,dy01);
1357             tz               = _mm256_mul_pd(fscal,dz01);
1358
1359             /* Update vectorial force */
1360             fix0             = _mm256_add_pd(fix0,tx);
1361             fiy0             = _mm256_add_pd(fiy0,ty);
1362             fiz0             = _mm256_add_pd(fiz0,tz);
1363
1364             fjx1             = _mm256_add_pd(fjx1,tx);
1365             fjy1             = _mm256_add_pd(fjy1,ty);
1366             fjz1             = _mm256_add_pd(fjz1,tz);
1367
1368             /**************************
1369              * CALCULATE INTERACTIONS *
1370              **************************/
1371
1372             /* COULOMB ELECTROSTATICS */
1373             velec            = _mm256_mul_pd(qq02,rinv02);
1374             felec            = _mm256_mul_pd(velec,rinvsq02);
1375
1376             fscal            = felec;
1377
1378             /* Calculate temporary vectorial force */
1379             tx               = _mm256_mul_pd(fscal,dx02);
1380             ty               = _mm256_mul_pd(fscal,dy02);
1381             tz               = _mm256_mul_pd(fscal,dz02);
1382
1383             /* Update vectorial force */
1384             fix0             = _mm256_add_pd(fix0,tx);
1385             fiy0             = _mm256_add_pd(fiy0,ty);
1386             fiz0             = _mm256_add_pd(fiz0,tz);
1387
1388             fjx2             = _mm256_add_pd(fjx2,tx);
1389             fjy2             = _mm256_add_pd(fjy2,ty);
1390             fjz2             = _mm256_add_pd(fjz2,tz);
1391
1392             /**************************
1393              * CALCULATE INTERACTIONS *
1394              **************************/
1395
1396             /* COULOMB ELECTROSTATICS */
1397             velec            = _mm256_mul_pd(qq10,rinv10);
1398             felec            = _mm256_mul_pd(velec,rinvsq10);
1399
1400             fscal            = felec;
1401
1402             /* Calculate temporary vectorial force */
1403             tx               = _mm256_mul_pd(fscal,dx10);
1404             ty               = _mm256_mul_pd(fscal,dy10);
1405             tz               = _mm256_mul_pd(fscal,dz10);
1406
1407             /* Update vectorial force */
1408             fix1             = _mm256_add_pd(fix1,tx);
1409             fiy1             = _mm256_add_pd(fiy1,ty);
1410             fiz1             = _mm256_add_pd(fiz1,tz);
1411
1412             fjx0             = _mm256_add_pd(fjx0,tx);
1413             fjy0             = _mm256_add_pd(fjy0,ty);
1414             fjz0             = _mm256_add_pd(fjz0,tz);
1415
1416             /**************************
1417              * CALCULATE INTERACTIONS *
1418              **************************/
1419
1420             /* COULOMB ELECTROSTATICS */
1421             velec            = _mm256_mul_pd(qq11,rinv11);
1422             felec            = _mm256_mul_pd(velec,rinvsq11);
1423
1424             fscal            = felec;
1425
1426             /* Calculate temporary vectorial force */
1427             tx               = _mm256_mul_pd(fscal,dx11);
1428             ty               = _mm256_mul_pd(fscal,dy11);
1429             tz               = _mm256_mul_pd(fscal,dz11);
1430
1431             /* Update vectorial force */
1432             fix1             = _mm256_add_pd(fix1,tx);
1433             fiy1             = _mm256_add_pd(fiy1,ty);
1434             fiz1             = _mm256_add_pd(fiz1,tz);
1435
1436             fjx1             = _mm256_add_pd(fjx1,tx);
1437             fjy1             = _mm256_add_pd(fjy1,ty);
1438             fjz1             = _mm256_add_pd(fjz1,tz);
1439
1440             /**************************
1441              * CALCULATE INTERACTIONS *
1442              **************************/
1443
1444             /* COULOMB ELECTROSTATICS */
1445             velec            = _mm256_mul_pd(qq12,rinv12);
1446             felec            = _mm256_mul_pd(velec,rinvsq12);
1447
1448             fscal            = felec;
1449
1450             /* Calculate temporary vectorial force */
1451             tx               = _mm256_mul_pd(fscal,dx12);
1452             ty               = _mm256_mul_pd(fscal,dy12);
1453             tz               = _mm256_mul_pd(fscal,dz12);
1454
1455             /* Update vectorial force */
1456             fix1             = _mm256_add_pd(fix1,tx);
1457             fiy1             = _mm256_add_pd(fiy1,ty);
1458             fiz1             = _mm256_add_pd(fiz1,tz);
1459
1460             fjx2             = _mm256_add_pd(fjx2,tx);
1461             fjy2             = _mm256_add_pd(fjy2,ty);
1462             fjz2             = _mm256_add_pd(fjz2,tz);
1463
1464             /**************************
1465              * CALCULATE INTERACTIONS *
1466              **************************/
1467
1468             /* COULOMB ELECTROSTATICS */
1469             velec            = _mm256_mul_pd(qq20,rinv20);
1470             felec            = _mm256_mul_pd(velec,rinvsq20);
1471
1472             fscal            = felec;
1473
1474             /* Calculate temporary vectorial force */
1475             tx               = _mm256_mul_pd(fscal,dx20);
1476             ty               = _mm256_mul_pd(fscal,dy20);
1477             tz               = _mm256_mul_pd(fscal,dz20);
1478
1479             /* Update vectorial force */
1480             fix2             = _mm256_add_pd(fix2,tx);
1481             fiy2             = _mm256_add_pd(fiy2,ty);
1482             fiz2             = _mm256_add_pd(fiz2,tz);
1483
1484             fjx0             = _mm256_add_pd(fjx0,tx);
1485             fjy0             = _mm256_add_pd(fjy0,ty);
1486             fjz0             = _mm256_add_pd(fjz0,tz);
1487
1488             /**************************
1489              * CALCULATE INTERACTIONS *
1490              **************************/
1491
1492             /* COULOMB ELECTROSTATICS */
1493             velec            = _mm256_mul_pd(qq21,rinv21);
1494             felec            = _mm256_mul_pd(velec,rinvsq21);
1495
1496             fscal            = felec;
1497
1498             /* Calculate temporary vectorial force */
1499             tx               = _mm256_mul_pd(fscal,dx21);
1500             ty               = _mm256_mul_pd(fscal,dy21);
1501             tz               = _mm256_mul_pd(fscal,dz21);
1502
1503             /* Update vectorial force */
1504             fix2             = _mm256_add_pd(fix2,tx);
1505             fiy2             = _mm256_add_pd(fiy2,ty);
1506             fiz2             = _mm256_add_pd(fiz2,tz);
1507
1508             fjx1             = _mm256_add_pd(fjx1,tx);
1509             fjy1             = _mm256_add_pd(fjy1,ty);
1510             fjz1             = _mm256_add_pd(fjz1,tz);
1511
1512             /**************************
1513              * CALCULATE INTERACTIONS *
1514              **************************/
1515
1516             /* COULOMB ELECTROSTATICS */
1517             velec            = _mm256_mul_pd(qq22,rinv22);
1518             felec            = _mm256_mul_pd(velec,rinvsq22);
1519
1520             fscal            = felec;
1521
1522             /* Calculate temporary vectorial force */
1523             tx               = _mm256_mul_pd(fscal,dx22);
1524             ty               = _mm256_mul_pd(fscal,dy22);
1525             tz               = _mm256_mul_pd(fscal,dz22);
1526
1527             /* Update vectorial force */
1528             fix2             = _mm256_add_pd(fix2,tx);
1529             fiy2             = _mm256_add_pd(fiy2,ty);
1530             fiz2             = _mm256_add_pd(fiz2,tz);
1531
1532             fjx2             = _mm256_add_pd(fjx2,tx);
1533             fjy2             = _mm256_add_pd(fjy2,ty);
1534             fjz2             = _mm256_add_pd(fjz2,tz);
1535
1536             fjptrA             = f+j_coord_offsetA;
1537             fjptrB             = f+j_coord_offsetB;
1538             fjptrC             = f+j_coord_offsetC;
1539             fjptrD             = f+j_coord_offsetD;
1540
1541             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1542                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1543
1544             /* Inner loop uses 261 flops */
1545         }
1546
1547         if(jidx<j_index_end)
1548         {
1549
1550             /* Get j neighbor index, and coordinate index */
1551             jnrlistA         = jjnr[jidx];
1552             jnrlistB         = jjnr[jidx+1];
1553             jnrlistC         = jjnr[jidx+2];
1554             jnrlistD         = jjnr[jidx+3];
1555             /* Sign of each element will be negative for non-real atoms.
1556              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1557              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1558              */
1559             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1560
1561             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1562             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1563             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1564
1565             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1566             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1567             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1568             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1569             j_coord_offsetA  = DIM*jnrA;
1570             j_coord_offsetB  = DIM*jnrB;
1571             j_coord_offsetC  = DIM*jnrC;
1572             j_coord_offsetD  = DIM*jnrD;
1573
1574             /* load j atom coordinates */
1575             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1576                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1577                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1578
1579             /* Calculate displacement vector */
1580             dx00             = _mm256_sub_pd(ix0,jx0);
1581             dy00             = _mm256_sub_pd(iy0,jy0);
1582             dz00             = _mm256_sub_pd(iz0,jz0);
1583             dx01             = _mm256_sub_pd(ix0,jx1);
1584             dy01             = _mm256_sub_pd(iy0,jy1);
1585             dz01             = _mm256_sub_pd(iz0,jz1);
1586             dx02             = _mm256_sub_pd(ix0,jx2);
1587             dy02             = _mm256_sub_pd(iy0,jy2);
1588             dz02             = _mm256_sub_pd(iz0,jz2);
1589             dx10             = _mm256_sub_pd(ix1,jx0);
1590             dy10             = _mm256_sub_pd(iy1,jy0);
1591             dz10             = _mm256_sub_pd(iz1,jz0);
1592             dx11             = _mm256_sub_pd(ix1,jx1);
1593             dy11             = _mm256_sub_pd(iy1,jy1);
1594             dz11             = _mm256_sub_pd(iz1,jz1);
1595             dx12             = _mm256_sub_pd(ix1,jx2);
1596             dy12             = _mm256_sub_pd(iy1,jy2);
1597             dz12             = _mm256_sub_pd(iz1,jz2);
1598             dx20             = _mm256_sub_pd(ix2,jx0);
1599             dy20             = _mm256_sub_pd(iy2,jy0);
1600             dz20             = _mm256_sub_pd(iz2,jz0);
1601             dx21             = _mm256_sub_pd(ix2,jx1);
1602             dy21             = _mm256_sub_pd(iy2,jy1);
1603             dz21             = _mm256_sub_pd(iz2,jz1);
1604             dx22             = _mm256_sub_pd(ix2,jx2);
1605             dy22             = _mm256_sub_pd(iy2,jy2);
1606             dz22             = _mm256_sub_pd(iz2,jz2);
1607
1608             /* Calculate squared distance and things based on it */
1609             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1610             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1611             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1612             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1613             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1614             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1615             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1616             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1617             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1618
1619             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1620             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1621             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1622             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1623             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1624             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1625             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1626             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1627             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1628
1629             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1630             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1631             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1632             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1633             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1634             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1635             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1636             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1637             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1638
1639             fjx0             = _mm256_setzero_pd();
1640             fjy0             = _mm256_setzero_pd();
1641             fjz0             = _mm256_setzero_pd();
1642             fjx1             = _mm256_setzero_pd();
1643             fjy1             = _mm256_setzero_pd();
1644             fjz1             = _mm256_setzero_pd();
1645             fjx2             = _mm256_setzero_pd();
1646             fjy2             = _mm256_setzero_pd();
1647             fjz2             = _mm256_setzero_pd();
1648
1649             /**************************
1650              * CALCULATE INTERACTIONS *
1651              **************************/
1652
1653             r00              = _mm256_mul_pd(rsq00,rinv00);
1654             r00              = _mm256_andnot_pd(dummy_mask,r00);
1655
1656             /* Calculate table index by multiplying r with table scale and truncate to integer */
1657             rt               = _mm256_mul_pd(r00,vftabscale);
1658             vfitab           = _mm256_cvttpd_epi32(rt);
1659             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1660             vfitab           = _mm_slli_epi32(vfitab,3);
1661
1662             /* COULOMB ELECTROSTATICS */
1663             velec            = _mm256_mul_pd(qq00,rinv00);
1664             felec            = _mm256_mul_pd(velec,rinvsq00);
1665
1666             /* CUBIC SPLINE TABLE DISPERSION */
1667             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1668             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1669             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1670             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1671             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1672             Heps             = _mm256_mul_pd(vfeps,H);
1673             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1674             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1675             fvdw6            = _mm256_mul_pd(c6_00,FF);
1676
1677             /* CUBIC SPLINE TABLE REPULSION */
1678             vfitab           = _mm_add_epi32(vfitab,ifour);
1679             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1680             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1681             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1682             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1683             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1684             Heps             = _mm256_mul_pd(vfeps,H);
1685             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1686             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1687             fvdw12           = _mm256_mul_pd(c12_00,FF);
1688             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1689
1690             fscal            = _mm256_add_pd(felec,fvdw);
1691
1692             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1693
1694             /* Calculate temporary vectorial force */
1695             tx               = _mm256_mul_pd(fscal,dx00);
1696             ty               = _mm256_mul_pd(fscal,dy00);
1697             tz               = _mm256_mul_pd(fscal,dz00);
1698
1699             /* Update vectorial force */
1700             fix0             = _mm256_add_pd(fix0,tx);
1701             fiy0             = _mm256_add_pd(fiy0,ty);
1702             fiz0             = _mm256_add_pd(fiz0,tz);
1703
1704             fjx0             = _mm256_add_pd(fjx0,tx);
1705             fjy0             = _mm256_add_pd(fjy0,ty);
1706             fjz0             = _mm256_add_pd(fjz0,tz);
1707
1708             /**************************
1709              * CALCULATE INTERACTIONS *
1710              **************************/
1711
1712             /* COULOMB ELECTROSTATICS */
1713             velec            = _mm256_mul_pd(qq01,rinv01);
1714             felec            = _mm256_mul_pd(velec,rinvsq01);
1715
1716             fscal            = felec;
1717
1718             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1719
1720             /* Calculate temporary vectorial force */
1721             tx               = _mm256_mul_pd(fscal,dx01);
1722             ty               = _mm256_mul_pd(fscal,dy01);
1723             tz               = _mm256_mul_pd(fscal,dz01);
1724
1725             /* Update vectorial force */
1726             fix0             = _mm256_add_pd(fix0,tx);
1727             fiy0             = _mm256_add_pd(fiy0,ty);
1728             fiz0             = _mm256_add_pd(fiz0,tz);
1729
1730             fjx1             = _mm256_add_pd(fjx1,tx);
1731             fjy1             = _mm256_add_pd(fjy1,ty);
1732             fjz1             = _mm256_add_pd(fjz1,tz);
1733
1734             /**************************
1735              * CALCULATE INTERACTIONS *
1736              **************************/
1737
1738             /* COULOMB ELECTROSTATICS */
1739             velec            = _mm256_mul_pd(qq02,rinv02);
1740             felec            = _mm256_mul_pd(velec,rinvsq02);
1741
1742             fscal            = felec;
1743
1744             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1745
1746             /* Calculate temporary vectorial force */
1747             tx               = _mm256_mul_pd(fscal,dx02);
1748             ty               = _mm256_mul_pd(fscal,dy02);
1749             tz               = _mm256_mul_pd(fscal,dz02);
1750
1751             /* Update vectorial force */
1752             fix0             = _mm256_add_pd(fix0,tx);
1753             fiy0             = _mm256_add_pd(fiy0,ty);
1754             fiz0             = _mm256_add_pd(fiz0,tz);
1755
1756             fjx2             = _mm256_add_pd(fjx2,tx);
1757             fjy2             = _mm256_add_pd(fjy2,ty);
1758             fjz2             = _mm256_add_pd(fjz2,tz);
1759
1760             /**************************
1761              * CALCULATE INTERACTIONS *
1762              **************************/
1763
1764             /* COULOMB ELECTROSTATICS */
1765             velec            = _mm256_mul_pd(qq10,rinv10);
1766             felec            = _mm256_mul_pd(velec,rinvsq10);
1767
1768             fscal            = felec;
1769
1770             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1771
1772             /* Calculate temporary vectorial force */
1773             tx               = _mm256_mul_pd(fscal,dx10);
1774             ty               = _mm256_mul_pd(fscal,dy10);
1775             tz               = _mm256_mul_pd(fscal,dz10);
1776
1777             /* Update vectorial force */
1778             fix1             = _mm256_add_pd(fix1,tx);
1779             fiy1             = _mm256_add_pd(fiy1,ty);
1780             fiz1             = _mm256_add_pd(fiz1,tz);
1781
1782             fjx0             = _mm256_add_pd(fjx0,tx);
1783             fjy0             = _mm256_add_pd(fjy0,ty);
1784             fjz0             = _mm256_add_pd(fjz0,tz);
1785
1786             /**************************
1787              * CALCULATE INTERACTIONS *
1788              **************************/
1789
1790             /* COULOMB ELECTROSTATICS */
1791             velec            = _mm256_mul_pd(qq11,rinv11);
1792             felec            = _mm256_mul_pd(velec,rinvsq11);
1793
1794             fscal            = felec;
1795
1796             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1797
1798             /* Calculate temporary vectorial force */
1799             tx               = _mm256_mul_pd(fscal,dx11);
1800             ty               = _mm256_mul_pd(fscal,dy11);
1801             tz               = _mm256_mul_pd(fscal,dz11);
1802
1803             /* Update vectorial force */
1804             fix1             = _mm256_add_pd(fix1,tx);
1805             fiy1             = _mm256_add_pd(fiy1,ty);
1806             fiz1             = _mm256_add_pd(fiz1,tz);
1807
1808             fjx1             = _mm256_add_pd(fjx1,tx);
1809             fjy1             = _mm256_add_pd(fjy1,ty);
1810             fjz1             = _mm256_add_pd(fjz1,tz);
1811
1812             /**************************
1813              * CALCULATE INTERACTIONS *
1814              **************************/
1815
1816             /* COULOMB ELECTROSTATICS */
1817             velec            = _mm256_mul_pd(qq12,rinv12);
1818             felec            = _mm256_mul_pd(velec,rinvsq12);
1819
1820             fscal            = felec;
1821
1822             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1823
1824             /* Calculate temporary vectorial force */
1825             tx               = _mm256_mul_pd(fscal,dx12);
1826             ty               = _mm256_mul_pd(fscal,dy12);
1827             tz               = _mm256_mul_pd(fscal,dz12);
1828
1829             /* Update vectorial force */
1830             fix1             = _mm256_add_pd(fix1,tx);
1831             fiy1             = _mm256_add_pd(fiy1,ty);
1832             fiz1             = _mm256_add_pd(fiz1,tz);
1833
1834             fjx2             = _mm256_add_pd(fjx2,tx);
1835             fjy2             = _mm256_add_pd(fjy2,ty);
1836             fjz2             = _mm256_add_pd(fjz2,tz);
1837
1838             /**************************
1839              * CALCULATE INTERACTIONS *
1840              **************************/
1841
1842             /* COULOMB ELECTROSTATICS */
1843             velec            = _mm256_mul_pd(qq20,rinv20);
1844             felec            = _mm256_mul_pd(velec,rinvsq20);
1845
1846             fscal            = felec;
1847
1848             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1849
1850             /* Calculate temporary vectorial force */
1851             tx               = _mm256_mul_pd(fscal,dx20);
1852             ty               = _mm256_mul_pd(fscal,dy20);
1853             tz               = _mm256_mul_pd(fscal,dz20);
1854
1855             /* Update vectorial force */
1856             fix2             = _mm256_add_pd(fix2,tx);
1857             fiy2             = _mm256_add_pd(fiy2,ty);
1858             fiz2             = _mm256_add_pd(fiz2,tz);
1859
1860             fjx0             = _mm256_add_pd(fjx0,tx);
1861             fjy0             = _mm256_add_pd(fjy0,ty);
1862             fjz0             = _mm256_add_pd(fjz0,tz);
1863
1864             /**************************
1865              * CALCULATE INTERACTIONS *
1866              **************************/
1867
1868             /* COULOMB ELECTROSTATICS */
1869             velec            = _mm256_mul_pd(qq21,rinv21);
1870             felec            = _mm256_mul_pd(velec,rinvsq21);
1871
1872             fscal            = felec;
1873
1874             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1875
1876             /* Calculate temporary vectorial force */
1877             tx               = _mm256_mul_pd(fscal,dx21);
1878             ty               = _mm256_mul_pd(fscal,dy21);
1879             tz               = _mm256_mul_pd(fscal,dz21);
1880
1881             /* Update vectorial force */
1882             fix2             = _mm256_add_pd(fix2,tx);
1883             fiy2             = _mm256_add_pd(fiy2,ty);
1884             fiz2             = _mm256_add_pd(fiz2,tz);
1885
1886             fjx1             = _mm256_add_pd(fjx1,tx);
1887             fjy1             = _mm256_add_pd(fjy1,ty);
1888             fjz1             = _mm256_add_pd(fjz1,tz);
1889
1890             /**************************
1891              * CALCULATE INTERACTIONS *
1892              **************************/
1893
1894             /* COULOMB ELECTROSTATICS */
1895             velec            = _mm256_mul_pd(qq22,rinv22);
1896             felec            = _mm256_mul_pd(velec,rinvsq22);
1897
1898             fscal            = felec;
1899
1900             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1901
1902             /* Calculate temporary vectorial force */
1903             tx               = _mm256_mul_pd(fscal,dx22);
1904             ty               = _mm256_mul_pd(fscal,dy22);
1905             tz               = _mm256_mul_pd(fscal,dz22);
1906
1907             /* Update vectorial force */
1908             fix2             = _mm256_add_pd(fix2,tx);
1909             fiy2             = _mm256_add_pd(fiy2,ty);
1910             fiz2             = _mm256_add_pd(fiz2,tz);
1911
1912             fjx2             = _mm256_add_pd(fjx2,tx);
1913             fjy2             = _mm256_add_pd(fjy2,ty);
1914             fjz2             = _mm256_add_pd(fjz2,tz);
1915
1916             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1917             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1918             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1919             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1920
1921             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1922                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1923
1924             /* Inner loop uses 262 flops */
1925         }
1926
1927         /* End of innermost loop */
1928
1929         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1930                                                  f+i_coord_offset,fshift+i_shift_offset);
1931
1932         /* Increment number of inner iterations */
1933         inneriter                  += j_index_end - j_index_start;
1934
1935         /* Outer loop uses 18 flops */
1936     }
1937
1938     /* Increment number of outer iterations */
1939     outeriter        += nri;
1940
1941     /* Update outer/inner flops */
1942
1943     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*262);
1944 }