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) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_double.h"
  48 #include "kernelutil_x86_avx_256_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_VF_avx_256_double
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water3-Water3
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_VF_avx_256_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  82     real             scratch[4*DIM];
  83     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  84     real *           vdwioffsetptr0;
  85     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  86     real *           vdwioffsetptr1;
  87     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  88     real *           vdwioffsetptr2;
  89     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  90     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  91     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  93     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  94     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  95     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  96     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  97     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  98     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  99     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 100     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 101     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 102     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 103     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 104     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 105     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 106     real             *charge;
 107     int              nvdwtype;
 108     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 109     int              *vdwtype;
 110     real             *vdwparam;
 111     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 112     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 113     __m128i          vfitab;
 114     __m128i          ifour       = _mm_set1_epi32(4);
 115     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 116     real             *vftab;
 117     __m256d          dummy_mask,cutoff_mask;
 118     __m128           tmpmask0,tmpmask1;
 119     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 120     __m256d          one     = _mm256_set1_pd(1.0);
 121     __m256d          two     = _mm256_set1_pd(2.0);
 122     x                = xx[0];
 123     f                = ff[0];
 124
 125     nri              = nlist->nri;
 126     iinr             = nlist->iinr;
 127     jindex           = nlist->jindex;
 128     jjnr             = nlist->jjnr;
 129     shiftidx         = nlist->shift;
 130     gid              = nlist->gid;
 131     shiftvec         = fr->shift_vec[0];
 132     fshift           = fr->fshift[0];
 133     facel            = _mm256_set1_pd(fr->epsfac);
 134     charge           = mdatoms->chargeA;
 135     nvdwtype         = fr->ntype;
 136     vdwparam         = fr->nbfp;
 137     vdwtype          = mdatoms->typeA;
 138
 139     vftab            = kernel_data->table_vdw->data;
 140     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
 141
 142     /* Setup water-specific parameters */
 143     inr              = nlist->iinr[0];
 144     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 145     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 146     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 147     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 148
 149     jq0              = _mm256_set1_pd(charge[inr+0]);
 150     jq1              = _mm256_set1_pd(charge[inr+1]);
 151     jq2              = _mm256_set1_pd(charge[inr+2]);
 152     vdwjidx0A        = 2*vdwtype[inr+0];
 153     qq00             = _mm256_mul_pd(iq0,jq0);
 154     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 155     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 156     qq01             = _mm256_mul_pd(iq0,jq1);
 157     qq02             = _mm256_mul_pd(iq0,jq2);
 158     qq10             = _mm256_mul_pd(iq1,jq0);
 159     qq11             = _mm256_mul_pd(iq1,jq1);
 160     qq12             = _mm256_mul_pd(iq1,jq2);
 161     qq20             = _mm256_mul_pd(iq2,jq0);
 162     qq21             = _mm256_mul_pd(iq2,jq1);
 163     qq22             = _mm256_mul_pd(iq2,jq2);
 164
 165     /* Avoid stupid compiler warnings */
 166     jnrA = jnrB = jnrC = jnrD = 0;
 167     j_coord_offsetA = 0;
 168     j_coord_offsetB = 0;
 169     j_coord_offsetC = 0;
 170     j_coord_offsetD = 0;
 171
 172     outeriter        = 0;
 173     inneriter        = 0;
 174
 175     for(iidx=0;iidx<4*DIM;iidx++)
 176     {
 177         scratch[iidx] = 0.0;
 178     }
 179
 180     /* Start outer loop over neighborlists */
 181     for(iidx=0; iidx<nri; iidx++)
 182     {
 183         /* Load shift vector for this list */
 184         i_shift_offset   = DIM*shiftidx[iidx];
 185
 186         /* Load limits for loop over neighbors */
 187         j_index_start    = jindex[iidx];
 188         j_index_end      = jindex[iidx+1];
 189
 190         /* Get outer coordinate index */
 191         inr              = iinr[iidx];
 192         i_coord_offset   = DIM*inr;
 193
 194         /* Load i particle coords and add shift vector */
 195         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 196                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 197
 198         fix0             = _mm256_setzero_pd();
 199         fiy0             = _mm256_setzero_pd();
 200         fiz0             = _mm256_setzero_pd();
 201         fix1             = _mm256_setzero_pd();
 202         fiy1             = _mm256_setzero_pd();
 203         fiz1             = _mm256_setzero_pd();
 204         fix2             = _mm256_setzero_pd();
 205         fiy2             = _mm256_setzero_pd();
 206         fiz2             = _mm256_setzero_pd();
 207
 208         /* Reset potential sums */
 209         velecsum         = _mm256_setzero_pd();
 210         vvdwsum          = _mm256_setzero_pd();
 211
 212         /* Start inner kernel loop */
 213         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 214         {
 215
 216             /* Get j neighbor index, and coordinate index */
 217             jnrA             = jjnr[jidx];
 218             jnrB             = jjnr[jidx+1];
 219             jnrC             = jjnr[jidx+2];
 220             jnrD             = jjnr[jidx+3];
 221             j_coord_offsetA  = DIM*jnrA;
 222             j_coord_offsetB  = DIM*jnrB;
 223             j_coord_offsetC  = DIM*jnrC;
 224             j_coord_offsetD  = DIM*jnrD;
 225
 226             /* load j atom coordinates */
 227             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 228                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 229                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 230
 231             /* Calculate displacement vector */
 232             dx00             = _mm256_sub_pd(ix0,jx0);
 233             dy00             = _mm256_sub_pd(iy0,jy0);
 234             dz00             = _mm256_sub_pd(iz0,jz0);
 235             dx01             = _mm256_sub_pd(ix0,jx1);
 236             dy01             = _mm256_sub_pd(iy0,jy1);
 237             dz01             = _mm256_sub_pd(iz0,jz1);
 238             dx02             = _mm256_sub_pd(ix0,jx2);
 239             dy02             = _mm256_sub_pd(iy0,jy2);
 240             dz02             = _mm256_sub_pd(iz0,jz2);
 241             dx10             = _mm256_sub_pd(ix1,jx0);
 242             dy10             = _mm256_sub_pd(iy1,jy0);
 243             dz10             = _mm256_sub_pd(iz1,jz0);
 244             dx11             = _mm256_sub_pd(ix1,jx1);
 245             dy11             = _mm256_sub_pd(iy1,jy1);
 246             dz11             = _mm256_sub_pd(iz1,jz1);
 247             dx12             = _mm256_sub_pd(ix1,jx2);
 248             dy12             = _mm256_sub_pd(iy1,jy2);
 249             dz12             = _mm256_sub_pd(iz1,jz2);
 250             dx20             = _mm256_sub_pd(ix2,jx0);
 251             dy20             = _mm256_sub_pd(iy2,jy0);
 252             dz20             = _mm256_sub_pd(iz2,jz0);
 253             dx21             = _mm256_sub_pd(ix2,jx1);
 254             dy21             = _mm256_sub_pd(iy2,jy1);
 255             dz21             = _mm256_sub_pd(iz2,jz1);
 256             dx22             = _mm256_sub_pd(ix2,jx2);
 257             dy22             = _mm256_sub_pd(iy2,jy2);
 258             dz22             = _mm256_sub_pd(iz2,jz2);
 259
 260             /* Calculate squared distance and things based on it */
 261             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 262             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 263             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 264             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 265             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 266             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 267             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 268             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 269             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 270
 271             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 272             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 273             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 274             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 275             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 276             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 277             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 278             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 279             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 280
 281             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 282             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 283             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 284             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 285             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 286             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 287             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 288             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 289             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 290
 291             fjx0             = _mm256_setzero_pd();
 292             fjy0             = _mm256_setzero_pd();
 293             fjz0             = _mm256_setzero_pd();
 294             fjx1             = _mm256_setzero_pd();
 295             fjy1             = _mm256_setzero_pd();
 296             fjz1             = _mm256_setzero_pd();
 297             fjx2             = _mm256_setzero_pd();
 298             fjy2             = _mm256_setzero_pd();
 299             fjz2             = _mm256_setzero_pd();
 300
 301             /**************************
 302              * CALCULATE INTERACTIONS *
 303              **************************/
 304
 305             r00              = _mm256_mul_pd(rsq00,rinv00);
 306
 307             /* Calculate table index by multiplying r with table scale and truncate to integer */
 308             rt               = _mm256_mul_pd(r00,vftabscale);
 309             vfitab           = _mm256_cvttpd_epi32(rt);
 310             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 311             vfitab           = _mm_slli_epi32(vfitab,3);
 312
 313             /* COULOMB ELECTROSTATICS */
 314             velec            = _mm256_mul_pd(qq00,rinv00);
 315             felec            = _mm256_mul_pd(velec,rinvsq00);
 316
 317             /* CUBIC SPLINE TABLE DISPERSION */
 318             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 319             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 320             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 321             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 322             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 323             Heps             = _mm256_mul_pd(vfeps,H);
 324             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 325             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 326             vvdw6            = _mm256_mul_pd(c6_00,VV);
 327             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 328             fvdw6            = _mm256_mul_pd(c6_00,FF);
 329
 330             /* CUBIC SPLINE TABLE REPULSION */
 331             vfitab           = _mm_add_epi32(vfitab,ifour);
 332             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 333             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 334             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 335             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 336             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 337             Heps             = _mm256_mul_pd(vfeps,H);
 338             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 339             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 340             vvdw12           = _mm256_mul_pd(c12_00,VV);
 341             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 342             fvdw12           = _mm256_mul_pd(c12_00,FF);
 343             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 344             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 345
 346             /* Update potential sum for this i atom from the interaction with this j atom. */
 347             velecsum         = _mm256_add_pd(velecsum,velec);
 348             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 349
 350             fscal            = _mm256_add_pd(felec,fvdw);
 351
 352             /* Calculate temporary vectorial force */
 353             tx               = _mm256_mul_pd(fscal,dx00);
 354             ty               = _mm256_mul_pd(fscal,dy00);
 355             tz               = _mm256_mul_pd(fscal,dz00);
 356
 357             /* Update vectorial force */
 358             fix0             = _mm256_add_pd(fix0,tx);
 359             fiy0             = _mm256_add_pd(fiy0,ty);
 360             fiz0             = _mm256_add_pd(fiz0,tz);
 361
 362             fjx0             = _mm256_add_pd(fjx0,tx);
 363             fjy0             = _mm256_add_pd(fjy0,ty);
 364             fjz0             = _mm256_add_pd(fjz0,tz);
 365
 366             /**************************
 367              * CALCULATE INTERACTIONS *
 368              **************************/
 369
 370             /* COULOMB ELECTROSTATICS */
 371             velec            = _mm256_mul_pd(qq01,rinv01);
 372             felec            = _mm256_mul_pd(velec,rinvsq01);
 373
 374             /* Update potential sum for this i atom from the interaction with this j atom. */
 375             velecsum         = _mm256_add_pd(velecsum,velec);
 376
 377             fscal            = felec;
 378
 379             /* Calculate temporary vectorial force */
 380             tx               = _mm256_mul_pd(fscal,dx01);
 381             ty               = _mm256_mul_pd(fscal,dy01);
 382             tz               = _mm256_mul_pd(fscal,dz01);
 383
 384             /* Update vectorial force */
 385             fix0             = _mm256_add_pd(fix0,tx);
 386             fiy0             = _mm256_add_pd(fiy0,ty);
 387             fiz0             = _mm256_add_pd(fiz0,tz);
 388
 389             fjx1             = _mm256_add_pd(fjx1,tx);
 390             fjy1             = _mm256_add_pd(fjy1,ty);
 391             fjz1             = _mm256_add_pd(fjz1,tz);
 392
 393             /**************************
 394              * CALCULATE INTERACTIONS *
 395              **************************/
 396
 397             /* COULOMB ELECTROSTATICS */
 398             velec            = _mm256_mul_pd(qq02,rinv02);
 399             felec            = _mm256_mul_pd(velec,rinvsq02);
 400
 401             /* Update potential sum for this i atom from the interaction with this j atom. */
 402             velecsum         = _mm256_add_pd(velecsum,velec);
 403
 404             fscal            = felec;
 405
 406             /* Calculate temporary vectorial force */
 407             tx               = _mm256_mul_pd(fscal,dx02);
 408             ty               = _mm256_mul_pd(fscal,dy02);
 409             tz               = _mm256_mul_pd(fscal,dz02);
 410
 411             /* Update vectorial force */
 412             fix0             = _mm256_add_pd(fix0,tx);
 413             fiy0             = _mm256_add_pd(fiy0,ty);
 414             fiz0             = _mm256_add_pd(fiz0,tz);
 415
 416             fjx2             = _mm256_add_pd(fjx2,tx);
 417             fjy2             = _mm256_add_pd(fjy2,ty);
 418             fjz2             = _mm256_add_pd(fjz2,tz);
 419
 420             /**************************
 421              * CALCULATE INTERACTIONS *
 422              **************************/
 423
 424             /* COULOMB ELECTROSTATICS */
 425             velec            = _mm256_mul_pd(qq10,rinv10);
 426             felec            = _mm256_mul_pd(velec,rinvsq10);
 427
 428             /* Update potential sum for this i atom from the interaction with this j atom. */
 429             velecsum         = _mm256_add_pd(velecsum,velec);
 430
 431             fscal            = felec;
 432
 433             /* Calculate temporary vectorial force */
 434             tx               = _mm256_mul_pd(fscal,dx10);
 435             ty               = _mm256_mul_pd(fscal,dy10);
 436             tz               = _mm256_mul_pd(fscal,dz10);
 437
 438             /* Update vectorial force */
 439             fix1             = _mm256_add_pd(fix1,tx);
 440             fiy1             = _mm256_add_pd(fiy1,ty);
 441             fiz1             = _mm256_add_pd(fiz1,tz);
 442
 443             fjx0             = _mm256_add_pd(fjx0,tx);
 444             fjy0             = _mm256_add_pd(fjy0,ty);
 445             fjz0             = _mm256_add_pd(fjz0,tz);
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             /* COULOMB ELECTROSTATICS */
 452             velec            = _mm256_mul_pd(qq11,rinv11);
 453             felec            = _mm256_mul_pd(velec,rinvsq11);
 454
 455             /* Update potential sum for this i atom from the interaction with this j atom. */
 456             velecsum         = _mm256_add_pd(velecsum,velec);
 457
 458             fscal            = felec;
 459
 460             /* Calculate temporary vectorial force */
 461             tx               = _mm256_mul_pd(fscal,dx11);
 462             ty               = _mm256_mul_pd(fscal,dy11);
 463             tz               = _mm256_mul_pd(fscal,dz11);
 464
 465             /* Update vectorial force */
 466             fix1             = _mm256_add_pd(fix1,tx);
 467             fiy1             = _mm256_add_pd(fiy1,ty);
 468             fiz1             = _mm256_add_pd(fiz1,tz);
 469
 470             fjx1             = _mm256_add_pd(fjx1,tx);
 471             fjy1             = _mm256_add_pd(fjy1,ty);
 472             fjz1             = _mm256_add_pd(fjz1,tz);
 473
 474             /**************************
 475              * CALCULATE INTERACTIONS *
 476              **************************/
 477
 478             /* COULOMB ELECTROSTATICS */
 479             velec            = _mm256_mul_pd(qq12,rinv12);
 480             felec            = _mm256_mul_pd(velec,rinvsq12);
 481
 482             /* Update potential sum for this i atom from the interaction with this j atom. */
 483             velecsum         = _mm256_add_pd(velecsum,velec);
 484
 485             fscal            = felec;
 486
 487             /* Calculate temporary vectorial force */
 488             tx               = _mm256_mul_pd(fscal,dx12);
 489             ty               = _mm256_mul_pd(fscal,dy12);
 490             tz               = _mm256_mul_pd(fscal,dz12);
 491
 492             /* Update vectorial force */
 493             fix1             = _mm256_add_pd(fix1,tx);
 494             fiy1             = _mm256_add_pd(fiy1,ty);
 495             fiz1             = _mm256_add_pd(fiz1,tz);
 496
 497             fjx2             = _mm256_add_pd(fjx2,tx);
 498             fjy2             = _mm256_add_pd(fjy2,ty);
 499             fjz2             = _mm256_add_pd(fjz2,tz);
 500
 501             /**************************
 502              * CALCULATE INTERACTIONS *
 503              **************************/
 504
 505             /* COULOMB ELECTROSTATICS */
 506             velec            = _mm256_mul_pd(qq20,rinv20);
 507             felec            = _mm256_mul_pd(velec,rinvsq20);
 508
 509             /* Update potential sum for this i atom from the interaction with this j atom. */
 510             velecsum         = _mm256_add_pd(velecsum,velec);
 511
 512             fscal            = felec;
 513
 514             /* Calculate temporary vectorial force */
 515             tx               = _mm256_mul_pd(fscal,dx20);
 516             ty               = _mm256_mul_pd(fscal,dy20);
 517             tz               = _mm256_mul_pd(fscal,dz20);
 518
 519             /* Update vectorial force */
 520             fix2             = _mm256_add_pd(fix2,tx);
 521             fiy2             = _mm256_add_pd(fiy2,ty);
 522             fiz2             = _mm256_add_pd(fiz2,tz);
 523
 524             fjx0             = _mm256_add_pd(fjx0,tx);
 525             fjy0             = _mm256_add_pd(fjy0,ty);
 526             fjz0             = _mm256_add_pd(fjz0,tz);
 527
 528             /**************************
 529              * CALCULATE INTERACTIONS *
 530              **************************/
 531
 532             /* COULOMB ELECTROSTATICS */
 533             velec            = _mm256_mul_pd(qq21,rinv21);
 534             felec            = _mm256_mul_pd(velec,rinvsq21);
 535
 536             /* Update potential sum for this i atom from the interaction with this j atom. */
 537             velecsum         = _mm256_add_pd(velecsum,velec);
 538
 539             fscal            = felec;
 540
 541             /* Calculate temporary vectorial force */
 542             tx               = _mm256_mul_pd(fscal,dx21);
 543             ty               = _mm256_mul_pd(fscal,dy21);
 544             tz               = _mm256_mul_pd(fscal,dz21);
 545
 546             /* Update vectorial force */
 547             fix2             = _mm256_add_pd(fix2,tx);
 548             fiy2             = _mm256_add_pd(fiy2,ty);
 549             fiz2             = _mm256_add_pd(fiz2,tz);
 550
 551             fjx1             = _mm256_add_pd(fjx1,tx);
 552             fjy1             = _mm256_add_pd(fjy1,ty);
 553             fjz1             = _mm256_add_pd(fjz1,tz);
 554
 555             /**************************
 556              * CALCULATE INTERACTIONS *
 557              **************************/
 558
 559             /* COULOMB ELECTROSTATICS */
 560             velec            = _mm256_mul_pd(qq22,rinv22);
 561             felec            = _mm256_mul_pd(velec,rinvsq22);
 562
 563             /* Update potential sum for this i atom from the interaction with this j atom. */
 564             velecsum         = _mm256_add_pd(velecsum,velec);
 565
 566             fscal            = felec;
 567
 568             /* Calculate temporary vectorial force */
 569             tx               = _mm256_mul_pd(fscal,dx22);
 570             ty               = _mm256_mul_pd(fscal,dy22);
 571             tz               = _mm256_mul_pd(fscal,dz22);
 572
 573             /* Update vectorial force */
 574             fix2             = _mm256_add_pd(fix2,tx);
 575             fiy2             = _mm256_add_pd(fiy2,ty);
 576             fiz2             = _mm256_add_pd(fiz2,tz);
 577
 578             fjx2             = _mm256_add_pd(fjx2,tx);
 579             fjy2             = _mm256_add_pd(fjy2,ty);
 580             fjz2             = _mm256_add_pd(fjz2,tz);
 581
 582             fjptrA             = f+j_coord_offsetA;
 583             fjptrB             = f+j_coord_offsetB;
 584             fjptrC             = f+j_coord_offsetC;
 585             fjptrD             = f+j_coord_offsetD;
 586
 587             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 588                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 589
 590             /* Inner loop uses 278 flops */
 591         }
 592
 593         if(jidx<j_index_end)
 594         {
 595
 596             /* Get j neighbor index, and coordinate index */
 597             jnrlistA         = jjnr[jidx];
 598             jnrlistB         = jjnr[jidx+1];
 599             jnrlistC         = jjnr[jidx+2];
 600             jnrlistD         = jjnr[jidx+3];
 601             /* Sign of each element will be negative for non-real atoms.
 602              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 603              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 604              */
 605             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 606
 607             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 608             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 609             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 610
 611             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 612             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 613             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 614             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 615             j_coord_offsetA  = DIM*jnrA;
 616             j_coord_offsetB  = DIM*jnrB;
 617             j_coord_offsetC  = DIM*jnrC;
 618             j_coord_offsetD  = DIM*jnrD;
 619
 620             /* load j atom coordinates */
 621             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 622                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 623                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 624
 625             /* Calculate displacement vector */
 626             dx00             = _mm256_sub_pd(ix0,jx0);
 627             dy00             = _mm256_sub_pd(iy0,jy0);
 628             dz00             = _mm256_sub_pd(iz0,jz0);
 629             dx01             = _mm256_sub_pd(ix0,jx1);
 630             dy01             = _mm256_sub_pd(iy0,jy1);
 631             dz01             = _mm256_sub_pd(iz0,jz1);
 632             dx02             = _mm256_sub_pd(ix0,jx2);
 633             dy02             = _mm256_sub_pd(iy0,jy2);
 634             dz02             = _mm256_sub_pd(iz0,jz2);
 635             dx10             = _mm256_sub_pd(ix1,jx0);
 636             dy10             = _mm256_sub_pd(iy1,jy0);
 637             dz10             = _mm256_sub_pd(iz1,jz0);
 638             dx11             = _mm256_sub_pd(ix1,jx1);
 639             dy11             = _mm256_sub_pd(iy1,jy1);
 640             dz11             = _mm256_sub_pd(iz1,jz1);
 641             dx12             = _mm256_sub_pd(ix1,jx2);
 642             dy12             = _mm256_sub_pd(iy1,jy2);
 643             dz12             = _mm256_sub_pd(iz1,jz2);
 644             dx20             = _mm256_sub_pd(ix2,jx0);
 645             dy20             = _mm256_sub_pd(iy2,jy0);
 646             dz20             = _mm256_sub_pd(iz2,jz0);
 647             dx21             = _mm256_sub_pd(ix2,jx1);
 648             dy21             = _mm256_sub_pd(iy2,jy1);
 649             dz21             = _mm256_sub_pd(iz2,jz1);
 650             dx22             = _mm256_sub_pd(ix2,jx2);
 651             dy22             = _mm256_sub_pd(iy2,jy2);
 652             dz22             = _mm256_sub_pd(iz2,jz2);
 653
 654             /* Calculate squared distance and things based on it */
 655             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 656             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 657             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 658             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 659             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 660             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 661             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 662             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 663             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 664
 665             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 666             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 667             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 668             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 669             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 670             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 671             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 672             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 673             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 674
 675             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 676             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 677             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 678             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 679             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 680             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 681             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 682             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 683             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 684
 685             fjx0             = _mm256_setzero_pd();
 686             fjy0             = _mm256_setzero_pd();
 687             fjz0             = _mm256_setzero_pd();
 688             fjx1             = _mm256_setzero_pd();
 689             fjy1             = _mm256_setzero_pd();
 690             fjz1             = _mm256_setzero_pd();
 691             fjx2             = _mm256_setzero_pd();
 692             fjy2             = _mm256_setzero_pd();
 693             fjz2             = _mm256_setzero_pd();
 694
 695             /**************************
 696              * CALCULATE INTERACTIONS *
 697              **************************/
 698
 699             r00              = _mm256_mul_pd(rsq00,rinv00);
 700             r00              = _mm256_andnot_pd(dummy_mask,r00);
 701
 702             /* Calculate table index by multiplying r with table scale and truncate to integer */
 703             rt               = _mm256_mul_pd(r00,vftabscale);
 704             vfitab           = _mm256_cvttpd_epi32(rt);
 705             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 706             vfitab           = _mm_slli_epi32(vfitab,3);
 707
 708             /* COULOMB ELECTROSTATICS */
 709             velec            = _mm256_mul_pd(qq00,rinv00);
 710             felec            = _mm256_mul_pd(velec,rinvsq00);
 711
 712             /* CUBIC SPLINE TABLE DISPERSION */
 713             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 714             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 715             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 716             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 717             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 718             Heps             = _mm256_mul_pd(vfeps,H);
 719             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 720             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 721             vvdw6            = _mm256_mul_pd(c6_00,VV);
 722             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 723             fvdw6            = _mm256_mul_pd(c6_00,FF);
 724
 725             /* CUBIC SPLINE TABLE REPULSION */
 726             vfitab           = _mm_add_epi32(vfitab,ifour);
 727             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 728             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 729             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 730             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 731             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 732             Heps             = _mm256_mul_pd(vfeps,H);
 733             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 734             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 735             vvdw12           = _mm256_mul_pd(c12_00,VV);
 736             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 737             fvdw12           = _mm256_mul_pd(c12_00,FF);
 738             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 739             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 740
 741             /* Update potential sum for this i atom from the interaction with this j atom. */
 742             velec            = _mm256_andnot_pd(dummy_mask,velec);
 743             velecsum         = _mm256_add_pd(velecsum,velec);
 744             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 745             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 746
 747             fscal            = _mm256_add_pd(felec,fvdw);
 748
 749             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 750
 751             /* Calculate temporary vectorial force */
 752             tx               = _mm256_mul_pd(fscal,dx00);
 753             ty               = _mm256_mul_pd(fscal,dy00);
 754             tz               = _mm256_mul_pd(fscal,dz00);
 755
 756             /* Update vectorial force */
 757             fix0             = _mm256_add_pd(fix0,tx);
 758             fiy0             = _mm256_add_pd(fiy0,ty);
 759             fiz0             = _mm256_add_pd(fiz0,tz);
 760
 761             fjx0             = _mm256_add_pd(fjx0,tx);
 762             fjy0             = _mm256_add_pd(fjy0,ty);
 763             fjz0             = _mm256_add_pd(fjz0,tz);
 764
 765             /**************************
 766              * CALCULATE INTERACTIONS *
 767              **************************/
 768
 769             /* COULOMB ELECTROSTATICS */
 770             velec            = _mm256_mul_pd(qq01,rinv01);
 771             felec            = _mm256_mul_pd(velec,rinvsq01);
 772
 773             /* Update potential sum for this i atom from the interaction with this j atom. */
 774             velec            = _mm256_andnot_pd(dummy_mask,velec);
 775             velecsum         = _mm256_add_pd(velecsum,velec);
 776
 777             fscal            = felec;
 778
 779             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 780
 781             /* Calculate temporary vectorial force */
 782             tx               = _mm256_mul_pd(fscal,dx01);
 783             ty               = _mm256_mul_pd(fscal,dy01);
 784             tz               = _mm256_mul_pd(fscal,dz01);
 785
 786             /* Update vectorial force */
 787             fix0             = _mm256_add_pd(fix0,tx);
 788             fiy0             = _mm256_add_pd(fiy0,ty);
 789             fiz0             = _mm256_add_pd(fiz0,tz);
 790
 791             fjx1             = _mm256_add_pd(fjx1,tx);
 792             fjy1             = _mm256_add_pd(fjy1,ty);
 793             fjz1             = _mm256_add_pd(fjz1,tz);
 794
 795             /**************************
 796              * CALCULATE INTERACTIONS *
 797              **************************/
 798
 799             /* COULOMB ELECTROSTATICS */
 800             velec            = _mm256_mul_pd(qq02,rinv02);
 801             felec            = _mm256_mul_pd(velec,rinvsq02);
 802
 803             /* Update potential sum for this i atom from the interaction with this j atom. */
 804             velec            = _mm256_andnot_pd(dummy_mask,velec);
 805             velecsum         = _mm256_add_pd(velecsum,velec);
 806
 807             fscal            = felec;
 808
 809             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 810
 811             /* Calculate temporary vectorial force */
 812             tx               = _mm256_mul_pd(fscal,dx02);
 813             ty               = _mm256_mul_pd(fscal,dy02);
 814             tz               = _mm256_mul_pd(fscal,dz02);
 815
 816             /* Update vectorial force */
 817             fix0             = _mm256_add_pd(fix0,tx);
 818             fiy0             = _mm256_add_pd(fiy0,ty);
 819             fiz0             = _mm256_add_pd(fiz0,tz);
 820
 821             fjx2             = _mm256_add_pd(fjx2,tx);
 822             fjy2             = _mm256_add_pd(fjy2,ty);
 823             fjz2             = _mm256_add_pd(fjz2,tz);
 824
 825             /**************************
 826              * CALCULATE INTERACTIONS *
 827              **************************/
 828
 829             /* COULOMB ELECTROSTATICS */
 830             velec            = _mm256_mul_pd(qq10,rinv10);
 831             felec            = _mm256_mul_pd(velec,rinvsq10);
 832
 833             /* Update potential sum for this i atom from the interaction with this j atom. */
 834             velec            = _mm256_andnot_pd(dummy_mask,velec);
 835             velecsum         = _mm256_add_pd(velecsum,velec);
 836
 837             fscal            = felec;
 838
 839             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 840
 841             /* Calculate temporary vectorial force */
 842             tx               = _mm256_mul_pd(fscal,dx10);
 843             ty               = _mm256_mul_pd(fscal,dy10);
 844             tz               = _mm256_mul_pd(fscal,dz10);
 845
 846             /* Update vectorial force */
 847             fix1             = _mm256_add_pd(fix1,tx);
 848             fiy1             = _mm256_add_pd(fiy1,ty);
 849             fiz1             = _mm256_add_pd(fiz1,tz);
 850
 851             fjx0             = _mm256_add_pd(fjx0,tx);
 852             fjy0             = _mm256_add_pd(fjy0,ty);
 853             fjz0             = _mm256_add_pd(fjz0,tz);
 854
 855             /**************************
 856              * CALCULATE INTERACTIONS *
 857              **************************/
 858
 859             /* COULOMB ELECTROSTATICS */
 860             velec            = _mm256_mul_pd(qq11,rinv11);
 861             felec            = _mm256_mul_pd(velec,rinvsq11);
 862
 863             /* Update potential sum for this i atom from the interaction with this j atom. */
 864             velec            = _mm256_andnot_pd(dummy_mask,velec);
 865             velecsum         = _mm256_add_pd(velecsum,velec);
 866
 867             fscal            = felec;
 868
 869             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 870
 871             /* Calculate temporary vectorial force */
 872             tx               = _mm256_mul_pd(fscal,dx11);
 873             ty               = _mm256_mul_pd(fscal,dy11);
 874             tz               = _mm256_mul_pd(fscal,dz11);
 875
 876             /* Update vectorial force */
 877             fix1             = _mm256_add_pd(fix1,tx);
 878             fiy1             = _mm256_add_pd(fiy1,ty);
 879             fiz1             = _mm256_add_pd(fiz1,tz);
 880
 881             fjx1             = _mm256_add_pd(fjx1,tx);
 882             fjy1             = _mm256_add_pd(fjy1,ty);
 883             fjz1             = _mm256_add_pd(fjz1,tz);
 884
 885             /**************************
 886              * CALCULATE INTERACTIONS *
 887              **************************/
 888
 889             /* COULOMB ELECTROSTATICS */
 890             velec            = _mm256_mul_pd(qq12,rinv12);
 891             felec            = _mm256_mul_pd(velec,rinvsq12);
 892
 893             /* Update potential sum for this i atom from the interaction with this j atom. */
 894             velec            = _mm256_andnot_pd(dummy_mask,velec);
 895             velecsum         = _mm256_add_pd(velecsum,velec);
 896
 897             fscal            = felec;
 898
 899             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 900
 901             /* Calculate temporary vectorial force */
 902             tx               = _mm256_mul_pd(fscal,dx12);
 903             ty               = _mm256_mul_pd(fscal,dy12);
 904             tz               = _mm256_mul_pd(fscal,dz12);
 905
 906             /* Update vectorial force */
 907             fix1             = _mm256_add_pd(fix1,tx);
 908             fiy1             = _mm256_add_pd(fiy1,ty);
 909             fiz1             = _mm256_add_pd(fiz1,tz);
 910
 911             fjx2             = _mm256_add_pd(fjx2,tx);
 912             fjy2             = _mm256_add_pd(fjy2,ty);
 913             fjz2             = _mm256_add_pd(fjz2,tz);
 914
 915             /**************************
 916              * CALCULATE INTERACTIONS *
 917              **************************/
 918
 919             /* COULOMB ELECTROSTATICS */
 920             velec            = _mm256_mul_pd(qq20,rinv20);
 921             felec            = _mm256_mul_pd(velec,rinvsq20);
 922
 923             /* Update potential sum for this i atom from the interaction with this j atom. */
 924             velec            = _mm256_andnot_pd(dummy_mask,velec);
 925             velecsum         = _mm256_add_pd(velecsum,velec);
 926
 927             fscal            = felec;
 928
 929             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 930
 931             /* Calculate temporary vectorial force */
 932             tx               = _mm256_mul_pd(fscal,dx20);
 933             ty               = _mm256_mul_pd(fscal,dy20);
 934             tz               = _mm256_mul_pd(fscal,dz20);
 935
 936             /* Update vectorial force */
 937             fix2             = _mm256_add_pd(fix2,tx);
 938             fiy2             = _mm256_add_pd(fiy2,ty);
 939             fiz2             = _mm256_add_pd(fiz2,tz);
 940
 941             fjx0             = _mm256_add_pd(fjx0,tx);
 942             fjy0             = _mm256_add_pd(fjy0,ty);
 943             fjz0             = _mm256_add_pd(fjz0,tz);
 944
 945             /**************************
 946              * CALCULATE INTERACTIONS *
 947              **************************/
 948
 949             /* COULOMB ELECTROSTATICS */
 950             velec            = _mm256_mul_pd(qq21,rinv21);
 951             felec            = _mm256_mul_pd(velec,rinvsq21);
 952
 953             /* Update potential sum for this i atom from the interaction with this j atom. */
 954             velec            = _mm256_andnot_pd(dummy_mask,velec);
 955             velecsum         = _mm256_add_pd(velecsum,velec);
 956
 957             fscal            = felec;
 958
 959             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 960
 961             /* Calculate temporary vectorial force */
 962             tx               = _mm256_mul_pd(fscal,dx21);
 963             ty               = _mm256_mul_pd(fscal,dy21);
 964             tz               = _mm256_mul_pd(fscal,dz21);
 965
 966             /* Update vectorial force */
 967             fix2             = _mm256_add_pd(fix2,tx);
 968             fiy2             = _mm256_add_pd(fiy2,ty);
 969             fiz2             = _mm256_add_pd(fiz2,tz);
 970
 971             fjx1             = _mm256_add_pd(fjx1,tx);
 972             fjy1             = _mm256_add_pd(fjy1,ty);
 973             fjz1             = _mm256_add_pd(fjz1,tz);
 974
 975             /**************************
 976              * CALCULATE INTERACTIONS *
 977              **************************/
 978
 979             /* COULOMB ELECTROSTATICS */
 980             velec            = _mm256_mul_pd(qq22,rinv22);
 981             felec            = _mm256_mul_pd(velec,rinvsq22);
 982
 983             /* Update potential sum for this i atom from the interaction with this j atom. */
 984             velec            = _mm256_andnot_pd(dummy_mask,velec);
 985             velecsum         = _mm256_add_pd(velecsum,velec);
 986
 987             fscal            = felec;
 988
 989             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 990
 991             /* Calculate temporary vectorial force */
 992             tx               = _mm256_mul_pd(fscal,dx22);
 993             ty               = _mm256_mul_pd(fscal,dy22);
 994             tz               = _mm256_mul_pd(fscal,dz22);
 995
 996             /* Update vectorial force */
 997             fix2             = _mm256_add_pd(fix2,tx);
 998             fiy2             = _mm256_add_pd(fiy2,ty);
 999             fiz2             = _mm256_add_pd(fiz2,tz);
1000
1001             fjx2             = _mm256_add_pd(fjx2,tx);
1002             fjy2             = _mm256_add_pd(fjy2,ty);
1003             fjz2             = _mm256_add_pd(fjz2,tz);
1004
1005             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1006             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1007             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1008             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1009
1010             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1011                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1012
1013             /* Inner loop uses 279 flops */
1014         }
1015
1016         /* End of innermost loop */
1017
1018         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1019                                                  f+i_coord_offset,fshift+i_shift_offset);
1020
1021         ggid                        = gid[iidx];
1022         /* Update potential energies */
1023         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1024         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1025
1026         /* Increment number of inner iterations */
1027         inneriter                  += j_index_end - j_index_start;
1028
1029         /* Outer loop uses 20 flops */
1030     }
1031
1032     /* Increment number of outer iterations */
1033     outeriter        += nri;
1034
1035     /* Update outer/inner flops */
1036
1037     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*279);
1038 }
1039 /*
1040  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_F_avx_256_double
1041  * Electrostatics interaction: Coulomb
1042  * VdW interaction:            CubicSplineTable
1043  * Geometry:                   Water3-Water3
1044  * Calculate force/pot:        Force
1045  */
1046 void
1047 nb_kernel_ElecCoul_VdwCSTab_GeomW3W3_F_avx_256_double
1048                     (t_nblist                    * gmx_restrict       nlist,
1049                      rvec                        * gmx_restrict          xx,
1050                      rvec                        * gmx_restrict          ff,
1051                      t_forcerec                  * gmx_restrict          fr,
1052                      t_mdatoms                   * gmx_restrict     mdatoms,
1053                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1054                      t_nrnb                      * gmx_restrict        nrnb)
1055 {
1056     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1057      * just 0 for non-waters.
1058      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1059      * jnr indices corresponding to data put in the four positions in the SIMD register.
1060      */
1061     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1062     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1063     int              jnrA,jnrB,jnrC,jnrD;
1064     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1065     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1066     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1067     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1068     real             rcutoff_scalar;
1069     real             *shiftvec,*fshift,*x,*f;
1070     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1071     real             scratch[4*DIM];
1072     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1073     real *           vdwioffsetptr0;
1074     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1075     real *           vdwioffsetptr1;
1076     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1077     real *           vdwioffsetptr2;
1078     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1079     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1080     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1081     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1082     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1083     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1084     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1085     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1086     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1087     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1088     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1089     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1090     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1091     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1092     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1093     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1094     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1095     real             *charge;
1096     int              nvdwtype;
1097     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1098     int              *vdwtype;
1099     real             *vdwparam;
1100     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1101     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1102     __m128i          vfitab;
1103     __m128i          ifour       = _mm_set1_epi32(4);
1104     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1105     real             *vftab;
1106     __m256d          dummy_mask,cutoff_mask;
1107     __m128           tmpmask0,tmpmask1;
1108     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1109     __m256d          one     = _mm256_set1_pd(1.0);
1110     __m256d          two     = _mm256_set1_pd(2.0);
1111     x                = xx[0];
1112     f                = ff[0];
1113
1114     nri              = nlist->nri;
1115     iinr             = nlist->iinr;
1116     jindex           = nlist->jindex;
1117     jjnr             = nlist->jjnr;
1118     shiftidx         = nlist->shift;
1119     gid              = nlist->gid;
1120     shiftvec         = fr->shift_vec[0];
1121     fshift           = fr->fshift[0];
1122     facel            = _mm256_set1_pd(fr->epsfac);
1123     charge           = mdatoms->chargeA;
1124     nvdwtype         = fr->ntype;
1125     vdwparam         = fr->nbfp;
1126     vdwtype          = mdatoms->typeA;
1127
1128     vftab            = kernel_data->table_vdw->data;
1129     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
1130
1131     /* Setup water-specific parameters */
1132     inr              = nlist->iinr[0];
1133     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1134     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1135     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1136     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1137
1138     jq0              = _mm256_set1_pd(charge[inr+0]);
1139     jq1              = _mm256_set1_pd(charge[inr+1]);
1140     jq2              = _mm256_set1_pd(charge[inr+2]);
1141     vdwjidx0A        = 2*vdwtype[inr+0];
1142     qq00             = _mm256_mul_pd(iq0,jq0);
1143     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1144     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1145     qq01             = _mm256_mul_pd(iq0,jq1);
1146     qq02             = _mm256_mul_pd(iq0,jq2);
1147     qq10             = _mm256_mul_pd(iq1,jq0);
1148     qq11             = _mm256_mul_pd(iq1,jq1);
1149     qq12             = _mm256_mul_pd(iq1,jq2);
1150     qq20             = _mm256_mul_pd(iq2,jq0);
1151     qq21             = _mm256_mul_pd(iq2,jq1);
1152     qq22             = _mm256_mul_pd(iq2,jq2);
1153
1154     /* Avoid stupid compiler warnings */
1155     jnrA = jnrB = jnrC = jnrD = 0;
1156     j_coord_offsetA = 0;
1157     j_coord_offsetB = 0;
1158     j_coord_offsetC = 0;
1159     j_coord_offsetD = 0;
1160
1161     outeriter        = 0;
1162     inneriter        = 0;
1163
1164     for(iidx=0;iidx<4*DIM;iidx++)
1165     {
1166         scratch[iidx] = 0.0;
1167     }
1168
1169     /* Start outer loop over neighborlists */
1170     for(iidx=0; iidx<nri; iidx++)
1171     {
1172         /* Load shift vector for this list */
1173         i_shift_offset   = DIM*shiftidx[iidx];
1174
1175         /* Load limits for loop over neighbors */
1176         j_index_start    = jindex[iidx];
1177         j_index_end      = jindex[iidx+1];
1178
1179         /* Get outer coordinate index */
1180         inr              = iinr[iidx];
1181         i_coord_offset   = DIM*inr;
1182
1183         /* Load i particle coords and add shift vector */
1184         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1185                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1186
1187         fix0             = _mm256_setzero_pd();
1188         fiy0             = _mm256_setzero_pd();
1189         fiz0             = _mm256_setzero_pd();
1190         fix1             = _mm256_setzero_pd();
1191         fiy1             = _mm256_setzero_pd();
1192         fiz1             = _mm256_setzero_pd();
1193         fix2             = _mm256_setzero_pd();
1194         fiy2             = _mm256_setzero_pd();
1195         fiz2             = _mm256_setzero_pd();
1196
1197         /* Start inner kernel loop */
1198         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1199         {
1200
1201             /* Get j neighbor index, and coordinate index */
1202             jnrA             = jjnr[jidx];
1203             jnrB             = jjnr[jidx+1];
1204             jnrC             = jjnr[jidx+2];
1205             jnrD             = jjnr[jidx+3];
1206             j_coord_offsetA  = DIM*jnrA;
1207             j_coord_offsetB  = DIM*jnrB;
1208             j_coord_offsetC  = DIM*jnrC;
1209             j_coord_offsetD  = DIM*jnrD;
1210
1211             /* load j atom coordinates */
1212             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1213                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1214                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1215
1216             /* Calculate displacement vector */
1217             dx00             = _mm256_sub_pd(ix0,jx0);
1218             dy00             = _mm256_sub_pd(iy0,jy0);
1219             dz00             = _mm256_sub_pd(iz0,jz0);
1220             dx01             = _mm256_sub_pd(ix0,jx1);
1221             dy01             = _mm256_sub_pd(iy0,jy1);
1222             dz01             = _mm256_sub_pd(iz0,jz1);
1223             dx02             = _mm256_sub_pd(ix0,jx2);
1224             dy02             = _mm256_sub_pd(iy0,jy2);
1225             dz02             = _mm256_sub_pd(iz0,jz2);
1226             dx10             = _mm256_sub_pd(ix1,jx0);
1227             dy10             = _mm256_sub_pd(iy1,jy0);
1228             dz10             = _mm256_sub_pd(iz1,jz0);
1229             dx11             = _mm256_sub_pd(ix1,jx1);
1230             dy11             = _mm256_sub_pd(iy1,jy1);
1231             dz11             = _mm256_sub_pd(iz1,jz1);
1232             dx12             = _mm256_sub_pd(ix1,jx2);
1233             dy12             = _mm256_sub_pd(iy1,jy2);
1234             dz12             = _mm256_sub_pd(iz1,jz2);
1235             dx20             = _mm256_sub_pd(ix2,jx0);
1236             dy20             = _mm256_sub_pd(iy2,jy0);
1237             dz20             = _mm256_sub_pd(iz2,jz0);
1238             dx21             = _mm256_sub_pd(ix2,jx1);
1239             dy21             = _mm256_sub_pd(iy2,jy1);
1240             dz21             = _mm256_sub_pd(iz2,jz1);
1241             dx22             = _mm256_sub_pd(ix2,jx2);
1242             dy22             = _mm256_sub_pd(iy2,jy2);
1243             dz22             = _mm256_sub_pd(iz2,jz2);
1244
1245             /* Calculate squared distance and things based on it */
1246             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1247             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1248             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1249             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1250             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1251             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1252             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1253             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1254             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1255
1256             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1257             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1258             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1259             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1260             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1261             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1262             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1263             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1264             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1265
1266             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1267             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1268             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1269             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1270             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1271             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1272             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1273             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1274             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1275
1276             fjx0             = _mm256_setzero_pd();
1277             fjy0             = _mm256_setzero_pd();
1278             fjz0             = _mm256_setzero_pd();
1279             fjx1             = _mm256_setzero_pd();
1280             fjy1             = _mm256_setzero_pd();
1281             fjz1             = _mm256_setzero_pd();
1282             fjx2             = _mm256_setzero_pd();
1283             fjy2             = _mm256_setzero_pd();
1284             fjz2             = _mm256_setzero_pd();
1285
1286             /**************************
1287              * CALCULATE INTERACTIONS *
1288              **************************/
1289
1290             r00              = _mm256_mul_pd(rsq00,rinv00);
1291
1292             /* Calculate table index by multiplying r with table scale and truncate to integer */
1293             rt               = _mm256_mul_pd(r00,vftabscale);
1294             vfitab           = _mm256_cvttpd_epi32(rt);
1295             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1296             vfitab           = _mm_slli_epi32(vfitab,3);
1297
1298             /* COULOMB ELECTROSTATICS */
1299             velec            = _mm256_mul_pd(qq00,rinv00);
1300             felec            = _mm256_mul_pd(velec,rinvsq00);
1301
1302             /* CUBIC SPLINE TABLE DISPERSION */
1303             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1304             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1305             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1306             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1307             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1308             Heps             = _mm256_mul_pd(vfeps,H);
1309             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1310             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1311             fvdw6            = _mm256_mul_pd(c6_00,FF);
1312
1313             /* CUBIC SPLINE TABLE REPULSION */
1314             vfitab           = _mm_add_epi32(vfitab,ifour);
1315             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1316             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1317             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1318             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1319             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1320             Heps             = _mm256_mul_pd(vfeps,H);
1321             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1322             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1323             fvdw12           = _mm256_mul_pd(c12_00,FF);
1324             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1325
1326             fscal            = _mm256_add_pd(felec,fvdw);
1327
1328             /* Calculate temporary vectorial force */
1329             tx               = _mm256_mul_pd(fscal,dx00);
1330             ty               = _mm256_mul_pd(fscal,dy00);
1331             tz               = _mm256_mul_pd(fscal,dz00);
1332
1333             /* Update vectorial force */
1334             fix0             = _mm256_add_pd(fix0,tx);
1335             fiy0             = _mm256_add_pd(fiy0,ty);
1336             fiz0             = _mm256_add_pd(fiz0,tz);
1337
1338             fjx0             = _mm256_add_pd(fjx0,tx);
1339             fjy0             = _mm256_add_pd(fjy0,ty);
1340             fjz0             = _mm256_add_pd(fjz0,tz);
1341
1342             /**************************
1343              * CALCULATE INTERACTIONS *
1344              **************************/
1345
1346             /* COULOMB ELECTROSTATICS */
1347             velec            = _mm256_mul_pd(qq01,rinv01);
1348             felec            = _mm256_mul_pd(velec,rinvsq01);
1349
1350             fscal            = felec;
1351
1352             /* Calculate temporary vectorial force */
1353             tx               = _mm256_mul_pd(fscal,dx01);
1354             ty               = _mm256_mul_pd(fscal,dy01);
1355             tz               = _mm256_mul_pd(fscal,dz01);
1356
1357             /* Update vectorial force */
1358             fix0             = _mm256_add_pd(fix0,tx);
1359             fiy0             = _mm256_add_pd(fiy0,ty);
1360             fiz0             = _mm256_add_pd(fiz0,tz);
1361
1362             fjx1             = _mm256_add_pd(fjx1,tx);
1363             fjy1             = _mm256_add_pd(fjy1,ty);
1364             fjz1             = _mm256_add_pd(fjz1,tz);
1365
1366             /**************************
1367              * CALCULATE INTERACTIONS *
1368              **************************/
1369
1370             /* COULOMB ELECTROSTATICS */
1371             velec            = _mm256_mul_pd(qq02,rinv02);
1372             felec            = _mm256_mul_pd(velec,rinvsq02);
1373
1374             fscal            = felec;
1375
1376             /* Calculate temporary vectorial force */
1377             tx               = _mm256_mul_pd(fscal,dx02);
1378             ty               = _mm256_mul_pd(fscal,dy02);
1379             tz               = _mm256_mul_pd(fscal,dz02);
1380
1381             /* Update vectorial force */
1382             fix0             = _mm256_add_pd(fix0,tx);
1383             fiy0             = _mm256_add_pd(fiy0,ty);
1384             fiz0             = _mm256_add_pd(fiz0,tz);
1385
1386             fjx2             = _mm256_add_pd(fjx2,tx);
1387             fjy2             = _mm256_add_pd(fjy2,ty);
1388             fjz2             = _mm256_add_pd(fjz2,tz);
1389
1390             /**************************
1391              * CALCULATE INTERACTIONS *
1392              **************************/
1393
1394             /* COULOMB ELECTROSTATICS */
1395             velec            = _mm256_mul_pd(qq10,rinv10);
1396             felec            = _mm256_mul_pd(velec,rinvsq10);
1397
1398             fscal            = felec;
1399
1400             /* Calculate temporary vectorial force */
1401             tx               = _mm256_mul_pd(fscal,dx10);
1402             ty               = _mm256_mul_pd(fscal,dy10);
1403             tz               = _mm256_mul_pd(fscal,dz10);
1404
1405             /* Update vectorial force */
1406             fix1             = _mm256_add_pd(fix1,tx);
1407             fiy1             = _mm256_add_pd(fiy1,ty);
1408             fiz1             = _mm256_add_pd(fiz1,tz);
1409
1410             fjx0             = _mm256_add_pd(fjx0,tx);
1411             fjy0             = _mm256_add_pd(fjy0,ty);
1412             fjz0             = _mm256_add_pd(fjz0,tz);
1413
1414             /**************************
1415              * CALCULATE INTERACTIONS *
1416              **************************/
1417
1418             /* COULOMB ELECTROSTATICS */
1419             velec            = _mm256_mul_pd(qq11,rinv11);
1420             felec            = _mm256_mul_pd(velec,rinvsq11);
1421
1422             fscal            = felec;
1423
1424             /* Calculate temporary vectorial force */
1425             tx               = _mm256_mul_pd(fscal,dx11);
1426             ty               = _mm256_mul_pd(fscal,dy11);
1427             tz               = _mm256_mul_pd(fscal,dz11);
1428
1429             /* Update vectorial force */
1430             fix1             = _mm256_add_pd(fix1,tx);
1431             fiy1             = _mm256_add_pd(fiy1,ty);
1432             fiz1             = _mm256_add_pd(fiz1,tz);
1433
1434             fjx1             = _mm256_add_pd(fjx1,tx);
1435             fjy1             = _mm256_add_pd(fjy1,ty);
1436             fjz1             = _mm256_add_pd(fjz1,tz);
1437
1438             /**************************
1439              * CALCULATE INTERACTIONS *
1440              **************************/
1441
1442             /* COULOMB ELECTROSTATICS */
1443             velec            = _mm256_mul_pd(qq12,rinv12);
1444             felec            = _mm256_mul_pd(velec,rinvsq12);
1445
1446             fscal            = felec;
1447
1448             /* Calculate temporary vectorial force */
1449             tx               = _mm256_mul_pd(fscal,dx12);
1450             ty               = _mm256_mul_pd(fscal,dy12);
1451             tz               = _mm256_mul_pd(fscal,dz12);
1452
1453             /* Update vectorial force */
1454             fix1             = _mm256_add_pd(fix1,tx);
1455             fiy1             = _mm256_add_pd(fiy1,ty);
1456             fiz1             = _mm256_add_pd(fiz1,tz);
1457
1458             fjx2             = _mm256_add_pd(fjx2,tx);
1459             fjy2             = _mm256_add_pd(fjy2,ty);
1460             fjz2             = _mm256_add_pd(fjz2,tz);
1461
1462             /**************************
1463              * CALCULATE INTERACTIONS *
1464              **************************/
1465
1466             /* COULOMB ELECTROSTATICS */
1467             velec            = _mm256_mul_pd(qq20,rinv20);
1468             felec            = _mm256_mul_pd(velec,rinvsq20);
1469
1470             fscal            = felec;
1471
1472             /* Calculate temporary vectorial force */
1473             tx               = _mm256_mul_pd(fscal,dx20);
1474             ty               = _mm256_mul_pd(fscal,dy20);
1475             tz               = _mm256_mul_pd(fscal,dz20);
1476
1477             /* Update vectorial force */
1478             fix2             = _mm256_add_pd(fix2,tx);
1479             fiy2             = _mm256_add_pd(fiy2,ty);
1480             fiz2             = _mm256_add_pd(fiz2,tz);
1481
1482             fjx0             = _mm256_add_pd(fjx0,tx);
1483             fjy0             = _mm256_add_pd(fjy0,ty);
1484             fjz0             = _mm256_add_pd(fjz0,tz);
1485
1486             /**************************
1487              * CALCULATE INTERACTIONS *
1488              **************************/
1489
1490             /* COULOMB ELECTROSTATICS */
1491             velec            = _mm256_mul_pd(qq21,rinv21);
1492             felec            = _mm256_mul_pd(velec,rinvsq21);
1493
1494             fscal            = felec;
1495
1496             /* Calculate temporary vectorial force */
1497             tx               = _mm256_mul_pd(fscal,dx21);
1498             ty               = _mm256_mul_pd(fscal,dy21);
1499             tz               = _mm256_mul_pd(fscal,dz21);
1500
1501             /* Update vectorial force */
1502             fix2             = _mm256_add_pd(fix2,tx);
1503             fiy2             = _mm256_add_pd(fiy2,ty);
1504             fiz2             = _mm256_add_pd(fiz2,tz);
1505
1506             fjx1             = _mm256_add_pd(fjx1,tx);
1507             fjy1             = _mm256_add_pd(fjy1,ty);
1508             fjz1             = _mm256_add_pd(fjz1,tz);
1509
1510             /**************************
1511              * CALCULATE INTERACTIONS *
1512              **************************/
1513
1514             /* COULOMB ELECTROSTATICS */
1515             velec            = _mm256_mul_pd(qq22,rinv22);
1516             felec            = _mm256_mul_pd(velec,rinvsq22);
1517
1518             fscal            = felec;
1519
1520             /* Calculate temporary vectorial force */
1521             tx               = _mm256_mul_pd(fscal,dx22);
1522             ty               = _mm256_mul_pd(fscal,dy22);
1523             tz               = _mm256_mul_pd(fscal,dz22);
1524
1525             /* Update vectorial force */
1526             fix2             = _mm256_add_pd(fix2,tx);
1527             fiy2             = _mm256_add_pd(fiy2,ty);
1528             fiz2             = _mm256_add_pd(fiz2,tz);
1529
1530             fjx2             = _mm256_add_pd(fjx2,tx);
1531             fjy2             = _mm256_add_pd(fjy2,ty);
1532             fjz2             = _mm256_add_pd(fjz2,tz);
1533
1534             fjptrA             = f+j_coord_offsetA;
1535             fjptrB             = f+j_coord_offsetB;
1536             fjptrC             = f+j_coord_offsetC;
1537             fjptrD             = f+j_coord_offsetD;
1538
1539             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1540                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1541
1542             /* Inner loop uses 261 flops */
1543         }
1544
1545         if(jidx<j_index_end)
1546         {
1547
1548             /* Get j neighbor index, and coordinate index */
1549             jnrlistA         = jjnr[jidx];
1550             jnrlistB         = jjnr[jidx+1];
1551             jnrlistC         = jjnr[jidx+2];
1552             jnrlistD         = jjnr[jidx+3];
1553             /* Sign of each element will be negative for non-real atoms.
1554              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1555              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1556              */
1557             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1558
1559             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1560             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1561             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1562
1563             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1564             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1565             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1566             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1567             j_coord_offsetA  = DIM*jnrA;
1568             j_coord_offsetB  = DIM*jnrB;
1569             j_coord_offsetC  = DIM*jnrC;
1570             j_coord_offsetD  = DIM*jnrD;
1571
1572             /* load j atom coordinates */
1573             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1574                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1575                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1576
1577             /* Calculate displacement vector */
1578             dx00             = _mm256_sub_pd(ix0,jx0);
1579             dy00             = _mm256_sub_pd(iy0,jy0);
1580             dz00             = _mm256_sub_pd(iz0,jz0);
1581             dx01             = _mm256_sub_pd(ix0,jx1);
1582             dy01             = _mm256_sub_pd(iy0,jy1);
1583             dz01             = _mm256_sub_pd(iz0,jz1);
1584             dx02             = _mm256_sub_pd(ix0,jx2);
1585             dy02             = _mm256_sub_pd(iy0,jy2);
1586             dz02             = _mm256_sub_pd(iz0,jz2);
1587             dx10             = _mm256_sub_pd(ix1,jx0);
1588             dy10             = _mm256_sub_pd(iy1,jy0);
1589             dz10             = _mm256_sub_pd(iz1,jz0);
1590             dx11             = _mm256_sub_pd(ix1,jx1);
1591             dy11             = _mm256_sub_pd(iy1,jy1);
1592             dz11             = _mm256_sub_pd(iz1,jz1);
1593             dx12             = _mm256_sub_pd(ix1,jx2);
1594             dy12             = _mm256_sub_pd(iy1,jy2);
1595             dz12             = _mm256_sub_pd(iz1,jz2);
1596             dx20             = _mm256_sub_pd(ix2,jx0);
1597             dy20             = _mm256_sub_pd(iy2,jy0);
1598             dz20             = _mm256_sub_pd(iz2,jz0);
1599             dx21             = _mm256_sub_pd(ix2,jx1);
1600             dy21             = _mm256_sub_pd(iy2,jy1);
1601             dz21             = _mm256_sub_pd(iz2,jz1);
1602             dx22             = _mm256_sub_pd(ix2,jx2);
1603             dy22             = _mm256_sub_pd(iy2,jy2);
1604             dz22             = _mm256_sub_pd(iz2,jz2);
1605
1606             /* Calculate squared distance and things based on it */
1607             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1608             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1609             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1610             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1611             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1612             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1613             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1614             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1615             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1616
1617             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1618             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1619             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1620             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1621             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1622             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1623             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1624             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1625             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1626
1627             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1628             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1629             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1630             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1631             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1632             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1633             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1634             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1635             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1636
1637             fjx0             = _mm256_setzero_pd();
1638             fjy0             = _mm256_setzero_pd();
1639             fjz0             = _mm256_setzero_pd();
1640             fjx1             = _mm256_setzero_pd();
1641             fjy1             = _mm256_setzero_pd();
1642             fjz1             = _mm256_setzero_pd();
1643             fjx2             = _mm256_setzero_pd();
1644             fjy2             = _mm256_setzero_pd();
1645             fjz2             = _mm256_setzero_pd();
1646
1647             /**************************
1648              * CALCULATE INTERACTIONS *
1649              **************************/
1650
1651             r00              = _mm256_mul_pd(rsq00,rinv00);
1652             r00              = _mm256_andnot_pd(dummy_mask,r00);
1653
1654             /* Calculate table index by multiplying r with table scale and truncate to integer */
1655             rt               = _mm256_mul_pd(r00,vftabscale);
1656             vfitab           = _mm256_cvttpd_epi32(rt);
1657             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1658             vfitab           = _mm_slli_epi32(vfitab,3);
1659
1660             /* COULOMB ELECTROSTATICS */
1661             velec            = _mm256_mul_pd(qq00,rinv00);
1662             felec            = _mm256_mul_pd(velec,rinvsq00);
1663
1664             /* CUBIC SPLINE TABLE DISPERSION */
1665             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1666             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1667             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1668             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1669             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1670             Heps             = _mm256_mul_pd(vfeps,H);
1671             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1672             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1673             fvdw6            = _mm256_mul_pd(c6_00,FF);
1674
1675             /* CUBIC SPLINE TABLE REPULSION */
1676             vfitab           = _mm_add_epi32(vfitab,ifour);
1677             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1678             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1679             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1680             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1681             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1682             Heps             = _mm256_mul_pd(vfeps,H);
1683             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1684             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1685             fvdw12           = _mm256_mul_pd(c12_00,FF);
1686             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1687
1688             fscal            = _mm256_add_pd(felec,fvdw);
1689
1690             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1691
1692             /* Calculate temporary vectorial force */
1693             tx               = _mm256_mul_pd(fscal,dx00);
1694             ty               = _mm256_mul_pd(fscal,dy00);
1695             tz               = _mm256_mul_pd(fscal,dz00);
1696
1697             /* Update vectorial force */
1698             fix0             = _mm256_add_pd(fix0,tx);
1699             fiy0             = _mm256_add_pd(fiy0,ty);
1700             fiz0             = _mm256_add_pd(fiz0,tz);
1701
1702             fjx0             = _mm256_add_pd(fjx0,tx);
1703             fjy0             = _mm256_add_pd(fjy0,ty);
1704             fjz0             = _mm256_add_pd(fjz0,tz);
1705
1706             /**************************
1707              * CALCULATE INTERACTIONS *
1708              **************************/
1709
1710             /* COULOMB ELECTROSTATICS */
1711             velec            = _mm256_mul_pd(qq01,rinv01);
1712             felec            = _mm256_mul_pd(velec,rinvsq01);
1713
1714             fscal            = felec;
1715
1716             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1717
1718             /* Calculate temporary vectorial force */
1719             tx               = _mm256_mul_pd(fscal,dx01);
1720             ty               = _mm256_mul_pd(fscal,dy01);
1721             tz               = _mm256_mul_pd(fscal,dz01);
1722
1723             /* Update vectorial force */
1724             fix0             = _mm256_add_pd(fix0,tx);
1725             fiy0             = _mm256_add_pd(fiy0,ty);
1726             fiz0             = _mm256_add_pd(fiz0,tz);
1727
1728             fjx1             = _mm256_add_pd(fjx1,tx);
1729             fjy1             = _mm256_add_pd(fjy1,ty);
1730             fjz1             = _mm256_add_pd(fjz1,tz);
1731
1732             /**************************
1733              * CALCULATE INTERACTIONS *
1734              **************************/
1735
1736             /* COULOMB ELECTROSTATICS */
1737             velec            = _mm256_mul_pd(qq02,rinv02);
1738             felec            = _mm256_mul_pd(velec,rinvsq02);
1739
1740             fscal            = felec;
1741
1742             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1743
1744             /* Calculate temporary vectorial force */
1745             tx               = _mm256_mul_pd(fscal,dx02);
1746             ty               = _mm256_mul_pd(fscal,dy02);
1747             tz               = _mm256_mul_pd(fscal,dz02);
1748
1749             /* Update vectorial force */
1750             fix0             = _mm256_add_pd(fix0,tx);
1751             fiy0             = _mm256_add_pd(fiy0,ty);
1752             fiz0             = _mm256_add_pd(fiz0,tz);
1753
1754             fjx2             = _mm256_add_pd(fjx2,tx);
1755             fjy2             = _mm256_add_pd(fjy2,ty);
1756             fjz2             = _mm256_add_pd(fjz2,tz);
1757
1758             /**************************
1759              * CALCULATE INTERACTIONS *
1760              **************************/
1761
1762             /* COULOMB ELECTROSTATICS */
1763             velec            = _mm256_mul_pd(qq10,rinv10);
1764             felec            = _mm256_mul_pd(velec,rinvsq10);
1765
1766             fscal            = felec;
1767
1768             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1769
1770             /* Calculate temporary vectorial force */
1771             tx               = _mm256_mul_pd(fscal,dx10);
1772             ty               = _mm256_mul_pd(fscal,dy10);
1773             tz               = _mm256_mul_pd(fscal,dz10);
1774
1775             /* Update vectorial force */
1776             fix1             = _mm256_add_pd(fix1,tx);
1777             fiy1             = _mm256_add_pd(fiy1,ty);
1778             fiz1             = _mm256_add_pd(fiz1,tz);
1779
1780             fjx0             = _mm256_add_pd(fjx0,tx);
1781             fjy0             = _mm256_add_pd(fjy0,ty);
1782             fjz0             = _mm256_add_pd(fjz0,tz);
1783
1784             /**************************
1785              * CALCULATE INTERACTIONS *
1786              **************************/
1787
1788             /* COULOMB ELECTROSTATICS */
1789             velec            = _mm256_mul_pd(qq11,rinv11);
1790             felec            = _mm256_mul_pd(velec,rinvsq11);
1791
1792             fscal            = felec;
1793
1794             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1795
1796             /* Calculate temporary vectorial force */
1797             tx               = _mm256_mul_pd(fscal,dx11);
1798             ty               = _mm256_mul_pd(fscal,dy11);
1799             tz               = _mm256_mul_pd(fscal,dz11);
1800
1801             /* Update vectorial force */
1802             fix1             = _mm256_add_pd(fix1,tx);
1803             fiy1             = _mm256_add_pd(fiy1,ty);
1804             fiz1             = _mm256_add_pd(fiz1,tz);
1805
1806             fjx1             = _mm256_add_pd(fjx1,tx);
1807             fjy1             = _mm256_add_pd(fjy1,ty);
1808             fjz1             = _mm256_add_pd(fjz1,tz);
1809
1810             /**************************
1811              * CALCULATE INTERACTIONS *
1812              **************************/
1813
1814             /* COULOMB ELECTROSTATICS */
1815             velec            = _mm256_mul_pd(qq12,rinv12);
1816             felec            = _mm256_mul_pd(velec,rinvsq12);
1817
1818             fscal            = felec;
1819
1820             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1821
1822             /* Calculate temporary vectorial force */
1823             tx               = _mm256_mul_pd(fscal,dx12);
1824             ty               = _mm256_mul_pd(fscal,dy12);
1825             tz               = _mm256_mul_pd(fscal,dz12);
1826
1827             /* Update vectorial force */
1828             fix1             = _mm256_add_pd(fix1,tx);
1829             fiy1             = _mm256_add_pd(fiy1,ty);
1830             fiz1             = _mm256_add_pd(fiz1,tz);
1831
1832             fjx2             = _mm256_add_pd(fjx2,tx);
1833             fjy2             = _mm256_add_pd(fjy2,ty);
1834             fjz2             = _mm256_add_pd(fjz2,tz);
1835
1836             /**************************
1837              * CALCULATE INTERACTIONS *
1838              **************************/
1839
1840             /* COULOMB ELECTROSTATICS */
1841             velec            = _mm256_mul_pd(qq20,rinv20);
1842             felec            = _mm256_mul_pd(velec,rinvsq20);
1843
1844             fscal            = felec;
1845
1846             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1847
1848             /* Calculate temporary vectorial force */
1849             tx               = _mm256_mul_pd(fscal,dx20);
1850             ty               = _mm256_mul_pd(fscal,dy20);
1851             tz               = _mm256_mul_pd(fscal,dz20);
1852
1853             /* Update vectorial force */
1854             fix2             = _mm256_add_pd(fix2,tx);
1855             fiy2             = _mm256_add_pd(fiy2,ty);
1856             fiz2             = _mm256_add_pd(fiz2,tz);
1857
1858             fjx0             = _mm256_add_pd(fjx0,tx);
1859             fjy0             = _mm256_add_pd(fjy0,ty);
1860             fjz0             = _mm256_add_pd(fjz0,tz);
1861
1862             /**************************
1863              * CALCULATE INTERACTIONS *
1864              **************************/
1865
1866             /* COULOMB ELECTROSTATICS */
1867             velec            = _mm256_mul_pd(qq21,rinv21);
1868             felec            = _mm256_mul_pd(velec,rinvsq21);
1869
1870             fscal            = felec;
1871
1872             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1873
1874             /* Calculate temporary vectorial force */
1875             tx               = _mm256_mul_pd(fscal,dx21);
1876             ty               = _mm256_mul_pd(fscal,dy21);
1877             tz               = _mm256_mul_pd(fscal,dz21);
1878
1879             /* Update vectorial force */
1880             fix2             = _mm256_add_pd(fix2,tx);
1881             fiy2             = _mm256_add_pd(fiy2,ty);
1882             fiz2             = _mm256_add_pd(fiz2,tz);
1883
1884             fjx1             = _mm256_add_pd(fjx1,tx);
1885             fjy1             = _mm256_add_pd(fjy1,ty);
1886             fjz1             = _mm256_add_pd(fjz1,tz);
1887
1888             /**************************
1889              * CALCULATE INTERACTIONS *
1890              **************************/
1891
1892             /* COULOMB ELECTROSTATICS */
1893             velec            = _mm256_mul_pd(qq22,rinv22);
1894             felec            = _mm256_mul_pd(velec,rinvsq22);
1895
1896             fscal            = felec;
1897
1898             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1899
1900             /* Calculate temporary vectorial force */
1901             tx               = _mm256_mul_pd(fscal,dx22);
1902             ty               = _mm256_mul_pd(fscal,dy22);
1903             tz               = _mm256_mul_pd(fscal,dz22);
1904
1905             /* Update vectorial force */
1906             fix2             = _mm256_add_pd(fix2,tx);
1907             fiy2             = _mm256_add_pd(fiy2,ty);
1908             fiz2             = _mm256_add_pd(fiz2,tz);
1909
1910             fjx2             = _mm256_add_pd(fjx2,tx);
1911             fjy2             = _mm256_add_pd(fjy2,ty);
1912             fjz2             = _mm256_add_pd(fjz2,tz);
1913
1914             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1915             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1916             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1917             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1918
1919             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1920                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1921
1922             /* Inner loop uses 262 flops */
1923         }
1924
1925         /* End of innermost loop */
1926
1927         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1928                                                  f+i_coord_offset,fshift+i_shift_offset);
1929
1930         /* Increment number of inner iterations */
1931         inneriter                  += j_index_end - j_index_start;
1932
1933         /* Outer loop uses 18 flops */
1934     }
1935
1936     /* Increment number of outer iterations */
1937     outeriter        += nri;
1938
1939     /* Update outer/inner flops */
1940
1941     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*262);
1942 }