src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_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 "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "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_GeomW4W4_VF_avx_256_double
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_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     real *           vdwioffsetptr3;
  91     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  95     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  97     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  99     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 100     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 101     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 102     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 103     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 104     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 105     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 106     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 107     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 108     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 109     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 110     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 111     real             *charge;
 112     int              nvdwtype;
 113     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 114     int              *vdwtype;
 115     real             *vdwparam;
 116     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 117     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 118     __m128i          vfitab;
 119     __m128i          ifour       = _mm_set1_epi32(4);
 120     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 121     real             *vftab;
 122     __m256d          dummy_mask,cutoff_mask;
 123     __m128           tmpmask0,tmpmask1;
 124     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 125     __m256d          one     = _mm256_set1_pd(1.0);
 126     __m256d          two     = _mm256_set1_pd(2.0);
 127     x                = xx[0];
 128     f                = ff[0];
 129
 130     nri              = nlist->nri;
 131     iinr             = nlist->iinr;
 132     jindex           = nlist->jindex;
 133     jjnr             = nlist->jjnr;
 134     shiftidx         = nlist->shift;
 135     gid              = nlist->gid;
 136     shiftvec         = fr->shift_vec[0];
 137     fshift           = fr->fshift[0];
 138     facel            = _mm256_set1_pd(fr->epsfac);
 139     charge           = mdatoms->chargeA;
 140     nvdwtype         = fr->ntype;
 141     vdwparam         = fr->nbfp;
 142     vdwtype          = mdatoms->typeA;
 143
 144     vftab            = kernel_data->table_vdw->data;
 145     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
 146
 147     /* Setup water-specific parameters */
 148     inr              = nlist->iinr[0];
 149     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 150     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 151     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
 152     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 153
 154     jq1              = _mm256_set1_pd(charge[inr+1]);
 155     jq2              = _mm256_set1_pd(charge[inr+2]);
 156     jq3              = _mm256_set1_pd(charge[inr+3]);
 157     vdwjidx0A        = 2*vdwtype[inr+0];
 158     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 159     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 160     qq11             = _mm256_mul_pd(iq1,jq1);
 161     qq12             = _mm256_mul_pd(iq1,jq2);
 162     qq13             = _mm256_mul_pd(iq1,jq3);
 163     qq21             = _mm256_mul_pd(iq2,jq1);
 164     qq22             = _mm256_mul_pd(iq2,jq2);
 165     qq23             = _mm256_mul_pd(iq2,jq3);
 166     qq31             = _mm256_mul_pd(iq3,jq1);
 167     qq32             = _mm256_mul_pd(iq3,jq2);
 168     qq33             = _mm256_mul_pd(iq3,jq3);
 169
 170     /* Avoid stupid compiler warnings */
 171     jnrA = jnrB = jnrC = jnrD = 0;
 172     j_coord_offsetA = 0;
 173     j_coord_offsetB = 0;
 174     j_coord_offsetC = 0;
 175     j_coord_offsetD = 0;
 176
 177     outeriter        = 0;
 178     inneriter        = 0;
 179
 180     for(iidx=0;iidx<4*DIM;iidx++)
 181     {
 182         scratch[iidx] = 0.0;
 183     }
 184
 185     /* Start outer loop over neighborlists */
 186     for(iidx=0; iidx<nri; iidx++)
 187     {
 188         /* Load shift vector for this list */
 189         i_shift_offset   = DIM*shiftidx[iidx];
 190
 191         /* Load limits for loop over neighbors */
 192         j_index_start    = jindex[iidx];
 193         j_index_end      = jindex[iidx+1];
 194
 195         /* Get outer coordinate index */
 196         inr              = iinr[iidx];
 197         i_coord_offset   = DIM*inr;
 198
 199         /* Load i particle coords and add shift vector */
 200         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 201                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 202
 203         fix0             = _mm256_setzero_pd();
 204         fiy0             = _mm256_setzero_pd();
 205         fiz0             = _mm256_setzero_pd();
 206         fix1             = _mm256_setzero_pd();
 207         fiy1             = _mm256_setzero_pd();
 208         fiz1             = _mm256_setzero_pd();
 209         fix2             = _mm256_setzero_pd();
 210         fiy2             = _mm256_setzero_pd();
 211         fiz2             = _mm256_setzero_pd();
 212         fix3             = _mm256_setzero_pd();
 213         fiy3             = _mm256_setzero_pd();
 214         fiz3             = _mm256_setzero_pd();
 215
 216         /* Reset potential sums */
 217         velecsum         = _mm256_setzero_pd();
 218         vvdwsum          = _mm256_setzero_pd();
 219
 220         /* Start inner kernel loop */
 221         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 222         {
 223
 224             /* Get j neighbor index, and coordinate index */
 225             jnrA             = jjnr[jidx];
 226             jnrB             = jjnr[jidx+1];
 227             jnrC             = jjnr[jidx+2];
 228             jnrD             = jjnr[jidx+3];
 229             j_coord_offsetA  = DIM*jnrA;
 230             j_coord_offsetB  = DIM*jnrB;
 231             j_coord_offsetC  = DIM*jnrC;
 232             j_coord_offsetD  = DIM*jnrD;
 233
 234             /* load j atom coordinates */
 235             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 236                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 237                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 238                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 239
 240             /* Calculate displacement vector */
 241             dx00             = _mm256_sub_pd(ix0,jx0);
 242             dy00             = _mm256_sub_pd(iy0,jy0);
 243             dz00             = _mm256_sub_pd(iz0,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             dx13             = _mm256_sub_pd(ix1,jx3);
 251             dy13             = _mm256_sub_pd(iy1,jy3);
 252             dz13             = _mm256_sub_pd(iz1,jz3);
 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             dx23             = _mm256_sub_pd(ix2,jx3);
 260             dy23             = _mm256_sub_pd(iy2,jy3);
 261             dz23             = _mm256_sub_pd(iz2,jz3);
 262             dx31             = _mm256_sub_pd(ix3,jx1);
 263             dy31             = _mm256_sub_pd(iy3,jy1);
 264             dz31             = _mm256_sub_pd(iz3,jz1);
 265             dx32             = _mm256_sub_pd(ix3,jx2);
 266             dy32             = _mm256_sub_pd(iy3,jy2);
 267             dz32             = _mm256_sub_pd(iz3,jz2);
 268             dx33             = _mm256_sub_pd(ix3,jx3);
 269             dy33             = _mm256_sub_pd(iy3,jy3);
 270             dz33             = _mm256_sub_pd(iz3,jz3);
 271
 272             /* Calculate squared distance and things based on it */
 273             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 274             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 275             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 276             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 277             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 278             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 279             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 280             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 281             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 282             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 283
 284             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 285             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 286             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 287             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 288             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 289             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 290             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 291             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 292             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 293             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 294
 295             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 296             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 297             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 298             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 299             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 300             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 301             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 302             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 303             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 304
 305             fjx0             = _mm256_setzero_pd();
 306             fjy0             = _mm256_setzero_pd();
 307             fjz0             = _mm256_setzero_pd();
 308             fjx1             = _mm256_setzero_pd();
 309             fjy1             = _mm256_setzero_pd();
 310             fjz1             = _mm256_setzero_pd();
 311             fjx2             = _mm256_setzero_pd();
 312             fjy2             = _mm256_setzero_pd();
 313             fjz2             = _mm256_setzero_pd();
 314             fjx3             = _mm256_setzero_pd();
 315             fjy3             = _mm256_setzero_pd();
 316             fjz3             = _mm256_setzero_pd();
 317
 318             /**************************
 319              * CALCULATE INTERACTIONS *
 320              **************************/
 321
 322             r00              = _mm256_mul_pd(rsq00,rinv00);
 323
 324             /* Calculate table index by multiplying r with table scale and truncate to integer */
 325             rt               = _mm256_mul_pd(r00,vftabscale);
 326             vfitab           = _mm256_cvttpd_epi32(rt);
 327             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 328             vfitab           = _mm_slli_epi32(vfitab,3);
 329
 330             /* CUBIC SPLINE TABLE DISPERSION */
 331             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 332             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 333             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 334             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 335             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 336             Heps             = _mm256_mul_pd(vfeps,H);
 337             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 338             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 339             vvdw6            = _mm256_mul_pd(c6_00,VV);
 340             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 341             fvdw6            = _mm256_mul_pd(c6_00,FF);
 342
 343             /* CUBIC SPLINE TABLE REPULSION */
 344             vfitab           = _mm_add_epi32(vfitab,ifour);
 345             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 346             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 347             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 348             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 349             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 350             Heps             = _mm256_mul_pd(vfeps,H);
 351             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 352             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 353             vvdw12           = _mm256_mul_pd(c12_00,VV);
 354             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 355             fvdw12           = _mm256_mul_pd(c12_00,FF);
 356             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 357             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 358
 359             /* Update potential sum for this i atom from the interaction with this j atom. */
 360             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 361
 362             fscal            = fvdw;
 363
 364             /* Calculate temporary vectorial force */
 365             tx               = _mm256_mul_pd(fscal,dx00);
 366             ty               = _mm256_mul_pd(fscal,dy00);
 367             tz               = _mm256_mul_pd(fscal,dz00);
 368
 369             /* Update vectorial force */
 370             fix0             = _mm256_add_pd(fix0,tx);
 371             fiy0             = _mm256_add_pd(fiy0,ty);
 372             fiz0             = _mm256_add_pd(fiz0,tz);
 373
 374             fjx0             = _mm256_add_pd(fjx0,tx);
 375             fjy0             = _mm256_add_pd(fjy0,ty);
 376             fjz0             = _mm256_add_pd(fjz0,tz);
 377
 378             /**************************
 379              * CALCULATE INTERACTIONS *
 380              **************************/
 381
 382             /* COULOMB ELECTROSTATICS */
 383             velec            = _mm256_mul_pd(qq11,rinv11);
 384             felec            = _mm256_mul_pd(velec,rinvsq11);
 385
 386             /* Update potential sum for this i atom from the interaction with this j atom. */
 387             velecsum         = _mm256_add_pd(velecsum,velec);
 388
 389             fscal            = felec;
 390
 391             /* Calculate temporary vectorial force */
 392             tx               = _mm256_mul_pd(fscal,dx11);
 393             ty               = _mm256_mul_pd(fscal,dy11);
 394             tz               = _mm256_mul_pd(fscal,dz11);
 395
 396             /* Update vectorial force */
 397             fix1             = _mm256_add_pd(fix1,tx);
 398             fiy1             = _mm256_add_pd(fiy1,ty);
 399             fiz1             = _mm256_add_pd(fiz1,tz);
 400
 401             fjx1             = _mm256_add_pd(fjx1,tx);
 402             fjy1             = _mm256_add_pd(fjy1,ty);
 403             fjz1             = _mm256_add_pd(fjz1,tz);
 404
 405             /**************************
 406              * CALCULATE INTERACTIONS *
 407              **************************/
 408
 409             /* COULOMB ELECTROSTATICS */
 410             velec            = _mm256_mul_pd(qq12,rinv12);
 411             felec            = _mm256_mul_pd(velec,rinvsq12);
 412
 413             /* Update potential sum for this i atom from the interaction with this j atom. */
 414             velecsum         = _mm256_add_pd(velecsum,velec);
 415
 416             fscal            = felec;
 417
 418             /* Calculate temporary vectorial force */
 419             tx               = _mm256_mul_pd(fscal,dx12);
 420             ty               = _mm256_mul_pd(fscal,dy12);
 421             tz               = _mm256_mul_pd(fscal,dz12);
 422
 423             /* Update vectorial force */
 424             fix1             = _mm256_add_pd(fix1,tx);
 425             fiy1             = _mm256_add_pd(fiy1,ty);
 426             fiz1             = _mm256_add_pd(fiz1,tz);
 427
 428             fjx2             = _mm256_add_pd(fjx2,tx);
 429             fjy2             = _mm256_add_pd(fjy2,ty);
 430             fjz2             = _mm256_add_pd(fjz2,tz);
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             /* COULOMB ELECTROSTATICS */
 437             velec            = _mm256_mul_pd(qq13,rinv13);
 438             felec            = _mm256_mul_pd(velec,rinvsq13);
 439
 440             /* Update potential sum for this i atom from the interaction with this j atom. */
 441             velecsum         = _mm256_add_pd(velecsum,velec);
 442
 443             fscal            = felec;
 444
 445             /* Calculate temporary vectorial force */
 446             tx               = _mm256_mul_pd(fscal,dx13);
 447             ty               = _mm256_mul_pd(fscal,dy13);
 448             tz               = _mm256_mul_pd(fscal,dz13);
 449
 450             /* Update vectorial force */
 451             fix1             = _mm256_add_pd(fix1,tx);
 452             fiy1             = _mm256_add_pd(fiy1,ty);
 453             fiz1             = _mm256_add_pd(fiz1,tz);
 454
 455             fjx3             = _mm256_add_pd(fjx3,tx);
 456             fjy3             = _mm256_add_pd(fjy3,ty);
 457             fjz3             = _mm256_add_pd(fjz3,tz);
 458
 459             /**************************
 460              * CALCULATE INTERACTIONS *
 461              **************************/
 462
 463             /* COULOMB ELECTROSTATICS */
 464             velec            = _mm256_mul_pd(qq21,rinv21);
 465             felec            = _mm256_mul_pd(velec,rinvsq21);
 466
 467             /* Update potential sum for this i atom from the interaction with this j atom. */
 468             velecsum         = _mm256_add_pd(velecsum,velec);
 469
 470             fscal            = felec;
 471
 472             /* Calculate temporary vectorial force */
 473             tx               = _mm256_mul_pd(fscal,dx21);
 474             ty               = _mm256_mul_pd(fscal,dy21);
 475             tz               = _mm256_mul_pd(fscal,dz21);
 476
 477             /* Update vectorial force */
 478             fix2             = _mm256_add_pd(fix2,tx);
 479             fiy2             = _mm256_add_pd(fiy2,ty);
 480             fiz2             = _mm256_add_pd(fiz2,tz);
 481
 482             fjx1             = _mm256_add_pd(fjx1,tx);
 483             fjy1             = _mm256_add_pd(fjy1,ty);
 484             fjz1             = _mm256_add_pd(fjz1,tz);
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             /* COULOMB ELECTROSTATICS */
 491             velec            = _mm256_mul_pd(qq22,rinv22);
 492             felec            = _mm256_mul_pd(velec,rinvsq22);
 493
 494             /* Update potential sum for this i atom from the interaction with this j atom. */
 495             velecsum         = _mm256_add_pd(velecsum,velec);
 496
 497             fscal            = felec;
 498
 499             /* Calculate temporary vectorial force */
 500             tx               = _mm256_mul_pd(fscal,dx22);
 501             ty               = _mm256_mul_pd(fscal,dy22);
 502             tz               = _mm256_mul_pd(fscal,dz22);
 503
 504             /* Update vectorial force */
 505             fix2             = _mm256_add_pd(fix2,tx);
 506             fiy2             = _mm256_add_pd(fiy2,ty);
 507             fiz2             = _mm256_add_pd(fiz2,tz);
 508
 509             fjx2             = _mm256_add_pd(fjx2,tx);
 510             fjy2             = _mm256_add_pd(fjy2,ty);
 511             fjz2             = _mm256_add_pd(fjz2,tz);
 512
 513             /**************************
 514              * CALCULATE INTERACTIONS *
 515              **************************/
 516
 517             /* COULOMB ELECTROSTATICS */
 518             velec            = _mm256_mul_pd(qq23,rinv23);
 519             felec            = _mm256_mul_pd(velec,rinvsq23);
 520
 521             /* Update potential sum for this i atom from the interaction with this j atom. */
 522             velecsum         = _mm256_add_pd(velecsum,velec);
 523
 524             fscal            = felec;
 525
 526             /* Calculate temporary vectorial force */
 527             tx               = _mm256_mul_pd(fscal,dx23);
 528             ty               = _mm256_mul_pd(fscal,dy23);
 529             tz               = _mm256_mul_pd(fscal,dz23);
 530
 531             /* Update vectorial force */
 532             fix2             = _mm256_add_pd(fix2,tx);
 533             fiy2             = _mm256_add_pd(fiy2,ty);
 534             fiz2             = _mm256_add_pd(fiz2,tz);
 535
 536             fjx3             = _mm256_add_pd(fjx3,tx);
 537             fjy3             = _mm256_add_pd(fjy3,ty);
 538             fjz3             = _mm256_add_pd(fjz3,tz);
 539
 540             /**************************
 541              * CALCULATE INTERACTIONS *
 542              **************************/
 543
 544             /* COULOMB ELECTROSTATICS */
 545             velec            = _mm256_mul_pd(qq31,rinv31);
 546             felec            = _mm256_mul_pd(velec,rinvsq31);
 547
 548             /* Update potential sum for this i atom from the interaction with this j atom. */
 549             velecsum         = _mm256_add_pd(velecsum,velec);
 550
 551             fscal            = felec;
 552
 553             /* Calculate temporary vectorial force */
 554             tx               = _mm256_mul_pd(fscal,dx31);
 555             ty               = _mm256_mul_pd(fscal,dy31);
 556             tz               = _mm256_mul_pd(fscal,dz31);
 557
 558             /* Update vectorial force */
 559             fix3             = _mm256_add_pd(fix3,tx);
 560             fiy3             = _mm256_add_pd(fiy3,ty);
 561             fiz3             = _mm256_add_pd(fiz3,tz);
 562
 563             fjx1             = _mm256_add_pd(fjx1,tx);
 564             fjy1             = _mm256_add_pd(fjy1,ty);
 565             fjz1             = _mm256_add_pd(fjz1,tz);
 566
 567             /**************************
 568              * CALCULATE INTERACTIONS *
 569              **************************/
 570
 571             /* COULOMB ELECTROSTATICS */
 572             velec            = _mm256_mul_pd(qq32,rinv32);
 573             felec            = _mm256_mul_pd(velec,rinvsq32);
 574
 575             /* Update potential sum for this i atom from the interaction with this j atom. */
 576             velecsum         = _mm256_add_pd(velecsum,velec);
 577
 578             fscal            = felec;
 579
 580             /* Calculate temporary vectorial force */
 581             tx               = _mm256_mul_pd(fscal,dx32);
 582             ty               = _mm256_mul_pd(fscal,dy32);
 583             tz               = _mm256_mul_pd(fscal,dz32);
 584
 585             /* Update vectorial force */
 586             fix3             = _mm256_add_pd(fix3,tx);
 587             fiy3             = _mm256_add_pd(fiy3,ty);
 588             fiz3             = _mm256_add_pd(fiz3,tz);
 589
 590             fjx2             = _mm256_add_pd(fjx2,tx);
 591             fjy2             = _mm256_add_pd(fjy2,ty);
 592             fjz2             = _mm256_add_pd(fjz2,tz);
 593
 594             /**************************
 595              * CALCULATE INTERACTIONS *
 596              **************************/
 597
 598             /* COULOMB ELECTROSTATICS */
 599             velec            = _mm256_mul_pd(qq33,rinv33);
 600             felec            = _mm256_mul_pd(velec,rinvsq33);
 601
 602             /* Update potential sum for this i atom from the interaction with this j atom. */
 603             velecsum         = _mm256_add_pd(velecsum,velec);
 604
 605             fscal            = felec;
 606
 607             /* Calculate temporary vectorial force */
 608             tx               = _mm256_mul_pd(fscal,dx33);
 609             ty               = _mm256_mul_pd(fscal,dy33);
 610             tz               = _mm256_mul_pd(fscal,dz33);
 611
 612             /* Update vectorial force */
 613             fix3             = _mm256_add_pd(fix3,tx);
 614             fiy3             = _mm256_add_pd(fiy3,ty);
 615             fiz3             = _mm256_add_pd(fiz3,tz);
 616
 617             fjx3             = _mm256_add_pd(fjx3,tx);
 618             fjy3             = _mm256_add_pd(fjy3,ty);
 619             fjz3             = _mm256_add_pd(fjz3,tz);
 620
 621             fjptrA             = f+j_coord_offsetA;
 622             fjptrB             = f+j_coord_offsetB;
 623             fjptrC             = f+j_coord_offsetC;
 624             fjptrD             = f+j_coord_offsetD;
 625
 626             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 627                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 628                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 629
 630             /* Inner loop uses 302 flops */
 631         }
 632
 633         if(jidx<j_index_end)
 634         {
 635
 636             /* Get j neighbor index, and coordinate index */
 637             jnrlistA         = jjnr[jidx];
 638             jnrlistB         = jjnr[jidx+1];
 639             jnrlistC         = jjnr[jidx+2];
 640             jnrlistD         = jjnr[jidx+3];
 641             /* Sign of each element will be negative for non-real atoms.
 642              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 643              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 644              */
 645             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 646
 647             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 648             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 649             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 650
 651             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 652             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 653             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 654             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 655             j_coord_offsetA  = DIM*jnrA;
 656             j_coord_offsetB  = DIM*jnrB;
 657             j_coord_offsetC  = DIM*jnrC;
 658             j_coord_offsetD  = DIM*jnrD;
 659
 660             /* load j atom coordinates */
 661             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 662                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 663                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 664                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 665
 666             /* Calculate displacement vector */
 667             dx00             = _mm256_sub_pd(ix0,jx0);
 668             dy00             = _mm256_sub_pd(iy0,jy0);
 669             dz00             = _mm256_sub_pd(iz0,jz0);
 670             dx11             = _mm256_sub_pd(ix1,jx1);
 671             dy11             = _mm256_sub_pd(iy1,jy1);
 672             dz11             = _mm256_sub_pd(iz1,jz1);
 673             dx12             = _mm256_sub_pd(ix1,jx2);
 674             dy12             = _mm256_sub_pd(iy1,jy2);
 675             dz12             = _mm256_sub_pd(iz1,jz2);
 676             dx13             = _mm256_sub_pd(ix1,jx3);
 677             dy13             = _mm256_sub_pd(iy1,jy3);
 678             dz13             = _mm256_sub_pd(iz1,jz3);
 679             dx21             = _mm256_sub_pd(ix2,jx1);
 680             dy21             = _mm256_sub_pd(iy2,jy1);
 681             dz21             = _mm256_sub_pd(iz2,jz1);
 682             dx22             = _mm256_sub_pd(ix2,jx2);
 683             dy22             = _mm256_sub_pd(iy2,jy2);
 684             dz22             = _mm256_sub_pd(iz2,jz2);
 685             dx23             = _mm256_sub_pd(ix2,jx3);
 686             dy23             = _mm256_sub_pd(iy2,jy3);
 687             dz23             = _mm256_sub_pd(iz2,jz3);
 688             dx31             = _mm256_sub_pd(ix3,jx1);
 689             dy31             = _mm256_sub_pd(iy3,jy1);
 690             dz31             = _mm256_sub_pd(iz3,jz1);
 691             dx32             = _mm256_sub_pd(ix3,jx2);
 692             dy32             = _mm256_sub_pd(iy3,jy2);
 693             dz32             = _mm256_sub_pd(iz3,jz2);
 694             dx33             = _mm256_sub_pd(ix3,jx3);
 695             dy33             = _mm256_sub_pd(iy3,jy3);
 696             dz33             = _mm256_sub_pd(iz3,jz3);
 697
 698             /* Calculate squared distance and things based on it */
 699             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 700             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 701             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 702             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
 703             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 704             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 705             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
 706             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
 707             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
 708             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
 709
 710             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 711             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 712             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 713             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
 714             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 715             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 716             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
 717             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
 718             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
 719             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
 720
 721             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 722             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 723             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
 724             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 725             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 726             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
 727             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
 728             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
 729             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
 730
 731             fjx0             = _mm256_setzero_pd();
 732             fjy0             = _mm256_setzero_pd();
 733             fjz0             = _mm256_setzero_pd();
 734             fjx1             = _mm256_setzero_pd();
 735             fjy1             = _mm256_setzero_pd();
 736             fjz1             = _mm256_setzero_pd();
 737             fjx2             = _mm256_setzero_pd();
 738             fjy2             = _mm256_setzero_pd();
 739             fjz2             = _mm256_setzero_pd();
 740             fjx3             = _mm256_setzero_pd();
 741             fjy3             = _mm256_setzero_pd();
 742             fjz3             = _mm256_setzero_pd();
 743
 744             /**************************
 745              * CALCULATE INTERACTIONS *
 746              **************************/
 747
 748             r00              = _mm256_mul_pd(rsq00,rinv00);
 749             r00              = _mm256_andnot_pd(dummy_mask,r00);
 750
 751             /* Calculate table index by multiplying r with table scale and truncate to integer */
 752             rt               = _mm256_mul_pd(r00,vftabscale);
 753             vfitab           = _mm256_cvttpd_epi32(rt);
 754             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 755             vfitab           = _mm_slli_epi32(vfitab,3);
 756
 757             /* CUBIC SPLINE TABLE DISPERSION */
 758             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 759             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 760             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 761             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 762             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 763             Heps             = _mm256_mul_pd(vfeps,H);
 764             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 765             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 766             vvdw6            = _mm256_mul_pd(c6_00,VV);
 767             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 768             fvdw6            = _mm256_mul_pd(c6_00,FF);
 769
 770             /* CUBIC SPLINE TABLE REPULSION */
 771             vfitab           = _mm_add_epi32(vfitab,ifour);
 772             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 773             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 774             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 775             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 776             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 777             Heps             = _mm256_mul_pd(vfeps,H);
 778             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 779             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 780             vvdw12           = _mm256_mul_pd(c12_00,VV);
 781             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 782             fvdw12           = _mm256_mul_pd(c12_00,FF);
 783             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 784             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 785
 786             /* Update potential sum for this i atom from the interaction with this j atom. */
 787             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 788             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 789
 790             fscal            = fvdw;
 791
 792             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 793
 794             /* Calculate temporary vectorial force */
 795             tx               = _mm256_mul_pd(fscal,dx00);
 796             ty               = _mm256_mul_pd(fscal,dy00);
 797             tz               = _mm256_mul_pd(fscal,dz00);
 798
 799             /* Update vectorial force */
 800             fix0             = _mm256_add_pd(fix0,tx);
 801             fiy0             = _mm256_add_pd(fiy0,ty);
 802             fiz0             = _mm256_add_pd(fiz0,tz);
 803
 804             fjx0             = _mm256_add_pd(fjx0,tx);
 805             fjy0             = _mm256_add_pd(fjy0,ty);
 806             fjz0             = _mm256_add_pd(fjz0,tz);
 807
 808             /**************************
 809              * CALCULATE INTERACTIONS *
 810              **************************/
 811
 812             /* COULOMB ELECTROSTATICS */
 813             velec            = _mm256_mul_pd(qq11,rinv11);
 814             felec            = _mm256_mul_pd(velec,rinvsq11);
 815
 816             /* Update potential sum for this i atom from the interaction with this j atom. */
 817             velec            = _mm256_andnot_pd(dummy_mask,velec);
 818             velecsum         = _mm256_add_pd(velecsum,velec);
 819
 820             fscal            = felec;
 821
 822             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 823
 824             /* Calculate temporary vectorial force */
 825             tx               = _mm256_mul_pd(fscal,dx11);
 826             ty               = _mm256_mul_pd(fscal,dy11);
 827             tz               = _mm256_mul_pd(fscal,dz11);
 828
 829             /* Update vectorial force */
 830             fix1             = _mm256_add_pd(fix1,tx);
 831             fiy1             = _mm256_add_pd(fiy1,ty);
 832             fiz1             = _mm256_add_pd(fiz1,tz);
 833
 834             fjx1             = _mm256_add_pd(fjx1,tx);
 835             fjy1             = _mm256_add_pd(fjy1,ty);
 836             fjz1             = _mm256_add_pd(fjz1,tz);
 837
 838             /**************************
 839              * CALCULATE INTERACTIONS *
 840              **************************/
 841
 842             /* COULOMB ELECTROSTATICS */
 843             velec            = _mm256_mul_pd(qq12,rinv12);
 844             felec            = _mm256_mul_pd(velec,rinvsq12);
 845
 846             /* Update potential sum for this i atom from the interaction with this j atom. */
 847             velec            = _mm256_andnot_pd(dummy_mask,velec);
 848             velecsum         = _mm256_add_pd(velecsum,velec);
 849
 850             fscal            = felec;
 851
 852             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 853
 854             /* Calculate temporary vectorial force */
 855             tx               = _mm256_mul_pd(fscal,dx12);
 856             ty               = _mm256_mul_pd(fscal,dy12);
 857             tz               = _mm256_mul_pd(fscal,dz12);
 858
 859             /* Update vectorial force */
 860             fix1             = _mm256_add_pd(fix1,tx);
 861             fiy1             = _mm256_add_pd(fiy1,ty);
 862             fiz1             = _mm256_add_pd(fiz1,tz);
 863
 864             fjx2             = _mm256_add_pd(fjx2,tx);
 865             fjy2             = _mm256_add_pd(fjy2,ty);
 866             fjz2             = _mm256_add_pd(fjz2,tz);
 867
 868             /**************************
 869              * CALCULATE INTERACTIONS *
 870              **************************/
 871
 872             /* COULOMB ELECTROSTATICS */
 873             velec            = _mm256_mul_pd(qq13,rinv13);
 874             felec            = _mm256_mul_pd(velec,rinvsq13);
 875
 876             /* Update potential sum for this i atom from the interaction with this j atom. */
 877             velec            = _mm256_andnot_pd(dummy_mask,velec);
 878             velecsum         = _mm256_add_pd(velecsum,velec);
 879
 880             fscal            = felec;
 881
 882             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 883
 884             /* Calculate temporary vectorial force */
 885             tx               = _mm256_mul_pd(fscal,dx13);
 886             ty               = _mm256_mul_pd(fscal,dy13);
 887             tz               = _mm256_mul_pd(fscal,dz13);
 888
 889             /* Update vectorial force */
 890             fix1             = _mm256_add_pd(fix1,tx);
 891             fiy1             = _mm256_add_pd(fiy1,ty);
 892             fiz1             = _mm256_add_pd(fiz1,tz);
 893
 894             fjx3             = _mm256_add_pd(fjx3,tx);
 895             fjy3             = _mm256_add_pd(fjy3,ty);
 896             fjz3             = _mm256_add_pd(fjz3,tz);
 897
 898             /**************************
 899              * CALCULATE INTERACTIONS *
 900              **************************/
 901
 902             /* COULOMB ELECTROSTATICS */
 903             velec            = _mm256_mul_pd(qq21,rinv21);
 904             felec            = _mm256_mul_pd(velec,rinvsq21);
 905
 906             /* Update potential sum for this i atom from the interaction with this j atom. */
 907             velec            = _mm256_andnot_pd(dummy_mask,velec);
 908             velecsum         = _mm256_add_pd(velecsum,velec);
 909
 910             fscal            = felec;
 911
 912             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 913
 914             /* Calculate temporary vectorial force */
 915             tx               = _mm256_mul_pd(fscal,dx21);
 916             ty               = _mm256_mul_pd(fscal,dy21);
 917             tz               = _mm256_mul_pd(fscal,dz21);
 918
 919             /* Update vectorial force */
 920             fix2             = _mm256_add_pd(fix2,tx);
 921             fiy2             = _mm256_add_pd(fiy2,ty);
 922             fiz2             = _mm256_add_pd(fiz2,tz);
 923
 924             fjx1             = _mm256_add_pd(fjx1,tx);
 925             fjy1             = _mm256_add_pd(fjy1,ty);
 926             fjz1             = _mm256_add_pd(fjz1,tz);
 927
 928             /**************************
 929              * CALCULATE INTERACTIONS *
 930              **************************/
 931
 932             /* COULOMB ELECTROSTATICS */
 933             velec            = _mm256_mul_pd(qq22,rinv22);
 934             felec            = _mm256_mul_pd(velec,rinvsq22);
 935
 936             /* Update potential sum for this i atom from the interaction with this j atom. */
 937             velec            = _mm256_andnot_pd(dummy_mask,velec);
 938             velecsum         = _mm256_add_pd(velecsum,velec);
 939
 940             fscal            = felec;
 941
 942             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 943
 944             /* Calculate temporary vectorial force */
 945             tx               = _mm256_mul_pd(fscal,dx22);
 946             ty               = _mm256_mul_pd(fscal,dy22);
 947             tz               = _mm256_mul_pd(fscal,dz22);
 948
 949             /* Update vectorial force */
 950             fix2             = _mm256_add_pd(fix2,tx);
 951             fiy2             = _mm256_add_pd(fiy2,ty);
 952             fiz2             = _mm256_add_pd(fiz2,tz);
 953
 954             fjx2             = _mm256_add_pd(fjx2,tx);
 955             fjy2             = _mm256_add_pd(fjy2,ty);
 956             fjz2             = _mm256_add_pd(fjz2,tz);
 957
 958             /**************************
 959              * CALCULATE INTERACTIONS *
 960              **************************/
 961
 962             /* COULOMB ELECTROSTATICS */
 963             velec            = _mm256_mul_pd(qq23,rinv23);
 964             felec            = _mm256_mul_pd(velec,rinvsq23);
 965
 966             /* Update potential sum for this i atom from the interaction with this j atom. */
 967             velec            = _mm256_andnot_pd(dummy_mask,velec);
 968             velecsum         = _mm256_add_pd(velecsum,velec);
 969
 970             fscal            = felec;
 971
 972             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 973
 974             /* Calculate temporary vectorial force */
 975             tx               = _mm256_mul_pd(fscal,dx23);
 976             ty               = _mm256_mul_pd(fscal,dy23);
 977             tz               = _mm256_mul_pd(fscal,dz23);
 978
 979             /* Update vectorial force */
 980             fix2             = _mm256_add_pd(fix2,tx);
 981             fiy2             = _mm256_add_pd(fiy2,ty);
 982             fiz2             = _mm256_add_pd(fiz2,tz);
 983
 984             fjx3             = _mm256_add_pd(fjx3,tx);
 985             fjy3             = _mm256_add_pd(fjy3,ty);
 986             fjz3             = _mm256_add_pd(fjz3,tz);
 987
 988             /**************************
 989              * CALCULATE INTERACTIONS *
 990              **************************/
 991
 992             /* COULOMB ELECTROSTATICS */
 993             velec            = _mm256_mul_pd(qq31,rinv31);
 994             felec            = _mm256_mul_pd(velec,rinvsq31);
 995
 996             /* Update potential sum for this i atom from the interaction with this j atom. */
 997             velec            = _mm256_andnot_pd(dummy_mask,velec);
 998             velecsum         = _mm256_add_pd(velecsum,velec);
 999
1000             fscal            = felec;
1001
1002             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1003
1004             /* Calculate temporary vectorial force */
1005             tx               = _mm256_mul_pd(fscal,dx31);
1006             ty               = _mm256_mul_pd(fscal,dy31);
1007             tz               = _mm256_mul_pd(fscal,dz31);
1008
1009             /* Update vectorial force */
1010             fix3             = _mm256_add_pd(fix3,tx);
1011             fiy3             = _mm256_add_pd(fiy3,ty);
1012             fiz3             = _mm256_add_pd(fiz3,tz);
1013
1014             fjx1             = _mm256_add_pd(fjx1,tx);
1015             fjy1             = _mm256_add_pd(fjy1,ty);
1016             fjz1             = _mm256_add_pd(fjz1,tz);
1017
1018             /**************************
1019              * CALCULATE INTERACTIONS *
1020              **************************/
1021
1022             /* COULOMB ELECTROSTATICS */
1023             velec            = _mm256_mul_pd(qq32,rinv32);
1024             felec            = _mm256_mul_pd(velec,rinvsq32);
1025
1026             /* Update potential sum for this i atom from the interaction with this j atom. */
1027             velec            = _mm256_andnot_pd(dummy_mask,velec);
1028             velecsum         = _mm256_add_pd(velecsum,velec);
1029
1030             fscal            = felec;
1031
1032             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1033
1034             /* Calculate temporary vectorial force */
1035             tx               = _mm256_mul_pd(fscal,dx32);
1036             ty               = _mm256_mul_pd(fscal,dy32);
1037             tz               = _mm256_mul_pd(fscal,dz32);
1038
1039             /* Update vectorial force */
1040             fix3             = _mm256_add_pd(fix3,tx);
1041             fiy3             = _mm256_add_pd(fiy3,ty);
1042             fiz3             = _mm256_add_pd(fiz3,tz);
1043
1044             fjx2             = _mm256_add_pd(fjx2,tx);
1045             fjy2             = _mm256_add_pd(fjy2,ty);
1046             fjz2             = _mm256_add_pd(fjz2,tz);
1047
1048             /**************************
1049              * CALCULATE INTERACTIONS *
1050              **************************/
1051
1052             /* COULOMB ELECTROSTATICS */
1053             velec            = _mm256_mul_pd(qq33,rinv33);
1054             felec            = _mm256_mul_pd(velec,rinvsq33);
1055
1056             /* Update potential sum for this i atom from the interaction with this j atom. */
1057             velec            = _mm256_andnot_pd(dummy_mask,velec);
1058             velecsum         = _mm256_add_pd(velecsum,velec);
1059
1060             fscal            = felec;
1061
1062             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1063
1064             /* Calculate temporary vectorial force */
1065             tx               = _mm256_mul_pd(fscal,dx33);
1066             ty               = _mm256_mul_pd(fscal,dy33);
1067             tz               = _mm256_mul_pd(fscal,dz33);
1068
1069             /* Update vectorial force */
1070             fix3             = _mm256_add_pd(fix3,tx);
1071             fiy3             = _mm256_add_pd(fiy3,ty);
1072             fiz3             = _mm256_add_pd(fiz3,tz);
1073
1074             fjx3             = _mm256_add_pd(fjx3,tx);
1075             fjy3             = _mm256_add_pd(fjy3,ty);
1076             fjz3             = _mm256_add_pd(fjz3,tz);
1077
1078             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1079             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1080             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1081             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1082
1083             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1084                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1085                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1086
1087             /* Inner loop uses 303 flops */
1088         }
1089
1090         /* End of innermost loop */
1091
1092         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1093                                                  f+i_coord_offset,fshift+i_shift_offset);
1094
1095         ggid                        = gid[iidx];
1096         /* Update potential energies */
1097         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1098         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1099
1100         /* Increment number of inner iterations */
1101         inneriter                  += j_index_end - j_index_start;
1102
1103         /* Outer loop uses 26 flops */
1104     }
1105
1106     /* Increment number of outer iterations */
1107     outeriter        += nri;
1108
1109     /* Update outer/inner flops */
1110
1111     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*303);
1112 }
1113 /*
1114  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_256_double
1115  * Electrostatics interaction: Coulomb
1116  * VdW interaction:            CubicSplineTable
1117  * Geometry:                   Water4-Water4
1118  * Calculate force/pot:        Force
1119  */
1120 void
1121 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_256_double
1122                     (t_nblist                    * gmx_restrict       nlist,
1123                      rvec                        * gmx_restrict          xx,
1124                      rvec                        * gmx_restrict          ff,
1125                      t_forcerec                  * gmx_restrict          fr,
1126                      t_mdatoms                   * gmx_restrict     mdatoms,
1127                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1128                      t_nrnb                      * gmx_restrict        nrnb)
1129 {
1130     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1131      * just 0 for non-waters.
1132      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1133      * jnr indices corresponding to data put in the four positions in the SIMD register.
1134      */
1135     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1136     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1137     int              jnrA,jnrB,jnrC,jnrD;
1138     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1139     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1140     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1141     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1142     real             rcutoff_scalar;
1143     real             *shiftvec,*fshift,*x,*f;
1144     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1145     real             scratch[4*DIM];
1146     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1147     real *           vdwioffsetptr0;
1148     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1149     real *           vdwioffsetptr1;
1150     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1151     real *           vdwioffsetptr2;
1152     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1153     real *           vdwioffsetptr3;
1154     __m256d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1155     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1156     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1157     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1158     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1159     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1160     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1161     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1162     __m256d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1163     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1164     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1165     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1166     __m256d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1167     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1168     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1169     __m256d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1170     __m256d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1171     __m256d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1172     __m256d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1173     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1174     real             *charge;
1175     int              nvdwtype;
1176     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1177     int              *vdwtype;
1178     real             *vdwparam;
1179     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1180     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1181     __m128i          vfitab;
1182     __m128i          ifour       = _mm_set1_epi32(4);
1183     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1184     real             *vftab;
1185     __m256d          dummy_mask,cutoff_mask;
1186     __m128           tmpmask0,tmpmask1;
1187     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1188     __m256d          one     = _mm256_set1_pd(1.0);
1189     __m256d          two     = _mm256_set1_pd(2.0);
1190     x                = xx[0];
1191     f                = ff[0];
1192
1193     nri              = nlist->nri;
1194     iinr             = nlist->iinr;
1195     jindex           = nlist->jindex;
1196     jjnr             = nlist->jjnr;
1197     shiftidx         = nlist->shift;
1198     gid              = nlist->gid;
1199     shiftvec         = fr->shift_vec[0];
1200     fshift           = fr->fshift[0];
1201     facel            = _mm256_set1_pd(fr->epsfac);
1202     charge           = mdatoms->chargeA;
1203     nvdwtype         = fr->ntype;
1204     vdwparam         = fr->nbfp;
1205     vdwtype          = mdatoms->typeA;
1206
1207     vftab            = kernel_data->table_vdw->data;
1208     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
1209
1210     /* Setup water-specific parameters */
1211     inr              = nlist->iinr[0];
1212     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1213     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1214     iq3              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+3]));
1215     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1216
1217     jq1              = _mm256_set1_pd(charge[inr+1]);
1218     jq2              = _mm256_set1_pd(charge[inr+2]);
1219     jq3              = _mm256_set1_pd(charge[inr+3]);
1220     vdwjidx0A        = 2*vdwtype[inr+0];
1221     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1222     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1223     qq11             = _mm256_mul_pd(iq1,jq1);
1224     qq12             = _mm256_mul_pd(iq1,jq2);
1225     qq13             = _mm256_mul_pd(iq1,jq3);
1226     qq21             = _mm256_mul_pd(iq2,jq1);
1227     qq22             = _mm256_mul_pd(iq2,jq2);
1228     qq23             = _mm256_mul_pd(iq2,jq3);
1229     qq31             = _mm256_mul_pd(iq3,jq1);
1230     qq32             = _mm256_mul_pd(iq3,jq2);
1231     qq33             = _mm256_mul_pd(iq3,jq3);
1232
1233     /* Avoid stupid compiler warnings */
1234     jnrA = jnrB = jnrC = jnrD = 0;
1235     j_coord_offsetA = 0;
1236     j_coord_offsetB = 0;
1237     j_coord_offsetC = 0;
1238     j_coord_offsetD = 0;
1239
1240     outeriter        = 0;
1241     inneriter        = 0;
1242
1243     for(iidx=0;iidx<4*DIM;iidx++)
1244     {
1245         scratch[iidx] = 0.0;
1246     }
1247
1248     /* Start outer loop over neighborlists */
1249     for(iidx=0; iidx<nri; iidx++)
1250     {
1251         /* Load shift vector for this list */
1252         i_shift_offset   = DIM*shiftidx[iidx];
1253
1254         /* Load limits for loop over neighbors */
1255         j_index_start    = jindex[iidx];
1256         j_index_end      = jindex[iidx+1];
1257
1258         /* Get outer coordinate index */
1259         inr              = iinr[iidx];
1260         i_coord_offset   = DIM*inr;
1261
1262         /* Load i particle coords and add shift vector */
1263         gmx_mm256_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1264                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1265
1266         fix0             = _mm256_setzero_pd();
1267         fiy0             = _mm256_setzero_pd();
1268         fiz0             = _mm256_setzero_pd();
1269         fix1             = _mm256_setzero_pd();
1270         fiy1             = _mm256_setzero_pd();
1271         fiz1             = _mm256_setzero_pd();
1272         fix2             = _mm256_setzero_pd();
1273         fiy2             = _mm256_setzero_pd();
1274         fiz2             = _mm256_setzero_pd();
1275         fix3             = _mm256_setzero_pd();
1276         fiy3             = _mm256_setzero_pd();
1277         fiz3             = _mm256_setzero_pd();
1278
1279         /* Start inner kernel loop */
1280         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1281         {
1282
1283             /* Get j neighbor index, and coordinate index */
1284             jnrA             = jjnr[jidx];
1285             jnrB             = jjnr[jidx+1];
1286             jnrC             = jjnr[jidx+2];
1287             jnrD             = jjnr[jidx+3];
1288             j_coord_offsetA  = DIM*jnrA;
1289             j_coord_offsetB  = DIM*jnrB;
1290             j_coord_offsetC  = DIM*jnrC;
1291             j_coord_offsetD  = DIM*jnrD;
1292
1293             /* load j atom coordinates */
1294             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1295                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1296                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1297                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1298
1299             /* Calculate displacement vector */
1300             dx00             = _mm256_sub_pd(ix0,jx0);
1301             dy00             = _mm256_sub_pd(iy0,jy0);
1302             dz00             = _mm256_sub_pd(iz0,jz0);
1303             dx11             = _mm256_sub_pd(ix1,jx1);
1304             dy11             = _mm256_sub_pd(iy1,jy1);
1305             dz11             = _mm256_sub_pd(iz1,jz1);
1306             dx12             = _mm256_sub_pd(ix1,jx2);
1307             dy12             = _mm256_sub_pd(iy1,jy2);
1308             dz12             = _mm256_sub_pd(iz1,jz2);
1309             dx13             = _mm256_sub_pd(ix1,jx3);
1310             dy13             = _mm256_sub_pd(iy1,jy3);
1311             dz13             = _mm256_sub_pd(iz1,jz3);
1312             dx21             = _mm256_sub_pd(ix2,jx1);
1313             dy21             = _mm256_sub_pd(iy2,jy1);
1314             dz21             = _mm256_sub_pd(iz2,jz1);
1315             dx22             = _mm256_sub_pd(ix2,jx2);
1316             dy22             = _mm256_sub_pd(iy2,jy2);
1317             dz22             = _mm256_sub_pd(iz2,jz2);
1318             dx23             = _mm256_sub_pd(ix2,jx3);
1319             dy23             = _mm256_sub_pd(iy2,jy3);
1320             dz23             = _mm256_sub_pd(iz2,jz3);
1321             dx31             = _mm256_sub_pd(ix3,jx1);
1322             dy31             = _mm256_sub_pd(iy3,jy1);
1323             dz31             = _mm256_sub_pd(iz3,jz1);
1324             dx32             = _mm256_sub_pd(ix3,jx2);
1325             dy32             = _mm256_sub_pd(iy3,jy2);
1326             dz32             = _mm256_sub_pd(iz3,jz2);
1327             dx33             = _mm256_sub_pd(ix3,jx3);
1328             dy33             = _mm256_sub_pd(iy3,jy3);
1329             dz33             = _mm256_sub_pd(iz3,jz3);
1330
1331             /* Calculate squared distance and things based on it */
1332             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1333             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1334             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1335             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1336             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1337             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1338             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1339             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1340             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1341             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1342
1343             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1344             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1345             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1346             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1347             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1348             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1349             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1350             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1351             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1352             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1353
1354             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1355             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1356             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1357             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1358             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1359             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1360             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1361             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1362             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1363
1364             fjx0             = _mm256_setzero_pd();
1365             fjy0             = _mm256_setzero_pd();
1366             fjz0             = _mm256_setzero_pd();
1367             fjx1             = _mm256_setzero_pd();
1368             fjy1             = _mm256_setzero_pd();
1369             fjz1             = _mm256_setzero_pd();
1370             fjx2             = _mm256_setzero_pd();
1371             fjy2             = _mm256_setzero_pd();
1372             fjz2             = _mm256_setzero_pd();
1373             fjx3             = _mm256_setzero_pd();
1374             fjy3             = _mm256_setzero_pd();
1375             fjz3             = _mm256_setzero_pd();
1376
1377             /**************************
1378              * CALCULATE INTERACTIONS *
1379              **************************/
1380
1381             r00              = _mm256_mul_pd(rsq00,rinv00);
1382
1383             /* Calculate table index by multiplying r with table scale and truncate to integer */
1384             rt               = _mm256_mul_pd(r00,vftabscale);
1385             vfitab           = _mm256_cvttpd_epi32(rt);
1386             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1387             vfitab           = _mm_slli_epi32(vfitab,3);
1388
1389             /* CUBIC SPLINE TABLE DISPERSION */
1390             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1391             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1392             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1393             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1394             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1395             Heps             = _mm256_mul_pd(vfeps,H);
1396             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1397             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1398             fvdw6            = _mm256_mul_pd(c6_00,FF);
1399
1400             /* CUBIC SPLINE TABLE REPULSION */
1401             vfitab           = _mm_add_epi32(vfitab,ifour);
1402             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1403             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1404             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1405             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1406             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1407             Heps             = _mm256_mul_pd(vfeps,H);
1408             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1409             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1410             fvdw12           = _mm256_mul_pd(c12_00,FF);
1411             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1412
1413             fscal            = fvdw;
1414
1415             /* Calculate temporary vectorial force */
1416             tx               = _mm256_mul_pd(fscal,dx00);
1417             ty               = _mm256_mul_pd(fscal,dy00);
1418             tz               = _mm256_mul_pd(fscal,dz00);
1419
1420             /* Update vectorial force */
1421             fix0             = _mm256_add_pd(fix0,tx);
1422             fiy0             = _mm256_add_pd(fiy0,ty);
1423             fiz0             = _mm256_add_pd(fiz0,tz);
1424
1425             fjx0             = _mm256_add_pd(fjx0,tx);
1426             fjy0             = _mm256_add_pd(fjy0,ty);
1427             fjz0             = _mm256_add_pd(fjz0,tz);
1428
1429             /**************************
1430              * CALCULATE INTERACTIONS *
1431              **************************/
1432
1433             /* COULOMB ELECTROSTATICS */
1434             velec            = _mm256_mul_pd(qq11,rinv11);
1435             felec            = _mm256_mul_pd(velec,rinvsq11);
1436
1437             fscal            = felec;
1438
1439             /* Calculate temporary vectorial force */
1440             tx               = _mm256_mul_pd(fscal,dx11);
1441             ty               = _mm256_mul_pd(fscal,dy11);
1442             tz               = _mm256_mul_pd(fscal,dz11);
1443
1444             /* Update vectorial force */
1445             fix1             = _mm256_add_pd(fix1,tx);
1446             fiy1             = _mm256_add_pd(fiy1,ty);
1447             fiz1             = _mm256_add_pd(fiz1,tz);
1448
1449             fjx1             = _mm256_add_pd(fjx1,tx);
1450             fjy1             = _mm256_add_pd(fjy1,ty);
1451             fjz1             = _mm256_add_pd(fjz1,tz);
1452
1453             /**************************
1454              * CALCULATE INTERACTIONS *
1455              **************************/
1456
1457             /* COULOMB ELECTROSTATICS */
1458             velec            = _mm256_mul_pd(qq12,rinv12);
1459             felec            = _mm256_mul_pd(velec,rinvsq12);
1460
1461             fscal            = felec;
1462
1463             /* Calculate temporary vectorial force */
1464             tx               = _mm256_mul_pd(fscal,dx12);
1465             ty               = _mm256_mul_pd(fscal,dy12);
1466             tz               = _mm256_mul_pd(fscal,dz12);
1467
1468             /* Update vectorial force */
1469             fix1             = _mm256_add_pd(fix1,tx);
1470             fiy1             = _mm256_add_pd(fiy1,ty);
1471             fiz1             = _mm256_add_pd(fiz1,tz);
1472
1473             fjx2             = _mm256_add_pd(fjx2,tx);
1474             fjy2             = _mm256_add_pd(fjy2,ty);
1475             fjz2             = _mm256_add_pd(fjz2,tz);
1476
1477             /**************************
1478              * CALCULATE INTERACTIONS *
1479              **************************/
1480
1481             /* COULOMB ELECTROSTATICS */
1482             velec            = _mm256_mul_pd(qq13,rinv13);
1483             felec            = _mm256_mul_pd(velec,rinvsq13);
1484
1485             fscal            = felec;
1486
1487             /* Calculate temporary vectorial force */
1488             tx               = _mm256_mul_pd(fscal,dx13);
1489             ty               = _mm256_mul_pd(fscal,dy13);
1490             tz               = _mm256_mul_pd(fscal,dz13);
1491
1492             /* Update vectorial force */
1493             fix1             = _mm256_add_pd(fix1,tx);
1494             fiy1             = _mm256_add_pd(fiy1,ty);
1495             fiz1             = _mm256_add_pd(fiz1,tz);
1496
1497             fjx3             = _mm256_add_pd(fjx3,tx);
1498             fjy3             = _mm256_add_pd(fjy3,ty);
1499             fjz3             = _mm256_add_pd(fjz3,tz);
1500
1501             /**************************
1502              * CALCULATE INTERACTIONS *
1503              **************************/
1504
1505             /* COULOMB ELECTROSTATICS */
1506             velec            = _mm256_mul_pd(qq21,rinv21);
1507             felec            = _mm256_mul_pd(velec,rinvsq21);
1508
1509             fscal            = felec;
1510
1511             /* Calculate temporary vectorial force */
1512             tx               = _mm256_mul_pd(fscal,dx21);
1513             ty               = _mm256_mul_pd(fscal,dy21);
1514             tz               = _mm256_mul_pd(fscal,dz21);
1515
1516             /* Update vectorial force */
1517             fix2             = _mm256_add_pd(fix2,tx);
1518             fiy2             = _mm256_add_pd(fiy2,ty);
1519             fiz2             = _mm256_add_pd(fiz2,tz);
1520
1521             fjx1             = _mm256_add_pd(fjx1,tx);
1522             fjy1             = _mm256_add_pd(fjy1,ty);
1523             fjz1             = _mm256_add_pd(fjz1,tz);
1524
1525             /**************************
1526              * CALCULATE INTERACTIONS *
1527              **************************/
1528
1529             /* COULOMB ELECTROSTATICS */
1530             velec            = _mm256_mul_pd(qq22,rinv22);
1531             felec            = _mm256_mul_pd(velec,rinvsq22);
1532
1533             fscal            = felec;
1534
1535             /* Calculate temporary vectorial force */
1536             tx               = _mm256_mul_pd(fscal,dx22);
1537             ty               = _mm256_mul_pd(fscal,dy22);
1538             tz               = _mm256_mul_pd(fscal,dz22);
1539
1540             /* Update vectorial force */
1541             fix2             = _mm256_add_pd(fix2,tx);
1542             fiy2             = _mm256_add_pd(fiy2,ty);
1543             fiz2             = _mm256_add_pd(fiz2,tz);
1544
1545             fjx2             = _mm256_add_pd(fjx2,tx);
1546             fjy2             = _mm256_add_pd(fjy2,ty);
1547             fjz2             = _mm256_add_pd(fjz2,tz);
1548
1549             /**************************
1550              * CALCULATE INTERACTIONS *
1551              **************************/
1552
1553             /* COULOMB ELECTROSTATICS */
1554             velec            = _mm256_mul_pd(qq23,rinv23);
1555             felec            = _mm256_mul_pd(velec,rinvsq23);
1556
1557             fscal            = felec;
1558
1559             /* Calculate temporary vectorial force */
1560             tx               = _mm256_mul_pd(fscal,dx23);
1561             ty               = _mm256_mul_pd(fscal,dy23);
1562             tz               = _mm256_mul_pd(fscal,dz23);
1563
1564             /* Update vectorial force */
1565             fix2             = _mm256_add_pd(fix2,tx);
1566             fiy2             = _mm256_add_pd(fiy2,ty);
1567             fiz2             = _mm256_add_pd(fiz2,tz);
1568
1569             fjx3             = _mm256_add_pd(fjx3,tx);
1570             fjy3             = _mm256_add_pd(fjy3,ty);
1571             fjz3             = _mm256_add_pd(fjz3,tz);
1572
1573             /**************************
1574              * CALCULATE INTERACTIONS *
1575              **************************/
1576
1577             /* COULOMB ELECTROSTATICS */
1578             velec            = _mm256_mul_pd(qq31,rinv31);
1579             felec            = _mm256_mul_pd(velec,rinvsq31);
1580
1581             fscal            = felec;
1582
1583             /* Calculate temporary vectorial force */
1584             tx               = _mm256_mul_pd(fscal,dx31);
1585             ty               = _mm256_mul_pd(fscal,dy31);
1586             tz               = _mm256_mul_pd(fscal,dz31);
1587
1588             /* Update vectorial force */
1589             fix3             = _mm256_add_pd(fix3,tx);
1590             fiy3             = _mm256_add_pd(fiy3,ty);
1591             fiz3             = _mm256_add_pd(fiz3,tz);
1592
1593             fjx1             = _mm256_add_pd(fjx1,tx);
1594             fjy1             = _mm256_add_pd(fjy1,ty);
1595             fjz1             = _mm256_add_pd(fjz1,tz);
1596
1597             /**************************
1598              * CALCULATE INTERACTIONS *
1599              **************************/
1600
1601             /* COULOMB ELECTROSTATICS */
1602             velec            = _mm256_mul_pd(qq32,rinv32);
1603             felec            = _mm256_mul_pd(velec,rinvsq32);
1604
1605             fscal            = felec;
1606
1607             /* Calculate temporary vectorial force */
1608             tx               = _mm256_mul_pd(fscal,dx32);
1609             ty               = _mm256_mul_pd(fscal,dy32);
1610             tz               = _mm256_mul_pd(fscal,dz32);
1611
1612             /* Update vectorial force */
1613             fix3             = _mm256_add_pd(fix3,tx);
1614             fiy3             = _mm256_add_pd(fiy3,ty);
1615             fiz3             = _mm256_add_pd(fiz3,tz);
1616
1617             fjx2             = _mm256_add_pd(fjx2,tx);
1618             fjy2             = _mm256_add_pd(fjy2,ty);
1619             fjz2             = _mm256_add_pd(fjz2,tz);
1620
1621             /**************************
1622              * CALCULATE INTERACTIONS *
1623              **************************/
1624
1625             /* COULOMB ELECTROSTATICS */
1626             velec            = _mm256_mul_pd(qq33,rinv33);
1627             felec            = _mm256_mul_pd(velec,rinvsq33);
1628
1629             fscal            = felec;
1630
1631             /* Calculate temporary vectorial force */
1632             tx               = _mm256_mul_pd(fscal,dx33);
1633             ty               = _mm256_mul_pd(fscal,dy33);
1634             tz               = _mm256_mul_pd(fscal,dz33);
1635
1636             /* Update vectorial force */
1637             fix3             = _mm256_add_pd(fix3,tx);
1638             fiy3             = _mm256_add_pd(fiy3,ty);
1639             fiz3             = _mm256_add_pd(fiz3,tz);
1640
1641             fjx3             = _mm256_add_pd(fjx3,tx);
1642             fjy3             = _mm256_add_pd(fjy3,ty);
1643             fjz3             = _mm256_add_pd(fjz3,tz);
1644
1645             fjptrA             = f+j_coord_offsetA;
1646             fjptrB             = f+j_coord_offsetB;
1647             fjptrC             = f+j_coord_offsetC;
1648             fjptrD             = f+j_coord_offsetD;
1649
1650             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1651                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1652                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1653
1654             /* Inner loop uses 285 flops */
1655         }
1656
1657         if(jidx<j_index_end)
1658         {
1659
1660             /* Get j neighbor index, and coordinate index */
1661             jnrlistA         = jjnr[jidx];
1662             jnrlistB         = jjnr[jidx+1];
1663             jnrlistC         = jjnr[jidx+2];
1664             jnrlistD         = jjnr[jidx+3];
1665             /* Sign of each element will be negative for non-real atoms.
1666              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1667              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1668              */
1669             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1670
1671             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1672             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1673             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1674
1675             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1676             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1677             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1678             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1679             j_coord_offsetA  = DIM*jnrA;
1680             j_coord_offsetB  = DIM*jnrB;
1681             j_coord_offsetC  = DIM*jnrC;
1682             j_coord_offsetD  = DIM*jnrD;
1683
1684             /* load j atom coordinates */
1685             gmx_mm256_load_4rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1686                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1687                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1688                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1689
1690             /* Calculate displacement vector */
1691             dx00             = _mm256_sub_pd(ix0,jx0);
1692             dy00             = _mm256_sub_pd(iy0,jy0);
1693             dz00             = _mm256_sub_pd(iz0,jz0);
1694             dx11             = _mm256_sub_pd(ix1,jx1);
1695             dy11             = _mm256_sub_pd(iy1,jy1);
1696             dz11             = _mm256_sub_pd(iz1,jz1);
1697             dx12             = _mm256_sub_pd(ix1,jx2);
1698             dy12             = _mm256_sub_pd(iy1,jy2);
1699             dz12             = _mm256_sub_pd(iz1,jz2);
1700             dx13             = _mm256_sub_pd(ix1,jx3);
1701             dy13             = _mm256_sub_pd(iy1,jy3);
1702             dz13             = _mm256_sub_pd(iz1,jz3);
1703             dx21             = _mm256_sub_pd(ix2,jx1);
1704             dy21             = _mm256_sub_pd(iy2,jy1);
1705             dz21             = _mm256_sub_pd(iz2,jz1);
1706             dx22             = _mm256_sub_pd(ix2,jx2);
1707             dy22             = _mm256_sub_pd(iy2,jy2);
1708             dz22             = _mm256_sub_pd(iz2,jz2);
1709             dx23             = _mm256_sub_pd(ix2,jx3);
1710             dy23             = _mm256_sub_pd(iy2,jy3);
1711             dz23             = _mm256_sub_pd(iz2,jz3);
1712             dx31             = _mm256_sub_pd(ix3,jx1);
1713             dy31             = _mm256_sub_pd(iy3,jy1);
1714             dz31             = _mm256_sub_pd(iz3,jz1);
1715             dx32             = _mm256_sub_pd(ix3,jx2);
1716             dy32             = _mm256_sub_pd(iy3,jy2);
1717             dz32             = _mm256_sub_pd(iz3,jz2);
1718             dx33             = _mm256_sub_pd(ix3,jx3);
1719             dy33             = _mm256_sub_pd(iy3,jy3);
1720             dz33             = _mm256_sub_pd(iz3,jz3);
1721
1722             /* Calculate squared distance and things based on it */
1723             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1724             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1725             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1726             rsq13            = gmx_mm256_calc_rsq_pd(dx13,dy13,dz13);
1727             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1728             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1729             rsq23            = gmx_mm256_calc_rsq_pd(dx23,dy23,dz23);
1730             rsq31            = gmx_mm256_calc_rsq_pd(dx31,dy31,dz31);
1731             rsq32            = gmx_mm256_calc_rsq_pd(dx32,dy32,dz32);
1732             rsq33            = gmx_mm256_calc_rsq_pd(dx33,dy33,dz33);
1733
1734             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1735             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1736             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1737             rinv13           = gmx_mm256_invsqrt_pd(rsq13);
1738             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1739             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1740             rinv23           = gmx_mm256_invsqrt_pd(rsq23);
1741             rinv31           = gmx_mm256_invsqrt_pd(rsq31);
1742             rinv32           = gmx_mm256_invsqrt_pd(rsq32);
1743             rinv33           = gmx_mm256_invsqrt_pd(rsq33);
1744
1745             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1746             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1747             rinvsq13         = _mm256_mul_pd(rinv13,rinv13);
1748             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1749             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1750             rinvsq23         = _mm256_mul_pd(rinv23,rinv23);
1751             rinvsq31         = _mm256_mul_pd(rinv31,rinv31);
1752             rinvsq32         = _mm256_mul_pd(rinv32,rinv32);
1753             rinvsq33         = _mm256_mul_pd(rinv33,rinv33);
1754
1755             fjx0             = _mm256_setzero_pd();
1756             fjy0             = _mm256_setzero_pd();
1757             fjz0             = _mm256_setzero_pd();
1758             fjx1             = _mm256_setzero_pd();
1759             fjy1             = _mm256_setzero_pd();
1760             fjz1             = _mm256_setzero_pd();
1761             fjx2             = _mm256_setzero_pd();
1762             fjy2             = _mm256_setzero_pd();
1763             fjz2             = _mm256_setzero_pd();
1764             fjx3             = _mm256_setzero_pd();
1765             fjy3             = _mm256_setzero_pd();
1766             fjz3             = _mm256_setzero_pd();
1767
1768             /**************************
1769              * CALCULATE INTERACTIONS *
1770              **************************/
1771
1772             r00              = _mm256_mul_pd(rsq00,rinv00);
1773             r00              = _mm256_andnot_pd(dummy_mask,r00);
1774
1775             /* Calculate table index by multiplying r with table scale and truncate to integer */
1776             rt               = _mm256_mul_pd(r00,vftabscale);
1777             vfitab           = _mm256_cvttpd_epi32(rt);
1778             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1779             vfitab           = _mm_slli_epi32(vfitab,3);
1780
1781             /* CUBIC SPLINE TABLE DISPERSION */
1782             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1783             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1784             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1785             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1786             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1787             Heps             = _mm256_mul_pd(vfeps,H);
1788             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1789             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1790             fvdw6            = _mm256_mul_pd(c6_00,FF);
1791
1792             /* CUBIC SPLINE TABLE REPULSION */
1793             vfitab           = _mm_add_epi32(vfitab,ifour);
1794             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1795             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1796             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1797             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1798             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1799             Heps             = _mm256_mul_pd(vfeps,H);
1800             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1801             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1802             fvdw12           = _mm256_mul_pd(c12_00,FF);
1803             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1804
1805             fscal            = fvdw;
1806
1807             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1808
1809             /* Calculate temporary vectorial force */
1810             tx               = _mm256_mul_pd(fscal,dx00);
1811             ty               = _mm256_mul_pd(fscal,dy00);
1812             tz               = _mm256_mul_pd(fscal,dz00);
1813
1814             /* Update vectorial force */
1815             fix0             = _mm256_add_pd(fix0,tx);
1816             fiy0             = _mm256_add_pd(fiy0,ty);
1817             fiz0             = _mm256_add_pd(fiz0,tz);
1818
1819             fjx0             = _mm256_add_pd(fjx0,tx);
1820             fjy0             = _mm256_add_pd(fjy0,ty);
1821             fjz0             = _mm256_add_pd(fjz0,tz);
1822
1823             /**************************
1824              * CALCULATE INTERACTIONS *
1825              **************************/
1826
1827             /* COULOMB ELECTROSTATICS */
1828             velec            = _mm256_mul_pd(qq11,rinv11);
1829             felec            = _mm256_mul_pd(velec,rinvsq11);
1830
1831             fscal            = felec;
1832
1833             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1834
1835             /* Calculate temporary vectorial force */
1836             tx               = _mm256_mul_pd(fscal,dx11);
1837             ty               = _mm256_mul_pd(fscal,dy11);
1838             tz               = _mm256_mul_pd(fscal,dz11);
1839
1840             /* Update vectorial force */
1841             fix1             = _mm256_add_pd(fix1,tx);
1842             fiy1             = _mm256_add_pd(fiy1,ty);
1843             fiz1             = _mm256_add_pd(fiz1,tz);
1844
1845             fjx1             = _mm256_add_pd(fjx1,tx);
1846             fjy1             = _mm256_add_pd(fjy1,ty);
1847             fjz1             = _mm256_add_pd(fjz1,tz);
1848
1849             /**************************
1850              * CALCULATE INTERACTIONS *
1851              **************************/
1852
1853             /* COULOMB ELECTROSTATICS */
1854             velec            = _mm256_mul_pd(qq12,rinv12);
1855             felec            = _mm256_mul_pd(velec,rinvsq12);
1856
1857             fscal            = felec;
1858
1859             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1860
1861             /* Calculate temporary vectorial force */
1862             tx               = _mm256_mul_pd(fscal,dx12);
1863             ty               = _mm256_mul_pd(fscal,dy12);
1864             tz               = _mm256_mul_pd(fscal,dz12);
1865
1866             /* Update vectorial force */
1867             fix1             = _mm256_add_pd(fix1,tx);
1868             fiy1             = _mm256_add_pd(fiy1,ty);
1869             fiz1             = _mm256_add_pd(fiz1,tz);
1870
1871             fjx2             = _mm256_add_pd(fjx2,tx);
1872             fjy2             = _mm256_add_pd(fjy2,ty);
1873             fjz2             = _mm256_add_pd(fjz2,tz);
1874
1875             /**************************
1876              * CALCULATE INTERACTIONS *
1877              **************************/
1878
1879             /* COULOMB ELECTROSTATICS */
1880             velec            = _mm256_mul_pd(qq13,rinv13);
1881             felec            = _mm256_mul_pd(velec,rinvsq13);
1882
1883             fscal            = felec;
1884
1885             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1886
1887             /* Calculate temporary vectorial force */
1888             tx               = _mm256_mul_pd(fscal,dx13);
1889             ty               = _mm256_mul_pd(fscal,dy13);
1890             tz               = _mm256_mul_pd(fscal,dz13);
1891
1892             /* Update vectorial force */
1893             fix1             = _mm256_add_pd(fix1,tx);
1894             fiy1             = _mm256_add_pd(fiy1,ty);
1895             fiz1             = _mm256_add_pd(fiz1,tz);
1896
1897             fjx3             = _mm256_add_pd(fjx3,tx);
1898             fjy3             = _mm256_add_pd(fjy3,ty);
1899             fjz3             = _mm256_add_pd(fjz3,tz);
1900
1901             /**************************
1902              * CALCULATE INTERACTIONS *
1903              **************************/
1904
1905             /* COULOMB ELECTROSTATICS */
1906             velec            = _mm256_mul_pd(qq21,rinv21);
1907             felec            = _mm256_mul_pd(velec,rinvsq21);
1908
1909             fscal            = felec;
1910
1911             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1912
1913             /* Calculate temporary vectorial force */
1914             tx               = _mm256_mul_pd(fscal,dx21);
1915             ty               = _mm256_mul_pd(fscal,dy21);
1916             tz               = _mm256_mul_pd(fscal,dz21);
1917
1918             /* Update vectorial force */
1919             fix2             = _mm256_add_pd(fix2,tx);
1920             fiy2             = _mm256_add_pd(fiy2,ty);
1921             fiz2             = _mm256_add_pd(fiz2,tz);
1922
1923             fjx1             = _mm256_add_pd(fjx1,tx);
1924             fjy1             = _mm256_add_pd(fjy1,ty);
1925             fjz1             = _mm256_add_pd(fjz1,tz);
1926
1927             /**************************
1928              * CALCULATE INTERACTIONS *
1929              **************************/
1930
1931             /* COULOMB ELECTROSTATICS */
1932             velec            = _mm256_mul_pd(qq22,rinv22);
1933             felec            = _mm256_mul_pd(velec,rinvsq22);
1934
1935             fscal            = felec;
1936
1937             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1938
1939             /* Calculate temporary vectorial force */
1940             tx               = _mm256_mul_pd(fscal,dx22);
1941             ty               = _mm256_mul_pd(fscal,dy22);
1942             tz               = _mm256_mul_pd(fscal,dz22);
1943
1944             /* Update vectorial force */
1945             fix2             = _mm256_add_pd(fix2,tx);
1946             fiy2             = _mm256_add_pd(fiy2,ty);
1947             fiz2             = _mm256_add_pd(fiz2,tz);
1948
1949             fjx2             = _mm256_add_pd(fjx2,tx);
1950             fjy2             = _mm256_add_pd(fjy2,ty);
1951             fjz2             = _mm256_add_pd(fjz2,tz);
1952
1953             /**************************
1954              * CALCULATE INTERACTIONS *
1955              **************************/
1956
1957             /* COULOMB ELECTROSTATICS */
1958             velec            = _mm256_mul_pd(qq23,rinv23);
1959             felec            = _mm256_mul_pd(velec,rinvsq23);
1960
1961             fscal            = felec;
1962
1963             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1964
1965             /* Calculate temporary vectorial force */
1966             tx               = _mm256_mul_pd(fscal,dx23);
1967             ty               = _mm256_mul_pd(fscal,dy23);
1968             tz               = _mm256_mul_pd(fscal,dz23);
1969
1970             /* Update vectorial force */
1971             fix2             = _mm256_add_pd(fix2,tx);
1972             fiy2             = _mm256_add_pd(fiy2,ty);
1973             fiz2             = _mm256_add_pd(fiz2,tz);
1974
1975             fjx3             = _mm256_add_pd(fjx3,tx);
1976             fjy3             = _mm256_add_pd(fjy3,ty);
1977             fjz3             = _mm256_add_pd(fjz3,tz);
1978
1979             /**************************
1980              * CALCULATE INTERACTIONS *
1981              **************************/
1982
1983             /* COULOMB ELECTROSTATICS */
1984             velec            = _mm256_mul_pd(qq31,rinv31);
1985             felec            = _mm256_mul_pd(velec,rinvsq31);
1986
1987             fscal            = felec;
1988
1989             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1990
1991             /* Calculate temporary vectorial force */
1992             tx               = _mm256_mul_pd(fscal,dx31);
1993             ty               = _mm256_mul_pd(fscal,dy31);
1994             tz               = _mm256_mul_pd(fscal,dz31);
1995
1996             /* Update vectorial force */
1997             fix3             = _mm256_add_pd(fix3,tx);
1998             fiy3             = _mm256_add_pd(fiy3,ty);
1999             fiz3             = _mm256_add_pd(fiz3,tz);
2000
2001             fjx1             = _mm256_add_pd(fjx1,tx);
2002             fjy1             = _mm256_add_pd(fjy1,ty);
2003             fjz1             = _mm256_add_pd(fjz1,tz);
2004
2005             /**************************
2006              * CALCULATE INTERACTIONS *
2007              **************************/
2008
2009             /* COULOMB ELECTROSTATICS */
2010             velec            = _mm256_mul_pd(qq32,rinv32);
2011             felec            = _mm256_mul_pd(velec,rinvsq32);
2012
2013             fscal            = felec;
2014
2015             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2016
2017             /* Calculate temporary vectorial force */
2018             tx               = _mm256_mul_pd(fscal,dx32);
2019             ty               = _mm256_mul_pd(fscal,dy32);
2020             tz               = _mm256_mul_pd(fscal,dz32);
2021
2022             /* Update vectorial force */
2023             fix3             = _mm256_add_pd(fix3,tx);
2024             fiy3             = _mm256_add_pd(fiy3,ty);
2025             fiz3             = _mm256_add_pd(fiz3,tz);
2026
2027             fjx2             = _mm256_add_pd(fjx2,tx);
2028             fjy2             = _mm256_add_pd(fjy2,ty);
2029             fjz2             = _mm256_add_pd(fjz2,tz);
2030
2031             /**************************
2032              * CALCULATE INTERACTIONS *
2033              **************************/
2034
2035             /* COULOMB ELECTROSTATICS */
2036             velec            = _mm256_mul_pd(qq33,rinv33);
2037             felec            = _mm256_mul_pd(velec,rinvsq33);
2038
2039             fscal            = felec;
2040
2041             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
2042
2043             /* Calculate temporary vectorial force */
2044             tx               = _mm256_mul_pd(fscal,dx33);
2045             ty               = _mm256_mul_pd(fscal,dy33);
2046             tz               = _mm256_mul_pd(fscal,dz33);
2047
2048             /* Update vectorial force */
2049             fix3             = _mm256_add_pd(fix3,tx);
2050             fiy3             = _mm256_add_pd(fiy3,ty);
2051             fiz3             = _mm256_add_pd(fiz3,tz);
2052
2053             fjx3             = _mm256_add_pd(fjx3,tx);
2054             fjy3             = _mm256_add_pd(fjy3,ty);
2055             fjz3             = _mm256_add_pd(fjz3,tz);
2056
2057             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2058             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2059             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2060             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2061
2062             gmx_mm256_decrement_4rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
2063                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2064                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2065
2066             /* Inner loop uses 286 flops */
2067         }
2068
2069         /* End of innermost loop */
2070
2071         gmx_mm256_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2072                                                  f+i_coord_offset,fshift+i_shift_offset);
2073
2074         /* Increment number of inner iterations */
2075         inneriter                  += j_index_end - j_index_start;
2076
2077         /* Outer loop uses 24 flops */
2078     }
2079
2080     /* Increment number of outer iterations */
2081     outeriter        += nri;
2082
2083     /* Update outer/inner flops */
2084
2085     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*286);
2086 }