src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecRF_VdwCSTab_GeomW3W3_avx_256_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_double.h"
  50 #include "kernelutil_x86_avx_256_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW3W3_VF_avx_256_double
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRF_VdwCSTab_GeomW3W3_VF_avx_256_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  79     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  84     real             scratch[4*DIM];
  85     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  95     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  97     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 100     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 101     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 102     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 103     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 104     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 105     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 106     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 107     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 108     real             *charge;
 109     int              nvdwtype;
 110     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 111     int              *vdwtype;
 112     real             *vdwparam;
 113     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 114     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 115     __m128i          vfitab;
 116     __m128i          ifour       = _mm_set1_epi32(4);
 117     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 118     real             *vftab;
 119     __m256d          dummy_mask,cutoff_mask;
 120     __m128           tmpmask0,tmpmask1;
 121     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 122     __m256d          one     = _mm256_set1_pd(1.0);
 123     __m256d          two     = _mm256_set1_pd(2.0);
 124     x                = xx[0];
 125     f                = ff[0];
 126
 127     nri              = nlist->nri;
 128     iinr             = nlist->iinr;
 129     jindex           = nlist->jindex;
 130     jjnr             = nlist->jjnr;
 131     shiftidx         = nlist->shift;
 132     gid              = nlist->gid;
 133     shiftvec         = fr->shift_vec[0];
 134     fshift           = fr->fshift[0];
 135     facel            = _mm256_set1_pd(fr->epsfac);
 136     charge           = mdatoms->chargeA;
 137     krf              = _mm256_set1_pd(fr->ic->k_rf);
 138     krf2             = _mm256_set1_pd(fr->ic->k_rf*2.0);
 139     crf              = _mm256_set1_pd(fr->ic->c_rf);
 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     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 150     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 151     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 152     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 153
 154     jq0              = _mm256_set1_pd(charge[inr+0]);
 155     jq1              = _mm256_set1_pd(charge[inr+1]);
 156     jq2              = _mm256_set1_pd(charge[inr+2]);
 157     vdwjidx0A        = 2*vdwtype[inr+0];
 158     qq00             = _mm256_mul_pd(iq0,jq0);
 159     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
 160     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
 161     qq01             = _mm256_mul_pd(iq0,jq1);
 162     qq02             = _mm256_mul_pd(iq0,jq2);
 163     qq10             = _mm256_mul_pd(iq1,jq0);
 164     qq11             = _mm256_mul_pd(iq1,jq1);
 165     qq12             = _mm256_mul_pd(iq1,jq2);
 166     qq20             = _mm256_mul_pd(iq2,jq0);
 167     qq21             = _mm256_mul_pd(iq2,jq1);
 168     qq22             = _mm256_mul_pd(iq2,jq2);
 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_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 201                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 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
 213         /* Reset potential sums */
 214         velecsum         = _mm256_setzero_pd();
 215         vvdwsum          = _mm256_setzero_pd();
 216
 217         /* Start inner kernel loop */
 218         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 219         {
 220
 221             /* Get j neighbor index, and coordinate index */
 222             jnrA             = jjnr[jidx];
 223             jnrB             = jjnr[jidx+1];
 224             jnrC             = jjnr[jidx+2];
 225             jnrD             = jjnr[jidx+3];
 226             j_coord_offsetA  = DIM*jnrA;
 227             j_coord_offsetB  = DIM*jnrB;
 228             j_coord_offsetC  = DIM*jnrC;
 229             j_coord_offsetD  = DIM*jnrD;
 230
 231             /* load j atom coordinates */
 232             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 233                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 234                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 235
 236             /* Calculate displacement vector */
 237             dx00             = _mm256_sub_pd(ix0,jx0);
 238             dy00             = _mm256_sub_pd(iy0,jy0);
 239             dz00             = _mm256_sub_pd(iz0,jz0);
 240             dx01             = _mm256_sub_pd(ix0,jx1);
 241             dy01             = _mm256_sub_pd(iy0,jy1);
 242             dz01             = _mm256_sub_pd(iz0,jz1);
 243             dx02             = _mm256_sub_pd(ix0,jx2);
 244             dy02             = _mm256_sub_pd(iy0,jy2);
 245             dz02             = _mm256_sub_pd(iz0,jz2);
 246             dx10             = _mm256_sub_pd(ix1,jx0);
 247             dy10             = _mm256_sub_pd(iy1,jy0);
 248             dz10             = _mm256_sub_pd(iz1,jz0);
 249             dx11             = _mm256_sub_pd(ix1,jx1);
 250             dy11             = _mm256_sub_pd(iy1,jy1);
 251             dz11             = _mm256_sub_pd(iz1,jz1);
 252             dx12             = _mm256_sub_pd(ix1,jx2);
 253             dy12             = _mm256_sub_pd(iy1,jy2);
 254             dz12             = _mm256_sub_pd(iz1,jz2);
 255             dx20             = _mm256_sub_pd(ix2,jx0);
 256             dy20             = _mm256_sub_pd(iy2,jy0);
 257             dz20             = _mm256_sub_pd(iz2,jz0);
 258             dx21             = _mm256_sub_pd(ix2,jx1);
 259             dy21             = _mm256_sub_pd(iy2,jy1);
 260             dz21             = _mm256_sub_pd(iz2,jz1);
 261             dx22             = _mm256_sub_pd(ix2,jx2);
 262             dy22             = _mm256_sub_pd(iy2,jy2);
 263             dz22             = _mm256_sub_pd(iz2,jz2);
 264
 265             /* Calculate squared distance and things based on it */
 266             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 267             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 268             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 269             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 270             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 271             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 272             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 273             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 274             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 275
 276             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 277             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 278             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 279             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 280             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 281             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 282             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 283             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 284             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 285
 286             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 287             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 288             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 289             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 290             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 291             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 292             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 293             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 294             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 295
 296             fjx0             = _mm256_setzero_pd();
 297             fjy0             = _mm256_setzero_pd();
 298             fjz0             = _mm256_setzero_pd();
 299             fjx1             = _mm256_setzero_pd();
 300             fjy1             = _mm256_setzero_pd();
 301             fjz1             = _mm256_setzero_pd();
 302             fjx2             = _mm256_setzero_pd();
 303             fjy2             = _mm256_setzero_pd();
 304             fjz2             = _mm256_setzero_pd();
 305
 306             /**************************
 307              * CALCULATE INTERACTIONS *
 308              **************************/
 309
 310             r00              = _mm256_mul_pd(rsq00,rinv00);
 311
 312             /* Calculate table index by multiplying r with table scale and truncate to integer */
 313             rt               = _mm256_mul_pd(r00,vftabscale);
 314             vfitab           = _mm256_cvttpd_epi32(rt);
 315             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 316             vfitab           = _mm_slli_epi32(vfitab,3);
 317
 318             /* REACTION-FIELD ELECTROSTATICS */
 319             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
 320             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
 321
 322             /* CUBIC SPLINE TABLE DISPERSION */
 323             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 324             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 325             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 326             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 327             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 328             Heps             = _mm256_mul_pd(vfeps,H);
 329             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 330             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 331             vvdw6            = _mm256_mul_pd(c6_00,VV);
 332             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 333             fvdw6            = _mm256_mul_pd(c6_00,FF);
 334
 335             /* CUBIC SPLINE TABLE REPULSION */
 336             vfitab           = _mm_add_epi32(vfitab,ifour);
 337             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 338             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 339             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 340             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 341             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 342             Heps             = _mm256_mul_pd(vfeps,H);
 343             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 344             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 345             vvdw12           = _mm256_mul_pd(c12_00,VV);
 346             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 347             fvdw12           = _mm256_mul_pd(c12_00,FF);
 348             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 349             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 350
 351             /* Update potential sum for this i atom from the interaction with this j atom. */
 352             velecsum         = _mm256_add_pd(velecsum,velec);
 353             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 354
 355             fscal            = _mm256_add_pd(felec,fvdw);
 356
 357             /* Calculate temporary vectorial force */
 358             tx               = _mm256_mul_pd(fscal,dx00);
 359             ty               = _mm256_mul_pd(fscal,dy00);
 360             tz               = _mm256_mul_pd(fscal,dz00);
 361
 362             /* Update vectorial force */
 363             fix0             = _mm256_add_pd(fix0,tx);
 364             fiy0             = _mm256_add_pd(fiy0,ty);
 365             fiz0             = _mm256_add_pd(fiz0,tz);
 366
 367             fjx0             = _mm256_add_pd(fjx0,tx);
 368             fjy0             = _mm256_add_pd(fjy0,ty);
 369             fjz0             = _mm256_add_pd(fjz0,tz);
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             /* REACTION-FIELD ELECTROSTATICS */
 376             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
 377             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
 378
 379             /* Update potential sum for this i atom from the interaction with this j atom. */
 380             velecsum         = _mm256_add_pd(velecsum,velec);
 381
 382             fscal            = felec;
 383
 384             /* Calculate temporary vectorial force */
 385             tx               = _mm256_mul_pd(fscal,dx01);
 386             ty               = _mm256_mul_pd(fscal,dy01);
 387             tz               = _mm256_mul_pd(fscal,dz01);
 388
 389             /* Update vectorial force */
 390             fix0             = _mm256_add_pd(fix0,tx);
 391             fiy0             = _mm256_add_pd(fiy0,ty);
 392             fiz0             = _mm256_add_pd(fiz0,tz);
 393
 394             fjx1             = _mm256_add_pd(fjx1,tx);
 395             fjy1             = _mm256_add_pd(fjy1,ty);
 396             fjz1             = _mm256_add_pd(fjz1,tz);
 397
 398             /**************************
 399              * CALCULATE INTERACTIONS *
 400              **************************/
 401
 402             /* REACTION-FIELD ELECTROSTATICS */
 403             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
 404             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
 405
 406             /* Update potential sum for this i atom from the interaction with this j atom. */
 407             velecsum         = _mm256_add_pd(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             /* Calculate temporary vectorial force */
 412             tx               = _mm256_mul_pd(fscal,dx02);
 413             ty               = _mm256_mul_pd(fscal,dy02);
 414             tz               = _mm256_mul_pd(fscal,dz02);
 415
 416             /* Update vectorial force */
 417             fix0             = _mm256_add_pd(fix0,tx);
 418             fiy0             = _mm256_add_pd(fiy0,ty);
 419             fiz0             = _mm256_add_pd(fiz0,tz);
 420
 421             fjx2             = _mm256_add_pd(fjx2,tx);
 422             fjy2             = _mm256_add_pd(fjy2,ty);
 423             fjz2             = _mm256_add_pd(fjz2,tz);
 424
 425             /**************************
 426              * CALCULATE INTERACTIONS *
 427              **************************/
 428
 429             /* REACTION-FIELD ELECTROSTATICS */
 430             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
 431             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
 432
 433             /* Update potential sum for this i atom from the interaction with this j atom. */
 434             velecsum         = _mm256_add_pd(velecsum,velec);
 435
 436             fscal            = felec;
 437
 438             /* Calculate temporary vectorial force */
 439             tx               = _mm256_mul_pd(fscal,dx10);
 440             ty               = _mm256_mul_pd(fscal,dy10);
 441             tz               = _mm256_mul_pd(fscal,dz10);
 442
 443             /* Update vectorial force */
 444             fix1             = _mm256_add_pd(fix1,tx);
 445             fiy1             = _mm256_add_pd(fiy1,ty);
 446             fiz1             = _mm256_add_pd(fiz1,tz);
 447
 448             fjx0             = _mm256_add_pd(fjx0,tx);
 449             fjy0             = _mm256_add_pd(fjy0,ty);
 450             fjz0             = _mm256_add_pd(fjz0,tz);
 451
 452             /**************************
 453              * CALCULATE INTERACTIONS *
 454              **************************/
 455
 456             /* REACTION-FIELD ELECTROSTATICS */
 457             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
 458             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
 459
 460             /* Update potential sum for this i atom from the interaction with this j atom. */
 461             velecsum         = _mm256_add_pd(velecsum,velec);
 462
 463             fscal            = felec;
 464
 465             /* Calculate temporary vectorial force */
 466             tx               = _mm256_mul_pd(fscal,dx11);
 467             ty               = _mm256_mul_pd(fscal,dy11);
 468             tz               = _mm256_mul_pd(fscal,dz11);
 469
 470             /* Update vectorial force */
 471             fix1             = _mm256_add_pd(fix1,tx);
 472             fiy1             = _mm256_add_pd(fiy1,ty);
 473             fiz1             = _mm256_add_pd(fiz1,tz);
 474
 475             fjx1             = _mm256_add_pd(fjx1,tx);
 476             fjy1             = _mm256_add_pd(fjy1,ty);
 477             fjz1             = _mm256_add_pd(fjz1,tz);
 478
 479             /**************************
 480              * CALCULATE INTERACTIONS *
 481              **************************/
 482
 483             /* REACTION-FIELD ELECTROSTATICS */
 484             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
 485             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
 486
 487             /* Update potential sum for this i atom from the interaction with this j atom. */
 488             velecsum         = _mm256_add_pd(velecsum,velec);
 489
 490             fscal            = felec;
 491
 492             /* Calculate temporary vectorial force */
 493             tx               = _mm256_mul_pd(fscal,dx12);
 494             ty               = _mm256_mul_pd(fscal,dy12);
 495             tz               = _mm256_mul_pd(fscal,dz12);
 496
 497             /* Update vectorial force */
 498             fix1             = _mm256_add_pd(fix1,tx);
 499             fiy1             = _mm256_add_pd(fiy1,ty);
 500             fiz1             = _mm256_add_pd(fiz1,tz);
 501
 502             fjx2             = _mm256_add_pd(fjx2,tx);
 503             fjy2             = _mm256_add_pd(fjy2,ty);
 504             fjz2             = _mm256_add_pd(fjz2,tz);
 505
 506             /**************************
 507              * CALCULATE INTERACTIONS *
 508              **************************/
 509
 510             /* REACTION-FIELD ELECTROSTATICS */
 511             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
 512             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
 513
 514             /* Update potential sum for this i atom from the interaction with this j atom. */
 515             velecsum         = _mm256_add_pd(velecsum,velec);
 516
 517             fscal            = felec;
 518
 519             /* Calculate temporary vectorial force */
 520             tx               = _mm256_mul_pd(fscal,dx20);
 521             ty               = _mm256_mul_pd(fscal,dy20);
 522             tz               = _mm256_mul_pd(fscal,dz20);
 523
 524             /* Update vectorial force */
 525             fix2             = _mm256_add_pd(fix2,tx);
 526             fiy2             = _mm256_add_pd(fiy2,ty);
 527             fiz2             = _mm256_add_pd(fiz2,tz);
 528
 529             fjx0             = _mm256_add_pd(fjx0,tx);
 530             fjy0             = _mm256_add_pd(fjy0,ty);
 531             fjz0             = _mm256_add_pd(fjz0,tz);
 532
 533             /**************************
 534              * CALCULATE INTERACTIONS *
 535              **************************/
 536
 537             /* REACTION-FIELD ELECTROSTATICS */
 538             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
 539             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
 540
 541             /* Update potential sum for this i atom from the interaction with this j atom. */
 542             velecsum         = _mm256_add_pd(velecsum,velec);
 543
 544             fscal            = felec;
 545
 546             /* Calculate temporary vectorial force */
 547             tx               = _mm256_mul_pd(fscal,dx21);
 548             ty               = _mm256_mul_pd(fscal,dy21);
 549             tz               = _mm256_mul_pd(fscal,dz21);
 550
 551             /* Update vectorial force */
 552             fix2             = _mm256_add_pd(fix2,tx);
 553             fiy2             = _mm256_add_pd(fiy2,ty);
 554             fiz2             = _mm256_add_pd(fiz2,tz);
 555
 556             fjx1             = _mm256_add_pd(fjx1,tx);
 557             fjy1             = _mm256_add_pd(fjy1,ty);
 558             fjz1             = _mm256_add_pd(fjz1,tz);
 559
 560             /**************************
 561              * CALCULATE INTERACTIONS *
 562              **************************/
 563
 564             /* REACTION-FIELD ELECTROSTATICS */
 565             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
 566             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
 567
 568             /* Update potential sum for this i atom from the interaction with this j atom. */
 569             velecsum         = _mm256_add_pd(velecsum,velec);
 570
 571             fscal            = felec;
 572
 573             /* Calculate temporary vectorial force */
 574             tx               = _mm256_mul_pd(fscal,dx22);
 575             ty               = _mm256_mul_pd(fscal,dy22);
 576             tz               = _mm256_mul_pd(fscal,dz22);
 577
 578             /* Update vectorial force */
 579             fix2             = _mm256_add_pd(fix2,tx);
 580             fiy2             = _mm256_add_pd(fiy2,ty);
 581             fiz2             = _mm256_add_pd(fiz2,tz);
 582
 583             fjx2             = _mm256_add_pd(fjx2,tx);
 584             fjy2             = _mm256_add_pd(fjy2,ty);
 585             fjz2             = _mm256_add_pd(fjz2,tz);
 586
 587             fjptrA             = f+j_coord_offsetA;
 588             fjptrB             = f+j_coord_offsetB;
 589             fjptrC             = f+j_coord_offsetC;
 590             fjptrD             = f+j_coord_offsetD;
 591
 592             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
 593                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 594
 595             /* Inner loop uses 323 flops */
 596         }
 597
 598         if(jidx<j_index_end)
 599         {
 600
 601             /* Get j neighbor index, and coordinate index */
 602             jnrlistA         = jjnr[jidx];
 603             jnrlistB         = jjnr[jidx+1];
 604             jnrlistC         = jjnr[jidx+2];
 605             jnrlistD         = jjnr[jidx+3];
 606             /* Sign of each element will be negative for non-real atoms.
 607              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 608              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 609              */
 610             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 611
 612             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 613             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 614             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 615
 616             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 617             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 618             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 619             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 620             j_coord_offsetA  = DIM*jnrA;
 621             j_coord_offsetB  = DIM*jnrB;
 622             j_coord_offsetC  = DIM*jnrC;
 623             j_coord_offsetD  = DIM*jnrD;
 624
 625             /* load j atom coordinates */
 626             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 627                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 628                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 629
 630             /* Calculate displacement vector */
 631             dx00             = _mm256_sub_pd(ix0,jx0);
 632             dy00             = _mm256_sub_pd(iy0,jy0);
 633             dz00             = _mm256_sub_pd(iz0,jz0);
 634             dx01             = _mm256_sub_pd(ix0,jx1);
 635             dy01             = _mm256_sub_pd(iy0,jy1);
 636             dz01             = _mm256_sub_pd(iz0,jz1);
 637             dx02             = _mm256_sub_pd(ix0,jx2);
 638             dy02             = _mm256_sub_pd(iy0,jy2);
 639             dz02             = _mm256_sub_pd(iz0,jz2);
 640             dx10             = _mm256_sub_pd(ix1,jx0);
 641             dy10             = _mm256_sub_pd(iy1,jy0);
 642             dz10             = _mm256_sub_pd(iz1,jz0);
 643             dx11             = _mm256_sub_pd(ix1,jx1);
 644             dy11             = _mm256_sub_pd(iy1,jy1);
 645             dz11             = _mm256_sub_pd(iz1,jz1);
 646             dx12             = _mm256_sub_pd(ix1,jx2);
 647             dy12             = _mm256_sub_pd(iy1,jy2);
 648             dz12             = _mm256_sub_pd(iz1,jz2);
 649             dx20             = _mm256_sub_pd(ix2,jx0);
 650             dy20             = _mm256_sub_pd(iy2,jy0);
 651             dz20             = _mm256_sub_pd(iz2,jz0);
 652             dx21             = _mm256_sub_pd(ix2,jx1);
 653             dy21             = _mm256_sub_pd(iy2,jy1);
 654             dz21             = _mm256_sub_pd(iz2,jz1);
 655             dx22             = _mm256_sub_pd(ix2,jx2);
 656             dy22             = _mm256_sub_pd(iy2,jy2);
 657             dz22             = _mm256_sub_pd(iz2,jz2);
 658
 659             /* Calculate squared distance and things based on it */
 660             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 661             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
 662             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
 663             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 664             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
 665             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
 666             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 667             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
 668             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
 669
 670             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 671             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
 672             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
 673             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 674             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
 675             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
 676             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 677             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
 678             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
 679
 680             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 681             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
 682             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
 683             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 684             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
 685             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
 686             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 687             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
 688             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
 689
 690             fjx0             = _mm256_setzero_pd();
 691             fjy0             = _mm256_setzero_pd();
 692             fjz0             = _mm256_setzero_pd();
 693             fjx1             = _mm256_setzero_pd();
 694             fjy1             = _mm256_setzero_pd();
 695             fjz1             = _mm256_setzero_pd();
 696             fjx2             = _mm256_setzero_pd();
 697             fjy2             = _mm256_setzero_pd();
 698             fjz2             = _mm256_setzero_pd();
 699
 700             /**************************
 701              * CALCULATE INTERACTIONS *
 702              **************************/
 703
 704             r00              = _mm256_mul_pd(rsq00,rinv00);
 705             r00              = _mm256_andnot_pd(dummy_mask,r00);
 706
 707             /* Calculate table index by multiplying r with table scale and truncate to integer */
 708             rt               = _mm256_mul_pd(r00,vftabscale);
 709             vfitab           = _mm256_cvttpd_epi32(rt);
 710             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
 711             vfitab           = _mm_slli_epi32(vfitab,3);
 712
 713             /* REACTION-FIELD ELECTROSTATICS */
 714             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_add_pd(rinv00,_mm256_mul_pd(krf,rsq00)),crf));
 715             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
 716
 717             /* CUBIC SPLINE TABLE DISPERSION */
 718             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 719             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 720             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 721             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 722             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 723             Heps             = _mm256_mul_pd(vfeps,H);
 724             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 725             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 726             vvdw6            = _mm256_mul_pd(c6_00,VV);
 727             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 728             fvdw6            = _mm256_mul_pd(c6_00,FF);
 729
 730             /* CUBIC SPLINE TABLE REPULSION */
 731             vfitab           = _mm_add_epi32(vfitab,ifour);
 732             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 733             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 734             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
 735             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
 736             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
 737             Heps             = _mm256_mul_pd(vfeps,H);
 738             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
 739             VV               = _mm256_add_pd(Y,_mm256_mul_pd(vfeps,Fp));
 740             vvdw12           = _mm256_mul_pd(c12_00,VV);
 741             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
 742             fvdw12           = _mm256_mul_pd(c12_00,FF);
 743             vvdw             = _mm256_add_pd(vvdw12,vvdw6);
 744             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
 745
 746             /* Update potential sum for this i atom from the interaction with this j atom. */
 747             velec            = _mm256_andnot_pd(dummy_mask,velec);
 748             velecsum         = _mm256_add_pd(velecsum,velec);
 749             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 750             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 751
 752             fscal            = _mm256_add_pd(felec,fvdw);
 753
 754             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 755
 756             /* Calculate temporary vectorial force */
 757             tx               = _mm256_mul_pd(fscal,dx00);
 758             ty               = _mm256_mul_pd(fscal,dy00);
 759             tz               = _mm256_mul_pd(fscal,dz00);
 760
 761             /* Update vectorial force */
 762             fix0             = _mm256_add_pd(fix0,tx);
 763             fiy0             = _mm256_add_pd(fiy0,ty);
 764             fiz0             = _mm256_add_pd(fiz0,tz);
 765
 766             fjx0             = _mm256_add_pd(fjx0,tx);
 767             fjy0             = _mm256_add_pd(fjy0,ty);
 768             fjz0             = _mm256_add_pd(fjz0,tz);
 769
 770             /**************************
 771              * CALCULATE INTERACTIONS *
 772              **************************/
 773
 774             /* REACTION-FIELD ELECTROSTATICS */
 775             velec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_add_pd(rinv01,_mm256_mul_pd(krf,rsq01)),crf));
 776             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
 777
 778             /* Update potential sum for this i atom from the interaction with this j atom. */
 779             velec            = _mm256_andnot_pd(dummy_mask,velec);
 780             velecsum         = _mm256_add_pd(velecsum,velec);
 781
 782             fscal            = felec;
 783
 784             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 785
 786             /* Calculate temporary vectorial force */
 787             tx               = _mm256_mul_pd(fscal,dx01);
 788             ty               = _mm256_mul_pd(fscal,dy01);
 789             tz               = _mm256_mul_pd(fscal,dz01);
 790
 791             /* Update vectorial force */
 792             fix0             = _mm256_add_pd(fix0,tx);
 793             fiy0             = _mm256_add_pd(fiy0,ty);
 794             fiz0             = _mm256_add_pd(fiz0,tz);
 795
 796             fjx1             = _mm256_add_pd(fjx1,tx);
 797             fjy1             = _mm256_add_pd(fjy1,ty);
 798             fjz1             = _mm256_add_pd(fjz1,tz);
 799
 800             /**************************
 801              * CALCULATE INTERACTIONS *
 802              **************************/
 803
 804             /* REACTION-FIELD ELECTROSTATICS */
 805             velec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_add_pd(rinv02,_mm256_mul_pd(krf,rsq02)),crf));
 806             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
 807
 808             /* Update potential sum for this i atom from the interaction with this j atom. */
 809             velec            = _mm256_andnot_pd(dummy_mask,velec);
 810             velecsum         = _mm256_add_pd(velecsum,velec);
 811
 812             fscal            = felec;
 813
 814             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 815
 816             /* Calculate temporary vectorial force */
 817             tx               = _mm256_mul_pd(fscal,dx02);
 818             ty               = _mm256_mul_pd(fscal,dy02);
 819             tz               = _mm256_mul_pd(fscal,dz02);
 820
 821             /* Update vectorial force */
 822             fix0             = _mm256_add_pd(fix0,tx);
 823             fiy0             = _mm256_add_pd(fiy0,ty);
 824             fiz0             = _mm256_add_pd(fiz0,tz);
 825
 826             fjx2             = _mm256_add_pd(fjx2,tx);
 827             fjy2             = _mm256_add_pd(fjy2,ty);
 828             fjz2             = _mm256_add_pd(fjz2,tz);
 829
 830             /**************************
 831              * CALCULATE INTERACTIONS *
 832              **************************/
 833
 834             /* REACTION-FIELD ELECTROSTATICS */
 835             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_add_pd(rinv10,_mm256_mul_pd(krf,rsq10)),crf));
 836             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
 837
 838             /* Update potential sum for this i atom from the interaction with this j atom. */
 839             velec            = _mm256_andnot_pd(dummy_mask,velec);
 840             velecsum         = _mm256_add_pd(velecsum,velec);
 841
 842             fscal            = felec;
 843
 844             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 845
 846             /* Calculate temporary vectorial force */
 847             tx               = _mm256_mul_pd(fscal,dx10);
 848             ty               = _mm256_mul_pd(fscal,dy10);
 849             tz               = _mm256_mul_pd(fscal,dz10);
 850
 851             /* Update vectorial force */
 852             fix1             = _mm256_add_pd(fix1,tx);
 853             fiy1             = _mm256_add_pd(fiy1,ty);
 854             fiz1             = _mm256_add_pd(fiz1,tz);
 855
 856             fjx0             = _mm256_add_pd(fjx0,tx);
 857             fjy0             = _mm256_add_pd(fjy0,ty);
 858             fjz0             = _mm256_add_pd(fjz0,tz);
 859
 860             /**************************
 861              * CALCULATE INTERACTIONS *
 862              **************************/
 863
 864             /* REACTION-FIELD ELECTROSTATICS */
 865             velec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_add_pd(rinv11,_mm256_mul_pd(krf,rsq11)),crf));
 866             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
 867
 868             /* Update potential sum for this i atom from the interaction with this j atom. */
 869             velec            = _mm256_andnot_pd(dummy_mask,velec);
 870             velecsum         = _mm256_add_pd(velecsum,velec);
 871
 872             fscal            = felec;
 873
 874             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 875
 876             /* Calculate temporary vectorial force */
 877             tx               = _mm256_mul_pd(fscal,dx11);
 878             ty               = _mm256_mul_pd(fscal,dy11);
 879             tz               = _mm256_mul_pd(fscal,dz11);
 880
 881             /* Update vectorial force */
 882             fix1             = _mm256_add_pd(fix1,tx);
 883             fiy1             = _mm256_add_pd(fiy1,ty);
 884             fiz1             = _mm256_add_pd(fiz1,tz);
 885
 886             fjx1             = _mm256_add_pd(fjx1,tx);
 887             fjy1             = _mm256_add_pd(fjy1,ty);
 888             fjz1             = _mm256_add_pd(fjz1,tz);
 889
 890             /**************************
 891              * CALCULATE INTERACTIONS *
 892              **************************/
 893
 894             /* REACTION-FIELD ELECTROSTATICS */
 895             velec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_add_pd(rinv12,_mm256_mul_pd(krf,rsq12)),crf));
 896             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
 897
 898             /* Update potential sum for this i atom from the interaction with this j atom. */
 899             velec            = _mm256_andnot_pd(dummy_mask,velec);
 900             velecsum         = _mm256_add_pd(velecsum,velec);
 901
 902             fscal            = felec;
 903
 904             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 905
 906             /* Calculate temporary vectorial force */
 907             tx               = _mm256_mul_pd(fscal,dx12);
 908             ty               = _mm256_mul_pd(fscal,dy12);
 909             tz               = _mm256_mul_pd(fscal,dz12);
 910
 911             /* Update vectorial force */
 912             fix1             = _mm256_add_pd(fix1,tx);
 913             fiy1             = _mm256_add_pd(fiy1,ty);
 914             fiz1             = _mm256_add_pd(fiz1,tz);
 915
 916             fjx2             = _mm256_add_pd(fjx2,tx);
 917             fjy2             = _mm256_add_pd(fjy2,ty);
 918             fjz2             = _mm256_add_pd(fjz2,tz);
 919
 920             /**************************
 921              * CALCULATE INTERACTIONS *
 922              **************************/
 923
 924             /* REACTION-FIELD ELECTROSTATICS */
 925             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_add_pd(rinv20,_mm256_mul_pd(krf,rsq20)),crf));
 926             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
 927
 928             /* Update potential sum for this i atom from the interaction with this j atom. */
 929             velec            = _mm256_andnot_pd(dummy_mask,velec);
 930             velecsum         = _mm256_add_pd(velecsum,velec);
 931
 932             fscal            = felec;
 933
 934             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 935
 936             /* Calculate temporary vectorial force */
 937             tx               = _mm256_mul_pd(fscal,dx20);
 938             ty               = _mm256_mul_pd(fscal,dy20);
 939             tz               = _mm256_mul_pd(fscal,dz20);
 940
 941             /* Update vectorial force */
 942             fix2             = _mm256_add_pd(fix2,tx);
 943             fiy2             = _mm256_add_pd(fiy2,ty);
 944             fiz2             = _mm256_add_pd(fiz2,tz);
 945
 946             fjx0             = _mm256_add_pd(fjx0,tx);
 947             fjy0             = _mm256_add_pd(fjy0,ty);
 948             fjz0             = _mm256_add_pd(fjz0,tz);
 949
 950             /**************************
 951              * CALCULATE INTERACTIONS *
 952              **************************/
 953
 954             /* REACTION-FIELD ELECTROSTATICS */
 955             velec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_add_pd(rinv21,_mm256_mul_pd(krf,rsq21)),crf));
 956             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
 957
 958             /* Update potential sum for this i atom from the interaction with this j atom. */
 959             velec            = _mm256_andnot_pd(dummy_mask,velec);
 960             velecsum         = _mm256_add_pd(velecsum,velec);
 961
 962             fscal            = felec;
 963
 964             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 965
 966             /* Calculate temporary vectorial force */
 967             tx               = _mm256_mul_pd(fscal,dx21);
 968             ty               = _mm256_mul_pd(fscal,dy21);
 969             tz               = _mm256_mul_pd(fscal,dz21);
 970
 971             /* Update vectorial force */
 972             fix2             = _mm256_add_pd(fix2,tx);
 973             fiy2             = _mm256_add_pd(fiy2,ty);
 974             fiz2             = _mm256_add_pd(fiz2,tz);
 975
 976             fjx1             = _mm256_add_pd(fjx1,tx);
 977             fjy1             = _mm256_add_pd(fjy1,ty);
 978             fjz1             = _mm256_add_pd(fjz1,tz);
 979
 980             /**************************
 981              * CALCULATE INTERACTIONS *
 982              **************************/
 983
 984             /* REACTION-FIELD ELECTROSTATICS */
 985             velec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_add_pd(rinv22,_mm256_mul_pd(krf,rsq22)),crf));
 986             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
 987
 988             /* Update potential sum for this i atom from the interaction with this j atom. */
 989             velec            = _mm256_andnot_pd(dummy_mask,velec);
 990             velecsum         = _mm256_add_pd(velecsum,velec);
 991
 992             fscal            = felec;
 993
 994             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 995
 996             /* Calculate temporary vectorial force */
 997             tx               = _mm256_mul_pd(fscal,dx22);
 998             ty               = _mm256_mul_pd(fscal,dy22);
 999             tz               = _mm256_mul_pd(fscal,dz22);
1000
1001             /* Update vectorial force */
1002             fix2             = _mm256_add_pd(fix2,tx);
1003             fiy2             = _mm256_add_pd(fiy2,ty);
1004             fiz2             = _mm256_add_pd(fiz2,tz);
1005
1006             fjx2             = _mm256_add_pd(fjx2,tx);
1007             fjy2             = _mm256_add_pd(fjy2,ty);
1008             fjz2             = _mm256_add_pd(fjz2,tz);
1009
1010             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1011             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1012             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1013             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1014
1015             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1016                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1017
1018             /* Inner loop uses 324 flops */
1019         }
1020
1021         /* End of innermost loop */
1022
1023         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1024                                                  f+i_coord_offset,fshift+i_shift_offset);
1025
1026         ggid                        = gid[iidx];
1027         /* Update potential energies */
1028         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1029         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1030
1031         /* Increment number of inner iterations */
1032         inneriter                  += j_index_end - j_index_start;
1033
1034         /* Outer loop uses 20 flops */
1035     }
1036
1037     /* Increment number of outer iterations */
1038     outeriter        += nri;
1039
1040     /* Update outer/inner flops */
1041
1042     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*324);
1043 }
1044 /*
1045  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW3W3_F_avx_256_double
1046  * Electrostatics interaction: ReactionField
1047  * VdW interaction:            CubicSplineTable
1048  * Geometry:                   Water3-Water3
1049  * Calculate force/pot:        Force
1050  */
1051 void
1052 nb_kernel_ElecRF_VdwCSTab_GeomW3W3_F_avx_256_double
1053                     (t_nblist                    * gmx_restrict       nlist,
1054                      rvec                        * gmx_restrict          xx,
1055                      rvec                        * gmx_restrict          ff,
1056                      t_forcerec                  * gmx_restrict          fr,
1057                      t_mdatoms                   * gmx_restrict     mdatoms,
1058                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1059                      t_nrnb                      * gmx_restrict        nrnb)
1060 {
1061     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1062      * just 0 for non-waters.
1063      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
1064      * jnr indices corresponding to data put in the four positions in the SIMD register.
1065      */
1066     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1067     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1068     int              jnrA,jnrB,jnrC,jnrD;
1069     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1070     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1071     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1072     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1073     real             rcutoff_scalar;
1074     real             *shiftvec,*fshift,*x,*f;
1075     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1076     real             scratch[4*DIM];
1077     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1078     real *           vdwioffsetptr0;
1079     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1080     real *           vdwioffsetptr1;
1081     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1082     real *           vdwioffsetptr2;
1083     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1084     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1085     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1086     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1087     __m256d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1088     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1089     __m256d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1090     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1091     __m256d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1092     __m256d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1093     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1094     __m256d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1095     __m256d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1096     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1097     __m256d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1098     __m256d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1099     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
1100     real             *charge;
1101     int              nvdwtype;
1102     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1103     int              *vdwtype;
1104     real             *vdwparam;
1105     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
1106     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
1107     __m128i          vfitab;
1108     __m128i          ifour       = _mm_set1_epi32(4);
1109     __m256d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1110     real             *vftab;
1111     __m256d          dummy_mask,cutoff_mask;
1112     __m128           tmpmask0,tmpmask1;
1113     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
1114     __m256d          one     = _mm256_set1_pd(1.0);
1115     __m256d          two     = _mm256_set1_pd(2.0);
1116     x                = xx[0];
1117     f                = ff[0];
1118
1119     nri              = nlist->nri;
1120     iinr             = nlist->iinr;
1121     jindex           = nlist->jindex;
1122     jjnr             = nlist->jjnr;
1123     shiftidx         = nlist->shift;
1124     gid              = nlist->gid;
1125     shiftvec         = fr->shift_vec[0];
1126     fshift           = fr->fshift[0];
1127     facel            = _mm256_set1_pd(fr->epsfac);
1128     charge           = mdatoms->chargeA;
1129     krf              = _mm256_set1_pd(fr->ic->k_rf);
1130     krf2             = _mm256_set1_pd(fr->ic->k_rf*2.0);
1131     crf              = _mm256_set1_pd(fr->ic->c_rf);
1132     nvdwtype         = fr->ntype;
1133     vdwparam         = fr->nbfp;
1134     vdwtype          = mdatoms->typeA;
1135
1136     vftab            = kernel_data->table_vdw->data;
1137     vftabscale       = _mm256_set1_pd(kernel_data->table_vdw->scale);
1138
1139     /* Setup water-specific parameters */
1140     inr              = nlist->iinr[0];
1141     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
1142     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
1143     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
1144     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1145
1146     jq0              = _mm256_set1_pd(charge[inr+0]);
1147     jq1              = _mm256_set1_pd(charge[inr+1]);
1148     jq2              = _mm256_set1_pd(charge[inr+2]);
1149     vdwjidx0A        = 2*vdwtype[inr+0];
1150     qq00             = _mm256_mul_pd(iq0,jq0);
1151     c6_00            = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A]);
1152     c12_00           = _mm256_set1_pd(vdwioffsetptr0[vdwjidx0A+1]);
1153     qq01             = _mm256_mul_pd(iq0,jq1);
1154     qq02             = _mm256_mul_pd(iq0,jq2);
1155     qq10             = _mm256_mul_pd(iq1,jq0);
1156     qq11             = _mm256_mul_pd(iq1,jq1);
1157     qq12             = _mm256_mul_pd(iq1,jq2);
1158     qq20             = _mm256_mul_pd(iq2,jq0);
1159     qq21             = _mm256_mul_pd(iq2,jq1);
1160     qq22             = _mm256_mul_pd(iq2,jq2);
1161
1162     /* Avoid stupid compiler warnings */
1163     jnrA = jnrB = jnrC = jnrD = 0;
1164     j_coord_offsetA = 0;
1165     j_coord_offsetB = 0;
1166     j_coord_offsetC = 0;
1167     j_coord_offsetD = 0;
1168
1169     outeriter        = 0;
1170     inneriter        = 0;
1171
1172     for(iidx=0;iidx<4*DIM;iidx++)
1173     {
1174         scratch[iidx] = 0.0;
1175     }
1176
1177     /* Start outer loop over neighborlists */
1178     for(iidx=0; iidx<nri; iidx++)
1179     {
1180         /* Load shift vector for this list */
1181         i_shift_offset   = DIM*shiftidx[iidx];
1182
1183         /* Load limits for loop over neighbors */
1184         j_index_start    = jindex[iidx];
1185         j_index_end      = jindex[iidx+1];
1186
1187         /* Get outer coordinate index */
1188         inr              = iinr[iidx];
1189         i_coord_offset   = DIM*inr;
1190
1191         /* Load i particle coords and add shift vector */
1192         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1193                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1194
1195         fix0             = _mm256_setzero_pd();
1196         fiy0             = _mm256_setzero_pd();
1197         fiz0             = _mm256_setzero_pd();
1198         fix1             = _mm256_setzero_pd();
1199         fiy1             = _mm256_setzero_pd();
1200         fiz1             = _mm256_setzero_pd();
1201         fix2             = _mm256_setzero_pd();
1202         fiy2             = _mm256_setzero_pd();
1203         fiz2             = _mm256_setzero_pd();
1204
1205         /* Start inner kernel loop */
1206         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1207         {
1208
1209             /* Get j neighbor index, and coordinate index */
1210             jnrA             = jjnr[jidx];
1211             jnrB             = jjnr[jidx+1];
1212             jnrC             = jjnr[jidx+2];
1213             jnrD             = jjnr[jidx+3];
1214             j_coord_offsetA  = DIM*jnrA;
1215             j_coord_offsetB  = DIM*jnrB;
1216             j_coord_offsetC  = DIM*jnrC;
1217             j_coord_offsetD  = DIM*jnrD;
1218
1219             /* load j atom coordinates */
1220             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1221                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1222                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1223
1224             /* Calculate displacement vector */
1225             dx00             = _mm256_sub_pd(ix0,jx0);
1226             dy00             = _mm256_sub_pd(iy0,jy0);
1227             dz00             = _mm256_sub_pd(iz0,jz0);
1228             dx01             = _mm256_sub_pd(ix0,jx1);
1229             dy01             = _mm256_sub_pd(iy0,jy1);
1230             dz01             = _mm256_sub_pd(iz0,jz1);
1231             dx02             = _mm256_sub_pd(ix0,jx2);
1232             dy02             = _mm256_sub_pd(iy0,jy2);
1233             dz02             = _mm256_sub_pd(iz0,jz2);
1234             dx10             = _mm256_sub_pd(ix1,jx0);
1235             dy10             = _mm256_sub_pd(iy1,jy0);
1236             dz10             = _mm256_sub_pd(iz1,jz0);
1237             dx11             = _mm256_sub_pd(ix1,jx1);
1238             dy11             = _mm256_sub_pd(iy1,jy1);
1239             dz11             = _mm256_sub_pd(iz1,jz1);
1240             dx12             = _mm256_sub_pd(ix1,jx2);
1241             dy12             = _mm256_sub_pd(iy1,jy2);
1242             dz12             = _mm256_sub_pd(iz1,jz2);
1243             dx20             = _mm256_sub_pd(ix2,jx0);
1244             dy20             = _mm256_sub_pd(iy2,jy0);
1245             dz20             = _mm256_sub_pd(iz2,jz0);
1246             dx21             = _mm256_sub_pd(ix2,jx1);
1247             dy21             = _mm256_sub_pd(iy2,jy1);
1248             dz21             = _mm256_sub_pd(iz2,jz1);
1249             dx22             = _mm256_sub_pd(ix2,jx2);
1250             dy22             = _mm256_sub_pd(iy2,jy2);
1251             dz22             = _mm256_sub_pd(iz2,jz2);
1252
1253             /* Calculate squared distance and things based on it */
1254             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1255             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1256             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1257             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1258             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1259             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1260             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1261             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1262             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1263
1264             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1265             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1266             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1267             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1268             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1269             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1270             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1271             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1272             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1273
1274             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1275             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1276             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1277             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1278             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1279             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1280             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1281             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1282             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1283
1284             fjx0             = _mm256_setzero_pd();
1285             fjy0             = _mm256_setzero_pd();
1286             fjz0             = _mm256_setzero_pd();
1287             fjx1             = _mm256_setzero_pd();
1288             fjy1             = _mm256_setzero_pd();
1289             fjz1             = _mm256_setzero_pd();
1290             fjx2             = _mm256_setzero_pd();
1291             fjy2             = _mm256_setzero_pd();
1292             fjz2             = _mm256_setzero_pd();
1293
1294             /**************************
1295              * CALCULATE INTERACTIONS *
1296              **************************/
1297
1298             r00              = _mm256_mul_pd(rsq00,rinv00);
1299
1300             /* Calculate table index by multiplying r with table scale and truncate to integer */
1301             rt               = _mm256_mul_pd(r00,vftabscale);
1302             vfitab           = _mm256_cvttpd_epi32(rt);
1303             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1304             vfitab           = _mm_slli_epi32(vfitab,3);
1305
1306             /* REACTION-FIELD ELECTROSTATICS */
1307             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1308
1309             /* CUBIC SPLINE TABLE DISPERSION */
1310             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1311             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1312             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1313             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1314             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1315             Heps             = _mm256_mul_pd(vfeps,H);
1316             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1317             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1318             fvdw6            = _mm256_mul_pd(c6_00,FF);
1319
1320             /* CUBIC SPLINE TABLE REPULSION */
1321             vfitab           = _mm_add_epi32(vfitab,ifour);
1322             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1323             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1324             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1325             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1326             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1327             Heps             = _mm256_mul_pd(vfeps,H);
1328             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1329             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1330             fvdw12           = _mm256_mul_pd(c12_00,FF);
1331             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1332
1333             fscal            = _mm256_add_pd(felec,fvdw);
1334
1335             /* Calculate temporary vectorial force */
1336             tx               = _mm256_mul_pd(fscal,dx00);
1337             ty               = _mm256_mul_pd(fscal,dy00);
1338             tz               = _mm256_mul_pd(fscal,dz00);
1339
1340             /* Update vectorial force */
1341             fix0             = _mm256_add_pd(fix0,tx);
1342             fiy0             = _mm256_add_pd(fiy0,ty);
1343             fiz0             = _mm256_add_pd(fiz0,tz);
1344
1345             fjx0             = _mm256_add_pd(fjx0,tx);
1346             fjy0             = _mm256_add_pd(fjy0,ty);
1347             fjz0             = _mm256_add_pd(fjz0,tz);
1348
1349             /**************************
1350              * CALCULATE INTERACTIONS *
1351              **************************/
1352
1353             /* REACTION-FIELD ELECTROSTATICS */
1354             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1355
1356             fscal            = felec;
1357
1358             /* Calculate temporary vectorial force */
1359             tx               = _mm256_mul_pd(fscal,dx01);
1360             ty               = _mm256_mul_pd(fscal,dy01);
1361             tz               = _mm256_mul_pd(fscal,dz01);
1362
1363             /* Update vectorial force */
1364             fix0             = _mm256_add_pd(fix0,tx);
1365             fiy0             = _mm256_add_pd(fiy0,ty);
1366             fiz0             = _mm256_add_pd(fiz0,tz);
1367
1368             fjx1             = _mm256_add_pd(fjx1,tx);
1369             fjy1             = _mm256_add_pd(fjy1,ty);
1370             fjz1             = _mm256_add_pd(fjz1,tz);
1371
1372             /**************************
1373              * CALCULATE INTERACTIONS *
1374              **************************/
1375
1376             /* REACTION-FIELD ELECTROSTATICS */
1377             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1378
1379             fscal            = felec;
1380
1381             /* Calculate temporary vectorial force */
1382             tx               = _mm256_mul_pd(fscal,dx02);
1383             ty               = _mm256_mul_pd(fscal,dy02);
1384             tz               = _mm256_mul_pd(fscal,dz02);
1385
1386             /* Update vectorial force */
1387             fix0             = _mm256_add_pd(fix0,tx);
1388             fiy0             = _mm256_add_pd(fiy0,ty);
1389             fiz0             = _mm256_add_pd(fiz0,tz);
1390
1391             fjx2             = _mm256_add_pd(fjx2,tx);
1392             fjy2             = _mm256_add_pd(fjy2,ty);
1393             fjz2             = _mm256_add_pd(fjz2,tz);
1394
1395             /**************************
1396              * CALCULATE INTERACTIONS *
1397              **************************/
1398
1399             /* REACTION-FIELD ELECTROSTATICS */
1400             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1401
1402             fscal            = felec;
1403
1404             /* Calculate temporary vectorial force */
1405             tx               = _mm256_mul_pd(fscal,dx10);
1406             ty               = _mm256_mul_pd(fscal,dy10);
1407             tz               = _mm256_mul_pd(fscal,dz10);
1408
1409             /* Update vectorial force */
1410             fix1             = _mm256_add_pd(fix1,tx);
1411             fiy1             = _mm256_add_pd(fiy1,ty);
1412             fiz1             = _mm256_add_pd(fiz1,tz);
1413
1414             fjx0             = _mm256_add_pd(fjx0,tx);
1415             fjy0             = _mm256_add_pd(fjy0,ty);
1416             fjz0             = _mm256_add_pd(fjz0,tz);
1417
1418             /**************************
1419              * CALCULATE INTERACTIONS *
1420              **************************/
1421
1422             /* REACTION-FIELD ELECTROSTATICS */
1423             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1424
1425             fscal            = felec;
1426
1427             /* Calculate temporary vectorial force */
1428             tx               = _mm256_mul_pd(fscal,dx11);
1429             ty               = _mm256_mul_pd(fscal,dy11);
1430             tz               = _mm256_mul_pd(fscal,dz11);
1431
1432             /* Update vectorial force */
1433             fix1             = _mm256_add_pd(fix1,tx);
1434             fiy1             = _mm256_add_pd(fiy1,ty);
1435             fiz1             = _mm256_add_pd(fiz1,tz);
1436
1437             fjx1             = _mm256_add_pd(fjx1,tx);
1438             fjy1             = _mm256_add_pd(fjy1,ty);
1439             fjz1             = _mm256_add_pd(fjz1,tz);
1440
1441             /**************************
1442              * CALCULATE INTERACTIONS *
1443              **************************/
1444
1445             /* REACTION-FIELD ELECTROSTATICS */
1446             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1447
1448             fscal            = felec;
1449
1450             /* Calculate temporary vectorial force */
1451             tx               = _mm256_mul_pd(fscal,dx12);
1452             ty               = _mm256_mul_pd(fscal,dy12);
1453             tz               = _mm256_mul_pd(fscal,dz12);
1454
1455             /* Update vectorial force */
1456             fix1             = _mm256_add_pd(fix1,tx);
1457             fiy1             = _mm256_add_pd(fiy1,ty);
1458             fiz1             = _mm256_add_pd(fiz1,tz);
1459
1460             fjx2             = _mm256_add_pd(fjx2,tx);
1461             fjy2             = _mm256_add_pd(fjy2,ty);
1462             fjz2             = _mm256_add_pd(fjz2,tz);
1463
1464             /**************************
1465              * CALCULATE INTERACTIONS *
1466              **************************/
1467
1468             /* REACTION-FIELD ELECTROSTATICS */
1469             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1470
1471             fscal            = felec;
1472
1473             /* Calculate temporary vectorial force */
1474             tx               = _mm256_mul_pd(fscal,dx20);
1475             ty               = _mm256_mul_pd(fscal,dy20);
1476             tz               = _mm256_mul_pd(fscal,dz20);
1477
1478             /* Update vectorial force */
1479             fix2             = _mm256_add_pd(fix2,tx);
1480             fiy2             = _mm256_add_pd(fiy2,ty);
1481             fiz2             = _mm256_add_pd(fiz2,tz);
1482
1483             fjx0             = _mm256_add_pd(fjx0,tx);
1484             fjy0             = _mm256_add_pd(fjy0,ty);
1485             fjz0             = _mm256_add_pd(fjz0,tz);
1486
1487             /**************************
1488              * CALCULATE INTERACTIONS *
1489              **************************/
1490
1491             /* REACTION-FIELD ELECTROSTATICS */
1492             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1493
1494             fscal            = felec;
1495
1496             /* Calculate temporary vectorial force */
1497             tx               = _mm256_mul_pd(fscal,dx21);
1498             ty               = _mm256_mul_pd(fscal,dy21);
1499             tz               = _mm256_mul_pd(fscal,dz21);
1500
1501             /* Update vectorial force */
1502             fix2             = _mm256_add_pd(fix2,tx);
1503             fiy2             = _mm256_add_pd(fiy2,ty);
1504             fiz2             = _mm256_add_pd(fiz2,tz);
1505
1506             fjx1             = _mm256_add_pd(fjx1,tx);
1507             fjy1             = _mm256_add_pd(fjy1,ty);
1508             fjz1             = _mm256_add_pd(fjz1,tz);
1509
1510             /**************************
1511              * CALCULATE INTERACTIONS *
1512              **************************/
1513
1514             /* REACTION-FIELD ELECTROSTATICS */
1515             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1516
1517             fscal            = felec;
1518
1519             /* Calculate temporary vectorial force */
1520             tx               = _mm256_mul_pd(fscal,dx22);
1521             ty               = _mm256_mul_pd(fscal,dy22);
1522             tz               = _mm256_mul_pd(fscal,dz22);
1523
1524             /* Update vectorial force */
1525             fix2             = _mm256_add_pd(fix2,tx);
1526             fiy2             = _mm256_add_pd(fiy2,ty);
1527             fiz2             = _mm256_add_pd(fiz2,tz);
1528
1529             fjx2             = _mm256_add_pd(fjx2,tx);
1530             fjy2             = _mm256_add_pd(fjy2,ty);
1531             fjz2             = _mm256_add_pd(fjz2,tz);
1532
1533             fjptrA             = f+j_coord_offsetA;
1534             fjptrB             = f+j_coord_offsetB;
1535             fjptrC             = f+j_coord_offsetC;
1536             fjptrD             = f+j_coord_offsetD;
1537
1538             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1539                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1540
1541             /* Inner loop uses 270 flops */
1542         }
1543
1544         if(jidx<j_index_end)
1545         {
1546
1547             /* Get j neighbor index, and coordinate index */
1548             jnrlistA         = jjnr[jidx];
1549             jnrlistB         = jjnr[jidx+1];
1550             jnrlistC         = jjnr[jidx+2];
1551             jnrlistD         = jjnr[jidx+3];
1552             /* Sign of each element will be negative for non-real atoms.
1553              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1554              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1555              */
1556             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1557
1558             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1559             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1560             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1561
1562             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1563             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1564             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1565             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1566             j_coord_offsetA  = DIM*jnrA;
1567             j_coord_offsetB  = DIM*jnrB;
1568             j_coord_offsetC  = DIM*jnrC;
1569             j_coord_offsetD  = DIM*jnrD;
1570
1571             /* load j atom coordinates */
1572             gmx_mm256_load_3rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1573                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1574                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1575
1576             /* Calculate displacement vector */
1577             dx00             = _mm256_sub_pd(ix0,jx0);
1578             dy00             = _mm256_sub_pd(iy0,jy0);
1579             dz00             = _mm256_sub_pd(iz0,jz0);
1580             dx01             = _mm256_sub_pd(ix0,jx1);
1581             dy01             = _mm256_sub_pd(iy0,jy1);
1582             dz01             = _mm256_sub_pd(iz0,jz1);
1583             dx02             = _mm256_sub_pd(ix0,jx2);
1584             dy02             = _mm256_sub_pd(iy0,jy2);
1585             dz02             = _mm256_sub_pd(iz0,jz2);
1586             dx10             = _mm256_sub_pd(ix1,jx0);
1587             dy10             = _mm256_sub_pd(iy1,jy0);
1588             dz10             = _mm256_sub_pd(iz1,jz0);
1589             dx11             = _mm256_sub_pd(ix1,jx1);
1590             dy11             = _mm256_sub_pd(iy1,jy1);
1591             dz11             = _mm256_sub_pd(iz1,jz1);
1592             dx12             = _mm256_sub_pd(ix1,jx2);
1593             dy12             = _mm256_sub_pd(iy1,jy2);
1594             dz12             = _mm256_sub_pd(iz1,jz2);
1595             dx20             = _mm256_sub_pd(ix2,jx0);
1596             dy20             = _mm256_sub_pd(iy2,jy0);
1597             dz20             = _mm256_sub_pd(iz2,jz0);
1598             dx21             = _mm256_sub_pd(ix2,jx1);
1599             dy21             = _mm256_sub_pd(iy2,jy1);
1600             dz21             = _mm256_sub_pd(iz2,jz1);
1601             dx22             = _mm256_sub_pd(ix2,jx2);
1602             dy22             = _mm256_sub_pd(iy2,jy2);
1603             dz22             = _mm256_sub_pd(iz2,jz2);
1604
1605             /* Calculate squared distance and things based on it */
1606             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1607             rsq01            = gmx_mm256_calc_rsq_pd(dx01,dy01,dz01);
1608             rsq02            = gmx_mm256_calc_rsq_pd(dx02,dy02,dz02);
1609             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1610             rsq11            = gmx_mm256_calc_rsq_pd(dx11,dy11,dz11);
1611             rsq12            = gmx_mm256_calc_rsq_pd(dx12,dy12,dz12);
1612             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1613             rsq21            = gmx_mm256_calc_rsq_pd(dx21,dy21,dz21);
1614             rsq22            = gmx_mm256_calc_rsq_pd(dx22,dy22,dz22);
1615
1616             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1617             rinv01           = gmx_mm256_invsqrt_pd(rsq01);
1618             rinv02           = gmx_mm256_invsqrt_pd(rsq02);
1619             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1620             rinv11           = gmx_mm256_invsqrt_pd(rsq11);
1621             rinv12           = gmx_mm256_invsqrt_pd(rsq12);
1622             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1623             rinv21           = gmx_mm256_invsqrt_pd(rsq21);
1624             rinv22           = gmx_mm256_invsqrt_pd(rsq22);
1625
1626             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1627             rinvsq01         = _mm256_mul_pd(rinv01,rinv01);
1628             rinvsq02         = _mm256_mul_pd(rinv02,rinv02);
1629             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1630             rinvsq11         = _mm256_mul_pd(rinv11,rinv11);
1631             rinvsq12         = _mm256_mul_pd(rinv12,rinv12);
1632             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1633             rinvsq21         = _mm256_mul_pd(rinv21,rinv21);
1634             rinvsq22         = _mm256_mul_pd(rinv22,rinv22);
1635
1636             fjx0             = _mm256_setzero_pd();
1637             fjy0             = _mm256_setzero_pd();
1638             fjz0             = _mm256_setzero_pd();
1639             fjx1             = _mm256_setzero_pd();
1640             fjy1             = _mm256_setzero_pd();
1641             fjz1             = _mm256_setzero_pd();
1642             fjx2             = _mm256_setzero_pd();
1643             fjy2             = _mm256_setzero_pd();
1644             fjz2             = _mm256_setzero_pd();
1645
1646             /**************************
1647              * CALCULATE INTERACTIONS *
1648              **************************/
1649
1650             r00              = _mm256_mul_pd(rsq00,rinv00);
1651             r00              = _mm256_andnot_pd(dummy_mask,r00);
1652
1653             /* Calculate table index by multiplying r with table scale and truncate to integer */
1654             rt               = _mm256_mul_pd(r00,vftabscale);
1655             vfitab           = _mm256_cvttpd_epi32(rt);
1656             vfeps            = _mm256_sub_pd(rt,_mm256_round_pd(rt, _MM_FROUND_FLOOR));
1657             vfitab           = _mm_slli_epi32(vfitab,3);
1658
1659             /* REACTION-FIELD ELECTROSTATICS */
1660             felec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_mul_pd(rinv00,rinvsq00),krf2));
1661
1662             /* CUBIC SPLINE TABLE DISPERSION */
1663             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1664             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1665             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1666             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1667             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1668             Heps             = _mm256_mul_pd(vfeps,H);
1669             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1670             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1671             fvdw6            = _mm256_mul_pd(c6_00,FF);
1672
1673             /* CUBIC SPLINE TABLE REPULSION */
1674             vfitab           = _mm_add_epi32(vfitab,ifour);
1675             Y                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1676             F                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1677             G                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,2) );
1678             H                = _mm256_load_pd( vftab + _mm_extract_epi32(vfitab,3) );
1679             GMX_MM256_FULLTRANSPOSE4_PD(Y,F,G,H);
1680             Heps             = _mm256_mul_pd(vfeps,H);
1681             Fp               = _mm256_add_pd(F,_mm256_mul_pd(vfeps,_mm256_add_pd(G,Heps)));
1682             FF               = _mm256_add_pd(Fp,_mm256_mul_pd(vfeps,_mm256_add_pd(G,_mm256_add_pd(Heps,Heps))));
1683             fvdw12           = _mm256_mul_pd(c12_00,FF);
1684             fvdw             = _mm256_xor_pd(signbit,_mm256_mul_pd(_mm256_add_pd(fvdw6,fvdw12),_mm256_mul_pd(vftabscale,rinv00)));
1685
1686             fscal            = _mm256_add_pd(felec,fvdw);
1687
1688             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1689
1690             /* Calculate temporary vectorial force */
1691             tx               = _mm256_mul_pd(fscal,dx00);
1692             ty               = _mm256_mul_pd(fscal,dy00);
1693             tz               = _mm256_mul_pd(fscal,dz00);
1694
1695             /* Update vectorial force */
1696             fix0             = _mm256_add_pd(fix0,tx);
1697             fiy0             = _mm256_add_pd(fiy0,ty);
1698             fiz0             = _mm256_add_pd(fiz0,tz);
1699
1700             fjx0             = _mm256_add_pd(fjx0,tx);
1701             fjy0             = _mm256_add_pd(fjy0,ty);
1702             fjz0             = _mm256_add_pd(fjz0,tz);
1703
1704             /**************************
1705              * CALCULATE INTERACTIONS *
1706              **************************/
1707
1708             /* REACTION-FIELD ELECTROSTATICS */
1709             felec            = _mm256_mul_pd(qq01,_mm256_sub_pd(_mm256_mul_pd(rinv01,rinvsq01),krf2));
1710
1711             fscal            = felec;
1712
1713             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1714
1715             /* Calculate temporary vectorial force */
1716             tx               = _mm256_mul_pd(fscal,dx01);
1717             ty               = _mm256_mul_pd(fscal,dy01);
1718             tz               = _mm256_mul_pd(fscal,dz01);
1719
1720             /* Update vectorial force */
1721             fix0             = _mm256_add_pd(fix0,tx);
1722             fiy0             = _mm256_add_pd(fiy0,ty);
1723             fiz0             = _mm256_add_pd(fiz0,tz);
1724
1725             fjx1             = _mm256_add_pd(fjx1,tx);
1726             fjy1             = _mm256_add_pd(fjy1,ty);
1727             fjz1             = _mm256_add_pd(fjz1,tz);
1728
1729             /**************************
1730              * CALCULATE INTERACTIONS *
1731              **************************/
1732
1733             /* REACTION-FIELD ELECTROSTATICS */
1734             felec            = _mm256_mul_pd(qq02,_mm256_sub_pd(_mm256_mul_pd(rinv02,rinvsq02),krf2));
1735
1736             fscal            = felec;
1737
1738             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1739
1740             /* Calculate temporary vectorial force */
1741             tx               = _mm256_mul_pd(fscal,dx02);
1742             ty               = _mm256_mul_pd(fscal,dy02);
1743             tz               = _mm256_mul_pd(fscal,dz02);
1744
1745             /* Update vectorial force */
1746             fix0             = _mm256_add_pd(fix0,tx);
1747             fiy0             = _mm256_add_pd(fiy0,ty);
1748             fiz0             = _mm256_add_pd(fiz0,tz);
1749
1750             fjx2             = _mm256_add_pd(fjx2,tx);
1751             fjy2             = _mm256_add_pd(fjy2,ty);
1752             fjz2             = _mm256_add_pd(fjz2,tz);
1753
1754             /**************************
1755              * CALCULATE INTERACTIONS *
1756              **************************/
1757
1758             /* REACTION-FIELD ELECTROSTATICS */
1759             felec            = _mm256_mul_pd(qq10,_mm256_sub_pd(_mm256_mul_pd(rinv10,rinvsq10),krf2));
1760
1761             fscal            = felec;
1762
1763             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1764
1765             /* Calculate temporary vectorial force */
1766             tx               = _mm256_mul_pd(fscal,dx10);
1767             ty               = _mm256_mul_pd(fscal,dy10);
1768             tz               = _mm256_mul_pd(fscal,dz10);
1769
1770             /* Update vectorial force */
1771             fix1             = _mm256_add_pd(fix1,tx);
1772             fiy1             = _mm256_add_pd(fiy1,ty);
1773             fiz1             = _mm256_add_pd(fiz1,tz);
1774
1775             fjx0             = _mm256_add_pd(fjx0,tx);
1776             fjy0             = _mm256_add_pd(fjy0,ty);
1777             fjz0             = _mm256_add_pd(fjz0,tz);
1778
1779             /**************************
1780              * CALCULATE INTERACTIONS *
1781              **************************/
1782
1783             /* REACTION-FIELD ELECTROSTATICS */
1784             felec            = _mm256_mul_pd(qq11,_mm256_sub_pd(_mm256_mul_pd(rinv11,rinvsq11),krf2));
1785
1786             fscal            = felec;
1787
1788             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1789
1790             /* Calculate temporary vectorial force */
1791             tx               = _mm256_mul_pd(fscal,dx11);
1792             ty               = _mm256_mul_pd(fscal,dy11);
1793             tz               = _mm256_mul_pd(fscal,dz11);
1794
1795             /* Update vectorial force */
1796             fix1             = _mm256_add_pd(fix1,tx);
1797             fiy1             = _mm256_add_pd(fiy1,ty);
1798             fiz1             = _mm256_add_pd(fiz1,tz);
1799
1800             fjx1             = _mm256_add_pd(fjx1,tx);
1801             fjy1             = _mm256_add_pd(fjy1,ty);
1802             fjz1             = _mm256_add_pd(fjz1,tz);
1803
1804             /**************************
1805              * CALCULATE INTERACTIONS *
1806              **************************/
1807
1808             /* REACTION-FIELD ELECTROSTATICS */
1809             felec            = _mm256_mul_pd(qq12,_mm256_sub_pd(_mm256_mul_pd(rinv12,rinvsq12),krf2));
1810
1811             fscal            = felec;
1812
1813             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1814
1815             /* Calculate temporary vectorial force */
1816             tx               = _mm256_mul_pd(fscal,dx12);
1817             ty               = _mm256_mul_pd(fscal,dy12);
1818             tz               = _mm256_mul_pd(fscal,dz12);
1819
1820             /* Update vectorial force */
1821             fix1             = _mm256_add_pd(fix1,tx);
1822             fiy1             = _mm256_add_pd(fiy1,ty);
1823             fiz1             = _mm256_add_pd(fiz1,tz);
1824
1825             fjx2             = _mm256_add_pd(fjx2,tx);
1826             fjy2             = _mm256_add_pd(fjy2,ty);
1827             fjz2             = _mm256_add_pd(fjz2,tz);
1828
1829             /**************************
1830              * CALCULATE INTERACTIONS *
1831              **************************/
1832
1833             /* REACTION-FIELD ELECTROSTATICS */
1834             felec            = _mm256_mul_pd(qq20,_mm256_sub_pd(_mm256_mul_pd(rinv20,rinvsq20),krf2));
1835
1836             fscal            = felec;
1837
1838             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1839
1840             /* Calculate temporary vectorial force */
1841             tx               = _mm256_mul_pd(fscal,dx20);
1842             ty               = _mm256_mul_pd(fscal,dy20);
1843             tz               = _mm256_mul_pd(fscal,dz20);
1844
1845             /* Update vectorial force */
1846             fix2             = _mm256_add_pd(fix2,tx);
1847             fiy2             = _mm256_add_pd(fiy2,ty);
1848             fiz2             = _mm256_add_pd(fiz2,tz);
1849
1850             fjx0             = _mm256_add_pd(fjx0,tx);
1851             fjy0             = _mm256_add_pd(fjy0,ty);
1852             fjz0             = _mm256_add_pd(fjz0,tz);
1853
1854             /**************************
1855              * CALCULATE INTERACTIONS *
1856              **************************/
1857
1858             /* REACTION-FIELD ELECTROSTATICS */
1859             felec            = _mm256_mul_pd(qq21,_mm256_sub_pd(_mm256_mul_pd(rinv21,rinvsq21),krf2));
1860
1861             fscal            = felec;
1862
1863             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1864
1865             /* Calculate temporary vectorial force */
1866             tx               = _mm256_mul_pd(fscal,dx21);
1867             ty               = _mm256_mul_pd(fscal,dy21);
1868             tz               = _mm256_mul_pd(fscal,dz21);
1869
1870             /* Update vectorial force */
1871             fix2             = _mm256_add_pd(fix2,tx);
1872             fiy2             = _mm256_add_pd(fiy2,ty);
1873             fiz2             = _mm256_add_pd(fiz2,tz);
1874
1875             fjx1             = _mm256_add_pd(fjx1,tx);
1876             fjy1             = _mm256_add_pd(fjy1,ty);
1877             fjz1             = _mm256_add_pd(fjz1,tz);
1878
1879             /**************************
1880              * CALCULATE INTERACTIONS *
1881              **************************/
1882
1883             /* REACTION-FIELD ELECTROSTATICS */
1884             felec            = _mm256_mul_pd(qq22,_mm256_sub_pd(_mm256_mul_pd(rinv22,rinvsq22),krf2));
1885
1886             fscal            = felec;
1887
1888             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1889
1890             /* Calculate temporary vectorial force */
1891             tx               = _mm256_mul_pd(fscal,dx22);
1892             ty               = _mm256_mul_pd(fscal,dy22);
1893             tz               = _mm256_mul_pd(fscal,dz22);
1894
1895             /* Update vectorial force */
1896             fix2             = _mm256_add_pd(fix2,tx);
1897             fiy2             = _mm256_add_pd(fiy2,ty);
1898             fiz2             = _mm256_add_pd(fiz2,tz);
1899
1900             fjx2             = _mm256_add_pd(fjx2,tx);
1901             fjy2             = _mm256_add_pd(fjy2,ty);
1902             fjz2             = _mm256_add_pd(fjz2,tz);
1903
1904             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1905             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1906             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1907             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1908
1909             gmx_mm256_decrement_3rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,
1910                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1911
1912             /* Inner loop uses 271 flops */
1913         }
1914
1915         /* End of innermost loop */
1916
1917         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1918                                                  f+i_coord_offset,fshift+i_shift_offset);
1919
1920         /* Increment number of inner iterations */
1921         inneriter                  += j_index_end - j_index_start;
1922
1923         /* Outer loop uses 18 flops */
1924     }
1925
1926     /* Increment number of outer iterations */
1927     outeriter        += nri;
1928
1929     /* Update outer/inner flops */
1930
1931     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*271);
1932 }