src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_avx_256_single.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_single 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 "gmx_math_x86_avx_256_single.h"
  50 #include "kernelutil_x86_avx_256_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_VF_avx_256_single
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water3-Particle
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_VF_avx_256_single
  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,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight 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              jnrE,jnrF,jnrG,jnrH;
  78     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  79     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  80     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  81     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  82     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  83     real             rcutoff_scalar;
  84     real             *shiftvec,*fshift,*x,*f;
  85     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  86     real             scratch[4*DIM];
  87     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  88     real *           vdwioffsetptr0;
  89     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  90     real *           vdwioffsetptr1;
  91     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  92     real *           vdwioffsetptr2;
  93     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  97     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  98     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  99     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 100     real             *charge;
 101     int              nvdwtype;
 102     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 103     int              *vdwtype;
 104     real             *vdwparam;
 105     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 106     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 107     __m256i          vfitab;
 108     __m128i          vfitab_lo,vfitab_hi;
 109     __m128i          ifour       = _mm_set1_epi32(4);
 110     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 111     real             *vftab;
 112     __m256           dummy_mask,cutoff_mask;
 113     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 114     __m256           one     = _mm256_set1_ps(1.0);
 115     __m256           two     = _mm256_set1_ps(2.0);
 116     x                = xx[0];
 117     f                = ff[0];
 118
 119     nri              = nlist->nri;
 120     iinr             = nlist->iinr;
 121     jindex           = nlist->jindex;
 122     jjnr             = nlist->jjnr;
 123     shiftidx         = nlist->shift;
 124     gid              = nlist->gid;
 125     shiftvec         = fr->shift_vec[0];
 126     fshift           = fr->fshift[0];
 127     facel            = _mm256_set1_ps(fr->epsfac);
 128     charge           = mdatoms->chargeA;
 129     krf              = _mm256_set1_ps(fr->ic->k_rf);
 130     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 131     crf              = _mm256_set1_ps(fr->ic->c_rf);
 132     nvdwtype         = fr->ntype;
 133     vdwparam         = fr->nbfp;
 134     vdwtype          = mdatoms->typeA;
 135
 136     vftab            = kernel_data->table_vdw->data;
 137     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
 138
 139     /* Setup water-specific parameters */
 140     inr              = nlist->iinr[0];
 141     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 142     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 143     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 144     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 145
 146     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 147     rcutoff_scalar   = fr->rcoulomb;
 148     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 149     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 150
 151     /* Avoid stupid compiler warnings */
 152     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 153     j_coord_offsetA = 0;
 154     j_coord_offsetB = 0;
 155     j_coord_offsetC = 0;
 156     j_coord_offsetD = 0;
 157     j_coord_offsetE = 0;
 158     j_coord_offsetF = 0;
 159     j_coord_offsetG = 0;
 160     j_coord_offsetH = 0;
 161
 162     outeriter        = 0;
 163     inneriter        = 0;
 164
 165     for(iidx=0;iidx<4*DIM;iidx++)
 166     {
 167         scratch[iidx] = 0.0;
 168     }
 169
 170     /* Start outer loop over neighborlists */
 171     for(iidx=0; iidx<nri; iidx++)
 172     {
 173         /* Load shift vector for this list */
 174         i_shift_offset   = DIM*shiftidx[iidx];
 175
 176         /* Load limits for loop over neighbors */
 177         j_index_start    = jindex[iidx];
 178         j_index_end      = jindex[iidx+1];
 179
 180         /* Get outer coordinate index */
 181         inr              = iinr[iidx];
 182         i_coord_offset   = DIM*inr;
 183
 184         /* Load i particle coords and add shift vector */
 185         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 186                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 187
 188         fix0             = _mm256_setzero_ps();
 189         fiy0             = _mm256_setzero_ps();
 190         fiz0             = _mm256_setzero_ps();
 191         fix1             = _mm256_setzero_ps();
 192         fiy1             = _mm256_setzero_ps();
 193         fiz1             = _mm256_setzero_ps();
 194         fix2             = _mm256_setzero_ps();
 195         fiy2             = _mm256_setzero_ps();
 196         fiz2             = _mm256_setzero_ps();
 197
 198         /* Reset potential sums */
 199         velecsum         = _mm256_setzero_ps();
 200         vvdwsum          = _mm256_setzero_ps();
 201
 202         /* Start inner kernel loop */
 203         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 204         {
 205
 206             /* Get j neighbor index, and coordinate index */
 207             jnrA             = jjnr[jidx];
 208             jnrB             = jjnr[jidx+1];
 209             jnrC             = jjnr[jidx+2];
 210             jnrD             = jjnr[jidx+3];
 211             jnrE             = jjnr[jidx+4];
 212             jnrF             = jjnr[jidx+5];
 213             jnrG             = jjnr[jidx+6];
 214             jnrH             = jjnr[jidx+7];
 215             j_coord_offsetA  = DIM*jnrA;
 216             j_coord_offsetB  = DIM*jnrB;
 217             j_coord_offsetC  = DIM*jnrC;
 218             j_coord_offsetD  = DIM*jnrD;
 219             j_coord_offsetE  = DIM*jnrE;
 220             j_coord_offsetF  = DIM*jnrF;
 221             j_coord_offsetG  = DIM*jnrG;
 222             j_coord_offsetH  = DIM*jnrH;
 223
 224             /* load j atom coordinates */
 225             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 226                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 227                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 228                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 229                                                  &jx0,&jy0,&jz0);
 230
 231             /* Calculate displacement vector */
 232             dx00             = _mm256_sub_ps(ix0,jx0);
 233             dy00             = _mm256_sub_ps(iy0,jy0);
 234             dz00             = _mm256_sub_ps(iz0,jz0);
 235             dx10             = _mm256_sub_ps(ix1,jx0);
 236             dy10             = _mm256_sub_ps(iy1,jy0);
 237             dz10             = _mm256_sub_ps(iz1,jz0);
 238             dx20             = _mm256_sub_ps(ix2,jx0);
 239             dy20             = _mm256_sub_ps(iy2,jy0);
 240             dz20             = _mm256_sub_ps(iz2,jz0);
 241
 242             /* Calculate squared distance and things based on it */
 243             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 244             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 245             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 246
 247             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 248             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 249             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 250
 251             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 252             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 253             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 254
 255             /* Load parameters for j particles */
 256             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 257                                                                  charge+jnrC+0,charge+jnrD+0,
 258                                                                  charge+jnrE+0,charge+jnrF+0,
 259                                                                  charge+jnrG+0,charge+jnrH+0);
 260             vdwjidx0A        = 2*vdwtype[jnrA+0];
 261             vdwjidx0B        = 2*vdwtype[jnrB+0];
 262             vdwjidx0C        = 2*vdwtype[jnrC+0];
 263             vdwjidx0D        = 2*vdwtype[jnrD+0];
 264             vdwjidx0E        = 2*vdwtype[jnrE+0];
 265             vdwjidx0F        = 2*vdwtype[jnrF+0];
 266             vdwjidx0G        = 2*vdwtype[jnrG+0];
 267             vdwjidx0H        = 2*vdwtype[jnrH+0];
 268
 269             fjx0             = _mm256_setzero_ps();
 270             fjy0             = _mm256_setzero_ps();
 271             fjz0             = _mm256_setzero_ps();
 272
 273             /**************************
 274              * CALCULATE INTERACTIONS *
 275              **************************/
 276
 277             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 278             {
 279
 280             r00              = _mm256_mul_ps(rsq00,rinv00);
 281
 282             /* Compute parameters for interactions between i and j atoms */
 283             qq00             = _mm256_mul_ps(iq0,jq0);
 284             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 285                                             vdwioffsetptr0+vdwjidx0B,
 286                                             vdwioffsetptr0+vdwjidx0C,
 287                                             vdwioffsetptr0+vdwjidx0D,
 288                                             vdwioffsetptr0+vdwjidx0E,
 289                                             vdwioffsetptr0+vdwjidx0F,
 290                                             vdwioffsetptr0+vdwjidx0G,
 291                                             vdwioffsetptr0+vdwjidx0H,
 292                                             &c6_00,&c12_00);
 293
 294             /* Calculate table index by multiplying r with table scale and truncate to integer */
 295             rt               = _mm256_mul_ps(r00,vftabscale);
 296             vfitab           = _mm256_cvttps_epi32(rt);
 297             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 298             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 299             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 300             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 301             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 302             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 303
 304             /* REACTION-FIELD ELECTROSTATICS */
 305             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 306             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 307
 308             /* CUBIC SPLINE TABLE DISPERSION */
 309             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 310                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 311             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 312                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 313             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 314                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 315             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 316                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 317             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 318             Heps             = _mm256_mul_ps(vfeps,H);
 319             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 320             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 321             vvdw6            = _mm256_mul_ps(c6_00,VV);
 322             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 323             fvdw6            = _mm256_mul_ps(c6_00,FF);
 324
 325             /* CUBIC SPLINE TABLE REPULSION */
 326             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 327             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 328             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 329                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 330             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 331                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 332             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 333                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 334             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 335                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 336             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 337             Heps             = _mm256_mul_ps(vfeps,H);
 338             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 339             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 340             vvdw12           = _mm256_mul_ps(c12_00,VV);
 341             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 342             fvdw12           = _mm256_mul_ps(c12_00,FF);
 343             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 344             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 345
 346             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 347
 348             /* Update potential sum for this i atom from the interaction with this j atom. */
 349             velec            = _mm256_and_ps(velec,cutoff_mask);
 350             velecsum         = _mm256_add_ps(velecsum,velec);
 351             vvdw             = _mm256_and_ps(vvdw,cutoff_mask);
 352             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 353
 354             fscal            = _mm256_add_ps(felec,fvdw);
 355
 356             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 357
 358             /* Calculate temporary vectorial force */
 359             tx               = _mm256_mul_ps(fscal,dx00);
 360             ty               = _mm256_mul_ps(fscal,dy00);
 361             tz               = _mm256_mul_ps(fscal,dz00);
 362
 363             /* Update vectorial force */
 364             fix0             = _mm256_add_ps(fix0,tx);
 365             fiy0             = _mm256_add_ps(fiy0,ty);
 366             fiz0             = _mm256_add_ps(fiz0,tz);
 367
 368             fjx0             = _mm256_add_ps(fjx0,tx);
 369             fjy0             = _mm256_add_ps(fjy0,ty);
 370             fjz0             = _mm256_add_ps(fjz0,tz);
 371
 372             }
 373
 374             /**************************
 375              * CALCULATE INTERACTIONS *
 376              **************************/
 377
 378             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 379             {
 380
 381             /* Compute parameters for interactions between i and j atoms */
 382             qq10             = _mm256_mul_ps(iq1,jq0);
 383
 384             /* REACTION-FIELD ELECTROSTATICS */
 385             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 386             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 387
 388             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velec            = _mm256_and_ps(velec,cutoff_mask);
 392             velecsum         = _mm256_add_ps(velecsum,velec);
 393
 394             fscal            = felec;
 395
 396             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 397
 398             /* Calculate temporary vectorial force */
 399             tx               = _mm256_mul_ps(fscal,dx10);
 400             ty               = _mm256_mul_ps(fscal,dy10);
 401             tz               = _mm256_mul_ps(fscal,dz10);
 402
 403             /* Update vectorial force */
 404             fix1             = _mm256_add_ps(fix1,tx);
 405             fiy1             = _mm256_add_ps(fiy1,ty);
 406             fiz1             = _mm256_add_ps(fiz1,tz);
 407
 408             fjx0             = _mm256_add_ps(fjx0,tx);
 409             fjy0             = _mm256_add_ps(fjy0,ty);
 410             fjz0             = _mm256_add_ps(fjz0,tz);
 411
 412             }
 413
 414             /**************************
 415              * CALCULATE INTERACTIONS *
 416              **************************/
 417
 418             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 419             {
 420
 421             /* Compute parameters for interactions between i and j atoms */
 422             qq20             = _mm256_mul_ps(iq2,jq0);
 423
 424             /* REACTION-FIELD ELECTROSTATICS */
 425             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 426             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 427
 428             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
 429
 430             /* Update potential sum for this i atom from the interaction with this j atom. */
 431             velec            = _mm256_and_ps(velec,cutoff_mask);
 432             velecsum         = _mm256_add_ps(velecsum,velec);
 433
 434             fscal            = felec;
 435
 436             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 437
 438             /* Calculate temporary vectorial force */
 439             tx               = _mm256_mul_ps(fscal,dx20);
 440             ty               = _mm256_mul_ps(fscal,dy20);
 441             tz               = _mm256_mul_ps(fscal,dz20);
 442
 443             /* Update vectorial force */
 444             fix2             = _mm256_add_ps(fix2,tx);
 445             fiy2             = _mm256_add_ps(fiy2,ty);
 446             fiz2             = _mm256_add_ps(fiz2,tz);
 447
 448             fjx0             = _mm256_add_ps(fjx0,tx);
 449             fjy0             = _mm256_add_ps(fjy0,ty);
 450             fjz0             = _mm256_add_ps(fjz0,tz);
 451
 452             }
 453
 454             fjptrA             = f+j_coord_offsetA;
 455             fjptrB             = f+j_coord_offsetB;
 456             fjptrC             = f+j_coord_offsetC;
 457             fjptrD             = f+j_coord_offsetD;
 458             fjptrE             = f+j_coord_offsetE;
 459             fjptrF             = f+j_coord_offsetF;
 460             fjptrG             = f+j_coord_offsetG;
 461             fjptrH             = f+j_coord_offsetH;
 462
 463             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 464
 465             /* Inner loop uses 147 flops */
 466         }
 467
 468         if(jidx<j_index_end)
 469         {
 470
 471             /* Get j neighbor index, and coordinate index */
 472             jnrlistA         = jjnr[jidx];
 473             jnrlistB         = jjnr[jidx+1];
 474             jnrlistC         = jjnr[jidx+2];
 475             jnrlistD         = jjnr[jidx+3];
 476             jnrlistE         = jjnr[jidx+4];
 477             jnrlistF         = jjnr[jidx+5];
 478             jnrlistG         = jjnr[jidx+6];
 479             jnrlistH         = jjnr[jidx+7];
 480             /* Sign of each element will be negative for non-real atoms.
 481              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 482              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 483              */
 484             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 485                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 486
 487             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 488             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 489             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 490             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 491             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 492             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 493             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 494             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 495             j_coord_offsetA  = DIM*jnrA;
 496             j_coord_offsetB  = DIM*jnrB;
 497             j_coord_offsetC  = DIM*jnrC;
 498             j_coord_offsetD  = DIM*jnrD;
 499             j_coord_offsetE  = DIM*jnrE;
 500             j_coord_offsetF  = DIM*jnrF;
 501             j_coord_offsetG  = DIM*jnrG;
 502             j_coord_offsetH  = DIM*jnrH;
 503
 504             /* load j atom coordinates */
 505             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 506                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 507                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 508                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 509                                                  &jx0,&jy0,&jz0);
 510
 511             /* Calculate displacement vector */
 512             dx00             = _mm256_sub_ps(ix0,jx0);
 513             dy00             = _mm256_sub_ps(iy0,jy0);
 514             dz00             = _mm256_sub_ps(iz0,jz0);
 515             dx10             = _mm256_sub_ps(ix1,jx0);
 516             dy10             = _mm256_sub_ps(iy1,jy0);
 517             dz10             = _mm256_sub_ps(iz1,jz0);
 518             dx20             = _mm256_sub_ps(ix2,jx0);
 519             dy20             = _mm256_sub_ps(iy2,jy0);
 520             dz20             = _mm256_sub_ps(iz2,jz0);
 521
 522             /* Calculate squared distance and things based on it */
 523             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 524             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 525             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 526
 527             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 528             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 529             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 530
 531             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 532             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 533             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 534
 535             /* Load parameters for j particles */
 536             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 537                                                                  charge+jnrC+0,charge+jnrD+0,
 538                                                                  charge+jnrE+0,charge+jnrF+0,
 539                                                                  charge+jnrG+0,charge+jnrH+0);
 540             vdwjidx0A        = 2*vdwtype[jnrA+0];
 541             vdwjidx0B        = 2*vdwtype[jnrB+0];
 542             vdwjidx0C        = 2*vdwtype[jnrC+0];
 543             vdwjidx0D        = 2*vdwtype[jnrD+0];
 544             vdwjidx0E        = 2*vdwtype[jnrE+0];
 545             vdwjidx0F        = 2*vdwtype[jnrF+0];
 546             vdwjidx0G        = 2*vdwtype[jnrG+0];
 547             vdwjidx0H        = 2*vdwtype[jnrH+0];
 548
 549             fjx0             = _mm256_setzero_ps();
 550             fjy0             = _mm256_setzero_ps();
 551             fjz0             = _mm256_setzero_ps();
 552
 553             /**************************
 554              * CALCULATE INTERACTIONS *
 555              **************************/
 556
 557             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 558             {
 559
 560             r00              = _mm256_mul_ps(rsq00,rinv00);
 561             r00              = _mm256_andnot_ps(dummy_mask,r00);
 562
 563             /* Compute parameters for interactions between i and j atoms */
 564             qq00             = _mm256_mul_ps(iq0,jq0);
 565             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 566                                             vdwioffsetptr0+vdwjidx0B,
 567                                             vdwioffsetptr0+vdwjidx0C,
 568                                             vdwioffsetptr0+vdwjidx0D,
 569                                             vdwioffsetptr0+vdwjidx0E,
 570                                             vdwioffsetptr0+vdwjidx0F,
 571                                             vdwioffsetptr0+vdwjidx0G,
 572                                             vdwioffsetptr0+vdwjidx0H,
 573                                             &c6_00,&c12_00);
 574
 575             /* Calculate table index by multiplying r with table scale and truncate to integer */
 576             rt               = _mm256_mul_ps(r00,vftabscale);
 577             vfitab           = _mm256_cvttps_epi32(rt);
 578             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 579             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 580             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 581             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 582             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
 583             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
 584
 585             /* REACTION-FIELD ELECTROSTATICS */
 586             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 587             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 588
 589             /* CUBIC SPLINE TABLE DISPERSION */
 590             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 591                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 592             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 593                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 594             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 595                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 596             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 597                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 598             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 599             Heps             = _mm256_mul_ps(vfeps,H);
 600             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 601             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 602             vvdw6            = _mm256_mul_ps(c6_00,VV);
 603             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 604             fvdw6            = _mm256_mul_ps(c6_00,FF);
 605
 606             /* CUBIC SPLINE TABLE REPULSION */
 607             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 608             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 609             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 610                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 611             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 612                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 613             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 614                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 615             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 616                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 617             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 618             Heps             = _mm256_mul_ps(vfeps,H);
 619             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 620             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 621             vvdw12           = _mm256_mul_ps(c12_00,VV);
 622             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 623             fvdw12           = _mm256_mul_ps(c12_00,FF);
 624             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 625             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 626
 627             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 628
 629             /* Update potential sum for this i atom from the interaction with this j atom. */
 630             velec            = _mm256_and_ps(velec,cutoff_mask);
 631             velec            = _mm256_andnot_ps(dummy_mask,velec);
 632             velecsum         = _mm256_add_ps(velecsum,velec);
 633             vvdw             = _mm256_and_ps(vvdw,cutoff_mask);
 634             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 635             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 636
 637             fscal            = _mm256_add_ps(felec,fvdw);
 638
 639             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 640
 641             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 642
 643             /* Calculate temporary vectorial force */
 644             tx               = _mm256_mul_ps(fscal,dx00);
 645             ty               = _mm256_mul_ps(fscal,dy00);
 646             tz               = _mm256_mul_ps(fscal,dz00);
 647
 648             /* Update vectorial force */
 649             fix0             = _mm256_add_ps(fix0,tx);
 650             fiy0             = _mm256_add_ps(fiy0,ty);
 651             fiz0             = _mm256_add_ps(fiz0,tz);
 652
 653             fjx0             = _mm256_add_ps(fjx0,tx);
 654             fjy0             = _mm256_add_ps(fjy0,ty);
 655             fjz0             = _mm256_add_ps(fjz0,tz);
 656
 657             }
 658
 659             /**************************
 660              * CALCULATE INTERACTIONS *
 661              **************************/
 662
 663             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 664             {
 665
 666             /* Compute parameters for interactions between i and j atoms */
 667             qq10             = _mm256_mul_ps(iq1,jq0);
 668
 669             /* REACTION-FIELD ELECTROSTATICS */
 670             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 671             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 672
 673             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 674
 675             /* Update potential sum for this i atom from the interaction with this j atom. */
 676             velec            = _mm256_and_ps(velec,cutoff_mask);
 677             velec            = _mm256_andnot_ps(dummy_mask,velec);
 678             velecsum         = _mm256_add_ps(velecsum,velec);
 679
 680             fscal            = felec;
 681
 682             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 683
 684             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 685
 686             /* Calculate temporary vectorial force */
 687             tx               = _mm256_mul_ps(fscal,dx10);
 688             ty               = _mm256_mul_ps(fscal,dy10);
 689             tz               = _mm256_mul_ps(fscal,dz10);
 690
 691             /* Update vectorial force */
 692             fix1             = _mm256_add_ps(fix1,tx);
 693             fiy1             = _mm256_add_ps(fiy1,ty);
 694             fiz1             = _mm256_add_ps(fiz1,tz);
 695
 696             fjx0             = _mm256_add_ps(fjx0,tx);
 697             fjy0             = _mm256_add_ps(fjy0,ty);
 698             fjz0             = _mm256_add_ps(fjz0,tz);
 699
 700             }
 701
 702             /**************************
 703              * CALCULATE INTERACTIONS *
 704              **************************/
 705
 706             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 707             {
 708
 709             /* Compute parameters for interactions between i and j atoms */
 710             qq20             = _mm256_mul_ps(iq2,jq0);
 711
 712             /* REACTION-FIELD ELECTROSTATICS */
 713             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 714             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 715
 716             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
 717
 718             /* Update potential sum for this i atom from the interaction with this j atom. */
 719             velec            = _mm256_and_ps(velec,cutoff_mask);
 720             velec            = _mm256_andnot_ps(dummy_mask,velec);
 721             velecsum         = _mm256_add_ps(velecsum,velec);
 722
 723             fscal            = felec;
 724
 725             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 726
 727             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 728
 729             /* Calculate temporary vectorial force */
 730             tx               = _mm256_mul_ps(fscal,dx20);
 731             ty               = _mm256_mul_ps(fscal,dy20);
 732             tz               = _mm256_mul_ps(fscal,dz20);
 733
 734             /* Update vectorial force */
 735             fix2             = _mm256_add_ps(fix2,tx);
 736             fiy2             = _mm256_add_ps(fiy2,ty);
 737             fiz2             = _mm256_add_ps(fiz2,tz);
 738
 739             fjx0             = _mm256_add_ps(fjx0,tx);
 740             fjy0             = _mm256_add_ps(fjy0,ty);
 741             fjz0             = _mm256_add_ps(fjz0,tz);
 742
 743             }
 744
 745             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 746             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 747             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 748             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 749             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 750             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 751             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 752             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 753
 754             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 755
 756             /* Inner loop uses 148 flops */
 757         }
 758
 759         /* End of innermost loop */
 760
 761         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 762                                                  f+i_coord_offset,fshift+i_shift_offset);
 763
 764         ggid                        = gid[iidx];
 765         /* Update potential energies */
 766         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 767         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 768
 769         /* Increment number of inner iterations */
 770         inneriter                  += j_index_end - j_index_start;
 771
 772         /* Outer loop uses 20 flops */
 773     }
 774
 775     /* Increment number of outer iterations */
 776     outeriter        += nri;
 777
 778     /* Update outer/inner flops */
 779
 780     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*148);
 781 }
 782 /*
 783  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_avx_256_single
 784  * Electrostatics interaction: ReactionField
 785  * VdW interaction:            CubicSplineTable
 786  * Geometry:                   Water3-Particle
 787  * Calculate force/pot:        Force
 788  */
 789 void
 790 nb_kernel_ElecRFCut_VdwCSTab_GeomW3P1_F_avx_256_single
 791                     (t_nblist                    * gmx_restrict       nlist,
 792                      rvec                        * gmx_restrict          xx,
 793                      rvec                        * gmx_restrict          ff,
 794                      t_forcerec                  * gmx_restrict          fr,
 795                      t_mdatoms                   * gmx_restrict     mdatoms,
 796                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 797                      t_nrnb                      * gmx_restrict        nrnb)
 798 {
 799     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 800      * just 0 for non-waters.
 801      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 802      * jnr indices corresponding to data put in the four positions in the SIMD register.
 803      */
 804     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 805     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 806     int              jnrA,jnrB,jnrC,jnrD;
 807     int              jnrE,jnrF,jnrG,jnrH;
 808     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 809     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 810     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 811     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 812     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 813     real             rcutoff_scalar;
 814     real             *shiftvec,*fshift,*x,*f;
 815     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 816     real             scratch[4*DIM];
 817     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 818     real *           vdwioffsetptr0;
 819     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 820     real *           vdwioffsetptr1;
 821     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 822     real *           vdwioffsetptr2;
 823     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 824     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
 825     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 826     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 827     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 828     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 829     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 830     real             *charge;
 831     int              nvdwtype;
 832     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 833     int              *vdwtype;
 834     real             *vdwparam;
 835     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 836     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 837     __m256i          vfitab;
 838     __m128i          vfitab_lo,vfitab_hi;
 839     __m128i          ifour       = _mm_set1_epi32(4);
 840     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 841     real             *vftab;
 842     __m256           dummy_mask,cutoff_mask;
 843     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 844     __m256           one     = _mm256_set1_ps(1.0);
 845     __m256           two     = _mm256_set1_ps(2.0);
 846     x                = xx[0];
 847     f                = ff[0];
 848
 849     nri              = nlist->nri;
 850     iinr             = nlist->iinr;
 851     jindex           = nlist->jindex;
 852     jjnr             = nlist->jjnr;
 853     shiftidx         = nlist->shift;
 854     gid              = nlist->gid;
 855     shiftvec         = fr->shift_vec[0];
 856     fshift           = fr->fshift[0];
 857     facel            = _mm256_set1_ps(fr->epsfac);
 858     charge           = mdatoms->chargeA;
 859     krf              = _mm256_set1_ps(fr->ic->k_rf);
 860     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 861     crf              = _mm256_set1_ps(fr->ic->c_rf);
 862     nvdwtype         = fr->ntype;
 863     vdwparam         = fr->nbfp;
 864     vdwtype          = mdatoms->typeA;
 865
 866     vftab            = kernel_data->table_vdw->data;
 867     vftabscale       = _mm256_set1_ps(kernel_data->table_vdw->scale);
 868
 869     /* Setup water-specific parameters */
 870     inr              = nlist->iinr[0];
 871     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 872     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 873     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 874     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 875
 876     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 877     rcutoff_scalar   = fr->rcoulomb;
 878     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 879     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 880
 881     /* Avoid stupid compiler warnings */
 882     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 883     j_coord_offsetA = 0;
 884     j_coord_offsetB = 0;
 885     j_coord_offsetC = 0;
 886     j_coord_offsetD = 0;
 887     j_coord_offsetE = 0;
 888     j_coord_offsetF = 0;
 889     j_coord_offsetG = 0;
 890     j_coord_offsetH = 0;
 891
 892     outeriter        = 0;
 893     inneriter        = 0;
 894
 895     for(iidx=0;iidx<4*DIM;iidx++)
 896     {
 897         scratch[iidx] = 0.0;
 898     }
 899
 900     /* Start outer loop over neighborlists */
 901     for(iidx=0; iidx<nri; iidx++)
 902     {
 903         /* Load shift vector for this list */
 904         i_shift_offset   = DIM*shiftidx[iidx];
 905
 906         /* Load limits for loop over neighbors */
 907         j_index_start    = jindex[iidx];
 908         j_index_end      = jindex[iidx+1];
 909
 910         /* Get outer coordinate index */
 911         inr              = iinr[iidx];
 912         i_coord_offset   = DIM*inr;
 913
 914         /* Load i particle coords and add shift vector */
 915         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 916                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 917
 918         fix0             = _mm256_setzero_ps();
 919         fiy0             = _mm256_setzero_ps();
 920         fiz0             = _mm256_setzero_ps();
 921         fix1             = _mm256_setzero_ps();
 922         fiy1             = _mm256_setzero_ps();
 923         fiz1             = _mm256_setzero_ps();
 924         fix2             = _mm256_setzero_ps();
 925         fiy2             = _mm256_setzero_ps();
 926         fiz2             = _mm256_setzero_ps();
 927
 928         /* Start inner kernel loop */
 929         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 930         {
 931
 932             /* Get j neighbor index, and coordinate index */
 933             jnrA             = jjnr[jidx];
 934             jnrB             = jjnr[jidx+1];
 935             jnrC             = jjnr[jidx+2];
 936             jnrD             = jjnr[jidx+3];
 937             jnrE             = jjnr[jidx+4];
 938             jnrF             = jjnr[jidx+5];
 939             jnrG             = jjnr[jidx+6];
 940             jnrH             = jjnr[jidx+7];
 941             j_coord_offsetA  = DIM*jnrA;
 942             j_coord_offsetB  = DIM*jnrB;
 943             j_coord_offsetC  = DIM*jnrC;
 944             j_coord_offsetD  = DIM*jnrD;
 945             j_coord_offsetE  = DIM*jnrE;
 946             j_coord_offsetF  = DIM*jnrF;
 947             j_coord_offsetG  = DIM*jnrG;
 948             j_coord_offsetH  = DIM*jnrH;
 949
 950             /* load j atom coordinates */
 951             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 952                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 953                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 954                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 955                                                  &jx0,&jy0,&jz0);
 956
 957             /* Calculate displacement vector */
 958             dx00             = _mm256_sub_ps(ix0,jx0);
 959             dy00             = _mm256_sub_ps(iy0,jy0);
 960             dz00             = _mm256_sub_ps(iz0,jz0);
 961             dx10             = _mm256_sub_ps(ix1,jx0);
 962             dy10             = _mm256_sub_ps(iy1,jy0);
 963             dz10             = _mm256_sub_ps(iz1,jz0);
 964             dx20             = _mm256_sub_ps(ix2,jx0);
 965             dy20             = _mm256_sub_ps(iy2,jy0);
 966             dz20             = _mm256_sub_ps(iz2,jz0);
 967
 968             /* Calculate squared distance and things based on it */
 969             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 970             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 971             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 972
 973             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 974             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 975             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 976
 977             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 978             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 979             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 980
 981             /* Load parameters for j particles */
 982             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 983                                                                  charge+jnrC+0,charge+jnrD+0,
 984                                                                  charge+jnrE+0,charge+jnrF+0,
 985                                                                  charge+jnrG+0,charge+jnrH+0);
 986             vdwjidx0A        = 2*vdwtype[jnrA+0];
 987             vdwjidx0B        = 2*vdwtype[jnrB+0];
 988             vdwjidx0C        = 2*vdwtype[jnrC+0];
 989             vdwjidx0D        = 2*vdwtype[jnrD+0];
 990             vdwjidx0E        = 2*vdwtype[jnrE+0];
 991             vdwjidx0F        = 2*vdwtype[jnrF+0];
 992             vdwjidx0G        = 2*vdwtype[jnrG+0];
 993             vdwjidx0H        = 2*vdwtype[jnrH+0];
 994
 995             fjx0             = _mm256_setzero_ps();
 996             fjy0             = _mm256_setzero_ps();
 997             fjz0             = _mm256_setzero_ps();
 998
 999             /**************************
1000              * CALCULATE INTERACTIONS *
1001              **************************/
1002
1003             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1004             {
1005
1006             r00              = _mm256_mul_ps(rsq00,rinv00);
1007
1008             /* Compute parameters for interactions between i and j atoms */
1009             qq00             = _mm256_mul_ps(iq0,jq0);
1010             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1011                                             vdwioffsetptr0+vdwjidx0B,
1012                                             vdwioffsetptr0+vdwjidx0C,
1013                                             vdwioffsetptr0+vdwjidx0D,
1014                                             vdwioffsetptr0+vdwjidx0E,
1015                                             vdwioffsetptr0+vdwjidx0F,
1016                                             vdwioffsetptr0+vdwjidx0G,
1017                                             vdwioffsetptr0+vdwjidx0H,
1018                                             &c6_00,&c12_00);
1019
1020             /* Calculate table index by multiplying r with table scale and truncate to integer */
1021             rt               = _mm256_mul_ps(r00,vftabscale);
1022             vfitab           = _mm256_cvttps_epi32(rt);
1023             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1024             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1025             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1026             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1027             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1028             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1029
1030             /* REACTION-FIELD ELECTROSTATICS */
1031             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1032
1033             /* CUBIC SPLINE TABLE DISPERSION */
1034             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1035                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1036             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1037                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1038             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1039                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1040             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1041                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1042             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1043             Heps             = _mm256_mul_ps(vfeps,H);
1044             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1045             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1046             fvdw6            = _mm256_mul_ps(c6_00,FF);
1047
1048             /* CUBIC SPLINE TABLE REPULSION */
1049             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1050             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1051             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1052                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1053             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1054                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1055             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1056                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1057             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1058                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1059             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1060             Heps             = _mm256_mul_ps(vfeps,H);
1061             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1062             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1063             fvdw12           = _mm256_mul_ps(c12_00,FF);
1064             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1065
1066             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1067
1068             fscal            = _mm256_add_ps(felec,fvdw);
1069
1070             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1071
1072             /* Calculate temporary vectorial force */
1073             tx               = _mm256_mul_ps(fscal,dx00);
1074             ty               = _mm256_mul_ps(fscal,dy00);
1075             tz               = _mm256_mul_ps(fscal,dz00);
1076
1077             /* Update vectorial force */
1078             fix0             = _mm256_add_ps(fix0,tx);
1079             fiy0             = _mm256_add_ps(fiy0,ty);
1080             fiz0             = _mm256_add_ps(fiz0,tz);
1081
1082             fjx0             = _mm256_add_ps(fjx0,tx);
1083             fjy0             = _mm256_add_ps(fjy0,ty);
1084             fjz0             = _mm256_add_ps(fjz0,tz);
1085
1086             }
1087
1088             /**************************
1089              * CALCULATE INTERACTIONS *
1090              **************************/
1091
1092             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1093             {
1094
1095             /* Compute parameters for interactions between i and j atoms */
1096             qq10             = _mm256_mul_ps(iq1,jq0);
1097
1098             /* REACTION-FIELD ELECTROSTATICS */
1099             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1100
1101             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1102
1103             fscal            = felec;
1104
1105             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1106
1107             /* Calculate temporary vectorial force */
1108             tx               = _mm256_mul_ps(fscal,dx10);
1109             ty               = _mm256_mul_ps(fscal,dy10);
1110             tz               = _mm256_mul_ps(fscal,dz10);
1111
1112             /* Update vectorial force */
1113             fix1             = _mm256_add_ps(fix1,tx);
1114             fiy1             = _mm256_add_ps(fiy1,ty);
1115             fiz1             = _mm256_add_ps(fiz1,tz);
1116
1117             fjx0             = _mm256_add_ps(fjx0,tx);
1118             fjy0             = _mm256_add_ps(fjy0,ty);
1119             fjz0             = _mm256_add_ps(fjz0,tz);
1120
1121             }
1122
1123             /**************************
1124              * CALCULATE INTERACTIONS *
1125              **************************/
1126
1127             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1128             {
1129
1130             /* Compute parameters for interactions between i and j atoms */
1131             qq20             = _mm256_mul_ps(iq2,jq0);
1132
1133             /* REACTION-FIELD ELECTROSTATICS */
1134             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1135
1136             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1137
1138             fscal            = felec;
1139
1140             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1141
1142             /* Calculate temporary vectorial force */
1143             tx               = _mm256_mul_ps(fscal,dx20);
1144             ty               = _mm256_mul_ps(fscal,dy20);
1145             tz               = _mm256_mul_ps(fscal,dz20);
1146
1147             /* Update vectorial force */
1148             fix2             = _mm256_add_ps(fix2,tx);
1149             fiy2             = _mm256_add_ps(fiy2,ty);
1150             fiz2             = _mm256_add_ps(fiz2,tz);
1151
1152             fjx0             = _mm256_add_ps(fjx0,tx);
1153             fjy0             = _mm256_add_ps(fjy0,ty);
1154             fjz0             = _mm256_add_ps(fjz0,tz);
1155
1156             }
1157
1158             fjptrA             = f+j_coord_offsetA;
1159             fjptrB             = f+j_coord_offsetB;
1160             fjptrC             = f+j_coord_offsetC;
1161             fjptrD             = f+j_coord_offsetD;
1162             fjptrE             = f+j_coord_offsetE;
1163             fjptrF             = f+j_coord_offsetF;
1164             fjptrG             = f+j_coord_offsetG;
1165             fjptrH             = f+j_coord_offsetH;
1166
1167             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1168
1169             /* Inner loop uses 120 flops */
1170         }
1171
1172         if(jidx<j_index_end)
1173         {
1174
1175             /* Get j neighbor index, and coordinate index */
1176             jnrlistA         = jjnr[jidx];
1177             jnrlistB         = jjnr[jidx+1];
1178             jnrlistC         = jjnr[jidx+2];
1179             jnrlistD         = jjnr[jidx+3];
1180             jnrlistE         = jjnr[jidx+4];
1181             jnrlistF         = jjnr[jidx+5];
1182             jnrlistG         = jjnr[jidx+6];
1183             jnrlistH         = jjnr[jidx+7];
1184             /* Sign of each element will be negative for non-real atoms.
1185              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1186              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1187              */
1188             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1189                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1190
1191             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1192             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1193             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1194             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1195             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1196             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1197             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1198             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1199             j_coord_offsetA  = DIM*jnrA;
1200             j_coord_offsetB  = DIM*jnrB;
1201             j_coord_offsetC  = DIM*jnrC;
1202             j_coord_offsetD  = DIM*jnrD;
1203             j_coord_offsetE  = DIM*jnrE;
1204             j_coord_offsetF  = DIM*jnrF;
1205             j_coord_offsetG  = DIM*jnrG;
1206             j_coord_offsetH  = DIM*jnrH;
1207
1208             /* load j atom coordinates */
1209             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1210                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1211                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1212                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1213                                                  &jx0,&jy0,&jz0);
1214
1215             /* Calculate displacement vector */
1216             dx00             = _mm256_sub_ps(ix0,jx0);
1217             dy00             = _mm256_sub_ps(iy0,jy0);
1218             dz00             = _mm256_sub_ps(iz0,jz0);
1219             dx10             = _mm256_sub_ps(ix1,jx0);
1220             dy10             = _mm256_sub_ps(iy1,jy0);
1221             dz10             = _mm256_sub_ps(iz1,jz0);
1222             dx20             = _mm256_sub_ps(ix2,jx0);
1223             dy20             = _mm256_sub_ps(iy2,jy0);
1224             dz20             = _mm256_sub_ps(iz2,jz0);
1225
1226             /* Calculate squared distance and things based on it */
1227             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1228             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1229             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1230
1231             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1232             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1233             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1234
1235             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1236             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1237             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1238
1239             /* Load parameters for j particles */
1240             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1241                                                                  charge+jnrC+0,charge+jnrD+0,
1242                                                                  charge+jnrE+0,charge+jnrF+0,
1243                                                                  charge+jnrG+0,charge+jnrH+0);
1244             vdwjidx0A        = 2*vdwtype[jnrA+0];
1245             vdwjidx0B        = 2*vdwtype[jnrB+0];
1246             vdwjidx0C        = 2*vdwtype[jnrC+0];
1247             vdwjidx0D        = 2*vdwtype[jnrD+0];
1248             vdwjidx0E        = 2*vdwtype[jnrE+0];
1249             vdwjidx0F        = 2*vdwtype[jnrF+0];
1250             vdwjidx0G        = 2*vdwtype[jnrG+0];
1251             vdwjidx0H        = 2*vdwtype[jnrH+0];
1252
1253             fjx0             = _mm256_setzero_ps();
1254             fjy0             = _mm256_setzero_ps();
1255             fjz0             = _mm256_setzero_ps();
1256
1257             /**************************
1258              * CALCULATE INTERACTIONS *
1259              **************************/
1260
1261             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1262             {
1263
1264             r00              = _mm256_mul_ps(rsq00,rinv00);
1265             r00              = _mm256_andnot_ps(dummy_mask,r00);
1266
1267             /* Compute parameters for interactions between i and j atoms */
1268             qq00             = _mm256_mul_ps(iq0,jq0);
1269             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
1270                                             vdwioffsetptr0+vdwjidx0B,
1271                                             vdwioffsetptr0+vdwjidx0C,
1272                                             vdwioffsetptr0+vdwjidx0D,
1273                                             vdwioffsetptr0+vdwjidx0E,
1274                                             vdwioffsetptr0+vdwjidx0F,
1275                                             vdwioffsetptr0+vdwjidx0G,
1276                                             vdwioffsetptr0+vdwjidx0H,
1277                                             &c6_00,&c12_00);
1278
1279             /* Calculate table index by multiplying r with table scale and truncate to integer */
1280             rt               = _mm256_mul_ps(r00,vftabscale);
1281             vfitab           = _mm256_cvttps_epi32(rt);
1282             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1283             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1284             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1285             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1286             vfitab_lo        = _mm_slli_epi32(vfitab_lo,3);
1287             vfitab_hi        = _mm_slli_epi32(vfitab_hi,3);
1288
1289             /* REACTION-FIELD ELECTROSTATICS */
1290             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1291
1292             /* CUBIC SPLINE TABLE DISPERSION */
1293             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1294                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1295             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1296                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1297             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1298                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1299             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1300                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1301             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1302             Heps             = _mm256_mul_ps(vfeps,H);
1303             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1304             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1305             fvdw6            = _mm256_mul_ps(c6_00,FF);
1306
1307             /* CUBIC SPLINE TABLE REPULSION */
1308             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
1309             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
1310             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1311                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1312             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1313                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1314             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1315                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1316             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1317                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1318             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1319             Heps             = _mm256_mul_ps(vfeps,H);
1320             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1321             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1322             fvdw12           = _mm256_mul_ps(c12_00,FF);
1323             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
1324
1325             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1326
1327             fscal            = _mm256_add_ps(felec,fvdw);
1328
1329             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1330
1331             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1332
1333             /* Calculate temporary vectorial force */
1334             tx               = _mm256_mul_ps(fscal,dx00);
1335             ty               = _mm256_mul_ps(fscal,dy00);
1336             tz               = _mm256_mul_ps(fscal,dz00);
1337
1338             /* Update vectorial force */
1339             fix0             = _mm256_add_ps(fix0,tx);
1340             fiy0             = _mm256_add_ps(fiy0,ty);
1341             fiz0             = _mm256_add_ps(fiz0,tz);
1342
1343             fjx0             = _mm256_add_ps(fjx0,tx);
1344             fjy0             = _mm256_add_ps(fjy0,ty);
1345             fjz0             = _mm256_add_ps(fjz0,tz);
1346
1347             }
1348
1349             /**************************
1350              * CALCULATE INTERACTIONS *
1351              **************************/
1352
1353             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1354             {
1355
1356             /* Compute parameters for interactions between i and j atoms */
1357             qq10             = _mm256_mul_ps(iq1,jq0);
1358
1359             /* REACTION-FIELD ELECTROSTATICS */
1360             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1361
1362             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1363
1364             fscal            = felec;
1365
1366             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1367
1368             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1369
1370             /* Calculate temporary vectorial force */
1371             tx               = _mm256_mul_ps(fscal,dx10);
1372             ty               = _mm256_mul_ps(fscal,dy10);
1373             tz               = _mm256_mul_ps(fscal,dz10);
1374
1375             /* Update vectorial force */
1376             fix1             = _mm256_add_ps(fix1,tx);
1377             fiy1             = _mm256_add_ps(fiy1,ty);
1378             fiz1             = _mm256_add_ps(fiz1,tz);
1379
1380             fjx0             = _mm256_add_ps(fjx0,tx);
1381             fjy0             = _mm256_add_ps(fjy0,ty);
1382             fjz0             = _mm256_add_ps(fjz0,tz);
1383
1384             }
1385
1386             /**************************
1387              * CALCULATE INTERACTIONS *
1388              **************************/
1389
1390             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1391             {
1392
1393             /* Compute parameters for interactions between i and j atoms */
1394             qq20             = _mm256_mul_ps(iq2,jq0);
1395
1396             /* REACTION-FIELD ELECTROSTATICS */
1397             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1398
1399             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1400
1401             fscal            = felec;
1402
1403             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1404
1405             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1406
1407             /* Calculate temporary vectorial force */
1408             tx               = _mm256_mul_ps(fscal,dx20);
1409             ty               = _mm256_mul_ps(fscal,dy20);
1410             tz               = _mm256_mul_ps(fscal,dz20);
1411
1412             /* Update vectorial force */
1413             fix2             = _mm256_add_ps(fix2,tx);
1414             fiy2             = _mm256_add_ps(fiy2,ty);
1415             fiz2             = _mm256_add_ps(fiz2,tz);
1416
1417             fjx0             = _mm256_add_ps(fjx0,tx);
1418             fjy0             = _mm256_add_ps(fjy0,ty);
1419             fjz0             = _mm256_add_ps(fjz0,tz);
1420
1421             }
1422
1423             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1424             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1425             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1426             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1427             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1428             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1429             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1430             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1431
1432             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1433
1434             /* Inner loop uses 121 flops */
1435         }
1436
1437         /* End of innermost loop */
1438
1439         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1440                                                  f+i_coord_offset,fshift+i_shift_offset);
1441
1442         /* Increment number of inner iterations */
1443         inneriter                  += j_index_end - j_index_start;
1444
1445         /* Outer loop uses 18 flops */
1446     }
1447
1448     /* Increment number of outer iterations */
1449     outeriter        += nri;
1450
1451     /* Update outer/inner flops */
1452
1453     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*121);
1454 }