src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCSTab_VdwNone_GeomW4P1_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_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 "gromacs/math/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_single.h"
  50 #include "kernelutil_x86_avx_256_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW4P1_VF_avx_256_single
  54  * Electrostatics interaction: CubicSplineTable
  55  * VdW interaction:            None
  56  * Geometry:                   Water4-Particle
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecCSTab_VdwNone_GeomW4P1_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 *           vdwioffsetptr1;
  89     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  97     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  98     __m256           dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  99     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 100     real             *charge;
 101     __m256i          vfitab;
 102     __m128i          vfitab_lo,vfitab_hi;
 103     __m128i          ifour       = _mm_set1_epi32(4);
 104     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 105     real             *vftab;
 106     __m256           dummy_mask,cutoff_mask;
 107     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 108     __m256           one     = _mm256_set1_ps(1.0);
 109     __m256           two     = _mm256_set1_ps(2.0);
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = _mm256_set1_ps(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123
 124     vftab            = kernel_data->table_elec->data;
 125     vftabscale       = _mm256_set1_ps(kernel_data->table_elec->scale);
 126
 127     /* Setup water-specific parameters */
 128     inr              = nlist->iinr[0];
 129     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 130     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 131     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 132
 133     /* Avoid stupid compiler warnings */
 134     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 135     j_coord_offsetA = 0;
 136     j_coord_offsetB = 0;
 137     j_coord_offsetC = 0;
 138     j_coord_offsetD = 0;
 139     j_coord_offsetE = 0;
 140     j_coord_offsetF = 0;
 141     j_coord_offsetG = 0;
 142     j_coord_offsetH = 0;
 143
 144     outeriter        = 0;
 145     inneriter        = 0;
 146
 147     for(iidx=0;iidx<4*DIM;iidx++)
 148     {
 149         scratch[iidx] = 0.0;
 150     }
 151
 152     /* Start outer loop over neighborlists */
 153     for(iidx=0; iidx<nri; iidx++)
 154     {
 155         /* Load shift vector for this list */
 156         i_shift_offset   = DIM*shiftidx[iidx];
 157
 158         /* Load limits for loop over neighbors */
 159         j_index_start    = jindex[iidx];
 160         j_index_end      = jindex[iidx+1];
 161
 162         /* Get outer coordinate index */
 163         inr              = iinr[iidx];
 164         i_coord_offset   = DIM*inr;
 165
 166         /* Load i particle coords and add shift vector */
 167         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 168                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 169
 170         fix1             = _mm256_setzero_ps();
 171         fiy1             = _mm256_setzero_ps();
 172         fiz1             = _mm256_setzero_ps();
 173         fix2             = _mm256_setzero_ps();
 174         fiy2             = _mm256_setzero_ps();
 175         fiz2             = _mm256_setzero_ps();
 176         fix3             = _mm256_setzero_ps();
 177         fiy3             = _mm256_setzero_ps();
 178         fiz3             = _mm256_setzero_ps();
 179
 180         /* Reset potential sums */
 181         velecsum         = _mm256_setzero_ps();
 182
 183         /* Start inner kernel loop */
 184         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 185         {
 186
 187             /* Get j neighbor index, and coordinate index */
 188             jnrA             = jjnr[jidx];
 189             jnrB             = jjnr[jidx+1];
 190             jnrC             = jjnr[jidx+2];
 191             jnrD             = jjnr[jidx+3];
 192             jnrE             = jjnr[jidx+4];
 193             jnrF             = jjnr[jidx+5];
 194             jnrG             = jjnr[jidx+6];
 195             jnrH             = jjnr[jidx+7];
 196             j_coord_offsetA  = DIM*jnrA;
 197             j_coord_offsetB  = DIM*jnrB;
 198             j_coord_offsetC  = DIM*jnrC;
 199             j_coord_offsetD  = DIM*jnrD;
 200             j_coord_offsetE  = DIM*jnrE;
 201             j_coord_offsetF  = DIM*jnrF;
 202             j_coord_offsetG  = DIM*jnrG;
 203             j_coord_offsetH  = DIM*jnrH;
 204
 205             /* load j atom coordinates */
 206             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 207                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 208                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 209                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 210                                                  &jx0,&jy0,&jz0);
 211
 212             /* Calculate displacement vector */
 213             dx10             = _mm256_sub_ps(ix1,jx0);
 214             dy10             = _mm256_sub_ps(iy1,jy0);
 215             dz10             = _mm256_sub_ps(iz1,jz0);
 216             dx20             = _mm256_sub_ps(ix2,jx0);
 217             dy20             = _mm256_sub_ps(iy2,jy0);
 218             dz20             = _mm256_sub_ps(iz2,jz0);
 219             dx30             = _mm256_sub_ps(ix3,jx0);
 220             dy30             = _mm256_sub_ps(iy3,jy0);
 221             dz30             = _mm256_sub_ps(iz3,jz0);
 222
 223             /* Calculate squared distance and things based on it */
 224             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 225             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 226             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
 227
 228             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 229             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 230             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
 231
 232             /* Load parameters for j particles */
 233             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 234                                                                  charge+jnrC+0,charge+jnrD+0,
 235                                                                  charge+jnrE+0,charge+jnrF+0,
 236                                                                  charge+jnrG+0,charge+jnrH+0);
 237
 238             fjx0             = _mm256_setzero_ps();
 239             fjy0             = _mm256_setzero_ps();
 240             fjz0             = _mm256_setzero_ps();
 241
 242             /**************************
 243              * CALCULATE INTERACTIONS *
 244              **************************/
 245
 246             r10              = _mm256_mul_ps(rsq10,rinv10);
 247
 248             /* Compute parameters for interactions between i and j atoms */
 249             qq10             = _mm256_mul_ps(iq1,jq0);
 250
 251             /* Calculate table index by multiplying r with table scale and truncate to integer */
 252             rt               = _mm256_mul_ps(r10,vftabscale);
 253             vfitab           = _mm256_cvttps_epi32(rt);
 254             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 255             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 256             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 257             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 258             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 259             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 260
 261             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 262             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 263                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 264             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 265                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 266             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 267                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 268             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 269                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 270             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 271             Heps             = _mm256_mul_ps(vfeps,H);
 272             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 273             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 274             velec            = _mm256_mul_ps(qq10,VV);
 275             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 276             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq10,FF),_mm256_mul_ps(vftabscale,rinv10)));
 277
 278             /* Update potential sum for this i atom from the interaction with this j atom. */
 279             velecsum         = _mm256_add_ps(velecsum,velec);
 280
 281             fscal            = felec;
 282
 283             /* Calculate temporary vectorial force */
 284             tx               = _mm256_mul_ps(fscal,dx10);
 285             ty               = _mm256_mul_ps(fscal,dy10);
 286             tz               = _mm256_mul_ps(fscal,dz10);
 287
 288             /* Update vectorial force */
 289             fix1             = _mm256_add_ps(fix1,tx);
 290             fiy1             = _mm256_add_ps(fiy1,ty);
 291             fiz1             = _mm256_add_ps(fiz1,tz);
 292
 293             fjx0             = _mm256_add_ps(fjx0,tx);
 294             fjy0             = _mm256_add_ps(fjy0,ty);
 295             fjz0             = _mm256_add_ps(fjz0,tz);
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             r20              = _mm256_mul_ps(rsq20,rinv20);
 302
 303             /* Compute parameters for interactions between i and j atoms */
 304             qq20             = _mm256_mul_ps(iq2,jq0);
 305
 306             /* Calculate table index by multiplying r with table scale and truncate to integer */
 307             rt               = _mm256_mul_ps(r20,vftabscale);
 308             vfitab           = _mm256_cvttps_epi32(rt);
 309             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 310             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 311             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 312             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 313             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 314             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 315
 316             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 317             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 318                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 319             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 320                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 321             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 322                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 323             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 324                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 325             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 326             Heps             = _mm256_mul_ps(vfeps,H);
 327             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 328             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 329             velec            = _mm256_mul_ps(qq20,VV);
 330             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 331             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq20,FF),_mm256_mul_ps(vftabscale,rinv20)));
 332
 333             /* Update potential sum for this i atom from the interaction with this j atom. */
 334             velecsum         = _mm256_add_ps(velecsum,velec);
 335
 336             fscal            = felec;
 337
 338             /* Calculate temporary vectorial force */
 339             tx               = _mm256_mul_ps(fscal,dx20);
 340             ty               = _mm256_mul_ps(fscal,dy20);
 341             tz               = _mm256_mul_ps(fscal,dz20);
 342
 343             /* Update vectorial force */
 344             fix2             = _mm256_add_ps(fix2,tx);
 345             fiy2             = _mm256_add_ps(fiy2,ty);
 346             fiz2             = _mm256_add_ps(fiz2,tz);
 347
 348             fjx0             = _mm256_add_ps(fjx0,tx);
 349             fjy0             = _mm256_add_ps(fjy0,ty);
 350             fjz0             = _mm256_add_ps(fjz0,tz);
 351
 352             /**************************
 353              * CALCULATE INTERACTIONS *
 354              **************************/
 355
 356             r30              = _mm256_mul_ps(rsq30,rinv30);
 357
 358             /* Compute parameters for interactions between i and j atoms */
 359             qq30             = _mm256_mul_ps(iq3,jq0);
 360
 361             /* Calculate table index by multiplying r with table scale and truncate to integer */
 362             rt               = _mm256_mul_ps(r30,vftabscale);
 363             vfitab           = _mm256_cvttps_epi32(rt);
 364             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 365             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 366             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 367             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 368             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 369             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 370
 371             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 372             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 373                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 374             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 375                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 376             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 377                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 378             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 379                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 380             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 381             Heps             = _mm256_mul_ps(vfeps,H);
 382             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 383             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 384             velec            = _mm256_mul_ps(qq30,VV);
 385             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 386             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq30,FF),_mm256_mul_ps(vftabscale,rinv30)));
 387
 388             /* Update potential sum for this i atom from the interaction with this j atom. */
 389             velecsum         = _mm256_add_ps(velecsum,velec);
 390
 391             fscal            = felec;
 392
 393             /* Calculate temporary vectorial force */
 394             tx               = _mm256_mul_ps(fscal,dx30);
 395             ty               = _mm256_mul_ps(fscal,dy30);
 396             tz               = _mm256_mul_ps(fscal,dz30);
 397
 398             /* Update vectorial force */
 399             fix3             = _mm256_add_ps(fix3,tx);
 400             fiy3             = _mm256_add_ps(fiy3,ty);
 401             fiz3             = _mm256_add_ps(fiz3,tz);
 402
 403             fjx0             = _mm256_add_ps(fjx0,tx);
 404             fjy0             = _mm256_add_ps(fjy0,ty);
 405             fjz0             = _mm256_add_ps(fjz0,tz);
 406
 407             fjptrA             = f+j_coord_offsetA;
 408             fjptrB             = f+j_coord_offsetB;
 409             fjptrC             = f+j_coord_offsetC;
 410             fjptrD             = f+j_coord_offsetD;
 411             fjptrE             = f+j_coord_offsetE;
 412             fjptrF             = f+j_coord_offsetF;
 413             fjptrG             = f+j_coord_offsetG;
 414             fjptrH             = f+j_coord_offsetH;
 415
 416             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 417
 418             /* Inner loop uses 132 flops */
 419         }
 420
 421         if(jidx<j_index_end)
 422         {
 423
 424             /* Get j neighbor index, and coordinate index */
 425             jnrlistA         = jjnr[jidx];
 426             jnrlistB         = jjnr[jidx+1];
 427             jnrlistC         = jjnr[jidx+2];
 428             jnrlistD         = jjnr[jidx+3];
 429             jnrlistE         = jjnr[jidx+4];
 430             jnrlistF         = jjnr[jidx+5];
 431             jnrlistG         = jjnr[jidx+6];
 432             jnrlistH         = jjnr[jidx+7];
 433             /* Sign of each element will be negative for non-real atoms.
 434              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 435              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 436              */
 437             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 438                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 439
 440             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 441             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 442             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 443             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 444             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 445             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 446             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 447             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 448             j_coord_offsetA  = DIM*jnrA;
 449             j_coord_offsetB  = DIM*jnrB;
 450             j_coord_offsetC  = DIM*jnrC;
 451             j_coord_offsetD  = DIM*jnrD;
 452             j_coord_offsetE  = DIM*jnrE;
 453             j_coord_offsetF  = DIM*jnrF;
 454             j_coord_offsetG  = DIM*jnrG;
 455             j_coord_offsetH  = DIM*jnrH;
 456
 457             /* load j atom coordinates */
 458             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 459                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 460                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 461                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 462                                                  &jx0,&jy0,&jz0);
 463
 464             /* Calculate displacement vector */
 465             dx10             = _mm256_sub_ps(ix1,jx0);
 466             dy10             = _mm256_sub_ps(iy1,jy0);
 467             dz10             = _mm256_sub_ps(iz1,jz0);
 468             dx20             = _mm256_sub_ps(ix2,jx0);
 469             dy20             = _mm256_sub_ps(iy2,jy0);
 470             dz20             = _mm256_sub_ps(iz2,jz0);
 471             dx30             = _mm256_sub_ps(ix3,jx0);
 472             dy30             = _mm256_sub_ps(iy3,jy0);
 473             dz30             = _mm256_sub_ps(iz3,jz0);
 474
 475             /* Calculate squared distance and things based on it */
 476             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 477             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 478             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
 479
 480             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 481             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 482             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
 483
 484             /* Load parameters for j particles */
 485             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 486                                                                  charge+jnrC+0,charge+jnrD+0,
 487                                                                  charge+jnrE+0,charge+jnrF+0,
 488                                                                  charge+jnrG+0,charge+jnrH+0);
 489
 490             fjx0             = _mm256_setzero_ps();
 491             fjy0             = _mm256_setzero_ps();
 492             fjz0             = _mm256_setzero_ps();
 493
 494             /**************************
 495              * CALCULATE INTERACTIONS *
 496              **************************/
 497
 498             r10              = _mm256_mul_ps(rsq10,rinv10);
 499             r10              = _mm256_andnot_ps(dummy_mask,r10);
 500
 501             /* Compute parameters for interactions between i and j atoms */
 502             qq10             = _mm256_mul_ps(iq1,jq0);
 503
 504             /* Calculate table index by multiplying r with table scale and truncate to integer */
 505             rt               = _mm256_mul_ps(r10,vftabscale);
 506             vfitab           = _mm256_cvttps_epi32(rt);
 507             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 508             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 509             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 510             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 511             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 512             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 513
 514             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 515             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 516                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 517             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 518                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 519             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 520                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 521             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 522                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 523             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 524             Heps             = _mm256_mul_ps(vfeps,H);
 525             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 526             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 527             velec            = _mm256_mul_ps(qq10,VV);
 528             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 529             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq10,FF),_mm256_mul_ps(vftabscale,rinv10)));
 530
 531             /* Update potential sum for this i atom from the interaction with this j atom. */
 532             velec            = _mm256_andnot_ps(dummy_mask,velec);
 533             velecsum         = _mm256_add_ps(velecsum,velec);
 534
 535             fscal            = felec;
 536
 537             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 538
 539             /* Calculate temporary vectorial force */
 540             tx               = _mm256_mul_ps(fscal,dx10);
 541             ty               = _mm256_mul_ps(fscal,dy10);
 542             tz               = _mm256_mul_ps(fscal,dz10);
 543
 544             /* Update vectorial force */
 545             fix1             = _mm256_add_ps(fix1,tx);
 546             fiy1             = _mm256_add_ps(fiy1,ty);
 547             fiz1             = _mm256_add_ps(fiz1,tz);
 548
 549             fjx0             = _mm256_add_ps(fjx0,tx);
 550             fjy0             = _mm256_add_ps(fjy0,ty);
 551             fjz0             = _mm256_add_ps(fjz0,tz);
 552
 553             /**************************
 554              * CALCULATE INTERACTIONS *
 555              **************************/
 556
 557             r20              = _mm256_mul_ps(rsq20,rinv20);
 558             r20              = _mm256_andnot_ps(dummy_mask,r20);
 559
 560             /* Compute parameters for interactions between i and j atoms */
 561             qq20             = _mm256_mul_ps(iq2,jq0);
 562
 563             /* Calculate table index by multiplying r with table scale and truncate to integer */
 564             rt               = _mm256_mul_ps(r20,vftabscale);
 565             vfitab           = _mm256_cvttps_epi32(rt);
 566             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 567             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 568             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 569             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 570             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 571             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 572
 573             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 574             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 575                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 576             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 577                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 578             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 579                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 580             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 581                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 582             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 583             Heps             = _mm256_mul_ps(vfeps,H);
 584             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 585             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 586             velec            = _mm256_mul_ps(qq20,VV);
 587             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 588             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq20,FF),_mm256_mul_ps(vftabscale,rinv20)));
 589
 590             /* Update potential sum for this i atom from the interaction with this j atom. */
 591             velec            = _mm256_andnot_ps(dummy_mask,velec);
 592             velecsum         = _mm256_add_ps(velecsum,velec);
 593
 594             fscal            = felec;
 595
 596             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 597
 598             /* Calculate temporary vectorial force */
 599             tx               = _mm256_mul_ps(fscal,dx20);
 600             ty               = _mm256_mul_ps(fscal,dy20);
 601             tz               = _mm256_mul_ps(fscal,dz20);
 602
 603             /* Update vectorial force */
 604             fix2             = _mm256_add_ps(fix2,tx);
 605             fiy2             = _mm256_add_ps(fiy2,ty);
 606             fiz2             = _mm256_add_ps(fiz2,tz);
 607
 608             fjx0             = _mm256_add_ps(fjx0,tx);
 609             fjy0             = _mm256_add_ps(fjy0,ty);
 610             fjz0             = _mm256_add_ps(fjz0,tz);
 611
 612             /**************************
 613              * CALCULATE INTERACTIONS *
 614              **************************/
 615
 616             r30              = _mm256_mul_ps(rsq30,rinv30);
 617             r30              = _mm256_andnot_ps(dummy_mask,r30);
 618
 619             /* Compute parameters for interactions between i and j atoms */
 620             qq30             = _mm256_mul_ps(iq3,jq0);
 621
 622             /* Calculate table index by multiplying r with table scale and truncate to integer */
 623             rt               = _mm256_mul_ps(r30,vftabscale);
 624             vfitab           = _mm256_cvttps_epi32(rt);
 625             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 626             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 627             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 628             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 629             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 630             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 631
 632             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 633             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 634                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 635             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 636                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 637             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 638                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 639             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 640                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 641             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 642             Heps             = _mm256_mul_ps(vfeps,H);
 643             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 644             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 645             velec            = _mm256_mul_ps(qq30,VV);
 646             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 647             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq30,FF),_mm256_mul_ps(vftabscale,rinv30)));
 648
 649             /* Update potential sum for this i atom from the interaction with this j atom. */
 650             velec            = _mm256_andnot_ps(dummy_mask,velec);
 651             velecsum         = _mm256_add_ps(velecsum,velec);
 652
 653             fscal            = felec;
 654
 655             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 656
 657             /* Calculate temporary vectorial force */
 658             tx               = _mm256_mul_ps(fscal,dx30);
 659             ty               = _mm256_mul_ps(fscal,dy30);
 660             tz               = _mm256_mul_ps(fscal,dz30);
 661
 662             /* Update vectorial force */
 663             fix3             = _mm256_add_ps(fix3,tx);
 664             fiy3             = _mm256_add_ps(fiy3,ty);
 665             fiz3             = _mm256_add_ps(fiz3,tz);
 666
 667             fjx0             = _mm256_add_ps(fjx0,tx);
 668             fjy0             = _mm256_add_ps(fjy0,ty);
 669             fjz0             = _mm256_add_ps(fjz0,tz);
 670
 671             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 672             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 673             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 674             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 675             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 676             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 677             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 678             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 679
 680             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 681
 682             /* Inner loop uses 135 flops */
 683         }
 684
 685         /* End of innermost loop */
 686
 687         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 688                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 689
 690         ggid                        = gid[iidx];
 691         /* Update potential energies */
 692         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 693
 694         /* Increment number of inner iterations */
 695         inneriter                  += j_index_end - j_index_start;
 696
 697         /* Outer loop uses 19 flops */
 698     }
 699
 700     /* Increment number of outer iterations */
 701     outeriter        += nri;
 702
 703     /* Update outer/inner flops */
 704
 705     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*135);
 706 }
 707 /*
 708  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW4P1_F_avx_256_single
 709  * Electrostatics interaction: CubicSplineTable
 710  * VdW interaction:            None
 711  * Geometry:                   Water4-Particle
 712  * Calculate force/pot:        Force
 713  */
 714 void
 715 nb_kernel_ElecCSTab_VdwNone_GeomW4P1_F_avx_256_single
 716                     (t_nblist                    * gmx_restrict       nlist,
 717                      rvec                        * gmx_restrict          xx,
 718                      rvec                        * gmx_restrict          ff,
 719                      t_forcerec                  * gmx_restrict          fr,
 720                      t_mdatoms                   * gmx_restrict     mdatoms,
 721                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 722                      t_nrnb                      * gmx_restrict        nrnb)
 723 {
 724     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 725      * just 0 for non-waters.
 726      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 727      * jnr indices corresponding to data put in the four positions in the SIMD register.
 728      */
 729     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 730     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 731     int              jnrA,jnrB,jnrC,jnrD;
 732     int              jnrE,jnrF,jnrG,jnrH;
 733     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 734     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 735     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 736     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 737     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 738     real             rcutoff_scalar;
 739     real             *shiftvec,*fshift,*x,*f;
 740     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 741     real             scratch[4*DIM];
 742     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 743     real *           vdwioffsetptr1;
 744     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 745     real *           vdwioffsetptr2;
 746     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 747     real *           vdwioffsetptr3;
 748     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 749     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
 750     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 751     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 752     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 753     __m256           dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 754     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 755     real             *charge;
 756     __m256i          vfitab;
 757     __m128i          vfitab_lo,vfitab_hi;
 758     __m128i          ifour       = _mm_set1_epi32(4);
 759     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 760     real             *vftab;
 761     __m256           dummy_mask,cutoff_mask;
 762     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 763     __m256           one     = _mm256_set1_ps(1.0);
 764     __m256           two     = _mm256_set1_ps(2.0);
 765     x                = xx[0];
 766     f                = ff[0];
 767
 768     nri              = nlist->nri;
 769     iinr             = nlist->iinr;
 770     jindex           = nlist->jindex;
 771     jjnr             = nlist->jjnr;
 772     shiftidx         = nlist->shift;
 773     gid              = nlist->gid;
 774     shiftvec         = fr->shift_vec[0];
 775     fshift           = fr->fshift[0];
 776     facel            = _mm256_set1_ps(fr->epsfac);
 777     charge           = mdatoms->chargeA;
 778
 779     vftab            = kernel_data->table_elec->data;
 780     vftabscale       = _mm256_set1_ps(kernel_data->table_elec->scale);
 781
 782     /* Setup water-specific parameters */
 783     inr              = nlist->iinr[0];
 784     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 785     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 786     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 787
 788     /* Avoid stupid compiler warnings */
 789     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 790     j_coord_offsetA = 0;
 791     j_coord_offsetB = 0;
 792     j_coord_offsetC = 0;
 793     j_coord_offsetD = 0;
 794     j_coord_offsetE = 0;
 795     j_coord_offsetF = 0;
 796     j_coord_offsetG = 0;
 797     j_coord_offsetH = 0;
 798
 799     outeriter        = 0;
 800     inneriter        = 0;
 801
 802     for(iidx=0;iidx<4*DIM;iidx++)
 803     {
 804         scratch[iidx] = 0.0;
 805     }
 806
 807     /* Start outer loop over neighborlists */
 808     for(iidx=0; iidx<nri; iidx++)
 809     {
 810         /* Load shift vector for this list */
 811         i_shift_offset   = DIM*shiftidx[iidx];
 812
 813         /* Load limits for loop over neighbors */
 814         j_index_start    = jindex[iidx];
 815         j_index_end      = jindex[iidx+1];
 816
 817         /* Get outer coordinate index */
 818         inr              = iinr[iidx];
 819         i_coord_offset   = DIM*inr;
 820
 821         /* Load i particle coords and add shift vector */
 822         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 823                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 824
 825         fix1             = _mm256_setzero_ps();
 826         fiy1             = _mm256_setzero_ps();
 827         fiz1             = _mm256_setzero_ps();
 828         fix2             = _mm256_setzero_ps();
 829         fiy2             = _mm256_setzero_ps();
 830         fiz2             = _mm256_setzero_ps();
 831         fix3             = _mm256_setzero_ps();
 832         fiy3             = _mm256_setzero_ps();
 833         fiz3             = _mm256_setzero_ps();
 834
 835         /* Start inner kernel loop */
 836         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 837         {
 838
 839             /* Get j neighbor index, and coordinate index */
 840             jnrA             = jjnr[jidx];
 841             jnrB             = jjnr[jidx+1];
 842             jnrC             = jjnr[jidx+2];
 843             jnrD             = jjnr[jidx+3];
 844             jnrE             = jjnr[jidx+4];
 845             jnrF             = jjnr[jidx+5];
 846             jnrG             = jjnr[jidx+6];
 847             jnrH             = jjnr[jidx+7];
 848             j_coord_offsetA  = DIM*jnrA;
 849             j_coord_offsetB  = DIM*jnrB;
 850             j_coord_offsetC  = DIM*jnrC;
 851             j_coord_offsetD  = DIM*jnrD;
 852             j_coord_offsetE  = DIM*jnrE;
 853             j_coord_offsetF  = DIM*jnrF;
 854             j_coord_offsetG  = DIM*jnrG;
 855             j_coord_offsetH  = DIM*jnrH;
 856
 857             /* load j atom coordinates */
 858             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 859                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 860                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 861                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 862                                                  &jx0,&jy0,&jz0);
 863
 864             /* Calculate displacement vector */
 865             dx10             = _mm256_sub_ps(ix1,jx0);
 866             dy10             = _mm256_sub_ps(iy1,jy0);
 867             dz10             = _mm256_sub_ps(iz1,jz0);
 868             dx20             = _mm256_sub_ps(ix2,jx0);
 869             dy20             = _mm256_sub_ps(iy2,jy0);
 870             dz20             = _mm256_sub_ps(iz2,jz0);
 871             dx30             = _mm256_sub_ps(ix3,jx0);
 872             dy30             = _mm256_sub_ps(iy3,jy0);
 873             dz30             = _mm256_sub_ps(iz3,jz0);
 874
 875             /* Calculate squared distance and things based on it */
 876             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 877             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 878             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
 879
 880             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 881             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 882             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
 883
 884             /* Load parameters for j particles */
 885             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 886                                                                  charge+jnrC+0,charge+jnrD+0,
 887                                                                  charge+jnrE+0,charge+jnrF+0,
 888                                                                  charge+jnrG+0,charge+jnrH+0);
 889
 890             fjx0             = _mm256_setzero_ps();
 891             fjy0             = _mm256_setzero_ps();
 892             fjz0             = _mm256_setzero_ps();
 893
 894             /**************************
 895              * CALCULATE INTERACTIONS *
 896              **************************/
 897
 898             r10              = _mm256_mul_ps(rsq10,rinv10);
 899
 900             /* Compute parameters for interactions between i and j atoms */
 901             qq10             = _mm256_mul_ps(iq1,jq0);
 902
 903             /* Calculate table index by multiplying r with table scale and truncate to integer */
 904             rt               = _mm256_mul_ps(r10,vftabscale);
 905             vfitab           = _mm256_cvttps_epi32(rt);
 906             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 907             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 908             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 909             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 910             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 911             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 912
 913             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 914             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 915                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 916             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 917                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 918             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 919                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 920             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 921                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 922             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 923             Heps             = _mm256_mul_ps(vfeps,H);
 924             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 925             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 926             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq10,FF),_mm256_mul_ps(vftabscale,rinv10)));
 927
 928             fscal            = felec;
 929
 930             /* Calculate temporary vectorial force */
 931             tx               = _mm256_mul_ps(fscal,dx10);
 932             ty               = _mm256_mul_ps(fscal,dy10);
 933             tz               = _mm256_mul_ps(fscal,dz10);
 934
 935             /* Update vectorial force */
 936             fix1             = _mm256_add_ps(fix1,tx);
 937             fiy1             = _mm256_add_ps(fiy1,ty);
 938             fiz1             = _mm256_add_ps(fiz1,tz);
 939
 940             fjx0             = _mm256_add_ps(fjx0,tx);
 941             fjy0             = _mm256_add_ps(fjy0,ty);
 942             fjz0             = _mm256_add_ps(fjz0,tz);
 943
 944             /**************************
 945              * CALCULATE INTERACTIONS *
 946              **************************/
 947
 948             r20              = _mm256_mul_ps(rsq20,rinv20);
 949
 950             /* Compute parameters for interactions between i and j atoms */
 951             qq20             = _mm256_mul_ps(iq2,jq0);
 952
 953             /* Calculate table index by multiplying r with table scale and truncate to integer */
 954             rt               = _mm256_mul_ps(r20,vftabscale);
 955             vfitab           = _mm256_cvttps_epi32(rt);
 956             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 957             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 958             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 959             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 960             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
 961             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
 962
 963             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 964             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 965                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 966             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 967                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 968             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 969                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 970             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 971                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 972             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 973             Heps             = _mm256_mul_ps(vfeps,H);
 974             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 975             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 976             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq20,FF),_mm256_mul_ps(vftabscale,rinv20)));
 977
 978             fscal            = felec;
 979
 980             /* Calculate temporary vectorial force */
 981             tx               = _mm256_mul_ps(fscal,dx20);
 982             ty               = _mm256_mul_ps(fscal,dy20);
 983             tz               = _mm256_mul_ps(fscal,dz20);
 984
 985             /* Update vectorial force */
 986             fix2             = _mm256_add_ps(fix2,tx);
 987             fiy2             = _mm256_add_ps(fiy2,ty);
 988             fiz2             = _mm256_add_ps(fiz2,tz);
 989
 990             fjx0             = _mm256_add_ps(fjx0,tx);
 991             fjy0             = _mm256_add_ps(fjy0,ty);
 992             fjz0             = _mm256_add_ps(fjz0,tz);
 993
 994             /**************************
 995              * CALCULATE INTERACTIONS *
 996              **************************/
 997
 998             r30              = _mm256_mul_ps(rsq30,rinv30);
 999
1000             /* Compute parameters for interactions between i and j atoms */
1001             qq30             = _mm256_mul_ps(iq3,jq0);
1002
1003             /* Calculate table index by multiplying r with table scale and truncate to integer */
1004             rt               = _mm256_mul_ps(r30,vftabscale);
1005             vfitab           = _mm256_cvttps_epi32(rt);
1006             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1007             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1008             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1009             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1010             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
1011             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
1012
1013             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1014             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1015                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1016             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1017                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1018             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1019                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1020             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1021                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1022             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1023             Heps             = _mm256_mul_ps(vfeps,H);
1024             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1025             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1026             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq30,FF),_mm256_mul_ps(vftabscale,rinv30)));
1027
1028             fscal            = felec;
1029
1030             /* Calculate temporary vectorial force */
1031             tx               = _mm256_mul_ps(fscal,dx30);
1032             ty               = _mm256_mul_ps(fscal,dy30);
1033             tz               = _mm256_mul_ps(fscal,dz30);
1034
1035             /* Update vectorial force */
1036             fix3             = _mm256_add_ps(fix3,tx);
1037             fiy3             = _mm256_add_ps(fiy3,ty);
1038             fiz3             = _mm256_add_ps(fiz3,tz);
1039
1040             fjx0             = _mm256_add_ps(fjx0,tx);
1041             fjy0             = _mm256_add_ps(fjy0,ty);
1042             fjz0             = _mm256_add_ps(fjz0,tz);
1043
1044             fjptrA             = f+j_coord_offsetA;
1045             fjptrB             = f+j_coord_offsetB;
1046             fjptrC             = f+j_coord_offsetC;
1047             fjptrD             = f+j_coord_offsetD;
1048             fjptrE             = f+j_coord_offsetE;
1049             fjptrF             = f+j_coord_offsetF;
1050             fjptrG             = f+j_coord_offsetG;
1051             fjptrH             = f+j_coord_offsetH;
1052
1053             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1054
1055             /* Inner loop uses 120 flops */
1056         }
1057
1058         if(jidx<j_index_end)
1059         {
1060
1061             /* Get j neighbor index, and coordinate index */
1062             jnrlistA         = jjnr[jidx];
1063             jnrlistB         = jjnr[jidx+1];
1064             jnrlistC         = jjnr[jidx+2];
1065             jnrlistD         = jjnr[jidx+3];
1066             jnrlistE         = jjnr[jidx+4];
1067             jnrlistF         = jjnr[jidx+5];
1068             jnrlistG         = jjnr[jidx+6];
1069             jnrlistH         = jjnr[jidx+7];
1070             /* Sign of each element will be negative for non-real atoms.
1071              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1072              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1073              */
1074             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1075                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1076
1077             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1078             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1079             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1080             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1081             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1082             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1083             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1084             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1085             j_coord_offsetA  = DIM*jnrA;
1086             j_coord_offsetB  = DIM*jnrB;
1087             j_coord_offsetC  = DIM*jnrC;
1088             j_coord_offsetD  = DIM*jnrD;
1089             j_coord_offsetE  = DIM*jnrE;
1090             j_coord_offsetF  = DIM*jnrF;
1091             j_coord_offsetG  = DIM*jnrG;
1092             j_coord_offsetH  = DIM*jnrH;
1093
1094             /* load j atom coordinates */
1095             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1096                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1097                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1098                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1099                                                  &jx0,&jy0,&jz0);
1100
1101             /* Calculate displacement vector */
1102             dx10             = _mm256_sub_ps(ix1,jx0);
1103             dy10             = _mm256_sub_ps(iy1,jy0);
1104             dz10             = _mm256_sub_ps(iz1,jz0);
1105             dx20             = _mm256_sub_ps(ix2,jx0);
1106             dy20             = _mm256_sub_ps(iy2,jy0);
1107             dz20             = _mm256_sub_ps(iz2,jz0);
1108             dx30             = _mm256_sub_ps(ix3,jx0);
1109             dy30             = _mm256_sub_ps(iy3,jy0);
1110             dz30             = _mm256_sub_ps(iz3,jz0);
1111
1112             /* Calculate squared distance and things based on it */
1113             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1114             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1115             rsq30            = gmx_mm256_calc_rsq_ps(dx30,dy30,dz30);
1116
1117             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1118             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1119             rinv30           = gmx_mm256_invsqrt_ps(rsq30);
1120
1121             /* Load parameters for j particles */
1122             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1123                                                                  charge+jnrC+0,charge+jnrD+0,
1124                                                                  charge+jnrE+0,charge+jnrF+0,
1125                                                                  charge+jnrG+0,charge+jnrH+0);
1126
1127             fjx0             = _mm256_setzero_ps();
1128             fjy0             = _mm256_setzero_ps();
1129             fjz0             = _mm256_setzero_ps();
1130
1131             /**************************
1132              * CALCULATE INTERACTIONS *
1133              **************************/
1134
1135             r10              = _mm256_mul_ps(rsq10,rinv10);
1136             r10              = _mm256_andnot_ps(dummy_mask,r10);
1137
1138             /* Compute parameters for interactions between i and j atoms */
1139             qq10             = _mm256_mul_ps(iq1,jq0);
1140
1141             /* Calculate table index by multiplying r with table scale and truncate to integer */
1142             rt               = _mm256_mul_ps(r10,vftabscale);
1143             vfitab           = _mm256_cvttps_epi32(rt);
1144             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1145             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1146             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1147             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1148             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
1149             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
1150
1151             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1152             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1153                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1154             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1155                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1156             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1157                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1158             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1159                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1160             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1161             Heps             = _mm256_mul_ps(vfeps,H);
1162             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1163             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1164             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq10,FF),_mm256_mul_ps(vftabscale,rinv10)));
1165
1166             fscal            = felec;
1167
1168             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1169
1170             /* Calculate temporary vectorial force */
1171             tx               = _mm256_mul_ps(fscal,dx10);
1172             ty               = _mm256_mul_ps(fscal,dy10);
1173             tz               = _mm256_mul_ps(fscal,dz10);
1174
1175             /* Update vectorial force */
1176             fix1             = _mm256_add_ps(fix1,tx);
1177             fiy1             = _mm256_add_ps(fiy1,ty);
1178             fiz1             = _mm256_add_ps(fiz1,tz);
1179
1180             fjx0             = _mm256_add_ps(fjx0,tx);
1181             fjy0             = _mm256_add_ps(fjy0,ty);
1182             fjz0             = _mm256_add_ps(fjz0,tz);
1183
1184             /**************************
1185              * CALCULATE INTERACTIONS *
1186              **************************/
1187
1188             r20              = _mm256_mul_ps(rsq20,rinv20);
1189             r20              = _mm256_andnot_ps(dummy_mask,r20);
1190
1191             /* Compute parameters for interactions between i and j atoms */
1192             qq20             = _mm256_mul_ps(iq2,jq0);
1193
1194             /* Calculate table index by multiplying r with table scale and truncate to integer */
1195             rt               = _mm256_mul_ps(r20,vftabscale);
1196             vfitab           = _mm256_cvttps_epi32(rt);
1197             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1198             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1199             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1200             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1201             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
1202             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
1203
1204             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1205             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1206                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1207             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1208                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1209             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1210                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1211             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1212                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1213             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1214             Heps             = _mm256_mul_ps(vfeps,H);
1215             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1216             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1217             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq20,FF),_mm256_mul_ps(vftabscale,rinv20)));
1218
1219             fscal            = felec;
1220
1221             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1222
1223             /* Calculate temporary vectorial force */
1224             tx               = _mm256_mul_ps(fscal,dx20);
1225             ty               = _mm256_mul_ps(fscal,dy20);
1226             tz               = _mm256_mul_ps(fscal,dz20);
1227
1228             /* Update vectorial force */
1229             fix2             = _mm256_add_ps(fix2,tx);
1230             fiy2             = _mm256_add_ps(fiy2,ty);
1231             fiz2             = _mm256_add_ps(fiz2,tz);
1232
1233             fjx0             = _mm256_add_ps(fjx0,tx);
1234             fjy0             = _mm256_add_ps(fjy0,ty);
1235             fjz0             = _mm256_add_ps(fjz0,tz);
1236
1237             /**************************
1238              * CALCULATE INTERACTIONS *
1239              **************************/
1240
1241             r30              = _mm256_mul_ps(rsq30,rinv30);
1242             r30              = _mm256_andnot_ps(dummy_mask,r30);
1243
1244             /* Compute parameters for interactions between i and j atoms */
1245             qq30             = _mm256_mul_ps(iq3,jq0);
1246
1247             /* Calculate table index by multiplying r with table scale and truncate to integer */
1248             rt               = _mm256_mul_ps(r30,vftabscale);
1249             vfitab           = _mm256_cvttps_epi32(rt);
1250             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
1251             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
1252             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
1253             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
1254             vfitab_lo        = _mm_slli_epi32(vfitab_lo,2);
1255             vfitab_hi        = _mm_slli_epi32(vfitab_hi,2);
1256
1257             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1258             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
1259                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
1260             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
1261                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
1262             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
1263                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
1264             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
1265                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
1266             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
1267             Heps             = _mm256_mul_ps(vfeps,H);
1268             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
1269             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
1270             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq30,FF),_mm256_mul_ps(vftabscale,rinv30)));
1271
1272             fscal            = felec;
1273
1274             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1275
1276             /* Calculate temporary vectorial force */
1277             tx               = _mm256_mul_ps(fscal,dx30);
1278             ty               = _mm256_mul_ps(fscal,dy30);
1279             tz               = _mm256_mul_ps(fscal,dz30);
1280
1281             /* Update vectorial force */
1282             fix3             = _mm256_add_ps(fix3,tx);
1283             fiy3             = _mm256_add_ps(fiy3,ty);
1284             fiz3             = _mm256_add_ps(fiz3,tz);
1285
1286             fjx0             = _mm256_add_ps(fjx0,tx);
1287             fjy0             = _mm256_add_ps(fjy0,ty);
1288             fjz0             = _mm256_add_ps(fjz0,tz);
1289
1290             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1291             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1292             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1293             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1294             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1295             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1296             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1297             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1298
1299             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1300
1301             /* Inner loop uses 123 flops */
1302         }
1303
1304         /* End of innermost loop */
1305
1306         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1307                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1308
1309         /* Increment number of inner iterations */
1310         inneriter                  += j_index_end - j_index_start;
1311
1312         /* Outer loop uses 18 flops */
1313     }
1314
1315     /* Increment number of outer iterations */
1316     outeriter        += nri;
1317
1318     /* Update outer/inner flops */
1319
1320     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*123);
1321 }