src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCSTab_VdwCSTab_GeomP1P1_avx_256_single.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_avx_256_single.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomP1P1_VF_avx_256_single
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwCSTab_GeomP1P1_VF_avx_256_single
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrE,jnrF,jnrG,jnrH;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  79     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  80     real             rcutoff_scalar;
  81     real             *shiftvec,*fshift,*x,*f;
  82     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  83     real             scratch[4*DIM];
  84     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  85     real *           vdwioffsetptr0;
  86     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  87     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  88     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
  97     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
  98     __m256i          vfitab;
  99     __m128i          vfitab_lo,vfitab_hi;
 100     __m128i          ifour       = _mm_set1_epi32(4);
 101     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 102     real             *vftab;
 103     __m256           dummy_mask,cutoff_mask;
 104     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 105     __m256           one     = _mm256_set1_ps(1.0);
 106     __m256           two     = _mm256_set1_ps(2.0);
 107     x                = xx[0];
 108     f                = ff[0];
 109
 110     nri              = nlist->nri;
 111     iinr             = nlist->iinr;
 112     jindex           = nlist->jindex;
 113     jjnr             = nlist->jjnr;
 114     shiftidx         = nlist->shift;
 115     gid              = nlist->gid;
 116     shiftvec         = fr->shift_vec[0];
 117     fshift           = fr->fshift[0];
 118     facel            = _mm256_set1_ps(fr->ic->epsfac);
 119     charge           = mdatoms->chargeA;
 120     nvdwtype         = fr->ntype;
 121     vdwparam         = fr->nbfp;
 122     vdwtype          = mdatoms->typeA;
 123
 124     vftab            = kernel_data->table_elec_vdw->data;
 125     vftabscale       = _mm256_set1_ps(kernel_data->table_elec_vdw->scale);
 126
 127     /* Avoid stupid compiler warnings */
 128     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 129     j_coord_offsetA = 0;
 130     j_coord_offsetB = 0;
 131     j_coord_offsetC = 0;
 132     j_coord_offsetD = 0;
 133     j_coord_offsetE = 0;
 134     j_coord_offsetF = 0;
 135     j_coord_offsetG = 0;
 136     j_coord_offsetH = 0;
 137
 138     outeriter        = 0;
 139     inneriter        = 0;
 140
 141     for(iidx=0;iidx<4*DIM;iidx++)
 142     {
 143         scratch[iidx] = 0.0;
 144     }
 145
 146     /* Start outer loop over neighborlists */
 147     for(iidx=0; iidx<nri; iidx++)
 148     {
 149         /* Load shift vector for this list */
 150         i_shift_offset   = DIM*shiftidx[iidx];
 151
 152         /* Load limits for loop over neighbors */
 153         j_index_start    = jindex[iidx];
 154         j_index_end      = jindex[iidx+1];
 155
 156         /* Get outer coordinate index */
 157         inr              = iinr[iidx];
 158         i_coord_offset   = DIM*inr;
 159
 160         /* Load i particle coords and add shift vector */
 161         gmx_mm256_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 162
 163         fix0             = _mm256_setzero_ps();
 164         fiy0             = _mm256_setzero_ps();
 165         fiz0             = _mm256_setzero_ps();
 166
 167         /* Load parameters for i particles */
 168         iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 169         vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 170
 171         /* Reset potential sums */
 172         velecsum         = _mm256_setzero_ps();
 173         vvdwsum          = _mm256_setzero_ps();
 174
 175         /* Start inner kernel loop */
 176         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 177         {
 178
 179             /* Get j neighbor index, and coordinate index */
 180             jnrA             = jjnr[jidx];
 181             jnrB             = jjnr[jidx+1];
 182             jnrC             = jjnr[jidx+2];
 183             jnrD             = jjnr[jidx+3];
 184             jnrE             = jjnr[jidx+4];
 185             jnrF             = jjnr[jidx+5];
 186             jnrG             = jjnr[jidx+6];
 187             jnrH             = jjnr[jidx+7];
 188             j_coord_offsetA  = DIM*jnrA;
 189             j_coord_offsetB  = DIM*jnrB;
 190             j_coord_offsetC  = DIM*jnrC;
 191             j_coord_offsetD  = DIM*jnrD;
 192             j_coord_offsetE  = DIM*jnrE;
 193             j_coord_offsetF  = DIM*jnrF;
 194             j_coord_offsetG  = DIM*jnrG;
 195             j_coord_offsetH  = DIM*jnrH;
 196
 197             /* load j atom coordinates */
 198             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 199                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 200                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 201                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 202                                                  &jx0,&jy0,&jz0);
 203
 204             /* Calculate displacement vector */
 205             dx00             = _mm256_sub_ps(ix0,jx0);
 206             dy00             = _mm256_sub_ps(iy0,jy0);
 207             dz00             = _mm256_sub_ps(iz0,jz0);
 208
 209             /* Calculate squared distance and things based on it */
 210             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 211
 212             rinv00           = avx256_invsqrt_f(rsq00);
 213
 214             /* Load parameters for j particles */
 215             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 216                                                                  charge+jnrC+0,charge+jnrD+0,
 217                                                                  charge+jnrE+0,charge+jnrF+0,
 218                                                                  charge+jnrG+0,charge+jnrH+0);
 219             vdwjidx0A        = 2*vdwtype[jnrA+0];
 220             vdwjidx0B        = 2*vdwtype[jnrB+0];
 221             vdwjidx0C        = 2*vdwtype[jnrC+0];
 222             vdwjidx0D        = 2*vdwtype[jnrD+0];
 223             vdwjidx0E        = 2*vdwtype[jnrE+0];
 224             vdwjidx0F        = 2*vdwtype[jnrF+0];
 225             vdwjidx0G        = 2*vdwtype[jnrG+0];
 226             vdwjidx0H        = 2*vdwtype[jnrH+0];
 227
 228             /**************************
 229              * CALCULATE INTERACTIONS *
 230              **************************/
 231
 232             r00              = _mm256_mul_ps(rsq00,rinv00);
 233
 234             /* Compute parameters for interactions between i and j atoms */
 235             qq00             = _mm256_mul_ps(iq0,jq0);
 236             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 237                                             vdwioffsetptr0+vdwjidx0B,
 238                                             vdwioffsetptr0+vdwjidx0C,
 239                                             vdwioffsetptr0+vdwjidx0D,
 240                                             vdwioffsetptr0+vdwjidx0E,
 241                                             vdwioffsetptr0+vdwjidx0F,
 242                                             vdwioffsetptr0+vdwjidx0G,
 243                                             vdwioffsetptr0+vdwjidx0H,
 244                                             &c6_00,&c12_00);
 245
 246             /* Calculate table index by multiplying r with table scale and truncate to integer */
 247             rt               = _mm256_mul_ps(r00,vftabscale);
 248             vfitab           = _mm256_cvttps_epi32(rt);
 249             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 250             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 251             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 252             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 253             vfitab_lo        = _mm_slli_epi32(_mm_add_epi32(vfitab_lo,_mm_slli_epi32(vfitab_lo,1)),2);
 254             vfitab_hi        = _mm_slli_epi32(_mm_add_epi32(vfitab_hi,_mm_slli_epi32(vfitab_hi,1)),2);
 255
 256             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 257             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 258                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 259             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 260                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 261             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 262                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 263             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 264                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 265             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 266             Heps             = _mm256_mul_ps(vfeps,H);
 267             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 268             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 269             velec            = _mm256_mul_ps(qq00,VV);
 270             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 271             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq00,FF),_mm256_mul_ps(vftabscale,rinv00)));
 272
 273             /* CUBIC SPLINE TABLE DISPERSION */
 274             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 275             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 276             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 277                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 278             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 279                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 280             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 281                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 282             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 283                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 284             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 285             Heps             = _mm256_mul_ps(vfeps,H);
 286             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 287             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 288             vvdw6            = _mm256_mul_ps(c6_00,VV);
 289             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 290             fvdw6            = _mm256_mul_ps(c6_00,FF);
 291
 292             /* CUBIC SPLINE TABLE REPULSION */
 293             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 294             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 295             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 296                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 297             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 298                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 299             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 300                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 301             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 302                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 303             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 304             Heps             = _mm256_mul_ps(vfeps,H);
 305             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 306             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 307             vvdw12           = _mm256_mul_ps(c12_00,VV);
 308             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 309             fvdw12           = _mm256_mul_ps(c12_00,FF);
 310             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 311             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 312
 313             /* Update potential sum for this i atom from the interaction with this j atom. */
 314             velecsum         = _mm256_add_ps(velecsum,velec);
 315             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 316
 317             fscal            = _mm256_add_ps(felec,fvdw);
 318
 319             /* Calculate temporary vectorial force */
 320             tx               = _mm256_mul_ps(fscal,dx00);
 321             ty               = _mm256_mul_ps(fscal,dy00);
 322             tz               = _mm256_mul_ps(fscal,dz00);
 323
 324             /* Update vectorial force */
 325             fix0             = _mm256_add_ps(fix0,tx);
 326             fiy0             = _mm256_add_ps(fiy0,ty);
 327             fiz0             = _mm256_add_ps(fiz0,tz);
 328
 329             fjptrA             = f+j_coord_offsetA;
 330             fjptrB             = f+j_coord_offsetB;
 331             fjptrC             = f+j_coord_offsetC;
 332             fjptrD             = f+j_coord_offsetD;
 333             fjptrE             = f+j_coord_offsetE;
 334             fjptrF             = f+j_coord_offsetF;
 335             fjptrG             = f+j_coord_offsetG;
 336             fjptrH             = f+j_coord_offsetH;
 337             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
 338
 339             /* Inner loop uses 73 flops */
 340         }
 341
 342         if(jidx<j_index_end)
 343         {
 344
 345             /* Get j neighbor index, and coordinate index */
 346             jnrlistA         = jjnr[jidx];
 347             jnrlistB         = jjnr[jidx+1];
 348             jnrlistC         = jjnr[jidx+2];
 349             jnrlistD         = jjnr[jidx+3];
 350             jnrlistE         = jjnr[jidx+4];
 351             jnrlistF         = jjnr[jidx+5];
 352             jnrlistG         = jjnr[jidx+6];
 353             jnrlistH         = jjnr[jidx+7];
 354             /* Sign of each element will be negative for non-real atoms.
 355              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 356              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 357              */
 358             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 359                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 360
 361             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 362             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 363             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 364             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 365             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 366             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 367             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 368             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 369             j_coord_offsetA  = DIM*jnrA;
 370             j_coord_offsetB  = DIM*jnrB;
 371             j_coord_offsetC  = DIM*jnrC;
 372             j_coord_offsetD  = DIM*jnrD;
 373             j_coord_offsetE  = DIM*jnrE;
 374             j_coord_offsetF  = DIM*jnrF;
 375             j_coord_offsetG  = DIM*jnrG;
 376             j_coord_offsetH  = DIM*jnrH;
 377
 378             /* load j atom coordinates */
 379             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 380                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 381                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 382                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 383                                                  &jx0,&jy0,&jz0);
 384
 385             /* Calculate displacement vector */
 386             dx00             = _mm256_sub_ps(ix0,jx0);
 387             dy00             = _mm256_sub_ps(iy0,jy0);
 388             dz00             = _mm256_sub_ps(iz0,jz0);
 389
 390             /* Calculate squared distance and things based on it */
 391             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 392
 393             rinv00           = avx256_invsqrt_f(rsq00);
 394
 395             /* Load parameters for j particles */
 396             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 397                                                                  charge+jnrC+0,charge+jnrD+0,
 398                                                                  charge+jnrE+0,charge+jnrF+0,
 399                                                                  charge+jnrG+0,charge+jnrH+0);
 400             vdwjidx0A        = 2*vdwtype[jnrA+0];
 401             vdwjidx0B        = 2*vdwtype[jnrB+0];
 402             vdwjidx0C        = 2*vdwtype[jnrC+0];
 403             vdwjidx0D        = 2*vdwtype[jnrD+0];
 404             vdwjidx0E        = 2*vdwtype[jnrE+0];
 405             vdwjidx0F        = 2*vdwtype[jnrF+0];
 406             vdwjidx0G        = 2*vdwtype[jnrG+0];
 407             vdwjidx0H        = 2*vdwtype[jnrH+0];
 408
 409             /**************************
 410              * CALCULATE INTERACTIONS *
 411              **************************/
 412
 413             r00              = _mm256_mul_ps(rsq00,rinv00);
 414             r00              = _mm256_andnot_ps(dummy_mask,r00);
 415
 416             /* Compute parameters for interactions between i and j atoms */
 417             qq00             = _mm256_mul_ps(iq0,jq0);
 418             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 419                                             vdwioffsetptr0+vdwjidx0B,
 420                                             vdwioffsetptr0+vdwjidx0C,
 421                                             vdwioffsetptr0+vdwjidx0D,
 422                                             vdwioffsetptr0+vdwjidx0E,
 423                                             vdwioffsetptr0+vdwjidx0F,
 424                                             vdwioffsetptr0+vdwjidx0G,
 425                                             vdwioffsetptr0+vdwjidx0H,
 426                                             &c6_00,&c12_00);
 427
 428             /* Calculate table index by multiplying r with table scale and truncate to integer */
 429             rt               = _mm256_mul_ps(r00,vftabscale);
 430             vfitab           = _mm256_cvttps_epi32(rt);
 431             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 432             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 433             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 434             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 435             vfitab_lo        = _mm_slli_epi32(_mm_add_epi32(vfitab_lo,_mm_slli_epi32(vfitab_lo,1)),2);
 436             vfitab_hi        = _mm_slli_epi32(_mm_add_epi32(vfitab_hi,_mm_slli_epi32(vfitab_hi,1)),2);
 437
 438             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 439             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 440                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 441             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 442                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 443             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 444                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 445             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 446                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 447             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 448             Heps             = _mm256_mul_ps(vfeps,H);
 449             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 450             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 451             velec            = _mm256_mul_ps(qq00,VV);
 452             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 453             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq00,FF),_mm256_mul_ps(vftabscale,rinv00)));
 454
 455             /* CUBIC SPLINE TABLE DISPERSION */
 456             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 457             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 458             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 459                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 460             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 461                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 462             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 463                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 464             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 465                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 466             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 467             Heps             = _mm256_mul_ps(vfeps,H);
 468             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 469             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 470             vvdw6            = _mm256_mul_ps(c6_00,VV);
 471             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 472             fvdw6            = _mm256_mul_ps(c6_00,FF);
 473
 474             /* CUBIC SPLINE TABLE REPULSION */
 475             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 476             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 477             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 478                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 479             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 480                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 481             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 482                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 483             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 484                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 485             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 486             Heps             = _mm256_mul_ps(vfeps,H);
 487             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 488             VV               = _mm256_add_ps(Y,_mm256_mul_ps(vfeps,Fp));
 489             vvdw12           = _mm256_mul_ps(c12_00,VV);
 490             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 491             fvdw12           = _mm256_mul_ps(c12_00,FF);
 492             vvdw             = _mm256_add_ps(vvdw12,vvdw6);
 493             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 494
 495             /* Update potential sum for this i atom from the interaction with this j atom. */
 496             velec            = _mm256_andnot_ps(dummy_mask,velec);
 497             velecsum         = _mm256_add_ps(velecsum,velec);
 498             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 499             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 500
 501             fscal            = _mm256_add_ps(felec,fvdw);
 502
 503             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 504
 505             /* Calculate temporary vectorial force */
 506             tx               = _mm256_mul_ps(fscal,dx00);
 507             ty               = _mm256_mul_ps(fscal,dy00);
 508             tz               = _mm256_mul_ps(fscal,dz00);
 509
 510             /* Update vectorial force */
 511             fix0             = _mm256_add_ps(fix0,tx);
 512             fiy0             = _mm256_add_ps(fiy0,ty);
 513             fiz0             = _mm256_add_ps(fiz0,tz);
 514
 515             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 516             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 517             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 518             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 519             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 520             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 521             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 522             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 523             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
 524
 525             /* Inner loop uses 74 flops */
 526         }
 527
 528         /* End of innermost loop */
 529
 530         gmx_mm256_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 531                                                  f+i_coord_offset,fshift+i_shift_offset);
 532
 533         ggid                        = gid[iidx];
 534         /* Update potential energies */
 535         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 536         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 537
 538         /* Increment number of inner iterations */
 539         inneriter                  += j_index_end - j_index_start;
 540
 541         /* Outer loop uses 9 flops */
 542     }
 543
 544     /* Increment number of outer iterations */
 545     outeriter        += nri;
 546
 547     /* Update outer/inner flops */
 548
 549     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*74);
 550 }
 551 /*
 552  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomP1P1_F_avx_256_single
 553  * Electrostatics interaction: CubicSplineTable
 554  * VdW interaction:            CubicSplineTable
 555  * Geometry:                   Particle-Particle
 556  * Calculate force/pot:        Force
 557  */
 558 void
 559 nb_kernel_ElecCSTab_VdwCSTab_GeomP1P1_F_avx_256_single
 560                     (t_nblist                    * gmx_restrict       nlist,
 561                      rvec                        * gmx_restrict          xx,
 562                      rvec                        * gmx_restrict          ff,
 563                      struct t_forcerec           * gmx_restrict          fr,
 564                      t_mdatoms                   * gmx_restrict     mdatoms,
 565                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 566                      t_nrnb                      * gmx_restrict        nrnb)
 567 {
 568     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 569      * just 0 for non-waters.
 570      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 571      * jnr indices corresponding to data put in the four positions in the SIMD register.
 572      */
 573     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 574     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 575     int              jnrA,jnrB,jnrC,jnrD;
 576     int              jnrE,jnrF,jnrG,jnrH;
 577     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 578     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 579     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 580     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 581     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 582     real             rcutoff_scalar;
 583     real             *shiftvec,*fshift,*x,*f;
 584     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 585     real             scratch[4*DIM];
 586     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 587     real *           vdwioffsetptr0;
 588     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 589     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
 590     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 591     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 592     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 593     real             *charge;
 594     int              nvdwtype;
 595     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 596     int              *vdwtype;
 597     real             *vdwparam;
 598     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 599     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 600     __m256i          vfitab;
 601     __m128i          vfitab_lo,vfitab_hi;
 602     __m128i          ifour       = _mm_set1_epi32(4);
 603     __m256           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 604     real             *vftab;
 605     __m256           dummy_mask,cutoff_mask;
 606     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 607     __m256           one     = _mm256_set1_ps(1.0);
 608     __m256           two     = _mm256_set1_ps(2.0);
 609     x                = xx[0];
 610     f                = ff[0];
 611
 612     nri              = nlist->nri;
 613     iinr             = nlist->iinr;
 614     jindex           = nlist->jindex;
 615     jjnr             = nlist->jjnr;
 616     shiftidx         = nlist->shift;
 617     gid              = nlist->gid;
 618     shiftvec         = fr->shift_vec[0];
 619     fshift           = fr->fshift[0];
 620     facel            = _mm256_set1_ps(fr->ic->epsfac);
 621     charge           = mdatoms->chargeA;
 622     nvdwtype         = fr->ntype;
 623     vdwparam         = fr->nbfp;
 624     vdwtype          = mdatoms->typeA;
 625
 626     vftab            = kernel_data->table_elec_vdw->data;
 627     vftabscale       = _mm256_set1_ps(kernel_data->table_elec_vdw->scale);
 628
 629     /* Avoid stupid compiler warnings */
 630     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 631     j_coord_offsetA = 0;
 632     j_coord_offsetB = 0;
 633     j_coord_offsetC = 0;
 634     j_coord_offsetD = 0;
 635     j_coord_offsetE = 0;
 636     j_coord_offsetF = 0;
 637     j_coord_offsetG = 0;
 638     j_coord_offsetH = 0;
 639
 640     outeriter        = 0;
 641     inneriter        = 0;
 642
 643     for(iidx=0;iidx<4*DIM;iidx++)
 644     {
 645         scratch[iidx] = 0.0;
 646     }
 647
 648     /* Start outer loop over neighborlists */
 649     for(iidx=0; iidx<nri; iidx++)
 650     {
 651         /* Load shift vector for this list */
 652         i_shift_offset   = DIM*shiftidx[iidx];
 653
 654         /* Load limits for loop over neighbors */
 655         j_index_start    = jindex[iidx];
 656         j_index_end      = jindex[iidx+1];
 657
 658         /* Get outer coordinate index */
 659         inr              = iinr[iidx];
 660         i_coord_offset   = DIM*inr;
 661
 662         /* Load i particle coords and add shift vector */
 663         gmx_mm256_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 664
 665         fix0             = _mm256_setzero_ps();
 666         fiy0             = _mm256_setzero_ps();
 667         fiz0             = _mm256_setzero_ps();
 668
 669         /* Load parameters for i particles */
 670         iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 671         vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 672
 673         /* Start inner kernel loop */
 674         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 675         {
 676
 677             /* Get j neighbor index, and coordinate index */
 678             jnrA             = jjnr[jidx];
 679             jnrB             = jjnr[jidx+1];
 680             jnrC             = jjnr[jidx+2];
 681             jnrD             = jjnr[jidx+3];
 682             jnrE             = jjnr[jidx+4];
 683             jnrF             = jjnr[jidx+5];
 684             jnrG             = jjnr[jidx+6];
 685             jnrH             = jjnr[jidx+7];
 686             j_coord_offsetA  = DIM*jnrA;
 687             j_coord_offsetB  = DIM*jnrB;
 688             j_coord_offsetC  = DIM*jnrC;
 689             j_coord_offsetD  = DIM*jnrD;
 690             j_coord_offsetE  = DIM*jnrE;
 691             j_coord_offsetF  = DIM*jnrF;
 692             j_coord_offsetG  = DIM*jnrG;
 693             j_coord_offsetH  = DIM*jnrH;
 694
 695             /* load j atom coordinates */
 696             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 697                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 698                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 699                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 700                                                  &jx0,&jy0,&jz0);
 701
 702             /* Calculate displacement vector */
 703             dx00             = _mm256_sub_ps(ix0,jx0);
 704             dy00             = _mm256_sub_ps(iy0,jy0);
 705             dz00             = _mm256_sub_ps(iz0,jz0);
 706
 707             /* Calculate squared distance and things based on it */
 708             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 709
 710             rinv00           = avx256_invsqrt_f(rsq00);
 711
 712             /* Load parameters for j particles */
 713             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 714                                                                  charge+jnrC+0,charge+jnrD+0,
 715                                                                  charge+jnrE+0,charge+jnrF+0,
 716                                                                  charge+jnrG+0,charge+jnrH+0);
 717             vdwjidx0A        = 2*vdwtype[jnrA+0];
 718             vdwjidx0B        = 2*vdwtype[jnrB+0];
 719             vdwjidx0C        = 2*vdwtype[jnrC+0];
 720             vdwjidx0D        = 2*vdwtype[jnrD+0];
 721             vdwjidx0E        = 2*vdwtype[jnrE+0];
 722             vdwjidx0F        = 2*vdwtype[jnrF+0];
 723             vdwjidx0G        = 2*vdwtype[jnrG+0];
 724             vdwjidx0H        = 2*vdwtype[jnrH+0];
 725
 726             /**************************
 727              * CALCULATE INTERACTIONS *
 728              **************************/
 729
 730             r00              = _mm256_mul_ps(rsq00,rinv00);
 731
 732             /* Compute parameters for interactions between i and j atoms */
 733             qq00             = _mm256_mul_ps(iq0,jq0);
 734             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 735                                             vdwioffsetptr0+vdwjidx0B,
 736                                             vdwioffsetptr0+vdwjidx0C,
 737                                             vdwioffsetptr0+vdwjidx0D,
 738                                             vdwioffsetptr0+vdwjidx0E,
 739                                             vdwioffsetptr0+vdwjidx0F,
 740                                             vdwioffsetptr0+vdwjidx0G,
 741                                             vdwioffsetptr0+vdwjidx0H,
 742                                             &c6_00,&c12_00);
 743
 744             /* Calculate table index by multiplying r with table scale and truncate to integer */
 745             rt               = _mm256_mul_ps(r00,vftabscale);
 746             vfitab           = _mm256_cvttps_epi32(rt);
 747             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 748             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 749             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 750             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 751             vfitab_lo        = _mm_slli_epi32(_mm_add_epi32(vfitab_lo,_mm_slli_epi32(vfitab_lo,1)),2);
 752             vfitab_hi        = _mm_slli_epi32(_mm_add_epi32(vfitab_hi,_mm_slli_epi32(vfitab_hi,1)),2);
 753
 754             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 755             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 756                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 757             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 758                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 759             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 760                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 761             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 762                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 763             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 764             Heps             = _mm256_mul_ps(vfeps,H);
 765             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 766             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 767             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq00,FF),_mm256_mul_ps(vftabscale,rinv00)));
 768
 769             /* CUBIC SPLINE TABLE DISPERSION */
 770             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 771             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 772             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 773                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 774             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 775                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 776             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 777                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 778             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 779                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 780             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 781             Heps             = _mm256_mul_ps(vfeps,H);
 782             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 783             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 784             fvdw6            = _mm256_mul_ps(c6_00,FF);
 785
 786             /* CUBIC SPLINE TABLE REPULSION */
 787             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 788             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 789             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 790                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 791             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 792                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 793             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 794                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 795             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 796                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 797             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 798             Heps             = _mm256_mul_ps(vfeps,H);
 799             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 800             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 801             fvdw12           = _mm256_mul_ps(c12_00,FF);
 802             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 803
 804             fscal            = _mm256_add_ps(felec,fvdw);
 805
 806             /* Calculate temporary vectorial force */
 807             tx               = _mm256_mul_ps(fscal,dx00);
 808             ty               = _mm256_mul_ps(fscal,dy00);
 809             tz               = _mm256_mul_ps(fscal,dz00);
 810
 811             /* Update vectorial force */
 812             fix0             = _mm256_add_ps(fix0,tx);
 813             fiy0             = _mm256_add_ps(fiy0,ty);
 814             fiz0             = _mm256_add_ps(fiz0,tz);
 815
 816             fjptrA             = f+j_coord_offsetA;
 817             fjptrB             = f+j_coord_offsetB;
 818             fjptrC             = f+j_coord_offsetC;
 819             fjptrD             = f+j_coord_offsetD;
 820             fjptrE             = f+j_coord_offsetE;
 821             fjptrF             = f+j_coord_offsetF;
 822             fjptrG             = f+j_coord_offsetG;
 823             fjptrH             = f+j_coord_offsetH;
 824             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
 825
 826             /* Inner loop uses 61 flops */
 827         }
 828
 829         if(jidx<j_index_end)
 830         {
 831
 832             /* Get j neighbor index, and coordinate index */
 833             jnrlistA         = jjnr[jidx];
 834             jnrlistB         = jjnr[jidx+1];
 835             jnrlistC         = jjnr[jidx+2];
 836             jnrlistD         = jjnr[jidx+3];
 837             jnrlistE         = jjnr[jidx+4];
 838             jnrlistF         = jjnr[jidx+5];
 839             jnrlistG         = jjnr[jidx+6];
 840             jnrlistH         = jjnr[jidx+7];
 841             /* Sign of each element will be negative for non-real atoms.
 842              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 843              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 844              */
 845             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 846                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 847
 848             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 849             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 850             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 851             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 852             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 853             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 854             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 855             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 856             j_coord_offsetA  = DIM*jnrA;
 857             j_coord_offsetB  = DIM*jnrB;
 858             j_coord_offsetC  = DIM*jnrC;
 859             j_coord_offsetD  = DIM*jnrD;
 860             j_coord_offsetE  = DIM*jnrE;
 861             j_coord_offsetF  = DIM*jnrF;
 862             j_coord_offsetG  = DIM*jnrG;
 863             j_coord_offsetH  = DIM*jnrH;
 864
 865             /* load j atom coordinates */
 866             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 867                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 868                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 869                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 870                                                  &jx0,&jy0,&jz0);
 871
 872             /* Calculate displacement vector */
 873             dx00             = _mm256_sub_ps(ix0,jx0);
 874             dy00             = _mm256_sub_ps(iy0,jy0);
 875             dz00             = _mm256_sub_ps(iz0,jz0);
 876
 877             /* Calculate squared distance and things based on it */
 878             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 879
 880             rinv00           = avx256_invsqrt_f(rsq00);
 881
 882             /* Load parameters for j particles */
 883             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 884                                                                  charge+jnrC+0,charge+jnrD+0,
 885                                                                  charge+jnrE+0,charge+jnrF+0,
 886                                                                  charge+jnrG+0,charge+jnrH+0);
 887             vdwjidx0A        = 2*vdwtype[jnrA+0];
 888             vdwjidx0B        = 2*vdwtype[jnrB+0];
 889             vdwjidx0C        = 2*vdwtype[jnrC+0];
 890             vdwjidx0D        = 2*vdwtype[jnrD+0];
 891             vdwjidx0E        = 2*vdwtype[jnrE+0];
 892             vdwjidx0F        = 2*vdwtype[jnrF+0];
 893             vdwjidx0G        = 2*vdwtype[jnrG+0];
 894             vdwjidx0H        = 2*vdwtype[jnrH+0];
 895
 896             /**************************
 897              * CALCULATE INTERACTIONS *
 898              **************************/
 899
 900             r00              = _mm256_mul_ps(rsq00,rinv00);
 901             r00              = _mm256_andnot_ps(dummy_mask,r00);
 902
 903             /* Compute parameters for interactions between i and j atoms */
 904             qq00             = _mm256_mul_ps(iq0,jq0);
 905             gmx_mm256_load_8pair_swizzle_ps(vdwioffsetptr0+vdwjidx0A,
 906                                             vdwioffsetptr0+vdwjidx0B,
 907                                             vdwioffsetptr0+vdwjidx0C,
 908                                             vdwioffsetptr0+vdwjidx0D,
 909                                             vdwioffsetptr0+vdwjidx0E,
 910                                             vdwioffsetptr0+vdwjidx0F,
 911                                             vdwioffsetptr0+vdwjidx0G,
 912                                             vdwioffsetptr0+vdwjidx0H,
 913                                             &c6_00,&c12_00);
 914
 915             /* Calculate table index by multiplying r with table scale and truncate to integer */
 916             rt               = _mm256_mul_ps(r00,vftabscale);
 917             vfitab           = _mm256_cvttps_epi32(rt);
 918             vfeps            = _mm256_sub_ps(rt,_mm256_round_ps(rt, _MM_FROUND_FLOOR));
 919             /*         AVX1 does not support 256-bit integer operations, so now we go to 128-bit mode... */
 920             vfitab_lo        = _mm256_extractf128_si256(vfitab,0x0);
 921             vfitab_hi        = _mm256_extractf128_si256(vfitab,0x1);
 922             vfitab_lo        = _mm_slli_epi32(_mm_add_epi32(vfitab_lo,_mm_slli_epi32(vfitab_lo,1)),2);
 923             vfitab_hi        = _mm_slli_epi32(_mm_add_epi32(vfitab_hi,_mm_slli_epi32(vfitab_hi,1)),2);
 924
 925             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 926             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 927                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 928             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 929                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 930             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 931                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 932             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 933                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 934             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 935             Heps             = _mm256_mul_ps(vfeps,H);
 936             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 937             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 938             felec            = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_mul_ps(qq00,FF),_mm256_mul_ps(vftabscale,rinv00)));
 939
 940             /* CUBIC SPLINE TABLE DISPERSION */
 941             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 942             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 943             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 944                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 945             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 946                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 947             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 948                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 949             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 950                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 951             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 952             Heps             = _mm256_mul_ps(vfeps,H);
 953             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 954             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 955             fvdw6            = _mm256_mul_ps(c6_00,FF);
 956
 957             /* CUBIC SPLINE TABLE REPULSION */
 958             vfitab_lo        = _mm_add_epi32(vfitab_lo,ifour);
 959             vfitab_hi        = _mm_add_epi32(vfitab_hi,ifour);
 960             Y                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,0)),
 961                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,0)));
 962             F                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,1)),
 963                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,1)));
 964             G                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,2)),
 965                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,2)));
 966             H                = gmx_mm256_set_m128(_mm_load_ps(vftab + _mm_extract_epi32(vfitab_hi,3)),
 967                                                   _mm_load_ps(vftab + _mm_extract_epi32(vfitab_lo,3)));
 968             GMX_MM256_HALFTRANSPOSE4_PS(Y,F,G,H);
 969             Heps             = _mm256_mul_ps(vfeps,H);
 970             Fp               = _mm256_add_ps(F,_mm256_mul_ps(vfeps,_mm256_add_ps(G,Heps)));
 971             FF               = _mm256_add_ps(Fp,_mm256_mul_ps(vfeps,_mm256_add_ps(G,_mm256_add_ps(Heps,Heps))));
 972             fvdw12           = _mm256_mul_ps(c12_00,FF);
 973             fvdw             = _mm256_xor_ps(signbit,_mm256_mul_ps(_mm256_add_ps(fvdw6,fvdw12),_mm256_mul_ps(vftabscale,rinv00)));
 974
 975             fscal            = _mm256_add_ps(felec,fvdw);
 976
 977             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 978
 979             /* Calculate temporary vectorial force */
 980             tx               = _mm256_mul_ps(fscal,dx00);
 981             ty               = _mm256_mul_ps(fscal,dy00);
 982             tz               = _mm256_mul_ps(fscal,dz00);
 983
 984             /* Update vectorial force */
 985             fix0             = _mm256_add_ps(fix0,tx);
 986             fiy0             = _mm256_add_ps(fiy0,ty);
 987             fiz0             = _mm256_add_ps(fiz0,tz);
 988
 989             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 990             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 991             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 992             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 993             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 994             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 995             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 996             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 997             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,tx,ty,tz);
 998
 999             /* Inner loop uses 62 flops */
1000         }
1001
1002         /* End of innermost loop */
1003
1004         gmx_mm256_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
1005                                                  f+i_coord_offset,fshift+i_shift_offset);
1006
1007         /* Increment number of inner iterations */
1008         inneriter                  += j_index_end - j_index_start;
1009
1010         /* Outer loop uses 7 flops */
1011     }
1012
1013     /* Increment number of outer iterations */
1014     outeriter        += nri;
1015
1016     /* Update outer/inner flops */
1017
1018     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*62);
1019 }