src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecNone_VdwLJEw_GeomP1P1_avx_256_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_double.h"
  48 #include "kernelutil_x86_avx_256_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_avx_256_double
  52  * Electrostatics interaction: None
  53  * VdW interaction:            LJEwald
  54  * Geometry:                   Particle-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_VF_avx_256_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  82     real             scratch[4*DIM];
  83     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  84     real *           vdwioffsetptr0;
  85     real *           vdwgridioffsetptr0;
  86     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  87     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  88     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     int              nvdwtype;
  91     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  92     int              *vdwtype;
  93     real             *vdwparam;
  94     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
  95     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
  96     __m256d           c6grid_00;
  97     real             *vdwgridparam;
  98     __m256d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
  99     __m256d           one_half  = _mm256_set1_pd(0.5);
 100     __m256d           minus_one = _mm256_set1_pd(-1.0);
 101     __m256d          dummy_mask,cutoff_mask;
 102     __m128           tmpmask0,tmpmask1;
 103     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 104     __m256d          one     = _mm256_set1_pd(1.0);
 105     __m256d          two     = _mm256_set1_pd(2.0);
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     nvdwtype         = fr->ntype;
 118     vdwparam         = fr->nbfp;
 119     vdwtype          = mdatoms->typeA;
 120     vdwgridparam     = fr->ljpme_c6grid;
 121     sh_lj_ewald      = _mm256_set1_pd(fr->ic->sh_lj_ewald);
 122     ewclj            = _mm256_set1_pd(fr->ewaldcoeff_lj);
 123     ewclj2           = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
 124
 125     /* Avoid stupid compiler warnings */
 126     jnrA = jnrB = jnrC = jnrD = 0;
 127     j_coord_offsetA = 0;
 128     j_coord_offsetB = 0;
 129     j_coord_offsetC = 0;
 130     j_coord_offsetD = 0;
 131
 132     outeriter        = 0;
 133     inneriter        = 0;
 134
 135     for(iidx=0;iidx<4*DIM;iidx++)
 136     {
 137         scratch[iidx] = 0.0;
 138     }
 139
 140     /* Start outer loop over neighborlists */
 141     for(iidx=0; iidx<nri; iidx++)
 142     {
 143         /* Load shift vector for this list */
 144         i_shift_offset   = DIM*shiftidx[iidx];
 145
 146         /* Load limits for loop over neighbors */
 147         j_index_start    = jindex[iidx];
 148         j_index_end      = jindex[iidx+1];
 149
 150         /* Get outer coordinate index */
 151         inr              = iinr[iidx];
 152         i_coord_offset   = DIM*inr;
 153
 154         /* Load i particle coords and add shift vector */
 155         gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 156
 157         fix0             = _mm256_setzero_pd();
 158         fiy0             = _mm256_setzero_pd();
 159         fiz0             = _mm256_setzero_pd();
 160
 161         /* Load parameters for i particles */
 162         vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 163         vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
 164
 165         /* Reset potential sums */
 166         vvdwsum          = _mm256_setzero_pd();
 167
 168         /* Start inner kernel loop */
 169         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 170         {
 171
 172             /* Get j neighbor index, and coordinate index */
 173             jnrA             = jjnr[jidx];
 174             jnrB             = jjnr[jidx+1];
 175             jnrC             = jjnr[jidx+2];
 176             jnrD             = jjnr[jidx+3];
 177             j_coord_offsetA  = DIM*jnrA;
 178             j_coord_offsetB  = DIM*jnrB;
 179             j_coord_offsetC  = DIM*jnrC;
 180             j_coord_offsetD  = DIM*jnrD;
 181
 182             /* load j atom coordinates */
 183             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 184                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 185                                                  &jx0,&jy0,&jz0);
 186
 187             /* Calculate displacement vector */
 188             dx00             = _mm256_sub_pd(ix0,jx0);
 189             dy00             = _mm256_sub_pd(iy0,jy0);
 190             dz00             = _mm256_sub_pd(iz0,jz0);
 191
 192             /* Calculate squared distance and things based on it */
 193             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 194
 195             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 196
 197             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 198
 199             /* Load parameters for j particles */
 200             vdwjidx0A        = 2*vdwtype[jnrA+0];
 201             vdwjidx0B        = 2*vdwtype[jnrB+0];
 202             vdwjidx0C        = 2*vdwtype[jnrC+0];
 203             vdwjidx0D        = 2*vdwtype[jnrD+0];
 204
 205             /**************************
 206              * CALCULATE INTERACTIONS *
 207              **************************/
 208
 209             r00              = _mm256_mul_pd(rsq00,rinv00);
 210
 211             /* Compute parameters for interactions between i and j atoms */
 212             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
 213                                             vdwioffsetptr0+vdwjidx0B,
 214                                             vdwioffsetptr0+vdwjidx0C,
 215                                             vdwioffsetptr0+vdwjidx0D,
 216                                             &c6_00,&c12_00);
 217
 218             c6grid_00       = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
 219                                                                   vdwgridioffsetptr0+vdwjidx0B,
 220                                                                   vdwgridioffsetptr0+vdwjidx0C,
 221                                                                   vdwgridioffsetptr0+vdwjidx0D);
 222
 223             /* Analytical LJ-PME */
 224             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 225             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
 226             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
 227             exponent         = gmx_simd_exp_d(ewcljrsq);
 228             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 229             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 230             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 231             vvdw6            = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
 232             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 233             vvdw             = _mm256_sub_pd(_mm256_mul_pd(vvdw12,one_twelfth),_mm256_mul_pd(vvdw6,one_sixth));
 234             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 235             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
 236
 237             /* Update potential sum for this i atom from the interaction with this j atom. */
 238             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 239
 240             fscal            = fvdw;
 241
 242             /* Calculate temporary vectorial force */
 243             tx               = _mm256_mul_pd(fscal,dx00);
 244             ty               = _mm256_mul_pd(fscal,dy00);
 245             tz               = _mm256_mul_pd(fscal,dz00);
 246
 247             /* Update vectorial force */
 248             fix0             = _mm256_add_pd(fix0,tx);
 249             fiy0             = _mm256_add_pd(fiy0,ty);
 250             fiz0             = _mm256_add_pd(fiz0,tz);
 251
 252             fjptrA             = f+j_coord_offsetA;
 253             fjptrB             = f+j_coord_offsetB;
 254             fjptrC             = f+j_coord_offsetC;
 255             fjptrD             = f+j_coord_offsetD;
 256             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 257
 258             /* Inner loop uses 54 flops */
 259         }
 260
 261         if(jidx<j_index_end)
 262         {
 263
 264             /* Get j neighbor index, and coordinate index */
 265             jnrlistA         = jjnr[jidx];
 266             jnrlistB         = jjnr[jidx+1];
 267             jnrlistC         = jjnr[jidx+2];
 268             jnrlistD         = jjnr[jidx+3];
 269             /* Sign of each element will be negative for non-real atoms.
 270              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 271              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 272              */
 273             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 274
 275             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 276             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 277             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 278
 279             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 280             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 281             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 282             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 283             j_coord_offsetA  = DIM*jnrA;
 284             j_coord_offsetB  = DIM*jnrB;
 285             j_coord_offsetC  = DIM*jnrC;
 286             j_coord_offsetD  = DIM*jnrD;
 287
 288             /* load j atom coordinates */
 289             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 290                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 291                                                  &jx0,&jy0,&jz0);
 292
 293             /* Calculate displacement vector */
 294             dx00             = _mm256_sub_pd(ix0,jx0);
 295             dy00             = _mm256_sub_pd(iy0,jy0);
 296             dz00             = _mm256_sub_pd(iz0,jz0);
 297
 298             /* Calculate squared distance and things based on it */
 299             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 300
 301             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 302
 303             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 304
 305             /* Load parameters for j particles */
 306             vdwjidx0A        = 2*vdwtype[jnrA+0];
 307             vdwjidx0B        = 2*vdwtype[jnrB+0];
 308             vdwjidx0C        = 2*vdwtype[jnrC+0];
 309             vdwjidx0D        = 2*vdwtype[jnrD+0];
 310
 311             /**************************
 312              * CALCULATE INTERACTIONS *
 313              **************************/
 314
 315             r00              = _mm256_mul_pd(rsq00,rinv00);
 316             r00              = _mm256_andnot_pd(dummy_mask,r00);
 317
 318             /* Compute parameters for interactions between i and j atoms */
 319             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
 320                                             vdwioffsetptr0+vdwjidx0B,
 321                                             vdwioffsetptr0+vdwjidx0C,
 322                                             vdwioffsetptr0+vdwjidx0D,
 323                                             &c6_00,&c12_00);
 324
 325             c6grid_00       = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
 326                                                                   vdwgridioffsetptr0+vdwjidx0B,
 327                                                                   vdwgridioffsetptr0+vdwjidx0C,
 328                                                                   vdwgridioffsetptr0+vdwjidx0D);
 329
 330             /* Analytical LJ-PME */
 331             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 332             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
 333             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
 334             exponent         = gmx_simd_exp_d(ewcljrsq);
 335             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 336             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 337             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 338             vvdw6            = _mm256_mul_pd(_mm256_sub_pd(c6_00,_mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly))),rinvsix);
 339             vvdw12           = _mm256_mul_pd(c12_00,_mm256_mul_pd(rinvsix,rinvsix));
 340             vvdw             = _mm256_sub_pd(_mm256_mul_pd(vvdw12,one_twelfth),_mm256_mul_pd(vvdw6,one_sixth));
 341             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 342             fvdw             = _mm256_mul_pd(_mm256_sub_pd(vvdw12,_mm256_sub_pd(vvdw6,_mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6)))),rinvsq00);
 343
 344             /* Update potential sum for this i atom from the interaction with this j atom. */
 345             vvdw             = _mm256_andnot_pd(dummy_mask,vvdw);
 346             vvdwsum          = _mm256_add_pd(vvdwsum,vvdw);
 347
 348             fscal            = fvdw;
 349
 350             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 351
 352             /* Calculate temporary vectorial force */
 353             tx               = _mm256_mul_pd(fscal,dx00);
 354             ty               = _mm256_mul_pd(fscal,dy00);
 355             tz               = _mm256_mul_pd(fscal,dz00);
 356
 357             /* Update vectorial force */
 358             fix0             = _mm256_add_pd(fix0,tx);
 359             fiy0             = _mm256_add_pd(fiy0,ty);
 360             fiz0             = _mm256_add_pd(fiz0,tz);
 361
 362             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 363             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 364             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 365             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 366             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 367
 368             /* Inner loop uses 55 flops */
 369         }
 370
 371         /* End of innermost loop */
 372
 373         gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 374                                                  f+i_coord_offset,fshift+i_shift_offset);
 375
 376         ggid                        = gid[iidx];
 377         /* Update potential energies */
 378         gmx_mm256_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 379
 380         /* Increment number of inner iterations */
 381         inneriter                  += j_index_end - j_index_start;
 382
 383         /* Outer loop uses 7 flops */
 384     }
 385
 386     /* Increment number of outer iterations */
 387     outeriter        += nri;
 388
 389     /* Update outer/inner flops */
 390
 391     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_VF,outeriter*7 + inneriter*55);
 392 }
 393 /*
 394  * Gromacs nonbonded kernel:   nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_avx_256_double
 395  * Electrostatics interaction: None
 396  * VdW interaction:            LJEwald
 397  * Geometry:                   Particle-Particle
 398  * Calculate force/pot:        Force
 399  */
 400 void
 401 nb_kernel_ElecNone_VdwLJEw_GeomP1P1_F_avx_256_double
 402                     (t_nblist                    * gmx_restrict       nlist,
 403                      rvec                        * gmx_restrict          xx,
 404                      rvec                        * gmx_restrict          ff,
 405                      t_forcerec                  * gmx_restrict          fr,
 406                      t_mdatoms                   * gmx_restrict     mdatoms,
 407                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 408                      t_nrnb                      * gmx_restrict        nrnb)
 409 {
 410     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 411      * just 0 for non-waters.
 412      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
 413      * jnr indices corresponding to data put in the four positions in the SIMD register.
 414      */
 415     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 416     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 417     int              jnrA,jnrB,jnrC,jnrD;
 418     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 419     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 420     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 421     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 422     real             rcutoff_scalar;
 423     real             *shiftvec,*fshift,*x,*f;
 424     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 425     real             scratch[4*DIM];
 426     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 427     real *           vdwioffsetptr0;
 428     real *           vdwgridioffsetptr0;
 429     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 430     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 431     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 432     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 433     int              nvdwtype;
 434     __m256d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 435     int              *vdwtype;
 436     real             *vdwparam;
 437     __m256d          one_sixth   = _mm256_set1_pd(1.0/6.0);
 438     __m256d          one_twelfth = _mm256_set1_pd(1.0/12.0);
 439     __m256d           c6grid_00;
 440     real             *vdwgridparam;
 441     __m256d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 442     __m256d           one_half  = _mm256_set1_pd(0.5);
 443     __m256d           minus_one = _mm256_set1_pd(-1.0);
 444     __m256d          dummy_mask,cutoff_mask;
 445     __m128           tmpmask0,tmpmask1;
 446     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 447     __m256d          one     = _mm256_set1_pd(1.0);
 448     __m256d          two     = _mm256_set1_pd(2.0);
 449     x                = xx[0];
 450     f                = ff[0];
 451
 452     nri              = nlist->nri;
 453     iinr             = nlist->iinr;
 454     jindex           = nlist->jindex;
 455     jjnr             = nlist->jjnr;
 456     shiftidx         = nlist->shift;
 457     gid              = nlist->gid;
 458     shiftvec         = fr->shift_vec[0];
 459     fshift           = fr->fshift[0];
 460     nvdwtype         = fr->ntype;
 461     vdwparam         = fr->nbfp;
 462     vdwtype          = mdatoms->typeA;
 463     vdwgridparam     = fr->ljpme_c6grid;
 464     sh_lj_ewald      = _mm256_set1_pd(fr->ic->sh_lj_ewald);
 465     ewclj            = _mm256_set1_pd(fr->ewaldcoeff_lj);
 466     ewclj2           = _mm256_mul_pd(minus_one,_mm256_mul_pd(ewclj,ewclj));
 467
 468     /* Avoid stupid compiler warnings */
 469     jnrA = jnrB = jnrC = jnrD = 0;
 470     j_coord_offsetA = 0;
 471     j_coord_offsetB = 0;
 472     j_coord_offsetC = 0;
 473     j_coord_offsetD = 0;
 474
 475     outeriter        = 0;
 476     inneriter        = 0;
 477
 478     for(iidx=0;iidx<4*DIM;iidx++)
 479     {
 480         scratch[iidx] = 0.0;
 481     }
 482
 483     /* Start outer loop over neighborlists */
 484     for(iidx=0; iidx<nri; iidx++)
 485     {
 486         /* Load shift vector for this list */
 487         i_shift_offset   = DIM*shiftidx[iidx];
 488
 489         /* Load limits for loop over neighbors */
 490         j_index_start    = jindex[iidx];
 491         j_index_end      = jindex[iidx+1];
 492
 493         /* Get outer coordinate index */
 494         inr              = iinr[iidx];
 495         i_coord_offset   = DIM*inr;
 496
 497         /* Load i particle coords and add shift vector */
 498         gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 499
 500         fix0             = _mm256_setzero_pd();
 501         fiy0             = _mm256_setzero_pd();
 502         fiz0             = _mm256_setzero_pd();
 503
 504         /* Load parameters for i particles */
 505         vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 506         vdwgridioffsetptr0 = vdwgridparam+2*nvdwtype*vdwtype[inr+0];
 507
 508         /* Start inner kernel loop */
 509         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 510         {
 511
 512             /* Get j neighbor index, and coordinate index */
 513             jnrA             = jjnr[jidx];
 514             jnrB             = jjnr[jidx+1];
 515             jnrC             = jjnr[jidx+2];
 516             jnrD             = jjnr[jidx+3];
 517             j_coord_offsetA  = DIM*jnrA;
 518             j_coord_offsetB  = DIM*jnrB;
 519             j_coord_offsetC  = DIM*jnrC;
 520             j_coord_offsetD  = DIM*jnrD;
 521
 522             /* load j atom coordinates */
 523             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 524                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 525                                                  &jx0,&jy0,&jz0);
 526
 527             /* Calculate displacement vector */
 528             dx00             = _mm256_sub_pd(ix0,jx0);
 529             dy00             = _mm256_sub_pd(iy0,jy0);
 530             dz00             = _mm256_sub_pd(iz0,jz0);
 531
 532             /* Calculate squared distance and things based on it */
 533             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 534
 535             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 536
 537             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 538
 539             /* Load parameters for j particles */
 540             vdwjidx0A        = 2*vdwtype[jnrA+0];
 541             vdwjidx0B        = 2*vdwtype[jnrB+0];
 542             vdwjidx0C        = 2*vdwtype[jnrC+0];
 543             vdwjidx0D        = 2*vdwtype[jnrD+0];
 544
 545             /**************************
 546              * CALCULATE INTERACTIONS *
 547              **************************/
 548
 549             r00              = _mm256_mul_pd(rsq00,rinv00);
 550
 551             /* Compute parameters for interactions between i and j atoms */
 552             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
 553                                             vdwioffsetptr0+vdwjidx0B,
 554                                             vdwioffsetptr0+vdwjidx0C,
 555                                             vdwioffsetptr0+vdwjidx0D,
 556                                             &c6_00,&c12_00);
 557
 558             c6grid_00       = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
 559                                                                   vdwgridioffsetptr0+vdwjidx0B,
 560                                                                   vdwgridioffsetptr0+vdwjidx0C,
 561                                                                   vdwgridioffsetptr0+vdwjidx0D);
 562
 563             /* Analytical LJ-PME */
 564             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 565             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
 566             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
 567             exponent         = gmx_simd_exp_d(ewcljrsq);
 568             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 569             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 570             /* f6A = 6 * C6grid * (1 - poly) */
 571             f6A              = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
 572             /* f6B = C6grid * exponent * beta^6 */
 573             f6B              = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
 574             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 575             fvdw              = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
 576
 577             fscal            = fvdw;
 578
 579             /* Calculate temporary vectorial force */
 580             tx               = _mm256_mul_pd(fscal,dx00);
 581             ty               = _mm256_mul_pd(fscal,dy00);
 582             tz               = _mm256_mul_pd(fscal,dz00);
 583
 584             /* Update vectorial force */
 585             fix0             = _mm256_add_pd(fix0,tx);
 586             fiy0             = _mm256_add_pd(fiy0,ty);
 587             fiz0             = _mm256_add_pd(fiz0,tz);
 588
 589             fjptrA             = f+j_coord_offsetA;
 590             fjptrB             = f+j_coord_offsetB;
 591             fjptrC             = f+j_coord_offsetC;
 592             fjptrD             = f+j_coord_offsetD;
 593             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 594
 595             /* Inner loop uses 46 flops */
 596         }
 597
 598         if(jidx<j_index_end)
 599         {
 600
 601             /* Get j neighbor index, and coordinate index */
 602             jnrlistA         = jjnr[jidx];
 603             jnrlistB         = jjnr[jidx+1];
 604             jnrlistC         = jjnr[jidx+2];
 605             jnrlistD         = jjnr[jidx+3];
 606             /* Sign of each element will be negative for non-real atoms.
 607              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 608              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 609              */
 610             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 611
 612             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 613             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 614             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 615
 616             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 617             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 618             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 619             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 620             j_coord_offsetA  = DIM*jnrA;
 621             j_coord_offsetB  = DIM*jnrB;
 622             j_coord_offsetC  = DIM*jnrC;
 623             j_coord_offsetD  = DIM*jnrD;
 624
 625             /* load j atom coordinates */
 626             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 627                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 628                                                  &jx0,&jy0,&jz0);
 629
 630             /* Calculate displacement vector */
 631             dx00             = _mm256_sub_pd(ix0,jx0);
 632             dy00             = _mm256_sub_pd(iy0,jy0);
 633             dz00             = _mm256_sub_pd(iz0,jz0);
 634
 635             /* Calculate squared distance and things based on it */
 636             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 637
 638             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 639
 640             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 641
 642             /* Load parameters for j particles */
 643             vdwjidx0A        = 2*vdwtype[jnrA+0];
 644             vdwjidx0B        = 2*vdwtype[jnrB+0];
 645             vdwjidx0C        = 2*vdwtype[jnrC+0];
 646             vdwjidx0D        = 2*vdwtype[jnrD+0];
 647
 648             /**************************
 649              * CALCULATE INTERACTIONS *
 650              **************************/
 651
 652             r00              = _mm256_mul_pd(rsq00,rinv00);
 653             r00              = _mm256_andnot_pd(dummy_mask,r00);
 654
 655             /* Compute parameters for interactions between i and j atoms */
 656             gmx_mm256_load_4pair_swizzle_pd(vdwioffsetptr0+vdwjidx0A,
 657                                             vdwioffsetptr0+vdwjidx0B,
 658                                             vdwioffsetptr0+vdwjidx0C,
 659                                             vdwioffsetptr0+vdwjidx0D,
 660                                             &c6_00,&c12_00);
 661
 662             c6grid_00       = gmx_mm256_load_4real_swizzle_pd(vdwgridioffsetptr0+vdwjidx0A,
 663                                                                   vdwgridioffsetptr0+vdwjidx0B,
 664                                                                   vdwgridioffsetptr0+vdwjidx0C,
 665                                                                   vdwgridioffsetptr0+vdwjidx0D);
 666
 667             /* Analytical LJ-PME */
 668             rinvsix          = _mm256_mul_pd(_mm256_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 669             ewcljrsq         = _mm256_mul_pd(ewclj2,rsq00);
 670             ewclj6           = _mm256_mul_pd(ewclj2,_mm256_mul_pd(ewclj2,ewclj2));
 671             exponent         = gmx_simd_exp_d(ewcljrsq);
 672             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 673             poly             = _mm256_mul_pd(exponent,_mm256_add_pd(_mm256_sub_pd(one,ewcljrsq),_mm256_mul_pd(_mm256_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 674             /* f6A = 6 * C6grid * (1 - poly) */
 675             f6A              = _mm256_mul_pd(c6grid_00,_mm256_sub_pd(one,poly));
 676             /* f6B = C6grid * exponent * beta^6 */
 677             f6B              = _mm256_mul_pd(_mm256_mul_pd(c6grid_00,one_sixth),_mm256_mul_pd(exponent,ewclj6));
 678             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 679             fvdw              = _mm256_mul_pd(_mm256_add_pd(_mm256_mul_pd(_mm256_sub_pd(_mm256_mul_pd(c12_00,rinvsix),_mm256_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
 680
 681             fscal            = fvdw;
 682
 683             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 684
 685             /* Calculate temporary vectorial force */
 686             tx               = _mm256_mul_pd(fscal,dx00);
 687             ty               = _mm256_mul_pd(fscal,dy00);
 688             tz               = _mm256_mul_pd(fscal,dz00);
 689
 690             /* Update vectorial force */
 691             fix0             = _mm256_add_pd(fix0,tx);
 692             fiy0             = _mm256_add_pd(fiy0,ty);
 693             fiz0             = _mm256_add_pd(fiz0,tz);
 694
 695             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 696             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 697             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 698             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 699             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 700
 701             /* Inner loop uses 47 flops */
 702         }
 703
 704         /* End of innermost loop */
 705
 706         gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 707                                                  f+i_coord_offset,fshift+i_shift_offset);
 708
 709         /* Increment number of inner iterations */
 710         inneriter                  += j_index_end - j_index_start;
 711
 712         /* Outer loop uses 6 flops */
 713     }
 714
 715     /* Increment number of outer iterations */
 716     outeriter        += nri;
 717
 718     /* Update outer/inner flops */
 719
 720     inc_nrnb(nrnb,eNR_NBKERNEL_VDW_F,outeriter*6 + inneriter*47);
 721 }