src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEwSh_VdwNone_GeomP1P1_avx_256_double.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_double 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_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwNone_GeomP1P1_VF_avx_256_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            None
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSh_VdwNone_GeomP1P1_VF_avx_256_double
  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 refer to j loop unrolling done with AVX, e.g. for the four 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              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  75     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  83     real *           vdwioffsetptr0;
  84     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  85     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  86     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
  89     real             *charge;
  90     __m128i          ewitab;
  91     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  92     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
  93     real             *ewtab;
  94     __m256d          dummy_mask,cutoff_mask;
  95     __m128           tmpmask0,tmpmask1;
  96     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
  97     __m256d          one     = _mm256_set1_pd(1.0);
  98     __m256d          two     = _mm256_set1_pd(2.0);
  99     x                = xx[0];
 100     f                = ff[0];
 101
 102     nri              = nlist->nri;
 103     iinr             = nlist->iinr;
 104     jindex           = nlist->jindex;
 105     jjnr             = nlist->jjnr;
 106     shiftidx         = nlist->shift;
 107     gid              = nlist->gid;
 108     shiftvec         = fr->shift_vec[0];
 109     fshift           = fr->fshift[0];
 110     facel            = _mm256_set1_pd(fr->ic->epsfac);
 111     charge           = mdatoms->chargeA;
 112
 113     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 114     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 115     beta2            = _mm256_mul_pd(beta,beta);
 116     beta3            = _mm256_mul_pd(beta,beta2);
 117
 118     ewtab            = fr->ic->tabq_coul_FDV0;
 119     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 120     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 121
 122     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 123     rcutoff_scalar   = fr->ic->rcoulomb;
 124     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 125     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 126
 127     /* Avoid stupid compiler warnings */
 128     jnrA = jnrB = jnrC = jnrD = 0;
 129     j_coord_offsetA = 0;
 130     j_coord_offsetB = 0;
 131     j_coord_offsetC = 0;
 132     j_coord_offsetD = 0;
 133
 134     outeriter        = 0;
 135     inneriter        = 0;
 136
 137     for(iidx=0;iidx<4*DIM;iidx++)
 138     {
 139         scratch[iidx] = 0.0;
 140     }
 141
 142     /* Start outer loop over neighborlists */
 143     for(iidx=0; iidx<nri; iidx++)
 144     {
 145         /* Load shift vector for this list */
 146         i_shift_offset   = DIM*shiftidx[iidx];
 147
 148         /* Load limits for loop over neighbors */
 149         j_index_start    = jindex[iidx];
 150         j_index_end      = jindex[iidx+1];
 151
 152         /* Get outer coordinate index */
 153         inr              = iinr[iidx];
 154         i_coord_offset   = DIM*inr;
 155
 156         /* Load i particle coords and add shift vector */
 157         gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 158
 159         fix0             = _mm256_setzero_pd();
 160         fiy0             = _mm256_setzero_pd();
 161         fiz0             = _mm256_setzero_pd();
 162
 163         /* Load parameters for i particles */
 164         iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 165
 166         /* Reset potential sums */
 167         velecsum         = _mm256_setzero_pd();
 168
 169         /* Start inner kernel loop */
 170         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 171         {
 172
 173             /* Get j neighbor index, and coordinate index */
 174             jnrA             = jjnr[jidx];
 175             jnrB             = jjnr[jidx+1];
 176             jnrC             = jjnr[jidx+2];
 177             jnrD             = jjnr[jidx+3];
 178             j_coord_offsetA  = DIM*jnrA;
 179             j_coord_offsetB  = DIM*jnrB;
 180             j_coord_offsetC  = DIM*jnrC;
 181             j_coord_offsetD  = DIM*jnrD;
 182
 183             /* load j atom coordinates */
 184             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 185                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 186                                                  &jx0,&jy0,&jz0);
 187
 188             /* Calculate displacement vector */
 189             dx00             = _mm256_sub_pd(ix0,jx0);
 190             dy00             = _mm256_sub_pd(iy0,jy0);
 191             dz00             = _mm256_sub_pd(iz0,jz0);
 192
 193             /* Calculate squared distance and things based on it */
 194             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 195
 196             rinv00           = avx256_invsqrt_d(rsq00);
 197
 198             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 199
 200             /* Load parameters for j particles */
 201             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 202                                                                  charge+jnrC+0,charge+jnrD+0);
 203
 204             /**************************
 205              * CALCULATE INTERACTIONS *
 206              **************************/
 207
 208             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 209             {
 210
 211             r00              = _mm256_mul_pd(rsq00,rinv00);
 212
 213             /* Compute parameters for interactions between i and j atoms */
 214             qq00             = _mm256_mul_pd(iq0,jq0);
 215
 216             /* EWALD ELECTROSTATICS */
 217
 218             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 219             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 220             ewitab           = _mm256_cvttpd_epi32(ewrt);
 221             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 222             ewitab           = _mm_slli_epi32(ewitab,2);
 223             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 224             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 225             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 226             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 227             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 228             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 229             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 230             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 231             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 232
 233             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 234
 235             /* Update potential sum for this i atom from the interaction with this j atom. */
 236             velec            = _mm256_and_pd(velec,cutoff_mask);
 237             velecsum         = _mm256_add_pd(velecsum,velec);
 238
 239             fscal            = felec;
 240
 241             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 242
 243             /* Calculate temporary vectorial force */
 244             tx               = _mm256_mul_pd(fscal,dx00);
 245             ty               = _mm256_mul_pd(fscal,dy00);
 246             tz               = _mm256_mul_pd(fscal,dz00);
 247
 248             /* Update vectorial force */
 249             fix0             = _mm256_add_pd(fix0,tx);
 250             fiy0             = _mm256_add_pd(fiy0,ty);
 251             fiz0             = _mm256_add_pd(fiz0,tz);
 252
 253             fjptrA             = f+j_coord_offsetA;
 254             fjptrB             = f+j_coord_offsetB;
 255             fjptrC             = f+j_coord_offsetC;
 256             fjptrD             = f+j_coord_offsetD;
 257             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 258
 259             }
 260
 261             /* Inner loop uses 46 flops */
 262         }
 263
 264         if(jidx<j_index_end)
 265         {
 266
 267             /* Get j neighbor index, and coordinate index */
 268             jnrlistA         = jjnr[jidx];
 269             jnrlistB         = jjnr[jidx+1];
 270             jnrlistC         = jjnr[jidx+2];
 271             jnrlistD         = jjnr[jidx+3];
 272             /* Sign of each element will be negative for non-real atoms.
 273              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 274              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 275              */
 276             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 277
 278             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 279             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 280             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 281
 282             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 283             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 284             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 285             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 286             j_coord_offsetA  = DIM*jnrA;
 287             j_coord_offsetB  = DIM*jnrB;
 288             j_coord_offsetC  = DIM*jnrC;
 289             j_coord_offsetD  = DIM*jnrD;
 290
 291             /* load j atom coordinates */
 292             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 293                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 294                                                  &jx0,&jy0,&jz0);
 295
 296             /* Calculate displacement vector */
 297             dx00             = _mm256_sub_pd(ix0,jx0);
 298             dy00             = _mm256_sub_pd(iy0,jy0);
 299             dz00             = _mm256_sub_pd(iz0,jz0);
 300
 301             /* Calculate squared distance and things based on it */
 302             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 303
 304             rinv00           = avx256_invsqrt_d(rsq00);
 305
 306             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 307
 308             /* Load parameters for j particles */
 309             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 310                                                                  charge+jnrC+0,charge+jnrD+0);
 311
 312             /**************************
 313              * CALCULATE INTERACTIONS *
 314              **************************/
 315
 316             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 317             {
 318
 319             r00              = _mm256_mul_pd(rsq00,rinv00);
 320             r00              = _mm256_andnot_pd(dummy_mask,r00);
 321
 322             /* Compute parameters for interactions between i and j atoms */
 323             qq00             = _mm256_mul_pd(iq0,jq0);
 324
 325             /* EWALD ELECTROSTATICS */
 326
 327             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 328             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 329             ewitab           = _mm256_cvttpd_epi32(ewrt);
 330             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 331             ewitab           = _mm_slli_epi32(ewitab,2);
 332             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 333             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 334             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 335             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 336             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 337             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 338             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 339             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(_mm256_sub_pd(rinv00,sh_ewald),velec));
 340             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 341
 342             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 343
 344             /* Update potential sum for this i atom from the interaction with this j atom. */
 345             velec            = _mm256_and_pd(velec,cutoff_mask);
 346             velec            = _mm256_andnot_pd(dummy_mask,velec);
 347             velecsum         = _mm256_add_pd(velecsum,velec);
 348
 349             fscal            = felec;
 350
 351             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 352
 353             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 354
 355             /* Calculate temporary vectorial force */
 356             tx               = _mm256_mul_pd(fscal,dx00);
 357             ty               = _mm256_mul_pd(fscal,dy00);
 358             tz               = _mm256_mul_pd(fscal,dz00);
 359
 360             /* Update vectorial force */
 361             fix0             = _mm256_add_pd(fix0,tx);
 362             fiy0             = _mm256_add_pd(fiy0,ty);
 363             fiz0             = _mm256_add_pd(fiz0,tz);
 364
 365             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 366             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 367             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 368             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 369             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 370
 371             }
 372
 373             /* Inner loop uses 47 flops */
 374         }
 375
 376         /* End of innermost loop */
 377
 378         gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 379                                                  f+i_coord_offset,fshift+i_shift_offset);
 380
 381         ggid                        = gid[iidx];
 382         /* Update potential energies */
 383         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 384
 385         /* Increment number of inner iterations */
 386         inneriter                  += j_index_end - j_index_start;
 387
 388         /* Outer loop uses 8 flops */
 389     }
 390
 391     /* Increment number of outer iterations */
 392     outeriter        += nri;
 393
 394     /* Update outer/inner flops */
 395
 396     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VF,outeriter*8 + inneriter*47);
 397 }
 398 /*
 399  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwNone_GeomP1P1_F_avx_256_double
 400  * Electrostatics interaction: Ewald
 401  * VdW interaction:            None
 402  * Geometry:                   Particle-Particle
 403  * Calculate force/pot:        Force
 404  */
 405 void
 406 nb_kernel_ElecEwSh_VdwNone_GeomP1P1_F_avx_256_double
 407                     (t_nblist                    * gmx_restrict       nlist,
 408                      rvec                        * gmx_restrict          xx,
 409                      rvec                        * gmx_restrict          ff,
 410                      struct t_forcerec           * gmx_restrict          fr,
 411                      t_mdatoms                   * gmx_restrict     mdatoms,
 412                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 413                      t_nrnb                      * gmx_restrict        nrnb)
 414 {
 415     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 416      * just 0 for non-waters.
 417      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
 418      * jnr indices corresponding to data put in the four positions in the SIMD register.
 419      */
 420     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 421     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 422     int              jnrA,jnrB,jnrC,jnrD;
 423     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 424     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 425     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 426     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 427     real             rcutoff_scalar;
 428     real             *shiftvec,*fshift,*x,*f;
 429     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 430     real             scratch[4*DIM];
 431     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 432     real *           vdwioffsetptr0;
 433     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 434     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 435     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 436     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 437     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 438     real             *charge;
 439     __m128i          ewitab;
 440     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 441     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 442     real             *ewtab;
 443     __m256d          dummy_mask,cutoff_mask;
 444     __m128           tmpmask0,tmpmask1;
 445     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 446     __m256d          one     = _mm256_set1_pd(1.0);
 447     __m256d          two     = _mm256_set1_pd(2.0);
 448     x                = xx[0];
 449     f                = ff[0];
 450
 451     nri              = nlist->nri;
 452     iinr             = nlist->iinr;
 453     jindex           = nlist->jindex;
 454     jjnr             = nlist->jjnr;
 455     shiftidx         = nlist->shift;
 456     gid              = nlist->gid;
 457     shiftvec         = fr->shift_vec[0];
 458     fshift           = fr->fshift[0];
 459     facel            = _mm256_set1_pd(fr->ic->epsfac);
 460     charge           = mdatoms->chargeA;
 461
 462     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 463     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 464     beta2            = _mm256_mul_pd(beta,beta);
 465     beta3            = _mm256_mul_pd(beta,beta2);
 466
 467     ewtab            = fr->ic->tabq_coul_F;
 468     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 469     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 470
 471     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 472     rcutoff_scalar   = fr->ic->rcoulomb;
 473     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 474     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 475
 476     /* Avoid stupid compiler warnings */
 477     jnrA = jnrB = jnrC = jnrD = 0;
 478     j_coord_offsetA = 0;
 479     j_coord_offsetB = 0;
 480     j_coord_offsetC = 0;
 481     j_coord_offsetD = 0;
 482
 483     outeriter        = 0;
 484     inneriter        = 0;
 485
 486     for(iidx=0;iidx<4*DIM;iidx++)
 487     {
 488         scratch[iidx] = 0.0;
 489     }
 490
 491     /* Start outer loop over neighborlists */
 492     for(iidx=0; iidx<nri; iidx++)
 493     {
 494         /* Load shift vector for this list */
 495         i_shift_offset   = DIM*shiftidx[iidx];
 496
 497         /* Load limits for loop over neighbors */
 498         j_index_start    = jindex[iidx];
 499         j_index_end      = jindex[iidx+1];
 500
 501         /* Get outer coordinate index */
 502         inr              = iinr[iidx];
 503         i_coord_offset   = DIM*inr;
 504
 505         /* Load i particle coords and add shift vector */
 506         gmx_mm256_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 507
 508         fix0             = _mm256_setzero_pd();
 509         fiy0             = _mm256_setzero_pd();
 510         fiz0             = _mm256_setzero_pd();
 511
 512         /* Load parameters for i particles */
 513         iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 514
 515         /* Start inner kernel loop */
 516         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 517         {
 518
 519             /* Get j neighbor index, and coordinate index */
 520             jnrA             = jjnr[jidx];
 521             jnrB             = jjnr[jidx+1];
 522             jnrC             = jjnr[jidx+2];
 523             jnrD             = jjnr[jidx+3];
 524             j_coord_offsetA  = DIM*jnrA;
 525             j_coord_offsetB  = DIM*jnrB;
 526             j_coord_offsetC  = DIM*jnrC;
 527             j_coord_offsetD  = DIM*jnrD;
 528
 529             /* load j atom coordinates */
 530             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 531                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 532                                                  &jx0,&jy0,&jz0);
 533
 534             /* Calculate displacement vector */
 535             dx00             = _mm256_sub_pd(ix0,jx0);
 536             dy00             = _mm256_sub_pd(iy0,jy0);
 537             dz00             = _mm256_sub_pd(iz0,jz0);
 538
 539             /* Calculate squared distance and things based on it */
 540             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 541
 542             rinv00           = avx256_invsqrt_d(rsq00);
 543
 544             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 545
 546             /* Load parameters for j particles */
 547             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 548                                                                  charge+jnrC+0,charge+jnrD+0);
 549
 550             /**************************
 551              * CALCULATE INTERACTIONS *
 552              **************************/
 553
 554             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 555             {
 556
 557             r00              = _mm256_mul_pd(rsq00,rinv00);
 558
 559             /* Compute parameters for interactions between i and j atoms */
 560             qq00             = _mm256_mul_pd(iq0,jq0);
 561
 562             /* EWALD ELECTROSTATICS */
 563
 564             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 565             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 566             ewitab           = _mm256_cvttpd_epi32(ewrt);
 567             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 568             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
 569                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
 570                                             &ewtabF,&ewtabFn);
 571             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
 572             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 573
 574             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 575
 576             fscal            = felec;
 577
 578             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 579
 580             /* Calculate temporary vectorial force */
 581             tx               = _mm256_mul_pd(fscal,dx00);
 582             ty               = _mm256_mul_pd(fscal,dy00);
 583             tz               = _mm256_mul_pd(fscal,dz00);
 584
 585             /* Update vectorial force */
 586             fix0             = _mm256_add_pd(fix0,tx);
 587             fiy0             = _mm256_add_pd(fiy0,ty);
 588             fiz0             = _mm256_add_pd(fiz0,tz);
 589
 590             fjptrA             = f+j_coord_offsetA;
 591             fjptrB             = f+j_coord_offsetB;
 592             fjptrC             = f+j_coord_offsetC;
 593             fjptrD             = f+j_coord_offsetD;
 594             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 595
 596             }
 597
 598             /* Inner loop uses 39 flops */
 599         }
 600
 601         if(jidx<j_index_end)
 602         {
 603
 604             /* Get j neighbor index, and coordinate index */
 605             jnrlistA         = jjnr[jidx];
 606             jnrlistB         = jjnr[jidx+1];
 607             jnrlistC         = jjnr[jidx+2];
 608             jnrlistD         = jjnr[jidx+3];
 609             /* Sign of each element will be negative for non-real atoms.
 610              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 611              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 612              */
 613             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 614
 615             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 616             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 617             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 618
 619             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 620             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 621             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 622             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 623             j_coord_offsetA  = DIM*jnrA;
 624             j_coord_offsetB  = DIM*jnrB;
 625             j_coord_offsetC  = DIM*jnrC;
 626             j_coord_offsetD  = DIM*jnrD;
 627
 628             /* load j atom coordinates */
 629             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 630                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 631                                                  &jx0,&jy0,&jz0);
 632
 633             /* Calculate displacement vector */
 634             dx00             = _mm256_sub_pd(ix0,jx0);
 635             dy00             = _mm256_sub_pd(iy0,jy0);
 636             dz00             = _mm256_sub_pd(iz0,jz0);
 637
 638             /* Calculate squared distance and things based on it */
 639             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 640
 641             rinv00           = avx256_invsqrt_d(rsq00);
 642
 643             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 644
 645             /* Load parameters for j particles */
 646             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 647                                                                  charge+jnrC+0,charge+jnrD+0);
 648
 649             /**************************
 650              * CALCULATE INTERACTIONS *
 651              **************************/
 652
 653             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 654             {
 655
 656             r00              = _mm256_mul_pd(rsq00,rinv00);
 657             r00              = _mm256_andnot_pd(dummy_mask,r00);
 658
 659             /* Compute parameters for interactions between i and j atoms */
 660             qq00             = _mm256_mul_pd(iq0,jq0);
 661
 662             /* EWALD ELECTROSTATICS */
 663
 664             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 665             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 666             ewitab           = _mm256_cvttpd_epi32(ewrt);
 667             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 668             gmx_mm256_load_4pair_swizzle_pd(ewtab + _mm_extract_epi32(ewitab,0),ewtab + _mm_extract_epi32(ewitab,1),
 669                                             ewtab + _mm_extract_epi32(ewitab,2),ewtab + _mm_extract_epi32(ewitab,3),
 670                                             &ewtabF,&ewtabFn);
 671             felec            = _mm256_add_pd(_mm256_mul_pd( _mm256_sub_pd(one,eweps),ewtabF),_mm256_mul_pd(eweps,ewtabFn));
 672             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 673
 674             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 675
 676             fscal            = felec;
 677
 678             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 679
 680             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 681
 682             /* Calculate temporary vectorial force */
 683             tx               = _mm256_mul_pd(fscal,dx00);
 684             ty               = _mm256_mul_pd(fscal,dy00);
 685             tz               = _mm256_mul_pd(fscal,dz00);
 686
 687             /* Update vectorial force */
 688             fix0             = _mm256_add_pd(fix0,tx);
 689             fiy0             = _mm256_add_pd(fiy0,ty);
 690             fiz0             = _mm256_add_pd(fiz0,tz);
 691
 692             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 693             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 694             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 695             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 696             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,tx,ty,tz);
 697
 698             }
 699
 700             /* Inner loop uses 40 flops */
 701         }
 702
 703         /* End of innermost loop */
 704
 705         gmx_mm256_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 706                                                  f+i_coord_offset,fshift+i_shift_offset);
 707
 708         /* Increment number of inner iterations */
 709         inneriter                  += j_index_end - j_index_start;
 710
 711         /* Outer loop uses 7 flops */
 712     }
 713
 714     /* Increment number of outer iterations */
 715     outeriter        += nri;
 716
 717     /* Update outer/inner flops */
 718
 719     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_F,outeriter*7 + inneriter*40);
 720 }