src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_double/nb_kernel_ElecEwSw_VdwNone_GeomW3P1_avx_256_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_double.h"
  50 #include "kernelutil_x86_avx_256_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_avx_256_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            None
  56  * Geometry:                   Water3-Particle
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_VF_avx_256_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  79     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  84     real             scratch[4*DIM];
  85     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  93     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  94     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  95     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  96     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  97     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
  98     real             *charge;
  99     __m128i          ewitab;
 100     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 101     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 102     real             *ewtab;
 103     __m256d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 104     real             rswitch_scalar,d_scalar;
 105     __m256d          dummy_mask,cutoff_mask;
 106     __m128           tmpmask0,tmpmask1;
 107     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 108     __m256d          one     = _mm256_set1_pd(1.0);
 109     __m256d          two     = _mm256_set1_pd(2.0);
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = _mm256_set1_pd(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123
 124     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 125     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 126     beta2            = _mm256_mul_pd(beta,beta);
 127     beta3            = _mm256_mul_pd(beta,beta2);
 128
 129     ewtab            = fr->ic->tabq_coul_FDV0;
 130     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 131     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 132
 133     /* Setup water-specific parameters */
 134     inr              = nlist->iinr[0];
 135     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 136     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 137     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 138
 139     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 140     rcutoff_scalar   = fr->rcoulomb;
 141     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 142     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 143
 144     rswitch_scalar   = fr->rcoulomb_switch;
 145     rswitch          = _mm256_set1_pd(rswitch_scalar);
 146     /* Setup switch parameters */
 147     d_scalar         = rcutoff_scalar-rswitch_scalar;
 148     d                = _mm256_set1_pd(d_scalar);
 149     swV3             = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 150     swV4             = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 151     swV5             = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 152     swF2             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 153     swF3             = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 154     swF4             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 155
 156     /* Avoid stupid compiler warnings */
 157     jnrA = jnrB = jnrC = jnrD = 0;
 158     j_coord_offsetA = 0;
 159     j_coord_offsetB = 0;
 160     j_coord_offsetC = 0;
 161     j_coord_offsetD = 0;
 162
 163     outeriter        = 0;
 164     inneriter        = 0;
 165
 166     for(iidx=0;iidx<4*DIM;iidx++)
 167     {
 168         scratch[iidx] = 0.0;
 169     }
 170
 171     /* Start outer loop over neighborlists */
 172     for(iidx=0; iidx<nri; iidx++)
 173     {
 174         /* Load shift vector for this list */
 175         i_shift_offset   = DIM*shiftidx[iidx];
 176
 177         /* Load limits for loop over neighbors */
 178         j_index_start    = jindex[iidx];
 179         j_index_end      = jindex[iidx+1];
 180
 181         /* Get outer coordinate index */
 182         inr              = iinr[iidx];
 183         i_coord_offset   = DIM*inr;
 184
 185         /* Load i particle coords and add shift vector */
 186         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 187                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 188
 189         fix0             = _mm256_setzero_pd();
 190         fiy0             = _mm256_setzero_pd();
 191         fiz0             = _mm256_setzero_pd();
 192         fix1             = _mm256_setzero_pd();
 193         fiy1             = _mm256_setzero_pd();
 194         fiz1             = _mm256_setzero_pd();
 195         fix2             = _mm256_setzero_pd();
 196         fiy2             = _mm256_setzero_pd();
 197         fiz2             = _mm256_setzero_pd();
 198
 199         /* Reset potential sums */
 200         velecsum         = _mm256_setzero_pd();
 201
 202         /* Start inner kernel loop */
 203         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 204         {
 205
 206             /* Get j neighbor index, and coordinate index */
 207             jnrA             = jjnr[jidx];
 208             jnrB             = jjnr[jidx+1];
 209             jnrC             = jjnr[jidx+2];
 210             jnrD             = jjnr[jidx+3];
 211             j_coord_offsetA  = DIM*jnrA;
 212             j_coord_offsetB  = DIM*jnrB;
 213             j_coord_offsetC  = DIM*jnrC;
 214             j_coord_offsetD  = DIM*jnrD;
 215
 216             /* load j atom coordinates */
 217             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 218                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 219                                                  &jx0,&jy0,&jz0);
 220
 221             /* Calculate displacement vector */
 222             dx00             = _mm256_sub_pd(ix0,jx0);
 223             dy00             = _mm256_sub_pd(iy0,jy0);
 224             dz00             = _mm256_sub_pd(iz0,jz0);
 225             dx10             = _mm256_sub_pd(ix1,jx0);
 226             dy10             = _mm256_sub_pd(iy1,jy0);
 227             dz10             = _mm256_sub_pd(iz1,jz0);
 228             dx20             = _mm256_sub_pd(ix2,jx0);
 229             dy20             = _mm256_sub_pd(iy2,jy0);
 230             dz20             = _mm256_sub_pd(iz2,jz0);
 231
 232             /* Calculate squared distance and things based on it */
 233             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 234             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 235             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 236
 237             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 238             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 239             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 240
 241             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 242             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 243             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 244
 245             /* Load parameters for j particles */
 246             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 247                                                                  charge+jnrC+0,charge+jnrD+0);
 248
 249             fjx0             = _mm256_setzero_pd();
 250             fjy0             = _mm256_setzero_pd();
 251             fjz0             = _mm256_setzero_pd();
 252
 253             /**************************
 254              * CALCULATE INTERACTIONS *
 255              **************************/
 256
 257             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 258             {
 259
 260             r00              = _mm256_mul_pd(rsq00,rinv00);
 261
 262             /* Compute parameters for interactions between i and j atoms */
 263             qq00             = _mm256_mul_pd(iq0,jq0);
 264
 265             /* EWALD ELECTROSTATICS */
 266
 267             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 268             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 269             ewitab           = _mm256_cvttpd_epi32(ewrt);
 270             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 271             ewitab           = _mm_slli_epi32(ewitab,2);
 272             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 273             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 274             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 275             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 276             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 277             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 278             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 279             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
 280             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 281
 282             d                = _mm256_sub_pd(r00,rswitch);
 283             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 284             d2               = _mm256_mul_pd(d,d);
 285             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 286
 287             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 288
 289             /* Evaluate switch function */
 290             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 291             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
 292             velec            = _mm256_mul_pd(velec,sw);
 293             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 294
 295             /* Update potential sum for this i atom from the interaction with this j atom. */
 296             velec            = _mm256_and_pd(velec,cutoff_mask);
 297             velecsum         = _mm256_add_pd(velecsum,velec);
 298
 299             fscal            = felec;
 300
 301             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 302
 303             /* Calculate temporary vectorial force */
 304             tx               = _mm256_mul_pd(fscal,dx00);
 305             ty               = _mm256_mul_pd(fscal,dy00);
 306             tz               = _mm256_mul_pd(fscal,dz00);
 307
 308             /* Update vectorial force */
 309             fix0             = _mm256_add_pd(fix0,tx);
 310             fiy0             = _mm256_add_pd(fiy0,ty);
 311             fiz0             = _mm256_add_pd(fiz0,tz);
 312
 313             fjx0             = _mm256_add_pd(fjx0,tx);
 314             fjy0             = _mm256_add_pd(fjy0,ty);
 315             fjz0             = _mm256_add_pd(fjz0,tz);
 316
 317             }
 318
 319             /**************************
 320              * CALCULATE INTERACTIONS *
 321              **************************/
 322
 323             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 324             {
 325
 326             r10              = _mm256_mul_pd(rsq10,rinv10);
 327
 328             /* Compute parameters for interactions between i and j atoms */
 329             qq10             = _mm256_mul_pd(iq1,jq0);
 330
 331             /* EWALD ELECTROSTATICS */
 332
 333             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 334             ewrt             = _mm256_mul_pd(r10,ewtabscale);
 335             ewitab           = _mm256_cvttpd_epi32(ewrt);
 336             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 337             ewitab           = _mm_slli_epi32(ewitab,2);
 338             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 339             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 340             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 341             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 342             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 343             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 344             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 345             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
 346             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
 347
 348             d                = _mm256_sub_pd(r10,rswitch);
 349             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 350             d2               = _mm256_mul_pd(d,d);
 351             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 352
 353             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 354
 355             /* Evaluate switch function */
 356             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 357             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
 358             velec            = _mm256_mul_pd(velec,sw);
 359             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
 360
 361             /* Update potential sum for this i atom from the interaction with this j atom. */
 362             velec            = _mm256_and_pd(velec,cutoff_mask);
 363             velecsum         = _mm256_add_pd(velecsum,velec);
 364
 365             fscal            = felec;
 366
 367             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 368
 369             /* Calculate temporary vectorial force */
 370             tx               = _mm256_mul_pd(fscal,dx10);
 371             ty               = _mm256_mul_pd(fscal,dy10);
 372             tz               = _mm256_mul_pd(fscal,dz10);
 373
 374             /* Update vectorial force */
 375             fix1             = _mm256_add_pd(fix1,tx);
 376             fiy1             = _mm256_add_pd(fiy1,ty);
 377             fiz1             = _mm256_add_pd(fiz1,tz);
 378
 379             fjx0             = _mm256_add_pd(fjx0,tx);
 380             fjy0             = _mm256_add_pd(fjy0,ty);
 381             fjz0             = _mm256_add_pd(fjz0,tz);
 382
 383             }
 384
 385             /**************************
 386              * CALCULATE INTERACTIONS *
 387              **************************/
 388
 389             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 390             {
 391
 392             r20              = _mm256_mul_pd(rsq20,rinv20);
 393
 394             /* Compute parameters for interactions between i and j atoms */
 395             qq20             = _mm256_mul_pd(iq2,jq0);
 396
 397             /* EWALD ELECTROSTATICS */
 398
 399             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 400             ewrt             = _mm256_mul_pd(r20,ewtabscale);
 401             ewitab           = _mm256_cvttpd_epi32(ewrt);
 402             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 403             ewitab           = _mm_slli_epi32(ewitab,2);
 404             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 405             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 406             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 407             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 408             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 409             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 410             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 411             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
 412             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
 413
 414             d                = _mm256_sub_pd(r20,rswitch);
 415             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 416             d2               = _mm256_mul_pd(d,d);
 417             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 418
 419             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 420
 421             /* Evaluate switch function */
 422             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 423             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
 424             velec            = _mm256_mul_pd(velec,sw);
 425             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
 426
 427             /* Update potential sum for this i atom from the interaction with this j atom. */
 428             velec            = _mm256_and_pd(velec,cutoff_mask);
 429             velecsum         = _mm256_add_pd(velecsum,velec);
 430
 431             fscal            = felec;
 432
 433             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 434
 435             /* Calculate temporary vectorial force */
 436             tx               = _mm256_mul_pd(fscal,dx20);
 437             ty               = _mm256_mul_pd(fscal,dy20);
 438             tz               = _mm256_mul_pd(fscal,dz20);
 439
 440             /* Update vectorial force */
 441             fix2             = _mm256_add_pd(fix2,tx);
 442             fiy2             = _mm256_add_pd(fiy2,ty);
 443             fiz2             = _mm256_add_pd(fiz2,tz);
 444
 445             fjx0             = _mm256_add_pd(fjx0,tx);
 446             fjy0             = _mm256_add_pd(fjy0,ty);
 447             fjz0             = _mm256_add_pd(fjz0,tz);
 448
 449             }
 450
 451             fjptrA             = f+j_coord_offsetA;
 452             fjptrB             = f+j_coord_offsetB;
 453             fjptrC             = f+j_coord_offsetC;
 454             fjptrD             = f+j_coord_offsetD;
 455
 456             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 457
 458             /* Inner loop uses 198 flops */
 459         }
 460
 461         if(jidx<j_index_end)
 462         {
 463
 464             /* Get j neighbor index, and coordinate index */
 465             jnrlistA         = jjnr[jidx];
 466             jnrlistB         = jjnr[jidx+1];
 467             jnrlistC         = jjnr[jidx+2];
 468             jnrlistD         = jjnr[jidx+3];
 469             /* Sign of each element will be negative for non-real atoms.
 470              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 471              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
 472              */
 473             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 474
 475             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
 476             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
 477             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
 478
 479             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 480             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 481             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 482             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 483             j_coord_offsetA  = DIM*jnrA;
 484             j_coord_offsetB  = DIM*jnrB;
 485             j_coord_offsetC  = DIM*jnrC;
 486             j_coord_offsetD  = DIM*jnrD;
 487
 488             /* load j atom coordinates */
 489             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 490                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 491                                                  &jx0,&jy0,&jz0);
 492
 493             /* Calculate displacement vector */
 494             dx00             = _mm256_sub_pd(ix0,jx0);
 495             dy00             = _mm256_sub_pd(iy0,jy0);
 496             dz00             = _mm256_sub_pd(iz0,jz0);
 497             dx10             = _mm256_sub_pd(ix1,jx0);
 498             dy10             = _mm256_sub_pd(iy1,jy0);
 499             dz10             = _mm256_sub_pd(iz1,jz0);
 500             dx20             = _mm256_sub_pd(ix2,jx0);
 501             dy20             = _mm256_sub_pd(iy2,jy0);
 502             dz20             = _mm256_sub_pd(iz2,jz0);
 503
 504             /* Calculate squared distance and things based on it */
 505             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 506             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 507             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 508
 509             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 510             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 511             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 512
 513             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 514             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 515             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 516
 517             /* Load parameters for j particles */
 518             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 519                                                                  charge+jnrC+0,charge+jnrD+0);
 520
 521             fjx0             = _mm256_setzero_pd();
 522             fjy0             = _mm256_setzero_pd();
 523             fjz0             = _mm256_setzero_pd();
 524
 525             /**************************
 526              * CALCULATE INTERACTIONS *
 527              **************************/
 528
 529             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 530             {
 531
 532             r00              = _mm256_mul_pd(rsq00,rinv00);
 533             r00              = _mm256_andnot_pd(dummy_mask,r00);
 534
 535             /* Compute parameters for interactions between i and j atoms */
 536             qq00             = _mm256_mul_pd(iq0,jq0);
 537
 538             /* EWALD ELECTROSTATICS */
 539
 540             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 541             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 542             ewitab           = _mm256_cvttpd_epi32(ewrt);
 543             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 544             ewitab           = _mm_slli_epi32(ewitab,2);
 545             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 546             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 547             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 548             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 549             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 550             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 551             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 552             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
 553             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 554
 555             d                = _mm256_sub_pd(r00,rswitch);
 556             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 557             d2               = _mm256_mul_pd(d,d);
 558             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 559
 560             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 561
 562             /* Evaluate switch function */
 563             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 564             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
 565             velec            = _mm256_mul_pd(velec,sw);
 566             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
 567
 568             /* Update potential sum for this i atom from the interaction with this j atom. */
 569             velec            = _mm256_and_pd(velec,cutoff_mask);
 570             velec            = _mm256_andnot_pd(dummy_mask,velec);
 571             velecsum         = _mm256_add_pd(velecsum,velec);
 572
 573             fscal            = felec;
 574
 575             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 576
 577             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 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             fjx0             = _mm256_add_pd(fjx0,tx);
 590             fjy0             = _mm256_add_pd(fjy0,ty);
 591             fjz0             = _mm256_add_pd(fjz0,tz);
 592
 593             }
 594
 595             /**************************
 596              * CALCULATE INTERACTIONS *
 597              **************************/
 598
 599             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 600             {
 601
 602             r10              = _mm256_mul_pd(rsq10,rinv10);
 603             r10              = _mm256_andnot_pd(dummy_mask,r10);
 604
 605             /* Compute parameters for interactions between i and j atoms */
 606             qq10             = _mm256_mul_pd(iq1,jq0);
 607
 608             /* EWALD ELECTROSTATICS */
 609
 610             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 611             ewrt             = _mm256_mul_pd(r10,ewtabscale);
 612             ewitab           = _mm256_cvttpd_epi32(ewrt);
 613             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 614             ewitab           = _mm_slli_epi32(ewitab,2);
 615             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 616             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 617             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 618             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 619             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 620             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 621             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 622             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
 623             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
 624
 625             d                = _mm256_sub_pd(r10,rswitch);
 626             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 627             d2               = _mm256_mul_pd(d,d);
 628             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 629
 630             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 631
 632             /* Evaluate switch function */
 633             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 634             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
 635             velec            = _mm256_mul_pd(velec,sw);
 636             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
 637
 638             /* Update potential sum for this i atom from the interaction with this j atom. */
 639             velec            = _mm256_and_pd(velec,cutoff_mask);
 640             velec            = _mm256_andnot_pd(dummy_mask,velec);
 641             velecsum         = _mm256_add_pd(velecsum,velec);
 642
 643             fscal            = felec;
 644
 645             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 646
 647             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 648
 649             /* Calculate temporary vectorial force */
 650             tx               = _mm256_mul_pd(fscal,dx10);
 651             ty               = _mm256_mul_pd(fscal,dy10);
 652             tz               = _mm256_mul_pd(fscal,dz10);
 653
 654             /* Update vectorial force */
 655             fix1             = _mm256_add_pd(fix1,tx);
 656             fiy1             = _mm256_add_pd(fiy1,ty);
 657             fiz1             = _mm256_add_pd(fiz1,tz);
 658
 659             fjx0             = _mm256_add_pd(fjx0,tx);
 660             fjy0             = _mm256_add_pd(fjy0,ty);
 661             fjz0             = _mm256_add_pd(fjz0,tz);
 662
 663             }
 664
 665             /**************************
 666              * CALCULATE INTERACTIONS *
 667              **************************/
 668
 669             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 670             {
 671
 672             r20              = _mm256_mul_pd(rsq20,rinv20);
 673             r20              = _mm256_andnot_pd(dummy_mask,r20);
 674
 675             /* Compute parameters for interactions between i and j atoms */
 676             qq20             = _mm256_mul_pd(iq2,jq0);
 677
 678             /* EWALD ELECTROSTATICS */
 679
 680             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 681             ewrt             = _mm256_mul_pd(r20,ewtabscale);
 682             ewitab           = _mm256_cvttpd_epi32(ewrt);
 683             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 684             ewitab           = _mm_slli_epi32(ewitab,2);
 685             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 686             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 687             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 688             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 689             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 690             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 691             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 692             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
 693             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
 694
 695             d                = _mm256_sub_pd(r20,rswitch);
 696             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 697             d2               = _mm256_mul_pd(d,d);
 698             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 699
 700             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
 701
 702             /* Evaluate switch function */
 703             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 704             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
 705             velec            = _mm256_mul_pd(velec,sw);
 706             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
 707
 708             /* Update potential sum for this i atom from the interaction with this j atom. */
 709             velec            = _mm256_and_pd(velec,cutoff_mask);
 710             velec            = _mm256_andnot_pd(dummy_mask,velec);
 711             velecsum         = _mm256_add_pd(velecsum,velec);
 712
 713             fscal            = felec;
 714
 715             fscal            = _mm256_and_pd(fscal,cutoff_mask);
 716
 717             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
 718
 719             /* Calculate temporary vectorial force */
 720             tx               = _mm256_mul_pd(fscal,dx20);
 721             ty               = _mm256_mul_pd(fscal,dy20);
 722             tz               = _mm256_mul_pd(fscal,dz20);
 723
 724             /* Update vectorial force */
 725             fix2             = _mm256_add_pd(fix2,tx);
 726             fiy2             = _mm256_add_pd(fiy2,ty);
 727             fiz2             = _mm256_add_pd(fiz2,tz);
 728
 729             fjx0             = _mm256_add_pd(fjx0,tx);
 730             fjy0             = _mm256_add_pd(fjy0,ty);
 731             fjz0             = _mm256_add_pd(fjz0,tz);
 732
 733             }
 734
 735             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 736             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 737             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 738             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 739
 740             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 741
 742             /* Inner loop uses 201 flops */
 743         }
 744
 745         /* End of innermost loop */
 746
 747         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 748                                                  f+i_coord_offset,fshift+i_shift_offset);
 749
 750         ggid                        = gid[iidx];
 751         /* Update potential energies */
 752         gmx_mm256_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 753
 754         /* Increment number of inner iterations */
 755         inneriter                  += j_index_end - j_index_start;
 756
 757         /* Outer loop uses 19 flops */
 758     }
 759
 760     /* Increment number of outer iterations */
 761     outeriter        += nri;
 762
 763     /* Update outer/inner flops */
 764
 765     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*201);
 766 }
 767 /*
 768  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_avx_256_double
 769  * Electrostatics interaction: Ewald
 770  * VdW interaction:            None
 771  * Geometry:                   Water3-Particle
 772  * Calculate force/pot:        Force
 773  */
 774 void
 775 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_avx_256_double
 776                     (t_nblist                    * gmx_restrict       nlist,
 777                      rvec                        * gmx_restrict          xx,
 778                      rvec                        * gmx_restrict          ff,
 779                      t_forcerec                  * gmx_restrict          fr,
 780                      t_mdatoms                   * gmx_restrict     mdatoms,
 781                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 782                      t_nrnb                      * gmx_restrict        nrnb)
 783 {
 784     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 785      * just 0 for non-waters.
 786      * Suffixes A,B,C,D refer to j loop unrolling done with AVX, e.g. for the four different
 787      * jnr indices corresponding to data put in the four positions in the SIMD register.
 788      */
 789     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 790     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 791     int              jnrA,jnrB,jnrC,jnrD;
 792     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 793     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 794     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 795     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 796     real             rcutoff_scalar;
 797     real             *shiftvec,*fshift,*x,*f;
 798     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 799     real             scratch[4*DIM];
 800     __m256d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 801     real *           vdwioffsetptr0;
 802     __m256d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 803     real *           vdwioffsetptr1;
 804     __m256d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 805     real *           vdwioffsetptr2;
 806     __m256d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 807     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 808     __m256d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 809     __m256d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 810     __m256d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 811     __m256d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 812     __m256d          velec,felec,velecsum,facel,crf,krf,krf2;
 813     real             *charge;
 814     __m128i          ewitab;
 815     __m256d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 816     __m256d          beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 817     real             *ewtab;
 818     __m256d          rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 819     real             rswitch_scalar,d_scalar;
 820     __m256d          dummy_mask,cutoff_mask;
 821     __m128           tmpmask0,tmpmask1;
 822     __m256d          signbit = _mm256_castsi256_pd( _mm256_set1_epi32(0x80000000) );
 823     __m256d          one     = _mm256_set1_pd(1.0);
 824     __m256d          two     = _mm256_set1_pd(2.0);
 825     x                = xx[0];
 826     f                = ff[0];
 827
 828     nri              = nlist->nri;
 829     iinr             = nlist->iinr;
 830     jindex           = nlist->jindex;
 831     jjnr             = nlist->jjnr;
 832     shiftidx         = nlist->shift;
 833     gid              = nlist->gid;
 834     shiftvec         = fr->shift_vec[0];
 835     fshift           = fr->fshift[0];
 836     facel            = _mm256_set1_pd(fr->epsfac);
 837     charge           = mdatoms->chargeA;
 838
 839     sh_ewald         = _mm256_set1_pd(fr->ic->sh_ewald);
 840     beta             = _mm256_set1_pd(fr->ic->ewaldcoeff_q);
 841     beta2            = _mm256_mul_pd(beta,beta);
 842     beta3            = _mm256_mul_pd(beta,beta2);
 843
 844     ewtab            = fr->ic->tabq_coul_FDV0;
 845     ewtabscale       = _mm256_set1_pd(fr->ic->tabq_scale);
 846     ewtabhalfspace   = _mm256_set1_pd(0.5/fr->ic->tabq_scale);
 847
 848     /* Setup water-specific parameters */
 849     inr              = nlist->iinr[0];
 850     iq0              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+0]));
 851     iq1              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+1]));
 852     iq2              = _mm256_mul_pd(facel,_mm256_set1_pd(charge[inr+2]));
 853
 854     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 855     rcutoff_scalar   = fr->rcoulomb;
 856     rcutoff          = _mm256_set1_pd(rcutoff_scalar);
 857     rcutoff2         = _mm256_mul_pd(rcutoff,rcutoff);
 858
 859     rswitch_scalar   = fr->rcoulomb_switch;
 860     rswitch          = _mm256_set1_pd(rswitch_scalar);
 861     /* Setup switch parameters */
 862     d_scalar         = rcutoff_scalar-rswitch_scalar;
 863     d                = _mm256_set1_pd(d_scalar);
 864     swV3             = _mm256_set1_pd(-10.0/(d_scalar*d_scalar*d_scalar));
 865     swV4             = _mm256_set1_pd( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 866     swV5             = _mm256_set1_pd( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 867     swF2             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar));
 868     swF3             = _mm256_set1_pd( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 869     swF4             = _mm256_set1_pd(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 870
 871     /* Avoid stupid compiler warnings */
 872     jnrA = jnrB = jnrC = jnrD = 0;
 873     j_coord_offsetA = 0;
 874     j_coord_offsetB = 0;
 875     j_coord_offsetC = 0;
 876     j_coord_offsetD = 0;
 877
 878     outeriter        = 0;
 879     inneriter        = 0;
 880
 881     for(iidx=0;iidx<4*DIM;iidx++)
 882     {
 883         scratch[iidx] = 0.0;
 884     }
 885
 886     /* Start outer loop over neighborlists */
 887     for(iidx=0; iidx<nri; iidx++)
 888     {
 889         /* Load shift vector for this list */
 890         i_shift_offset   = DIM*shiftidx[iidx];
 891
 892         /* Load limits for loop over neighbors */
 893         j_index_start    = jindex[iidx];
 894         j_index_end      = jindex[iidx+1];
 895
 896         /* Get outer coordinate index */
 897         inr              = iinr[iidx];
 898         i_coord_offset   = DIM*inr;
 899
 900         /* Load i particle coords and add shift vector */
 901         gmx_mm256_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 902                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 903
 904         fix0             = _mm256_setzero_pd();
 905         fiy0             = _mm256_setzero_pd();
 906         fiz0             = _mm256_setzero_pd();
 907         fix1             = _mm256_setzero_pd();
 908         fiy1             = _mm256_setzero_pd();
 909         fiz1             = _mm256_setzero_pd();
 910         fix2             = _mm256_setzero_pd();
 911         fiy2             = _mm256_setzero_pd();
 912         fiz2             = _mm256_setzero_pd();
 913
 914         /* Start inner kernel loop */
 915         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 916         {
 917
 918             /* Get j neighbor index, and coordinate index */
 919             jnrA             = jjnr[jidx];
 920             jnrB             = jjnr[jidx+1];
 921             jnrC             = jjnr[jidx+2];
 922             jnrD             = jjnr[jidx+3];
 923             j_coord_offsetA  = DIM*jnrA;
 924             j_coord_offsetB  = DIM*jnrB;
 925             j_coord_offsetC  = DIM*jnrC;
 926             j_coord_offsetD  = DIM*jnrD;
 927
 928             /* load j atom coordinates */
 929             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 930                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 931                                                  &jx0,&jy0,&jz0);
 932
 933             /* Calculate displacement vector */
 934             dx00             = _mm256_sub_pd(ix0,jx0);
 935             dy00             = _mm256_sub_pd(iy0,jy0);
 936             dz00             = _mm256_sub_pd(iz0,jz0);
 937             dx10             = _mm256_sub_pd(ix1,jx0);
 938             dy10             = _mm256_sub_pd(iy1,jy0);
 939             dz10             = _mm256_sub_pd(iz1,jz0);
 940             dx20             = _mm256_sub_pd(ix2,jx0);
 941             dy20             = _mm256_sub_pd(iy2,jy0);
 942             dz20             = _mm256_sub_pd(iz2,jz0);
 943
 944             /* Calculate squared distance and things based on it */
 945             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
 946             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
 947             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
 948
 949             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
 950             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
 951             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
 952
 953             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
 954             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
 955             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
 956
 957             /* Load parameters for j particles */
 958             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
 959                                                                  charge+jnrC+0,charge+jnrD+0);
 960
 961             fjx0             = _mm256_setzero_pd();
 962             fjy0             = _mm256_setzero_pd();
 963             fjz0             = _mm256_setzero_pd();
 964
 965             /**************************
 966              * CALCULATE INTERACTIONS *
 967              **************************/
 968
 969             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 970             {
 971
 972             r00              = _mm256_mul_pd(rsq00,rinv00);
 973
 974             /* Compute parameters for interactions between i and j atoms */
 975             qq00             = _mm256_mul_pd(iq0,jq0);
 976
 977             /* EWALD ELECTROSTATICS */
 978
 979             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 980             ewrt             = _mm256_mul_pd(r00,ewtabscale);
 981             ewitab           = _mm256_cvttpd_epi32(ewrt);
 982             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
 983             ewitab           = _mm_slli_epi32(ewitab,2);
 984             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 985             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 986             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
 987             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
 988             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
 989             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
 990             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
 991             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
 992             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
 993
 994             d                = _mm256_sub_pd(r00,rswitch);
 995             d                = _mm256_max_pd(d,_mm256_setzero_pd());
 996             d2               = _mm256_mul_pd(d,d);
 997             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
 998
 999             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1000
1001             /* Evaluate switch function */
1002             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1003             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1004             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1005
1006             fscal            = felec;
1007
1008             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1009
1010             /* Calculate temporary vectorial force */
1011             tx               = _mm256_mul_pd(fscal,dx00);
1012             ty               = _mm256_mul_pd(fscal,dy00);
1013             tz               = _mm256_mul_pd(fscal,dz00);
1014
1015             /* Update vectorial force */
1016             fix0             = _mm256_add_pd(fix0,tx);
1017             fiy0             = _mm256_add_pd(fiy0,ty);
1018             fiz0             = _mm256_add_pd(fiz0,tz);
1019
1020             fjx0             = _mm256_add_pd(fjx0,tx);
1021             fjy0             = _mm256_add_pd(fjy0,ty);
1022             fjz0             = _mm256_add_pd(fjz0,tz);
1023
1024             }
1025
1026             /**************************
1027              * CALCULATE INTERACTIONS *
1028              **************************/
1029
1030             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1031             {
1032
1033             r10              = _mm256_mul_pd(rsq10,rinv10);
1034
1035             /* Compute parameters for interactions between i and j atoms */
1036             qq10             = _mm256_mul_pd(iq1,jq0);
1037
1038             /* EWALD ELECTROSTATICS */
1039
1040             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1041             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1042             ewitab           = _mm256_cvttpd_epi32(ewrt);
1043             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1044             ewitab           = _mm_slli_epi32(ewitab,2);
1045             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1046             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1047             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1048             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1049             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1050             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1051             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1052             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1053             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1054
1055             d                = _mm256_sub_pd(r10,rswitch);
1056             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1057             d2               = _mm256_mul_pd(d,d);
1058             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1059
1060             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1061
1062             /* Evaluate switch function */
1063             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1064             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1065             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1066
1067             fscal            = felec;
1068
1069             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1070
1071             /* Calculate temporary vectorial force */
1072             tx               = _mm256_mul_pd(fscal,dx10);
1073             ty               = _mm256_mul_pd(fscal,dy10);
1074             tz               = _mm256_mul_pd(fscal,dz10);
1075
1076             /* Update vectorial force */
1077             fix1             = _mm256_add_pd(fix1,tx);
1078             fiy1             = _mm256_add_pd(fiy1,ty);
1079             fiz1             = _mm256_add_pd(fiz1,tz);
1080
1081             fjx0             = _mm256_add_pd(fjx0,tx);
1082             fjy0             = _mm256_add_pd(fjy0,ty);
1083             fjz0             = _mm256_add_pd(fjz0,tz);
1084
1085             }
1086
1087             /**************************
1088              * CALCULATE INTERACTIONS *
1089              **************************/
1090
1091             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1092             {
1093
1094             r20              = _mm256_mul_pd(rsq20,rinv20);
1095
1096             /* Compute parameters for interactions between i and j atoms */
1097             qq20             = _mm256_mul_pd(iq2,jq0);
1098
1099             /* EWALD ELECTROSTATICS */
1100
1101             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1102             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1103             ewitab           = _mm256_cvttpd_epi32(ewrt);
1104             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1105             ewitab           = _mm_slli_epi32(ewitab,2);
1106             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1107             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1108             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1109             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1110             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1111             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1112             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1113             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1114             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1115
1116             d                = _mm256_sub_pd(r20,rswitch);
1117             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1118             d2               = _mm256_mul_pd(d,d);
1119             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1120
1121             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1122
1123             /* Evaluate switch function */
1124             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1125             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1126             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1127
1128             fscal            = felec;
1129
1130             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1131
1132             /* Calculate temporary vectorial force */
1133             tx               = _mm256_mul_pd(fscal,dx20);
1134             ty               = _mm256_mul_pd(fscal,dy20);
1135             tz               = _mm256_mul_pd(fscal,dz20);
1136
1137             /* Update vectorial force */
1138             fix2             = _mm256_add_pd(fix2,tx);
1139             fiy2             = _mm256_add_pd(fiy2,ty);
1140             fiz2             = _mm256_add_pd(fiz2,tz);
1141
1142             fjx0             = _mm256_add_pd(fjx0,tx);
1143             fjy0             = _mm256_add_pd(fjy0,ty);
1144             fjz0             = _mm256_add_pd(fjz0,tz);
1145
1146             }
1147
1148             fjptrA             = f+j_coord_offsetA;
1149             fjptrB             = f+j_coord_offsetB;
1150             fjptrC             = f+j_coord_offsetC;
1151             fjptrD             = f+j_coord_offsetD;
1152
1153             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1154
1155             /* Inner loop uses 189 flops */
1156         }
1157
1158         if(jidx<j_index_end)
1159         {
1160
1161             /* Get j neighbor index, and coordinate index */
1162             jnrlistA         = jjnr[jidx];
1163             jnrlistB         = jjnr[jidx+1];
1164             jnrlistC         = jjnr[jidx+2];
1165             jnrlistD         = jjnr[jidx+3];
1166             /* Sign of each element will be negative for non-real atoms.
1167              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1168              * so use it as val = _mm_andnot_pd(mask,val) to clear dummy entries.
1169              */
1170             tmpmask0 = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1171
1172             tmpmask1 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(3,3,2,2));
1173             tmpmask0 = _mm_permute_ps(tmpmask0,_GMX_MM_PERMUTE(1,1,0,0));
1174             dummy_mask = _mm256_castps_pd(gmx_mm256_set_m128(tmpmask1,tmpmask0));
1175
1176             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1177             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1178             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1179             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1180             j_coord_offsetA  = DIM*jnrA;
1181             j_coord_offsetB  = DIM*jnrB;
1182             j_coord_offsetC  = DIM*jnrC;
1183             j_coord_offsetD  = DIM*jnrD;
1184
1185             /* load j atom coordinates */
1186             gmx_mm256_load_1rvec_4ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1187                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1188                                                  &jx0,&jy0,&jz0);
1189
1190             /* Calculate displacement vector */
1191             dx00             = _mm256_sub_pd(ix0,jx0);
1192             dy00             = _mm256_sub_pd(iy0,jy0);
1193             dz00             = _mm256_sub_pd(iz0,jz0);
1194             dx10             = _mm256_sub_pd(ix1,jx0);
1195             dy10             = _mm256_sub_pd(iy1,jy0);
1196             dz10             = _mm256_sub_pd(iz1,jz0);
1197             dx20             = _mm256_sub_pd(ix2,jx0);
1198             dy20             = _mm256_sub_pd(iy2,jy0);
1199             dz20             = _mm256_sub_pd(iz2,jz0);
1200
1201             /* Calculate squared distance and things based on it */
1202             rsq00            = gmx_mm256_calc_rsq_pd(dx00,dy00,dz00);
1203             rsq10            = gmx_mm256_calc_rsq_pd(dx10,dy10,dz10);
1204             rsq20            = gmx_mm256_calc_rsq_pd(dx20,dy20,dz20);
1205
1206             rinv00           = gmx_mm256_invsqrt_pd(rsq00);
1207             rinv10           = gmx_mm256_invsqrt_pd(rsq10);
1208             rinv20           = gmx_mm256_invsqrt_pd(rsq20);
1209
1210             rinvsq00         = _mm256_mul_pd(rinv00,rinv00);
1211             rinvsq10         = _mm256_mul_pd(rinv10,rinv10);
1212             rinvsq20         = _mm256_mul_pd(rinv20,rinv20);
1213
1214             /* Load parameters for j particles */
1215             jq0              = gmx_mm256_load_4real_swizzle_pd(charge+jnrA+0,charge+jnrB+0,
1216                                                                  charge+jnrC+0,charge+jnrD+0);
1217
1218             fjx0             = _mm256_setzero_pd();
1219             fjy0             = _mm256_setzero_pd();
1220             fjz0             = _mm256_setzero_pd();
1221
1222             /**************************
1223              * CALCULATE INTERACTIONS *
1224              **************************/
1225
1226             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1227             {
1228
1229             r00              = _mm256_mul_pd(rsq00,rinv00);
1230             r00              = _mm256_andnot_pd(dummy_mask,r00);
1231
1232             /* Compute parameters for interactions between i and j atoms */
1233             qq00             = _mm256_mul_pd(iq0,jq0);
1234
1235             /* EWALD ELECTROSTATICS */
1236
1237             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1238             ewrt             = _mm256_mul_pd(r00,ewtabscale);
1239             ewitab           = _mm256_cvttpd_epi32(ewrt);
1240             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1241             ewitab           = _mm_slli_epi32(ewitab,2);
1242             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1243             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1244             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1245             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1246             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1247             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1248             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1249             velec            = _mm256_mul_pd(qq00,_mm256_sub_pd(rinv00,velec));
1250             felec            = _mm256_mul_pd(_mm256_mul_pd(qq00,rinv00),_mm256_sub_pd(rinvsq00,felec));
1251
1252             d                = _mm256_sub_pd(r00,rswitch);
1253             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1254             d2               = _mm256_mul_pd(d,d);
1255             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1256
1257             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1258
1259             /* Evaluate switch function */
1260             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1261             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv00,_mm256_mul_pd(velec,dsw)) );
1262             cutoff_mask      = _mm256_cmp_pd(rsq00,rcutoff2,_CMP_LT_OQ);
1263
1264             fscal            = felec;
1265
1266             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1267
1268             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1269
1270             /* Calculate temporary vectorial force */
1271             tx               = _mm256_mul_pd(fscal,dx00);
1272             ty               = _mm256_mul_pd(fscal,dy00);
1273             tz               = _mm256_mul_pd(fscal,dz00);
1274
1275             /* Update vectorial force */
1276             fix0             = _mm256_add_pd(fix0,tx);
1277             fiy0             = _mm256_add_pd(fiy0,ty);
1278             fiz0             = _mm256_add_pd(fiz0,tz);
1279
1280             fjx0             = _mm256_add_pd(fjx0,tx);
1281             fjy0             = _mm256_add_pd(fjy0,ty);
1282             fjz0             = _mm256_add_pd(fjz0,tz);
1283
1284             }
1285
1286             /**************************
1287              * CALCULATE INTERACTIONS *
1288              **************************/
1289
1290             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1291             {
1292
1293             r10              = _mm256_mul_pd(rsq10,rinv10);
1294             r10              = _mm256_andnot_pd(dummy_mask,r10);
1295
1296             /* Compute parameters for interactions between i and j atoms */
1297             qq10             = _mm256_mul_pd(iq1,jq0);
1298
1299             /* EWALD ELECTROSTATICS */
1300
1301             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1302             ewrt             = _mm256_mul_pd(r10,ewtabscale);
1303             ewitab           = _mm256_cvttpd_epi32(ewrt);
1304             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1305             ewitab           = _mm_slli_epi32(ewitab,2);
1306             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1307             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1308             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1309             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1310             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1311             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1312             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1313             velec            = _mm256_mul_pd(qq10,_mm256_sub_pd(rinv10,velec));
1314             felec            = _mm256_mul_pd(_mm256_mul_pd(qq10,rinv10),_mm256_sub_pd(rinvsq10,felec));
1315
1316             d                = _mm256_sub_pd(r10,rswitch);
1317             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1318             d2               = _mm256_mul_pd(d,d);
1319             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1320
1321             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1322
1323             /* Evaluate switch function */
1324             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1325             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv10,_mm256_mul_pd(velec,dsw)) );
1326             cutoff_mask      = _mm256_cmp_pd(rsq10,rcutoff2,_CMP_LT_OQ);
1327
1328             fscal            = felec;
1329
1330             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1331
1332             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1333
1334             /* Calculate temporary vectorial force */
1335             tx               = _mm256_mul_pd(fscal,dx10);
1336             ty               = _mm256_mul_pd(fscal,dy10);
1337             tz               = _mm256_mul_pd(fscal,dz10);
1338
1339             /* Update vectorial force */
1340             fix1             = _mm256_add_pd(fix1,tx);
1341             fiy1             = _mm256_add_pd(fiy1,ty);
1342             fiz1             = _mm256_add_pd(fiz1,tz);
1343
1344             fjx0             = _mm256_add_pd(fjx0,tx);
1345             fjy0             = _mm256_add_pd(fjy0,ty);
1346             fjz0             = _mm256_add_pd(fjz0,tz);
1347
1348             }
1349
1350             /**************************
1351              * CALCULATE INTERACTIONS *
1352              **************************/
1353
1354             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1355             {
1356
1357             r20              = _mm256_mul_pd(rsq20,rinv20);
1358             r20              = _mm256_andnot_pd(dummy_mask,r20);
1359
1360             /* Compute parameters for interactions between i and j atoms */
1361             qq20             = _mm256_mul_pd(iq2,jq0);
1362
1363             /* EWALD ELECTROSTATICS */
1364
1365             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1366             ewrt             = _mm256_mul_pd(r20,ewtabscale);
1367             ewitab           = _mm256_cvttpd_epi32(ewrt);
1368             eweps            = _mm256_sub_pd(ewrt,_mm256_round_pd(ewrt, _MM_FROUND_FLOOR));
1369             ewitab           = _mm_slli_epi32(ewitab,2);
1370             ewtabF           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
1371             ewtabD           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
1372             ewtabV           = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,2) );
1373             ewtabFn          = _mm256_load_pd( ewtab + _mm_extract_epi32(ewitab,3) );
1374             GMX_MM256_FULLTRANSPOSE4_PD(ewtabF,ewtabD,ewtabV,ewtabFn);
1375             felec            = _mm256_add_pd(ewtabF,_mm256_mul_pd(eweps,ewtabD));
1376             velec            = _mm256_sub_pd(ewtabV,_mm256_mul_pd(_mm256_mul_pd(ewtabhalfspace,eweps),_mm256_add_pd(ewtabF,felec)));
1377             velec            = _mm256_mul_pd(qq20,_mm256_sub_pd(rinv20,velec));
1378             felec            = _mm256_mul_pd(_mm256_mul_pd(qq20,rinv20),_mm256_sub_pd(rinvsq20,felec));
1379
1380             d                = _mm256_sub_pd(r20,rswitch);
1381             d                = _mm256_max_pd(d,_mm256_setzero_pd());
1382             d2               = _mm256_mul_pd(d,d);
1383             sw               = _mm256_add_pd(one,_mm256_mul_pd(d2,_mm256_mul_pd(d,_mm256_add_pd(swV3,_mm256_mul_pd(d,_mm256_add_pd(swV4,_mm256_mul_pd(d,swV5)))))));
1384
1385             dsw              = _mm256_mul_pd(d2,_mm256_add_pd(swF2,_mm256_mul_pd(d,_mm256_add_pd(swF3,_mm256_mul_pd(d,swF4)))));
1386
1387             /* Evaluate switch function */
1388             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1389             felec            = _mm256_sub_pd( _mm256_mul_pd(felec,sw) , _mm256_mul_pd(rinv20,_mm256_mul_pd(velec,dsw)) );
1390             cutoff_mask      = _mm256_cmp_pd(rsq20,rcutoff2,_CMP_LT_OQ);
1391
1392             fscal            = felec;
1393
1394             fscal            = _mm256_and_pd(fscal,cutoff_mask);
1395
1396             fscal            = _mm256_andnot_pd(dummy_mask,fscal);
1397
1398             /* Calculate temporary vectorial force */
1399             tx               = _mm256_mul_pd(fscal,dx20);
1400             ty               = _mm256_mul_pd(fscal,dy20);
1401             tz               = _mm256_mul_pd(fscal,dz20);
1402
1403             /* Update vectorial force */
1404             fix2             = _mm256_add_pd(fix2,tx);
1405             fiy2             = _mm256_add_pd(fiy2,ty);
1406             fiz2             = _mm256_add_pd(fiz2,tz);
1407
1408             fjx0             = _mm256_add_pd(fjx0,tx);
1409             fjy0             = _mm256_add_pd(fjy0,ty);
1410             fjz0             = _mm256_add_pd(fjz0,tz);
1411
1412             }
1413
1414             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1415             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1416             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1417             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1418
1419             gmx_mm256_decrement_1rvec_4ptr_swizzle_pd(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1420
1421             /* Inner loop uses 192 flops */
1422         }
1423
1424         /* End of innermost loop */
1425
1426         gmx_mm256_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1427                                                  f+i_coord_offset,fshift+i_shift_offset);
1428
1429         /* Increment number of inner iterations */
1430         inneriter                  += j_index_end - j_index_start;
1431
1432         /* Outer loop uses 18 flops */
1433     }
1434
1435     /* Increment number of outer iterations */
1436     outeriter        += nri;
1437
1438     /* Update outer/inner flops */
1439
1440     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*192);
1441 }