src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRFCut_VdwNone_GeomW3P1_avx_256_single.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_single kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_avx_256_single.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwNone_GeomW3P1_VF_avx_256_single
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            None
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwNone_GeomW3P1_VF_avx_256_single
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrE,jnrF,jnrG,jnrH;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  77     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  78     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  79     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  80     real             rcutoff_scalar;
  81     real             *shiftvec,*fshift,*x,*f;
  82     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  83     real             scratch[4*DIM];
  84     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  85     real *           vdwioffsetptr0;
  86     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  87     real *           vdwioffsetptr1;
  88     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  89     real *           vdwioffsetptr2;
  90     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  92     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  94     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  95     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  96     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
  97     real             *charge;
  98     __m256           dummy_mask,cutoff_mask;
  99     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 100     __m256           one     = _mm256_set1_ps(1.0);
 101     __m256           two     = _mm256_set1_ps(2.0);
 102     x                = xx[0];
 103     f                = ff[0];
 104
 105     nri              = nlist->nri;
 106     iinr             = nlist->iinr;
 107     jindex           = nlist->jindex;
 108     jjnr             = nlist->jjnr;
 109     shiftidx         = nlist->shift;
 110     gid              = nlist->gid;
 111     shiftvec         = fr->shift_vec[0];
 112     fshift           = fr->fshift[0];
 113     facel            = _mm256_set1_ps(fr->ic->epsfac);
 114     charge           = mdatoms->chargeA;
 115     krf              = _mm256_set1_ps(fr->ic->k_rf);
 116     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 117     crf              = _mm256_set1_ps(fr->ic->c_rf);
 118
 119     /* Setup water-specific parameters */
 120     inr              = nlist->iinr[0];
 121     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 122     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 123     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 124
 125     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 126     rcutoff_scalar   = fr->ic->rcoulomb;
 127     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 128     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 129
 130     /* Avoid stupid compiler warnings */
 131     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 132     j_coord_offsetA = 0;
 133     j_coord_offsetB = 0;
 134     j_coord_offsetC = 0;
 135     j_coord_offsetD = 0;
 136     j_coord_offsetE = 0;
 137     j_coord_offsetF = 0;
 138     j_coord_offsetG = 0;
 139     j_coord_offsetH = 0;
 140
 141     outeriter        = 0;
 142     inneriter        = 0;
 143
 144     for(iidx=0;iidx<4*DIM;iidx++)
 145     {
 146         scratch[iidx] = 0.0;
 147     }
 148
 149     /* Start outer loop over neighborlists */
 150     for(iidx=0; iidx<nri; iidx++)
 151     {
 152         /* Load shift vector for this list */
 153         i_shift_offset   = DIM*shiftidx[iidx];
 154
 155         /* Load limits for loop over neighbors */
 156         j_index_start    = jindex[iidx];
 157         j_index_end      = jindex[iidx+1];
 158
 159         /* Get outer coordinate index */
 160         inr              = iinr[iidx];
 161         i_coord_offset   = DIM*inr;
 162
 163         /* Load i particle coords and add shift vector */
 164         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 165                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 166
 167         fix0             = _mm256_setzero_ps();
 168         fiy0             = _mm256_setzero_ps();
 169         fiz0             = _mm256_setzero_ps();
 170         fix1             = _mm256_setzero_ps();
 171         fiy1             = _mm256_setzero_ps();
 172         fiz1             = _mm256_setzero_ps();
 173         fix2             = _mm256_setzero_ps();
 174         fiy2             = _mm256_setzero_ps();
 175         fiz2             = _mm256_setzero_ps();
 176
 177         /* Reset potential sums */
 178         velecsum         = _mm256_setzero_ps();
 179
 180         /* Start inner kernel loop */
 181         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 182         {
 183
 184             /* Get j neighbor index, and coordinate index */
 185             jnrA             = jjnr[jidx];
 186             jnrB             = jjnr[jidx+1];
 187             jnrC             = jjnr[jidx+2];
 188             jnrD             = jjnr[jidx+3];
 189             jnrE             = jjnr[jidx+4];
 190             jnrF             = jjnr[jidx+5];
 191             jnrG             = jjnr[jidx+6];
 192             jnrH             = jjnr[jidx+7];
 193             j_coord_offsetA  = DIM*jnrA;
 194             j_coord_offsetB  = DIM*jnrB;
 195             j_coord_offsetC  = DIM*jnrC;
 196             j_coord_offsetD  = DIM*jnrD;
 197             j_coord_offsetE  = DIM*jnrE;
 198             j_coord_offsetF  = DIM*jnrF;
 199             j_coord_offsetG  = DIM*jnrG;
 200             j_coord_offsetH  = DIM*jnrH;
 201
 202             /* load j atom coordinates */
 203             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 204                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 205                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 206                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 207                                                  &jx0,&jy0,&jz0);
 208
 209             /* Calculate displacement vector */
 210             dx00             = _mm256_sub_ps(ix0,jx0);
 211             dy00             = _mm256_sub_ps(iy0,jy0);
 212             dz00             = _mm256_sub_ps(iz0,jz0);
 213             dx10             = _mm256_sub_ps(ix1,jx0);
 214             dy10             = _mm256_sub_ps(iy1,jy0);
 215             dz10             = _mm256_sub_ps(iz1,jz0);
 216             dx20             = _mm256_sub_ps(ix2,jx0);
 217             dy20             = _mm256_sub_ps(iy2,jy0);
 218             dz20             = _mm256_sub_ps(iz2,jz0);
 219
 220             /* Calculate squared distance and things based on it */
 221             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 222             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 223             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 224
 225             rinv00           = avx256_invsqrt_f(rsq00);
 226             rinv10           = avx256_invsqrt_f(rsq10);
 227             rinv20           = avx256_invsqrt_f(rsq20);
 228
 229             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 230             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 231             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 232
 233             /* Load parameters for j particles */
 234             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 235                                                                  charge+jnrC+0,charge+jnrD+0,
 236                                                                  charge+jnrE+0,charge+jnrF+0,
 237                                                                  charge+jnrG+0,charge+jnrH+0);
 238
 239             fjx0             = _mm256_setzero_ps();
 240             fjy0             = _mm256_setzero_ps();
 241             fjz0             = _mm256_setzero_ps();
 242
 243             /**************************
 244              * CALCULATE INTERACTIONS *
 245              **************************/
 246
 247             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 248             {
 249
 250             /* Compute parameters for interactions between i and j atoms */
 251             qq00             = _mm256_mul_ps(iq0,jq0);
 252
 253             /* REACTION-FIELD ELECTROSTATICS */
 254             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 255             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 256
 257             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 258
 259             /* Update potential sum for this i atom from the interaction with this j atom. */
 260             velec            = _mm256_and_ps(velec,cutoff_mask);
 261             velecsum         = _mm256_add_ps(velecsum,velec);
 262
 263             fscal            = felec;
 264
 265             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 266
 267             /* Calculate temporary vectorial force */
 268             tx               = _mm256_mul_ps(fscal,dx00);
 269             ty               = _mm256_mul_ps(fscal,dy00);
 270             tz               = _mm256_mul_ps(fscal,dz00);
 271
 272             /* Update vectorial force */
 273             fix0             = _mm256_add_ps(fix0,tx);
 274             fiy0             = _mm256_add_ps(fiy0,ty);
 275             fiz0             = _mm256_add_ps(fiz0,tz);
 276
 277             fjx0             = _mm256_add_ps(fjx0,tx);
 278             fjy0             = _mm256_add_ps(fjy0,ty);
 279             fjz0             = _mm256_add_ps(fjz0,tz);
 280
 281             }
 282
 283             /**************************
 284              * CALCULATE INTERACTIONS *
 285              **************************/
 286
 287             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 288             {
 289
 290             /* Compute parameters for interactions between i and j atoms */
 291             qq10             = _mm256_mul_ps(iq1,jq0);
 292
 293             /* REACTION-FIELD ELECTROSTATICS */
 294             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 295             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 296
 297             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 298
 299             /* Update potential sum for this i atom from the interaction with this j atom. */
 300             velec            = _mm256_and_ps(velec,cutoff_mask);
 301             velecsum         = _mm256_add_ps(velecsum,velec);
 302
 303             fscal            = felec;
 304
 305             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 306
 307             /* Calculate temporary vectorial force */
 308             tx               = _mm256_mul_ps(fscal,dx10);
 309             ty               = _mm256_mul_ps(fscal,dy10);
 310             tz               = _mm256_mul_ps(fscal,dz10);
 311
 312             /* Update vectorial force */
 313             fix1             = _mm256_add_ps(fix1,tx);
 314             fiy1             = _mm256_add_ps(fiy1,ty);
 315             fiz1             = _mm256_add_ps(fiz1,tz);
 316
 317             fjx0             = _mm256_add_ps(fjx0,tx);
 318             fjy0             = _mm256_add_ps(fjy0,ty);
 319             fjz0             = _mm256_add_ps(fjz0,tz);
 320
 321             }
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 328             {
 329
 330             /* Compute parameters for interactions between i and j atoms */
 331             qq20             = _mm256_mul_ps(iq2,jq0);
 332
 333             /* REACTION-FIELD ELECTROSTATICS */
 334             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 335             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 336
 337             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
 338
 339             /* Update potential sum for this i atom from the interaction with this j atom. */
 340             velec            = _mm256_and_ps(velec,cutoff_mask);
 341             velecsum         = _mm256_add_ps(velecsum,velec);
 342
 343             fscal            = felec;
 344
 345             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 346
 347             /* Calculate temporary vectorial force */
 348             tx               = _mm256_mul_ps(fscal,dx20);
 349             ty               = _mm256_mul_ps(fscal,dy20);
 350             tz               = _mm256_mul_ps(fscal,dz20);
 351
 352             /* Update vectorial force */
 353             fix2             = _mm256_add_ps(fix2,tx);
 354             fiy2             = _mm256_add_ps(fiy2,ty);
 355             fiz2             = _mm256_add_ps(fiz2,tz);
 356
 357             fjx0             = _mm256_add_ps(fjx0,tx);
 358             fjy0             = _mm256_add_ps(fjy0,ty);
 359             fjz0             = _mm256_add_ps(fjz0,tz);
 360
 361             }
 362
 363             fjptrA             = f+j_coord_offsetA;
 364             fjptrB             = f+j_coord_offsetB;
 365             fjptrC             = f+j_coord_offsetC;
 366             fjptrD             = f+j_coord_offsetD;
 367             fjptrE             = f+j_coord_offsetE;
 368             fjptrF             = f+j_coord_offsetF;
 369             fjptrG             = f+j_coord_offsetG;
 370             fjptrH             = f+j_coord_offsetH;
 371
 372             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 373
 374             /* Inner loop uses 111 flops */
 375         }
 376
 377         if(jidx<j_index_end)
 378         {
 379
 380             /* Get j neighbor index, and coordinate index */
 381             jnrlistA         = jjnr[jidx];
 382             jnrlistB         = jjnr[jidx+1];
 383             jnrlistC         = jjnr[jidx+2];
 384             jnrlistD         = jjnr[jidx+3];
 385             jnrlistE         = jjnr[jidx+4];
 386             jnrlistF         = jjnr[jidx+5];
 387             jnrlistG         = jjnr[jidx+6];
 388             jnrlistH         = jjnr[jidx+7];
 389             /* Sign of each element will be negative for non-real atoms.
 390              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 391              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 392              */
 393             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 394                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 395
 396             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 397             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 398             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 399             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 400             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 401             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 402             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 403             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 404             j_coord_offsetA  = DIM*jnrA;
 405             j_coord_offsetB  = DIM*jnrB;
 406             j_coord_offsetC  = DIM*jnrC;
 407             j_coord_offsetD  = DIM*jnrD;
 408             j_coord_offsetE  = DIM*jnrE;
 409             j_coord_offsetF  = DIM*jnrF;
 410             j_coord_offsetG  = DIM*jnrG;
 411             j_coord_offsetH  = DIM*jnrH;
 412
 413             /* load j atom coordinates */
 414             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 415                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 416                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 417                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 418                                                  &jx0,&jy0,&jz0);
 419
 420             /* Calculate displacement vector */
 421             dx00             = _mm256_sub_ps(ix0,jx0);
 422             dy00             = _mm256_sub_ps(iy0,jy0);
 423             dz00             = _mm256_sub_ps(iz0,jz0);
 424             dx10             = _mm256_sub_ps(ix1,jx0);
 425             dy10             = _mm256_sub_ps(iy1,jy0);
 426             dz10             = _mm256_sub_ps(iz1,jz0);
 427             dx20             = _mm256_sub_ps(ix2,jx0);
 428             dy20             = _mm256_sub_ps(iy2,jy0);
 429             dz20             = _mm256_sub_ps(iz2,jz0);
 430
 431             /* Calculate squared distance and things based on it */
 432             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 433             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 434             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 435
 436             rinv00           = avx256_invsqrt_f(rsq00);
 437             rinv10           = avx256_invsqrt_f(rsq10);
 438             rinv20           = avx256_invsqrt_f(rsq20);
 439
 440             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 441             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 442             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 443
 444             /* Load parameters for j particles */
 445             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 446                                                                  charge+jnrC+0,charge+jnrD+0,
 447                                                                  charge+jnrE+0,charge+jnrF+0,
 448                                                                  charge+jnrG+0,charge+jnrH+0);
 449
 450             fjx0             = _mm256_setzero_ps();
 451             fjy0             = _mm256_setzero_ps();
 452             fjz0             = _mm256_setzero_ps();
 453
 454             /**************************
 455              * CALCULATE INTERACTIONS *
 456              **************************/
 457
 458             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 459             {
 460
 461             /* Compute parameters for interactions between i and j atoms */
 462             qq00             = _mm256_mul_ps(iq0,jq0);
 463
 464             /* REACTION-FIELD ELECTROSTATICS */
 465             velec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_add_ps(rinv00,_mm256_mul_ps(krf,rsq00)),crf));
 466             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 467
 468             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 469
 470             /* Update potential sum for this i atom from the interaction with this j atom. */
 471             velec            = _mm256_and_ps(velec,cutoff_mask);
 472             velec            = _mm256_andnot_ps(dummy_mask,velec);
 473             velecsum         = _mm256_add_ps(velecsum,velec);
 474
 475             fscal            = felec;
 476
 477             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 478
 479             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 480
 481             /* Calculate temporary vectorial force */
 482             tx               = _mm256_mul_ps(fscal,dx00);
 483             ty               = _mm256_mul_ps(fscal,dy00);
 484             tz               = _mm256_mul_ps(fscal,dz00);
 485
 486             /* Update vectorial force */
 487             fix0             = _mm256_add_ps(fix0,tx);
 488             fiy0             = _mm256_add_ps(fiy0,ty);
 489             fiz0             = _mm256_add_ps(fiz0,tz);
 490
 491             fjx0             = _mm256_add_ps(fjx0,tx);
 492             fjy0             = _mm256_add_ps(fjy0,ty);
 493             fjz0             = _mm256_add_ps(fjz0,tz);
 494
 495             }
 496
 497             /**************************
 498              * CALCULATE INTERACTIONS *
 499              **************************/
 500
 501             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 502             {
 503
 504             /* Compute parameters for interactions between i and j atoms */
 505             qq10             = _mm256_mul_ps(iq1,jq0);
 506
 507             /* REACTION-FIELD ELECTROSTATICS */
 508             velec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_add_ps(rinv10,_mm256_mul_ps(krf,rsq10)),crf));
 509             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 510
 511             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 512
 513             /* Update potential sum for this i atom from the interaction with this j atom. */
 514             velec            = _mm256_and_ps(velec,cutoff_mask);
 515             velec            = _mm256_andnot_ps(dummy_mask,velec);
 516             velecsum         = _mm256_add_ps(velecsum,velec);
 517
 518             fscal            = felec;
 519
 520             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 521
 522             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 523
 524             /* Calculate temporary vectorial force */
 525             tx               = _mm256_mul_ps(fscal,dx10);
 526             ty               = _mm256_mul_ps(fscal,dy10);
 527             tz               = _mm256_mul_ps(fscal,dz10);
 528
 529             /* Update vectorial force */
 530             fix1             = _mm256_add_ps(fix1,tx);
 531             fiy1             = _mm256_add_ps(fiy1,ty);
 532             fiz1             = _mm256_add_ps(fiz1,tz);
 533
 534             fjx0             = _mm256_add_ps(fjx0,tx);
 535             fjy0             = _mm256_add_ps(fjy0,ty);
 536             fjz0             = _mm256_add_ps(fjz0,tz);
 537
 538             }
 539
 540             /**************************
 541              * CALCULATE INTERACTIONS *
 542              **************************/
 543
 544             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 545             {
 546
 547             /* Compute parameters for interactions between i and j atoms */
 548             qq20             = _mm256_mul_ps(iq2,jq0);
 549
 550             /* REACTION-FIELD ELECTROSTATICS */
 551             velec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_add_ps(rinv20,_mm256_mul_ps(krf,rsq20)),crf));
 552             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 553
 554             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
 555
 556             /* Update potential sum for this i atom from the interaction with this j atom. */
 557             velec            = _mm256_and_ps(velec,cutoff_mask);
 558             velec            = _mm256_andnot_ps(dummy_mask,velec);
 559             velecsum         = _mm256_add_ps(velecsum,velec);
 560
 561             fscal            = felec;
 562
 563             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 564
 565             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 566
 567             /* Calculate temporary vectorial force */
 568             tx               = _mm256_mul_ps(fscal,dx20);
 569             ty               = _mm256_mul_ps(fscal,dy20);
 570             tz               = _mm256_mul_ps(fscal,dz20);
 571
 572             /* Update vectorial force */
 573             fix2             = _mm256_add_ps(fix2,tx);
 574             fiy2             = _mm256_add_ps(fiy2,ty);
 575             fiz2             = _mm256_add_ps(fiz2,tz);
 576
 577             fjx0             = _mm256_add_ps(fjx0,tx);
 578             fjy0             = _mm256_add_ps(fjy0,ty);
 579             fjz0             = _mm256_add_ps(fjz0,tz);
 580
 581             }
 582
 583             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 584             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 585             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 586             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 587             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 588             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 589             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 590             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 591
 592             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 593
 594             /* Inner loop uses 111 flops */
 595         }
 596
 597         /* End of innermost loop */
 598
 599         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 600                                                  f+i_coord_offset,fshift+i_shift_offset);
 601
 602         ggid                        = gid[iidx];
 603         /* Update potential energies */
 604         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 605
 606         /* Increment number of inner iterations */
 607         inneriter                  += j_index_end - j_index_start;
 608
 609         /* Outer loop uses 19 flops */
 610     }
 611
 612     /* Increment number of outer iterations */
 613     outeriter        += nri;
 614
 615     /* Update outer/inner flops */
 616
 617     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*111);
 618 }
 619 /*
 620  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwNone_GeomW3P1_F_avx_256_single
 621  * Electrostatics interaction: ReactionField
 622  * VdW interaction:            None
 623  * Geometry:                   Water3-Particle
 624  * Calculate force/pot:        Force
 625  */
 626 void
 627 nb_kernel_ElecRFCut_VdwNone_GeomW3P1_F_avx_256_single
 628                     (t_nblist                    * gmx_restrict       nlist,
 629                      rvec                        * gmx_restrict          xx,
 630                      rvec                        * gmx_restrict          ff,
 631                      struct t_forcerec           * gmx_restrict          fr,
 632                      t_mdatoms                   * gmx_restrict     mdatoms,
 633                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 634                      t_nrnb                      * gmx_restrict        nrnb)
 635 {
 636     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 637      * just 0 for non-waters.
 638      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 639      * jnr indices corresponding to data put in the four positions in the SIMD register.
 640      */
 641     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 642     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 643     int              jnrA,jnrB,jnrC,jnrD;
 644     int              jnrE,jnrF,jnrG,jnrH;
 645     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 646     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
 647     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 648     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
 649     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 650     real             rcutoff_scalar;
 651     real             *shiftvec,*fshift,*x,*f;
 652     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
 653     real             scratch[4*DIM];
 654     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 655     real *           vdwioffsetptr0;
 656     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 657     real *           vdwioffsetptr1;
 658     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 659     real *           vdwioffsetptr2;
 660     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 661     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
 662     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 663     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 664     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 665     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 666     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 667     real             *charge;
 668     __m256           dummy_mask,cutoff_mask;
 669     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 670     __m256           one     = _mm256_set1_ps(1.0);
 671     __m256           two     = _mm256_set1_ps(2.0);
 672     x                = xx[0];
 673     f                = ff[0];
 674
 675     nri              = nlist->nri;
 676     iinr             = nlist->iinr;
 677     jindex           = nlist->jindex;
 678     jjnr             = nlist->jjnr;
 679     shiftidx         = nlist->shift;
 680     gid              = nlist->gid;
 681     shiftvec         = fr->shift_vec[0];
 682     fshift           = fr->fshift[0];
 683     facel            = _mm256_set1_ps(fr->ic->epsfac);
 684     charge           = mdatoms->chargeA;
 685     krf              = _mm256_set1_ps(fr->ic->k_rf);
 686     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 687     crf              = _mm256_set1_ps(fr->ic->c_rf);
 688
 689     /* Setup water-specific parameters */
 690     inr              = nlist->iinr[0];
 691     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 692     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 693     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 694
 695     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 696     rcutoff_scalar   = fr->ic->rcoulomb;
 697     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 698     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 699
 700     /* Avoid stupid compiler warnings */
 701     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 702     j_coord_offsetA = 0;
 703     j_coord_offsetB = 0;
 704     j_coord_offsetC = 0;
 705     j_coord_offsetD = 0;
 706     j_coord_offsetE = 0;
 707     j_coord_offsetF = 0;
 708     j_coord_offsetG = 0;
 709     j_coord_offsetH = 0;
 710
 711     outeriter        = 0;
 712     inneriter        = 0;
 713
 714     for(iidx=0;iidx<4*DIM;iidx++)
 715     {
 716         scratch[iidx] = 0.0;
 717     }
 718
 719     /* Start outer loop over neighborlists */
 720     for(iidx=0; iidx<nri; iidx++)
 721     {
 722         /* Load shift vector for this list */
 723         i_shift_offset   = DIM*shiftidx[iidx];
 724
 725         /* Load limits for loop over neighbors */
 726         j_index_start    = jindex[iidx];
 727         j_index_end      = jindex[iidx+1];
 728
 729         /* Get outer coordinate index */
 730         inr              = iinr[iidx];
 731         i_coord_offset   = DIM*inr;
 732
 733         /* Load i particle coords and add shift vector */
 734         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 735                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 736
 737         fix0             = _mm256_setzero_ps();
 738         fiy0             = _mm256_setzero_ps();
 739         fiz0             = _mm256_setzero_ps();
 740         fix1             = _mm256_setzero_ps();
 741         fiy1             = _mm256_setzero_ps();
 742         fiz1             = _mm256_setzero_ps();
 743         fix2             = _mm256_setzero_ps();
 744         fiy2             = _mm256_setzero_ps();
 745         fiz2             = _mm256_setzero_ps();
 746
 747         /* Start inner kernel loop */
 748         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 749         {
 750
 751             /* Get j neighbor index, and coordinate index */
 752             jnrA             = jjnr[jidx];
 753             jnrB             = jjnr[jidx+1];
 754             jnrC             = jjnr[jidx+2];
 755             jnrD             = jjnr[jidx+3];
 756             jnrE             = jjnr[jidx+4];
 757             jnrF             = jjnr[jidx+5];
 758             jnrG             = jjnr[jidx+6];
 759             jnrH             = jjnr[jidx+7];
 760             j_coord_offsetA  = DIM*jnrA;
 761             j_coord_offsetB  = DIM*jnrB;
 762             j_coord_offsetC  = DIM*jnrC;
 763             j_coord_offsetD  = DIM*jnrD;
 764             j_coord_offsetE  = DIM*jnrE;
 765             j_coord_offsetF  = DIM*jnrF;
 766             j_coord_offsetG  = DIM*jnrG;
 767             j_coord_offsetH  = DIM*jnrH;
 768
 769             /* load j atom coordinates */
 770             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 771                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 772                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 773                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 774                                                  &jx0,&jy0,&jz0);
 775
 776             /* Calculate displacement vector */
 777             dx00             = _mm256_sub_ps(ix0,jx0);
 778             dy00             = _mm256_sub_ps(iy0,jy0);
 779             dz00             = _mm256_sub_ps(iz0,jz0);
 780             dx10             = _mm256_sub_ps(ix1,jx0);
 781             dy10             = _mm256_sub_ps(iy1,jy0);
 782             dz10             = _mm256_sub_ps(iz1,jz0);
 783             dx20             = _mm256_sub_ps(ix2,jx0);
 784             dy20             = _mm256_sub_ps(iy2,jy0);
 785             dz20             = _mm256_sub_ps(iz2,jz0);
 786
 787             /* Calculate squared distance and things based on it */
 788             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 789             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 790             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 791
 792             rinv00           = avx256_invsqrt_f(rsq00);
 793             rinv10           = avx256_invsqrt_f(rsq10);
 794             rinv20           = avx256_invsqrt_f(rsq20);
 795
 796             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 797             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 798             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 799
 800             /* Load parameters for j particles */
 801             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 802                                                                  charge+jnrC+0,charge+jnrD+0,
 803                                                                  charge+jnrE+0,charge+jnrF+0,
 804                                                                  charge+jnrG+0,charge+jnrH+0);
 805
 806             fjx0             = _mm256_setzero_ps();
 807             fjy0             = _mm256_setzero_ps();
 808             fjz0             = _mm256_setzero_ps();
 809
 810             /**************************
 811              * CALCULATE INTERACTIONS *
 812              **************************/
 813
 814             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 815             {
 816
 817             /* Compute parameters for interactions between i and j atoms */
 818             qq00             = _mm256_mul_ps(iq0,jq0);
 819
 820             /* REACTION-FIELD ELECTROSTATICS */
 821             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
 822
 823             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 824
 825             fscal            = felec;
 826
 827             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 828
 829             /* Calculate temporary vectorial force */
 830             tx               = _mm256_mul_ps(fscal,dx00);
 831             ty               = _mm256_mul_ps(fscal,dy00);
 832             tz               = _mm256_mul_ps(fscal,dz00);
 833
 834             /* Update vectorial force */
 835             fix0             = _mm256_add_ps(fix0,tx);
 836             fiy0             = _mm256_add_ps(fiy0,ty);
 837             fiz0             = _mm256_add_ps(fiz0,tz);
 838
 839             fjx0             = _mm256_add_ps(fjx0,tx);
 840             fjy0             = _mm256_add_ps(fjy0,ty);
 841             fjz0             = _mm256_add_ps(fjz0,tz);
 842
 843             }
 844
 845             /**************************
 846              * CALCULATE INTERACTIONS *
 847              **************************/
 848
 849             if (gmx_mm256_any_lt(rsq10,rcutoff2))
 850             {
 851
 852             /* Compute parameters for interactions between i and j atoms */
 853             qq10             = _mm256_mul_ps(iq1,jq0);
 854
 855             /* REACTION-FIELD ELECTROSTATICS */
 856             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
 857
 858             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
 859
 860             fscal            = felec;
 861
 862             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 863
 864             /* Calculate temporary vectorial force */
 865             tx               = _mm256_mul_ps(fscal,dx10);
 866             ty               = _mm256_mul_ps(fscal,dy10);
 867             tz               = _mm256_mul_ps(fscal,dz10);
 868
 869             /* Update vectorial force */
 870             fix1             = _mm256_add_ps(fix1,tx);
 871             fiy1             = _mm256_add_ps(fiy1,ty);
 872             fiz1             = _mm256_add_ps(fiz1,tz);
 873
 874             fjx0             = _mm256_add_ps(fjx0,tx);
 875             fjy0             = _mm256_add_ps(fjy0,ty);
 876             fjz0             = _mm256_add_ps(fjz0,tz);
 877
 878             }
 879
 880             /**************************
 881              * CALCULATE INTERACTIONS *
 882              **************************/
 883
 884             if (gmx_mm256_any_lt(rsq20,rcutoff2))
 885             {
 886
 887             /* Compute parameters for interactions between i and j atoms */
 888             qq20             = _mm256_mul_ps(iq2,jq0);
 889
 890             /* REACTION-FIELD ELECTROSTATICS */
 891             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
 892
 893             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
 894
 895             fscal            = felec;
 896
 897             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 898
 899             /* Calculate temporary vectorial force */
 900             tx               = _mm256_mul_ps(fscal,dx20);
 901             ty               = _mm256_mul_ps(fscal,dy20);
 902             tz               = _mm256_mul_ps(fscal,dz20);
 903
 904             /* Update vectorial force */
 905             fix2             = _mm256_add_ps(fix2,tx);
 906             fiy2             = _mm256_add_ps(fiy2,ty);
 907             fiz2             = _mm256_add_ps(fiz2,tz);
 908
 909             fjx0             = _mm256_add_ps(fjx0,tx);
 910             fjy0             = _mm256_add_ps(fjy0,ty);
 911             fjz0             = _mm256_add_ps(fjz0,tz);
 912
 913             }
 914
 915             fjptrA             = f+j_coord_offsetA;
 916             fjptrB             = f+j_coord_offsetB;
 917             fjptrC             = f+j_coord_offsetC;
 918             fjptrD             = f+j_coord_offsetD;
 919             fjptrE             = f+j_coord_offsetE;
 920             fjptrF             = f+j_coord_offsetF;
 921             fjptrG             = f+j_coord_offsetG;
 922             fjptrH             = f+j_coord_offsetH;
 923
 924             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
 925
 926             /* Inner loop uses 93 flops */
 927         }
 928
 929         if(jidx<j_index_end)
 930         {
 931
 932             /* Get j neighbor index, and coordinate index */
 933             jnrlistA         = jjnr[jidx];
 934             jnrlistB         = jjnr[jidx+1];
 935             jnrlistC         = jjnr[jidx+2];
 936             jnrlistD         = jjnr[jidx+3];
 937             jnrlistE         = jjnr[jidx+4];
 938             jnrlistF         = jjnr[jidx+5];
 939             jnrlistG         = jjnr[jidx+6];
 940             jnrlistH         = jjnr[jidx+7];
 941             /* Sign of each element will be negative for non-real atoms.
 942              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 943              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 944              */
 945             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 946                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 947
 948             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 949             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 950             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 951             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 952             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 953             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 954             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 955             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 956             j_coord_offsetA  = DIM*jnrA;
 957             j_coord_offsetB  = DIM*jnrB;
 958             j_coord_offsetC  = DIM*jnrC;
 959             j_coord_offsetD  = DIM*jnrD;
 960             j_coord_offsetE  = DIM*jnrE;
 961             j_coord_offsetF  = DIM*jnrF;
 962             j_coord_offsetG  = DIM*jnrG;
 963             j_coord_offsetH  = DIM*jnrH;
 964
 965             /* load j atom coordinates */
 966             gmx_mm256_load_1rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 967                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 968                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 969                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 970                                                  &jx0,&jy0,&jz0);
 971
 972             /* Calculate displacement vector */
 973             dx00             = _mm256_sub_ps(ix0,jx0);
 974             dy00             = _mm256_sub_ps(iy0,jy0);
 975             dz00             = _mm256_sub_ps(iz0,jz0);
 976             dx10             = _mm256_sub_ps(ix1,jx0);
 977             dy10             = _mm256_sub_ps(iy1,jy0);
 978             dz10             = _mm256_sub_ps(iz1,jz0);
 979             dx20             = _mm256_sub_ps(ix2,jx0);
 980             dy20             = _mm256_sub_ps(iy2,jy0);
 981             dz20             = _mm256_sub_ps(iz2,jz0);
 982
 983             /* Calculate squared distance and things based on it */
 984             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 985             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 986             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 987
 988             rinv00           = avx256_invsqrt_f(rsq00);
 989             rinv10           = avx256_invsqrt_f(rsq10);
 990             rinv20           = avx256_invsqrt_f(rsq20);
 991
 992             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 993             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 994             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 995
 996             /* Load parameters for j particles */
 997             jq0              = gmx_mm256_load_8real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 998                                                                  charge+jnrC+0,charge+jnrD+0,
 999                                                                  charge+jnrE+0,charge+jnrF+0,
1000                                                                  charge+jnrG+0,charge+jnrH+0);
1001
1002             fjx0             = _mm256_setzero_ps();
1003             fjy0             = _mm256_setzero_ps();
1004             fjz0             = _mm256_setzero_ps();
1005
1006             /**************************
1007              * CALCULATE INTERACTIONS *
1008              **************************/
1009
1010             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1011             {
1012
1013             /* Compute parameters for interactions between i and j atoms */
1014             qq00             = _mm256_mul_ps(iq0,jq0);
1015
1016             /* REACTION-FIELD ELECTROSTATICS */
1017             felec            = _mm256_mul_ps(qq00,_mm256_sub_ps(_mm256_mul_ps(rinv00,rinvsq00),krf2));
1018
1019             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1020
1021             fscal            = felec;
1022
1023             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1024
1025             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1026
1027             /* Calculate temporary vectorial force */
1028             tx               = _mm256_mul_ps(fscal,dx00);
1029             ty               = _mm256_mul_ps(fscal,dy00);
1030             tz               = _mm256_mul_ps(fscal,dz00);
1031
1032             /* Update vectorial force */
1033             fix0             = _mm256_add_ps(fix0,tx);
1034             fiy0             = _mm256_add_ps(fiy0,ty);
1035             fiz0             = _mm256_add_ps(fiz0,tz);
1036
1037             fjx0             = _mm256_add_ps(fjx0,tx);
1038             fjy0             = _mm256_add_ps(fjy0,ty);
1039             fjz0             = _mm256_add_ps(fjz0,tz);
1040
1041             }
1042
1043             /**************************
1044              * CALCULATE INTERACTIONS *
1045              **************************/
1046
1047             if (gmx_mm256_any_lt(rsq10,rcutoff2))
1048             {
1049
1050             /* Compute parameters for interactions between i and j atoms */
1051             qq10             = _mm256_mul_ps(iq1,jq0);
1052
1053             /* REACTION-FIELD ELECTROSTATICS */
1054             felec            = _mm256_mul_ps(qq10,_mm256_sub_ps(_mm256_mul_ps(rinv10,rinvsq10),krf2));
1055
1056             cutoff_mask      = _mm256_cmp_ps(rsq10,rcutoff2,_CMP_LT_OQ);
1057
1058             fscal            = felec;
1059
1060             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1061
1062             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1063
1064             /* Calculate temporary vectorial force */
1065             tx               = _mm256_mul_ps(fscal,dx10);
1066             ty               = _mm256_mul_ps(fscal,dy10);
1067             tz               = _mm256_mul_ps(fscal,dz10);
1068
1069             /* Update vectorial force */
1070             fix1             = _mm256_add_ps(fix1,tx);
1071             fiy1             = _mm256_add_ps(fiy1,ty);
1072             fiz1             = _mm256_add_ps(fiz1,tz);
1073
1074             fjx0             = _mm256_add_ps(fjx0,tx);
1075             fjy0             = _mm256_add_ps(fjy0,ty);
1076             fjz0             = _mm256_add_ps(fjz0,tz);
1077
1078             }
1079
1080             /**************************
1081              * CALCULATE INTERACTIONS *
1082              **************************/
1083
1084             if (gmx_mm256_any_lt(rsq20,rcutoff2))
1085             {
1086
1087             /* Compute parameters for interactions between i and j atoms */
1088             qq20             = _mm256_mul_ps(iq2,jq0);
1089
1090             /* REACTION-FIELD ELECTROSTATICS */
1091             felec            = _mm256_mul_ps(qq20,_mm256_sub_ps(_mm256_mul_ps(rinv20,rinvsq20),krf2));
1092
1093             cutoff_mask      = _mm256_cmp_ps(rsq20,rcutoff2,_CMP_LT_OQ);
1094
1095             fscal            = felec;
1096
1097             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1098
1099             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1100
1101             /* Calculate temporary vectorial force */
1102             tx               = _mm256_mul_ps(fscal,dx20);
1103             ty               = _mm256_mul_ps(fscal,dy20);
1104             tz               = _mm256_mul_ps(fscal,dz20);
1105
1106             /* Update vectorial force */
1107             fix2             = _mm256_add_ps(fix2,tx);
1108             fiy2             = _mm256_add_ps(fiy2,ty);
1109             fiz2             = _mm256_add_ps(fiz2,tz);
1110
1111             fjx0             = _mm256_add_ps(fjx0,tx);
1112             fjy0             = _mm256_add_ps(fjy0,ty);
1113             fjz0             = _mm256_add_ps(fjz0,tz);
1114
1115             }
1116
1117             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1118             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1119             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1120             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1121             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1122             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1123             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1124             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1125
1126             gmx_mm256_decrement_1rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,fjx0,fjy0,fjz0);
1127
1128             /* Inner loop uses 93 flops */
1129         }
1130
1131         /* End of innermost loop */
1132
1133         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1134                                                  f+i_coord_offset,fshift+i_shift_offset);
1135
1136         /* Increment number of inner iterations */
1137         inneriter                  += j_index_end - j_index_start;
1138
1139         /* Outer loop uses 18 flops */
1140     }
1141
1142     /* Increment number of outer iterations */
1143     outeriter        += nri;
1144
1145     /* Update outer/inner flops */
1146
1147     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*93);
1148 }