src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwNone_GeomW3W3_avx_256_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_256_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/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_avx_256_single.h"
  50 #include "kernelutil_x86_avx_256_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW3W3_VF_avx_256_single
  54  * Electrostatics interaction: Coulomb
  55  * VdW interaction:            None
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecCoul_VdwNone_GeomW3W3_VF_avx_256_single
  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,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight 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              jnrE,jnrF,jnrG,jnrH;
  78     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  79     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  80     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  81     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  82     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  83     real             rcutoff_scalar;
  84     real             *shiftvec,*fshift,*x,*f;
  85     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  86     real             scratch[4*DIM];
  87     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  88     real *           vdwioffsetptr0;
  89     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  90     real *           vdwioffsetptr1;
  91     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  92     real *           vdwioffsetptr2;
  93     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  97     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  99     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 101     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
 102     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
 103     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 104     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 105     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 106     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 107     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 108     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 109     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 110     real             *charge;
 111     __m256           dummy_mask,cutoff_mask;
 112     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 113     __m256           one     = _mm256_set1_ps(1.0);
 114     __m256           two     = _mm256_set1_ps(2.0);
 115     x                = xx[0];
 116     f                = ff[0];
 117
 118     nri              = nlist->nri;
 119     iinr             = nlist->iinr;
 120     jindex           = nlist->jindex;
 121     jjnr             = nlist->jjnr;
 122     shiftidx         = nlist->shift;
 123     gid              = nlist->gid;
 124     shiftvec         = fr->shift_vec[0];
 125     fshift           = fr->fshift[0];
 126     facel            = _mm256_set1_ps(fr->epsfac);
 127     charge           = mdatoms->chargeA;
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
 132     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 133     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 134
 135     jq0              = _mm256_set1_ps(charge[inr+0]);
 136     jq1              = _mm256_set1_ps(charge[inr+1]);
 137     jq2              = _mm256_set1_ps(charge[inr+2]);
 138     qq00             = _mm256_mul_ps(iq0,jq0);
 139     qq01             = _mm256_mul_ps(iq0,jq1);
 140     qq02             = _mm256_mul_ps(iq0,jq2);
 141     qq10             = _mm256_mul_ps(iq1,jq0);
 142     qq11             = _mm256_mul_ps(iq1,jq1);
 143     qq12             = _mm256_mul_ps(iq1,jq2);
 144     qq20             = _mm256_mul_ps(iq2,jq0);
 145     qq21             = _mm256_mul_ps(iq2,jq1);
 146     qq22             = _mm256_mul_ps(iq2,jq2);
 147
 148     /* Avoid stupid compiler warnings */
 149     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 150     j_coord_offsetA = 0;
 151     j_coord_offsetB = 0;
 152     j_coord_offsetC = 0;
 153     j_coord_offsetD = 0;
 154     j_coord_offsetE = 0;
 155     j_coord_offsetF = 0;
 156     j_coord_offsetG = 0;
 157     j_coord_offsetH = 0;
 158
 159     outeriter        = 0;
 160     inneriter        = 0;
 161
 162     for(iidx=0;iidx<4*DIM;iidx++)
 163     {
 164         scratch[iidx] = 0.0;
 165     }
 166
 167     /* Start outer loop over neighborlists */
 168     for(iidx=0; iidx<nri; iidx++)
 169     {
 170         /* Load shift vector for this list */
 171         i_shift_offset   = DIM*shiftidx[iidx];
 172
 173         /* Load limits for loop over neighbors */
 174         j_index_start    = jindex[iidx];
 175         j_index_end      = jindex[iidx+1];
 176
 177         /* Get outer coordinate index */
 178         inr              = iinr[iidx];
 179         i_coord_offset   = DIM*inr;
 180
 181         /* Load i particle coords and add shift vector */
 182         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 183                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 184
 185         fix0             = _mm256_setzero_ps();
 186         fiy0             = _mm256_setzero_ps();
 187         fiz0             = _mm256_setzero_ps();
 188         fix1             = _mm256_setzero_ps();
 189         fiy1             = _mm256_setzero_ps();
 190         fiz1             = _mm256_setzero_ps();
 191         fix2             = _mm256_setzero_ps();
 192         fiy2             = _mm256_setzero_ps();
 193         fiz2             = _mm256_setzero_ps();
 194
 195         /* Reset potential sums */
 196         velecsum         = _mm256_setzero_ps();
 197
 198         /* Start inner kernel loop */
 199         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 200         {
 201
 202             /* Get j neighbor index, and coordinate index */
 203             jnrA             = jjnr[jidx];
 204             jnrB             = jjnr[jidx+1];
 205             jnrC             = jjnr[jidx+2];
 206             jnrD             = jjnr[jidx+3];
 207             jnrE             = jjnr[jidx+4];
 208             jnrF             = jjnr[jidx+5];
 209             jnrG             = jjnr[jidx+6];
 210             jnrH             = jjnr[jidx+7];
 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             j_coord_offsetE  = DIM*jnrE;
 216             j_coord_offsetF  = DIM*jnrF;
 217             j_coord_offsetG  = DIM*jnrG;
 218             j_coord_offsetH  = DIM*jnrH;
 219
 220             /* load j atom coordinates */
 221             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 222                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 223                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 224                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 225                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 226
 227             /* Calculate displacement vector */
 228             dx00             = _mm256_sub_ps(ix0,jx0);
 229             dy00             = _mm256_sub_ps(iy0,jy0);
 230             dz00             = _mm256_sub_ps(iz0,jz0);
 231             dx01             = _mm256_sub_ps(ix0,jx1);
 232             dy01             = _mm256_sub_ps(iy0,jy1);
 233             dz01             = _mm256_sub_ps(iz0,jz1);
 234             dx02             = _mm256_sub_ps(ix0,jx2);
 235             dy02             = _mm256_sub_ps(iy0,jy2);
 236             dz02             = _mm256_sub_ps(iz0,jz2);
 237             dx10             = _mm256_sub_ps(ix1,jx0);
 238             dy10             = _mm256_sub_ps(iy1,jy0);
 239             dz10             = _mm256_sub_ps(iz1,jz0);
 240             dx11             = _mm256_sub_ps(ix1,jx1);
 241             dy11             = _mm256_sub_ps(iy1,jy1);
 242             dz11             = _mm256_sub_ps(iz1,jz1);
 243             dx12             = _mm256_sub_ps(ix1,jx2);
 244             dy12             = _mm256_sub_ps(iy1,jy2);
 245             dz12             = _mm256_sub_ps(iz1,jz2);
 246             dx20             = _mm256_sub_ps(ix2,jx0);
 247             dy20             = _mm256_sub_ps(iy2,jy0);
 248             dz20             = _mm256_sub_ps(iz2,jz0);
 249             dx21             = _mm256_sub_ps(ix2,jx1);
 250             dy21             = _mm256_sub_ps(iy2,jy1);
 251             dz21             = _mm256_sub_ps(iz2,jz1);
 252             dx22             = _mm256_sub_ps(ix2,jx2);
 253             dy22             = _mm256_sub_ps(iy2,jy2);
 254             dz22             = _mm256_sub_ps(iz2,jz2);
 255
 256             /* Calculate squared distance and things based on it */
 257             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 258             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 259             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 260             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 261             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 262             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 263             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 264             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 265             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 266
 267             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 268             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 269             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 270             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 271             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 272             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 273             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 274             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 275             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 276
 277             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 278             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 279             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 280             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 281             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 282             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 283             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 284             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 285             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 286
 287             fjx0             = _mm256_setzero_ps();
 288             fjy0             = _mm256_setzero_ps();
 289             fjz0             = _mm256_setzero_ps();
 290             fjx1             = _mm256_setzero_ps();
 291             fjy1             = _mm256_setzero_ps();
 292             fjz1             = _mm256_setzero_ps();
 293             fjx2             = _mm256_setzero_ps();
 294             fjy2             = _mm256_setzero_ps();
 295             fjz2             = _mm256_setzero_ps();
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             /* COULOMB ELECTROSTATICS */
 302             velec            = _mm256_mul_ps(qq00,rinv00);
 303             felec            = _mm256_mul_ps(velec,rinvsq00);
 304
 305             /* Update potential sum for this i atom from the interaction with this j atom. */
 306             velecsum         = _mm256_add_ps(velecsum,velec);
 307
 308             fscal            = felec;
 309
 310             /* Calculate temporary vectorial force */
 311             tx               = _mm256_mul_ps(fscal,dx00);
 312             ty               = _mm256_mul_ps(fscal,dy00);
 313             tz               = _mm256_mul_ps(fscal,dz00);
 314
 315             /* Update vectorial force */
 316             fix0             = _mm256_add_ps(fix0,tx);
 317             fiy0             = _mm256_add_ps(fiy0,ty);
 318             fiz0             = _mm256_add_ps(fiz0,tz);
 319
 320             fjx0             = _mm256_add_ps(fjx0,tx);
 321             fjy0             = _mm256_add_ps(fjy0,ty);
 322             fjz0             = _mm256_add_ps(fjz0,tz);
 323
 324             /**************************
 325              * CALCULATE INTERACTIONS *
 326              **************************/
 327
 328             /* COULOMB ELECTROSTATICS */
 329             velec            = _mm256_mul_ps(qq01,rinv01);
 330             felec            = _mm256_mul_ps(velec,rinvsq01);
 331
 332             /* Update potential sum for this i atom from the interaction with this j atom. */
 333             velecsum         = _mm256_add_ps(velecsum,velec);
 334
 335             fscal            = felec;
 336
 337             /* Calculate temporary vectorial force */
 338             tx               = _mm256_mul_ps(fscal,dx01);
 339             ty               = _mm256_mul_ps(fscal,dy01);
 340             tz               = _mm256_mul_ps(fscal,dz01);
 341
 342             /* Update vectorial force */
 343             fix0             = _mm256_add_ps(fix0,tx);
 344             fiy0             = _mm256_add_ps(fiy0,ty);
 345             fiz0             = _mm256_add_ps(fiz0,tz);
 346
 347             fjx1             = _mm256_add_ps(fjx1,tx);
 348             fjy1             = _mm256_add_ps(fjy1,ty);
 349             fjz1             = _mm256_add_ps(fjz1,tz);
 350
 351             /**************************
 352              * CALCULATE INTERACTIONS *
 353              **************************/
 354
 355             /* COULOMB ELECTROSTATICS */
 356             velec            = _mm256_mul_ps(qq02,rinv02);
 357             felec            = _mm256_mul_ps(velec,rinvsq02);
 358
 359             /* Update potential sum for this i atom from the interaction with this j atom. */
 360             velecsum         = _mm256_add_ps(velecsum,velec);
 361
 362             fscal            = felec;
 363
 364             /* Calculate temporary vectorial force */
 365             tx               = _mm256_mul_ps(fscal,dx02);
 366             ty               = _mm256_mul_ps(fscal,dy02);
 367             tz               = _mm256_mul_ps(fscal,dz02);
 368
 369             /* Update vectorial force */
 370             fix0             = _mm256_add_ps(fix0,tx);
 371             fiy0             = _mm256_add_ps(fiy0,ty);
 372             fiz0             = _mm256_add_ps(fiz0,tz);
 373
 374             fjx2             = _mm256_add_ps(fjx2,tx);
 375             fjy2             = _mm256_add_ps(fjy2,ty);
 376             fjz2             = _mm256_add_ps(fjz2,tz);
 377
 378             /**************************
 379              * CALCULATE INTERACTIONS *
 380              **************************/
 381
 382             /* COULOMB ELECTROSTATICS */
 383             velec            = _mm256_mul_ps(qq10,rinv10);
 384             felec            = _mm256_mul_ps(velec,rinvsq10);
 385
 386             /* Update potential sum for this i atom from the interaction with this j atom. */
 387             velecsum         = _mm256_add_ps(velecsum,velec);
 388
 389             fscal            = felec;
 390
 391             /* Calculate temporary vectorial force */
 392             tx               = _mm256_mul_ps(fscal,dx10);
 393             ty               = _mm256_mul_ps(fscal,dy10);
 394             tz               = _mm256_mul_ps(fscal,dz10);
 395
 396             /* Update vectorial force */
 397             fix1             = _mm256_add_ps(fix1,tx);
 398             fiy1             = _mm256_add_ps(fiy1,ty);
 399             fiz1             = _mm256_add_ps(fiz1,tz);
 400
 401             fjx0             = _mm256_add_ps(fjx0,tx);
 402             fjy0             = _mm256_add_ps(fjy0,ty);
 403             fjz0             = _mm256_add_ps(fjz0,tz);
 404
 405             /**************************
 406              * CALCULATE INTERACTIONS *
 407              **************************/
 408
 409             /* COULOMB ELECTROSTATICS */
 410             velec            = _mm256_mul_ps(qq11,rinv11);
 411             felec            = _mm256_mul_ps(velec,rinvsq11);
 412
 413             /* Update potential sum for this i atom from the interaction with this j atom. */
 414             velecsum         = _mm256_add_ps(velecsum,velec);
 415
 416             fscal            = felec;
 417
 418             /* Calculate temporary vectorial force */
 419             tx               = _mm256_mul_ps(fscal,dx11);
 420             ty               = _mm256_mul_ps(fscal,dy11);
 421             tz               = _mm256_mul_ps(fscal,dz11);
 422
 423             /* Update vectorial force */
 424             fix1             = _mm256_add_ps(fix1,tx);
 425             fiy1             = _mm256_add_ps(fiy1,ty);
 426             fiz1             = _mm256_add_ps(fiz1,tz);
 427
 428             fjx1             = _mm256_add_ps(fjx1,tx);
 429             fjy1             = _mm256_add_ps(fjy1,ty);
 430             fjz1             = _mm256_add_ps(fjz1,tz);
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             /* COULOMB ELECTROSTATICS */
 437             velec            = _mm256_mul_ps(qq12,rinv12);
 438             felec            = _mm256_mul_ps(velec,rinvsq12);
 439
 440             /* Update potential sum for this i atom from the interaction with this j atom. */
 441             velecsum         = _mm256_add_ps(velecsum,velec);
 442
 443             fscal            = felec;
 444
 445             /* Calculate temporary vectorial force */
 446             tx               = _mm256_mul_ps(fscal,dx12);
 447             ty               = _mm256_mul_ps(fscal,dy12);
 448             tz               = _mm256_mul_ps(fscal,dz12);
 449
 450             /* Update vectorial force */
 451             fix1             = _mm256_add_ps(fix1,tx);
 452             fiy1             = _mm256_add_ps(fiy1,ty);
 453             fiz1             = _mm256_add_ps(fiz1,tz);
 454
 455             fjx2             = _mm256_add_ps(fjx2,tx);
 456             fjy2             = _mm256_add_ps(fjy2,ty);
 457             fjz2             = _mm256_add_ps(fjz2,tz);
 458
 459             /**************************
 460              * CALCULATE INTERACTIONS *
 461              **************************/
 462
 463             /* COULOMB ELECTROSTATICS */
 464             velec            = _mm256_mul_ps(qq20,rinv20);
 465             felec            = _mm256_mul_ps(velec,rinvsq20);
 466
 467             /* Update potential sum for this i atom from the interaction with this j atom. */
 468             velecsum         = _mm256_add_ps(velecsum,velec);
 469
 470             fscal            = felec;
 471
 472             /* Calculate temporary vectorial force */
 473             tx               = _mm256_mul_ps(fscal,dx20);
 474             ty               = _mm256_mul_ps(fscal,dy20);
 475             tz               = _mm256_mul_ps(fscal,dz20);
 476
 477             /* Update vectorial force */
 478             fix2             = _mm256_add_ps(fix2,tx);
 479             fiy2             = _mm256_add_ps(fiy2,ty);
 480             fiz2             = _mm256_add_ps(fiz2,tz);
 481
 482             fjx0             = _mm256_add_ps(fjx0,tx);
 483             fjy0             = _mm256_add_ps(fjy0,ty);
 484             fjz0             = _mm256_add_ps(fjz0,tz);
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             /* COULOMB ELECTROSTATICS */
 491             velec            = _mm256_mul_ps(qq21,rinv21);
 492             felec            = _mm256_mul_ps(velec,rinvsq21);
 493
 494             /* Update potential sum for this i atom from the interaction with this j atom. */
 495             velecsum         = _mm256_add_ps(velecsum,velec);
 496
 497             fscal            = felec;
 498
 499             /* Calculate temporary vectorial force */
 500             tx               = _mm256_mul_ps(fscal,dx21);
 501             ty               = _mm256_mul_ps(fscal,dy21);
 502             tz               = _mm256_mul_ps(fscal,dz21);
 503
 504             /* Update vectorial force */
 505             fix2             = _mm256_add_ps(fix2,tx);
 506             fiy2             = _mm256_add_ps(fiy2,ty);
 507             fiz2             = _mm256_add_ps(fiz2,tz);
 508
 509             fjx1             = _mm256_add_ps(fjx1,tx);
 510             fjy1             = _mm256_add_ps(fjy1,ty);
 511             fjz1             = _mm256_add_ps(fjz1,tz);
 512
 513             /**************************
 514              * CALCULATE INTERACTIONS *
 515              **************************/
 516
 517             /* COULOMB ELECTROSTATICS */
 518             velec            = _mm256_mul_ps(qq22,rinv22);
 519             felec            = _mm256_mul_ps(velec,rinvsq22);
 520
 521             /* Update potential sum for this i atom from the interaction with this j atom. */
 522             velecsum         = _mm256_add_ps(velecsum,velec);
 523
 524             fscal            = felec;
 525
 526             /* Calculate temporary vectorial force */
 527             tx               = _mm256_mul_ps(fscal,dx22);
 528             ty               = _mm256_mul_ps(fscal,dy22);
 529             tz               = _mm256_mul_ps(fscal,dz22);
 530
 531             /* Update vectorial force */
 532             fix2             = _mm256_add_ps(fix2,tx);
 533             fiy2             = _mm256_add_ps(fiy2,ty);
 534             fiz2             = _mm256_add_ps(fiz2,tz);
 535
 536             fjx2             = _mm256_add_ps(fjx2,tx);
 537             fjy2             = _mm256_add_ps(fjy2,ty);
 538             fjz2             = _mm256_add_ps(fjz2,tz);
 539
 540             fjptrA             = f+j_coord_offsetA;
 541             fjptrB             = f+j_coord_offsetB;
 542             fjptrC             = f+j_coord_offsetC;
 543             fjptrD             = f+j_coord_offsetD;
 544             fjptrE             = f+j_coord_offsetE;
 545             fjptrF             = f+j_coord_offsetF;
 546             fjptrG             = f+j_coord_offsetG;
 547             fjptrH             = f+j_coord_offsetH;
 548
 549             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 550                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 551
 552             /* Inner loop uses 243 flops */
 553         }
 554
 555         if(jidx<j_index_end)
 556         {
 557
 558             /* Get j neighbor index, and coordinate index */
 559             jnrlistA         = jjnr[jidx];
 560             jnrlistB         = jjnr[jidx+1];
 561             jnrlistC         = jjnr[jidx+2];
 562             jnrlistD         = jjnr[jidx+3];
 563             jnrlistE         = jjnr[jidx+4];
 564             jnrlistF         = jjnr[jidx+5];
 565             jnrlistG         = jjnr[jidx+6];
 566             jnrlistH         = jjnr[jidx+7];
 567             /* Sign of each element will be negative for non-real atoms.
 568              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 569              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 570              */
 571             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 572                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 573
 574             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 575             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 576             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 577             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 578             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 579             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 580             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 581             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 582             j_coord_offsetA  = DIM*jnrA;
 583             j_coord_offsetB  = DIM*jnrB;
 584             j_coord_offsetC  = DIM*jnrC;
 585             j_coord_offsetD  = DIM*jnrD;
 586             j_coord_offsetE  = DIM*jnrE;
 587             j_coord_offsetF  = DIM*jnrF;
 588             j_coord_offsetG  = DIM*jnrG;
 589             j_coord_offsetH  = DIM*jnrH;
 590
 591             /* load j atom coordinates */
 592             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 593                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 594                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 595                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 596                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 597
 598             /* Calculate displacement vector */
 599             dx00             = _mm256_sub_ps(ix0,jx0);
 600             dy00             = _mm256_sub_ps(iy0,jy0);
 601             dz00             = _mm256_sub_ps(iz0,jz0);
 602             dx01             = _mm256_sub_ps(ix0,jx1);
 603             dy01             = _mm256_sub_ps(iy0,jy1);
 604             dz01             = _mm256_sub_ps(iz0,jz1);
 605             dx02             = _mm256_sub_ps(ix0,jx2);
 606             dy02             = _mm256_sub_ps(iy0,jy2);
 607             dz02             = _mm256_sub_ps(iz0,jz2);
 608             dx10             = _mm256_sub_ps(ix1,jx0);
 609             dy10             = _mm256_sub_ps(iy1,jy0);
 610             dz10             = _mm256_sub_ps(iz1,jz0);
 611             dx11             = _mm256_sub_ps(ix1,jx1);
 612             dy11             = _mm256_sub_ps(iy1,jy1);
 613             dz11             = _mm256_sub_ps(iz1,jz1);
 614             dx12             = _mm256_sub_ps(ix1,jx2);
 615             dy12             = _mm256_sub_ps(iy1,jy2);
 616             dz12             = _mm256_sub_ps(iz1,jz2);
 617             dx20             = _mm256_sub_ps(ix2,jx0);
 618             dy20             = _mm256_sub_ps(iy2,jy0);
 619             dz20             = _mm256_sub_ps(iz2,jz0);
 620             dx21             = _mm256_sub_ps(ix2,jx1);
 621             dy21             = _mm256_sub_ps(iy2,jy1);
 622             dz21             = _mm256_sub_ps(iz2,jz1);
 623             dx22             = _mm256_sub_ps(ix2,jx2);
 624             dy22             = _mm256_sub_ps(iy2,jy2);
 625             dz22             = _mm256_sub_ps(iz2,jz2);
 626
 627             /* Calculate squared distance and things based on it */
 628             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 629             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
 630             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
 631             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
 632             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 633             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 634             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
 635             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 636             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 637
 638             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 639             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
 640             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
 641             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
 642             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 643             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 644             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
 645             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 646             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 647
 648             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 649             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
 650             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
 651             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
 652             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 653             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 654             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
 655             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 656             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 657
 658             fjx0             = _mm256_setzero_ps();
 659             fjy0             = _mm256_setzero_ps();
 660             fjz0             = _mm256_setzero_ps();
 661             fjx1             = _mm256_setzero_ps();
 662             fjy1             = _mm256_setzero_ps();
 663             fjz1             = _mm256_setzero_ps();
 664             fjx2             = _mm256_setzero_ps();
 665             fjy2             = _mm256_setzero_ps();
 666             fjz2             = _mm256_setzero_ps();
 667
 668             /**************************
 669              * CALCULATE INTERACTIONS *
 670              **************************/
 671
 672             /* COULOMB ELECTROSTATICS */
 673             velec            = _mm256_mul_ps(qq00,rinv00);
 674             felec            = _mm256_mul_ps(velec,rinvsq00);
 675
 676             /* Update potential sum for this i atom from the interaction with this j atom. */
 677             velec            = _mm256_andnot_ps(dummy_mask,velec);
 678             velecsum         = _mm256_add_ps(velecsum,velec);
 679
 680             fscal            = felec;
 681
 682             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 683
 684             /* Calculate temporary vectorial force */
 685             tx               = _mm256_mul_ps(fscal,dx00);
 686             ty               = _mm256_mul_ps(fscal,dy00);
 687             tz               = _mm256_mul_ps(fscal,dz00);
 688
 689             /* Update vectorial force */
 690             fix0             = _mm256_add_ps(fix0,tx);
 691             fiy0             = _mm256_add_ps(fiy0,ty);
 692             fiz0             = _mm256_add_ps(fiz0,tz);
 693
 694             fjx0             = _mm256_add_ps(fjx0,tx);
 695             fjy0             = _mm256_add_ps(fjy0,ty);
 696             fjz0             = _mm256_add_ps(fjz0,tz);
 697
 698             /**************************
 699              * CALCULATE INTERACTIONS *
 700              **************************/
 701
 702             /* COULOMB ELECTROSTATICS */
 703             velec            = _mm256_mul_ps(qq01,rinv01);
 704             felec            = _mm256_mul_ps(velec,rinvsq01);
 705
 706             /* Update potential sum for this i atom from the interaction with this j atom. */
 707             velec            = _mm256_andnot_ps(dummy_mask,velec);
 708             velecsum         = _mm256_add_ps(velecsum,velec);
 709
 710             fscal            = felec;
 711
 712             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 713
 714             /* Calculate temporary vectorial force */
 715             tx               = _mm256_mul_ps(fscal,dx01);
 716             ty               = _mm256_mul_ps(fscal,dy01);
 717             tz               = _mm256_mul_ps(fscal,dz01);
 718
 719             /* Update vectorial force */
 720             fix0             = _mm256_add_ps(fix0,tx);
 721             fiy0             = _mm256_add_ps(fiy0,ty);
 722             fiz0             = _mm256_add_ps(fiz0,tz);
 723
 724             fjx1             = _mm256_add_ps(fjx1,tx);
 725             fjy1             = _mm256_add_ps(fjy1,ty);
 726             fjz1             = _mm256_add_ps(fjz1,tz);
 727
 728             /**************************
 729              * CALCULATE INTERACTIONS *
 730              **************************/
 731
 732             /* COULOMB ELECTROSTATICS */
 733             velec            = _mm256_mul_ps(qq02,rinv02);
 734             felec            = _mm256_mul_ps(velec,rinvsq02);
 735
 736             /* Update potential sum for this i atom from the interaction with this j atom. */
 737             velec            = _mm256_andnot_ps(dummy_mask,velec);
 738             velecsum         = _mm256_add_ps(velecsum,velec);
 739
 740             fscal            = felec;
 741
 742             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 743
 744             /* Calculate temporary vectorial force */
 745             tx               = _mm256_mul_ps(fscal,dx02);
 746             ty               = _mm256_mul_ps(fscal,dy02);
 747             tz               = _mm256_mul_ps(fscal,dz02);
 748
 749             /* Update vectorial force */
 750             fix0             = _mm256_add_ps(fix0,tx);
 751             fiy0             = _mm256_add_ps(fiy0,ty);
 752             fiz0             = _mm256_add_ps(fiz0,tz);
 753
 754             fjx2             = _mm256_add_ps(fjx2,tx);
 755             fjy2             = _mm256_add_ps(fjy2,ty);
 756             fjz2             = _mm256_add_ps(fjz2,tz);
 757
 758             /**************************
 759              * CALCULATE INTERACTIONS *
 760              **************************/
 761
 762             /* COULOMB ELECTROSTATICS */
 763             velec            = _mm256_mul_ps(qq10,rinv10);
 764             felec            = _mm256_mul_ps(velec,rinvsq10);
 765
 766             /* Update potential sum for this i atom from the interaction with this j atom. */
 767             velec            = _mm256_andnot_ps(dummy_mask,velec);
 768             velecsum         = _mm256_add_ps(velecsum,velec);
 769
 770             fscal            = felec;
 771
 772             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 773
 774             /* Calculate temporary vectorial force */
 775             tx               = _mm256_mul_ps(fscal,dx10);
 776             ty               = _mm256_mul_ps(fscal,dy10);
 777             tz               = _mm256_mul_ps(fscal,dz10);
 778
 779             /* Update vectorial force */
 780             fix1             = _mm256_add_ps(fix1,tx);
 781             fiy1             = _mm256_add_ps(fiy1,ty);
 782             fiz1             = _mm256_add_ps(fiz1,tz);
 783
 784             fjx0             = _mm256_add_ps(fjx0,tx);
 785             fjy0             = _mm256_add_ps(fjy0,ty);
 786             fjz0             = _mm256_add_ps(fjz0,tz);
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             /* COULOMB ELECTROSTATICS */
 793             velec            = _mm256_mul_ps(qq11,rinv11);
 794             felec            = _mm256_mul_ps(velec,rinvsq11);
 795
 796             /* Update potential sum for this i atom from the interaction with this j atom. */
 797             velec            = _mm256_andnot_ps(dummy_mask,velec);
 798             velecsum         = _mm256_add_ps(velecsum,velec);
 799
 800             fscal            = felec;
 801
 802             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 803
 804             /* Calculate temporary vectorial force */
 805             tx               = _mm256_mul_ps(fscal,dx11);
 806             ty               = _mm256_mul_ps(fscal,dy11);
 807             tz               = _mm256_mul_ps(fscal,dz11);
 808
 809             /* Update vectorial force */
 810             fix1             = _mm256_add_ps(fix1,tx);
 811             fiy1             = _mm256_add_ps(fiy1,ty);
 812             fiz1             = _mm256_add_ps(fiz1,tz);
 813
 814             fjx1             = _mm256_add_ps(fjx1,tx);
 815             fjy1             = _mm256_add_ps(fjy1,ty);
 816             fjz1             = _mm256_add_ps(fjz1,tz);
 817
 818             /**************************
 819              * CALCULATE INTERACTIONS *
 820              **************************/
 821
 822             /* COULOMB ELECTROSTATICS */
 823             velec            = _mm256_mul_ps(qq12,rinv12);
 824             felec            = _mm256_mul_ps(velec,rinvsq12);
 825
 826             /* Update potential sum for this i atom from the interaction with this j atom. */
 827             velec            = _mm256_andnot_ps(dummy_mask,velec);
 828             velecsum         = _mm256_add_ps(velecsum,velec);
 829
 830             fscal            = felec;
 831
 832             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 833
 834             /* Calculate temporary vectorial force */
 835             tx               = _mm256_mul_ps(fscal,dx12);
 836             ty               = _mm256_mul_ps(fscal,dy12);
 837             tz               = _mm256_mul_ps(fscal,dz12);
 838
 839             /* Update vectorial force */
 840             fix1             = _mm256_add_ps(fix1,tx);
 841             fiy1             = _mm256_add_ps(fiy1,ty);
 842             fiz1             = _mm256_add_ps(fiz1,tz);
 843
 844             fjx2             = _mm256_add_ps(fjx2,tx);
 845             fjy2             = _mm256_add_ps(fjy2,ty);
 846             fjz2             = _mm256_add_ps(fjz2,tz);
 847
 848             /**************************
 849              * CALCULATE INTERACTIONS *
 850              **************************/
 851
 852             /* COULOMB ELECTROSTATICS */
 853             velec            = _mm256_mul_ps(qq20,rinv20);
 854             felec            = _mm256_mul_ps(velec,rinvsq20);
 855
 856             /* Update potential sum for this i atom from the interaction with this j atom. */
 857             velec            = _mm256_andnot_ps(dummy_mask,velec);
 858             velecsum         = _mm256_add_ps(velecsum,velec);
 859
 860             fscal            = felec;
 861
 862             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 863
 864             /* Calculate temporary vectorial force */
 865             tx               = _mm256_mul_ps(fscal,dx20);
 866             ty               = _mm256_mul_ps(fscal,dy20);
 867             tz               = _mm256_mul_ps(fscal,dz20);
 868
 869             /* Update vectorial force */
 870             fix2             = _mm256_add_ps(fix2,tx);
 871             fiy2             = _mm256_add_ps(fiy2,ty);
 872             fiz2             = _mm256_add_ps(fiz2,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              * CALCULATE INTERACTIONS *
 880              **************************/
 881
 882             /* COULOMB ELECTROSTATICS */
 883             velec            = _mm256_mul_ps(qq21,rinv21);
 884             felec            = _mm256_mul_ps(velec,rinvsq21);
 885
 886             /* Update potential sum for this i atom from the interaction with this j atom. */
 887             velec            = _mm256_andnot_ps(dummy_mask,velec);
 888             velecsum         = _mm256_add_ps(velecsum,velec);
 889
 890             fscal            = felec;
 891
 892             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 893
 894             /* Calculate temporary vectorial force */
 895             tx               = _mm256_mul_ps(fscal,dx21);
 896             ty               = _mm256_mul_ps(fscal,dy21);
 897             tz               = _mm256_mul_ps(fscal,dz21);
 898
 899             /* Update vectorial force */
 900             fix2             = _mm256_add_ps(fix2,tx);
 901             fiy2             = _mm256_add_ps(fiy2,ty);
 902             fiz2             = _mm256_add_ps(fiz2,tz);
 903
 904             fjx1             = _mm256_add_ps(fjx1,tx);
 905             fjy1             = _mm256_add_ps(fjy1,ty);
 906             fjz1             = _mm256_add_ps(fjz1,tz);
 907
 908             /**************************
 909              * CALCULATE INTERACTIONS *
 910              **************************/
 911
 912             /* COULOMB ELECTROSTATICS */
 913             velec            = _mm256_mul_ps(qq22,rinv22);
 914             felec            = _mm256_mul_ps(velec,rinvsq22);
 915
 916             /* Update potential sum for this i atom from the interaction with this j atom. */
 917             velec            = _mm256_andnot_ps(dummy_mask,velec);
 918             velecsum         = _mm256_add_ps(velecsum,velec);
 919
 920             fscal            = felec;
 921
 922             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 923
 924             /* Calculate temporary vectorial force */
 925             tx               = _mm256_mul_ps(fscal,dx22);
 926             ty               = _mm256_mul_ps(fscal,dy22);
 927             tz               = _mm256_mul_ps(fscal,dz22);
 928
 929             /* Update vectorial force */
 930             fix2             = _mm256_add_ps(fix2,tx);
 931             fiy2             = _mm256_add_ps(fiy2,ty);
 932             fiz2             = _mm256_add_ps(fiz2,tz);
 933
 934             fjx2             = _mm256_add_ps(fjx2,tx);
 935             fjy2             = _mm256_add_ps(fjy2,ty);
 936             fjz2             = _mm256_add_ps(fjz2,tz);
 937
 938             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 939             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 940             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 941             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 942             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 943             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 944             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 945             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 946
 947             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 948                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 949
 950             /* Inner loop uses 243 flops */
 951         }
 952
 953         /* End of innermost loop */
 954
 955         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 956                                                  f+i_coord_offset,fshift+i_shift_offset);
 957
 958         ggid                        = gid[iidx];
 959         /* Update potential energies */
 960         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 961
 962         /* Increment number of inner iterations */
 963         inneriter                  += j_index_end - j_index_start;
 964
 965         /* Outer loop uses 19 flops */
 966     }
 967
 968     /* Increment number of outer iterations */
 969     outeriter        += nri;
 970
 971     /* Update outer/inner flops */
 972
 973     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_VF,outeriter*19 + inneriter*243);
 974 }
 975 /*
 976  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW3W3_F_avx_256_single
 977  * Electrostatics interaction: Coulomb
 978  * VdW interaction:            None
 979  * Geometry:                   Water3-Water3
 980  * Calculate force/pot:        Force
 981  */
 982 void
 983 nb_kernel_ElecCoul_VdwNone_GeomW3W3_F_avx_256_single
 984                     (t_nblist                    * gmx_restrict       nlist,
 985                      rvec                        * gmx_restrict          xx,
 986                      rvec                        * gmx_restrict          ff,
 987                      t_forcerec                  * gmx_restrict          fr,
 988                      t_mdatoms                   * gmx_restrict     mdatoms,
 989                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 990                      t_nrnb                      * gmx_restrict        nrnb)
 991 {
 992     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 993      * just 0 for non-waters.
 994      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 995      * jnr indices corresponding to data put in the four positions in the SIMD register.
 996      */
 997     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 998     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 999     int              jnrA,jnrB,jnrC,jnrD;
1000     int              jnrE,jnrF,jnrG,jnrH;
1001     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1002     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1003     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1004     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1005     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1006     real             rcutoff_scalar;
1007     real             *shiftvec,*fshift,*x,*f;
1008     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1009     real             scratch[4*DIM];
1010     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1011     real *           vdwioffsetptr0;
1012     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1013     real *           vdwioffsetptr1;
1014     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1015     real *           vdwioffsetptr2;
1016     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1017     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1018     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1019     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1020     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1021     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1022     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1023     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1024     __m256           dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1025     __m256           dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1026     __m256           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1027     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1028     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1029     __m256           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1030     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1031     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1032     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1033     real             *charge;
1034     __m256           dummy_mask,cutoff_mask;
1035     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1036     __m256           one     = _mm256_set1_ps(1.0);
1037     __m256           two     = _mm256_set1_ps(2.0);
1038     x                = xx[0];
1039     f                = ff[0];
1040
1041     nri              = nlist->nri;
1042     iinr             = nlist->iinr;
1043     jindex           = nlist->jindex;
1044     jjnr             = nlist->jjnr;
1045     shiftidx         = nlist->shift;
1046     gid              = nlist->gid;
1047     shiftvec         = fr->shift_vec[0];
1048     fshift           = fr->fshift[0];
1049     facel            = _mm256_set1_ps(fr->epsfac);
1050     charge           = mdatoms->chargeA;
1051
1052     /* Setup water-specific parameters */
1053     inr              = nlist->iinr[0];
1054     iq0              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+0]));
1055     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1056     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1057
1058     jq0              = _mm256_set1_ps(charge[inr+0]);
1059     jq1              = _mm256_set1_ps(charge[inr+1]);
1060     jq2              = _mm256_set1_ps(charge[inr+2]);
1061     qq00             = _mm256_mul_ps(iq0,jq0);
1062     qq01             = _mm256_mul_ps(iq0,jq1);
1063     qq02             = _mm256_mul_ps(iq0,jq2);
1064     qq10             = _mm256_mul_ps(iq1,jq0);
1065     qq11             = _mm256_mul_ps(iq1,jq1);
1066     qq12             = _mm256_mul_ps(iq1,jq2);
1067     qq20             = _mm256_mul_ps(iq2,jq0);
1068     qq21             = _mm256_mul_ps(iq2,jq1);
1069     qq22             = _mm256_mul_ps(iq2,jq2);
1070
1071     /* Avoid stupid compiler warnings */
1072     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1073     j_coord_offsetA = 0;
1074     j_coord_offsetB = 0;
1075     j_coord_offsetC = 0;
1076     j_coord_offsetD = 0;
1077     j_coord_offsetE = 0;
1078     j_coord_offsetF = 0;
1079     j_coord_offsetG = 0;
1080     j_coord_offsetH = 0;
1081
1082     outeriter        = 0;
1083     inneriter        = 0;
1084
1085     for(iidx=0;iidx<4*DIM;iidx++)
1086     {
1087         scratch[iidx] = 0.0;
1088     }
1089
1090     /* Start outer loop over neighborlists */
1091     for(iidx=0; iidx<nri; iidx++)
1092     {
1093         /* Load shift vector for this list */
1094         i_shift_offset   = DIM*shiftidx[iidx];
1095
1096         /* Load limits for loop over neighbors */
1097         j_index_start    = jindex[iidx];
1098         j_index_end      = jindex[iidx+1];
1099
1100         /* Get outer coordinate index */
1101         inr              = iinr[iidx];
1102         i_coord_offset   = DIM*inr;
1103
1104         /* Load i particle coords and add shift vector */
1105         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1106                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1107
1108         fix0             = _mm256_setzero_ps();
1109         fiy0             = _mm256_setzero_ps();
1110         fiz0             = _mm256_setzero_ps();
1111         fix1             = _mm256_setzero_ps();
1112         fiy1             = _mm256_setzero_ps();
1113         fiz1             = _mm256_setzero_ps();
1114         fix2             = _mm256_setzero_ps();
1115         fiy2             = _mm256_setzero_ps();
1116         fiz2             = _mm256_setzero_ps();
1117
1118         /* Start inner kernel loop */
1119         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1120         {
1121
1122             /* Get j neighbor index, and coordinate index */
1123             jnrA             = jjnr[jidx];
1124             jnrB             = jjnr[jidx+1];
1125             jnrC             = jjnr[jidx+2];
1126             jnrD             = jjnr[jidx+3];
1127             jnrE             = jjnr[jidx+4];
1128             jnrF             = jjnr[jidx+5];
1129             jnrG             = jjnr[jidx+6];
1130             jnrH             = jjnr[jidx+7];
1131             j_coord_offsetA  = DIM*jnrA;
1132             j_coord_offsetB  = DIM*jnrB;
1133             j_coord_offsetC  = DIM*jnrC;
1134             j_coord_offsetD  = DIM*jnrD;
1135             j_coord_offsetE  = DIM*jnrE;
1136             j_coord_offsetF  = DIM*jnrF;
1137             j_coord_offsetG  = DIM*jnrG;
1138             j_coord_offsetH  = DIM*jnrH;
1139
1140             /* load j atom coordinates */
1141             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1142                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1143                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1144                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1145                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1146
1147             /* Calculate displacement vector */
1148             dx00             = _mm256_sub_ps(ix0,jx0);
1149             dy00             = _mm256_sub_ps(iy0,jy0);
1150             dz00             = _mm256_sub_ps(iz0,jz0);
1151             dx01             = _mm256_sub_ps(ix0,jx1);
1152             dy01             = _mm256_sub_ps(iy0,jy1);
1153             dz01             = _mm256_sub_ps(iz0,jz1);
1154             dx02             = _mm256_sub_ps(ix0,jx2);
1155             dy02             = _mm256_sub_ps(iy0,jy2);
1156             dz02             = _mm256_sub_ps(iz0,jz2);
1157             dx10             = _mm256_sub_ps(ix1,jx0);
1158             dy10             = _mm256_sub_ps(iy1,jy0);
1159             dz10             = _mm256_sub_ps(iz1,jz0);
1160             dx11             = _mm256_sub_ps(ix1,jx1);
1161             dy11             = _mm256_sub_ps(iy1,jy1);
1162             dz11             = _mm256_sub_ps(iz1,jz1);
1163             dx12             = _mm256_sub_ps(ix1,jx2);
1164             dy12             = _mm256_sub_ps(iy1,jy2);
1165             dz12             = _mm256_sub_ps(iz1,jz2);
1166             dx20             = _mm256_sub_ps(ix2,jx0);
1167             dy20             = _mm256_sub_ps(iy2,jy0);
1168             dz20             = _mm256_sub_ps(iz2,jz0);
1169             dx21             = _mm256_sub_ps(ix2,jx1);
1170             dy21             = _mm256_sub_ps(iy2,jy1);
1171             dz21             = _mm256_sub_ps(iz2,jz1);
1172             dx22             = _mm256_sub_ps(ix2,jx2);
1173             dy22             = _mm256_sub_ps(iy2,jy2);
1174             dz22             = _mm256_sub_ps(iz2,jz2);
1175
1176             /* Calculate squared distance and things based on it */
1177             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1178             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1179             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1180             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1181             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1182             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1183             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1184             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1185             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1186
1187             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1188             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1189             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1190             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1191             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1192             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1193             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1194             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1195             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1196
1197             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1198             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1199             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1200             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1201             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1202             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1203             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1204             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1205             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1206
1207             fjx0             = _mm256_setzero_ps();
1208             fjy0             = _mm256_setzero_ps();
1209             fjz0             = _mm256_setzero_ps();
1210             fjx1             = _mm256_setzero_ps();
1211             fjy1             = _mm256_setzero_ps();
1212             fjz1             = _mm256_setzero_ps();
1213             fjx2             = _mm256_setzero_ps();
1214             fjy2             = _mm256_setzero_ps();
1215             fjz2             = _mm256_setzero_ps();
1216
1217             /**************************
1218              * CALCULATE INTERACTIONS *
1219              **************************/
1220
1221             /* COULOMB ELECTROSTATICS */
1222             velec            = _mm256_mul_ps(qq00,rinv00);
1223             felec            = _mm256_mul_ps(velec,rinvsq00);
1224
1225             fscal            = felec;
1226
1227             /* Calculate temporary vectorial force */
1228             tx               = _mm256_mul_ps(fscal,dx00);
1229             ty               = _mm256_mul_ps(fscal,dy00);
1230             tz               = _mm256_mul_ps(fscal,dz00);
1231
1232             /* Update vectorial force */
1233             fix0             = _mm256_add_ps(fix0,tx);
1234             fiy0             = _mm256_add_ps(fiy0,ty);
1235             fiz0             = _mm256_add_ps(fiz0,tz);
1236
1237             fjx0             = _mm256_add_ps(fjx0,tx);
1238             fjy0             = _mm256_add_ps(fjy0,ty);
1239             fjz0             = _mm256_add_ps(fjz0,tz);
1240
1241             /**************************
1242              * CALCULATE INTERACTIONS *
1243              **************************/
1244
1245             /* COULOMB ELECTROSTATICS */
1246             velec            = _mm256_mul_ps(qq01,rinv01);
1247             felec            = _mm256_mul_ps(velec,rinvsq01);
1248
1249             fscal            = felec;
1250
1251             /* Calculate temporary vectorial force */
1252             tx               = _mm256_mul_ps(fscal,dx01);
1253             ty               = _mm256_mul_ps(fscal,dy01);
1254             tz               = _mm256_mul_ps(fscal,dz01);
1255
1256             /* Update vectorial force */
1257             fix0             = _mm256_add_ps(fix0,tx);
1258             fiy0             = _mm256_add_ps(fiy0,ty);
1259             fiz0             = _mm256_add_ps(fiz0,tz);
1260
1261             fjx1             = _mm256_add_ps(fjx1,tx);
1262             fjy1             = _mm256_add_ps(fjy1,ty);
1263             fjz1             = _mm256_add_ps(fjz1,tz);
1264
1265             /**************************
1266              * CALCULATE INTERACTIONS *
1267              **************************/
1268
1269             /* COULOMB ELECTROSTATICS */
1270             velec            = _mm256_mul_ps(qq02,rinv02);
1271             felec            = _mm256_mul_ps(velec,rinvsq02);
1272
1273             fscal            = felec;
1274
1275             /* Calculate temporary vectorial force */
1276             tx               = _mm256_mul_ps(fscal,dx02);
1277             ty               = _mm256_mul_ps(fscal,dy02);
1278             tz               = _mm256_mul_ps(fscal,dz02);
1279
1280             /* Update vectorial force */
1281             fix0             = _mm256_add_ps(fix0,tx);
1282             fiy0             = _mm256_add_ps(fiy0,ty);
1283             fiz0             = _mm256_add_ps(fiz0,tz);
1284
1285             fjx2             = _mm256_add_ps(fjx2,tx);
1286             fjy2             = _mm256_add_ps(fjy2,ty);
1287             fjz2             = _mm256_add_ps(fjz2,tz);
1288
1289             /**************************
1290              * CALCULATE INTERACTIONS *
1291              **************************/
1292
1293             /* COULOMB ELECTROSTATICS */
1294             velec            = _mm256_mul_ps(qq10,rinv10);
1295             felec            = _mm256_mul_ps(velec,rinvsq10);
1296
1297             fscal            = felec;
1298
1299             /* Calculate temporary vectorial force */
1300             tx               = _mm256_mul_ps(fscal,dx10);
1301             ty               = _mm256_mul_ps(fscal,dy10);
1302             tz               = _mm256_mul_ps(fscal,dz10);
1303
1304             /* Update vectorial force */
1305             fix1             = _mm256_add_ps(fix1,tx);
1306             fiy1             = _mm256_add_ps(fiy1,ty);
1307             fiz1             = _mm256_add_ps(fiz1,tz);
1308
1309             fjx0             = _mm256_add_ps(fjx0,tx);
1310             fjy0             = _mm256_add_ps(fjy0,ty);
1311             fjz0             = _mm256_add_ps(fjz0,tz);
1312
1313             /**************************
1314              * CALCULATE INTERACTIONS *
1315              **************************/
1316
1317             /* COULOMB ELECTROSTATICS */
1318             velec            = _mm256_mul_ps(qq11,rinv11);
1319             felec            = _mm256_mul_ps(velec,rinvsq11);
1320
1321             fscal            = felec;
1322
1323             /* Calculate temporary vectorial force */
1324             tx               = _mm256_mul_ps(fscal,dx11);
1325             ty               = _mm256_mul_ps(fscal,dy11);
1326             tz               = _mm256_mul_ps(fscal,dz11);
1327
1328             /* Update vectorial force */
1329             fix1             = _mm256_add_ps(fix1,tx);
1330             fiy1             = _mm256_add_ps(fiy1,ty);
1331             fiz1             = _mm256_add_ps(fiz1,tz);
1332
1333             fjx1             = _mm256_add_ps(fjx1,tx);
1334             fjy1             = _mm256_add_ps(fjy1,ty);
1335             fjz1             = _mm256_add_ps(fjz1,tz);
1336
1337             /**************************
1338              * CALCULATE INTERACTIONS *
1339              **************************/
1340
1341             /* COULOMB ELECTROSTATICS */
1342             velec            = _mm256_mul_ps(qq12,rinv12);
1343             felec            = _mm256_mul_ps(velec,rinvsq12);
1344
1345             fscal            = felec;
1346
1347             /* Calculate temporary vectorial force */
1348             tx               = _mm256_mul_ps(fscal,dx12);
1349             ty               = _mm256_mul_ps(fscal,dy12);
1350             tz               = _mm256_mul_ps(fscal,dz12);
1351
1352             /* Update vectorial force */
1353             fix1             = _mm256_add_ps(fix1,tx);
1354             fiy1             = _mm256_add_ps(fiy1,ty);
1355             fiz1             = _mm256_add_ps(fiz1,tz);
1356
1357             fjx2             = _mm256_add_ps(fjx2,tx);
1358             fjy2             = _mm256_add_ps(fjy2,ty);
1359             fjz2             = _mm256_add_ps(fjz2,tz);
1360
1361             /**************************
1362              * CALCULATE INTERACTIONS *
1363              **************************/
1364
1365             /* COULOMB ELECTROSTATICS */
1366             velec            = _mm256_mul_ps(qq20,rinv20);
1367             felec            = _mm256_mul_ps(velec,rinvsq20);
1368
1369             fscal            = felec;
1370
1371             /* Calculate temporary vectorial force */
1372             tx               = _mm256_mul_ps(fscal,dx20);
1373             ty               = _mm256_mul_ps(fscal,dy20);
1374             tz               = _mm256_mul_ps(fscal,dz20);
1375
1376             /* Update vectorial force */
1377             fix2             = _mm256_add_ps(fix2,tx);
1378             fiy2             = _mm256_add_ps(fiy2,ty);
1379             fiz2             = _mm256_add_ps(fiz2,tz);
1380
1381             fjx0             = _mm256_add_ps(fjx0,tx);
1382             fjy0             = _mm256_add_ps(fjy0,ty);
1383             fjz0             = _mm256_add_ps(fjz0,tz);
1384
1385             /**************************
1386              * CALCULATE INTERACTIONS *
1387              **************************/
1388
1389             /* COULOMB ELECTROSTATICS */
1390             velec            = _mm256_mul_ps(qq21,rinv21);
1391             felec            = _mm256_mul_ps(velec,rinvsq21);
1392
1393             fscal            = felec;
1394
1395             /* Calculate temporary vectorial force */
1396             tx               = _mm256_mul_ps(fscal,dx21);
1397             ty               = _mm256_mul_ps(fscal,dy21);
1398             tz               = _mm256_mul_ps(fscal,dz21);
1399
1400             /* Update vectorial force */
1401             fix2             = _mm256_add_ps(fix2,tx);
1402             fiy2             = _mm256_add_ps(fiy2,ty);
1403             fiz2             = _mm256_add_ps(fiz2,tz);
1404
1405             fjx1             = _mm256_add_ps(fjx1,tx);
1406             fjy1             = _mm256_add_ps(fjy1,ty);
1407             fjz1             = _mm256_add_ps(fjz1,tz);
1408
1409             /**************************
1410              * CALCULATE INTERACTIONS *
1411              **************************/
1412
1413             /* COULOMB ELECTROSTATICS */
1414             velec            = _mm256_mul_ps(qq22,rinv22);
1415             felec            = _mm256_mul_ps(velec,rinvsq22);
1416
1417             fscal            = felec;
1418
1419             /* Calculate temporary vectorial force */
1420             tx               = _mm256_mul_ps(fscal,dx22);
1421             ty               = _mm256_mul_ps(fscal,dy22);
1422             tz               = _mm256_mul_ps(fscal,dz22);
1423
1424             /* Update vectorial force */
1425             fix2             = _mm256_add_ps(fix2,tx);
1426             fiy2             = _mm256_add_ps(fiy2,ty);
1427             fiz2             = _mm256_add_ps(fiz2,tz);
1428
1429             fjx2             = _mm256_add_ps(fjx2,tx);
1430             fjy2             = _mm256_add_ps(fjy2,ty);
1431             fjz2             = _mm256_add_ps(fjz2,tz);
1432
1433             fjptrA             = f+j_coord_offsetA;
1434             fjptrB             = f+j_coord_offsetB;
1435             fjptrC             = f+j_coord_offsetC;
1436             fjptrD             = f+j_coord_offsetD;
1437             fjptrE             = f+j_coord_offsetE;
1438             fjptrF             = f+j_coord_offsetF;
1439             fjptrG             = f+j_coord_offsetG;
1440             fjptrH             = f+j_coord_offsetH;
1441
1442             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1443                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1444
1445             /* Inner loop uses 234 flops */
1446         }
1447
1448         if(jidx<j_index_end)
1449         {
1450
1451             /* Get j neighbor index, and coordinate index */
1452             jnrlistA         = jjnr[jidx];
1453             jnrlistB         = jjnr[jidx+1];
1454             jnrlistC         = jjnr[jidx+2];
1455             jnrlistD         = jjnr[jidx+3];
1456             jnrlistE         = jjnr[jidx+4];
1457             jnrlistF         = jjnr[jidx+5];
1458             jnrlistG         = jjnr[jidx+6];
1459             jnrlistH         = jjnr[jidx+7];
1460             /* Sign of each element will be negative for non-real atoms.
1461              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1462              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1463              */
1464             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1465                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1466
1467             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1468             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1469             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1470             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1471             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1472             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1473             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1474             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1475             j_coord_offsetA  = DIM*jnrA;
1476             j_coord_offsetB  = DIM*jnrB;
1477             j_coord_offsetC  = DIM*jnrC;
1478             j_coord_offsetD  = DIM*jnrD;
1479             j_coord_offsetE  = DIM*jnrE;
1480             j_coord_offsetF  = DIM*jnrF;
1481             j_coord_offsetG  = DIM*jnrG;
1482             j_coord_offsetH  = DIM*jnrH;
1483
1484             /* load j atom coordinates */
1485             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1486                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1487                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1488                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1489                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1490
1491             /* Calculate displacement vector */
1492             dx00             = _mm256_sub_ps(ix0,jx0);
1493             dy00             = _mm256_sub_ps(iy0,jy0);
1494             dz00             = _mm256_sub_ps(iz0,jz0);
1495             dx01             = _mm256_sub_ps(ix0,jx1);
1496             dy01             = _mm256_sub_ps(iy0,jy1);
1497             dz01             = _mm256_sub_ps(iz0,jz1);
1498             dx02             = _mm256_sub_ps(ix0,jx2);
1499             dy02             = _mm256_sub_ps(iy0,jy2);
1500             dz02             = _mm256_sub_ps(iz0,jz2);
1501             dx10             = _mm256_sub_ps(ix1,jx0);
1502             dy10             = _mm256_sub_ps(iy1,jy0);
1503             dz10             = _mm256_sub_ps(iz1,jz0);
1504             dx11             = _mm256_sub_ps(ix1,jx1);
1505             dy11             = _mm256_sub_ps(iy1,jy1);
1506             dz11             = _mm256_sub_ps(iz1,jz1);
1507             dx12             = _mm256_sub_ps(ix1,jx2);
1508             dy12             = _mm256_sub_ps(iy1,jy2);
1509             dz12             = _mm256_sub_ps(iz1,jz2);
1510             dx20             = _mm256_sub_ps(ix2,jx0);
1511             dy20             = _mm256_sub_ps(iy2,jy0);
1512             dz20             = _mm256_sub_ps(iz2,jz0);
1513             dx21             = _mm256_sub_ps(ix2,jx1);
1514             dy21             = _mm256_sub_ps(iy2,jy1);
1515             dz21             = _mm256_sub_ps(iz2,jz1);
1516             dx22             = _mm256_sub_ps(ix2,jx2);
1517             dy22             = _mm256_sub_ps(iy2,jy2);
1518             dz22             = _mm256_sub_ps(iz2,jz2);
1519
1520             /* Calculate squared distance and things based on it */
1521             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1522             rsq01            = gmx_mm256_calc_rsq_ps(dx01,dy01,dz01);
1523             rsq02            = gmx_mm256_calc_rsq_ps(dx02,dy02,dz02);
1524             rsq10            = gmx_mm256_calc_rsq_ps(dx10,dy10,dz10);
1525             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1526             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1527             rsq20            = gmx_mm256_calc_rsq_ps(dx20,dy20,dz20);
1528             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1529             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1530
1531             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1532             rinv01           = gmx_mm256_invsqrt_ps(rsq01);
1533             rinv02           = gmx_mm256_invsqrt_ps(rsq02);
1534             rinv10           = gmx_mm256_invsqrt_ps(rsq10);
1535             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1536             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1537             rinv20           = gmx_mm256_invsqrt_ps(rsq20);
1538             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1539             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1540
1541             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1542             rinvsq01         = _mm256_mul_ps(rinv01,rinv01);
1543             rinvsq02         = _mm256_mul_ps(rinv02,rinv02);
1544             rinvsq10         = _mm256_mul_ps(rinv10,rinv10);
1545             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1546             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1547             rinvsq20         = _mm256_mul_ps(rinv20,rinv20);
1548             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1549             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1550
1551             fjx0             = _mm256_setzero_ps();
1552             fjy0             = _mm256_setzero_ps();
1553             fjz0             = _mm256_setzero_ps();
1554             fjx1             = _mm256_setzero_ps();
1555             fjy1             = _mm256_setzero_ps();
1556             fjz1             = _mm256_setzero_ps();
1557             fjx2             = _mm256_setzero_ps();
1558             fjy2             = _mm256_setzero_ps();
1559             fjz2             = _mm256_setzero_ps();
1560
1561             /**************************
1562              * CALCULATE INTERACTIONS *
1563              **************************/
1564
1565             /* COULOMB ELECTROSTATICS */
1566             velec            = _mm256_mul_ps(qq00,rinv00);
1567             felec            = _mm256_mul_ps(velec,rinvsq00);
1568
1569             fscal            = felec;
1570
1571             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1572
1573             /* Calculate temporary vectorial force */
1574             tx               = _mm256_mul_ps(fscal,dx00);
1575             ty               = _mm256_mul_ps(fscal,dy00);
1576             tz               = _mm256_mul_ps(fscal,dz00);
1577
1578             /* Update vectorial force */
1579             fix0             = _mm256_add_ps(fix0,tx);
1580             fiy0             = _mm256_add_ps(fiy0,ty);
1581             fiz0             = _mm256_add_ps(fiz0,tz);
1582
1583             fjx0             = _mm256_add_ps(fjx0,tx);
1584             fjy0             = _mm256_add_ps(fjy0,ty);
1585             fjz0             = _mm256_add_ps(fjz0,tz);
1586
1587             /**************************
1588              * CALCULATE INTERACTIONS *
1589              **************************/
1590
1591             /* COULOMB ELECTROSTATICS */
1592             velec            = _mm256_mul_ps(qq01,rinv01);
1593             felec            = _mm256_mul_ps(velec,rinvsq01);
1594
1595             fscal            = felec;
1596
1597             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1598
1599             /* Calculate temporary vectorial force */
1600             tx               = _mm256_mul_ps(fscal,dx01);
1601             ty               = _mm256_mul_ps(fscal,dy01);
1602             tz               = _mm256_mul_ps(fscal,dz01);
1603
1604             /* Update vectorial force */
1605             fix0             = _mm256_add_ps(fix0,tx);
1606             fiy0             = _mm256_add_ps(fiy0,ty);
1607             fiz0             = _mm256_add_ps(fiz0,tz);
1608
1609             fjx1             = _mm256_add_ps(fjx1,tx);
1610             fjy1             = _mm256_add_ps(fjy1,ty);
1611             fjz1             = _mm256_add_ps(fjz1,tz);
1612
1613             /**************************
1614              * CALCULATE INTERACTIONS *
1615              **************************/
1616
1617             /* COULOMB ELECTROSTATICS */
1618             velec            = _mm256_mul_ps(qq02,rinv02);
1619             felec            = _mm256_mul_ps(velec,rinvsq02);
1620
1621             fscal            = felec;
1622
1623             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1624
1625             /* Calculate temporary vectorial force */
1626             tx               = _mm256_mul_ps(fscal,dx02);
1627             ty               = _mm256_mul_ps(fscal,dy02);
1628             tz               = _mm256_mul_ps(fscal,dz02);
1629
1630             /* Update vectorial force */
1631             fix0             = _mm256_add_ps(fix0,tx);
1632             fiy0             = _mm256_add_ps(fiy0,ty);
1633             fiz0             = _mm256_add_ps(fiz0,tz);
1634
1635             fjx2             = _mm256_add_ps(fjx2,tx);
1636             fjy2             = _mm256_add_ps(fjy2,ty);
1637             fjz2             = _mm256_add_ps(fjz2,tz);
1638
1639             /**************************
1640              * CALCULATE INTERACTIONS *
1641              **************************/
1642
1643             /* COULOMB ELECTROSTATICS */
1644             velec            = _mm256_mul_ps(qq10,rinv10);
1645             felec            = _mm256_mul_ps(velec,rinvsq10);
1646
1647             fscal            = felec;
1648
1649             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1650
1651             /* Calculate temporary vectorial force */
1652             tx               = _mm256_mul_ps(fscal,dx10);
1653             ty               = _mm256_mul_ps(fscal,dy10);
1654             tz               = _mm256_mul_ps(fscal,dz10);
1655
1656             /* Update vectorial force */
1657             fix1             = _mm256_add_ps(fix1,tx);
1658             fiy1             = _mm256_add_ps(fiy1,ty);
1659             fiz1             = _mm256_add_ps(fiz1,tz);
1660
1661             fjx0             = _mm256_add_ps(fjx0,tx);
1662             fjy0             = _mm256_add_ps(fjy0,ty);
1663             fjz0             = _mm256_add_ps(fjz0,tz);
1664
1665             /**************************
1666              * CALCULATE INTERACTIONS *
1667              **************************/
1668
1669             /* COULOMB ELECTROSTATICS */
1670             velec            = _mm256_mul_ps(qq11,rinv11);
1671             felec            = _mm256_mul_ps(velec,rinvsq11);
1672
1673             fscal            = felec;
1674
1675             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1676
1677             /* Calculate temporary vectorial force */
1678             tx               = _mm256_mul_ps(fscal,dx11);
1679             ty               = _mm256_mul_ps(fscal,dy11);
1680             tz               = _mm256_mul_ps(fscal,dz11);
1681
1682             /* Update vectorial force */
1683             fix1             = _mm256_add_ps(fix1,tx);
1684             fiy1             = _mm256_add_ps(fiy1,ty);
1685             fiz1             = _mm256_add_ps(fiz1,tz);
1686
1687             fjx1             = _mm256_add_ps(fjx1,tx);
1688             fjy1             = _mm256_add_ps(fjy1,ty);
1689             fjz1             = _mm256_add_ps(fjz1,tz);
1690
1691             /**************************
1692              * CALCULATE INTERACTIONS *
1693              **************************/
1694
1695             /* COULOMB ELECTROSTATICS */
1696             velec            = _mm256_mul_ps(qq12,rinv12);
1697             felec            = _mm256_mul_ps(velec,rinvsq12);
1698
1699             fscal            = felec;
1700
1701             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1702
1703             /* Calculate temporary vectorial force */
1704             tx               = _mm256_mul_ps(fscal,dx12);
1705             ty               = _mm256_mul_ps(fscal,dy12);
1706             tz               = _mm256_mul_ps(fscal,dz12);
1707
1708             /* Update vectorial force */
1709             fix1             = _mm256_add_ps(fix1,tx);
1710             fiy1             = _mm256_add_ps(fiy1,ty);
1711             fiz1             = _mm256_add_ps(fiz1,tz);
1712
1713             fjx2             = _mm256_add_ps(fjx2,tx);
1714             fjy2             = _mm256_add_ps(fjy2,ty);
1715             fjz2             = _mm256_add_ps(fjz2,tz);
1716
1717             /**************************
1718              * CALCULATE INTERACTIONS *
1719              **************************/
1720
1721             /* COULOMB ELECTROSTATICS */
1722             velec            = _mm256_mul_ps(qq20,rinv20);
1723             felec            = _mm256_mul_ps(velec,rinvsq20);
1724
1725             fscal            = felec;
1726
1727             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1728
1729             /* Calculate temporary vectorial force */
1730             tx               = _mm256_mul_ps(fscal,dx20);
1731             ty               = _mm256_mul_ps(fscal,dy20);
1732             tz               = _mm256_mul_ps(fscal,dz20);
1733
1734             /* Update vectorial force */
1735             fix2             = _mm256_add_ps(fix2,tx);
1736             fiy2             = _mm256_add_ps(fiy2,ty);
1737             fiz2             = _mm256_add_ps(fiz2,tz);
1738
1739             fjx0             = _mm256_add_ps(fjx0,tx);
1740             fjy0             = _mm256_add_ps(fjy0,ty);
1741             fjz0             = _mm256_add_ps(fjz0,tz);
1742
1743             /**************************
1744              * CALCULATE INTERACTIONS *
1745              **************************/
1746
1747             /* COULOMB ELECTROSTATICS */
1748             velec            = _mm256_mul_ps(qq21,rinv21);
1749             felec            = _mm256_mul_ps(velec,rinvsq21);
1750
1751             fscal            = felec;
1752
1753             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1754
1755             /* Calculate temporary vectorial force */
1756             tx               = _mm256_mul_ps(fscal,dx21);
1757             ty               = _mm256_mul_ps(fscal,dy21);
1758             tz               = _mm256_mul_ps(fscal,dz21);
1759
1760             /* Update vectorial force */
1761             fix2             = _mm256_add_ps(fix2,tx);
1762             fiy2             = _mm256_add_ps(fiy2,ty);
1763             fiz2             = _mm256_add_ps(fiz2,tz);
1764
1765             fjx1             = _mm256_add_ps(fjx1,tx);
1766             fjy1             = _mm256_add_ps(fjy1,ty);
1767             fjz1             = _mm256_add_ps(fjz1,tz);
1768
1769             /**************************
1770              * CALCULATE INTERACTIONS *
1771              **************************/
1772
1773             /* COULOMB ELECTROSTATICS */
1774             velec            = _mm256_mul_ps(qq22,rinv22);
1775             felec            = _mm256_mul_ps(velec,rinvsq22);
1776
1777             fscal            = felec;
1778
1779             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1780
1781             /* Calculate temporary vectorial force */
1782             tx               = _mm256_mul_ps(fscal,dx22);
1783             ty               = _mm256_mul_ps(fscal,dy22);
1784             tz               = _mm256_mul_ps(fscal,dz22);
1785
1786             /* Update vectorial force */
1787             fix2             = _mm256_add_ps(fix2,tx);
1788             fiy2             = _mm256_add_ps(fiy2,ty);
1789             fiz2             = _mm256_add_ps(fiz2,tz);
1790
1791             fjx2             = _mm256_add_ps(fjx2,tx);
1792             fjy2             = _mm256_add_ps(fjy2,ty);
1793             fjz2             = _mm256_add_ps(fjz2,tz);
1794
1795             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1796             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1797             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1798             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1799             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1800             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1801             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1802             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1803
1804             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1805                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1806
1807             /* Inner loop uses 234 flops */
1808         }
1809
1810         /* End of innermost loop */
1811
1812         gmx_mm256_update_iforce_3atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1813                                                  f+i_coord_offset,fshift+i_shift_offset);
1814
1815         /* Increment number of inner iterations */
1816         inneriter                  += j_index_end - j_index_start;
1817
1818         /* Outer loop uses 18 flops */
1819     }
1820
1821     /* Increment number of outer iterations */
1822     outeriter        += nri;
1823
1824     /* Update outer/inner flops */
1825
1826     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3W3_F,outeriter*18 + inneriter*234);
1827 }