src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwNone_GeomW4W4_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_ElecCoul_VdwNone_GeomW4W4_VF_avx_256_single
  51  * Electrostatics interaction: Coulomb
  52  * VdW interaction:            None
  53  * Geometry:                   Water4-Water4
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCoul_VdwNone_GeomW4W4_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 *           vdwioffsetptr1;
  86     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  87     real *           vdwioffsetptr2;
  88     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  89     real *           vdwioffsetptr3;
  90     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  91     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  92     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  93     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  94     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  95     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  96     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  97     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  98     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  99     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 100     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 101     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 102     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 103     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 104     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 105     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 106     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     __m256           dummy_mask,cutoff_mask;
 109     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 110     __m256           one     = _mm256_set1_ps(1.0);
 111     __m256           two     = _mm256_set1_ps(2.0);
 112     x                = xx[0];
 113     f                = ff[0];
 114
 115     nri              = nlist->nri;
 116     iinr             = nlist->iinr;
 117     jindex           = nlist->jindex;
 118     jjnr             = nlist->jjnr;
 119     shiftidx         = nlist->shift;
 120     gid              = nlist->gid;
 121     shiftvec         = fr->shift_vec[0];
 122     fshift           = fr->fshift[0];
 123     facel            = _mm256_set1_ps(fr->ic->epsfac);
 124     charge           = mdatoms->chargeA;
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 129     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 130     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 131
 132     jq1              = _mm256_set1_ps(charge[inr+1]);
 133     jq2              = _mm256_set1_ps(charge[inr+2]);
 134     jq3              = _mm256_set1_ps(charge[inr+3]);
 135     qq11             = _mm256_mul_ps(iq1,jq1);
 136     qq12             = _mm256_mul_ps(iq1,jq2);
 137     qq13             = _mm256_mul_ps(iq1,jq3);
 138     qq21             = _mm256_mul_ps(iq2,jq1);
 139     qq22             = _mm256_mul_ps(iq2,jq2);
 140     qq23             = _mm256_mul_ps(iq2,jq3);
 141     qq31             = _mm256_mul_ps(iq3,jq1);
 142     qq32             = _mm256_mul_ps(iq3,jq2);
 143     qq33             = _mm256_mul_ps(iq3,jq3);
 144
 145     /* Avoid stupid compiler warnings */
 146     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 147     j_coord_offsetA = 0;
 148     j_coord_offsetB = 0;
 149     j_coord_offsetC = 0;
 150     j_coord_offsetD = 0;
 151     j_coord_offsetE = 0;
 152     j_coord_offsetF = 0;
 153     j_coord_offsetG = 0;
 154     j_coord_offsetH = 0;
 155
 156     outeriter        = 0;
 157     inneriter        = 0;
 158
 159     for(iidx=0;iidx<4*DIM;iidx++)
 160     {
 161         scratch[iidx] = 0.0;
 162     }
 163
 164     /* Start outer loop over neighborlists */
 165     for(iidx=0; iidx<nri; iidx++)
 166     {
 167         /* Load shift vector for this list */
 168         i_shift_offset   = DIM*shiftidx[iidx];
 169
 170         /* Load limits for loop over neighbors */
 171         j_index_start    = jindex[iidx];
 172         j_index_end      = jindex[iidx+1];
 173
 174         /* Get outer coordinate index */
 175         inr              = iinr[iidx];
 176         i_coord_offset   = DIM*inr;
 177
 178         /* Load i particle coords and add shift vector */
 179         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 180                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 181
 182         fix1             = _mm256_setzero_ps();
 183         fiy1             = _mm256_setzero_ps();
 184         fiz1             = _mm256_setzero_ps();
 185         fix2             = _mm256_setzero_ps();
 186         fiy2             = _mm256_setzero_ps();
 187         fiz2             = _mm256_setzero_ps();
 188         fix3             = _mm256_setzero_ps();
 189         fiy3             = _mm256_setzero_ps();
 190         fiz3             = _mm256_setzero_ps();
 191
 192         /* Reset potential sums */
 193         velecsum         = _mm256_setzero_ps();
 194
 195         /* Start inner kernel loop */
 196         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 197         {
 198
 199             /* Get j neighbor index, and coordinate index */
 200             jnrA             = jjnr[jidx];
 201             jnrB             = jjnr[jidx+1];
 202             jnrC             = jjnr[jidx+2];
 203             jnrD             = jjnr[jidx+3];
 204             jnrE             = jjnr[jidx+4];
 205             jnrF             = jjnr[jidx+5];
 206             jnrG             = jjnr[jidx+6];
 207             jnrH             = jjnr[jidx+7];
 208             j_coord_offsetA  = DIM*jnrA;
 209             j_coord_offsetB  = DIM*jnrB;
 210             j_coord_offsetC  = DIM*jnrC;
 211             j_coord_offsetD  = DIM*jnrD;
 212             j_coord_offsetE  = DIM*jnrE;
 213             j_coord_offsetF  = DIM*jnrF;
 214             j_coord_offsetG  = DIM*jnrG;
 215             j_coord_offsetH  = DIM*jnrH;
 216
 217             /* load j atom coordinates */
 218             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 219                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 220                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 221                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 222                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 223
 224             /* Calculate displacement vector */
 225             dx11             = _mm256_sub_ps(ix1,jx1);
 226             dy11             = _mm256_sub_ps(iy1,jy1);
 227             dz11             = _mm256_sub_ps(iz1,jz1);
 228             dx12             = _mm256_sub_ps(ix1,jx2);
 229             dy12             = _mm256_sub_ps(iy1,jy2);
 230             dz12             = _mm256_sub_ps(iz1,jz2);
 231             dx13             = _mm256_sub_ps(ix1,jx3);
 232             dy13             = _mm256_sub_ps(iy1,jy3);
 233             dz13             = _mm256_sub_ps(iz1,jz3);
 234             dx21             = _mm256_sub_ps(ix2,jx1);
 235             dy21             = _mm256_sub_ps(iy2,jy1);
 236             dz21             = _mm256_sub_ps(iz2,jz1);
 237             dx22             = _mm256_sub_ps(ix2,jx2);
 238             dy22             = _mm256_sub_ps(iy2,jy2);
 239             dz22             = _mm256_sub_ps(iz2,jz2);
 240             dx23             = _mm256_sub_ps(ix2,jx3);
 241             dy23             = _mm256_sub_ps(iy2,jy3);
 242             dz23             = _mm256_sub_ps(iz2,jz3);
 243             dx31             = _mm256_sub_ps(ix3,jx1);
 244             dy31             = _mm256_sub_ps(iy3,jy1);
 245             dz31             = _mm256_sub_ps(iz3,jz1);
 246             dx32             = _mm256_sub_ps(ix3,jx2);
 247             dy32             = _mm256_sub_ps(iy3,jy2);
 248             dz32             = _mm256_sub_ps(iz3,jz2);
 249             dx33             = _mm256_sub_ps(ix3,jx3);
 250             dy33             = _mm256_sub_ps(iy3,jy3);
 251             dz33             = _mm256_sub_ps(iz3,jz3);
 252
 253             /* Calculate squared distance and things based on it */
 254             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 255             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 256             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 257             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 258             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 259             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 260             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 261             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 262             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 263
 264             rinv11           = avx256_invsqrt_f(rsq11);
 265             rinv12           = avx256_invsqrt_f(rsq12);
 266             rinv13           = avx256_invsqrt_f(rsq13);
 267             rinv21           = avx256_invsqrt_f(rsq21);
 268             rinv22           = avx256_invsqrt_f(rsq22);
 269             rinv23           = avx256_invsqrt_f(rsq23);
 270             rinv31           = avx256_invsqrt_f(rsq31);
 271             rinv32           = avx256_invsqrt_f(rsq32);
 272             rinv33           = avx256_invsqrt_f(rsq33);
 273
 274             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 275             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 276             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 277             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 278             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 279             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 280             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 281             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 282             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 283
 284             fjx1             = _mm256_setzero_ps();
 285             fjy1             = _mm256_setzero_ps();
 286             fjz1             = _mm256_setzero_ps();
 287             fjx2             = _mm256_setzero_ps();
 288             fjy2             = _mm256_setzero_ps();
 289             fjz2             = _mm256_setzero_ps();
 290             fjx3             = _mm256_setzero_ps();
 291             fjy3             = _mm256_setzero_ps();
 292             fjz3             = _mm256_setzero_ps();
 293
 294             /**************************
 295              * CALCULATE INTERACTIONS *
 296              **************************/
 297
 298             /* COULOMB ELECTROSTATICS */
 299             velec            = _mm256_mul_ps(qq11,rinv11);
 300             felec            = _mm256_mul_ps(velec,rinvsq11);
 301
 302             /* Update potential sum for this i atom from the interaction with this j atom. */
 303             velecsum         = _mm256_add_ps(velecsum,velec);
 304
 305             fscal            = felec;
 306
 307             /* Calculate temporary vectorial force */
 308             tx               = _mm256_mul_ps(fscal,dx11);
 309             ty               = _mm256_mul_ps(fscal,dy11);
 310             tz               = _mm256_mul_ps(fscal,dz11);
 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             fjx1             = _mm256_add_ps(fjx1,tx);
 318             fjy1             = _mm256_add_ps(fjy1,ty);
 319             fjz1             = _mm256_add_ps(fjz1,tz);
 320
 321             /**************************
 322              * CALCULATE INTERACTIONS *
 323              **************************/
 324
 325             /* COULOMB ELECTROSTATICS */
 326             velec            = _mm256_mul_ps(qq12,rinv12);
 327             felec            = _mm256_mul_ps(velec,rinvsq12);
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velecsum         = _mm256_add_ps(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             /* Calculate temporary vectorial force */
 335             tx               = _mm256_mul_ps(fscal,dx12);
 336             ty               = _mm256_mul_ps(fscal,dy12);
 337             tz               = _mm256_mul_ps(fscal,dz12);
 338
 339             /* Update vectorial force */
 340             fix1             = _mm256_add_ps(fix1,tx);
 341             fiy1             = _mm256_add_ps(fiy1,ty);
 342             fiz1             = _mm256_add_ps(fiz1,tz);
 343
 344             fjx2             = _mm256_add_ps(fjx2,tx);
 345             fjy2             = _mm256_add_ps(fjy2,ty);
 346             fjz2             = _mm256_add_ps(fjz2,tz);
 347
 348             /**************************
 349              * CALCULATE INTERACTIONS *
 350              **************************/
 351
 352             /* COULOMB ELECTROSTATICS */
 353             velec            = _mm256_mul_ps(qq13,rinv13);
 354             felec            = _mm256_mul_ps(velec,rinvsq13);
 355
 356             /* Update potential sum for this i atom from the interaction with this j atom. */
 357             velecsum         = _mm256_add_ps(velecsum,velec);
 358
 359             fscal            = felec;
 360
 361             /* Calculate temporary vectorial force */
 362             tx               = _mm256_mul_ps(fscal,dx13);
 363             ty               = _mm256_mul_ps(fscal,dy13);
 364             tz               = _mm256_mul_ps(fscal,dz13);
 365
 366             /* Update vectorial force */
 367             fix1             = _mm256_add_ps(fix1,tx);
 368             fiy1             = _mm256_add_ps(fiy1,ty);
 369             fiz1             = _mm256_add_ps(fiz1,tz);
 370
 371             fjx3             = _mm256_add_ps(fjx3,tx);
 372             fjy3             = _mm256_add_ps(fjy3,ty);
 373             fjz3             = _mm256_add_ps(fjz3,tz);
 374
 375             /**************************
 376              * CALCULATE INTERACTIONS *
 377              **************************/
 378
 379             /* COULOMB ELECTROSTATICS */
 380             velec            = _mm256_mul_ps(qq21,rinv21);
 381             felec            = _mm256_mul_ps(velec,rinvsq21);
 382
 383             /* Update potential sum for this i atom from the interaction with this j atom. */
 384             velecsum         = _mm256_add_ps(velecsum,velec);
 385
 386             fscal            = felec;
 387
 388             /* Calculate temporary vectorial force */
 389             tx               = _mm256_mul_ps(fscal,dx21);
 390             ty               = _mm256_mul_ps(fscal,dy21);
 391             tz               = _mm256_mul_ps(fscal,dz21);
 392
 393             /* Update vectorial force */
 394             fix2             = _mm256_add_ps(fix2,tx);
 395             fiy2             = _mm256_add_ps(fiy2,ty);
 396             fiz2             = _mm256_add_ps(fiz2,tz);
 397
 398             fjx1             = _mm256_add_ps(fjx1,tx);
 399             fjy1             = _mm256_add_ps(fjy1,ty);
 400             fjz1             = _mm256_add_ps(fjz1,tz);
 401
 402             /**************************
 403              * CALCULATE INTERACTIONS *
 404              **************************/
 405
 406             /* COULOMB ELECTROSTATICS */
 407             velec            = _mm256_mul_ps(qq22,rinv22);
 408             felec            = _mm256_mul_ps(velec,rinvsq22);
 409
 410             /* Update potential sum for this i atom from the interaction with this j atom. */
 411             velecsum         = _mm256_add_ps(velecsum,velec);
 412
 413             fscal            = felec;
 414
 415             /* Calculate temporary vectorial force */
 416             tx               = _mm256_mul_ps(fscal,dx22);
 417             ty               = _mm256_mul_ps(fscal,dy22);
 418             tz               = _mm256_mul_ps(fscal,dz22);
 419
 420             /* Update vectorial force */
 421             fix2             = _mm256_add_ps(fix2,tx);
 422             fiy2             = _mm256_add_ps(fiy2,ty);
 423             fiz2             = _mm256_add_ps(fiz2,tz);
 424
 425             fjx2             = _mm256_add_ps(fjx2,tx);
 426             fjy2             = _mm256_add_ps(fjy2,ty);
 427             fjz2             = _mm256_add_ps(fjz2,tz);
 428
 429             /**************************
 430              * CALCULATE INTERACTIONS *
 431              **************************/
 432
 433             /* COULOMB ELECTROSTATICS */
 434             velec            = _mm256_mul_ps(qq23,rinv23);
 435             felec            = _mm256_mul_ps(velec,rinvsq23);
 436
 437             /* Update potential sum for this i atom from the interaction with this j atom. */
 438             velecsum         = _mm256_add_ps(velecsum,velec);
 439
 440             fscal            = felec;
 441
 442             /* Calculate temporary vectorial force */
 443             tx               = _mm256_mul_ps(fscal,dx23);
 444             ty               = _mm256_mul_ps(fscal,dy23);
 445             tz               = _mm256_mul_ps(fscal,dz23);
 446
 447             /* Update vectorial force */
 448             fix2             = _mm256_add_ps(fix2,tx);
 449             fiy2             = _mm256_add_ps(fiy2,ty);
 450             fiz2             = _mm256_add_ps(fiz2,tz);
 451
 452             fjx3             = _mm256_add_ps(fjx3,tx);
 453             fjy3             = _mm256_add_ps(fjy3,ty);
 454             fjz3             = _mm256_add_ps(fjz3,tz);
 455
 456             /**************************
 457              * CALCULATE INTERACTIONS *
 458              **************************/
 459
 460             /* COULOMB ELECTROSTATICS */
 461             velec            = _mm256_mul_ps(qq31,rinv31);
 462             felec            = _mm256_mul_ps(velec,rinvsq31);
 463
 464             /* Update potential sum for this i atom from the interaction with this j atom. */
 465             velecsum         = _mm256_add_ps(velecsum,velec);
 466
 467             fscal            = felec;
 468
 469             /* Calculate temporary vectorial force */
 470             tx               = _mm256_mul_ps(fscal,dx31);
 471             ty               = _mm256_mul_ps(fscal,dy31);
 472             tz               = _mm256_mul_ps(fscal,dz31);
 473
 474             /* Update vectorial force */
 475             fix3             = _mm256_add_ps(fix3,tx);
 476             fiy3             = _mm256_add_ps(fiy3,ty);
 477             fiz3             = _mm256_add_ps(fiz3,tz);
 478
 479             fjx1             = _mm256_add_ps(fjx1,tx);
 480             fjy1             = _mm256_add_ps(fjy1,ty);
 481             fjz1             = _mm256_add_ps(fjz1,tz);
 482
 483             /**************************
 484              * CALCULATE INTERACTIONS *
 485              **************************/
 486
 487             /* COULOMB ELECTROSTATICS */
 488             velec            = _mm256_mul_ps(qq32,rinv32);
 489             felec            = _mm256_mul_ps(velec,rinvsq32);
 490
 491             /* Update potential sum for this i atom from the interaction with this j atom. */
 492             velecsum         = _mm256_add_ps(velecsum,velec);
 493
 494             fscal            = felec;
 495
 496             /* Calculate temporary vectorial force */
 497             tx               = _mm256_mul_ps(fscal,dx32);
 498             ty               = _mm256_mul_ps(fscal,dy32);
 499             tz               = _mm256_mul_ps(fscal,dz32);
 500
 501             /* Update vectorial force */
 502             fix3             = _mm256_add_ps(fix3,tx);
 503             fiy3             = _mm256_add_ps(fiy3,ty);
 504             fiz3             = _mm256_add_ps(fiz3,tz);
 505
 506             fjx2             = _mm256_add_ps(fjx2,tx);
 507             fjy2             = _mm256_add_ps(fjy2,ty);
 508             fjz2             = _mm256_add_ps(fjz2,tz);
 509
 510             /**************************
 511              * CALCULATE INTERACTIONS *
 512              **************************/
 513
 514             /* COULOMB ELECTROSTATICS */
 515             velec            = _mm256_mul_ps(qq33,rinv33);
 516             felec            = _mm256_mul_ps(velec,rinvsq33);
 517
 518             /* Update potential sum for this i atom from the interaction with this j atom. */
 519             velecsum         = _mm256_add_ps(velecsum,velec);
 520
 521             fscal            = felec;
 522
 523             /* Calculate temporary vectorial force */
 524             tx               = _mm256_mul_ps(fscal,dx33);
 525             ty               = _mm256_mul_ps(fscal,dy33);
 526             tz               = _mm256_mul_ps(fscal,dz33);
 527
 528             /* Update vectorial force */
 529             fix3             = _mm256_add_ps(fix3,tx);
 530             fiy3             = _mm256_add_ps(fiy3,ty);
 531             fiz3             = _mm256_add_ps(fiz3,tz);
 532
 533             fjx3             = _mm256_add_ps(fjx3,tx);
 534             fjy3             = _mm256_add_ps(fjy3,ty);
 535             fjz3             = _mm256_add_ps(fjz3,tz);
 536
 537             fjptrA             = f+j_coord_offsetA;
 538             fjptrB             = f+j_coord_offsetB;
 539             fjptrC             = f+j_coord_offsetC;
 540             fjptrD             = f+j_coord_offsetD;
 541             fjptrE             = f+j_coord_offsetE;
 542             fjptrF             = f+j_coord_offsetF;
 543             fjptrG             = f+j_coord_offsetG;
 544             fjptrH             = f+j_coord_offsetH;
 545
 546             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 547                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 548                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 549
 550             /* Inner loop uses 243 flops */
 551         }
 552
 553         if(jidx<j_index_end)
 554         {
 555
 556             /* Get j neighbor index, and coordinate index */
 557             jnrlistA         = jjnr[jidx];
 558             jnrlistB         = jjnr[jidx+1];
 559             jnrlistC         = jjnr[jidx+2];
 560             jnrlistD         = jjnr[jidx+3];
 561             jnrlistE         = jjnr[jidx+4];
 562             jnrlistF         = jjnr[jidx+5];
 563             jnrlistG         = jjnr[jidx+6];
 564             jnrlistH         = jjnr[jidx+7];
 565             /* Sign of each element will be negative for non-real atoms.
 566              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 567              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 568              */
 569             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 570                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 571
 572             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 573             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 574             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 575             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 576             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 577             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 578             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 579             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 580             j_coord_offsetA  = DIM*jnrA;
 581             j_coord_offsetB  = DIM*jnrB;
 582             j_coord_offsetC  = DIM*jnrC;
 583             j_coord_offsetD  = DIM*jnrD;
 584             j_coord_offsetE  = DIM*jnrE;
 585             j_coord_offsetF  = DIM*jnrF;
 586             j_coord_offsetG  = DIM*jnrG;
 587             j_coord_offsetH  = DIM*jnrH;
 588
 589             /* load j atom coordinates */
 590             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 591                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 592                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 593                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 594                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 595
 596             /* Calculate displacement vector */
 597             dx11             = _mm256_sub_ps(ix1,jx1);
 598             dy11             = _mm256_sub_ps(iy1,jy1);
 599             dz11             = _mm256_sub_ps(iz1,jz1);
 600             dx12             = _mm256_sub_ps(ix1,jx2);
 601             dy12             = _mm256_sub_ps(iy1,jy2);
 602             dz12             = _mm256_sub_ps(iz1,jz2);
 603             dx13             = _mm256_sub_ps(ix1,jx3);
 604             dy13             = _mm256_sub_ps(iy1,jy3);
 605             dz13             = _mm256_sub_ps(iz1,jz3);
 606             dx21             = _mm256_sub_ps(ix2,jx1);
 607             dy21             = _mm256_sub_ps(iy2,jy1);
 608             dz21             = _mm256_sub_ps(iz2,jz1);
 609             dx22             = _mm256_sub_ps(ix2,jx2);
 610             dy22             = _mm256_sub_ps(iy2,jy2);
 611             dz22             = _mm256_sub_ps(iz2,jz2);
 612             dx23             = _mm256_sub_ps(ix2,jx3);
 613             dy23             = _mm256_sub_ps(iy2,jy3);
 614             dz23             = _mm256_sub_ps(iz2,jz3);
 615             dx31             = _mm256_sub_ps(ix3,jx1);
 616             dy31             = _mm256_sub_ps(iy3,jy1);
 617             dz31             = _mm256_sub_ps(iz3,jz1);
 618             dx32             = _mm256_sub_ps(ix3,jx2);
 619             dy32             = _mm256_sub_ps(iy3,jy2);
 620             dz32             = _mm256_sub_ps(iz3,jz2);
 621             dx33             = _mm256_sub_ps(ix3,jx3);
 622             dy33             = _mm256_sub_ps(iy3,jy3);
 623             dz33             = _mm256_sub_ps(iz3,jz3);
 624
 625             /* Calculate squared distance and things based on it */
 626             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 627             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 628             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 629             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 630             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 631             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 632             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 633             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 634             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 635
 636             rinv11           = avx256_invsqrt_f(rsq11);
 637             rinv12           = avx256_invsqrt_f(rsq12);
 638             rinv13           = avx256_invsqrt_f(rsq13);
 639             rinv21           = avx256_invsqrt_f(rsq21);
 640             rinv22           = avx256_invsqrt_f(rsq22);
 641             rinv23           = avx256_invsqrt_f(rsq23);
 642             rinv31           = avx256_invsqrt_f(rsq31);
 643             rinv32           = avx256_invsqrt_f(rsq32);
 644             rinv33           = avx256_invsqrt_f(rsq33);
 645
 646             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 647             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 648             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 649             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 650             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 651             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 652             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 653             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 654             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 655
 656             fjx1             = _mm256_setzero_ps();
 657             fjy1             = _mm256_setzero_ps();
 658             fjz1             = _mm256_setzero_ps();
 659             fjx2             = _mm256_setzero_ps();
 660             fjy2             = _mm256_setzero_ps();
 661             fjz2             = _mm256_setzero_ps();
 662             fjx3             = _mm256_setzero_ps();
 663             fjy3             = _mm256_setzero_ps();
 664             fjz3             = _mm256_setzero_ps();
 665
 666             /**************************
 667              * CALCULATE INTERACTIONS *
 668              **************************/
 669
 670             /* COULOMB ELECTROSTATICS */
 671             velec            = _mm256_mul_ps(qq11,rinv11);
 672             felec            = _mm256_mul_ps(velec,rinvsq11);
 673
 674             /* Update potential sum for this i atom from the interaction with this j atom. */
 675             velec            = _mm256_andnot_ps(dummy_mask,velec);
 676             velecsum         = _mm256_add_ps(velecsum,velec);
 677
 678             fscal            = felec;
 679
 680             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 681
 682             /* Calculate temporary vectorial force */
 683             tx               = _mm256_mul_ps(fscal,dx11);
 684             ty               = _mm256_mul_ps(fscal,dy11);
 685             tz               = _mm256_mul_ps(fscal,dz11);
 686
 687             /* Update vectorial force */
 688             fix1             = _mm256_add_ps(fix1,tx);
 689             fiy1             = _mm256_add_ps(fiy1,ty);
 690             fiz1             = _mm256_add_ps(fiz1,tz);
 691
 692             fjx1             = _mm256_add_ps(fjx1,tx);
 693             fjy1             = _mm256_add_ps(fjy1,ty);
 694             fjz1             = _mm256_add_ps(fjz1,tz);
 695
 696             /**************************
 697              * CALCULATE INTERACTIONS *
 698              **************************/
 699
 700             /* COULOMB ELECTROSTATICS */
 701             velec            = _mm256_mul_ps(qq12,rinv12);
 702             felec            = _mm256_mul_ps(velec,rinvsq12);
 703
 704             /* Update potential sum for this i atom from the interaction with this j atom. */
 705             velec            = _mm256_andnot_ps(dummy_mask,velec);
 706             velecsum         = _mm256_add_ps(velecsum,velec);
 707
 708             fscal            = felec;
 709
 710             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 711
 712             /* Calculate temporary vectorial force */
 713             tx               = _mm256_mul_ps(fscal,dx12);
 714             ty               = _mm256_mul_ps(fscal,dy12);
 715             tz               = _mm256_mul_ps(fscal,dz12);
 716
 717             /* Update vectorial force */
 718             fix1             = _mm256_add_ps(fix1,tx);
 719             fiy1             = _mm256_add_ps(fiy1,ty);
 720             fiz1             = _mm256_add_ps(fiz1,tz);
 721
 722             fjx2             = _mm256_add_ps(fjx2,tx);
 723             fjy2             = _mm256_add_ps(fjy2,ty);
 724             fjz2             = _mm256_add_ps(fjz2,tz);
 725
 726             /**************************
 727              * CALCULATE INTERACTIONS *
 728              **************************/
 729
 730             /* COULOMB ELECTROSTATICS */
 731             velec            = _mm256_mul_ps(qq13,rinv13);
 732             felec            = _mm256_mul_ps(velec,rinvsq13);
 733
 734             /* Update potential sum for this i atom from the interaction with this j atom. */
 735             velec            = _mm256_andnot_ps(dummy_mask,velec);
 736             velecsum         = _mm256_add_ps(velecsum,velec);
 737
 738             fscal            = felec;
 739
 740             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 741
 742             /* Calculate temporary vectorial force */
 743             tx               = _mm256_mul_ps(fscal,dx13);
 744             ty               = _mm256_mul_ps(fscal,dy13);
 745             tz               = _mm256_mul_ps(fscal,dz13);
 746
 747             /* Update vectorial force */
 748             fix1             = _mm256_add_ps(fix1,tx);
 749             fiy1             = _mm256_add_ps(fiy1,ty);
 750             fiz1             = _mm256_add_ps(fiz1,tz);
 751
 752             fjx3             = _mm256_add_ps(fjx3,tx);
 753             fjy3             = _mm256_add_ps(fjy3,ty);
 754             fjz3             = _mm256_add_ps(fjz3,tz);
 755
 756             /**************************
 757              * CALCULATE INTERACTIONS *
 758              **************************/
 759
 760             /* COULOMB ELECTROSTATICS */
 761             velec            = _mm256_mul_ps(qq21,rinv21);
 762             felec            = _mm256_mul_ps(velec,rinvsq21);
 763
 764             /* Update potential sum for this i atom from the interaction with this j atom. */
 765             velec            = _mm256_andnot_ps(dummy_mask,velec);
 766             velecsum         = _mm256_add_ps(velecsum,velec);
 767
 768             fscal            = felec;
 769
 770             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 771
 772             /* Calculate temporary vectorial force */
 773             tx               = _mm256_mul_ps(fscal,dx21);
 774             ty               = _mm256_mul_ps(fscal,dy21);
 775             tz               = _mm256_mul_ps(fscal,dz21);
 776
 777             /* Update vectorial force */
 778             fix2             = _mm256_add_ps(fix2,tx);
 779             fiy2             = _mm256_add_ps(fiy2,ty);
 780             fiz2             = _mm256_add_ps(fiz2,tz);
 781
 782             fjx1             = _mm256_add_ps(fjx1,tx);
 783             fjy1             = _mm256_add_ps(fjy1,ty);
 784             fjz1             = _mm256_add_ps(fjz1,tz);
 785
 786             /**************************
 787              * CALCULATE INTERACTIONS *
 788              **************************/
 789
 790             /* COULOMB ELECTROSTATICS */
 791             velec            = _mm256_mul_ps(qq22,rinv22);
 792             felec            = _mm256_mul_ps(velec,rinvsq22);
 793
 794             /* Update potential sum for this i atom from the interaction with this j atom. */
 795             velec            = _mm256_andnot_ps(dummy_mask,velec);
 796             velecsum         = _mm256_add_ps(velecsum,velec);
 797
 798             fscal            = felec;
 799
 800             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 801
 802             /* Calculate temporary vectorial force */
 803             tx               = _mm256_mul_ps(fscal,dx22);
 804             ty               = _mm256_mul_ps(fscal,dy22);
 805             tz               = _mm256_mul_ps(fscal,dz22);
 806
 807             /* Update vectorial force */
 808             fix2             = _mm256_add_ps(fix2,tx);
 809             fiy2             = _mm256_add_ps(fiy2,ty);
 810             fiz2             = _mm256_add_ps(fiz2,tz);
 811
 812             fjx2             = _mm256_add_ps(fjx2,tx);
 813             fjy2             = _mm256_add_ps(fjy2,ty);
 814             fjz2             = _mm256_add_ps(fjz2,tz);
 815
 816             /**************************
 817              * CALCULATE INTERACTIONS *
 818              **************************/
 819
 820             /* COULOMB ELECTROSTATICS */
 821             velec            = _mm256_mul_ps(qq23,rinv23);
 822             felec            = _mm256_mul_ps(velec,rinvsq23);
 823
 824             /* Update potential sum for this i atom from the interaction with this j atom. */
 825             velec            = _mm256_andnot_ps(dummy_mask,velec);
 826             velecsum         = _mm256_add_ps(velecsum,velec);
 827
 828             fscal            = felec;
 829
 830             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 831
 832             /* Calculate temporary vectorial force */
 833             tx               = _mm256_mul_ps(fscal,dx23);
 834             ty               = _mm256_mul_ps(fscal,dy23);
 835             tz               = _mm256_mul_ps(fscal,dz23);
 836
 837             /* Update vectorial force */
 838             fix2             = _mm256_add_ps(fix2,tx);
 839             fiy2             = _mm256_add_ps(fiy2,ty);
 840             fiz2             = _mm256_add_ps(fiz2,tz);
 841
 842             fjx3             = _mm256_add_ps(fjx3,tx);
 843             fjy3             = _mm256_add_ps(fjy3,ty);
 844             fjz3             = _mm256_add_ps(fjz3,tz);
 845
 846             /**************************
 847              * CALCULATE INTERACTIONS *
 848              **************************/
 849
 850             /* COULOMB ELECTROSTATICS */
 851             velec            = _mm256_mul_ps(qq31,rinv31);
 852             felec            = _mm256_mul_ps(velec,rinvsq31);
 853
 854             /* Update potential sum for this i atom from the interaction with this j atom. */
 855             velec            = _mm256_andnot_ps(dummy_mask,velec);
 856             velecsum         = _mm256_add_ps(velecsum,velec);
 857
 858             fscal            = felec;
 859
 860             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 861
 862             /* Calculate temporary vectorial force */
 863             tx               = _mm256_mul_ps(fscal,dx31);
 864             ty               = _mm256_mul_ps(fscal,dy31);
 865             tz               = _mm256_mul_ps(fscal,dz31);
 866
 867             /* Update vectorial force */
 868             fix3             = _mm256_add_ps(fix3,tx);
 869             fiy3             = _mm256_add_ps(fiy3,ty);
 870             fiz3             = _mm256_add_ps(fiz3,tz);
 871
 872             fjx1             = _mm256_add_ps(fjx1,tx);
 873             fjy1             = _mm256_add_ps(fjy1,ty);
 874             fjz1             = _mm256_add_ps(fjz1,tz);
 875
 876             /**************************
 877              * CALCULATE INTERACTIONS *
 878              **************************/
 879
 880             /* COULOMB ELECTROSTATICS */
 881             velec            = _mm256_mul_ps(qq32,rinv32);
 882             felec            = _mm256_mul_ps(velec,rinvsq32);
 883
 884             /* Update potential sum for this i atom from the interaction with this j atom. */
 885             velec            = _mm256_andnot_ps(dummy_mask,velec);
 886             velecsum         = _mm256_add_ps(velecsum,velec);
 887
 888             fscal            = felec;
 889
 890             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 891
 892             /* Calculate temporary vectorial force */
 893             tx               = _mm256_mul_ps(fscal,dx32);
 894             ty               = _mm256_mul_ps(fscal,dy32);
 895             tz               = _mm256_mul_ps(fscal,dz32);
 896
 897             /* Update vectorial force */
 898             fix3             = _mm256_add_ps(fix3,tx);
 899             fiy3             = _mm256_add_ps(fiy3,ty);
 900             fiz3             = _mm256_add_ps(fiz3,tz);
 901
 902             fjx2             = _mm256_add_ps(fjx2,tx);
 903             fjy2             = _mm256_add_ps(fjy2,ty);
 904             fjz2             = _mm256_add_ps(fjz2,tz);
 905
 906             /**************************
 907              * CALCULATE INTERACTIONS *
 908              **************************/
 909
 910             /* COULOMB ELECTROSTATICS */
 911             velec            = _mm256_mul_ps(qq33,rinv33);
 912             felec            = _mm256_mul_ps(velec,rinvsq33);
 913
 914             /* Update potential sum for this i atom from the interaction with this j atom. */
 915             velec            = _mm256_andnot_ps(dummy_mask,velec);
 916             velecsum         = _mm256_add_ps(velecsum,velec);
 917
 918             fscal            = felec;
 919
 920             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 921
 922             /* Calculate temporary vectorial force */
 923             tx               = _mm256_mul_ps(fscal,dx33);
 924             ty               = _mm256_mul_ps(fscal,dy33);
 925             tz               = _mm256_mul_ps(fscal,dz33);
 926
 927             /* Update vectorial force */
 928             fix3             = _mm256_add_ps(fix3,tx);
 929             fiy3             = _mm256_add_ps(fiy3,ty);
 930             fiz3             = _mm256_add_ps(fiz3,tz);
 931
 932             fjx3             = _mm256_add_ps(fjx3,tx);
 933             fjy3             = _mm256_add_ps(fjy3,ty);
 934             fjz3             = _mm256_add_ps(fjz3,tz);
 935
 936             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 937             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 938             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 939             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 940             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 941             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 942             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 943             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 944
 945             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 946                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 947                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 948
 949             /* Inner loop uses 243 flops */
 950         }
 951
 952         /* End of innermost loop */
 953
 954         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 955                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 956
 957         ggid                        = gid[iidx];
 958         /* Update potential energies */
 959         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 960
 961         /* Increment number of inner iterations */
 962         inneriter                  += j_index_end - j_index_start;
 963
 964         /* Outer loop uses 19 flops */
 965     }
 966
 967     /* Increment number of outer iterations */
 968     outeriter        += nri;
 969
 970     /* Update outer/inner flops */
 971
 972     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*243);
 973 }
 974 /*
 975  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 976  * Electrostatics interaction: Coulomb
 977  * VdW interaction:            None
 978  * Geometry:                   Water4-Water4
 979  * Calculate force/pot:        Force
 980  */
 981 void
 982 nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 983                     (t_nblist                    * gmx_restrict       nlist,
 984                      rvec                        * gmx_restrict          xx,
 985                      rvec                        * gmx_restrict          ff,
 986                      struct t_forcerec           * gmx_restrict          fr,
 987                      t_mdatoms                   * gmx_restrict     mdatoms,
 988                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 989                      t_nrnb                      * gmx_restrict        nrnb)
 990 {
 991     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 992      * just 0 for non-waters.
 993      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 994      * jnr indices corresponding to data put in the four positions in the SIMD register.
 995      */
 996     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 997     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 998     int              jnrA,jnrB,jnrC,jnrD;
 999     int              jnrE,jnrF,jnrG,jnrH;
1000     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1001     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1002     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1003     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1004     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1005     real             rcutoff_scalar;
1006     real             *shiftvec,*fshift,*x,*f;
1007     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1008     real             scratch[4*DIM];
1009     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1010     real *           vdwioffsetptr1;
1011     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1012     real *           vdwioffsetptr2;
1013     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1014     real *           vdwioffsetptr3;
1015     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1016     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1017     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1018     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1019     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1020     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1021     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1022     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1023     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1024     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1025     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1026     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1027     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1028     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1029     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1030     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1031     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1032     real             *charge;
1033     __m256           dummy_mask,cutoff_mask;
1034     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1035     __m256           one     = _mm256_set1_ps(1.0);
1036     __m256           two     = _mm256_set1_ps(2.0);
1037     x                = xx[0];
1038     f                = ff[0];
1039
1040     nri              = nlist->nri;
1041     iinr             = nlist->iinr;
1042     jindex           = nlist->jindex;
1043     jjnr             = nlist->jjnr;
1044     shiftidx         = nlist->shift;
1045     gid              = nlist->gid;
1046     shiftvec         = fr->shift_vec[0];
1047     fshift           = fr->fshift[0];
1048     facel            = _mm256_set1_ps(fr->ic->epsfac);
1049     charge           = mdatoms->chargeA;
1050
1051     /* Setup water-specific parameters */
1052     inr              = nlist->iinr[0];
1053     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1054     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1055     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1056
1057     jq1              = _mm256_set1_ps(charge[inr+1]);
1058     jq2              = _mm256_set1_ps(charge[inr+2]);
1059     jq3              = _mm256_set1_ps(charge[inr+3]);
1060     qq11             = _mm256_mul_ps(iq1,jq1);
1061     qq12             = _mm256_mul_ps(iq1,jq2);
1062     qq13             = _mm256_mul_ps(iq1,jq3);
1063     qq21             = _mm256_mul_ps(iq2,jq1);
1064     qq22             = _mm256_mul_ps(iq2,jq2);
1065     qq23             = _mm256_mul_ps(iq2,jq3);
1066     qq31             = _mm256_mul_ps(iq3,jq1);
1067     qq32             = _mm256_mul_ps(iq3,jq2);
1068     qq33             = _mm256_mul_ps(iq3,jq3);
1069
1070     /* Avoid stupid compiler warnings */
1071     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1072     j_coord_offsetA = 0;
1073     j_coord_offsetB = 0;
1074     j_coord_offsetC = 0;
1075     j_coord_offsetD = 0;
1076     j_coord_offsetE = 0;
1077     j_coord_offsetF = 0;
1078     j_coord_offsetG = 0;
1079     j_coord_offsetH = 0;
1080
1081     outeriter        = 0;
1082     inneriter        = 0;
1083
1084     for(iidx=0;iidx<4*DIM;iidx++)
1085     {
1086         scratch[iidx] = 0.0;
1087     }
1088
1089     /* Start outer loop over neighborlists */
1090     for(iidx=0; iidx<nri; iidx++)
1091     {
1092         /* Load shift vector for this list */
1093         i_shift_offset   = DIM*shiftidx[iidx];
1094
1095         /* Load limits for loop over neighbors */
1096         j_index_start    = jindex[iidx];
1097         j_index_end      = jindex[iidx+1];
1098
1099         /* Get outer coordinate index */
1100         inr              = iinr[iidx];
1101         i_coord_offset   = DIM*inr;
1102
1103         /* Load i particle coords and add shift vector */
1104         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1105                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1106
1107         fix1             = _mm256_setzero_ps();
1108         fiy1             = _mm256_setzero_ps();
1109         fiz1             = _mm256_setzero_ps();
1110         fix2             = _mm256_setzero_ps();
1111         fiy2             = _mm256_setzero_ps();
1112         fiz2             = _mm256_setzero_ps();
1113         fix3             = _mm256_setzero_ps();
1114         fiy3             = _mm256_setzero_ps();
1115         fiz3             = _mm256_setzero_ps();
1116
1117         /* Start inner kernel loop */
1118         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1119         {
1120
1121             /* Get j neighbor index, and coordinate index */
1122             jnrA             = jjnr[jidx];
1123             jnrB             = jjnr[jidx+1];
1124             jnrC             = jjnr[jidx+2];
1125             jnrD             = jjnr[jidx+3];
1126             jnrE             = jjnr[jidx+4];
1127             jnrF             = jjnr[jidx+5];
1128             jnrG             = jjnr[jidx+6];
1129             jnrH             = jjnr[jidx+7];
1130             j_coord_offsetA  = DIM*jnrA;
1131             j_coord_offsetB  = DIM*jnrB;
1132             j_coord_offsetC  = DIM*jnrC;
1133             j_coord_offsetD  = DIM*jnrD;
1134             j_coord_offsetE  = DIM*jnrE;
1135             j_coord_offsetF  = DIM*jnrF;
1136             j_coord_offsetG  = DIM*jnrG;
1137             j_coord_offsetH  = DIM*jnrH;
1138
1139             /* load j atom coordinates */
1140             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1141                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1142                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1143                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1144                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1145
1146             /* Calculate displacement vector */
1147             dx11             = _mm256_sub_ps(ix1,jx1);
1148             dy11             = _mm256_sub_ps(iy1,jy1);
1149             dz11             = _mm256_sub_ps(iz1,jz1);
1150             dx12             = _mm256_sub_ps(ix1,jx2);
1151             dy12             = _mm256_sub_ps(iy1,jy2);
1152             dz12             = _mm256_sub_ps(iz1,jz2);
1153             dx13             = _mm256_sub_ps(ix1,jx3);
1154             dy13             = _mm256_sub_ps(iy1,jy3);
1155             dz13             = _mm256_sub_ps(iz1,jz3);
1156             dx21             = _mm256_sub_ps(ix2,jx1);
1157             dy21             = _mm256_sub_ps(iy2,jy1);
1158             dz21             = _mm256_sub_ps(iz2,jz1);
1159             dx22             = _mm256_sub_ps(ix2,jx2);
1160             dy22             = _mm256_sub_ps(iy2,jy2);
1161             dz22             = _mm256_sub_ps(iz2,jz2);
1162             dx23             = _mm256_sub_ps(ix2,jx3);
1163             dy23             = _mm256_sub_ps(iy2,jy3);
1164             dz23             = _mm256_sub_ps(iz2,jz3);
1165             dx31             = _mm256_sub_ps(ix3,jx1);
1166             dy31             = _mm256_sub_ps(iy3,jy1);
1167             dz31             = _mm256_sub_ps(iz3,jz1);
1168             dx32             = _mm256_sub_ps(ix3,jx2);
1169             dy32             = _mm256_sub_ps(iy3,jy2);
1170             dz32             = _mm256_sub_ps(iz3,jz2);
1171             dx33             = _mm256_sub_ps(ix3,jx3);
1172             dy33             = _mm256_sub_ps(iy3,jy3);
1173             dz33             = _mm256_sub_ps(iz3,jz3);
1174
1175             /* Calculate squared distance and things based on it */
1176             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1177             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1178             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1179             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1180             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1181             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1182             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1183             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1184             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1185
1186             rinv11           = avx256_invsqrt_f(rsq11);
1187             rinv12           = avx256_invsqrt_f(rsq12);
1188             rinv13           = avx256_invsqrt_f(rsq13);
1189             rinv21           = avx256_invsqrt_f(rsq21);
1190             rinv22           = avx256_invsqrt_f(rsq22);
1191             rinv23           = avx256_invsqrt_f(rsq23);
1192             rinv31           = avx256_invsqrt_f(rsq31);
1193             rinv32           = avx256_invsqrt_f(rsq32);
1194             rinv33           = avx256_invsqrt_f(rsq33);
1195
1196             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1197             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1198             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1199             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1200             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1201             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1202             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1203             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1204             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1205
1206             fjx1             = _mm256_setzero_ps();
1207             fjy1             = _mm256_setzero_ps();
1208             fjz1             = _mm256_setzero_ps();
1209             fjx2             = _mm256_setzero_ps();
1210             fjy2             = _mm256_setzero_ps();
1211             fjz2             = _mm256_setzero_ps();
1212             fjx3             = _mm256_setzero_ps();
1213             fjy3             = _mm256_setzero_ps();
1214             fjz3             = _mm256_setzero_ps();
1215
1216             /**************************
1217              * CALCULATE INTERACTIONS *
1218              **************************/
1219
1220             /* COULOMB ELECTROSTATICS */
1221             velec            = _mm256_mul_ps(qq11,rinv11);
1222             felec            = _mm256_mul_ps(velec,rinvsq11);
1223
1224             fscal            = felec;
1225
1226             /* Calculate temporary vectorial force */
1227             tx               = _mm256_mul_ps(fscal,dx11);
1228             ty               = _mm256_mul_ps(fscal,dy11);
1229             tz               = _mm256_mul_ps(fscal,dz11);
1230
1231             /* Update vectorial force */
1232             fix1             = _mm256_add_ps(fix1,tx);
1233             fiy1             = _mm256_add_ps(fiy1,ty);
1234             fiz1             = _mm256_add_ps(fiz1,tz);
1235
1236             fjx1             = _mm256_add_ps(fjx1,tx);
1237             fjy1             = _mm256_add_ps(fjy1,ty);
1238             fjz1             = _mm256_add_ps(fjz1,tz);
1239
1240             /**************************
1241              * CALCULATE INTERACTIONS *
1242              **************************/
1243
1244             /* COULOMB ELECTROSTATICS */
1245             velec            = _mm256_mul_ps(qq12,rinv12);
1246             felec            = _mm256_mul_ps(velec,rinvsq12);
1247
1248             fscal            = felec;
1249
1250             /* Calculate temporary vectorial force */
1251             tx               = _mm256_mul_ps(fscal,dx12);
1252             ty               = _mm256_mul_ps(fscal,dy12);
1253             tz               = _mm256_mul_ps(fscal,dz12);
1254
1255             /* Update vectorial force */
1256             fix1             = _mm256_add_ps(fix1,tx);
1257             fiy1             = _mm256_add_ps(fiy1,ty);
1258             fiz1             = _mm256_add_ps(fiz1,tz);
1259
1260             fjx2             = _mm256_add_ps(fjx2,tx);
1261             fjy2             = _mm256_add_ps(fjy2,ty);
1262             fjz2             = _mm256_add_ps(fjz2,tz);
1263
1264             /**************************
1265              * CALCULATE INTERACTIONS *
1266              **************************/
1267
1268             /* COULOMB ELECTROSTATICS */
1269             velec            = _mm256_mul_ps(qq13,rinv13);
1270             felec            = _mm256_mul_ps(velec,rinvsq13);
1271
1272             fscal            = felec;
1273
1274             /* Calculate temporary vectorial force */
1275             tx               = _mm256_mul_ps(fscal,dx13);
1276             ty               = _mm256_mul_ps(fscal,dy13);
1277             tz               = _mm256_mul_ps(fscal,dz13);
1278
1279             /* Update vectorial force */
1280             fix1             = _mm256_add_ps(fix1,tx);
1281             fiy1             = _mm256_add_ps(fiy1,ty);
1282             fiz1             = _mm256_add_ps(fiz1,tz);
1283
1284             fjx3             = _mm256_add_ps(fjx3,tx);
1285             fjy3             = _mm256_add_ps(fjy3,ty);
1286             fjz3             = _mm256_add_ps(fjz3,tz);
1287
1288             /**************************
1289              * CALCULATE INTERACTIONS *
1290              **************************/
1291
1292             /* COULOMB ELECTROSTATICS */
1293             velec            = _mm256_mul_ps(qq21,rinv21);
1294             felec            = _mm256_mul_ps(velec,rinvsq21);
1295
1296             fscal            = felec;
1297
1298             /* Calculate temporary vectorial force */
1299             tx               = _mm256_mul_ps(fscal,dx21);
1300             ty               = _mm256_mul_ps(fscal,dy21);
1301             tz               = _mm256_mul_ps(fscal,dz21);
1302
1303             /* Update vectorial force */
1304             fix2             = _mm256_add_ps(fix2,tx);
1305             fiy2             = _mm256_add_ps(fiy2,ty);
1306             fiz2             = _mm256_add_ps(fiz2,tz);
1307
1308             fjx1             = _mm256_add_ps(fjx1,tx);
1309             fjy1             = _mm256_add_ps(fjy1,ty);
1310             fjz1             = _mm256_add_ps(fjz1,tz);
1311
1312             /**************************
1313              * CALCULATE INTERACTIONS *
1314              **************************/
1315
1316             /* COULOMB ELECTROSTATICS */
1317             velec            = _mm256_mul_ps(qq22,rinv22);
1318             felec            = _mm256_mul_ps(velec,rinvsq22);
1319
1320             fscal            = felec;
1321
1322             /* Calculate temporary vectorial force */
1323             tx               = _mm256_mul_ps(fscal,dx22);
1324             ty               = _mm256_mul_ps(fscal,dy22);
1325             tz               = _mm256_mul_ps(fscal,dz22);
1326
1327             /* Update vectorial force */
1328             fix2             = _mm256_add_ps(fix2,tx);
1329             fiy2             = _mm256_add_ps(fiy2,ty);
1330             fiz2             = _mm256_add_ps(fiz2,tz);
1331
1332             fjx2             = _mm256_add_ps(fjx2,tx);
1333             fjy2             = _mm256_add_ps(fjy2,ty);
1334             fjz2             = _mm256_add_ps(fjz2,tz);
1335
1336             /**************************
1337              * CALCULATE INTERACTIONS *
1338              **************************/
1339
1340             /* COULOMB ELECTROSTATICS */
1341             velec            = _mm256_mul_ps(qq23,rinv23);
1342             felec            = _mm256_mul_ps(velec,rinvsq23);
1343
1344             fscal            = felec;
1345
1346             /* Calculate temporary vectorial force */
1347             tx               = _mm256_mul_ps(fscal,dx23);
1348             ty               = _mm256_mul_ps(fscal,dy23);
1349             tz               = _mm256_mul_ps(fscal,dz23);
1350
1351             /* Update vectorial force */
1352             fix2             = _mm256_add_ps(fix2,tx);
1353             fiy2             = _mm256_add_ps(fiy2,ty);
1354             fiz2             = _mm256_add_ps(fiz2,tz);
1355
1356             fjx3             = _mm256_add_ps(fjx3,tx);
1357             fjy3             = _mm256_add_ps(fjy3,ty);
1358             fjz3             = _mm256_add_ps(fjz3,tz);
1359
1360             /**************************
1361              * CALCULATE INTERACTIONS *
1362              **************************/
1363
1364             /* COULOMB ELECTROSTATICS */
1365             velec            = _mm256_mul_ps(qq31,rinv31);
1366             felec            = _mm256_mul_ps(velec,rinvsq31);
1367
1368             fscal            = felec;
1369
1370             /* Calculate temporary vectorial force */
1371             tx               = _mm256_mul_ps(fscal,dx31);
1372             ty               = _mm256_mul_ps(fscal,dy31);
1373             tz               = _mm256_mul_ps(fscal,dz31);
1374
1375             /* Update vectorial force */
1376             fix3             = _mm256_add_ps(fix3,tx);
1377             fiy3             = _mm256_add_ps(fiy3,ty);
1378             fiz3             = _mm256_add_ps(fiz3,tz);
1379
1380             fjx1             = _mm256_add_ps(fjx1,tx);
1381             fjy1             = _mm256_add_ps(fjy1,ty);
1382             fjz1             = _mm256_add_ps(fjz1,tz);
1383
1384             /**************************
1385              * CALCULATE INTERACTIONS *
1386              **************************/
1387
1388             /* COULOMB ELECTROSTATICS */
1389             velec            = _mm256_mul_ps(qq32,rinv32);
1390             felec            = _mm256_mul_ps(velec,rinvsq32);
1391
1392             fscal            = felec;
1393
1394             /* Calculate temporary vectorial force */
1395             tx               = _mm256_mul_ps(fscal,dx32);
1396             ty               = _mm256_mul_ps(fscal,dy32);
1397             tz               = _mm256_mul_ps(fscal,dz32);
1398
1399             /* Update vectorial force */
1400             fix3             = _mm256_add_ps(fix3,tx);
1401             fiy3             = _mm256_add_ps(fiy3,ty);
1402             fiz3             = _mm256_add_ps(fiz3,tz);
1403
1404             fjx2             = _mm256_add_ps(fjx2,tx);
1405             fjy2             = _mm256_add_ps(fjy2,ty);
1406             fjz2             = _mm256_add_ps(fjz2,tz);
1407
1408             /**************************
1409              * CALCULATE INTERACTIONS *
1410              **************************/
1411
1412             /* COULOMB ELECTROSTATICS */
1413             velec            = _mm256_mul_ps(qq33,rinv33);
1414             felec            = _mm256_mul_ps(velec,rinvsq33);
1415
1416             fscal            = felec;
1417
1418             /* Calculate temporary vectorial force */
1419             tx               = _mm256_mul_ps(fscal,dx33);
1420             ty               = _mm256_mul_ps(fscal,dy33);
1421             tz               = _mm256_mul_ps(fscal,dz33);
1422
1423             /* Update vectorial force */
1424             fix3             = _mm256_add_ps(fix3,tx);
1425             fiy3             = _mm256_add_ps(fiy3,ty);
1426             fiz3             = _mm256_add_ps(fiz3,tz);
1427
1428             fjx3             = _mm256_add_ps(fjx3,tx);
1429             fjy3             = _mm256_add_ps(fjy3,ty);
1430             fjz3             = _mm256_add_ps(fjz3,tz);
1431
1432             fjptrA             = f+j_coord_offsetA;
1433             fjptrB             = f+j_coord_offsetB;
1434             fjptrC             = f+j_coord_offsetC;
1435             fjptrD             = f+j_coord_offsetD;
1436             fjptrE             = f+j_coord_offsetE;
1437             fjptrF             = f+j_coord_offsetF;
1438             fjptrG             = f+j_coord_offsetG;
1439             fjptrH             = f+j_coord_offsetH;
1440
1441             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1442                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1443                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
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+DIM,x+j_coord_offsetB+DIM,
1486                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1487                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1488                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1489                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1490
1491             /* Calculate displacement vector */
1492             dx11             = _mm256_sub_ps(ix1,jx1);
1493             dy11             = _mm256_sub_ps(iy1,jy1);
1494             dz11             = _mm256_sub_ps(iz1,jz1);
1495             dx12             = _mm256_sub_ps(ix1,jx2);
1496             dy12             = _mm256_sub_ps(iy1,jy2);
1497             dz12             = _mm256_sub_ps(iz1,jz2);
1498             dx13             = _mm256_sub_ps(ix1,jx3);
1499             dy13             = _mm256_sub_ps(iy1,jy3);
1500             dz13             = _mm256_sub_ps(iz1,jz3);
1501             dx21             = _mm256_sub_ps(ix2,jx1);
1502             dy21             = _mm256_sub_ps(iy2,jy1);
1503             dz21             = _mm256_sub_ps(iz2,jz1);
1504             dx22             = _mm256_sub_ps(ix2,jx2);
1505             dy22             = _mm256_sub_ps(iy2,jy2);
1506             dz22             = _mm256_sub_ps(iz2,jz2);
1507             dx23             = _mm256_sub_ps(ix2,jx3);
1508             dy23             = _mm256_sub_ps(iy2,jy3);
1509             dz23             = _mm256_sub_ps(iz2,jz3);
1510             dx31             = _mm256_sub_ps(ix3,jx1);
1511             dy31             = _mm256_sub_ps(iy3,jy1);
1512             dz31             = _mm256_sub_ps(iz3,jz1);
1513             dx32             = _mm256_sub_ps(ix3,jx2);
1514             dy32             = _mm256_sub_ps(iy3,jy2);
1515             dz32             = _mm256_sub_ps(iz3,jz2);
1516             dx33             = _mm256_sub_ps(ix3,jx3);
1517             dy33             = _mm256_sub_ps(iy3,jy3);
1518             dz33             = _mm256_sub_ps(iz3,jz3);
1519
1520             /* Calculate squared distance and things based on it */
1521             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1522             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1523             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1524             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1525             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1526             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1527             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1528             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1529             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1530
1531             rinv11           = avx256_invsqrt_f(rsq11);
1532             rinv12           = avx256_invsqrt_f(rsq12);
1533             rinv13           = avx256_invsqrt_f(rsq13);
1534             rinv21           = avx256_invsqrt_f(rsq21);
1535             rinv22           = avx256_invsqrt_f(rsq22);
1536             rinv23           = avx256_invsqrt_f(rsq23);
1537             rinv31           = avx256_invsqrt_f(rsq31);
1538             rinv32           = avx256_invsqrt_f(rsq32);
1539             rinv33           = avx256_invsqrt_f(rsq33);
1540
1541             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1542             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1543             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1544             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1545             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1546             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1547             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1548             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1549             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1550
1551             fjx1             = _mm256_setzero_ps();
1552             fjy1             = _mm256_setzero_ps();
1553             fjz1             = _mm256_setzero_ps();
1554             fjx2             = _mm256_setzero_ps();
1555             fjy2             = _mm256_setzero_ps();
1556             fjz2             = _mm256_setzero_ps();
1557             fjx3             = _mm256_setzero_ps();
1558             fjy3             = _mm256_setzero_ps();
1559             fjz3             = _mm256_setzero_ps();
1560
1561             /**************************
1562              * CALCULATE INTERACTIONS *
1563              **************************/
1564
1565             /* COULOMB ELECTROSTATICS */
1566             velec            = _mm256_mul_ps(qq11,rinv11);
1567             felec            = _mm256_mul_ps(velec,rinvsq11);
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,dx11);
1575             ty               = _mm256_mul_ps(fscal,dy11);
1576             tz               = _mm256_mul_ps(fscal,dz11);
1577
1578             /* Update vectorial force */
1579             fix1             = _mm256_add_ps(fix1,tx);
1580             fiy1             = _mm256_add_ps(fiy1,ty);
1581             fiz1             = _mm256_add_ps(fiz1,tz);
1582
1583             fjx1             = _mm256_add_ps(fjx1,tx);
1584             fjy1             = _mm256_add_ps(fjy1,ty);
1585             fjz1             = _mm256_add_ps(fjz1,tz);
1586
1587             /**************************
1588              * CALCULATE INTERACTIONS *
1589              **************************/
1590
1591             /* COULOMB ELECTROSTATICS */
1592             velec            = _mm256_mul_ps(qq12,rinv12);
1593             felec            = _mm256_mul_ps(velec,rinvsq12);
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,dx12);
1601             ty               = _mm256_mul_ps(fscal,dy12);
1602             tz               = _mm256_mul_ps(fscal,dz12);
1603
1604             /* Update vectorial force */
1605             fix1             = _mm256_add_ps(fix1,tx);
1606             fiy1             = _mm256_add_ps(fiy1,ty);
1607             fiz1             = _mm256_add_ps(fiz1,tz);
1608
1609             fjx2             = _mm256_add_ps(fjx2,tx);
1610             fjy2             = _mm256_add_ps(fjy2,ty);
1611             fjz2             = _mm256_add_ps(fjz2,tz);
1612
1613             /**************************
1614              * CALCULATE INTERACTIONS *
1615              **************************/
1616
1617             /* COULOMB ELECTROSTATICS */
1618             velec            = _mm256_mul_ps(qq13,rinv13);
1619             felec            = _mm256_mul_ps(velec,rinvsq13);
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,dx13);
1627             ty               = _mm256_mul_ps(fscal,dy13);
1628             tz               = _mm256_mul_ps(fscal,dz13);
1629
1630             /* Update vectorial force */
1631             fix1             = _mm256_add_ps(fix1,tx);
1632             fiy1             = _mm256_add_ps(fiy1,ty);
1633             fiz1             = _mm256_add_ps(fiz1,tz);
1634
1635             fjx3             = _mm256_add_ps(fjx3,tx);
1636             fjy3             = _mm256_add_ps(fjy3,ty);
1637             fjz3             = _mm256_add_ps(fjz3,tz);
1638
1639             /**************************
1640              * CALCULATE INTERACTIONS *
1641              **************************/
1642
1643             /* COULOMB ELECTROSTATICS */
1644             velec            = _mm256_mul_ps(qq21,rinv21);
1645             felec            = _mm256_mul_ps(velec,rinvsq21);
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,dx21);
1653             ty               = _mm256_mul_ps(fscal,dy21);
1654             tz               = _mm256_mul_ps(fscal,dz21);
1655
1656             /* Update vectorial force */
1657             fix2             = _mm256_add_ps(fix2,tx);
1658             fiy2             = _mm256_add_ps(fiy2,ty);
1659             fiz2             = _mm256_add_ps(fiz2,tz);
1660
1661             fjx1             = _mm256_add_ps(fjx1,tx);
1662             fjy1             = _mm256_add_ps(fjy1,ty);
1663             fjz1             = _mm256_add_ps(fjz1,tz);
1664
1665             /**************************
1666              * CALCULATE INTERACTIONS *
1667              **************************/
1668
1669             /* COULOMB ELECTROSTATICS */
1670             velec            = _mm256_mul_ps(qq22,rinv22);
1671             felec            = _mm256_mul_ps(velec,rinvsq22);
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,dx22);
1679             ty               = _mm256_mul_ps(fscal,dy22);
1680             tz               = _mm256_mul_ps(fscal,dz22);
1681
1682             /* Update vectorial force */
1683             fix2             = _mm256_add_ps(fix2,tx);
1684             fiy2             = _mm256_add_ps(fiy2,ty);
1685             fiz2             = _mm256_add_ps(fiz2,tz);
1686
1687             fjx2             = _mm256_add_ps(fjx2,tx);
1688             fjy2             = _mm256_add_ps(fjy2,ty);
1689             fjz2             = _mm256_add_ps(fjz2,tz);
1690
1691             /**************************
1692              * CALCULATE INTERACTIONS *
1693              **************************/
1694
1695             /* COULOMB ELECTROSTATICS */
1696             velec            = _mm256_mul_ps(qq23,rinv23);
1697             felec            = _mm256_mul_ps(velec,rinvsq23);
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,dx23);
1705             ty               = _mm256_mul_ps(fscal,dy23);
1706             tz               = _mm256_mul_ps(fscal,dz23);
1707
1708             /* Update vectorial force */
1709             fix2             = _mm256_add_ps(fix2,tx);
1710             fiy2             = _mm256_add_ps(fiy2,ty);
1711             fiz2             = _mm256_add_ps(fiz2,tz);
1712
1713             fjx3             = _mm256_add_ps(fjx3,tx);
1714             fjy3             = _mm256_add_ps(fjy3,ty);
1715             fjz3             = _mm256_add_ps(fjz3,tz);
1716
1717             /**************************
1718              * CALCULATE INTERACTIONS *
1719              **************************/
1720
1721             /* COULOMB ELECTROSTATICS */
1722             velec            = _mm256_mul_ps(qq31,rinv31);
1723             felec            = _mm256_mul_ps(velec,rinvsq31);
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,dx31);
1731             ty               = _mm256_mul_ps(fscal,dy31);
1732             tz               = _mm256_mul_ps(fscal,dz31);
1733
1734             /* Update vectorial force */
1735             fix3             = _mm256_add_ps(fix3,tx);
1736             fiy3             = _mm256_add_ps(fiy3,ty);
1737             fiz3             = _mm256_add_ps(fiz3,tz);
1738
1739             fjx1             = _mm256_add_ps(fjx1,tx);
1740             fjy1             = _mm256_add_ps(fjy1,ty);
1741             fjz1             = _mm256_add_ps(fjz1,tz);
1742
1743             /**************************
1744              * CALCULATE INTERACTIONS *
1745              **************************/
1746
1747             /* COULOMB ELECTROSTATICS */
1748             velec            = _mm256_mul_ps(qq32,rinv32);
1749             felec            = _mm256_mul_ps(velec,rinvsq32);
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,dx32);
1757             ty               = _mm256_mul_ps(fscal,dy32);
1758             tz               = _mm256_mul_ps(fscal,dz32);
1759
1760             /* Update vectorial force */
1761             fix3             = _mm256_add_ps(fix3,tx);
1762             fiy3             = _mm256_add_ps(fiy3,ty);
1763             fiz3             = _mm256_add_ps(fiz3,tz);
1764
1765             fjx2             = _mm256_add_ps(fjx2,tx);
1766             fjy2             = _mm256_add_ps(fjy2,ty);
1767             fjz2             = _mm256_add_ps(fjz2,tz);
1768
1769             /**************************
1770              * CALCULATE INTERACTIONS *
1771              **************************/
1772
1773             /* COULOMB ELECTROSTATICS */
1774             velec            = _mm256_mul_ps(qq33,rinv33);
1775             felec            = _mm256_mul_ps(velec,rinvsq33);
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,dx33);
1783             ty               = _mm256_mul_ps(fscal,dy33);
1784             tz               = _mm256_mul_ps(fscal,dz33);
1785
1786             /* Update vectorial force */
1787             fix3             = _mm256_add_ps(fix3,tx);
1788             fiy3             = _mm256_add_ps(fiy3,ty);
1789             fiz3             = _mm256_add_ps(fiz3,tz);
1790
1791             fjx3             = _mm256_add_ps(fjx3,tx);
1792             fjy3             = _mm256_add_ps(fjy3,ty);
1793             fjz3             = _mm256_add_ps(fjz3,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+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1805                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1806                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1807
1808             /* Inner loop uses 234 flops */
1809         }
1810
1811         /* End of innermost loop */
1812
1813         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1814                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1815
1816         /* Increment number of inner iterations */
1817         inneriter                  += j_index_end - j_index_start;
1818
1819         /* Outer loop uses 18 flops */
1820     }
1821
1822     /* Increment number of outer iterations */
1823     outeriter        += nri;
1824
1825     /* Update outer/inner flops */
1826
1827     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*234);
1828 }