src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwNone_GeomW4W4_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_GeomW4W4_VF_avx_256_single
  54  * Electrostatics interaction: Coulomb
  55  * VdW interaction:            None
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecCoul_VdwNone_GeomW4W4_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 *           vdwioffsetptr1;
  89     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  95     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  96     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  97     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  98     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
  99     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 100     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 101     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 102     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 103     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 104     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 105     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 106     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 107     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 108     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 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     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 132     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 133     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 134
 135     jq1              = _mm256_set1_ps(charge[inr+1]);
 136     jq2              = _mm256_set1_ps(charge[inr+2]);
 137     jq3              = _mm256_set1_ps(charge[inr+3]);
 138     qq11             = _mm256_mul_ps(iq1,jq1);
 139     qq12             = _mm256_mul_ps(iq1,jq2);
 140     qq13             = _mm256_mul_ps(iq1,jq3);
 141     qq21             = _mm256_mul_ps(iq2,jq1);
 142     qq22             = _mm256_mul_ps(iq2,jq2);
 143     qq23             = _mm256_mul_ps(iq2,jq3);
 144     qq31             = _mm256_mul_ps(iq3,jq1);
 145     qq32             = _mm256_mul_ps(iq3,jq2);
 146     qq33             = _mm256_mul_ps(iq3,jq3);
 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+DIM,
 183                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 184
 185         fix1             = _mm256_setzero_ps();
 186         fiy1             = _mm256_setzero_ps();
 187         fiz1             = _mm256_setzero_ps();
 188         fix2             = _mm256_setzero_ps();
 189         fiy2             = _mm256_setzero_ps();
 190         fiz2             = _mm256_setzero_ps();
 191         fix3             = _mm256_setzero_ps();
 192         fiy3             = _mm256_setzero_ps();
 193         fiz3             = _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+DIM,x+j_coord_offsetB+DIM,
 222                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 223                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 224                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 225                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 226
 227             /* Calculate displacement vector */
 228             dx11             = _mm256_sub_ps(ix1,jx1);
 229             dy11             = _mm256_sub_ps(iy1,jy1);
 230             dz11             = _mm256_sub_ps(iz1,jz1);
 231             dx12             = _mm256_sub_ps(ix1,jx2);
 232             dy12             = _mm256_sub_ps(iy1,jy2);
 233             dz12             = _mm256_sub_ps(iz1,jz2);
 234             dx13             = _mm256_sub_ps(ix1,jx3);
 235             dy13             = _mm256_sub_ps(iy1,jy3);
 236             dz13             = _mm256_sub_ps(iz1,jz3);
 237             dx21             = _mm256_sub_ps(ix2,jx1);
 238             dy21             = _mm256_sub_ps(iy2,jy1);
 239             dz21             = _mm256_sub_ps(iz2,jz1);
 240             dx22             = _mm256_sub_ps(ix2,jx2);
 241             dy22             = _mm256_sub_ps(iy2,jy2);
 242             dz22             = _mm256_sub_ps(iz2,jz2);
 243             dx23             = _mm256_sub_ps(ix2,jx3);
 244             dy23             = _mm256_sub_ps(iy2,jy3);
 245             dz23             = _mm256_sub_ps(iz2,jz3);
 246             dx31             = _mm256_sub_ps(ix3,jx1);
 247             dy31             = _mm256_sub_ps(iy3,jy1);
 248             dz31             = _mm256_sub_ps(iz3,jz1);
 249             dx32             = _mm256_sub_ps(ix3,jx2);
 250             dy32             = _mm256_sub_ps(iy3,jy2);
 251             dz32             = _mm256_sub_ps(iz3,jz2);
 252             dx33             = _mm256_sub_ps(ix3,jx3);
 253             dy33             = _mm256_sub_ps(iy3,jy3);
 254             dz33             = _mm256_sub_ps(iz3,jz3);
 255
 256             /* Calculate squared distance and things based on it */
 257             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 258             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 259             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 260             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 261             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 262             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 263             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 264             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 265             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 266
 267             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 268             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 269             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 270             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 271             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 272             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 273             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 274             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 275             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 276
 277             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 278             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 279             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 280             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 281             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 282             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 283             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 284             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 285             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 286
 287             fjx1             = _mm256_setzero_ps();
 288             fjy1             = _mm256_setzero_ps();
 289             fjz1             = _mm256_setzero_ps();
 290             fjx2             = _mm256_setzero_ps();
 291             fjy2             = _mm256_setzero_ps();
 292             fjz2             = _mm256_setzero_ps();
 293             fjx3             = _mm256_setzero_ps();
 294             fjy3             = _mm256_setzero_ps();
 295             fjz3             = _mm256_setzero_ps();
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             /* COULOMB ELECTROSTATICS */
 302             velec            = _mm256_mul_ps(qq11,rinv11);
 303             felec            = _mm256_mul_ps(velec,rinvsq11);
 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,dx11);
 312             ty               = _mm256_mul_ps(fscal,dy11);
 313             tz               = _mm256_mul_ps(fscal,dz11);
 314
 315             /* Update vectorial force */
 316             fix1             = _mm256_add_ps(fix1,tx);
 317             fiy1             = _mm256_add_ps(fiy1,ty);
 318             fiz1             = _mm256_add_ps(fiz1,tz);
 319
 320             fjx1             = _mm256_add_ps(fjx1,tx);
 321             fjy1             = _mm256_add_ps(fjy1,ty);
 322             fjz1             = _mm256_add_ps(fjz1,tz);
 323
 324             /**************************
 325              * CALCULATE INTERACTIONS *
 326              **************************/
 327
 328             /* COULOMB ELECTROSTATICS */
 329             velec            = _mm256_mul_ps(qq12,rinv12);
 330             felec            = _mm256_mul_ps(velec,rinvsq12);
 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,dx12);
 339             ty               = _mm256_mul_ps(fscal,dy12);
 340             tz               = _mm256_mul_ps(fscal,dz12);
 341
 342             /* Update vectorial force */
 343             fix1             = _mm256_add_ps(fix1,tx);
 344             fiy1             = _mm256_add_ps(fiy1,ty);
 345             fiz1             = _mm256_add_ps(fiz1,tz);
 346
 347             fjx2             = _mm256_add_ps(fjx2,tx);
 348             fjy2             = _mm256_add_ps(fjy2,ty);
 349             fjz2             = _mm256_add_ps(fjz2,tz);
 350
 351             /**************************
 352              * CALCULATE INTERACTIONS *
 353              **************************/
 354
 355             /* COULOMB ELECTROSTATICS */
 356             velec            = _mm256_mul_ps(qq13,rinv13);
 357             felec            = _mm256_mul_ps(velec,rinvsq13);
 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,dx13);
 366             ty               = _mm256_mul_ps(fscal,dy13);
 367             tz               = _mm256_mul_ps(fscal,dz13);
 368
 369             /* Update vectorial force */
 370             fix1             = _mm256_add_ps(fix1,tx);
 371             fiy1             = _mm256_add_ps(fiy1,ty);
 372             fiz1             = _mm256_add_ps(fiz1,tz);
 373
 374             fjx3             = _mm256_add_ps(fjx3,tx);
 375             fjy3             = _mm256_add_ps(fjy3,ty);
 376             fjz3             = _mm256_add_ps(fjz3,tz);
 377
 378             /**************************
 379              * CALCULATE INTERACTIONS *
 380              **************************/
 381
 382             /* COULOMB ELECTROSTATICS */
 383             velec            = _mm256_mul_ps(qq21,rinv21);
 384             felec            = _mm256_mul_ps(velec,rinvsq21);
 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,dx21);
 393             ty               = _mm256_mul_ps(fscal,dy21);
 394             tz               = _mm256_mul_ps(fscal,dz21);
 395
 396             /* Update vectorial force */
 397             fix2             = _mm256_add_ps(fix2,tx);
 398             fiy2             = _mm256_add_ps(fiy2,ty);
 399             fiz2             = _mm256_add_ps(fiz2,tz);
 400
 401             fjx1             = _mm256_add_ps(fjx1,tx);
 402             fjy1             = _mm256_add_ps(fjy1,ty);
 403             fjz1             = _mm256_add_ps(fjz1,tz);
 404
 405             /**************************
 406              * CALCULATE INTERACTIONS *
 407              **************************/
 408
 409             /* COULOMB ELECTROSTATICS */
 410             velec            = _mm256_mul_ps(qq22,rinv22);
 411             felec            = _mm256_mul_ps(velec,rinvsq22);
 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,dx22);
 420             ty               = _mm256_mul_ps(fscal,dy22);
 421             tz               = _mm256_mul_ps(fscal,dz22);
 422
 423             /* Update vectorial force */
 424             fix2             = _mm256_add_ps(fix2,tx);
 425             fiy2             = _mm256_add_ps(fiy2,ty);
 426             fiz2             = _mm256_add_ps(fiz2,tz);
 427
 428             fjx2             = _mm256_add_ps(fjx2,tx);
 429             fjy2             = _mm256_add_ps(fjy2,ty);
 430             fjz2             = _mm256_add_ps(fjz2,tz);
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             /* COULOMB ELECTROSTATICS */
 437             velec            = _mm256_mul_ps(qq23,rinv23);
 438             felec            = _mm256_mul_ps(velec,rinvsq23);
 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,dx23);
 447             ty               = _mm256_mul_ps(fscal,dy23);
 448             tz               = _mm256_mul_ps(fscal,dz23);
 449
 450             /* Update vectorial force */
 451             fix2             = _mm256_add_ps(fix2,tx);
 452             fiy2             = _mm256_add_ps(fiy2,ty);
 453             fiz2             = _mm256_add_ps(fiz2,tz);
 454
 455             fjx3             = _mm256_add_ps(fjx3,tx);
 456             fjy3             = _mm256_add_ps(fjy3,ty);
 457             fjz3             = _mm256_add_ps(fjz3,tz);
 458
 459             /**************************
 460              * CALCULATE INTERACTIONS *
 461              **************************/
 462
 463             /* COULOMB ELECTROSTATICS */
 464             velec            = _mm256_mul_ps(qq31,rinv31);
 465             felec            = _mm256_mul_ps(velec,rinvsq31);
 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,dx31);
 474             ty               = _mm256_mul_ps(fscal,dy31);
 475             tz               = _mm256_mul_ps(fscal,dz31);
 476
 477             /* Update vectorial force */
 478             fix3             = _mm256_add_ps(fix3,tx);
 479             fiy3             = _mm256_add_ps(fiy3,ty);
 480             fiz3             = _mm256_add_ps(fiz3,tz);
 481
 482             fjx1             = _mm256_add_ps(fjx1,tx);
 483             fjy1             = _mm256_add_ps(fjy1,ty);
 484             fjz1             = _mm256_add_ps(fjz1,tz);
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             /* COULOMB ELECTROSTATICS */
 491             velec            = _mm256_mul_ps(qq32,rinv32);
 492             felec            = _mm256_mul_ps(velec,rinvsq32);
 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,dx32);
 501             ty               = _mm256_mul_ps(fscal,dy32);
 502             tz               = _mm256_mul_ps(fscal,dz32);
 503
 504             /* Update vectorial force */
 505             fix3             = _mm256_add_ps(fix3,tx);
 506             fiy3             = _mm256_add_ps(fiy3,ty);
 507             fiz3             = _mm256_add_ps(fiz3,tz);
 508
 509             fjx2             = _mm256_add_ps(fjx2,tx);
 510             fjy2             = _mm256_add_ps(fjy2,ty);
 511             fjz2             = _mm256_add_ps(fjz2,tz);
 512
 513             /**************************
 514              * CALCULATE INTERACTIONS *
 515              **************************/
 516
 517             /* COULOMB ELECTROSTATICS */
 518             velec            = _mm256_mul_ps(qq33,rinv33);
 519             felec            = _mm256_mul_ps(velec,rinvsq33);
 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,dx33);
 528             ty               = _mm256_mul_ps(fscal,dy33);
 529             tz               = _mm256_mul_ps(fscal,dz33);
 530
 531             /* Update vectorial force */
 532             fix3             = _mm256_add_ps(fix3,tx);
 533             fiy3             = _mm256_add_ps(fiy3,ty);
 534             fiz3             = _mm256_add_ps(fiz3,tz);
 535
 536             fjx3             = _mm256_add_ps(fjx3,tx);
 537             fjy3             = _mm256_add_ps(fjy3,ty);
 538             fjz3             = _mm256_add_ps(fjz3,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+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 550                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 551                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 552
 553             /* Inner loop uses 243 flops */
 554         }
 555
 556         if(jidx<j_index_end)
 557         {
 558
 559             /* Get j neighbor index, and coordinate index */
 560             jnrlistA         = jjnr[jidx];
 561             jnrlistB         = jjnr[jidx+1];
 562             jnrlistC         = jjnr[jidx+2];
 563             jnrlistD         = jjnr[jidx+3];
 564             jnrlistE         = jjnr[jidx+4];
 565             jnrlistF         = jjnr[jidx+5];
 566             jnrlistG         = jjnr[jidx+6];
 567             jnrlistH         = jjnr[jidx+7];
 568             /* Sign of each element will be negative for non-real atoms.
 569              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 570              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 571              */
 572             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 573                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 574
 575             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 576             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 577             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 578             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 579             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 580             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 581             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 582             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 583             j_coord_offsetA  = DIM*jnrA;
 584             j_coord_offsetB  = DIM*jnrB;
 585             j_coord_offsetC  = DIM*jnrC;
 586             j_coord_offsetD  = DIM*jnrD;
 587             j_coord_offsetE  = DIM*jnrE;
 588             j_coord_offsetF  = DIM*jnrF;
 589             j_coord_offsetG  = DIM*jnrG;
 590             j_coord_offsetH  = DIM*jnrH;
 591
 592             /* load j atom coordinates */
 593             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 594                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
 595                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
 596                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
 597                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 598
 599             /* Calculate displacement vector */
 600             dx11             = _mm256_sub_ps(ix1,jx1);
 601             dy11             = _mm256_sub_ps(iy1,jy1);
 602             dz11             = _mm256_sub_ps(iz1,jz1);
 603             dx12             = _mm256_sub_ps(ix1,jx2);
 604             dy12             = _mm256_sub_ps(iy1,jy2);
 605             dz12             = _mm256_sub_ps(iz1,jz2);
 606             dx13             = _mm256_sub_ps(ix1,jx3);
 607             dy13             = _mm256_sub_ps(iy1,jy3);
 608             dz13             = _mm256_sub_ps(iz1,jz3);
 609             dx21             = _mm256_sub_ps(ix2,jx1);
 610             dy21             = _mm256_sub_ps(iy2,jy1);
 611             dz21             = _mm256_sub_ps(iz2,jz1);
 612             dx22             = _mm256_sub_ps(ix2,jx2);
 613             dy22             = _mm256_sub_ps(iy2,jy2);
 614             dz22             = _mm256_sub_ps(iz2,jz2);
 615             dx23             = _mm256_sub_ps(ix2,jx3);
 616             dy23             = _mm256_sub_ps(iy2,jy3);
 617             dz23             = _mm256_sub_ps(iz2,jz3);
 618             dx31             = _mm256_sub_ps(ix3,jx1);
 619             dy31             = _mm256_sub_ps(iy3,jy1);
 620             dz31             = _mm256_sub_ps(iz3,jz1);
 621             dx32             = _mm256_sub_ps(ix3,jx2);
 622             dy32             = _mm256_sub_ps(iy3,jy2);
 623             dz32             = _mm256_sub_ps(iz3,jz2);
 624             dx33             = _mm256_sub_ps(ix3,jx3);
 625             dy33             = _mm256_sub_ps(iy3,jy3);
 626             dz33             = _mm256_sub_ps(iz3,jz3);
 627
 628             /* Calculate squared distance and things based on it */
 629             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 630             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 631             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 632             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 633             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 634             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 635             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 636             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 637             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 638
 639             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 640             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 641             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 642             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 643             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 644             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 645             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 646             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 647             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 648
 649             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 650             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 651             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 652             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 653             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 654             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 655             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 656             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 657             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 658
 659             fjx1             = _mm256_setzero_ps();
 660             fjy1             = _mm256_setzero_ps();
 661             fjz1             = _mm256_setzero_ps();
 662             fjx2             = _mm256_setzero_ps();
 663             fjy2             = _mm256_setzero_ps();
 664             fjz2             = _mm256_setzero_ps();
 665             fjx3             = _mm256_setzero_ps();
 666             fjy3             = _mm256_setzero_ps();
 667             fjz3             = _mm256_setzero_ps();
 668
 669             /**************************
 670              * CALCULATE INTERACTIONS *
 671              **************************/
 672
 673             /* COULOMB ELECTROSTATICS */
 674             velec            = _mm256_mul_ps(qq11,rinv11);
 675             felec            = _mm256_mul_ps(velec,rinvsq11);
 676
 677             /* Update potential sum for this i atom from the interaction with this j atom. */
 678             velec            = _mm256_andnot_ps(dummy_mask,velec);
 679             velecsum         = _mm256_add_ps(velecsum,velec);
 680
 681             fscal            = felec;
 682
 683             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 684
 685             /* Calculate temporary vectorial force */
 686             tx               = _mm256_mul_ps(fscal,dx11);
 687             ty               = _mm256_mul_ps(fscal,dy11);
 688             tz               = _mm256_mul_ps(fscal,dz11);
 689
 690             /* Update vectorial force */
 691             fix1             = _mm256_add_ps(fix1,tx);
 692             fiy1             = _mm256_add_ps(fiy1,ty);
 693             fiz1             = _mm256_add_ps(fiz1,tz);
 694
 695             fjx1             = _mm256_add_ps(fjx1,tx);
 696             fjy1             = _mm256_add_ps(fjy1,ty);
 697             fjz1             = _mm256_add_ps(fjz1,tz);
 698
 699             /**************************
 700              * CALCULATE INTERACTIONS *
 701              **************************/
 702
 703             /* COULOMB ELECTROSTATICS */
 704             velec            = _mm256_mul_ps(qq12,rinv12);
 705             felec            = _mm256_mul_ps(velec,rinvsq12);
 706
 707             /* Update potential sum for this i atom from the interaction with this j atom. */
 708             velec            = _mm256_andnot_ps(dummy_mask,velec);
 709             velecsum         = _mm256_add_ps(velecsum,velec);
 710
 711             fscal            = felec;
 712
 713             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 714
 715             /* Calculate temporary vectorial force */
 716             tx               = _mm256_mul_ps(fscal,dx12);
 717             ty               = _mm256_mul_ps(fscal,dy12);
 718             tz               = _mm256_mul_ps(fscal,dz12);
 719
 720             /* Update vectorial force */
 721             fix1             = _mm256_add_ps(fix1,tx);
 722             fiy1             = _mm256_add_ps(fiy1,ty);
 723             fiz1             = _mm256_add_ps(fiz1,tz);
 724
 725             fjx2             = _mm256_add_ps(fjx2,tx);
 726             fjy2             = _mm256_add_ps(fjy2,ty);
 727             fjz2             = _mm256_add_ps(fjz2,tz);
 728
 729             /**************************
 730              * CALCULATE INTERACTIONS *
 731              **************************/
 732
 733             /* COULOMB ELECTROSTATICS */
 734             velec            = _mm256_mul_ps(qq13,rinv13);
 735             felec            = _mm256_mul_ps(velec,rinvsq13);
 736
 737             /* Update potential sum for this i atom from the interaction with this j atom. */
 738             velec            = _mm256_andnot_ps(dummy_mask,velec);
 739             velecsum         = _mm256_add_ps(velecsum,velec);
 740
 741             fscal            = felec;
 742
 743             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 744
 745             /* Calculate temporary vectorial force */
 746             tx               = _mm256_mul_ps(fscal,dx13);
 747             ty               = _mm256_mul_ps(fscal,dy13);
 748             tz               = _mm256_mul_ps(fscal,dz13);
 749
 750             /* Update vectorial force */
 751             fix1             = _mm256_add_ps(fix1,tx);
 752             fiy1             = _mm256_add_ps(fiy1,ty);
 753             fiz1             = _mm256_add_ps(fiz1,tz);
 754
 755             fjx3             = _mm256_add_ps(fjx3,tx);
 756             fjy3             = _mm256_add_ps(fjy3,ty);
 757             fjz3             = _mm256_add_ps(fjz3,tz);
 758
 759             /**************************
 760              * CALCULATE INTERACTIONS *
 761              **************************/
 762
 763             /* COULOMB ELECTROSTATICS */
 764             velec            = _mm256_mul_ps(qq21,rinv21);
 765             felec            = _mm256_mul_ps(velec,rinvsq21);
 766
 767             /* Update potential sum for this i atom from the interaction with this j atom. */
 768             velec            = _mm256_andnot_ps(dummy_mask,velec);
 769             velecsum         = _mm256_add_ps(velecsum,velec);
 770
 771             fscal            = felec;
 772
 773             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 774
 775             /* Calculate temporary vectorial force */
 776             tx               = _mm256_mul_ps(fscal,dx21);
 777             ty               = _mm256_mul_ps(fscal,dy21);
 778             tz               = _mm256_mul_ps(fscal,dz21);
 779
 780             /* Update vectorial force */
 781             fix2             = _mm256_add_ps(fix2,tx);
 782             fiy2             = _mm256_add_ps(fiy2,ty);
 783             fiz2             = _mm256_add_ps(fiz2,tz);
 784
 785             fjx1             = _mm256_add_ps(fjx1,tx);
 786             fjy1             = _mm256_add_ps(fjy1,ty);
 787             fjz1             = _mm256_add_ps(fjz1,tz);
 788
 789             /**************************
 790              * CALCULATE INTERACTIONS *
 791              **************************/
 792
 793             /* COULOMB ELECTROSTATICS */
 794             velec            = _mm256_mul_ps(qq22,rinv22);
 795             felec            = _mm256_mul_ps(velec,rinvsq22);
 796
 797             /* Update potential sum for this i atom from the interaction with this j atom. */
 798             velec            = _mm256_andnot_ps(dummy_mask,velec);
 799             velecsum         = _mm256_add_ps(velecsum,velec);
 800
 801             fscal            = felec;
 802
 803             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 804
 805             /* Calculate temporary vectorial force */
 806             tx               = _mm256_mul_ps(fscal,dx22);
 807             ty               = _mm256_mul_ps(fscal,dy22);
 808             tz               = _mm256_mul_ps(fscal,dz22);
 809
 810             /* Update vectorial force */
 811             fix2             = _mm256_add_ps(fix2,tx);
 812             fiy2             = _mm256_add_ps(fiy2,ty);
 813             fiz2             = _mm256_add_ps(fiz2,tz);
 814
 815             fjx2             = _mm256_add_ps(fjx2,tx);
 816             fjy2             = _mm256_add_ps(fjy2,ty);
 817             fjz2             = _mm256_add_ps(fjz2,tz);
 818
 819             /**************************
 820              * CALCULATE INTERACTIONS *
 821              **************************/
 822
 823             /* COULOMB ELECTROSTATICS */
 824             velec            = _mm256_mul_ps(qq23,rinv23);
 825             felec            = _mm256_mul_ps(velec,rinvsq23);
 826
 827             /* Update potential sum for this i atom from the interaction with this j atom. */
 828             velec            = _mm256_andnot_ps(dummy_mask,velec);
 829             velecsum         = _mm256_add_ps(velecsum,velec);
 830
 831             fscal            = felec;
 832
 833             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 834
 835             /* Calculate temporary vectorial force */
 836             tx               = _mm256_mul_ps(fscal,dx23);
 837             ty               = _mm256_mul_ps(fscal,dy23);
 838             tz               = _mm256_mul_ps(fscal,dz23);
 839
 840             /* Update vectorial force */
 841             fix2             = _mm256_add_ps(fix2,tx);
 842             fiy2             = _mm256_add_ps(fiy2,ty);
 843             fiz2             = _mm256_add_ps(fiz2,tz);
 844
 845             fjx3             = _mm256_add_ps(fjx3,tx);
 846             fjy3             = _mm256_add_ps(fjy3,ty);
 847             fjz3             = _mm256_add_ps(fjz3,tz);
 848
 849             /**************************
 850              * CALCULATE INTERACTIONS *
 851              **************************/
 852
 853             /* COULOMB ELECTROSTATICS */
 854             velec            = _mm256_mul_ps(qq31,rinv31);
 855             felec            = _mm256_mul_ps(velec,rinvsq31);
 856
 857             /* Update potential sum for this i atom from the interaction with this j atom. */
 858             velec            = _mm256_andnot_ps(dummy_mask,velec);
 859             velecsum         = _mm256_add_ps(velecsum,velec);
 860
 861             fscal            = felec;
 862
 863             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 864
 865             /* Calculate temporary vectorial force */
 866             tx               = _mm256_mul_ps(fscal,dx31);
 867             ty               = _mm256_mul_ps(fscal,dy31);
 868             tz               = _mm256_mul_ps(fscal,dz31);
 869
 870             /* Update vectorial force */
 871             fix3             = _mm256_add_ps(fix3,tx);
 872             fiy3             = _mm256_add_ps(fiy3,ty);
 873             fiz3             = _mm256_add_ps(fiz3,tz);
 874
 875             fjx1             = _mm256_add_ps(fjx1,tx);
 876             fjy1             = _mm256_add_ps(fjy1,ty);
 877             fjz1             = _mm256_add_ps(fjz1,tz);
 878
 879             /**************************
 880              * CALCULATE INTERACTIONS *
 881              **************************/
 882
 883             /* COULOMB ELECTROSTATICS */
 884             velec            = _mm256_mul_ps(qq32,rinv32);
 885             felec            = _mm256_mul_ps(velec,rinvsq32);
 886
 887             /* Update potential sum for this i atom from the interaction with this j atom. */
 888             velec            = _mm256_andnot_ps(dummy_mask,velec);
 889             velecsum         = _mm256_add_ps(velecsum,velec);
 890
 891             fscal            = felec;
 892
 893             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 894
 895             /* Calculate temporary vectorial force */
 896             tx               = _mm256_mul_ps(fscal,dx32);
 897             ty               = _mm256_mul_ps(fscal,dy32);
 898             tz               = _mm256_mul_ps(fscal,dz32);
 899
 900             /* Update vectorial force */
 901             fix3             = _mm256_add_ps(fix3,tx);
 902             fiy3             = _mm256_add_ps(fiy3,ty);
 903             fiz3             = _mm256_add_ps(fiz3,tz);
 904
 905             fjx2             = _mm256_add_ps(fjx2,tx);
 906             fjy2             = _mm256_add_ps(fjy2,ty);
 907             fjz2             = _mm256_add_ps(fjz2,tz);
 908
 909             /**************************
 910              * CALCULATE INTERACTIONS *
 911              **************************/
 912
 913             /* COULOMB ELECTROSTATICS */
 914             velec            = _mm256_mul_ps(qq33,rinv33);
 915             felec            = _mm256_mul_ps(velec,rinvsq33);
 916
 917             /* Update potential sum for this i atom from the interaction with this j atom. */
 918             velec            = _mm256_andnot_ps(dummy_mask,velec);
 919             velecsum         = _mm256_add_ps(velecsum,velec);
 920
 921             fscal            = felec;
 922
 923             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 924
 925             /* Calculate temporary vectorial force */
 926             tx               = _mm256_mul_ps(fscal,dx33);
 927             ty               = _mm256_mul_ps(fscal,dy33);
 928             tz               = _mm256_mul_ps(fscal,dz33);
 929
 930             /* Update vectorial force */
 931             fix3             = _mm256_add_ps(fix3,tx);
 932             fiy3             = _mm256_add_ps(fiy3,ty);
 933             fiz3             = _mm256_add_ps(fiz3,tz);
 934
 935             fjx3             = _mm256_add_ps(fjx3,tx);
 936             fjy3             = _mm256_add_ps(fjy3,ty);
 937             fjz3             = _mm256_add_ps(fjz3,tz);
 938
 939             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 940             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 941             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 942             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 943             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
 944             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
 945             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
 946             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
 947
 948             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
 949                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
 950                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 951
 952             /* Inner loop uses 243 flops */
 953         }
 954
 955         /* End of innermost loop */
 956
 957         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 958                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
 959
 960         ggid                        = gid[iidx];
 961         /* Update potential energies */
 962         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 963
 964         /* Increment number of inner iterations */
 965         inneriter                  += j_index_end - j_index_start;
 966
 967         /* Outer loop uses 19 flops */
 968     }
 969
 970     /* Increment number of outer iterations */
 971     outeriter        += nri;
 972
 973     /* Update outer/inner flops */
 974
 975     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*243);
 976 }
 977 /*
 978  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 979  * Electrostatics interaction: Coulomb
 980  * VdW interaction:            None
 981  * Geometry:                   Water4-Water4
 982  * Calculate force/pot:        Force
 983  */
 984 void
 985 nb_kernel_ElecCoul_VdwNone_GeomW4W4_F_avx_256_single
 986                     (t_nblist                    * gmx_restrict       nlist,
 987                      rvec                        * gmx_restrict          xx,
 988                      rvec                        * gmx_restrict          ff,
 989                      t_forcerec                  * gmx_restrict          fr,
 990                      t_mdatoms                   * gmx_restrict     mdatoms,
 991                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 992                      t_nrnb                      * gmx_restrict        nrnb)
 993 {
 994     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 995      * just 0 for non-waters.
 996      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
 997      * jnr indices corresponding to data put in the four positions in the SIMD register.
 998      */
 999     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1000     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1001     int              jnrA,jnrB,jnrC,jnrD;
1002     int              jnrE,jnrF,jnrG,jnrH;
1003     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1004     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1005     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1006     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1007     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1008     real             rcutoff_scalar;
1009     real             *shiftvec,*fshift,*x,*f;
1010     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1011     real             scratch[4*DIM];
1012     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
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     real *           vdwioffsetptr3;
1018     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
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     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1024     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1025     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1026     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1027     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1028     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1029     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1030     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1031     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1032     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1033     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1034     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1035     real             *charge;
1036     __m256           dummy_mask,cutoff_mask;
1037     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1038     __m256           one     = _mm256_set1_ps(1.0);
1039     __m256           two     = _mm256_set1_ps(2.0);
1040     x                = xx[0];
1041     f                = ff[0];
1042
1043     nri              = nlist->nri;
1044     iinr             = nlist->iinr;
1045     jindex           = nlist->jindex;
1046     jjnr             = nlist->jjnr;
1047     shiftidx         = nlist->shift;
1048     gid              = nlist->gid;
1049     shiftvec         = fr->shift_vec[0];
1050     fshift           = fr->fshift[0];
1051     facel            = _mm256_set1_ps(fr->epsfac);
1052     charge           = mdatoms->chargeA;
1053
1054     /* Setup water-specific parameters */
1055     inr              = nlist->iinr[0];
1056     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1057     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1058     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1059
1060     jq1              = _mm256_set1_ps(charge[inr+1]);
1061     jq2              = _mm256_set1_ps(charge[inr+2]);
1062     jq3              = _mm256_set1_ps(charge[inr+3]);
1063     qq11             = _mm256_mul_ps(iq1,jq1);
1064     qq12             = _mm256_mul_ps(iq1,jq2);
1065     qq13             = _mm256_mul_ps(iq1,jq3);
1066     qq21             = _mm256_mul_ps(iq2,jq1);
1067     qq22             = _mm256_mul_ps(iq2,jq2);
1068     qq23             = _mm256_mul_ps(iq2,jq3);
1069     qq31             = _mm256_mul_ps(iq3,jq1);
1070     qq32             = _mm256_mul_ps(iq3,jq2);
1071     qq33             = _mm256_mul_ps(iq3,jq3);
1072
1073     /* Avoid stupid compiler warnings */
1074     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1075     j_coord_offsetA = 0;
1076     j_coord_offsetB = 0;
1077     j_coord_offsetC = 0;
1078     j_coord_offsetD = 0;
1079     j_coord_offsetE = 0;
1080     j_coord_offsetF = 0;
1081     j_coord_offsetG = 0;
1082     j_coord_offsetH = 0;
1083
1084     outeriter        = 0;
1085     inneriter        = 0;
1086
1087     for(iidx=0;iidx<4*DIM;iidx++)
1088     {
1089         scratch[iidx] = 0.0;
1090     }
1091
1092     /* Start outer loop over neighborlists */
1093     for(iidx=0; iidx<nri; iidx++)
1094     {
1095         /* Load shift vector for this list */
1096         i_shift_offset   = DIM*shiftidx[iidx];
1097
1098         /* Load limits for loop over neighbors */
1099         j_index_start    = jindex[iidx];
1100         j_index_end      = jindex[iidx+1];
1101
1102         /* Get outer coordinate index */
1103         inr              = iinr[iidx];
1104         i_coord_offset   = DIM*inr;
1105
1106         /* Load i particle coords and add shift vector */
1107         gmx_mm256_load_shift_and_3rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1108                                                     &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1109
1110         fix1             = _mm256_setzero_ps();
1111         fiy1             = _mm256_setzero_ps();
1112         fiz1             = _mm256_setzero_ps();
1113         fix2             = _mm256_setzero_ps();
1114         fiy2             = _mm256_setzero_ps();
1115         fiz2             = _mm256_setzero_ps();
1116         fix3             = _mm256_setzero_ps();
1117         fiy3             = _mm256_setzero_ps();
1118         fiz3             = _mm256_setzero_ps();
1119
1120         /* Start inner kernel loop */
1121         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1122         {
1123
1124             /* Get j neighbor index, and coordinate index */
1125             jnrA             = jjnr[jidx];
1126             jnrB             = jjnr[jidx+1];
1127             jnrC             = jjnr[jidx+2];
1128             jnrD             = jjnr[jidx+3];
1129             jnrE             = jjnr[jidx+4];
1130             jnrF             = jjnr[jidx+5];
1131             jnrG             = jjnr[jidx+6];
1132             jnrH             = jjnr[jidx+7];
1133             j_coord_offsetA  = DIM*jnrA;
1134             j_coord_offsetB  = DIM*jnrB;
1135             j_coord_offsetC  = DIM*jnrC;
1136             j_coord_offsetD  = DIM*jnrD;
1137             j_coord_offsetE  = DIM*jnrE;
1138             j_coord_offsetF  = DIM*jnrF;
1139             j_coord_offsetG  = DIM*jnrG;
1140             j_coord_offsetH  = DIM*jnrH;
1141
1142             /* load j atom coordinates */
1143             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1144                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1145                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1146                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1147                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1148
1149             /* Calculate displacement vector */
1150             dx11             = _mm256_sub_ps(ix1,jx1);
1151             dy11             = _mm256_sub_ps(iy1,jy1);
1152             dz11             = _mm256_sub_ps(iz1,jz1);
1153             dx12             = _mm256_sub_ps(ix1,jx2);
1154             dy12             = _mm256_sub_ps(iy1,jy2);
1155             dz12             = _mm256_sub_ps(iz1,jz2);
1156             dx13             = _mm256_sub_ps(ix1,jx3);
1157             dy13             = _mm256_sub_ps(iy1,jy3);
1158             dz13             = _mm256_sub_ps(iz1,jz3);
1159             dx21             = _mm256_sub_ps(ix2,jx1);
1160             dy21             = _mm256_sub_ps(iy2,jy1);
1161             dz21             = _mm256_sub_ps(iz2,jz1);
1162             dx22             = _mm256_sub_ps(ix2,jx2);
1163             dy22             = _mm256_sub_ps(iy2,jy2);
1164             dz22             = _mm256_sub_ps(iz2,jz2);
1165             dx23             = _mm256_sub_ps(ix2,jx3);
1166             dy23             = _mm256_sub_ps(iy2,jy3);
1167             dz23             = _mm256_sub_ps(iz2,jz3);
1168             dx31             = _mm256_sub_ps(ix3,jx1);
1169             dy31             = _mm256_sub_ps(iy3,jy1);
1170             dz31             = _mm256_sub_ps(iz3,jz1);
1171             dx32             = _mm256_sub_ps(ix3,jx2);
1172             dy32             = _mm256_sub_ps(iy3,jy2);
1173             dz32             = _mm256_sub_ps(iz3,jz2);
1174             dx33             = _mm256_sub_ps(ix3,jx3);
1175             dy33             = _mm256_sub_ps(iy3,jy3);
1176             dz33             = _mm256_sub_ps(iz3,jz3);
1177
1178             /* Calculate squared distance and things based on it */
1179             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1180             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1181             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1182             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1183             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1184             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1185             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1186             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1187             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1188
1189             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1190             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1191             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1192             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1193             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1194             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1195             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1196             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1197             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1198
1199             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1200             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1201             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1202             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1203             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1204             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1205             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1206             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1207             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1208
1209             fjx1             = _mm256_setzero_ps();
1210             fjy1             = _mm256_setzero_ps();
1211             fjz1             = _mm256_setzero_ps();
1212             fjx2             = _mm256_setzero_ps();
1213             fjy2             = _mm256_setzero_ps();
1214             fjz2             = _mm256_setzero_ps();
1215             fjx3             = _mm256_setzero_ps();
1216             fjy3             = _mm256_setzero_ps();
1217             fjz3             = _mm256_setzero_ps();
1218
1219             /**************************
1220              * CALCULATE INTERACTIONS *
1221              **************************/
1222
1223             /* COULOMB ELECTROSTATICS */
1224             velec            = _mm256_mul_ps(qq11,rinv11);
1225             felec            = _mm256_mul_ps(velec,rinvsq11);
1226
1227             fscal            = felec;
1228
1229             /* Calculate temporary vectorial force */
1230             tx               = _mm256_mul_ps(fscal,dx11);
1231             ty               = _mm256_mul_ps(fscal,dy11);
1232             tz               = _mm256_mul_ps(fscal,dz11);
1233
1234             /* Update vectorial force */
1235             fix1             = _mm256_add_ps(fix1,tx);
1236             fiy1             = _mm256_add_ps(fiy1,ty);
1237             fiz1             = _mm256_add_ps(fiz1,tz);
1238
1239             fjx1             = _mm256_add_ps(fjx1,tx);
1240             fjy1             = _mm256_add_ps(fjy1,ty);
1241             fjz1             = _mm256_add_ps(fjz1,tz);
1242
1243             /**************************
1244              * CALCULATE INTERACTIONS *
1245              **************************/
1246
1247             /* COULOMB ELECTROSTATICS */
1248             velec            = _mm256_mul_ps(qq12,rinv12);
1249             felec            = _mm256_mul_ps(velec,rinvsq12);
1250
1251             fscal            = felec;
1252
1253             /* Calculate temporary vectorial force */
1254             tx               = _mm256_mul_ps(fscal,dx12);
1255             ty               = _mm256_mul_ps(fscal,dy12);
1256             tz               = _mm256_mul_ps(fscal,dz12);
1257
1258             /* Update vectorial force */
1259             fix1             = _mm256_add_ps(fix1,tx);
1260             fiy1             = _mm256_add_ps(fiy1,ty);
1261             fiz1             = _mm256_add_ps(fiz1,tz);
1262
1263             fjx2             = _mm256_add_ps(fjx2,tx);
1264             fjy2             = _mm256_add_ps(fjy2,ty);
1265             fjz2             = _mm256_add_ps(fjz2,tz);
1266
1267             /**************************
1268              * CALCULATE INTERACTIONS *
1269              **************************/
1270
1271             /* COULOMB ELECTROSTATICS */
1272             velec            = _mm256_mul_ps(qq13,rinv13);
1273             felec            = _mm256_mul_ps(velec,rinvsq13);
1274
1275             fscal            = felec;
1276
1277             /* Calculate temporary vectorial force */
1278             tx               = _mm256_mul_ps(fscal,dx13);
1279             ty               = _mm256_mul_ps(fscal,dy13);
1280             tz               = _mm256_mul_ps(fscal,dz13);
1281
1282             /* Update vectorial force */
1283             fix1             = _mm256_add_ps(fix1,tx);
1284             fiy1             = _mm256_add_ps(fiy1,ty);
1285             fiz1             = _mm256_add_ps(fiz1,tz);
1286
1287             fjx3             = _mm256_add_ps(fjx3,tx);
1288             fjy3             = _mm256_add_ps(fjy3,ty);
1289             fjz3             = _mm256_add_ps(fjz3,tz);
1290
1291             /**************************
1292              * CALCULATE INTERACTIONS *
1293              **************************/
1294
1295             /* COULOMB ELECTROSTATICS */
1296             velec            = _mm256_mul_ps(qq21,rinv21);
1297             felec            = _mm256_mul_ps(velec,rinvsq21);
1298
1299             fscal            = felec;
1300
1301             /* Calculate temporary vectorial force */
1302             tx               = _mm256_mul_ps(fscal,dx21);
1303             ty               = _mm256_mul_ps(fscal,dy21);
1304             tz               = _mm256_mul_ps(fscal,dz21);
1305
1306             /* Update vectorial force */
1307             fix2             = _mm256_add_ps(fix2,tx);
1308             fiy2             = _mm256_add_ps(fiy2,ty);
1309             fiz2             = _mm256_add_ps(fiz2,tz);
1310
1311             fjx1             = _mm256_add_ps(fjx1,tx);
1312             fjy1             = _mm256_add_ps(fjy1,ty);
1313             fjz1             = _mm256_add_ps(fjz1,tz);
1314
1315             /**************************
1316              * CALCULATE INTERACTIONS *
1317              **************************/
1318
1319             /* COULOMB ELECTROSTATICS */
1320             velec            = _mm256_mul_ps(qq22,rinv22);
1321             felec            = _mm256_mul_ps(velec,rinvsq22);
1322
1323             fscal            = felec;
1324
1325             /* Calculate temporary vectorial force */
1326             tx               = _mm256_mul_ps(fscal,dx22);
1327             ty               = _mm256_mul_ps(fscal,dy22);
1328             tz               = _mm256_mul_ps(fscal,dz22);
1329
1330             /* Update vectorial force */
1331             fix2             = _mm256_add_ps(fix2,tx);
1332             fiy2             = _mm256_add_ps(fiy2,ty);
1333             fiz2             = _mm256_add_ps(fiz2,tz);
1334
1335             fjx2             = _mm256_add_ps(fjx2,tx);
1336             fjy2             = _mm256_add_ps(fjy2,ty);
1337             fjz2             = _mm256_add_ps(fjz2,tz);
1338
1339             /**************************
1340              * CALCULATE INTERACTIONS *
1341              **************************/
1342
1343             /* COULOMB ELECTROSTATICS */
1344             velec            = _mm256_mul_ps(qq23,rinv23);
1345             felec            = _mm256_mul_ps(velec,rinvsq23);
1346
1347             fscal            = felec;
1348
1349             /* Calculate temporary vectorial force */
1350             tx               = _mm256_mul_ps(fscal,dx23);
1351             ty               = _mm256_mul_ps(fscal,dy23);
1352             tz               = _mm256_mul_ps(fscal,dz23);
1353
1354             /* Update vectorial force */
1355             fix2             = _mm256_add_ps(fix2,tx);
1356             fiy2             = _mm256_add_ps(fiy2,ty);
1357             fiz2             = _mm256_add_ps(fiz2,tz);
1358
1359             fjx3             = _mm256_add_ps(fjx3,tx);
1360             fjy3             = _mm256_add_ps(fjy3,ty);
1361             fjz3             = _mm256_add_ps(fjz3,tz);
1362
1363             /**************************
1364              * CALCULATE INTERACTIONS *
1365              **************************/
1366
1367             /* COULOMB ELECTROSTATICS */
1368             velec            = _mm256_mul_ps(qq31,rinv31);
1369             felec            = _mm256_mul_ps(velec,rinvsq31);
1370
1371             fscal            = felec;
1372
1373             /* Calculate temporary vectorial force */
1374             tx               = _mm256_mul_ps(fscal,dx31);
1375             ty               = _mm256_mul_ps(fscal,dy31);
1376             tz               = _mm256_mul_ps(fscal,dz31);
1377
1378             /* Update vectorial force */
1379             fix3             = _mm256_add_ps(fix3,tx);
1380             fiy3             = _mm256_add_ps(fiy3,ty);
1381             fiz3             = _mm256_add_ps(fiz3,tz);
1382
1383             fjx1             = _mm256_add_ps(fjx1,tx);
1384             fjy1             = _mm256_add_ps(fjy1,ty);
1385             fjz1             = _mm256_add_ps(fjz1,tz);
1386
1387             /**************************
1388              * CALCULATE INTERACTIONS *
1389              **************************/
1390
1391             /* COULOMB ELECTROSTATICS */
1392             velec            = _mm256_mul_ps(qq32,rinv32);
1393             felec            = _mm256_mul_ps(velec,rinvsq32);
1394
1395             fscal            = felec;
1396
1397             /* Calculate temporary vectorial force */
1398             tx               = _mm256_mul_ps(fscal,dx32);
1399             ty               = _mm256_mul_ps(fscal,dy32);
1400             tz               = _mm256_mul_ps(fscal,dz32);
1401
1402             /* Update vectorial force */
1403             fix3             = _mm256_add_ps(fix3,tx);
1404             fiy3             = _mm256_add_ps(fiy3,ty);
1405             fiz3             = _mm256_add_ps(fiz3,tz);
1406
1407             fjx2             = _mm256_add_ps(fjx2,tx);
1408             fjy2             = _mm256_add_ps(fjy2,ty);
1409             fjz2             = _mm256_add_ps(fjz2,tz);
1410
1411             /**************************
1412              * CALCULATE INTERACTIONS *
1413              **************************/
1414
1415             /* COULOMB ELECTROSTATICS */
1416             velec            = _mm256_mul_ps(qq33,rinv33);
1417             felec            = _mm256_mul_ps(velec,rinvsq33);
1418
1419             fscal            = felec;
1420
1421             /* Calculate temporary vectorial force */
1422             tx               = _mm256_mul_ps(fscal,dx33);
1423             ty               = _mm256_mul_ps(fscal,dy33);
1424             tz               = _mm256_mul_ps(fscal,dz33);
1425
1426             /* Update vectorial force */
1427             fix3             = _mm256_add_ps(fix3,tx);
1428             fiy3             = _mm256_add_ps(fiy3,ty);
1429             fiz3             = _mm256_add_ps(fiz3,tz);
1430
1431             fjx3             = _mm256_add_ps(fjx3,tx);
1432             fjy3             = _mm256_add_ps(fjy3,ty);
1433             fjz3             = _mm256_add_ps(fjz3,tz);
1434
1435             fjptrA             = f+j_coord_offsetA;
1436             fjptrB             = f+j_coord_offsetB;
1437             fjptrC             = f+j_coord_offsetC;
1438             fjptrD             = f+j_coord_offsetD;
1439             fjptrE             = f+j_coord_offsetE;
1440             fjptrF             = f+j_coord_offsetF;
1441             fjptrG             = f+j_coord_offsetG;
1442             fjptrH             = f+j_coord_offsetH;
1443
1444             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1445                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1446                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1447
1448             /* Inner loop uses 234 flops */
1449         }
1450
1451         if(jidx<j_index_end)
1452         {
1453
1454             /* Get j neighbor index, and coordinate index */
1455             jnrlistA         = jjnr[jidx];
1456             jnrlistB         = jjnr[jidx+1];
1457             jnrlistC         = jjnr[jidx+2];
1458             jnrlistD         = jjnr[jidx+3];
1459             jnrlistE         = jjnr[jidx+4];
1460             jnrlistF         = jjnr[jidx+5];
1461             jnrlistG         = jjnr[jidx+6];
1462             jnrlistH         = jjnr[jidx+7];
1463             /* Sign of each element will be negative for non-real atoms.
1464              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1465              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1466              */
1467             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1468                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1469
1470             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1471             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1472             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1473             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1474             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1475             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1476             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1477             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1478             j_coord_offsetA  = DIM*jnrA;
1479             j_coord_offsetB  = DIM*jnrB;
1480             j_coord_offsetC  = DIM*jnrC;
1481             j_coord_offsetD  = DIM*jnrD;
1482             j_coord_offsetE  = DIM*jnrE;
1483             j_coord_offsetF  = DIM*jnrF;
1484             j_coord_offsetG  = DIM*jnrG;
1485             j_coord_offsetH  = DIM*jnrH;
1486
1487             /* load j atom coordinates */
1488             gmx_mm256_load_3rvec_8ptr_swizzle_ps(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1489                                                  x+j_coord_offsetC+DIM,x+j_coord_offsetD+DIM,
1490                                                  x+j_coord_offsetE+DIM,x+j_coord_offsetF+DIM,
1491                                                  x+j_coord_offsetG+DIM,x+j_coord_offsetH+DIM,
1492                                                  &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1493
1494             /* Calculate displacement vector */
1495             dx11             = _mm256_sub_ps(ix1,jx1);
1496             dy11             = _mm256_sub_ps(iy1,jy1);
1497             dz11             = _mm256_sub_ps(iz1,jz1);
1498             dx12             = _mm256_sub_ps(ix1,jx2);
1499             dy12             = _mm256_sub_ps(iy1,jy2);
1500             dz12             = _mm256_sub_ps(iz1,jz2);
1501             dx13             = _mm256_sub_ps(ix1,jx3);
1502             dy13             = _mm256_sub_ps(iy1,jy3);
1503             dz13             = _mm256_sub_ps(iz1,jz3);
1504             dx21             = _mm256_sub_ps(ix2,jx1);
1505             dy21             = _mm256_sub_ps(iy2,jy1);
1506             dz21             = _mm256_sub_ps(iz2,jz1);
1507             dx22             = _mm256_sub_ps(ix2,jx2);
1508             dy22             = _mm256_sub_ps(iy2,jy2);
1509             dz22             = _mm256_sub_ps(iz2,jz2);
1510             dx23             = _mm256_sub_ps(ix2,jx3);
1511             dy23             = _mm256_sub_ps(iy2,jy3);
1512             dz23             = _mm256_sub_ps(iz2,jz3);
1513             dx31             = _mm256_sub_ps(ix3,jx1);
1514             dy31             = _mm256_sub_ps(iy3,jy1);
1515             dz31             = _mm256_sub_ps(iz3,jz1);
1516             dx32             = _mm256_sub_ps(ix3,jx2);
1517             dy32             = _mm256_sub_ps(iy3,jy2);
1518             dz32             = _mm256_sub_ps(iz3,jz2);
1519             dx33             = _mm256_sub_ps(ix3,jx3);
1520             dy33             = _mm256_sub_ps(iy3,jy3);
1521             dz33             = _mm256_sub_ps(iz3,jz3);
1522
1523             /* Calculate squared distance and things based on it */
1524             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1525             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1526             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1527             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1528             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1529             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1530             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1531             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1532             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1533
1534             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1535             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1536             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1537             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1538             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1539             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1540             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1541             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1542             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1543
1544             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1545             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1546             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1547             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1548             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1549             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1550             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1551             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1552             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1553
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             fjx3             = _mm256_setzero_ps();
1561             fjy3             = _mm256_setzero_ps();
1562             fjz3             = _mm256_setzero_ps();
1563
1564             /**************************
1565              * CALCULATE INTERACTIONS *
1566              **************************/
1567
1568             /* COULOMB ELECTROSTATICS */
1569             velec            = _mm256_mul_ps(qq11,rinv11);
1570             felec            = _mm256_mul_ps(velec,rinvsq11);
1571
1572             fscal            = felec;
1573
1574             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1575
1576             /* Calculate temporary vectorial force */
1577             tx               = _mm256_mul_ps(fscal,dx11);
1578             ty               = _mm256_mul_ps(fscal,dy11);
1579             tz               = _mm256_mul_ps(fscal,dz11);
1580
1581             /* Update vectorial force */
1582             fix1             = _mm256_add_ps(fix1,tx);
1583             fiy1             = _mm256_add_ps(fiy1,ty);
1584             fiz1             = _mm256_add_ps(fiz1,tz);
1585
1586             fjx1             = _mm256_add_ps(fjx1,tx);
1587             fjy1             = _mm256_add_ps(fjy1,ty);
1588             fjz1             = _mm256_add_ps(fjz1,tz);
1589
1590             /**************************
1591              * CALCULATE INTERACTIONS *
1592              **************************/
1593
1594             /* COULOMB ELECTROSTATICS */
1595             velec            = _mm256_mul_ps(qq12,rinv12);
1596             felec            = _mm256_mul_ps(velec,rinvsq12);
1597
1598             fscal            = felec;
1599
1600             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1601
1602             /* Calculate temporary vectorial force */
1603             tx               = _mm256_mul_ps(fscal,dx12);
1604             ty               = _mm256_mul_ps(fscal,dy12);
1605             tz               = _mm256_mul_ps(fscal,dz12);
1606
1607             /* Update vectorial force */
1608             fix1             = _mm256_add_ps(fix1,tx);
1609             fiy1             = _mm256_add_ps(fiy1,ty);
1610             fiz1             = _mm256_add_ps(fiz1,tz);
1611
1612             fjx2             = _mm256_add_ps(fjx2,tx);
1613             fjy2             = _mm256_add_ps(fjy2,ty);
1614             fjz2             = _mm256_add_ps(fjz2,tz);
1615
1616             /**************************
1617              * CALCULATE INTERACTIONS *
1618              **************************/
1619
1620             /* COULOMB ELECTROSTATICS */
1621             velec            = _mm256_mul_ps(qq13,rinv13);
1622             felec            = _mm256_mul_ps(velec,rinvsq13);
1623
1624             fscal            = felec;
1625
1626             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1627
1628             /* Calculate temporary vectorial force */
1629             tx               = _mm256_mul_ps(fscal,dx13);
1630             ty               = _mm256_mul_ps(fscal,dy13);
1631             tz               = _mm256_mul_ps(fscal,dz13);
1632
1633             /* Update vectorial force */
1634             fix1             = _mm256_add_ps(fix1,tx);
1635             fiy1             = _mm256_add_ps(fiy1,ty);
1636             fiz1             = _mm256_add_ps(fiz1,tz);
1637
1638             fjx3             = _mm256_add_ps(fjx3,tx);
1639             fjy3             = _mm256_add_ps(fjy3,ty);
1640             fjz3             = _mm256_add_ps(fjz3,tz);
1641
1642             /**************************
1643              * CALCULATE INTERACTIONS *
1644              **************************/
1645
1646             /* COULOMB ELECTROSTATICS */
1647             velec            = _mm256_mul_ps(qq21,rinv21);
1648             felec            = _mm256_mul_ps(velec,rinvsq21);
1649
1650             fscal            = felec;
1651
1652             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1653
1654             /* Calculate temporary vectorial force */
1655             tx               = _mm256_mul_ps(fscal,dx21);
1656             ty               = _mm256_mul_ps(fscal,dy21);
1657             tz               = _mm256_mul_ps(fscal,dz21);
1658
1659             /* Update vectorial force */
1660             fix2             = _mm256_add_ps(fix2,tx);
1661             fiy2             = _mm256_add_ps(fiy2,ty);
1662             fiz2             = _mm256_add_ps(fiz2,tz);
1663
1664             fjx1             = _mm256_add_ps(fjx1,tx);
1665             fjy1             = _mm256_add_ps(fjy1,ty);
1666             fjz1             = _mm256_add_ps(fjz1,tz);
1667
1668             /**************************
1669              * CALCULATE INTERACTIONS *
1670              **************************/
1671
1672             /* COULOMB ELECTROSTATICS */
1673             velec            = _mm256_mul_ps(qq22,rinv22);
1674             felec            = _mm256_mul_ps(velec,rinvsq22);
1675
1676             fscal            = felec;
1677
1678             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1679
1680             /* Calculate temporary vectorial force */
1681             tx               = _mm256_mul_ps(fscal,dx22);
1682             ty               = _mm256_mul_ps(fscal,dy22);
1683             tz               = _mm256_mul_ps(fscal,dz22);
1684
1685             /* Update vectorial force */
1686             fix2             = _mm256_add_ps(fix2,tx);
1687             fiy2             = _mm256_add_ps(fiy2,ty);
1688             fiz2             = _mm256_add_ps(fiz2,tz);
1689
1690             fjx2             = _mm256_add_ps(fjx2,tx);
1691             fjy2             = _mm256_add_ps(fjy2,ty);
1692             fjz2             = _mm256_add_ps(fjz2,tz);
1693
1694             /**************************
1695              * CALCULATE INTERACTIONS *
1696              **************************/
1697
1698             /* COULOMB ELECTROSTATICS */
1699             velec            = _mm256_mul_ps(qq23,rinv23);
1700             felec            = _mm256_mul_ps(velec,rinvsq23);
1701
1702             fscal            = felec;
1703
1704             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1705
1706             /* Calculate temporary vectorial force */
1707             tx               = _mm256_mul_ps(fscal,dx23);
1708             ty               = _mm256_mul_ps(fscal,dy23);
1709             tz               = _mm256_mul_ps(fscal,dz23);
1710
1711             /* Update vectorial force */
1712             fix2             = _mm256_add_ps(fix2,tx);
1713             fiy2             = _mm256_add_ps(fiy2,ty);
1714             fiz2             = _mm256_add_ps(fiz2,tz);
1715
1716             fjx3             = _mm256_add_ps(fjx3,tx);
1717             fjy3             = _mm256_add_ps(fjy3,ty);
1718             fjz3             = _mm256_add_ps(fjz3,tz);
1719
1720             /**************************
1721              * CALCULATE INTERACTIONS *
1722              **************************/
1723
1724             /* COULOMB ELECTROSTATICS */
1725             velec            = _mm256_mul_ps(qq31,rinv31);
1726             felec            = _mm256_mul_ps(velec,rinvsq31);
1727
1728             fscal            = felec;
1729
1730             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1731
1732             /* Calculate temporary vectorial force */
1733             tx               = _mm256_mul_ps(fscal,dx31);
1734             ty               = _mm256_mul_ps(fscal,dy31);
1735             tz               = _mm256_mul_ps(fscal,dz31);
1736
1737             /* Update vectorial force */
1738             fix3             = _mm256_add_ps(fix3,tx);
1739             fiy3             = _mm256_add_ps(fiy3,ty);
1740             fiz3             = _mm256_add_ps(fiz3,tz);
1741
1742             fjx1             = _mm256_add_ps(fjx1,tx);
1743             fjy1             = _mm256_add_ps(fjy1,ty);
1744             fjz1             = _mm256_add_ps(fjz1,tz);
1745
1746             /**************************
1747              * CALCULATE INTERACTIONS *
1748              **************************/
1749
1750             /* COULOMB ELECTROSTATICS */
1751             velec            = _mm256_mul_ps(qq32,rinv32);
1752             felec            = _mm256_mul_ps(velec,rinvsq32);
1753
1754             fscal            = felec;
1755
1756             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1757
1758             /* Calculate temporary vectorial force */
1759             tx               = _mm256_mul_ps(fscal,dx32);
1760             ty               = _mm256_mul_ps(fscal,dy32);
1761             tz               = _mm256_mul_ps(fscal,dz32);
1762
1763             /* Update vectorial force */
1764             fix3             = _mm256_add_ps(fix3,tx);
1765             fiy3             = _mm256_add_ps(fiy3,ty);
1766             fiz3             = _mm256_add_ps(fiz3,tz);
1767
1768             fjx2             = _mm256_add_ps(fjx2,tx);
1769             fjy2             = _mm256_add_ps(fjy2,ty);
1770             fjz2             = _mm256_add_ps(fjz2,tz);
1771
1772             /**************************
1773              * CALCULATE INTERACTIONS *
1774              **************************/
1775
1776             /* COULOMB ELECTROSTATICS */
1777             velec            = _mm256_mul_ps(qq33,rinv33);
1778             felec            = _mm256_mul_ps(velec,rinvsq33);
1779
1780             fscal            = felec;
1781
1782             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1783
1784             /* Calculate temporary vectorial force */
1785             tx               = _mm256_mul_ps(fscal,dx33);
1786             ty               = _mm256_mul_ps(fscal,dy33);
1787             tz               = _mm256_mul_ps(fscal,dz33);
1788
1789             /* Update vectorial force */
1790             fix3             = _mm256_add_ps(fix3,tx);
1791             fiy3             = _mm256_add_ps(fiy3,ty);
1792             fiz3             = _mm256_add_ps(fiz3,tz);
1793
1794             fjx3             = _mm256_add_ps(fjx3,tx);
1795             fjy3             = _mm256_add_ps(fjy3,ty);
1796             fjz3             = _mm256_add_ps(fjz3,tz);
1797
1798             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1799             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1800             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1801             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1802             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1803             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1804             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1805             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1806
1807             gmx_mm256_decrement_3rvec_8ptr_swizzle_ps(fjptrA+DIM,fjptrB+DIM,fjptrC+DIM,fjptrD+DIM,
1808                                                       fjptrE+DIM,fjptrF+DIM,fjptrG+DIM,fjptrH+DIM,
1809                                                       fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1810
1811             /* Inner loop uses 234 flops */
1812         }
1813
1814         /* End of innermost loop */
1815
1816         gmx_mm256_update_iforce_3atom_swizzle_ps(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1817                                                  f+i_coord_offset+DIM,fshift+i_shift_offset);
1818
1819         /* Increment number of inner iterations */
1820         inneriter                  += j_index_end - j_index_start;
1821
1822         /* Outer loop uses 18 flops */
1823     }
1824
1825     /* Increment number of outer iterations */
1826     outeriter        += nri;
1827
1828     /* Update outer/inner flops */
1829
1830     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*234);
1831 }