src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecCoul_VdwLJ_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_VdwLJ_GeomW4W4_VF_avx_256_single
  54  * Electrostatics interaction: Coulomb
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecCoul_VdwLJ_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 *           vdwioffsetptr0;
  89     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  90     real *           vdwioffsetptr1;
  91     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  92     real *           vdwioffsetptr2;
  93     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  94     real *           vdwioffsetptr3;
  95     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  96     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  97     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  98     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  99     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
 100     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
 101     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 102     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
 103     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 104     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 105     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 106     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 107     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 108     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 109     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 110     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 111     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 112     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 113     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 114     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 115     real             *charge;
 116     int              nvdwtype;
 117     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 118     int              *vdwtype;
 119     real             *vdwparam;
 120     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 121     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 122     __m256           dummy_mask,cutoff_mask;
 123     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 124     __m256           one     = _mm256_set1_ps(1.0);
 125     __m256           two     = _mm256_set1_ps(2.0);
 126     x                = xx[0];
 127     f                = ff[0];
 128
 129     nri              = nlist->nri;
 130     iinr             = nlist->iinr;
 131     jindex           = nlist->jindex;
 132     jjnr             = nlist->jjnr;
 133     shiftidx         = nlist->shift;
 134     gid              = nlist->gid;
 135     shiftvec         = fr->shift_vec[0];
 136     fshift           = fr->fshift[0];
 137     facel            = _mm256_set1_ps(fr->epsfac);
 138     charge           = mdatoms->chargeA;
 139     nvdwtype         = fr->ntype;
 140     vdwparam         = fr->nbfp;
 141     vdwtype          = mdatoms->typeA;
 142
 143     /* Setup water-specific parameters */
 144     inr              = nlist->iinr[0];
 145     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 146     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 147     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 148     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 149
 150     jq1              = _mm256_set1_ps(charge[inr+1]);
 151     jq2              = _mm256_set1_ps(charge[inr+2]);
 152     jq3              = _mm256_set1_ps(charge[inr+3]);
 153     vdwjidx0A        = 2*vdwtype[inr+0];
 154     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 155     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 156     qq11             = _mm256_mul_ps(iq1,jq1);
 157     qq12             = _mm256_mul_ps(iq1,jq2);
 158     qq13             = _mm256_mul_ps(iq1,jq3);
 159     qq21             = _mm256_mul_ps(iq2,jq1);
 160     qq22             = _mm256_mul_ps(iq2,jq2);
 161     qq23             = _mm256_mul_ps(iq2,jq3);
 162     qq31             = _mm256_mul_ps(iq3,jq1);
 163     qq32             = _mm256_mul_ps(iq3,jq2);
 164     qq33             = _mm256_mul_ps(iq3,jq3);
 165
 166     /* Avoid stupid compiler warnings */
 167     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 168     j_coord_offsetA = 0;
 169     j_coord_offsetB = 0;
 170     j_coord_offsetC = 0;
 171     j_coord_offsetD = 0;
 172     j_coord_offsetE = 0;
 173     j_coord_offsetF = 0;
 174     j_coord_offsetG = 0;
 175     j_coord_offsetH = 0;
 176
 177     outeriter        = 0;
 178     inneriter        = 0;
 179
 180     for(iidx=0;iidx<4*DIM;iidx++)
 181     {
 182         scratch[iidx] = 0.0;
 183     }
 184
 185     /* Start outer loop over neighborlists */
 186     for(iidx=0; iidx<nri; iidx++)
 187     {
 188         /* Load shift vector for this list */
 189         i_shift_offset   = DIM*shiftidx[iidx];
 190
 191         /* Load limits for loop over neighbors */
 192         j_index_start    = jindex[iidx];
 193         j_index_end      = jindex[iidx+1];
 194
 195         /* Get outer coordinate index */
 196         inr              = iinr[iidx];
 197         i_coord_offset   = DIM*inr;
 198
 199         /* Load i particle coords and add shift vector */
 200         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 201                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 202
 203         fix0             = _mm256_setzero_ps();
 204         fiy0             = _mm256_setzero_ps();
 205         fiz0             = _mm256_setzero_ps();
 206         fix1             = _mm256_setzero_ps();
 207         fiy1             = _mm256_setzero_ps();
 208         fiz1             = _mm256_setzero_ps();
 209         fix2             = _mm256_setzero_ps();
 210         fiy2             = _mm256_setzero_ps();
 211         fiz2             = _mm256_setzero_ps();
 212         fix3             = _mm256_setzero_ps();
 213         fiy3             = _mm256_setzero_ps();
 214         fiz3             = _mm256_setzero_ps();
 215
 216         /* Reset potential sums */
 217         velecsum         = _mm256_setzero_ps();
 218         vvdwsum          = _mm256_setzero_ps();
 219
 220         /* Start inner kernel loop */
 221         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 222         {
 223
 224             /* Get j neighbor index, and coordinate index */
 225             jnrA             = jjnr[jidx];
 226             jnrB             = jjnr[jidx+1];
 227             jnrC             = jjnr[jidx+2];
 228             jnrD             = jjnr[jidx+3];
 229             jnrE             = jjnr[jidx+4];
 230             jnrF             = jjnr[jidx+5];
 231             jnrG             = jjnr[jidx+6];
 232             jnrH             = jjnr[jidx+7];
 233             j_coord_offsetA  = DIM*jnrA;
 234             j_coord_offsetB  = DIM*jnrB;
 235             j_coord_offsetC  = DIM*jnrC;
 236             j_coord_offsetD  = DIM*jnrD;
 237             j_coord_offsetE  = DIM*jnrE;
 238             j_coord_offsetF  = DIM*jnrF;
 239             j_coord_offsetG  = DIM*jnrG;
 240             j_coord_offsetH  = DIM*jnrH;
 241
 242             /* load j atom coordinates */
 243             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 244                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 245                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 246                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 247                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 248                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 249
 250             /* Calculate displacement vector */
 251             dx00             = _mm256_sub_ps(ix0,jx0);
 252             dy00             = _mm256_sub_ps(iy0,jy0);
 253             dz00             = _mm256_sub_ps(iz0,jz0);
 254             dx11             = _mm256_sub_ps(ix1,jx1);
 255             dy11             = _mm256_sub_ps(iy1,jy1);
 256             dz11             = _mm256_sub_ps(iz1,jz1);
 257             dx12             = _mm256_sub_ps(ix1,jx2);
 258             dy12             = _mm256_sub_ps(iy1,jy2);
 259             dz12             = _mm256_sub_ps(iz1,jz2);
 260             dx13             = _mm256_sub_ps(ix1,jx3);
 261             dy13             = _mm256_sub_ps(iy1,jy3);
 262             dz13             = _mm256_sub_ps(iz1,jz3);
 263             dx21             = _mm256_sub_ps(ix2,jx1);
 264             dy21             = _mm256_sub_ps(iy2,jy1);
 265             dz21             = _mm256_sub_ps(iz2,jz1);
 266             dx22             = _mm256_sub_ps(ix2,jx2);
 267             dy22             = _mm256_sub_ps(iy2,jy2);
 268             dz22             = _mm256_sub_ps(iz2,jz2);
 269             dx23             = _mm256_sub_ps(ix2,jx3);
 270             dy23             = _mm256_sub_ps(iy2,jy3);
 271             dz23             = _mm256_sub_ps(iz2,jz3);
 272             dx31             = _mm256_sub_ps(ix3,jx1);
 273             dy31             = _mm256_sub_ps(iy3,jy1);
 274             dz31             = _mm256_sub_ps(iz3,jz1);
 275             dx32             = _mm256_sub_ps(ix3,jx2);
 276             dy32             = _mm256_sub_ps(iy3,jy2);
 277             dz32             = _mm256_sub_ps(iz3,jz2);
 278             dx33             = _mm256_sub_ps(ix3,jx3);
 279             dy33             = _mm256_sub_ps(iy3,jy3);
 280             dz33             = _mm256_sub_ps(iz3,jz3);
 281
 282             /* Calculate squared distance and things based on it */
 283             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 284             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 285             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 286             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 287             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 288             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 289             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 290             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 291             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 292             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 293
 294             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 295             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 296             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 297             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 298             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 299             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 300             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 301             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 302             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 303
 304             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 305             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 306             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 307             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 308             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 309             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 310             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 311             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 312             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 313             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 314
 315             fjx0             = _mm256_setzero_ps();
 316             fjy0             = _mm256_setzero_ps();
 317             fjz0             = _mm256_setzero_ps();
 318             fjx1             = _mm256_setzero_ps();
 319             fjy1             = _mm256_setzero_ps();
 320             fjz1             = _mm256_setzero_ps();
 321             fjx2             = _mm256_setzero_ps();
 322             fjy2             = _mm256_setzero_ps();
 323             fjz2             = _mm256_setzero_ps();
 324             fjx3             = _mm256_setzero_ps();
 325             fjy3             = _mm256_setzero_ps();
 326             fjz3             = _mm256_setzero_ps();
 327
 328             /**************************
 329              * CALCULATE INTERACTIONS *
 330              **************************/
 331
 332             /* LENNARD-JONES DISPERSION/REPULSION */
 333
 334             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 335             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 336             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 337             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 338             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 339
 340             /* Update potential sum for this i atom from the interaction with this j atom. */
 341             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 342
 343             fscal            = fvdw;
 344
 345             /* Calculate temporary vectorial force */
 346             tx               = _mm256_mul_ps(fscal,dx00);
 347             ty               = _mm256_mul_ps(fscal,dy00);
 348             tz               = _mm256_mul_ps(fscal,dz00);
 349
 350             /* Update vectorial force */
 351             fix0             = _mm256_add_ps(fix0,tx);
 352             fiy0             = _mm256_add_ps(fiy0,ty);
 353             fiz0             = _mm256_add_ps(fiz0,tz);
 354
 355             fjx0             = _mm256_add_ps(fjx0,tx);
 356             fjy0             = _mm256_add_ps(fjy0,ty);
 357             fjz0             = _mm256_add_ps(fjz0,tz);
 358
 359             /**************************
 360              * CALCULATE INTERACTIONS *
 361              **************************/
 362
 363             /* COULOMB ELECTROSTATICS */
 364             velec            = _mm256_mul_ps(qq11,rinv11);
 365             felec            = _mm256_mul_ps(velec,rinvsq11);
 366
 367             /* Update potential sum for this i atom from the interaction with this j atom. */
 368             velecsum         = _mm256_add_ps(velecsum,velec);
 369
 370             fscal            = felec;
 371
 372             /* Calculate temporary vectorial force */
 373             tx               = _mm256_mul_ps(fscal,dx11);
 374             ty               = _mm256_mul_ps(fscal,dy11);
 375             tz               = _mm256_mul_ps(fscal,dz11);
 376
 377             /* Update vectorial force */
 378             fix1             = _mm256_add_ps(fix1,tx);
 379             fiy1             = _mm256_add_ps(fiy1,ty);
 380             fiz1             = _mm256_add_ps(fiz1,tz);
 381
 382             fjx1             = _mm256_add_ps(fjx1,tx);
 383             fjy1             = _mm256_add_ps(fjy1,ty);
 384             fjz1             = _mm256_add_ps(fjz1,tz);
 385
 386             /**************************
 387              * CALCULATE INTERACTIONS *
 388              **************************/
 389
 390             /* COULOMB ELECTROSTATICS */
 391             velec            = _mm256_mul_ps(qq12,rinv12);
 392             felec            = _mm256_mul_ps(velec,rinvsq12);
 393
 394             /* Update potential sum for this i atom from the interaction with this j atom. */
 395             velecsum         = _mm256_add_ps(velecsum,velec);
 396
 397             fscal            = felec;
 398
 399             /* Calculate temporary vectorial force */
 400             tx               = _mm256_mul_ps(fscal,dx12);
 401             ty               = _mm256_mul_ps(fscal,dy12);
 402             tz               = _mm256_mul_ps(fscal,dz12);
 403
 404             /* Update vectorial force */
 405             fix1             = _mm256_add_ps(fix1,tx);
 406             fiy1             = _mm256_add_ps(fiy1,ty);
 407             fiz1             = _mm256_add_ps(fiz1,tz);
 408
 409             fjx2             = _mm256_add_ps(fjx2,tx);
 410             fjy2             = _mm256_add_ps(fjy2,ty);
 411             fjz2             = _mm256_add_ps(fjz2,tz);
 412
 413             /**************************
 414              * CALCULATE INTERACTIONS *
 415              **************************/
 416
 417             /* COULOMB ELECTROSTATICS */
 418             velec            = _mm256_mul_ps(qq13,rinv13);
 419             felec            = _mm256_mul_ps(velec,rinvsq13);
 420
 421             /* Update potential sum for this i atom from the interaction with this j atom. */
 422             velecsum         = _mm256_add_ps(velecsum,velec);
 423
 424             fscal            = felec;
 425
 426             /* Calculate temporary vectorial force */
 427             tx               = _mm256_mul_ps(fscal,dx13);
 428             ty               = _mm256_mul_ps(fscal,dy13);
 429             tz               = _mm256_mul_ps(fscal,dz13);
 430
 431             /* Update vectorial force */
 432             fix1             = _mm256_add_ps(fix1,tx);
 433             fiy1             = _mm256_add_ps(fiy1,ty);
 434             fiz1             = _mm256_add_ps(fiz1,tz);
 435
 436             fjx3             = _mm256_add_ps(fjx3,tx);
 437             fjy3             = _mm256_add_ps(fjy3,ty);
 438             fjz3             = _mm256_add_ps(fjz3,tz);
 439
 440             /**************************
 441              * CALCULATE INTERACTIONS *
 442              **************************/
 443
 444             /* COULOMB ELECTROSTATICS */
 445             velec            = _mm256_mul_ps(qq21,rinv21);
 446             felec            = _mm256_mul_ps(velec,rinvsq21);
 447
 448             /* Update potential sum for this i atom from the interaction with this j atom. */
 449             velecsum         = _mm256_add_ps(velecsum,velec);
 450
 451             fscal            = felec;
 452
 453             /* Calculate temporary vectorial force */
 454             tx               = _mm256_mul_ps(fscal,dx21);
 455             ty               = _mm256_mul_ps(fscal,dy21);
 456             tz               = _mm256_mul_ps(fscal,dz21);
 457
 458             /* Update vectorial force */
 459             fix2             = _mm256_add_ps(fix2,tx);
 460             fiy2             = _mm256_add_ps(fiy2,ty);
 461             fiz2             = _mm256_add_ps(fiz2,tz);
 462
 463             fjx1             = _mm256_add_ps(fjx1,tx);
 464             fjy1             = _mm256_add_ps(fjy1,ty);
 465             fjz1             = _mm256_add_ps(fjz1,tz);
 466
 467             /**************************
 468              * CALCULATE INTERACTIONS *
 469              **************************/
 470
 471             /* COULOMB ELECTROSTATICS */
 472             velec            = _mm256_mul_ps(qq22,rinv22);
 473             felec            = _mm256_mul_ps(velec,rinvsq22);
 474
 475             /* Update potential sum for this i atom from the interaction with this j atom. */
 476             velecsum         = _mm256_add_ps(velecsum,velec);
 477
 478             fscal            = felec;
 479
 480             /* Calculate temporary vectorial force */
 481             tx               = _mm256_mul_ps(fscal,dx22);
 482             ty               = _mm256_mul_ps(fscal,dy22);
 483             tz               = _mm256_mul_ps(fscal,dz22);
 484
 485             /* Update vectorial force */
 486             fix2             = _mm256_add_ps(fix2,tx);
 487             fiy2             = _mm256_add_ps(fiy2,ty);
 488             fiz2             = _mm256_add_ps(fiz2,tz);
 489
 490             fjx2             = _mm256_add_ps(fjx2,tx);
 491             fjy2             = _mm256_add_ps(fjy2,ty);
 492             fjz2             = _mm256_add_ps(fjz2,tz);
 493
 494             /**************************
 495              * CALCULATE INTERACTIONS *
 496              **************************/
 497
 498             /* COULOMB ELECTROSTATICS */
 499             velec            = _mm256_mul_ps(qq23,rinv23);
 500             felec            = _mm256_mul_ps(velec,rinvsq23);
 501
 502             /* Update potential sum for this i atom from the interaction with this j atom. */
 503             velecsum         = _mm256_add_ps(velecsum,velec);
 504
 505             fscal            = felec;
 506
 507             /* Calculate temporary vectorial force */
 508             tx               = _mm256_mul_ps(fscal,dx23);
 509             ty               = _mm256_mul_ps(fscal,dy23);
 510             tz               = _mm256_mul_ps(fscal,dz23);
 511
 512             /* Update vectorial force */
 513             fix2             = _mm256_add_ps(fix2,tx);
 514             fiy2             = _mm256_add_ps(fiy2,ty);
 515             fiz2             = _mm256_add_ps(fiz2,tz);
 516
 517             fjx3             = _mm256_add_ps(fjx3,tx);
 518             fjy3             = _mm256_add_ps(fjy3,ty);
 519             fjz3             = _mm256_add_ps(fjz3,tz);
 520
 521             /**************************
 522              * CALCULATE INTERACTIONS *
 523              **************************/
 524
 525             /* COULOMB ELECTROSTATICS */
 526             velec            = _mm256_mul_ps(qq31,rinv31);
 527             felec            = _mm256_mul_ps(velec,rinvsq31);
 528
 529             /* Update potential sum for this i atom from the interaction with this j atom. */
 530             velecsum         = _mm256_add_ps(velecsum,velec);
 531
 532             fscal            = felec;
 533
 534             /* Calculate temporary vectorial force */
 535             tx               = _mm256_mul_ps(fscal,dx31);
 536             ty               = _mm256_mul_ps(fscal,dy31);
 537             tz               = _mm256_mul_ps(fscal,dz31);
 538
 539             /* Update vectorial force */
 540             fix3             = _mm256_add_ps(fix3,tx);
 541             fiy3             = _mm256_add_ps(fiy3,ty);
 542             fiz3             = _mm256_add_ps(fiz3,tz);
 543
 544             fjx1             = _mm256_add_ps(fjx1,tx);
 545             fjy1             = _mm256_add_ps(fjy1,ty);
 546             fjz1             = _mm256_add_ps(fjz1,tz);
 547
 548             /**************************
 549              * CALCULATE INTERACTIONS *
 550              **************************/
 551
 552             /* COULOMB ELECTROSTATICS */
 553             velec            = _mm256_mul_ps(qq32,rinv32);
 554             felec            = _mm256_mul_ps(velec,rinvsq32);
 555
 556             /* Update potential sum for this i atom from the interaction with this j atom. */
 557             velecsum         = _mm256_add_ps(velecsum,velec);
 558
 559             fscal            = felec;
 560
 561             /* Calculate temporary vectorial force */
 562             tx               = _mm256_mul_ps(fscal,dx32);
 563             ty               = _mm256_mul_ps(fscal,dy32);
 564             tz               = _mm256_mul_ps(fscal,dz32);
 565
 566             /* Update vectorial force */
 567             fix3             = _mm256_add_ps(fix3,tx);
 568             fiy3             = _mm256_add_ps(fiy3,ty);
 569             fiz3             = _mm256_add_ps(fiz3,tz);
 570
 571             fjx2             = _mm256_add_ps(fjx2,tx);
 572             fjy2             = _mm256_add_ps(fjy2,ty);
 573             fjz2             = _mm256_add_ps(fjz2,tz);
 574
 575             /**************************
 576              * CALCULATE INTERACTIONS *
 577              **************************/
 578
 579             /* COULOMB ELECTROSTATICS */
 580             velec            = _mm256_mul_ps(qq33,rinv33);
 581             felec            = _mm256_mul_ps(velec,rinvsq33);
 582
 583             /* Update potential sum for this i atom from the interaction with this j atom. */
 584             velecsum         = _mm256_add_ps(velecsum,velec);
 585
 586             fscal            = felec;
 587
 588             /* Calculate temporary vectorial force */
 589             tx               = _mm256_mul_ps(fscal,dx33);
 590             ty               = _mm256_mul_ps(fscal,dy33);
 591             tz               = _mm256_mul_ps(fscal,dz33);
 592
 593             /* Update vectorial force */
 594             fix3             = _mm256_add_ps(fix3,tx);
 595             fiy3             = _mm256_add_ps(fiy3,ty);
 596             fiz3             = _mm256_add_ps(fiz3,tz);
 597
 598             fjx3             = _mm256_add_ps(fjx3,tx);
 599             fjy3             = _mm256_add_ps(fjy3,ty);
 600             fjz3             = _mm256_add_ps(fjz3,tz);
 601
 602             fjptrA             = f+j_coord_offsetA;
 603             fjptrB             = f+j_coord_offsetB;
 604             fjptrC             = f+j_coord_offsetC;
 605             fjptrD             = f+j_coord_offsetD;
 606             fjptrE             = f+j_coord_offsetE;
 607             fjptrF             = f+j_coord_offsetF;
 608             fjptrG             = f+j_coord_offsetG;
 609             fjptrH             = f+j_coord_offsetH;
 610
 611             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 612                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 613                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 614
 615             /* Inner loop uses 278 flops */
 616         }
 617
 618         if(jidx<j_index_end)
 619         {
 620
 621             /* Get j neighbor index, and coordinate index */
 622             jnrlistA         = jjnr[jidx];
 623             jnrlistB         = jjnr[jidx+1];
 624             jnrlistC         = jjnr[jidx+2];
 625             jnrlistD         = jjnr[jidx+3];
 626             jnrlistE         = jjnr[jidx+4];
 627             jnrlistF         = jjnr[jidx+5];
 628             jnrlistG         = jjnr[jidx+6];
 629             jnrlistH         = jjnr[jidx+7];
 630             /* Sign of each element will be negative for non-real atoms.
 631              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 632              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 633              */
 634             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 635                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 636
 637             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 638             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 639             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 640             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 641             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 642             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 643             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 644             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 645             j_coord_offsetA  = DIM*jnrA;
 646             j_coord_offsetB  = DIM*jnrB;
 647             j_coord_offsetC  = DIM*jnrC;
 648             j_coord_offsetD  = DIM*jnrD;
 649             j_coord_offsetE  = DIM*jnrE;
 650             j_coord_offsetF  = DIM*jnrF;
 651             j_coord_offsetG  = DIM*jnrG;
 652             j_coord_offsetH  = DIM*jnrH;
 653
 654             /* load j atom coordinates */
 655             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 656                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 657                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 658                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 659                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 660                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 661
 662             /* Calculate displacement vector */
 663             dx00             = _mm256_sub_ps(ix0,jx0);
 664             dy00             = _mm256_sub_ps(iy0,jy0);
 665             dz00             = _mm256_sub_ps(iz0,jz0);
 666             dx11             = _mm256_sub_ps(ix1,jx1);
 667             dy11             = _mm256_sub_ps(iy1,jy1);
 668             dz11             = _mm256_sub_ps(iz1,jz1);
 669             dx12             = _mm256_sub_ps(ix1,jx2);
 670             dy12             = _mm256_sub_ps(iy1,jy2);
 671             dz12             = _mm256_sub_ps(iz1,jz2);
 672             dx13             = _mm256_sub_ps(ix1,jx3);
 673             dy13             = _mm256_sub_ps(iy1,jy3);
 674             dz13             = _mm256_sub_ps(iz1,jz3);
 675             dx21             = _mm256_sub_ps(ix2,jx1);
 676             dy21             = _mm256_sub_ps(iy2,jy1);
 677             dz21             = _mm256_sub_ps(iz2,jz1);
 678             dx22             = _mm256_sub_ps(ix2,jx2);
 679             dy22             = _mm256_sub_ps(iy2,jy2);
 680             dz22             = _mm256_sub_ps(iz2,jz2);
 681             dx23             = _mm256_sub_ps(ix2,jx3);
 682             dy23             = _mm256_sub_ps(iy2,jy3);
 683             dz23             = _mm256_sub_ps(iz2,jz3);
 684             dx31             = _mm256_sub_ps(ix3,jx1);
 685             dy31             = _mm256_sub_ps(iy3,jy1);
 686             dz31             = _mm256_sub_ps(iz3,jz1);
 687             dx32             = _mm256_sub_ps(ix3,jx2);
 688             dy32             = _mm256_sub_ps(iy3,jy2);
 689             dz32             = _mm256_sub_ps(iz3,jz2);
 690             dx33             = _mm256_sub_ps(ix3,jx3);
 691             dy33             = _mm256_sub_ps(iy3,jy3);
 692             dz33             = _mm256_sub_ps(iz3,jz3);
 693
 694             /* Calculate squared distance and things based on it */
 695             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 696             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 697             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 698             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 699             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 700             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 701             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 702             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 703             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 704             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 705
 706             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 707             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 708             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 709             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 710             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 711             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 712             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 713             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 714             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 715
 716             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 717             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 718             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 719             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 720             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 721             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 722             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 723             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 724             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 725             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 726
 727             fjx0             = _mm256_setzero_ps();
 728             fjy0             = _mm256_setzero_ps();
 729             fjz0             = _mm256_setzero_ps();
 730             fjx1             = _mm256_setzero_ps();
 731             fjy1             = _mm256_setzero_ps();
 732             fjz1             = _mm256_setzero_ps();
 733             fjx2             = _mm256_setzero_ps();
 734             fjy2             = _mm256_setzero_ps();
 735             fjz2             = _mm256_setzero_ps();
 736             fjx3             = _mm256_setzero_ps();
 737             fjy3             = _mm256_setzero_ps();
 738             fjz3             = _mm256_setzero_ps();
 739
 740             /**************************
 741              * CALCULATE INTERACTIONS *
 742              **************************/
 743
 744             /* LENNARD-JONES DISPERSION/REPULSION */
 745
 746             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 747             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 748             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 749             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 750             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 751
 752             /* Update potential sum for this i atom from the interaction with this j atom. */
 753             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 754             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 755
 756             fscal            = fvdw;
 757
 758             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 759
 760             /* Calculate temporary vectorial force */
 761             tx               = _mm256_mul_ps(fscal,dx00);
 762             ty               = _mm256_mul_ps(fscal,dy00);
 763             tz               = _mm256_mul_ps(fscal,dz00);
 764
 765             /* Update vectorial force */
 766             fix0             = _mm256_add_ps(fix0,tx);
 767             fiy0             = _mm256_add_ps(fiy0,ty);
 768             fiz0             = _mm256_add_ps(fiz0,tz);
 769
 770             fjx0             = _mm256_add_ps(fjx0,tx);
 771             fjy0             = _mm256_add_ps(fjy0,ty);
 772             fjz0             = _mm256_add_ps(fjz0,tz);
 773
 774             /**************************
 775              * CALCULATE INTERACTIONS *
 776              **************************/
 777
 778             /* COULOMB ELECTROSTATICS */
 779             velec            = _mm256_mul_ps(qq11,rinv11);
 780             felec            = _mm256_mul_ps(velec,rinvsq11);
 781
 782             /* Update potential sum for this i atom from the interaction with this j atom. */
 783             velec            = _mm256_andnot_ps(dummy_mask,velec);
 784             velecsum         = _mm256_add_ps(velecsum,velec);
 785
 786             fscal            = felec;
 787
 788             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 789
 790             /* Calculate temporary vectorial force */
 791             tx               = _mm256_mul_ps(fscal,dx11);
 792             ty               = _mm256_mul_ps(fscal,dy11);
 793             tz               = _mm256_mul_ps(fscal,dz11);
 794
 795             /* Update vectorial force */
 796             fix1             = _mm256_add_ps(fix1,tx);
 797             fiy1             = _mm256_add_ps(fiy1,ty);
 798             fiz1             = _mm256_add_ps(fiz1,tz);
 799
 800             fjx1             = _mm256_add_ps(fjx1,tx);
 801             fjy1             = _mm256_add_ps(fjy1,ty);
 802             fjz1             = _mm256_add_ps(fjz1,tz);
 803
 804             /**************************
 805              * CALCULATE INTERACTIONS *
 806              **************************/
 807
 808             /* COULOMB ELECTROSTATICS */
 809             velec            = _mm256_mul_ps(qq12,rinv12);
 810             felec            = _mm256_mul_ps(velec,rinvsq12);
 811
 812             /* Update potential sum for this i atom from the interaction with this j atom. */
 813             velec            = _mm256_andnot_ps(dummy_mask,velec);
 814             velecsum         = _mm256_add_ps(velecsum,velec);
 815
 816             fscal            = felec;
 817
 818             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 819
 820             /* Calculate temporary vectorial force */
 821             tx               = _mm256_mul_ps(fscal,dx12);
 822             ty               = _mm256_mul_ps(fscal,dy12);
 823             tz               = _mm256_mul_ps(fscal,dz12);
 824
 825             /* Update vectorial force */
 826             fix1             = _mm256_add_ps(fix1,tx);
 827             fiy1             = _mm256_add_ps(fiy1,ty);
 828             fiz1             = _mm256_add_ps(fiz1,tz);
 829
 830             fjx2             = _mm256_add_ps(fjx2,tx);
 831             fjy2             = _mm256_add_ps(fjy2,ty);
 832             fjz2             = _mm256_add_ps(fjz2,tz);
 833
 834             /**************************
 835              * CALCULATE INTERACTIONS *
 836              **************************/
 837
 838             /* COULOMB ELECTROSTATICS */
 839             velec            = _mm256_mul_ps(qq13,rinv13);
 840             felec            = _mm256_mul_ps(velec,rinvsq13);
 841
 842             /* Update potential sum for this i atom from the interaction with this j atom. */
 843             velec            = _mm256_andnot_ps(dummy_mask,velec);
 844             velecsum         = _mm256_add_ps(velecsum,velec);
 845
 846             fscal            = felec;
 847
 848             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 849
 850             /* Calculate temporary vectorial force */
 851             tx               = _mm256_mul_ps(fscal,dx13);
 852             ty               = _mm256_mul_ps(fscal,dy13);
 853             tz               = _mm256_mul_ps(fscal,dz13);
 854
 855             /* Update vectorial force */
 856             fix1             = _mm256_add_ps(fix1,tx);
 857             fiy1             = _mm256_add_ps(fiy1,ty);
 858             fiz1             = _mm256_add_ps(fiz1,tz);
 859
 860             fjx3             = _mm256_add_ps(fjx3,tx);
 861             fjy3             = _mm256_add_ps(fjy3,ty);
 862             fjz3             = _mm256_add_ps(fjz3,tz);
 863
 864             /**************************
 865              * CALCULATE INTERACTIONS *
 866              **************************/
 867
 868             /* COULOMB ELECTROSTATICS */
 869             velec            = _mm256_mul_ps(qq21,rinv21);
 870             felec            = _mm256_mul_ps(velec,rinvsq21);
 871
 872             /* Update potential sum for this i atom from the interaction with this j atom. */
 873             velec            = _mm256_andnot_ps(dummy_mask,velec);
 874             velecsum         = _mm256_add_ps(velecsum,velec);
 875
 876             fscal            = felec;
 877
 878             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 879
 880             /* Calculate temporary vectorial force */
 881             tx               = _mm256_mul_ps(fscal,dx21);
 882             ty               = _mm256_mul_ps(fscal,dy21);
 883             tz               = _mm256_mul_ps(fscal,dz21);
 884
 885             /* Update vectorial force */
 886             fix2             = _mm256_add_ps(fix2,tx);
 887             fiy2             = _mm256_add_ps(fiy2,ty);
 888             fiz2             = _mm256_add_ps(fiz2,tz);
 889
 890             fjx1             = _mm256_add_ps(fjx1,tx);
 891             fjy1             = _mm256_add_ps(fjy1,ty);
 892             fjz1             = _mm256_add_ps(fjz1,tz);
 893
 894             /**************************
 895              * CALCULATE INTERACTIONS *
 896              **************************/
 897
 898             /* COULOMB ELECTROSTATICS */
 899             velec            = _mm256_mul_ps(qq22,rinv22);
 900             felec            = _mm256_mul_ps(velec,rinvsq22);
 901
 902             /* Update potential sum for this i atom from the interaction with this j atom. */
 903             velec            = _mm256_andnot_ps(dummy_mask,velec);
 904             velecsum         = _mm256_add_ps(velecsum,velec);
 905
 906             fscal            = felec;
 907
 908             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 909
 910             /* Calculate temporary vectorial force */
 911             tx               = _mm256_mul_ps(fscal,dx22);
 912             ty               = _mm256_mul_ps(fscal,dy22);
 913             tz               = _mm256_mul_ps(fscal,dz22);
 914
 915             /* Update vectorial force */
 916             fix2             = _mm256_add_ps(fix2,tx);
 917             fiy2             = _mm256_add_ps(fiy2,ty);
 918             fiz2             = _mm256_add_ps(fiz2,tz);
 919
 920             fjx2             = _mm256_add_ps(fjx2,tx);
 921             fjy2             = _mm256_add_ps(fjy2,ty);
 922             fjz2             = _mm256_add_ps(fjz2,tz);
 923
 924             /**************************
 925              * CALCULATE INTERACTIONS *
 926              **************************/
 927
 928             /* COULOMB ELECTROSTATICS */
 929             velec            = _mm256_mul_ps(qq23,rinv23);
 930             felec            = _mm256_mul_ps(velec,rinvsq23);
 931
 932             /* Update potential sum for this i atom from the interaction with this j atom. */
 933             velec            = _mm256_andnot_ps(dummy_mask,velec);
 934             velecsum         = _mm256_add_ps(velecsum,velec);
 935
 936             fscal            = felec;
 937
 938             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 939
 940             /* Calculate temporary vectorial force */
 941             tx               = _mm256_mul_ps(fscal,dx23);
 942             ty               = _mm256_mul_ps(fscal,dy23);
 943             tz               = _mm256_mul_ps(fscal,dz23);
 944
 945             /* Update vectorial force */
 946             fix2             = _mm256_add_ps(fix2,tx);
 947             fiy2             = _mm256_add_ps(fiy2,ty);
 948             fiz2             = _mm256_add_ps(fiz2,tz);
 949
 950             fjx3             = _mm256_add_ps(fjx3,tx);
 951             fjy3             = _mm256_add_ps(fjy3,ty);
 952             fjz3             = _mm256_add_ps(fjz3,tz);
 953
 954             /**************************
 955              * CALCULATE INTERACTIONS *
 956              **************************/
 957
 958             /* COULOMB ELECTROSTATICS */
 959             velec            = _mm256_mul_ps(qq31,rinv31);
 960             felec            = _mm256_mul_ps(velec,rinvsq31);
 961
 962             /* Update potential sum for this i atom from the interaction with this j atom. */
 963             velec            = _mm256_andnot_ps(dummy_mask,velec);
 964             velecsum         = _mm256_add_ps(velecsum,velec);
 965
 966             fscal            = felec;
 967
 968             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 969
 970             /* Calculate temporary vectorial force */
 971             tx               = _mm256_mul_ps(fscal,dx31);
 972             ty               = _mm256_mul_ps(fscal,dy31);
 973             tz               = _mm256_mul_ps(fscal,dz31);
 974
 975             /* Update vectorial force */
 976             fix3             = _mm256_add_ps(fix3,tx);
 977             fiy3             = _mm256_add_ps(fiy3,ty);
 978             fiz3             = _mm256_add_ps(fiz3,tz);
 979
 980             fjx1             = _mm256_add_ps(fjx1,tx);
 981             fjy1             = _mm256_add_ps(fjy1,ty);
 982             fjz1             = _mm256_add_ps(fjz1,tz);
 983
 984             /**************************
 985              * CALCULATE INTERACTIONS *
 986              **************************/
 987
 988             /* COULOMB ELECTROSTATICS */
 989             velec            = _mm256_mul_ps(qq32,rinv32);
 990             felec            = _mm256_mul_ps(velec,rinvsq32);
 991
 992             /* Update potential sum for this i atom from the interaction with this j atom. */
 993             velec            = _mm256_andnot_ps(dummy_mask,velec);
 994             velecsum         = _mm256_add_ps(velecsum,velec);
 995
 996             fscal            = felec;
 997
 998             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 999
1000             /* Calculate temporary vectorial force */
1001             tx               = _mm256_mul_ps(fscal,dx32);
1002             ty               = _mm256_mul_ps(fscal,dy32);
1003             tz               = _mm256_mul_ps(fscal,dz32);
1004
1005             /* Update vectorial force */
1006             fix3             = _mm256_add_ps(fix3,tx);
1007             fiy3             = _mm256_add_ps(fiy3,ty);
1008             fiz3             = _mm256_add_ps(fiz3,tz);
1009
1010             fjx2             = _mm256_add_ps(fjx2,tx);
1011             fjy2             = _mm256_add_ps(fjy2,ty);
1012             fjz2             = _mm256_add_ps(fjz2,tz);
1013
1014             /**************************
1015              * CALCULATE INTERACTIONS *
1016              **************************/
1017
1018             /* COULOMB ELECTROSTATICS */
1019             velec            = _mm256_mul_ps(qq33,rinv33);
1020             felec            = _mm256_mul_ps(velec,rinvsq33);
1021
1022             /* Update potential sum for this i atom from the interaction with this j atom. */
1023             velec            = _mm256_andnot_ps(dummy_mask,velec);
1024             velecsum         = _mm256_add_ps(velecsum,velec);
1025
1026             fscal            = felec;
1027
1028             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1029
1030             /* Calculate temporary vectorial force */
1031             tx               = _mm256_mul_ps(fscal,dx33);
1032             ty               = _mm256_mul_ps(fscal,dy33);
1033             tz               = _mm256_mul_ps(fscal,dz33);
1034
1035             /* Update vectorial force */
1036             fix3             = _mm256_add_ps(fix3,tx);
1037             fiy3             = _mm256_add_ps(fiy3,ty);
1038             fiz3             = _mm256_add_ps(fiz3,tz);
1039
1040             fjx3             = _mm256_add_ps(fjx3,tx);
1041             fjy3             = _mm256_add_ps(fjy3,ty);
1042             fjz3             = _mm256_add_ps(fjz3,tz);
1043
1044             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1045             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1046             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1047             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1048             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1049             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1050             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1051             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1052
1053             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1054                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1055                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1056
1057             /* Inner loop uses 278 flops */
1058         }
1059
1060         /* End of innermost loop */
1061
1062         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1063                                                  f+i_coord_offset,fshift+i_shift_offset);
1064
1065         ggid                        = gid[iidx];
1066         /* Update potential energies */
1067         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1068         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1069
1070         /* Increment number of inner iterations */
1071         inneriter                  += j_index_end - j_index_start;
1072
1073         /* Outer loop uses 26 flops */
1074     }
1075
1076     /* Increment number of outer iterations */
1077     outeriter        += nri;
1078
1079     /* Update outer/inner flops */
1080
1081     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*278);
1082 }
1083 /*
1084  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_single
1085  * Electrostatics interaction: Coulomb
1086  * VdW interaction:            LennardJones
1087  * Geometry:                   Water4-Water4
1088  * Calculate force/pot:        Force
1089  */
1090 void
1091 nb_kernel_ElecCoul_VdwLJ_GeomW4W4_F_avx_256_single
1092                     (t_nblist                    * gmx_restrict       nlist,
1093                      rvec                        * gmx_restrict          xx,
1094                      rvec                        * gmx_restrict          ff,
1095                      t_forcerec                  * gmx_restrict          fr,
1096                      t_mdatoms                   * gmx_restrict     mdatoms,
1097                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1098                      t_nrnb                      * gmx_restrict        nrnb)
1099 {
1100     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1101      * just 0 for non-waters.
1102      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1103      * jnr indices corresponding to data put in the four positions in the SIMD register.
1104      */
1105     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1106     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1107     int              jnrA,jnrB,jnrC,jnrD;
1108     int              jnrE,jnrF,jnrG,jnrH;
1109     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1110     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1111     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1112     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1113     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1114     real             rcutoff_scalar;
1115     real             *shiftvec,*fshift,*x,*f;
1116     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1117     real             scratch[4*DIM];
1118     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1119     real *           vdwioffsetptr0;
1120     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1121     real *           vdwioffsetptr1;
1122     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1123     real *           vdwioffsetptr2;
1124     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1125     real *           vdwioffsetptr3;
1126     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1127     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1128     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1129     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1130     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1131     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1132     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1133     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1134     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1135     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1136     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1137     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1138     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1139     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1140     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1141     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1142     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1143     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1144     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1145     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1146     real             *charge;
1147     int              nvdwtype;
1148     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1149     int              *vdwtype;
1150     real             *vdwparam;
1151     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1152     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1153     __m256           dummy_mask,cutoff_mask;
1154     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1155     __m256           one     = _mm256_set1_ps(1.0);
1156     __m256           two     = _mm256_set1_ps(2.0);
1157     x                = xx[0];
1158     f                = ff[0];
1159
1160     nri              = nlist->nri;
1161     iinr             = nlist->iinr;
1162     jindex           = nlist->jindex;
1163     jjnr             = nlist->jjnr;
1164     shiftidx         = nlist->shift;
1165     gid              = nlist->gid;
1166     shiftvec         = fr->shift_vec[0];
1167     fshift           = fr->fshift[0];
1168     facel            = _mm256_set1_ps(fr->epsfac);
1169     charge           = mdatoms->chargeA;
1170     nvdwtype         = fr->ntype;
1171     vdwparam         = fr->nbfp;
1172     vdwtype          = mdatoms->typeA;
1173
1174     /* Setup water-specific parameters */
1175     inr              = nlist->iinr[0];
1176     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1177     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1178     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1179     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1180
1181     jq1              = _mm256_set1_ps(charge[inr+1]);
1182     jq2              = _mm256_set1_ps(charge[inr+2]);
1183     jq3              = _mm256_set1_ps(charge[inr+3]);
1184     vdwjidx0A        = 2*vdwtype[inr+0];
1185     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1186     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1187     qq11             = _mm256_mul_ps(iq1,jq1);
1188     qq12             = _mm256_mul_ps(iq1,jq2);
1189     qq13             = _mm256_mul_ps(iq1,jq3);
1190     qq21             = _mm256_mul_ps(iq2,jq1);
1191     qq22             = _mm256_mul_ps(iq2,jq2);
1192     qq23             = _mm256_mul_ps(iq2,jq3);
1193     qq31             = _mm256_mul_ps(iq3,jq1);
1194     qq32             = _mm256_mul_ps(iq3,jq2);
1195     qq33             = _mm256_mul_ps(iq3,jq3);
1196
1197     /* Avoid stupid compiler warnings */
1198     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1199     j_coord_offsetA = 0;
1200     j_coord_offsetB = 0;
1201     j_coord_offsetC = 0;
1202     j_coord_offsetD = 0;
1203     j_coord_offsetE = 0;
1204     j_coord_offsetF = 0;
1205     j_coord_offsetG = 0;
1206     j_coord_offsetH = 0;
1207
1208     outeriter        = 0;
1209     inneriter        = 0;
1210
1211     for(iidx=0;iidx<4*DIM;iidx++)
1212     {
1213         scratch[iidx] = 0.0;
1214     }
1215
1216     /* Start outer loop over neighborlists */
1217     for(iidx=0; iidx<nri; iidx++)
1218     {
1219         /* Load shift vector for this list */
1220         i_shift_offset   = DIM*shiftidx[iidx];
1221
1222         /* Load limits for loop over neighbors */
1223         j_index_start    = jindex[iidx];
1224         j_index_end      = jindex[iidx+1];
1225
1226         /* Get outer coordinate index */
1227         inr              = iinr[iidx];
1228         i_coord_offset   = DIM*inr;
1229
1230         /* Load i particle coords and add shift vector */
1231         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1232                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1233
1234         fix0             = _mm256_setzero_ps();
1235         fiy0             = _mm256_setzero_ps();
1236         fiz0             = _mm256_setzero_ps();
1237         fix1             = _mm256_setzero_ps();
1238         fiy1             = _mm256_setzero_ps();
1239         fiz1             = _mm256_setzero_ps();
1240         fix2             = _mm256_setzero_ps();
1241         fiy2             = _mm256_setzero_ps();
1242         fiz2             = _mm256_setzero_ps();
1243         fix3             = _mm256_setzero_ps();
1244         fiy3             = _mm256_setzero_ps();
1245         fiz3             = _mm256_setzero_ps();
1246
1247         /* Start inner kernel loop */
1248         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1249         {
1250
1251             /* Get j neighbor index, and coordinate index */
1252             jnrA             = jjnr[jidx];
1253             jnrB             = jjnr[jidx+1];
1254             jnrC             = jjnr[jidx+2];
1255             jnrD             = jjnr[jidx+3];
1256             jnrE             = jjnr[jidx+4];
1257             jnrF             = jjnr[jidx+5];
1258             jnrG             = jjnr[jidx+6];
1259             jnrH             = jjnr[jidx+7];
1260             j_coord_offsetA  = DIM*jnrA;
1261             j_coord_offsetB  = DIM*jnrB;
1262             j_coord_offsetC  = DIM*jnrC;
1263             j_coord_offsetD  = DIM*jnrD;
1264             j_coord_offsetE  = DIM*jnrE;
1265             j_coord_offsetF  = DIM*jnrF;
1266             j_coord_offsetG  = DIM*jnrG;
1267             j_coord_offsetH  = DIM*jnrH;
1268
1269             /* load j atom coordinates */
1270             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1271                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1272                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1273                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1274                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1275                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1276
1277             /* Calculate displacement vector */
1278             dx00             = _mm256_sub_ps(ix0,jx0);
1279             dy00             = _mm256_sub_ps(iy0,jy0);
1280             dz00             = _mm256_sub_ps(iz0,jz0);
1281             dx11             = _mm256_sub_ps(ix1,jx1);
1282             dy11             = _mm256_sub_ps(iy1,jy1);
1283             dz11             = _mm256_sub_ps(iz1,jz1);
1284             dx12             = _mm256_sub_ps(ix1,jx2);
1285             dy12             = _mm256_sub_ps(iy1,jy2);
1286             dz12             = _mm256_sub_ps(iz1,jz2);
1287             dx13             = _mm256_sub_ps(ix1,jx3);
1288             dy13             = _mm256_sub_ps(iy1,jy3);
1289             dz13             = _mm256_sub_ps(iz1,jz3);
1290             dx21             = _mm256_sub_ps(ix2,jx1);
1291             dy21             = _mm256_sub_ps(iy2,jy1);
1292             dz21             = _mm256_sub_ps(iz2,jz1);
1293             dx22             = _mm256_sub_ps(ix2,jx2);
1294             dy22             = _mm256_sub_ps(iy2,jy2);
1295             dz22             = _mm256_sub_ps(iz2,jz2);
1296             dx23             = _mm256_sub_ps(ix2,jx3);
1297             dy23             = _mm256_sub_ps(iy2,jy3);
1298             dz23             = _mm256_sub_ps(iz2,jz3);
1299             dx31             = _mm256_sub_ps(ix3,jx1);
1300             dy31             = _mm256_sub_ps(iy3,jy1);
1301             dz31             = _mm256_sub_ps(iz3,jz1);
1302             dx32             = _mm256_sub_ps(ix3,jx2);
1303             dy32             = _mm256_sub_ps(iy3,jy2);
1304             dz32             = _mm256_sub_ps(iz3,jz2);
1305             dx33             = _mm256_sub_ps(ix3,jx3);
1306             dy33             = _mm256_sub_ps(iy3,jy3);
1307             dz33             = _mm256_sub_ps(iz3,jz3);
1308
1309             /* Calculate squared distance and things based on it */
1310             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1311             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1312             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1313             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1314             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1315             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1316             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1317             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1318             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1319             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1320
1321             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1322             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1323             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1324             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1325             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1326             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1327             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1328             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1329             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1330
1331             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1332             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1333             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1334             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1335             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1336             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1337             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1338             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1339             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1340             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1341
1342             fjx0             = _mm256_setzero_ps();
1343             fjy0             = _mm256_setzero_ps();
1344             fjz0             = _mm256_setzero_ps();
1345             fjx1             = _mm256_setzero_ps();
1346             fjy1             = _mm256_setzero_ps();
1347             fjz1             = _mm256_setzero_ps();
1348             fjx2             = _mm256_setzero_ps();
1349             fjy2             = _mm256_setzero_ps();
1350             fjz2             = _mm256_setzero_ps();
1351             fjx3             = _mm256_setzero_ps();
1352             fjy3             = _mm256_setzero_ps();
1353             fjz3             = _mm256_setzero_ps();
1354
1355             /**************************
1356              * CALCULATE INTERACTIONS *
1357              **************************/
1358
1359             /* LENNARD-JONES DISPERSION/REPULSION */
1360
1361             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1362             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1363
1364             fscal            = fvdw;
1365
1366             /* Calculate temporary vectorial force */
1367             tx               = _mm256_mul_ps(fscal,dx00);
1368             ty               = _mm256_mul_ps(fscal,dy00);
1369             tz               = _mm256_mul_ps(fscal,dz00);
1370
1371             /* Update vectorial force */
1372             fix0             = _mm256_add_ps(fix0,tx);
1373             fiy0             = _mm256_add_ps(fiy0,ty);
1374             fiz0             = _mm256_add_ps(fiz0,tz);
1375
1376             fjx0             = _mm256_add_ps(fjx0,tx);
1377             fjy0             = _mm256_add_ps(fjy0,ty);
1378             fjz0             = _mm256_add_ps(fjz0,tz);
1379
1380             /**************************
1381              * CALCULATE INTERACTIONS *
1382              **************************/
1383
1384             /* COULOMB ELECTROSTATICS */
1385             velec            = _mm256_mul_ps(qq11,rinv11);
1386             felec            = _mm256_mul_ps(velec,rinvsq11);
1387
1388             fscal            = felec;
1389
1390             /* Calculate temporary vectorial force */
1391             tx               = _mm256_mul_ps(fscal,dx11);
1392             ty               = _mm256_mul_ps(fscal,dy11);
1393             tz               = _mm256_mul_ps(fscal,dz11);
1394
1395             /* Update vectorial force */
1396             fix1             = _mm256_add_ps(fix1,tx);
1397             fiy1             = _mm256_add_ps(fiy1,ty);
1398             fiz1             = _mm256_add_ps(fiz1,tz);
1399
1400             fjx1             = _mm256_add_ps(fjx1,tx);
1401             fjy1             = _mm256_add_ps(fjy1,ty);
1402             fjz1             = _mm256_add_ps(fjz1,tz);
1403
1404             /**************************
1405              * CALCULATE INTERACTIONS *
1406              **************************/
1407
1408             /* COULOMB ELECTROSTATICS */
1409             velec            = _mm256_mul_ps(qq12,rinv12);
1410             felec            = _mm256_mul_ps(velec,rinvsq12);
1411
1412             fscal            = felec;
1413
1414             /* Calculate temporary vectorial force */
1415             tx               = _mm256_mul_ps(fscal,dx12);
1416             ty               = _mm256_mul_ps(fscal,dy12);
1417             tz               = _mm256_mul_ps(fscal,dz12);
1418
1419             /* Update vectorial force */
1420             fix1             = _mm256_add_ps(fix1,tx);
1421             fiy1             = _mm256_add_ps(fiy1,ty);
1422             fiz1             = _mm256_add_ps(fiz1,tz);
1423
1424             fjx2             = _mm256_add_ps(fjx2,tx);
1425             fjy2             = _mm256_add_ps(fjy2,ty);
1426             fjz2             = _mm256_add_ps(fjz2,tz);
1427
1428             /**************************
1429              * CALCULATE INTERACTIONS *
1430              **************************/
1431
1432             /* COULOMB ELECTROSTATICS */
1433             velec            = _mm256_mul_ps(qq13,rinv13);
1434             felec            = _mm256_mul_ps(velec,rinvsq13);
1435
1436             fscal            = felec;
1437
1438             /* Calculate temporary vectorial force */
1439             tx               = _mm256_mul_ps(fscal,dx13);
1440             ty               = _mm256_mul_ps(fscal,dy13);
1441             tz               = _mm256_mul_ps(fscal,dz13);
1442
1443             /* Update vectorial force */
1444             fix1             = _mm256_add_ps(fix1,tx);
1445             fiy1             = _mm256_add_ps(fiy1,ty);
1446             fiz1             = _mm256_add_ps(fiz1,tz);
1447
1448             fjx3             = _mm256_add_ps(fjx3,tx);
1449             fjy3             = _mm256_add_ps(fjy3,ty);
1450             fjz3             = _mm256_add_ps(fjz3,tz);
1451
1452             /**************************
1453              * CALCULATE INTERACTIONS *
1454              **************************/
1455
1456             /* COULOMB ELECTROSTATICS */
1457             velec            = _mm256_mul_ps(qq21,rinv21);
1458             felec            = _mm256_mul_ps(velec,rinvsq21);
1459
1460             fscal            = felec;
1461
1462             /* Calculate temporary vectorial force */
1463             tx               = _mm256_mul_ps(fscal,dx21);
1464             ty               = _mm256_mul_ps(fscal,dy21);
1465             tz               = _mm256_mul_ps(fscal,dz21);
1466
1467             /* Update vectorial force */
1468             fix2             = _mm256_add_ps(fix2,tx);
1469             fiy2             = _mm256_add_ps(fiy2,ty);
1470             fiz2             = _mm256_add_ps(fiz2,tz);
1471
1472             fjx1             = _mm256_add_ps(fjx1,tx);
1473             fjy1             = _mm256_add_ps(fjy1,ty);
1474             fjz1             = _mm256_add_ps(fjz1,tz);
1475
1476             /**************************
1477              * CALCULATE INTERACTIONS *
1478              **************************/
1479
1480             /* COULOMB ELECTROSTATICS */
1481             velec            = _mm256_mul_ps(qq22,rinv22);
1482             felec            = _mm256_mul_ps(velec,rinvsq22);
1483
1484             fscal            = felec;
1485
1486             /* Calculate temporary vectorial force */
1487             tx               = _mm256_mul_ps(fscal,dx22);
1488             ty               = _mm256_mul_ps(fscal,dy22);
1489             tz               = _mm256_mul_ps(fscal,dz22);
1490
1491             /* Update vectorial force */
1492             fix2             = _mm256_add_ps(fix2,tx);
1493             fiy2             = _mm256_add_ps(fiy2,ty);
1494             fiz2             = _mm256_add_ps(fiz2,tz);
1495
1496             fjx2             = _mm256_add_ps(fjx2,tx);
1497             fjy2             = _mm256_add_ps(fjy2,ty);
1498             fjz2             = _mm256_add_ps(fjz2,tz);
1499
1500             /**************************
1501              * CALCULATE INTERACTIONS *
1502              **************************/
1503
1504             /* COULOMB ELECTROSTATICS */
1505             velec            = _mm256_mul_ps(qq23,rinv23);
1506             felec            = _mm256_mul_ps(velec,rinvsq23);
1507
1508             fscal            = felec;
1509
1510             /* Calculate temporary vectorial force */
1511             tx               = _mm256_mul_ps(fscal,dx23);
1512             ty               = _mm256_mul_ps(fscal,dy23);
1513             tz               = _mm256_mul_ps(fscal,dz23);
1514
1515             /* Update vectorial force */
1516             fix2             = _mm256_add_ps(fix2,tx);
1517             fiy2             = _mm256_add_ps(fiy2,ty);
1518             fiz2             = _mm256_add_ps(fiz2,tz);
1519
1520             fjx3             = _mm256_add_ps(fjx3,tx);
1521             fjy3             = _mm256_add_ps(fjy3,ty);
1522             fjz3             = _mm256_add_ps(fjz3,tz);
1523
1524             /**************************
1525              * CALCULATE INTERACTIONS *
1526              **************************/
1527
1528             /* COULOMB ELECTROSTATICS */
1529             velec            = _mm256_mul_ps(qq31,rinv31);
1530             felec            = _mm256_mul_ps(velec,rinvsq31);
1531
1532             fscal            = felec;
1533
1534             /* Calculate temporary vectorial force */
1535             tx               = _mm256_mul_ps(fscal,dx31);
1536             ty               = _mm256_mul_ps(fscal,dy31);
1537             tz               = _mm256_mul_ps(fscal,dz31);
1538
1539             /* Update vectorial force */
1540             fix3             = _mm256_add_ps(fix3,tx);
1541             fiy3             = _mm256_add_ps(fiy3,ty);
1542             fiz3             = _mm256_add_ps(fiz3,tz);
1543
1544             fjx1             = _mm256_add_ps(fjx1,tx);
1545             fjy1             = _mm256_add_ps(fjy1,ty);
1546             fjz1             = _mm256_add_ps(fjz1,tz);
1547
1548             /**************************
1549              * CALCULATE INTERACTIONS *
1550              **************************/
1551
1552             /* COULOMB ELECTROSTATICS */
1553             velec            = _mm256_mul_ps(qq32,rinv32);
1554             felec            = _mm256_mul_ps(velec,rinvsq32);
1555
1556             fscal            = felec;
1557
1558             /* Calculate temporary vectorial force */
1559             tx               = _mm256_mul_ps(fscal,dx32);
1560             ty               = _mm256_mul_ps(fscal,dy32);
1561             tz               = _mm256_mul_ps(fscal,dz32);
1562
1563             /* Update vectorial force */
1564             fix3             = _mm256_add_ps(fix3,tx);
1565             fiy3             = _mm256_add_ps(fiy3,ty);
1566             fiz3             = _mm256_add_ps(fiz3,tz);
1567
1568             fjx2             = _mm256_add_ps(fjx2,tx);
1569             fjy2             = _mm256_add_ps(fjy2,ty);
1570             fjz2             = _mm256_add_ps(fjz2,tz);
1571
1572             /**************************
1573              * CALCULATE INTERACTIONS *
1574              **************************/
1575
1576             /* COULOMB ELECTROSTATICS */
1577             velec            = _mm256_mul_ps(qq33,rinv33);
1578             felec            = _mm256_mul_ps(velec,rinvsq33);
1579
1580             fscal            = felec;
1581
1582             /* Calculate temporary vectorial force */
1583             tx               = _mm256_mul_ps(fscal,dx33);
1584             ty               = _mm256_mul_ps(fscal,dy33);
1585             tz               = _mm256_mul_ps(fscal,dz33);
1586
1587             /* Update vectorial force */
1588             fix3             = _mm256_add_ps(fix3,tx);
1589             fiy3             = _mm256_add_ps(fiy3,ty);
1590             fiz3             = _mm256_add_ps(fiz3,tz);
1591
1592             fjx3             = _mm256_add_ps(fjx3,tx);
1593             fjy3             = _mm256_add_ps(fjy3,ty);
1594             fjz3             = _mm256_add_ps(fjz3,tz);
1595
1596             fjptrA             = f+j_coord_offsetA;
1597             fjptrB             = f+j_coord_offsetB;
1598             fjptrC             = f+j_coord_offsetC;
1599             fjptrD             = f+j_coord_offsetD;
1600             fjptrE             = f+j_coord_offsetE;
1601             fjptrF             = f+j_coord_offsetF;
1602             fjptrG             = f+j_coord_offsetG;
1603             fjptrH             = f+j_coord_offsetH;
1604
1605             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1606                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1607                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1608
1609             /* Inner loop uses 264 flops */
1610         }
1611
1612         if(jidx<j_index_end)
1613         {
1614
1615             /* Get j neighbor index, and coordinate index */
1616             jnrlistA         = jjnr[jidx];
1617             jnrlistB         = jjnr[jidx+1];
1618             jnrlistC         = jjnr[jidx+2];
1619             jnrlistD         = jjnr[jidx+3];
1620             jnrlistE         = jjnr[jidx+4];
1621             jnrlistF         = jjnr[jidx+5];
1622             jnrlistG         = jjnr[jidx+6];
1623             jnrlistH         = jjnr[jidx+7];
1624             /* Sign of each element will be negative for non-real atoms.
1625              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1626              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1627              */
1628             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1629                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1630
1631             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1632             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1633             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1634             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1635             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1636             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1637             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1638             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1639             j_coord_offsetA  = DIM*jnrA;
1640             j_coord_offsetB  = DIM*jnrB;
1641             j_coord_offsetC  = DIM*jnrC;
1642             j_coord_offsetD  = DIM*jnrD;
1643             j_coord_offsetE  = DIM*jnrE;
1644             j_coord_offsetF  = DIM*jnrF;
1645             j_coord_offsetG  = DIM*jnrG;
1646             j_coord_offsetH  = DIM*jnrH;
1647
1648             /* load j atom coordinates */
1649             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1650                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1651                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1652                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1653                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1654                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1655
1656             /* Calculate displacement vector */
1657             dx00             = _mm256_sub_ps(ix0,jx0);
1658             dy00             = _mm256_sub_ps(iy0,jy0);
1659             dz00             = _mm256_sub_ps(iz0,jz0);
1660             dx11             = _mm256_sub_ps(ix1,jx1);
1661             dy11             = _mm256_sub_ps(iy1,jy1);
1662             dz11             = _mm256_sub_ps(iz1,jz1);
1663             dx12             = _mm256_sub_ps(ix1,jx2);
1664             dy12             = _mm256_sub_ps(iy1,jy2);
1665             dz12             = _mm256_sub_ps(iz1,jz2);
1666             dx13             = _mm256_sub_ps(ix1,jx3);
1667             dy13             = _mm256_sub_ps(iy1,jy3);
1668             dz13             = _mm256_sub_ps(iz1,jz3);
1669             dx21             = _mm256_sub_ps(ix2,jx1);
1670             dy21             = _mm256_sub_ps(iy2,jy1);
1671             dz21             = _mm256_sub_ps(iz2,jz1);
1672             dx22             = _mm256_sub_ps(ix2,jx2);
1673             dy22             = _mm256_sub_ps(iy2,jy2);
1674             dz22             = _mm256_sub_ps(iz2,jz2);
1675             dx23             = _mm256_sub_ps(ix2,jx3);
1676             dy23             = _mm256_sub_ps(iy2,jy3);
1677             dz23             = _mm256_sub_ps(iz2,jz3);
1678             dx31             = _mm256_sub_ps(ix3,jx1);
1679             dy31             = _mm256_sub_ps(iy3,jy1);
1680             dz31             = _mm256_sub_ps(iz3,jz1);
1681             dx32             = _mm256_sub_ps(ix3,jx2);
1682             dy32             = _mm256_sub_ps(iy3,jy2);
1683             dz32             = _mm256_sub_ps(iz3,jz2);
1684             dx33             = _mm256_sub_ps(ix3,jx3);
1685             dy33             = _mm256_sub_ps(iy3,jy3);
1686             dz33             = _mm256_sub_ps(iz3,jz3);
1687
1688             /* Calculate squared distance and things based on it */
1689             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1690             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1691             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1692             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1693             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1694             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1695             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1696             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1697             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1698             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1699
1700             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1701             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1702             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1703             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1704             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1705             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1706             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1707             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1708             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1709
1710             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1711             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1712             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1713             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1714             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1715             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1716             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1717             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1718             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1719             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1720
1721             fjx0             = _mm256_setzero_ps();
1722             fjy0             = _mm256_setzero_ps();
1723             fjz0             = _mm256_setzero_ps();
1724             fjx1             = _mm256_setzero_ps();
1725             fjy1             = _mm256_setzero_ps();
1726             fjz1             = _mm256_setzero_ps();
1727             fjx2             = _mm256_setzero_ps();
1728             fjy2             = _mm256_setzero_ps();
1729             fjz2             = _mm256_setzero_ps();
1730             fjx3             = _mm256_setzero_ps();
1731             fjy3             = _mm256_setzero_ps();
1732             fjz3             = _mm256_setzero_ps();
1733
1734             /**************************
1735              * CALCULATE INTERACTIONS *
1736              **************************/
1737
1738             /* LENNARD-JONES DISPERSION/REPULSION */
1739
1740             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1741             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1742
1743             fscal            = fvdw;
1744
1745             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1746
1747             /* Calculate temporary vectorial force */
1748             tx               = _mm256_mul_ps(fscal,dx00);
1749             ty               = _mm256_mul_ps(fscal,dy00);
1750             tz               = _mm256_mul_ps(fscal,dz00);
1751
1752             /* Update vectorial force */
1753             fix0             = _mm256_add_ps(fix0,tx);
1754             fiy0             = _mm256_add_ps(fiy0,ty);
1755             fiz0             = _mm256_add_ps(fiz0,tz);
1756
1757             fjx0             = _mm256_add_ps(fjx0,tx);
1758             fjy0             = _mm256_add_ps(fjy0,ty);
1759             fjz0             = _mm256_add_ps(fjz0,tz);
1760
1761             /**************************
1762              * CALCULATE INTERACTIONS *
1763              **************************/
1764
1765             /* COULOMB ELECTROSTATICS */
1766             velec            = _mm256_mul_ps(qq11,rinv11);
1767             felec            = _mm256_mul_ps(velec,rinvsq11);
1768
1769             fscal            = felec;
1770
1771             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1772
1773             /* Calculate temporary vectorial force */
1774             tx               = _mm256_mul_ps(fscal,dx11);
1775             ty               = _mm256_mul_ps(fscal,dy11);
1776             tz               = _mm256_mul_ps(fscal,dz11);
1777
1778             /* Update vectorial force */
1779             fix1             = _mm256_add_ps(fix1,tx);
1780             fiy1             = _mm256_add_ps(fiy1,ty);
1781             fiz1             = _mm256_add_ps(fiz1,tz);
1782
1783             fjx1             = _mm256_add_ps(fjx1,tx);
1784             fjy1             = _mm256_add_ps(fjy1,ty);
1785             fjz1             = _mm256_add_ps(fjz1,tz);
1786
1787             /**************************
1788              * CALCULATE INTERACTIONS *
1789              **************************/
1790
1791             /* COULOMB ELECTROSTATICS */
1792             velec            = _mm256_mul_ps(qq12,rinv12);
1793             felec            = _mm256_mul_ps(velec,rinvsq12);
1794
1795             fscal            = felec;
1796
1797             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1798
1799             /* Calculate temporary vectorial force */
1800             tx               = _mm256_mul_ps(fscal,dx12);
1801             ty               = _mm256_mul_ps(fscal,dy12);
1802             tz               = _mm256_mul_ps(fscal,dz12);
1803
1804             /* Update vectorial force */
1805             fix1             = _mm256_add_ps(fix1,tx);
1806             fiy1             = _mm256_add_ps(fiy1,ty);
1807             fiz1             = _mm256_add_ps(fiz1,tz);
1808
1809             fjx2             = _mm256_add_ps(fjx2,tx);
1810             fjy2             = _mm256_add_ps(fjy2,ty);
1811             fjz2             = _mm256_add_ps(fjz2,tz);
1812
1813             /**************************
1814              * CALCULATE INTERACTIONS *
1815              **************************/
1816
1817             /* COULOMB ELECTROSTATICS */
1818             velec            = _mm256_mul_ps(qq13,rinv13);
1819             felec            = _mm256_mul_ps(velec,rinvsq13);
1820
1821             fscal            = felec;
1822
1823             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1824
1825             /* Calculate temporary vectorial force */
1826             tx               = _mm256_mul_ps(fscal,dx13);
1827             ty               = _mm256_mul_ps(fscal,dy13);
1828             tz               = _mm256_mul_ps(fscal,dz13);
1829
1830             /* Update vectorial force */
1831             fix1             = _mm256_add_ps(fix1,tx);
1832             fiy1             = _mm256_add_ps(fiy1,ty);
1833             fiz1             = _mm256_add_ps(fiz1,tz);
1834
1835             fjx3             = _mm256_add_ps(fjx3,tx);
1836             fjy3             = _mm256_add_ps(fjy3,ty);
1837             fjz3             = _mm256_add_ps(fjz3,tz);
1838
1839             /**************************
1840              * CALCULATE INTERACTIONS *
1841              **************************/
1842
1843             /* COULOMB ELECTROSTATICS */
1844             velec            = _mm256_mul_ps(qq21,rinv21);
1845             felec            = _mm256_mul_ps(velec,rinvsq21);
1846
1847             fscal            = felec;
1848
1849             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1850
1851             /* Calculate temporary vectorial force */
1852             tx               = _mm256_mul_ps(fscal,dx21);
1853             ty               = _mm256_mul_ps(fscal,dy21);
1854             tz               = _mm256_mul_ps(fscal,dz21);
1855
1856             /* Update vectorial force */
1857             fix2             = _mm256_add_ps(fix2,tx);
1858             fiy2             = _mm256_add_ps(fiy2,ty);
1859             fiz2             = _mm256_add_ps(fiz2,tz);
1860
1861             fjx1             = _mm256_add_ps(fjx1,tx);
1862             fjy1             = _mm256_add_ps(fjy1,ty);
1863             fjz1             = _mm256_add_ps(fjz1,tz);
1864
1865             /**************************
1866              * CALCULATE INTERACTIONS *
1867              **************************/
1868
1869             /* COULOMB ELECTROSTATICS */
1870             velec            = _mm256_mul_ps(qq22,rinv22);
1871             felec            = _mm256_mul_ps(velec,rinvsq22);
1872
1873             fscal            = felec;
1874
1875             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1876
1877             /* Calculate temporary vectorial force */
1878             tx               = _mm256_mul_ps(fscal,dx22);
1879             ty               = _mm256_mul_ps(fscal,dy22);
1880             tz               = _mm256_mul_ps(fscal,dz22);
1881
1882             /* Update vectorial force */
1883             fix2             = _mm256_add_ps(fix2,tx);
1884             fiy2             = _mm256_add_ps(fiy2,ty);
1885             fiz2             = _mm256_add_ps(fiz2,tz);
1886
1887             fjx2             = _mm256_add_ps(fjx2,tx);
1888             fjy2             = _mm256_add_ps(fjy2,ty);
1889             fjz2             = _mm256_add_ps(fjz2,tz);
1890
1891             /**************************
1892              * CALCULATE INTERACTIONS *
1893              **************************/
1894
1895             /* COULOMB ELECTROSTATICS */
1896             velec            = _mm256_mul_ps(qq23,rinv23);
1897             felec            = _mm256_mul_ps(velec,rinvsq23);
1898
1899             fscal            = felec;
1900
1901             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1902
1903             /* Calculate temporary vectorial force */
1904             tx               = _mm256_mul_ps(fscal,dx23);
1905             ty               = _mm256_mul_ps(fscal,dy23);
1906             tz               = _mm256_mul_ps(fscal,dz23);
1907
1908             /* Update vectorial force */
1909             fix2             = _mm256_add_ps(fix2,tx);
1910             fiy2             = _mm256_add_ps(fiy2,ty);
1911             fiz2             = _mm256_add_ps(fiz2,tz);
1912
1913             fjx3             = _mm256_add_ps(fjx3,tx);
1914             fjy3             = _mm256_add_ps(fjy3,ty);
1915             fjz3             = _mm256_add_ps(fjz3,tz);
1916
1917             /**************************
1918              * CALCULATE INTERACTIONS *
1919              **************************/
1920
1921             /* COULOMB ELECTROSTATICS */
1922             velec            = _mm256_mul_ps(qq31,rinv31);
1923             felec            = _mm256_mul_ps(velec,rinvsq31);
1924
1925             fscal            = felec;
1926
1927             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1928
1929             /* Calculate temporary vectorial force */
1930             tx               = _mm256_mul_ps(fscal,dx31);
1931             ty               = _mm256_mul_ps(fscal,dy31);
1932             tz               = _mm256_mul_ps(fscal,dz31);
1933
1934             /* Update vectorial force */
1935             fix3             = _mm256_add_ps(fix3,tx);
1936             fiy3             = _mm256_add_ps(fiy3,ty);
1937             fiz3             = _mm256_add_ps(fiz3,tz);
1938
1939             fjx1             = _mm256_add_ps(fjx1,tx);
1940             fjy1             = _mm256_add_ps(fjy1,ty);
1941             fjz1             = _mm256_add_ps(fjz1,tz);
1942
1943             /**************************
1944              * CALCULATE INTERACTIONS *
1945              **************************/
1946
1947             /* COULOMB ELECTROSTATICS */
1948             velec            = _mm256_mul_ps(qq32,rinv32);
1949             felec            = _mm256_mul_ps(velec,rinvsq32);
1950
1951             fscal            = felec;
1952
1953             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1954
1955             /* Calculate temporary vectorial force */
1956             tx               = _mm256_mul_ps(fscal,dx32);
1957             ty               = _mm256_mul_ps(fscal,dy32);
1958             tz               = _mm256_mul_ps(fscal,dz32);
1959
1960             /* Update vectorial force */
1961             fix3             = _mm256_add_ps(fix3,tx);
1962             fiy3             = _mm256_add_ps(fiy3,ty);
1963             fiz3             = _mm256_add_ps(fiz3,tz);
1964
1965             fjx2             = _mm256_add_ps(fjx2,tx);
1966             fjy2             = _mm256_add_ps(fjy2,ty);
1967             fjz2             = _mm256_add_ps(fjz2,tz);
1968
1969             /**************************
1970              * CALCULATE INTERACTIONS *
1971              **************************/
1972
1973             /* COULOMB ELECTROSTATICS */
1974             velec            = _mm256_mul_ps(qq33,rinv33);
1975             felec            = _mm256_mul_ps(velec,rinvsq33);
1976
1977             fscal            = felec;
1978
1979             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1980
1981             /* Calculate temporary vectorial force */
1982             tx               = _mm256_mul_ps(fscal,dx33);
1983             ty               = _mm256_mul_ps(fscal,dy33);
1984             tz               = _mm256_mul_ps(fscal,dz33);
1985
1986             /* Update vectorial force */
1987             fix3             = _mm256_add_ps(fix3,tx);
1988             fiy3             = _mm256_add_ps(fiy3,ty);
1989             fiz3             = _mm256_add_ps(fiz3,tz);
1990
1991             fjx3             = _mm256_add_ps(fjx3,tx);
1992             fjy3             = _mm256_add_ps(fjy3,ty);
1993             fjz3             = _mm256_add_ps(fjz3,tz);
1994
1995             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1996             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1997             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1998             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1999             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2000             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2001             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2002             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2003
2004             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2005                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2006                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2007
2008             /* Inner loop uses 264 flops */
2009         }
2010
2011         /* End of innermost loop */
2012
2013         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2014                                                  f+i_coord_offset,fshift+i_shift_offset);
2015
2016         /* Increment number of inner iterations */
2017         inneriter                  += j_index_end - j_index_start;
2018
2019         /* Outer loop uses 24 flops */
2020     }
2021
2022     /* Increment number of outer iterations */
2023     outeriter        += nri;
2024
2025     /* Update outer/inner flops */
2026
2027     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*264);
2028 }