src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRF_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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_256_single.h"
  48 #include "kernelutil_x86_avx_256_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_avx_256_single
  52  * Electrostatics interaction: ReactionField
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecRF_VdwLJ_GeomW4W4_VF_avx_256_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrE,jnrF,jnrG,jnrH;
  76     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  77     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
  78     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  79     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
  80     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  81     real             rcutoff_scalar;
  82     real             *shiftvec,*fshift,*x,*f;
  83     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
  84     real             scratch[4*DIM];
  85     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  86     real *           vdwioffsetptr0;
  87     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  88     real *           vdwioffsetptr1;
  89     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  90     real *           vdwioffsetptr2;
  91     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  92     real *           vdwioffsetptr3;
  93     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  94     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
  95     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  96     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
  97     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  98     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
  99     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
 100     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
 101     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 102     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 103     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 104     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 105     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 106     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 107     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 108     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 109     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 110     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 111     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 112     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
 113     real             *charge;
 114     int              nvdwtype;
 115     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 116     int              *vdwtype;
 117     real             *vdwparam;
 118     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
 119     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
 120     __m256           dummy_mask,cutoff_mask;
 121     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 122     __m256           one     = _mm256_set1_ps(1.0);
 123     __m256           two     = _mm256_set1_ps(2.0);
 124     x                = xx[0];
 125     f                = ff[0];
 126
 127     nri              = nlist->nri;
 128     iinr             = nlist->iinr;
 129     jindex           = nlist->jindex;
 130     jjnr             = nlist->jjnr;
 131     shiftidx         = nlist->shift;
 132     gid              = nlist->gid;
 133     shiftvec         = fr->shift_vec[0];
 134     fshift           = fr->fshift[0];
 135     facel            = _mm256_set1_ps(fr->epsfac);
 136     charge           = mdatoms->chargeA;
 137     krf              = _mm256_set1_ps(fr->ic->k_rf);
 138     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 139     crf              = _mm256_set1_ps(fr->ic->c_rf);
 140     nvdwtype         = fr->ntype;
 141     vdwparam         = fr->nbfp;
 142     vdwtype          = mdatoms->typeA;
 143
 144     /* Setup water-specific parameters */
 145     inr              = nlist->iinr[0];
 146     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 147     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 148     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 149     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 150
 151     jq1              = _mm256_set1_ps(charge[inr+1]);
 152     jq2              = _mm256_set1_ps(charge[inr+2]);
 153     jq3              = _mm256_set1_ps(charge[inr+3]);
 154     vdwjidx0A        = 2*vdwtype[inr+0];
 155     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 156     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 157     qq11             = _mm256_mul_ps(iq1,jq1);
 158     qq12             = _mm256_mul_ps(iq1,jq2);
 159     qq13             = _mm256_mul_ps(iq1,jq3);
 160     qq21             = _mm256_mul_ps(iq2,jq1);
 161     qq22             = _mm256_mul_ps(iq2,jq2);
 162     qq23             = _mm256_mul_ps(iq2,jq3);
 163     qq31             = _mm256_mul_ps(iq3,jq1);
 164     qq32             = _mm256_mul_ps(iq3,jq2);
 165     qq33             = _mm256_mul_ps(iq3,jq3);
 166
 167     /* Avoid stupid compiler warnings */
 168     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 169     j_coord_offsetA = 0;
 170     j_coord_offsetB = 0;
 171     j_coord_offsetC = 0;
 172     j_coord_offsetD = 0;
 173     j_coord_offsetE = 0;
 174     j_coord_offsetF = 0;
 175     j_coord_offsetG = 0;
 176     j_coord_offsetH = 0;
 177
 178     outeriter        = 0;
 179     inneriter        = 0;
 180
 181     for(iidx=0;iidx<4*DIM;iidx++)
 182     {
 183         scratch[iidx] = 0.0;
 184     }
 185
 186     /* Start outer loop over neighborlists */
 187     for(iidx=0; iidx<nri; iidx++)
 188     {
 189         /* Load shift vector for this list */
 190         i_shift_offset   = DIM*shiftidx[iidx];
 191
 192         /* Load limits for loop over neighbors */
 193         j_index_start    = jindex[iidx];
 194         j_index_end      = jindex[iidx+1];
 195
 196         /* Get outer coordinate index */
 197         inr              = iinr[iidx];
 198         i_coord_offset   = DIM*inr;
 199
 200         /* Load i particle coords and add shift vector */
 201         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 202                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 203
 204         fix0             = _mm256_setzero_ps();
 205         fiy0             = _mm256_setzero_ps();
 206         fiz0             = _mm256_setzero_ps();
 207         fix1             = _mm256_setzero_ps();
 208         fiy1             = _mm256_setzero_ps();
 209         fiz1             = _mm256_setzero_ps();
 210         fix2             = _mm256_setzero_ps();
 211         fiy2             = _mm256_setzero_ps();
 212         fiz2             = _mm256_setzero_ps();
 213         fix3             = _mm256_setzero_ps();
 214         fiy3             = _mm256_setzero_ps();
 215         fiz3             = _mm256_setzero_ps();
 216
 217         /* Reset potential sums */
 218         velecsum         = _mm256_setzero_ps();
 219         vvdwsum          = _mm256_setzero_ps();
 220
 221         /* Start inner kernel loop */
 222         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 223         {
 224
 225             /* Get j neighbor index, and coordinate index */
 226             jnrA             = jjnr[jidx];
 227             jnrB             = jjnr[jidx+1];
 228             jnrC             = jjnr[jidx+2];
 229             jnrD             = jjnr[jidx+3];
 230             jnrE             = jjnr[jidx+4];
 231             jnrF             = jjnr[jidx+5];
 232             jnrG             = jjnr[jidx+6];
 233             jnrH             = jjnr[jidx+7];
 234             j_coord_offsetA  = DIM*jnrA;
 235             j_coord_offsetB  = DIM*jnrB;
 236             j_coord_offsetC  = DIM*jnrC;
 237             j_coord_offsetD  = DIM*jnrD;
 238             j_coord_offsetE  = DIM*jnrE;
 239             j_coord_offsetF  = DIM*jnrF;
 240             j_coord_offsetG  = DIM*jnrG;
 241             j_coord_offsetH  = DIM*jnrH;
 242
 243             /* load j atom coordinates */
 244             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 245                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 246                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 247                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 248                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 249                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 250
 251             /* Calculate displacement vector */
 252             dx00             = _mm256_sub_ps(ix0,jx0);
 253             dy00             = _mm256_sub_ps(iy0,jy0);
 254             dz00             = _mm256_sub_ps(iz0,jz0);
 255             dx11             = _mm256_sub_ps(ix1,jx1);
 256             dy11             = _mm256_sub_ps(iy1,jy1);
 257             dz11             = _mm256_sub_ps(iz1,jz1);
 258             dx12             = _mm256_sub_ps(ix1,jx2);
 259             dy12             = _mm256_sub_ps(iy1,jy2);
 260             dz12             = _mm256_sub_ps(iz1,jz2);
 261             dx13             = _mm256_sub_ps(ix1,jx3);
 262             dy13             = _mm256_sub_ps(iy1,jy3);
 263             dz13             = _mm256_sub_ps(iz1,jz3);
 264             dx21             = _mm256_sub_ps(ix2,jx1);
 265             dy21             = _mm256_sub_ps(iy2,jy1);
 266             dz21             = _mm256_sub_ps(iz2,jz1);
 267             dx22             = _mm256_sub_ps(ix2,jx2);
 268             dy22             = _mm256_sub_ps(iy2,jy2);
 269             dz22             = _mm256_sub_ps(iz2,jz2);
 270             dx23             = _mm256_sub_ps(ix2,jx3);
 271             dy23             = _mm256_sub_ps(iy2,jy3);
 272             dz23             = _mm256_sub_ps(iz2,jz3);
 273             dx31             = _mm256_sub_ps(ix3,jx1);
 274             dy31             = _mm256_sub_ps(iy3,jy1);
 275             dz31             = _mm256_sub_ps(iz3,jz1);
 276             dx32             = _mm256_sub_ps(ix3,jx2);
 277             dy32             = _mm256_sub_ps(iy3,jy2);
 278             dz32             = _mm256_sub_ps(iz3,jz2);
 279             dx33             = _mm256_sub_ps(ix3,jx3);
 280             dy33             = _mm256_sub_ps(iy3,jy3);
 281             dz33             = _mm256_sub_ps(iz3,jz3);
 282
 283             /* Calculate squared distance and things based on it */
 284             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 285             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 286             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 287             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 288             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 289             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 290             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 291             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 292             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 293             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 294
 295             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 296             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 297             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 298             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 299             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 300             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 301             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 302             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 303             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 304
 305             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 306             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 307             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 308             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 309             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 310             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 311             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 312             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 313             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 314             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 315
 316             fjx0             = _mm256_setzero_ps();
 317             fjy0             = _mm256_setzero_ps();
 318             fjz0             = _mm256_setzero_ps();
 319             fjx1             = _mm256_setzero_ps();
 320             fjy1             = _mm256_setzero_ps();
 321             fjz1             = _mm256_setzero_ps();
 322             fjx2             = _mm256_setzero_ps();
 323             fjy2             = _mm256_setzero_ps();
 324             fjz2             = _mm256_setzero_ps();
 325             fjx3             = _mm256_setzero_ps();
 326             fjy3             = _mm256_setzero_ps();
 327             fjz3             = _mm256_setzero_ps();
 328
 329             /**************************
 330              * CALCULATE INTERACTIONS *
 331              **************************/
 332
 333             /* LENNARD-JONES DISPERSION/REPULSION */
 334
 335             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 336             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 337             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 338             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 339             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 340
 341             /* Update potential sum for this i atom from the interaction with this j atom. */
 342             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 343
 344             fscal            = fvdw;
 345
 346             /* Calculate temporary vectorial force */
 347             tx               = _mm256_mul_ps(fscal,dx00);
 348             ty               = _mm256_mul_ps(fscal,dy00);
 349             tz               = _mm256_mul_ps(fscal,dz00);
 350
 351             /* Update vectorial force */
 352             fix0             = _mm256_add_ps(fix0,tx);
 353             fiy0             = _mm256_add_ps(fiy0,ty);
 354             fiz0             = _mm256_add_ps(fiz0,tz);
 355
 356             fjx0             = _mm256_add_ps(fjx0,tx);
 357             fjy0             = _mm256_add_ps(fjy0,ty);
 358             fjz0             = _mm256_add_ps(fjz0,tz);
 359
 360             /**************************
 361              * CALCULATE INTERACTIONS *
 362              **************************/
 363
 364             /* REACTION-FIELD ELECTROSTATICS */
 365             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 366             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 367
 368             /* Update potential sum for this i atom from the interaction with this j atom. */
 369             velecsum         = _mm256_add_ps(velecsum,velec);
 370
 371             fscal            = felec;
 372
 373             /* Calculate temporary vectorial force */
 374             tx               = _mm256_mul_ps(fscal,dx11);
 375             ty               = _mm256_mul_ps(fscal,dy11);
 376             tz               = _mm256_mul_ps(fscal,dz11);
 377
 378             /* Update vectorial force */
 379             fix1             = _mm256_add_ps(fix1,tx);
 380             fiy1             = _mm256_add_ps(fiy1,ty);
 381             fiz1             = _mm256_add_ps(fiz1,tz);
 382
 383             fjx1             = _mm256_add_ps(fjx1,tx);
 384             fjy1             = _mm256_add_ps(fjy1,ty);
 385             fjz1             = _mm256_add_ps(fjz1,tz);
 386
 387             /**************************
 388              * CALCULATE INTERACTIONS *
 389              **************************/
 390
 391             /* REACTION-FIELD ELECTROSTATICS */
 392             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 393             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 394
 395             /* Update potential sum for this i atom from the interaction with this j atom. */
 396             velecsum         = _mm256_add_ps(velecsum,velec);
 397
 398             fscal            = felec;
 399
 400             /* Calculate temporary vectorial force */
 401             tx               = _mm256_mul_ps(fscal,dx12);
 402             ty               = _mm256_mul_ps(fscal,dy12);
 403             tz               = _mm256_mul_ps(fscal,dz12);
 404
 405             /* Update vectorial force */
 406             fix1             = _mm256_add_ps(fix1,tx);
 407             fiy1             = _mm256_add_ps(fiy1,ty);
 408             fiz1             = _mm256_add_ps(fiz1,tz);
 409
 410             fjx2             = _mm256_add_ps(fjx2,tx);
 411             fjy2             = _mm256_add_ps(fjy2,ty);
 412             fjz2             = _mm256_add_ps(fjz2,tz);
 413
 414             /**************************
 415              * CALCULATE INTERACTIONS *
 416              **************************/
 417
 418             /* REACTION-FIELD ELECTROSTATICS */
 419             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 420             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 421
 422             /* Update potential sum for this i atom from the interaction with this j atom. */
 423             velecsum         = _mm256_add_ps(velecsum,velec);
 424
 425             fscal            = felec;
 426
 427             /* Calculate temporary vectorial force */
 428             tx               = _mm256_mul_ps(fscal,dx13);
 429             ty               = _mm256_mul_ps(fscal,dy13);
 430             tz               = _mm256_mul_ps(fscal,dz13);
 431
 432             /* Update vectorial force */
 433             fix1             = _mm256_add_ps(fix1,tx);
 434             fiy1             = _mm256_add_ps(fiy1,ty);
 435             fiz1             = _mm256_add_ps(fiz1,tz);
 436
 437             fjx3             = _mm256_add_ps(fjx3,tx);
 438             fjy3             = _mm256_add_ps(fjy3,ty);
 439             fjz3             = _mm256_add_ps(fjz3,tz);
 440
 441             /**************************
 442              * CALCULATE INTERACTIONS *
 443              **************************/
 444
 445             /* REACTION-FIELD ELECTROSTATICS */
 446             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 447             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 448
 449             /* Update potential sum for this i atom from the interaction with this j atom. */
 450             velecsum         = _mm256_add_ps(velecsum,velec);
 451
 452             fscal            = felec;
 453
 454             /* Calculate temporary vectorial force */
 455             tx               = _mm256_mul_ps(fscal,dx21);
 456             ty               = _mm256_mul_ps(fscal,dy21);
 457             tz               = _mm256_mul_ps(fscal,dz21);
 458
 459             /* Update vectorial force */
 460             fix2             = _mm256_add_ps(fix2,tx);
 461             fiy2             = _mm256_add_ps(fiy2,ty);
 462             fiz2             = _mm256_add_ps(fiz2,tz);
 463
 464             fjx1             = _mm256_add_ps(fjx1,tx);
 465             fjy1             = _mm256_add_ps(fjy1,ty);
 466             fjz1             = _mm256_add_ps(fjz1,tz);
 467
 468             /**************************
 469              * CALCULATE INTERACTIONS *
 470              **************************/
 471
 472             /* REACTION-FIELD ELECTROSTATICS */
 473             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 474             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 475
 476             /* Update potential sum for this i atom from the interaction with this j atom. */
 477             velecsum         = _mm256_add_ps(velecsum,velec);
 478
 479             fscal            = felec;
 480
 481             /* Calculate temporary vectorial force */
 482             tx               = _mm256_mul_ps(fscal,dx22);
 483             ty               = _mm256_mul_ps(fscal,dy22);
 484             tz               = _mm256_mul_ps(fscal,dz22);
 485
 486             /* Update vectorial force */
 487             fix2             = _mm256_add_ps(fix2,tx);
 488             fiy2             = _mm256_add_ps(fiy2,ty);
 489             fiz2             = _mm256_add_ps(fiz2,tz);
 490
 491             fjx2             = _mm256_add_ps(fjx2,tx);
 492             fjy2             = _mm256_add_ps(fjy2,ty);
 493             fjz2             = _mm256_add_ps(fjz2,tz);
 494
 495             /**************************
 496              * CALCULATE INTERACTIONS *
 497              **************************/
 498
 499             /* REACTION-FIELD ELECTROSTATICS */
 500             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 501             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 502
 503             /* Update potential sum for this i atom from the interaction with this j atom. */
 504             velecsum         = _mm256_add_ps(velecsum,velec);
 505
 506             fscal            = felec;
 507
 508             /* Calculate temporary vectorial force */
 509             tx               = _mm256_mul_ps(fscal,dx23);
 510             ty               = _mm256_mul_ps(fscal,dy23);
 511             tz               = _mm256_mul_ps(fscal,dz23);
 512
 513             /* Update vectorial force */
 514             fix2             = _mm256_add_ps(fix2,tx);
 515             fiy2             = _mm256_add_ps(fiy2,ty);
 516             fiz2             = _mm256_add_ps(fiz2,tz);
 517
 518             fjx3             = _mm256_add_ps(fjx3,tx);
 519             fjy3             = _mm256_add_ps(fjy3,ty);
 520             fjz3             = _mm256_add_ps(fjz3,tz);
 521
 522             /**************************
 523              * CALCULATE INTERACTIONS *
 524              **************************/
 525
 526             /* REACTION-FIELD ELECTROSTATICS */
 527             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 528             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 529
 530             /* Update potential sum for this i atom from the interaction with this j atom. */
 531             velecsum         = _mm256_add_ps(velecsum,velec);
 532
 533             fscal            = felec;
 534
 535             /* Calculate temporary vectorial force */
 536             tx               = _mm256_mul_ps(fscal,dx31);
 537             ty               = _mm256_mul_ps(fscal,dy31);
 538             tz               = _mm256_mul_ps(fscal,dz31);
 539
 540             /* Update vectorial force */
 541             fix3             = _mm256_add_ps(fix3,tx);
 542             fiy3             = _mm256_add_ps(fiy3,ty);
 543             fiz3             = _mm256_add_ps(fiz3,tz);
 544
 545             fjx1             = _mm256_add_ps(fjx1,tx);
 546             fjy1             = _mm256_add_ps(fjy1,ty);
 547             fjz1             = _mm256_add_ps(fjz1,tz);
 548
 549             /**************************
 550              * CALCULATE INTERACTIONS *
 551              **************************/
 552
 553             /* REACTION-FIELD ELECTROSTATICS */
 554             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 555             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 556
 557             /* Update potential sum for this i atom from the interaction with this j atom. */
 558             velecsum         = _mm256_add_ps(velecsum,velec);
 559
 560             fscal            = felec;
 561
 562             /* Calculate temporary vectorial force */
 563             tx               = _mm256_mul_ps(fscal,dx32);
 564             ty               = _mm256_mul_ps(fscal,dy32);
 565             tz               = _mm256_mul_ps(fscal,dz32);
 566
 567             /* Update vectorial force */
 568             fix3             = _mm256_add_ps(fix3,tx);
 569             fiy3             = _mm256_add_ps(fiy3,ty);
 570             fiz3             = _mm256_add_ps(fiz3,tz);
 571
 572             fjx2             = _mm256_add_ps(fjx2,tx);
 573             fjy2             = _mm256_add_ps(fjy2,ty);
 574             fjz2             = _mm256_add_ps(fjz2,tz);
 575
 576             /**************************
 577              * CALCULATE INTERACTIONS *
 578              **************************/
 579
 580             /* REACTION-FIELD ELECTROSTATICS */
 581             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 582             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 583
 584             /* Update potential sum for this i atom from the interaction with this j atom. */
 585             velecsum         = _mm256_add_ps(velecsum,velec);
 586
 587             fscal            = felec;
 588
 589             /* Calculate temporary vectorial force */
 590             tx               = _mm256_mul_ps(fscal,dx33);
 591             ty               = _mm256_mul_ps(fscal,dy33);
 592             tz               = _mm256_mul_ps(fscal,dz33);
 593
 594             /* Update vectorial force */
 595             fix3             = _mm256_add_ps(fix3,tx);
 596             fiy3             = _mm256_add_ps(fiy3,ty);
 597             fiz3             = _mm256_add_ps(fiz3,tz);
 598
 599             fjx3             = _mm256_add_ps(fjx3,tx);
 600             fjy3             = _mm256_add_ps(fjy3,ty);
 601             fjz3             = _mm256_add_ps(fjz3,tz);
 602
 603             fjptrA             = f+j_coord_offsetA;
 604             fjptrB             = f+j_coord_offsetB;
 605             fjptrC             = f+j_coord_offsetC;
 606             fjptrD             = f+j_coord_offsetD;
 607             fjptrE             = f+j_coord_offsetE;
 608             fjptrF             = f+j_coord_offsetF;
 609             fjptrG             = f+j_coord_offsetG;
 610             fjptrH             = f+j_coord_offsetH;
 611
 612             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 613                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 614                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 615
 616             /* Inner loop uses 323 flops */
 617         }
 618
 619         if(jidx<j_index_end)
 620         {
 621
 622             /* Get j neighbor index, and coordinate index */
 623             jnrlistA         = jjnr[jidx];
 624             jnrlistB         = jjnr[jidx+1];
 625             jnrlistC         = jjnr[jidx+2];
 626             jnrlistD         = jjnr[jidx+3];
 627             jnrlistE         = jjnr[jidx+4];
 628             jnrlistF         = jjnr[jidx+5];
 629             jnrlistG         = jjnr[jidx+6];
 630             jnrlistH         = jjnr[jidx+7];
 631             /* Sign of each element will be negative for non-real atoms.
 632              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 633              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 634              */
 635             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 636                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 637
 638             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 639             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 640             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 641             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 642             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 643             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 644             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 645             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 646             j_coord_offsetA  = DIM*jnrA;
 647             j_coord_offsetB  = DIM*jnrB;
 648             j_coord_offsetC  = DIM*jnrC;
 649             j_coord_offsetD  = DIM*jnrD;
 650             j_coord_offsetE  = DIM*jnrE;
 651             j_coord_offsetF  = DIM*jnrF;
 652             j_coord_offsetG  = DIM*jnrG;
 653             j_coord_offsetH  = DIM*jnrH;
 654
 655             /* load j atom coordinates */
 656             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 657                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 658                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 659                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 660                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 661                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 662
 663             /* Calculate displacement vector */
 664             dx00             = _mm256_sub_ps(ix0,jx0);
 665             dy00             = _mm256_sub_ps(iy0,jy0);
 666             dz00             = _mm256_sub_ps(iz0,jz0);
 667             dx11             = _mm256_sub_ps(ix1,jx1);
 668             dy11             = _mm256_sub_ps(iy1,jy1);
 669             dz11             = _mm256_sub_ps(iz1,jz1);
 670             dx12             = _mm256_sub_ps(ix1,jx2);
 671             dy12             = _mm256_sub_ps(iy1,jy2);
 672             dz12             = _mm256_sub_ps(iz1,jz2);
 673             dx13             = _mm256_sub_ps(ix1,jx3);
 674             dy13             = _mm256_sub_ps(iy1,jy3);
 675             dz13             = _mm256_sub_ps(iz1,jz3);
 676             dx21             = _mm256_sub_ps(ix2,jx1);
 677             dy21             = _mm256_sub_ps(iy2,jy1);
 678             dz21             = _mm256_sub_ps(iz2,jz1);
 679             dx22             = _mm256_sub_ps(ix2,jx2);
 680             dy22             = _mm256_sub_ps(iy2,jy2);
 681             dz22             = _mm256_sub_ps(iz2,jz2);
 682             dx23             = _mm256_sub_ps(ix2,jx3);
 683             dy23             = _mm256_sub_ps(iy2,jy3);
 684             dz23             = _mm256_sub_ps(iz2,jz3);
 685             dx31             = _mm256_sub_ps(ix3,jx1);
 686             dy31             = _mm256_sub_ps(iy3,jy1);
 687             dz31             = _mm256_sub_ps(iz3,jz1);
 688             dx32             = _mm256_sub_ps(ix3,jx2);
 689             dy32             = _mm256_sub_ps(iy3,jy2);
 690             dz32             = _mm256_sub_ps(iz3,jz2);
 691             dx33             = _mm256_sub_ps(ix3,jx3);
 692             dy33             = _mm256_sub_ps(iy3,jy3);
 693             dz33             = _mm256_sub_ps(iz3,jz3);
 694
 695             /* Calculate squared distance and things based on it */
 696             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 697             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 698             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 699             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 700             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 701             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 702             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 703             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 704             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 705             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 706
 707             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 708             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 709             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 710             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 711             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 712             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 713             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 714             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 715             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 716
 717             rinvsq00         = gmx_mm256_inv_ps(rsq00);
 718             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 719             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 720             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 721             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 722             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 723             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 724             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 725             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 726             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 727
 728             fjx0             = _mm256_setzero_ps();
 729             fjy0             = _mm256_setzero_ps();
 730             fjz0             = _mm256_setzero_ps();
 731             fjx1             = _mm256_setzero_ps();
 732             fjy1             = _mm256_setzero_ps();
 733             fjz1             = _mm256_setzero_ps();
 734             fjx2             = _mm256_setzero_ps();
 735             fjy2             = _mm256_setzero_ps();
 736             fjz2             = _mm256_setzero_ps();
 737             fjx3             = _mm256_setzero_ps();
 738             fjy3             = _mm256_setzero_ps();
 739             fjz3             = _mm256_setzero_ps();
 740
 741             /**************************
 742              * CALCULATE INTERACTIONS *
 743              **************************/
 744
 745             /* LENNARD-JONES DISPERSION/REPULSION */
 746
 747             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 748             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 749             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 750             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 751             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 752
 753             /* Update potential sum for this i atom from the interaction with this j atom. */
 754             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 755             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 756
 757             fscal            = fvdw;
 758
 759             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 760
 761             /* Calculate temporary vectorial force */
 762             tx               = _mm256_mul_ps(fscal,dx00);
 763             ty               = _mm256_mul_ps(fscal,dy00);
 764             tz               = _mm256_mul_ps(fscal,dz00);
 765
 766             /* Update vectorial force */
 767             fix0             = _mm256_add_ps(fix0,tx);
 768             fiy0             = _mm256_add_ps(fiy0,ty);
 769             fiz0             = _mm256_add_ps(fiz0,tz);
 770
 771             fjx0             = _mm256_add_ps(fjx0,tx);
 772             fjy0             = _mm256_add_ps(fjy0,ty);
 773             fjz0             = _mm256_add_ps(fjz0,tz);
 774
 775             /**************************
 776              * CALCULATE INTERACTIONS *
 777              **************************/
 778
 779             /* REACTION-FIELD ELECTROSTATICS */
 780             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 781             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 782
 783             /* Update potential sum for this i atom from the interaction with this j atom. */
 784             velec            = _mm256_andnot_ps(dummy_mask,velec);
 785             velecsum         = _mm256_add_ps(velecsum,velec);
 786
 787             fscal            = felec;
 788
 789             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 790
 791             /* Calculate temporary vectorial force */
 792             tx               = _mm256_mul_ps(fscal,dx11);
 793             ty               = _mm256_mul_ps(fscal,dy11);
 794             tz               = _mm256_mul_ps(fscal,dz11);
 795
 796             /* Update vectorial force */
 797             fix1             = _mm256_add_ps(fix1,tx);
 798             fiy1             = _mm256_add_ps(fiy1,ty);
 799             fiz1             = _mm256_add_ps(fiz1,tz);
 800
 801             fjx1             = _mm256_add_ps(fjx1,tx);
 802             fjy1             = _mm256_add_ps(fjy1,ty);
 803             fjz1             = _mm256_add_ps(fjz1,tz);
 804
 805             /**************************
 806              * CALCULATE INTERACTIONS *
 807              **************************/
 808
 809             /* REACTION-FIELD ELECTROSTATICS */
 810             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 811             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 812
 813             /* Update potential sum for this i atom from the interaction with this j atom. */
 814             velec            = _mm256_andnot_ps(dummy_mask,velec);
 815             velecsum         = _mm256_add_ps(velecsum,velec);
 816
 817             fscal            = felec;
 818
 819             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 820
 821             /* Calculate temporary vectorial force */
 822             tx               = _mm256_mul_ps(fscal,dx12);
 823             ty               = _mm256_mul_ps(fscal,dy12);
 824             tz               = _mm256_mul_ps(fscal,dz12);
 825
 826             /* Update vectorial force */
 827             fix1             = _mm256_add_ps(fix1,tx);
 828             fiy1             = _mm256_add_ps(fiy1,ty);
 829             fiz1             = _mm256_add_ps(fiz1,tz);
 830
 831             fjx2             = _mm256_add_ps(fjx2,tx);
 832             fjy2             = _mm256_add_ps(fjy2,ty);
 833             fjz2             = _mm256_add_ps(fjz2,tz);
 834
 835             /**************************
 836              * CALCULATE INTERACTIONS *
 837              **************************/
 838
 839             /* REACTION-FIELD ELECTROSTATICS */
 840             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 841             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 842
 843             /* Update potential sum for this i atom from the interaction with this j atom. */
 844             velec            = _mm256_andnot_ps(dummy_mask,velec);
 845             velecsum         = _mm256_add_ps(velecsum,velec);
 846
 847             fscal            = felec;
 848
 849             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 850
 851             /* Calculate temporary vectorial force */
 852             tx               = _mm256_mul_ps(fscal,dx13);
 853             ty               = _mm256_mul_ps(fscal,dy13);
 854             tz               = _mm256_mul_ps(fscal,dz13);
 855
 856             /* Update vectorial force */
 857             fix1             = _mm256_add_ps(fix1,tx);
 858             fiy1             = _mm256_add_ps(fiy1,ty);
 859             fiz1             = _mm256_add_ps(fiz1,tz);
 860
 861             fjx3             = _mm256_add_ps(fjx3,tx);
 862             fjy3             = _mm256_add_ps(fjy3,ty);
 863             fjz3             = _mm256_add_ps(fjz3,tz);
 864
 865             /**************************
 866              * CALCULATE INTERACTIONS *
 867              **************************/
 868
 869             /* REACTION-FIELD ELECTROSTATICS */
 870             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 871             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 872
 873             /* Update potential sum for this i atom from the interaction with this j atom. */
 874             velec            = _mm256_andnot_ps(dummy_mask,velec);
 875             velecsum         = _mm256_add_ps(velecsum,velec);
 876
 877             fscal            = felec;
 878
 879             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 880
 881             /* Calculate temporary vectorial force */
 882             tx               = _mm256_mul_ps(fscal,dx21);
 883             ty               = _mm256_mul_ps(fscal,dy21);
 884             tz               = _mm256_mul_ps(fscal,dz21);
 885
 886             /* Update vectorial force */
 887             fix2             = _mm256_add_ps(fix2,tx);
 888             fiy2             = _mm256_add_ps(fiy2,ty);
 889             fiz2             = _mm256_add_ps(fiz2,tz);
 890
 891             fjx1             = _mm256_add_ps(fjx1,tx);
 892             fjy1             = _mm256_add_ps(fjy1,ty);
 893             fjz1             = _mm256_add_ps(fjz1,tz);
 894
 895             /**************************
 896              * CALCULATE INTERACTIONS *
 897              **************************/
 898
 899             /* REACTION-FIELD ELECTROSTATICS */
 900             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 901             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 902
 903             /* Update potential sum for this i atom from the interaction with this j atom. */
 904             velec            = _mm256_andnot_ps(dummy_mask,velec);
 905             velecsum         = _mm256_add_ps(velecsum,velec);
 906
 907             fscal            = felec;
 908
 909             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 910
 911             /* Calculate temporary vectorial force */
 912             tx               = _mm256_mul_ps(fscal,dx22);
 913             ty               = _mm256_mul_ps(fscal,dy22);
 914             tz               = _mm256_mul_ps(fscal,dz22);
 915
 916             /* Update vectorial force */
 917             fix2             = _mm256_add_ps(fix2,tx);
 918             fiy2             = _mm256_add_ps(fiy2,ty);
 919             fiz2             = _mm256_add_ps(fiz2,tz);
 920
 921             fjx2             = _mm256_add_ps(fjx2,tx);
 922             fjy2             = _mm256_add_ps(fjy2,ty);
 923             fjz2             = _mm256_add_ps(fjz2,tz);
 924
 925             /**************************
 926              * CALCULATE INTERACTIONS *
 927              **************************/
 928
 929             /* REACTION-FIELD ELECTROSTATICS */
 930             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 931             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 932
 933             /* Update potential sum for this i atom from the interaction with this j atom. */
 934             velec            = _mm256_andnot_ps(dummy_mask,velec);
 935             velecsum         = _mm256_add_ps(velecsum,velec);
 936
 937             fscal            = felec;
 938
 939             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 940
 941             /* Calculate temporary vectorial force */
 942             tx               = _mm256_mul_ps(fscal,dx23);
 943             ty               = _mm256_mul_ps(fscal,dy23);
 944             tz               = _mm256_mul_ps(fscal,dz23);
 945
 946             /* Update vectorial force */
 947             fix2             = _mm256_add_ps(fix2,tx);
 948             fiy2             = _mm256_add_ps(fiy2,ty);
 949             fiz2             = _mm256_add_ps(fiz2,tz);
 950
 951             fjx3             = _mm256_add_ps(fjx3,tx);
 952             fjy3             = _mm256_add_ps(fjy3,ty);
 953             fjz3             = _mm256_add_ps(fjz3,tz);
 954
 955             /**************************
 956              * CALCULATE INTERACTIONS *
 957              **************************/
 958
 959             /* REACTION-FIELD ELECTROSTATICS */
 960             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 961             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 962
 963             /* Update potential sum for this i atom from the interaction with this j atom. */
 964             velec            = _mm256_andnot_ps(dummy_mask,velec);
 965             velecsum         = _mm256_add_ps(velecsum,velec);
 966
 967             fscal            = felec;
 968
 969             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 970
 971             /* Calculate temporary vectorial force */
 972             tx               = _mm256_mul_ps(fscal,dx31);
 973             ty               = _mm256_mul_ps(fscal,dy31);
 974             tz               = _mm256_mul_ps(fscal,dz31);
 975
 976             /* Update vectorial force */
 977             fix3             = _mm256_add_ps(fix3,tx);
 978             fiy3             = _mm256_add_ps(fiy3,ty);
 979             fiz3             = _mm256_add_ps(fiz3,tz);
 980
 981             fjx1             = _mm256_add_ps(fjx1,tx);
 982             fjy1             = _mm256_add_ps(fjy1,ty);
 983             fjz1             = _mm256_add_ps(fjz1,tz);
 984
 985             /**************************
 986              * CALCULATE INTERACTIONS *
 987              **************************/
 988
 989             /* REACTION-FIELD ELECTROSTATICS */
 990             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 991             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 992
 993             /* Update potential sum for this i atom from the interaction with this j atom. */
 994             velec            = _mm256_andnot_ps(dummy_mask,velec);
 995             velecsum         = _mm256_add_ps(velecsum,velec);
 996
 997             fscal            = felec;
 998
 999             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1000
1001             /* Calculate temporary vectorial force */
1002             tx               = _mm256_mul_ps(fscal,dx32);
1003             ty               = _mm256_mul_ps(fscal,dy32);
1004             tz               = _mm256_mul_ps(fscal,dz32);
1005
1006             /* Update vectorial force */
1007             fix3             = _mm256_add_ps(fix3,tx);
1008             fiy3             = _mm256_add_ps(fiy3,ty);
1009             fiz3             = _mm256_add_ps(fiz3,tz);
1010
1011             fjx2             = _mm256_add_ps(fjx2,tx);
1012             fjy2             = _mm256_add_ps(fjy2,ty);
1013             fjz2             = _mm256_add_ps(fjz2,tz);
1014
1015             /**************************
1016              * CALCULATE INTERACTIONS *
1017              **************************/
1018
1019             /* REACTION-FIELD ELECTROSTATICS */
1020             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
1021             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1022
1023             /* Update potential sum for this i atom from the interaction with this j atom. */
1024             velec            = _mm256_andnot_ps(dummy_mask,velec);
1025             velecsum         = _mm256_add_ps(velecsum,velec);
1026
1027             fscal            = felec;
1028
1029             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1030
1031             /* Calculate temporary vectorial force */
1032             tx               = _mm256_mul_ps(fscal,dx33);
1033             ty               = _mm256_mul_ps(fscal,dy33);
1034             tz               = _mm256_mul_ps(fscal,dz33);
1035
1036             /* Update vectorial force */
1037             fix3             = _mm256_add_ps(fix3,tx);
1038             fiy3             = _mm256_add_ps(fiy3,ty);
1039             fiz3             = _mm256_add_ps(fiz3,tz);
1040
1041             fjx3             = _mm256_add_ps(fjx3,tx);
1042             fjy3             = _mm256_add_ps(fjy3,ty);
1043             fjz3             = _mm256_add_ps(fjz3,tz);
1044
1045             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1046             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1047             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1048             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1049             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1050             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1051             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1052             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1053
1054             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1055                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1056                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1057
1058             /* Inner loop uses 323 flops */
1059         }
1060
1061         /* End of innermost loop */
1062
1063         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1064                                                  f+i_coord_offset,fshift+i_shift_offset);
1065
1066         ggid                        = gid[iidx];
1067         /* Update potential energies */
1068         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1069         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1070
1071         /* Increment number of inner iterations */
1072         inneriter                  += j_index_end - j_index_start;
1073
1074         /* Outer loop uses 26 flops */
1075     }
1076
1077     /* Increment number of outer iterations */
1078     outeriter        += nri;
1079
1080     /* Update outer/inner flops */
1081
1082     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*323);
1083 }
1084 /*
1085  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_single
1086  * Electrostatics interaction: ReactionField
1087  * VdW interaction:            LennardJones
1088  * Geometry:                   Water4-Water4
1089  * Calculate force/pot:        Force
1090  */
1091 void
1092 nb_kernel_ElecRF_VdwLJ_GeomW4W4_F_avx_256_single
1093                     (t_nblist                    * gmx_restrict       nlist,
1094                      rvec                        * gmx_restrict          xx,
1095                      rvec                        * gmx_restrict          ff,
1096                      t_forcerec                  * gmx_restrict          fr,
1097                      t_mdatoms                   * gmx_restrict     mdatoms,
1098                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1099                      t_nrnb                      * gmx_restrict        nrnb)
1100 {
1101     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1102      * just 0 for non-waters.
1103      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1104      * jnr indices corresponding to data put in the four positions in the SIMD register.
1105      */
1106     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1107     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1108     int              jnrA,jnrB,jnrC,jnrD;
1109     int              jnrE,jnrF,jnrG,jnrH;
1110     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1111     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1112     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1113     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1114     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1115     real             rcutoff_scalar;
1116     real             *shiftvec,*fshift,*x,*f;
1117     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1118     real             scratch[4*DIM];
1119     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1120     real *           vdwioffsetptr0;
1121     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1122     real *           vdwioffsetptr1;
1123     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1124     real *           vdwioffsetptr2;
1125     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1126     real *           vdwioffsetptr3;
1127     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1128     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1129     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1130     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1131     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1132     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1133     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1134     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1135     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1136     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1137     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1138     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1139     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1140     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1141     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1142     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1143     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1144     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1145     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1146     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1147     real             *charge;
1148     int              nvdwtype;
1149     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1150     int              *vdwtype;
1151     real             *vdwparam;
1152     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1153     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1154     __m256           dummy_mask,cutoff_mask;
1155     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1156     __m256           one     = _mm256_set1_ps(1.0);
1157     __m256           two     = _mm256_set1_ps(2.0);
1158     x                = xx[0];
1159     f                = ff[0];
1160
1161     nri              = nlist->nri;
1162     iinr             = nlist->iinr;
1163     jindex           = nlist->jindex;
1164     jjnr             = nlist->jjnr;
1165     shiftidx         = nlist->shift;
1166     gid              = nlist->gid;
1167     shiftvec         = fr->shift_vec[0];
1168     fshift           = fr->fshift[0];
1169     facel            = _mm256_set1_ps(fr->epsfac);
1170     charge           = mdatoms->chargeA;
1171     krf              = _mm256_set1_ps(fr->ic->k_rf);
1172     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1173     crf              = _mm256_set1_ps(fr->ic->c_rf);
1174     nvdwtype         = fr->ntype;
1175     vdwparam         = fr->nbfp;
1176     vdwtype          = mdatoms->typeA;
1177
1178     /* Setup water-specific parameters */
1179     inr              = nlist->iinr[0];
1180     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1181     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1182     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1183     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1184
1185     jq1              = _mm256_set1_ps(charge[inr+1]);
1186     jq2              = _mm256_set1_ps(charge[inr+2]);
1187     jq3              = _mm256_set1_ps(charge[inr+3]);
1188     vdwjidx0A        = 2*vdwtype[inr+0];
1189     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1190     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1191     qq11             = _mm256_mul_ps(iq1,jq1);
1192     qq12             = _mm256_mul_ps(iq1,jq2);
1193     qq13             = _mm256_mul_ps(iq1,jq3);
1194     qq21             = _mm256_mul_ps(iq2,jq1);
1195     qq22             = _mm256_mul_ps(iq2,jq2);
1196     qq23             = _mm256_mul_ps(iq2,jq3);
1197     qq31             = _mm256_mul_ps(iq3,jq1);
1198     qq32             = _mm256_mul_ps(iq3,jq2);
1199     qq33             = _mm256_mul_ps(iq3,jq3);
1200
1201     /* Avoid stupid compiler warnings */
1202     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1203     j_coord_offsetA = 0;
1204     j_coord_offsetB = 0;
1205     j_coord_offsetC = 0;
1206     j_coord_offsetD = 0;
1207     j_coord_offsetE = 0;
1208     j_coord_offsetF = 0;
1209     j_coord_offsetG = 0;
1210     j_coord_offsetH = 0;
1211
1212     outeriter        = 0;
1213     inneriter        = 0;
1214
1215     for(iidx=0;iidx<4*DIM;iidx++)
1216     {
1217         scratch[iidx] = 0.0;
1218     }
1219
1220     /* Start outer loop over neighborlists */
1221     for(iidx=0; iidx<nri; iidx++)
1222     {
1223         /* Load shift vector for this list */
1224         i_shift_offset   = DIM*shiftidx[iidx];
1225
1226         /* Load limits for loop over neighbors */
1227         j_index_start    = jindex[iidx];
1228         j_index_end      = jindex[iidx+1];
1229
1230         /* Get outer coordinate index */
1231         inr              = iinr[iidx];
1232         i_coord_offset   = DIM*inr;
1233
1234         /* Load i particle coords and add shift vector */
1235         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1236                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1237
1238         fix0             = _mm256_setzero_ps();
1239         fiy0             = _mm256_setzero_ps();
1240         fiz0             = _mm256_setzero_ps();
1241         fix1             = _mm256_setzero_ps();
1242         fiy1             = _mm256_setzero_ps();
1243         fiz1             = _mm256_setzero_ps();
1244         fix2             = _mm256_setzero_ps();
1245         fiy2             = _mm256_setzero_ps();
1246         fiz2             = _mm256_setzero_ps();
1247         fix3             = _mm256_setzero_ps();
1248         fiy3             = _mm256_setzero_ps();
1249         fiz3             = _mm256_setzero_ps();
1250
1251         /* Start inner kernel loop */
1252         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1253         {
1254
1255             /* Get j neighbor index, and coordinate index */
1256             jnrA             = jjnr[jidx];
1257             jnrB             = jjnr[jidx+1];
1258             jnrC             = jjnr[jidx+2];
1259             jnrD             = jjnr[jidx+3];
1260             jnrE             = jjnr[jidx+4];
1261             jnrF             = jjnr[jidx+5];
1262             jnrG             = jjnr[jidx+6];
1263             jnrH             = jjnr[jidx+7];
1264             j_coord_offsetA  = DIM*jnrA;
1265             j_coord_offsetB  = DIM*jnrB;
1266             j_coord_offsetC  = DIM*jnrC;
1267             j_coord_offsetD  = DIM*jnrD;
1268             j_coord_offsetE  = DIM*jnrE;
1269             j_coord_offsetF  = DIM*jnrF;
1270             j_coord_offsetG  = DIM*jnrG;
1271             j_coord_offsetH  = DIM*jnrH;
1272
1273             /* load j atom coordinates */
1274             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1275                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1276                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1277                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1278                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1279                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1280
1281             /* Calculate displacement vector */
1282             dx00             = _mm256_sub_ps(ix0,jx0);
1283             dy00             = _mm256_sub_ps(iy0,jy0);
1284             dz00             = _mm256_sub_ps(iz0,jz0);
1285             dx11             = _mm256_sub_ps(ix1,jx1);
1286             dy11             = _mm256_sub_ps(iy1,jy1);
1287             dz11             = _mm256_sub_ps(iz1,jz1);
1288             dx12             = _mm256_sub_ps(ix1,jx2);
1289             dy12             = _mm256_sub_ps(iy1,jy2);
1290             dz12             = _mm256_sub_ps(iz1,jz2);
1291             dx13             = _mm256_sub_ps(ix1,jx3);
1292             dy13             = _mm256_sub_ps(iy1,jy3);
1293             dz13             = _mm256_sub_ps(iz1,jz3);
1294             dx21             = _mm256_sub_ps(ix2,jx1);
1295             dy21             = _mm256_sub_ps(iy2,jy1);
1296             dz21             = _mm256_sub_ps(iz2,jz1);
1297             dx22             = _mm256_sub_ps(ix2,jx2);
1298             dy22             = _mm256_sub_ps(iy2,jy2);
1299             dz22             = _mm256_sub_ps(iz2,jz2);
1300             dx23             = _mm256_sub_ps(ix2,jx3);
1301             dy23             = _mm256_sub_ps(iy2,jy3);
1302             dz23             = _mm256_sub_ps(iz2,jz3);
1303             dx31             = _mm256_sub_ps(ix3,jx1);
1304             dy31             = _mm256_sub_ps(iy3,jy1);
1305             dz31             = _mm256_sub_ps(iz3,jz1);
1306             dx32             = _mm256_sub_ps(ix3,jx2);
1307             dy32             = _mm256_sub_ps(iy3,jy2);
1308             dz32             = _mm256_sub_ps(iz3,jz2);
1309             dx33             = _mm256_sub_ps(ix3,jx3);
1310             dy33             = _mm256_sub_ps(iy3,jy3);
1311             dz33             = _mm256_sub_ps(iz3,jz3);
1312
1313             /* Calculate squared distance and things based on it */
1314             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1315             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1316             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1317             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1318             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1319             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1320             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1321             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1322             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1323             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1324
1325             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1326             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1327             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1328             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1329             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1330             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1331             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1332             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1333             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1334
1335             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1336             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1337             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1338             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1339             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1340             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1341             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1342             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1343             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1344             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1345
1346             fjx0             = _mm256_setzero_ps();
1347             fjy0             = _mm256_setzero_ps();
1348             fjz0             = _mm256_setzero_ps();
1349             fjx1             = _mm256_setzero_ps();
1350             fjy1             = _mm256_setzero_ps();
1351             fjz1             = _mm256_setzero_ps();
1352             fjx2             = _mm256_setzero_ps();
1353             fjy2             = _mm256_setzero_ps();
1354             fjz2             = _mm256_setzero_ps();
1355             fjx3             = _mm256_setzero_ps();
1356             fjy3             = _mm256_setzero_ps();
1357             fjz3             = _mm256_setzero_ps();
1358
1359             /**************************
1360              * CALCULATE INTERACTIONS *
1361              **************************/
1362
1363             /* LENNARD-JONES DISPERSION/REPULSION */
1364
1365             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1366             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1367
1368             fscal            = fvdw;
1369
1370             /* Calculate temporary vectorial force */
1371             tx               = _mm256_mul_ps(fscal,dx00);
1372             ty               = _mm256_mul_ps(fscal,dy00);
1373             tz               = _mm256_mul_ps(fscal,dz00);
1374
1375             /* Update vectorial force */
1376             fix0             = _mm256_add_ps(fix0,tx);
1377             fiy0             = _mm256_add_ps(fiy0,ty);
1378             fiz0             = _mm256_add_ps(fiz0,tz);
1379
1380             fjx0             = _mm256_add_ps(fjx0,tx);
1381             fjy0             = _mm256_add_ps(fjy0,ty);
1382             fjz0             = _mm256_add_ps(fjz0,tz);
1383
1384             /**************************
1385              * CALCULATE INTERACTIONS *
1386              **************************/
1387
1388             /* REACTION-FIELD ELECTROSTATICS */
1389             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1390
1391             fscal            = felec;
1392
1393             /* Calculate temporary vectorial force */
1394             tx               = _mm256_mul_ps(fscal,dx11);
1395             ty               = _mm256_mul_ps(fscal,dy11);
1396             tz               = _mm256_mul_ps(fscal,dz11);
1397
1398             /* Update vectorial force */
1399             fix1             = _mm256_add_ps(fix1,tx);
1400             fiy1             = _mm256_add_ps(fiy1,ty);
1401             fiz1             = _mm256_add_ps(fiz1,tz);
1402
1403             fjx1             = _mm256_add_ps(fjx1,tx);
1404             fjy1             = _mm256_add_ps(fjy1,ty);
1405             fjz1             = _mm256_add_ps(fjz1,tz);
1406
1407             /**************************
1408              * CALCULATE INTERACTIONS *
1409              **************************/
1410
1411             /* REACTION-FIELD ELECTROSTATICS */
1412             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1413
1414             fscal            = felec;
1415
1416             /* Calculate temporary vectorial force */
1417             tx               = _mm256_mul_ps(fscal,dx12);
1418             ty               = _mm256_mul_ps(fscal,dy12);
1419             tz               = _mm256_mul_ps(fscal,dz12);
1420
1421             /* Update vectorial force */
1422             fix1             = _mm256_add_ps(fix1,tx);
1423             fiy1             = _mm256_add_ps(fiy1,ty);
1424             fiz1             = _mm256_add_ps(fiz1,tz);
1425
1426             fjx2             = _mm256_add_ps(fjx2,tx);
1427             fjy2             = _mm256_add_ps(fjy2,ty);
1428             fjz2             = _mm256_add_ps(fjz2,tz);
1429
1430             /**************************
1431              * CALCULATE INTERACTIONS *
1432              **************************/
1433
1434             /* REACTION-FIELD ELECTROSTATICS */
1435             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1436
1437             fscal            = felec;
1438
1439             /* Calculate temporary vectorial force */
1440             tx               = _mm256_mul_ps(fscal,dx13);
1441             ty               = _mm256_mul_ps(fscal,dy13);
1442             tz               = _mm256_mul_ps(fscal,dz13);
1443
1444             /* Update vectorial force */
1445             fix1             = _mm256_add_ps(fix1,tx);
1446             fiy1             = _mm256_add_ps(fiy1,ty);
1447             fiz1             = _mm256_add_ps(fiz1,tz);
1448
1449             fjx3             = _mm256_add_ps(fjx3,tx);
1450             fjy3             = _mm256_add_ps(fjy3,ty);
1451             fjz3             = _mm256_add_ps(fjz3,tz);
1452
1453             /**************************
1454              * CALCULATE INTERACTIONS *
1455              **************************/
1456
1457             /* REACTION-FIELD ELECTROSTATICS */
1458             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
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             /* REACTION-FIELD ELECTROSTATICS */
1481             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1482
1483             fscal            = felec;
1484
1485             /* Calculate temporary vectorial force */
1486             tx               = _mm256_mul_ps(fscal,dx22);
1487             ty               = _mm256_mul_ps(fscal,dy22);
1488             tz               = _mm256_mul_ps(fscal,dz22);
1489
1490             /* Update vectorial force */
1491             fix2             = _mm256_add_ps(fix2,tx);
1492             fiy2             = _mm256_add_ps(fiy2,ty);
1493             fiz2             = _mm256_add_ps(fiz2,tz);
1494
1495             fjx2             = _mm256_add_ps(fjx2,tx);
1496             fjy2             = _mm256_add_ps(fjy2,ty);
1497             fjz2             = _mm256_add_ps(fjz2,tz);
1498
1499             /**************************
1500              * CALCULATE INTERACTIONS *
1501              **************************/
1502
1503             /* REACTION-FIELD ELECTROSTATICS */
1504             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1505
1506             fscal            = felec;
1507
1508             /* Calculate temporary vectorial force */
1509             tx               = _mm256_mul_ps(fscal,dx23);
1510             ty               = _mm256_mul_ps(fscal,dy23);
1511             tz               = _mm256_mul_ps(fscal,dz23);
1512
1513             /* Update vectorial force */
1514             fix2             = _mm256_add_ps(fix2,tx);
1515             fiy2             = _mm256_add_ps(fiy2,ty);
1516             fiz2             = _mm256_add_ps(fiz2,tz);
1517
1518             fjx3             = _mm256_add_ps(fjx3,tx);
1519             fjy3             = _mm256_add_ps(fjy3,ty);
1520             fjz3             = _mm256_add_ps(fjz3,tz);
1521
1522             /**************************
1523              * CALCULATE INTERACTIONS *
1524              **************************/
1525
1526             /* REACTION-FIELD ELECTROSTATICS */
1527             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1528
1529             fscal            = felec;
1530
1531             /* Calculate temporary vectorial force */
1532             tx               = _mm256_mul_ps(fscal,dx31);
1533             ty               = _mm256_mul_ps(fscal,dy31);
1534             tz               = _mm256_mul_ps(fscal,dz31);
1535
1536             /* Update vectorial force */
1537             fix3             = _mm256_add_ps(fix3,tx);
1538             fiy3             = _mm256_add_ps(fiy3,ty);
1539             fiz3             = _mm256_add_ps(fiz3,tz);
1540
1541             fjx1             = _mm256_add_ps(fjx1,tx);
1542             fjy1             = _mm256_add_ps(fjy1,ty);
1543             fjz1             = _mm256_add_ps(fjz1,tz);
1544
1545             /**************************
1546              * CALCULATE INTERACTIONS *
1547              **************************/
1548
1549             /* REACTION-FIELD ELECTROSTATICS */
1550             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1551
1552             fscal            = felec;
1553
1554             /* Calculate temporary vectorial force */
1555             tx               = _mm256_mul_ps(fscal,dx32);
1556             ty               = _mm256_mul_ps(fscal,dy32);
1557             tz               = _mm256_mul_ps(fscal,dz32);
1558
1559             /* Update vectorial force */
1560             fix3             = _mm256_add_ps(fix3,tx);
1561             fiy3             = _mm256_add_ps(fiy3,ty);
1562             fiz3             = _mm256_add_ps(fiz3,tz);
1563
1564             fjx2             = _mm256_add_ps(fjx2,tx);
1565             fjy2             = _mm256_add_ps(fjy2,ty);
1566             fjz2             = _mm256_add_ps(fjz2,tz);
1567
1568             /**************************
1569              * CALCULATE INTERACTIONS *
1570              **************************/
1571
1572             /* REACTION-FIELD ELECTROSTATICS */
1573             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1574
1575             fscal            = felec;
1576
1577             /* Calculate temporary vectorial force */
1578             tx               = _mm256_mul_ps(fscal,dx33);
1579             ty               = _mm256_mul_ps(fscal,dy33);
1580             tz               = _mm256_mul_ps(fscal,dz33);
1581
1582             /* Update vectorial force */
1583             fix3             = _mm256_add_ps(fix3,tx);
1584             fiy3             = _mm256_add_ps(fiy3,ty);
1585             fiz3             = _mm256_add_ps(fiz3,tz);
1586
1587             fjx3             = _mm256_add_ps(fjx3,tx);
1588             fjy3             = _mm256_add_ps(fjy3,ty);
1589             fjz3             = _mm256_add_ps(fjz3,tz);
1590
1591             fjptrA             = f+j_coord_offsetA;
1592             fjptrB             = f+j_coord_offsetB;
1593             fjptrC             = f+j_coord_offsetC;
1594             fjptrD             = f+j_coord_offsetD;
1595             fjptrE             = f+j_coord_offsetE;
1596             fjptrF             = f+j_coord_offsetF;
1597             fjptrG             = f+j_coord_offsetG;
1598             fjptrH             = f+j_coord_offsetH;
1599
1600             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1601                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1602                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1603
1604             /* Inner loop uses 273 flops */
1605         }
1606
1607         if(jidx<j_index_end)
1608         {
1609
1610             /* Get j neighbor index, and coordinate index */
1611             jnrlistA         = jjnr[jidx];
1612             jnrlistB         = jjnr[jidx+1];
1613             jnrlistC         = jjnr[jidx+2];
1614             jnrlistD         = jjnr[jidx+3];
1615             jnrlistE         = jjnr[jidx+4];
1616             jnrlistF         = jjnr[jidx+5];
1617             jnrlistG         = jjnr[jidx+6];
1618             jnrlistH         = jjnr[jidx+7];
1619             /* Sign of each element will be negative for non-real atoms.
1620              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1621              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1622              */
1623             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1624                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1625
1626             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1627             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1628             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1629             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1630             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
1631             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
1632             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
1633             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
1634             j_coord_offsetA  = DIM*jnrA;
1635             j_coord_offsetB  = DIM*jnrB;
1636             j_coord_offsetC  = DIM*jnrC;
1637             j_coord_offsetD  = DIM*jnrD;
1638             j_coord_offsetE  = DIM*jnrE;
1639             j_coord_offsetF  = DIM*jnrF;
1640             j_coord_offsetG  = DIM*jnrG;
1641             j_coord_offsetH  = DIM*jnrH;
1642
1643             /* load j atom coordinates */
1644             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1645                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1646                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1647                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1648                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1649                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1650
1651             /* Calculate displacement vector */
1652             dx00             = _mm256_sub_ps(ix0,jx0);
1653             dy00             = _mm256_sub_ps(iy0,jy0);
1654             dz00             = _mm256_sub_ps(iz0,jz0);
1655             dx11             = _mm256_sub_ps(ix1,jx1);
1656             dy11             = _mm256_sub_ps(iy1,jy1);
1657             dz11             = _mm256_sub_ps(iz1,jz1);
1658             dx12             = _mm256_sub_ps(ix1,jx2);
1659             dy12             = _mm256_sub_ps(iy1,jy2);
1660             dz12             = _mm256_sub_ps(iz1,jz2);
1661             dx13             = _mm256_sub_ps(ix1,jx3);
1662             dy13             = _mm256_sub_ps(iy1,jy3);
1663             dz13             = _mm256_sub_ps(iz1,jz3);
1664             dx21             = _mm256_sub_ps(ix2,jx1);
1665             dy21             = _mm256_sub_ps(iy2,jy1);
1666             dz21             = _mm256_sub_ps(iz2,jz1);
1667             dx22             = _mm256_sub_ps(ix2,jx2);
1668             dy22             = _mm256_sub_ps(iy2,jy2);
1669             dz22             = _mm256_sub_ps(iz2,jz2);
1670             dx23             = _mm256_sub_ps(ix2,jx3);
1671             dy23             = _mm256_sub_ps(iy2,jy3);
1672             dz23             = _mm256_sub_ps(iz2,jz3);
1673             dx31             = _mm256_sub_ps(ix3,jx1);
1674             dy31             = _mm256_sub_ps(iy3,jy1);
1675             dz31             = _mm256_sub_ps(iz3,jz1);
1676             dx32             = _mm256_sub_ps(ix3,jx2);
1677             dy32             = _mm256_sub_ps(iy3,jy2);
1678             dz32             = _mm256_sub_ps(iz3,jz2);
1679             dx33             = _mm256_sub_ps(ix3,jx3);
1680             dy33             = _mm256_sub_ps(iy3,jy3);
1681             dz33             = _mm256_sub_ps(iz3,jz3);
1682
1683             /* Calculate squared distance and things based on it */
1684             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1685             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1686             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1687             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1688             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1689             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1690             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1691             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1692             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1693             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1694
1695             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1696             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1697             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1698             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1699             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1700             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1701             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1702             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1703             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1704
1705             rinvsq00         = gmx_mm256_inv_ps(rsq00);
1706             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1707             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1708             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1709             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1710             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1711             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1712             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1713             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1714             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1715
1716             fjx0             = _mm256_setzero_ps();
1717             fjy0             = _mm256_setzero_ps();
1718             fjz0             = _mm256_setzero_ps();
1719             fjx1             = _mm256_setzero_ps();
1720             fjy1             = _mm256_setzero_ps();
1721             fjz1             = _mm256_setzero_ps();
1722             fjx2             = _mm256_setzero_ps();
1723             fjy2             = _mm256_setzero_ps();
1724             fjz2             = _mm256_setzero_ps();
1725             fjx3             = _mm256_setzero_ps();
1726             fjy3             = _mm256_setzero_ps();
1727             fjz3             = _mm256_setzero_ps();
1728
1729             /**************************
1730              * CALCULATE INTERACTIONS *
1731              **************************/
1732
1733             /* LENNARD-JONES DISPERSION/REPULSION */
1734
1735             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1736             fvdw             = _mm256_mul_ps(_mm256_sub_ps(_mm256_mul_ps(c12_00,rinvsix),c6_00),_mm256_mul_ps(rinvsix,rinvsq00));
1737
1738             fscal            = fvdw;
1739
1740             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1741
1742             /* Calculate temporary vectorial force */
1743             tx               = _mm256_mul_ps(fscal,dx00);
1744             ty               = _mm256_mul_ps(fscal,dy00);
1745             tz               = _mm256_mul_ps(fscal,dz00);
1746
1747             /* Update vectorial force */
1748             fix0             = _mm256_add_ps(fix0,tx);
1749             fiy0             = _mm256_add_ps(fiy0,ty);
1750             fiz0             = _mm256_add_ps(fiz0,tz);
1751
1752             fjx0             = _mm256_add_ps(fjx0,tx);
1753             fjy0             = _mm256_add_ps(fjy0,ty);
1754             fjz0             = _mm256_add_ps(fjz0,tz);
1755
1756             /**************************
1757              * CALCULATE INTERACTIONS *
1758              **************************/
1759
1760             /* REACTION-FIELD ELECTROSTATICS */
1761             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1762
1763             fscal            = felec;
1764
1765             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1766
1767             /* Calculate temporary vectorial force */
1768             tx               = _mm256_mul_ps(fscal,dx11);
1769             ty               = _mm256_mul_ps(fscal,dy11);
1770             tz               = _mm256_mul_ps(fscal,dz11);
1771
1772             /* Update vectorial force */
1773             fix1             = _mm256_add_ps(fix1,tx);
1774             fiy1             = _mm256_add_ps(fiy1,ty);
1775             fiz1             = _mm256_add_ps(fiz1,tz);
1776
1777             fjx1             = _mm256_add_ps(fjx1,tx);
1778             fjy1             = _mm256_add_ps(fjy1,ty);
1779             fjz1             = _mm256_add_ps(fjz1,tz);
1780
1781             /**************************
1782              * CALCULATE INTERACTIONS *
1783              **************************/
1784
1785             /* REACTION-FIELD ELECTROSTATICS */
1786             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1787
1788             fscal            = felec;
1789
1790             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1791
1792             /* Calculate temporary vectorial force */
1793             tx               = _mm256_mul_ps(fscal,dx12);
1794             ty               = _mm256_mul_ps(fscal,dy12);
1795             tz               = _mm256_mul_ps(fscal,dz12);
1796
1797             /* Update vectorial force */
1798             fix1             = _mm256_add_ps(fix1,tx);
1799             fiy1             = _mm256_add_ps(fiy1,ty);
1800             fiz1             = _mm256_add_ps(fiz1,tz);
1801
1802             fjx2             = _mm256_add_ps(fjx2,tx);
1803             fjy2             = _mm256_add_ps(fjy2,ty);
1804             fjz2             = _mm256_add_ps(fjz2,tz);
1805
1806             /**************************
1807              * CALCULATE INTERACTIONS *
1808              **************************/
1809
1810             /* REACTION-FIELD ELECTROSTATICS */
1811             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1812
1813             fscal            = felec;
1814
1815             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1816
1817             /* Calculate temporary vectorial force */
1818             tx               = _mm256_mul_ps(fscal,dx13);
1819             ty               = _mm256_mul_ps(fscal,dy13);
1820             tz               = _mm256_mul_ps(fscal,dz13);
1821
1822             /* Update vectorial force */
1823             fix1             = _mm256_add_ps(fix1,tx);
1824             fiy1             = _mm256_add_ps(fiy1,ty);
1825             fiz1             = _mm256_add_ps(fiz1,tz);
1826
1827             fjx3             = _mm256_add_ps(fjx3,tx);
1828             fjy3             = _mm256_add_ps(fjy3,ty);
1829             fjz3             = _mm256_add_ps(fjz3,tz);
1830
1831             /**************************
1832              * CALCULATE INTERACTIONS *
1833              **************************/
1834
1835             /* REACTION-FIELD ELECTROSTATICS */
1836             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1837
1838             fscal            = felec;
1839
1840             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1841
1842             /* Calculate temporary vectorial force */
1843             tx               = _mm256_mul_ps(fscal,dx21);
1844             ty               = _mm256_mul_ps(fscal,dy21);
1845             tz               = _mm256_mul_ps(fscal,dz21);
1846
1847             /* Update vectorial force */
1848             fix2             = _mm256_add_ps(fix2,tx);
1849             fiy2             = _mm256_add_ps(fiy2,ty);
1850             fiz2             = _mm256_add_ps(fiz2,tz);
1851
1852             fjx1             = _mm256_add_ps(fjx1,tx);
1853             fjy1             = _mm256_add_ps(fjy1,ty);
1854             fjz1             = _mm256_add_ps(fjz1,tz);
1855
1856             /**************************
1857              * CALCULATE INTERACTIONS *
1858              **************************/
1859
1860             /* REACTION-FIELD ELECTROSTATICS */
1861             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1862
1863             fscal            = felec;
1864
1865             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1866
1867             /* Calculate temporary vectorial force */
1868             tx               = _mm256_mul_ps(fscal,dx22);
1869             ty               = _mm256_mul_ps(fscal,dy22);
1870             tz               = _mm256_mul_ps(fscal,dz22);
1871
1872             /* Update vectorial force */
1873             fix2             = _mm256_add_ps(fix2,tx);
1874             fiy2             = _mm256_add_ps(fiy2,ty);
1875             fiz2             = _mm256_add_ps(fiz2,tz);
1876
1877             fjx2             = _mm256_add_ps(fjx2,tx);
1878             fjy2             = _mm256_add_ps(fjy2,ty);
1879             fjz2             = _mm256_add_ps(fjz2,tz);
1880
1881             /**************************
1882              * CALCULATE INTERACTIONS *
1883              **************************/
1884
1885             /* REACTION-FIELD ELECTROSTATICS */
1886             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1887
1888             fscal            = felec;
1889
1890             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1891
1892             /* Calculate temporary vectorial force */
1893             tx               = _mm256_mul_ps(fscal,dx23);
1894             ty               = _mm256_mul_ps(fscal,dy23);
1895             tz               = _mm256_mul_ps(fscal,dz23);
1896
1897             /* Update vectorial force */
1898             fix2             = _mm256_add_ps(fix2,tx);
1899             fiy2             = _mm256_add_ps(fiy2,ty);
1900             fiz2             = _mm256_add_ps(fiz2,tz);
1901
1902             fjx3             = _mm256_add_ps(fjx3,tx);
1903             fjy3             = _mm256_add_ps(fjy3,ty);
1904             fjz3             = _mm256_add_ps(fjz3,tz);
1905
1906             /**************************
1907              * CALCULATE INTERACTIONS *
1908              **************************/
1909
1910             /* REACTION-FIELD ELECTROSTATICS */
1911             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1912
1913             fscal            = felec;
1914
1915             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1916
1917             /* Calculate temporary vectorial force */
1918             tx               = _mm256_mul_ps(fscal,dx31);
1919             ty               = _mm256_mul_ps(fscal,dy31);
1920             tz               = _mm256_mul_ps(fscal,dz31);
1921
1922             /* Update vectorial force */
1923             fix3             = _mm256_add_ps(fix3,tx);
1924             fiy3             = _mm256_add_ps(fiy3,ty);
1925             fiz3             = _mm256_add_ps(fiz3,tz);
1926
1927             fjx1             = _mm256_add_ps(fjx1,tx);
1928             fjy1             = _mm256_add_ps(fjy1,ty);
1929             fjz1             = _mm256_add_ps(fjz1,tz);
1930
1931             /**************************
1932              * CALCULATE INTERACTIONS *
1933              **************************/
1934
1935             /* REACTION-FIELD ELECTROSTATICS */
1936             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1937
1938             fscal            = felec;
1939
1940             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1941
1942             /* Calculate temporary vectorial force */
1943             tx               = _mm256_mul_ps(fscal,dx32);
1944             ty               = _mm256_mul_ps(fscal,dy32);
1945             tz               = _mm256_mul_ps(fscal,dz32);
1946
1947             /* Update vectorial force */
1948             fix3             = _mm256_add_ps(fix3,tx);
1949             fiy3             = _mm256_add_ps(fiy3,ty);
1950             fiz3             = _mm256_add_ps(fiz3,tz);
1951
1952             fjx2             = _mm256_add_ps(fjx2,tx);
1953             fjy2             = _mm256_add_ps(fjy2,ty);
1954             fjz2             = _mm256_add_ps(fjz2,tz);
1955
1956             /**************************
1957              * CALCULATE INTERACTIONS *
1958              **************************/
1959
1960             /* REACTION-FIELD ELECTROSTATICS */
1961             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1962
1963             fscal            = felec;
1964
1965             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1966
1967             /* Calculate temporary vectorial force */
1968             tx               = _mm256_mul_ps(fscal,dx33);
1969             ty               = _mm256_mul_ps(fscal,dy33);
1970             tz               = _mm256_mul_ps(fscal,dz33);
1971
1972             /* Update vectorial force */
1973             fix3             = _mm256_add_ps(fix3,tx);
1974             fiy3             = _mm256_add_ps(fiy3,ty);
1975             fiz3             = _mm256_add_ps(fiz3,tz);
1976
1977             fjx3             = _mm256_add_ps(fjx3,tx);
1978             fjy3             = _mm256_add_ps(fjy3,ty);
1979             fjz3             = _mm256_add_ps(fjz3,tz);
1980
1981             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1982             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1983             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1984             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1985             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1986             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1987             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1988             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1989
1990             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1991                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1992                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1993
1994             /* Inner loop uses 273 flops */
1995         }
1996
1997         /* End of innermost loop */
1998
1999         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2000                                                  f+i_coord_offset,fshift+i_shift_offset);
2001
2002         /* Increment number of inner iterations */
2003         inneriter                  += j_index_end - j_index_start;
2004
2005         /* Outer loop uses 24 flops */
2006     }
2007
2008     /* Increment number of outer iterations */
2009     outeriter        += nri;
2010
2011     /* Update outer/inner flops */
2012
2013     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*273);
2014 }