src/gromacs/gmxlib/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecRFCut_VdwLJSw_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 "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/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_ElecRFCut_VdwLJSw_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_ElecRFCut_VdwLJSw_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           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 121     real             rswitch_scalar,d_scalar;
 122     __m256           dummy_mask,cutoff_mask;
 123     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
 124     __m256           one     = _mm256_set1_ps(1.0);
 125     __m256           two     = _mm256_set1_ps(2.0);
 126     x                = xx[0];
 127     f                = ff[0];
 128
 129     nri              = nlist->nri;
 130     iinr             = nlist->iinr;
 131     jindex           = nlist->jindex;
 132     jjnr             = nlist->jjnr;
 133     shiftidx         = nlist->shift;
 134     gid              = nlist->gid;
 135     shiftvec         = fr->shift_vec[0];
 136     fshift           = fr->fshift[0];
 137     facel            = _mm256_set1_ps(fr->epsfac);
 138     charge           = mdatoms->chargeA;
 139     krf              = _mm256_set1_ps(fr->ic->k_rf);
 140     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
 141     crf              = _mm256_set1_ps(fr->ic->c_rf);
 142     nvdwtype         = fr->ntype;
 143     vdwparam         = fr->nbfp;
 144     vdwtype          = mdatoms->typeA;
 145
 146     /* Setup water-specific parameters */
 147     inr              = nlist->iinr[0];
 148     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
 149     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
 150     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
 151     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
 152
 153     jq1              = _mm256_set1_ps(charge[inr+1]);
 154     jq2              = _mm256_set1_ps(charge[inr+2]);
 155     jq3              = _mm256_set1_ps(charge[inr+3]);
 156     vdwjidx0A        = 2*vdwtype[inr+0];
 157     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
 158     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
 159     qq11             = _mm256_mul_ps(iq1,jq1);
 160     qq12             = _mm256_mul_ps(iq1,jq2);
 161     qq13             = _mm256_mul_ps(iq1,jq3);
 162     qq21             = _mm256_mul_ps(iq2,jq1);
 163     qq22             = _mm256_mul_ps(iq2,jq2);
 164     qq23             = _mm256_mul_ps(iq2,jq3);
 165     qq31             = _mm256_mul_ps(iq3,jq1);
 166     qq32             = _mm256_mul_ps(iq3,jq2);
 167     qq33             = _mm256_mul_ps(iq3,jq3);
 168
 169     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 170     rcutoff_scalar   = fr->rcoulomb;
 171     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
 172     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
 173
 174     rswitch_scalar   = fr->rvdw_switch;
 175     rswitch          = _mm256_set1_ps(rswitch_scalar);
 176     /* Setup switch parameters */
 177     d_scalar         = rcutoff_scalar-rswitch_scalar;
 178     d                = _mm256_set1_ps(d_scalar);
 179     swV3             = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
 180     swV4             = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 181     swV5             = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 182     swF2             = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
 183     swF3             = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 184     swF4             = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 185
 186     /* Avoid stupid compiler warnings */
 187     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
 188     j_coord_offsetA = 0;
 189     j_coord_offsetB = 0;
 190     j_coord_offsetC = 0;
 191     j_coord_offsetD = 0;
 192     j_coord_offsetE = 0;
 193     j_coord_offsetF = 0;
 194     j_coord_offsetG = 0;
 195     j_coord_offsetH = 0;
 196
 197     outeriter        = 0;
 198     inneriter        = 0;
 199
 200     for(iidx=0;iidx<4*DIM;iidx++)
 201     {
 202         scratch[iidx] = 0.0;
 203     }
 204
 205     /* Start outer loop over neighborlists */
 206     for(iidx=0; iidx<nri; iidx++)
 207     {
 208         /* Load shift vector for this list */
 209         i_shift_offset   = DIM*shiftidx[iidx];
 210
 211         /* Load limits for loop over neighbors */
 212         j_index_start    = jindex[iidx];
 213         j_index_end      = jindex[iidx+1];
 214
 215         /* Get outer coordinate index */
 216         inr              = iinr[iidx];
 217         i_coord_offset   = DIM*inr;
 218
 219         /* Load i particle coords and add shift vector */
 220         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 221                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 222
 223         fix0             = _mm256_setzero_ps();
 224         fiy0             = _mm256_setzero_ps();
 225         fiz0             = _mm256_setzero_ps();
 226         fix1             = _mm256_setzero_ps();
 227         fiy1             = _mm256_setzero_ps();
 228         fiz1             = _mm256_setzero_ps();
 229         fix2             = _mm256_setzero_ps();
 230         fiy2             = _mm256_setzero_ps();
 231         fiz2             = _mm256_setzero_ps();
 232         fix3             = _mm256_setzero_ps();
 233         fiy3             = _mm256_setzero_ps();
 234         fiz3             = _mm256_setzero_ps();
 235
 236         /* Reset potential sums */
 237         velecsum         = _mm256_setzero_ps();
 238         vvdwsum          = _mm256_setzero_ps();
 239
 240         /* Start inner kernel loop */
 241         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
 242         {
 243
 244             /* Get j neighbor index, and coordinate index */
 245             jnrA             = jjnr[jidx];
 246             jnrB             = jjnr[jidx+1];
 247             jnrC             = jjnr[jidx+2];
 248             jnrD             = jjnr[jidx+3];
 249             jnrE             = jjnr[jidx+4];
 250             jnrF             = jjnr[jidx+5];
 251             jnrG             = jjnr[jidx+6];
 252             jnrH             = jjnr[jidx+7];
 253             j_coord_offsetA  = DIM*jnrA;
 254             j_coord_offsetB  = DIM*jnrB;
 255             j_coord_offsetC  = DIM*jnrC;
 256             j_coord_offsetD  = DIM*jnrD;
 257             j_coord_offsetE  = DIM*jnrE;
 258             j_coord_offsetF  = DIM*jnrF;
 259             j_coord_offsetG  = DIM*jnrG;
 260             j_coord_offsetH  = DIM*jnrH;
 261
 262             /* load j atom coordinates */
 263             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 264                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 265                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 266                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 267                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 268                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 269
 270             /* Calculate displacement vector */
 271             dx00             = _mm256_sub_ps(ix0,jx0);
 272             dy00             = _mm256_sub_ps(iy0,jy0);
 273             dz00             = _mm256_sub_ps(iz0,jz0);
 274             dx11             = _mm256_sub_ps(ix1,jx1);
 275             dy11             = _mm256_sub_ps(iy1,jy1);
 276             dz11             = _mm256_sub_ps(iz1,jz1);
 277             dx12             = _mm256_sub_ps(ix1,jx2);
 278             dy12             = _mm256_sub_ps(iy1,jy2);
 279             dz12             = _mm256_sub_ps(iz1,jz2);
 280             dx13             = _mm256_sub_ps(ix1,jx3);
 281             dy13             = _mm256_sub_ps(iy1,jy3);
 282             dz13             = _mm256_sub_ps(iz1,jz3);
 283             dx21             = _mm256_sub_ps(ix2,jx1);
 284             dy21             = _mm256_sub_ps(iy2,jy1);
 285             dz21             = _mm256_sub_ps(iz2,jz1);
 286             dx22             = _mm256_sub_ps(ix2,jx2);
 287             dy22             = _mm256_sub_ps(iy2,jy2);
 288             dz22             = _mm256_sub_ps(iz2,jz2);
 289             dx23             = _mm256_sub_ps(ix2,jx3);
 290             dy23             = _mm256_sub_ps(iy2,jy3);
 291             dz23             = _mm256_sub_ps(iz2,jz3);
 292             dx31             = _mm256_sub_ps(ix3,jx1);
 293             dy31             = _mm256_sub_ps(iy3,jy1);
 294             dz31             = _mm256_sub_ps(iz3,jz1);
 295             dx32             = _mm256_sub_ps(ix3,jx2);
 296             dy32             = _mm256_sub_ps(iy3,jy2);
 297             dz32             = _mm256_sub_ps(iz3,jz2);
 298             dx33             = _mm256_sub_ps(ix3,jx3);
 299             dy33             = _mm256_sub_ps(iy3,jy3);
 300             dz33             = _mm256_sub_ps(iz3,jz3);
 301
 302             /* Calculate squared distance and things based on it */
 303             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 304             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 305             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 306             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 307             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 308             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 309             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 310             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 311             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 312             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 313
 314             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 315             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 316             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 317             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 318             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 319             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 320             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 321             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 322             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 323             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 324
 325             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 326             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 327             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 328             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 329             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 330             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 331             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 332             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 333             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 334             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 335
 336             fjx0             = _mm256_setzero_ps();
 337             fjy0             = _mm256_setzero_ps();
 338             fjz0             = _mm256_setzero_ps();
 339             fjx1             = _mm256_setzero_ps();
 340             fjy1             = _mm256_setzero_ps();
 341             fjz1             = _mm256_setzero_ps();
 342             fjx2             = _mm256_setzero_ps();
 343             fjy2             = _mm256_setzero_ps();
 344             fjz2             = _mm256_setzero_ps();
 345             fjx3             = _mm256_setzero_ps();
 346             fjy3             = _mm256_setzero_ps();
 347             fjz3             = _mm256_setzero_ps();
 348
 349             /**************************
 350              * CALCULATE INTERACTIONS *
 351              **************************/
 352
 353             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 354             {
 355
 356             r00              = _mm256_mul_ps(rsq00,rinv00);
 357
 358             /* LENNARD-JONES DISPERSION/REPULSION */
 359
 360             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 361             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 362             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 363             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 364             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 365
 366             d                = _mm256_sub_ps(r00,rswitch);
 367             d                = _mm256_max_ps(d,_mm256_setzero_ps());
 368             d2               = _mm256_mul_ps(d,d);
 369             sw               = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
 370
 371             dsw              = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
 372
 373             /* Evaluate switch function */
 374             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 375             fvdw             = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
 376             vvdw             = _mm256_mul_ps(vvdw,sw);
 377             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 378
 379             /* Update potential sum for this i atom from the interaction with this j atom. */
 380             vvdw             = _mm256_and_ps(vvdw,cutoff_mask);
 381             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 382
 383             fscal            = fvdw;
 384
 385             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 386
 387             /* Calculate temporary vectorial force */
 388             tx               = _mm256_mul_ps(fscal,dx00);
 389             ty               = _mm256_mul_ps(fscal,dy00);
 390             tz               = _mm256_mul_ps(fscal,dz00);
 391
 392             /* Update vectorial force */
 393             fix0             = _mm256_add_ps(fix0,tx);
 394             fiy0             = _mm256_add_ps(fiy0,ty);
 395             fiz0             = _mm256_add_ps(fiz0,tz);
 396
 397             fjx0             = _mm256_add_ps(fjx0,tx);
 398             fjy0             = _mm256_add_ps(fjy0,ty);
 399             fjz0             = _mm256_add_ps(fjz0,tz);
 400
 401             }
 402
 403             /**************************
 404              * CALCULATE INTERACTIONS *
 405              **************************/
 406
 407             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 408             {
 409
 410             /* REACTION-FIELD ELECTROSTATICS */
 411             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 412             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 413
 414             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
 415
 416             /* Update potential sum for this i atom from the interaction with this j atom. */
 417             velec            = _mm256_and_ps(velec,cutoff_mask);
 418             velecsum         = _mm256_add_ps(velecsum,velec);
 419
 420             fscal            = felec;
 421
 422             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 423
 424             /* Calculate temporary vectorial force */
 425             tx               = _mm256_mul_ps(fscal,dx11);
 426             ty               = _mm256_mul_ps(fscal,dy11);
 427             tz               = _mm256_mul_ps(fscal,dz11);
 428
 429             /* Update vectorial force */
 430             fix1             = _mm256_add_ps(fix1,tx);
 431             fiy1             = _mm256_add_ps(fiy1,ty);
 432             fiz1             = _mm256_add_ps(fiz1,tz);
 433
 434             fjx1             = _mm256_add_ps(fjx1,tx);
 435             fjy1             = _mm256_add_ps(fjy1,ty);
 436             fjz1             = _mm256_add_ps(fjz1,tz);
 437
 438             }
 439
 440             /**************************
 441              * CALCULATE INTERACTIONS *
 442              **************************/
 443
 444             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 445             {
 446
 447             /* REACTION-FIELD ELECTROSTATICS */
 448             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 449             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 450
 451             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
 452
 453             /* Update potential sum for this i atom from the interaction with this j atom. */
 454             velec            = _mm256_and_ps(velec,cutoff_mask);
 455             velecsum         = _mm256_add_ps(velecsum,velec);
 456
 457             fscal            = felec;
 458
 459             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 460
 461             /* Calculate temporary vectorial force */
 462             tx               = _mm256_mul_ps(fscal,dx12);
 463             ty               = _mm256_mul_ps(fscal,dy12);
 464             tz               = _mm256_mul_ps(fscal,dz12);
 465
 466             /* Update vectorial force */
 467             fix1             = _mm256_add_ps(fix1,tx);
 468             fiy1             = _mm256_add_ps(fiy1,ty);
 469             fiz1             = _mm256_add_ps(fiz1,tz);
 470
 471             fjx2             = _mm256_add_ps(fjx2,tx);
 472             fjy2             = _mm256_add_ps(fjy2,ty);
 473             fjz2             = _mm256_add_ps(fjz2,tz);
 474
 475             }
 476
 477             /**************************
 478              * CALCULATE INTERACTIONS *
 479              **************************/
 480
 481             if (gmx_mm256_any_lt(rsq13,rcutoff2))
 482             {
 483
 484             /* REACTION-FIELD ELECTROSTATICS */
 485             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
 486             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
 487
 488             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
 489
 490             /* Update potential sum for this i atom from the interaction with this j atom. */
 491             velec            = _mm256_and_ps(velec,cutoff_mask);
 492             velecsum         = _mm256_add_ps(velecsum,velec);
 493
 494             fscal            = felec;
 495
 496             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 497
 498             /* Calculate temporary vectorial force */
 499             tx               = _mm256_mul_ps(fscal,dx13);
 500             ty               = _mm256_mul_ps(fscal,dy13);
 501             tz               = _mm256_mul_ps(fscal,dz13);
 502
 503             /* Update vectorial force */
 504             fix1             = _mm256_add_ps(fix1,tx);
 505             fiy1             = _mm256_add_ps(fiy1,ty);
 506             fiz1             = _mm256_add_ps(fiz1,tz);
 507
 508             fjx3             = _mm256_add_ps(fjx3,tx);
 509             fjy3             = _mm256_add_ps(fjy3,ty);
 510             fjz3             = _mm256_add_ps(fjz3,tz);
 511
 512             }
 513
 514             /**************************
 515              * CALCULATE INTERACTIONS *
 516              **************************/
 517
 518             if (gmx_mm256_any_lt(rsq21,rcutoff2))
 519             {
 520
 521             /* REACTION-FIELD ELECTROSTATICS */
 522             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
 523             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
 524
 525             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
 526
 527             /* Update potential sum for this i atom from the interaction with this j atom. */
 528             velec            = _mm256_and_ps(velec,cutoff_mask);
 529             velecsum         = _mm256_add_ps(velecsum,velec);
 530
 531             fscal            = felec;
 532
 533             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 534
 535             /* Calculate temporary vectorial force */
 536             tx               = _mm256_mul_ps(fscal,dx21);
 537             ty               = _mm256_mul_ps(fscal,dy21);
 538             tz               = _mm256_mul_ps(fscal,dz21);
 539
 540             /* Update vectorial force */
 541             fix2             = _mm256_add_ps(fix2,tx);
 542             fiy2             = _mm256_add_ps(fiy2,ty);
 543             fiz2             = _mm256_add_ps(fiz2,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
 551             /**************************
 552              * CALCULATE INTERACTIONS *
 553              **************************/
 554
 555             if (gmx_mm256_any_lt(rsq22,rcutoff2))
 556             {
 557
 558             /* REACTION-FIELD ELECTROSTATICS */
 559             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
 560             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
 561
 562             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
 563
 564             /* Update potential sum for this i atom from the interaction with this j atom. */
 565             velec            = _mm256_and_ps(velec,cutoff_mask);
 566             velecsum         = _mm256_add_ps(velecsum,velec);
 567
 568             fscal            = felec;
 569
 570             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 571
 572             /* Calculate temporary vectorial force */
 573             tx               = _mm256_mul_ps(fscal,dx22);
 574             ty               = _mm256_mul_ps(fscal,dy22);
 575             tz               = _mm256_mul_ps(fscal,dz22);
 576
 577             /* Update vectorial force */
 578             fix2             = _mm256_add_ps(fix2,tx);
 579             fiy2             = _mm256_add_ps(fiy2,ty);
 580             fiz2             = _mm256_add_ps(fiz2,tz);
 581
 582             fjx2             = _mm256_add_ps(fjx2,tx);
 583             fjy2             = _mm256_add_ps(fjy2,ty);
 584             fjz2             = _mm256_add_ps(fjz2,tz);
 585
 586             }
 587
 588             /**************************
 589              * CALCULATE INTERACTIONS *
 590              **************************/
 591
 592             if (gmx_mm256_any_lt(rsq23,rcutoff2))
 593             {
 594
 595             /* REACTION-FIELD ELECTROSTATICS */
 596             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
 597             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
 598
 599             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
 600
 601             /* Update potential sum for this i atom from the interaction with this j atom. */
 602             velec            = _mm256_and_ps(velec,cutoff_mask);
 603             velecsum         = _mm256_add_ps(velecsum,velec);
 604
 605             fscal            = felec;
 606
 607             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 608
 609             /* Calculate temporary vectorial force */
 610             tx               = _mm256_mul_ps(fscal,dx23);
 611             ty               = _mm256_mul_ps(fscal,dy23);
 612             tz               = _mm256_mul_ps(fscal,dz23);
 613
 614             /* Update vectorial force */
 615             fix2             = _mm256_add_ps(fix2,tx);
 616             fiy2             = _mm256_add_ps(fiy2,ty);
 617             fiz2             = _mm256_add_ps(fiz2,tz);
 618
 619             fjx3             = _mm256_add_ps(fjx3,tx);
 620             fjy3             = _mm256_add_ps(fjy3,ty);
 621             fjz3             = _mm256_add_ps(fjz3,tz);
 622
 623             }
 624
 625             /**************************
 626              * CALCULATE INTERACTIONS *
 627              **************************/
 628
 629             if (gmx_mm256_any_lt(rsq31,rcutoff2))
 630             {
 631
 632             /* REACTION-FIELD ELECTROSTATICS */
 633             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
 634             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
 635
 636             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
 637
 638             /* Update potential sum for this i atom from the interaction with this j atom. */
 639             velec            = _mm256_and_ps(velec,cutoff_mask);
 640             velecsum         = _mm256_add_ps(velecsum,velec);
 641
 642             fscal            = felec;
 643
 644             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 645
 646             /* Calculate temporary vectorial force */
 647             tx               = _mm256_mul_ps(fscal,dx31);
 648             ty               = _mm256_mul_ps(fscal,dy31);
 649             tz               = _mm256_mul_ps(fscal,dz31);
 650
 651             /* Update vectorial force */
 652             fix3             = _mm256_add_ps(fix3,tx);
 653             fiy3             = _mm256_add_ps(fiy3,ty);
 654             fiz3             = _mm256_add_ps(fiz3,tz);
 655
 656             fjx1             = _mm256_add_ps(fjx1,tx);
 657             fjy1             = _mm256_add_ps(fjy1,ty);
 658             fjz1             = _mm256_add_ps(fjz1,tz);
 659
 660             }
 661
 662             /**************************
 663              * CALCULATE INTERACTIONS *
 664              **************************/
 665
 666             if (gmx_mm256_any_lt(rsq32,rcutoff2))
 667             {
 668
 669             /* REACTION-FIELD ELECTROSTATICS */
 670             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
 671             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
 672
 673             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
 674
 675             /* Update potential sum for this i atom from the interaction with this j atom. */
 676             velec            = _mm256_and_ps(velec,cutoff_mask);
 677             velecsum         = _mm256_add_ps(velecsum,velec);
 678
 679             fscal            = felec;
 680
 681             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 682
 683             /* Calculate temporary vectorial force */
 684             tx               = _mm256_mul_ps(fscal,dx32);
 685             ty               = _mm256_mul_ps(fscal,dy32);
 686             tz               = _mm256_mul_ps(fscal,dz32);
 687
 688             /* Update vectorial force */
 689             fix3             = _mm256_add_ps(fix3,tx);
 690             fiy3             = _mm256_add_ps(fiy3,ty);
 691             fiz3             = _mm256_add_ps(fiz3,tz);
 692
 693             fjx2             = _mm256_add_ps(fjx2,tx);
 694             fjy2             = _mm256_add_ps(fjy2,ty);
 695             fjz2             = _mm256_add_ps(fjz2,tz);
 696
 697             }
 698
 699             /**************************
 700              * CALCULATE INTERACTIONS *
 701              **************************/
 702
 703             if (gmx_mm256_any_lt(rsq33,rcutoff2))
 704             {
 705
 706             /* REACTION-FIELD ELECTROSTATICS */
 707             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
 708             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
 709
 710             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
 711
 712             /* Update potential sum for this i atom from the interaction with this j atom. */
 713             velec            = _mm256_and_ps(velec,cutoff_mask);
 714             velecsum         = _mm256_add_ps(velecsum,velec);
 715
 716             fscal            = felec;
 717
 718             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 719
 720             /* Calculate temporary vectorial force */
 721             tx               = _mm256_mul_ps(fscal,dx33);
 722             ty               = _mm256_mul_ps(fscal,dy33);
 723             tz               = _mm256_mul_ps(fscal,dz33);
 724
 725             /* Update vectorial force */
 726             fix3             = _mm256_add_ps(fix3,tx);
 727             fiy3             = _mm256_add_ps(fiy3,ty);
 728             fiz3             = _mm256_add_ps(fiz3,tz);
 729
 730             fjx3             = _mm256_add_ps(fjx3,tx);
 731             fjy3             = _mm256_add_ps(fjy3,ty);
 732             fjz3             = _mm256_add_ps(fjz3,tz);
 733
 734             }
 735
 736             fjptrA             = f+j_coord_offsetA;
 737             fjptrB             = f+j_coord_offsetB;
 738             fjptrC             = f+j_coord_offsetC;
 739             fjptrD             = f+j_coord_offsetD;
 740             fjptrE             = f+j_coord_offsetE;
 741             fjptrF             = f+j_coord_offsetF;
 742             fjptrG             = f+j_coord_offsetG;
 743             fjptrH             = f+j_coord_offsetH;
 744
 745             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
 746                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 747                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 748
 749             /* Inner loop uses 386 flops */
 750         }
 751
 752         if(jidx<j_index_end)
 753         {
 754
 755             /* Get j neighbor index, and coordinate index */
 756             jnrlistA         = jjnr[jidx];
 757             jnrlistB         = jjnr[jidx+1];
 758             jnrlistC         = jjnr[jidx+2];
 759             jnrlistD         = jjnr[jidx+3];
 760             jnrlistE         = jjnr[jidx+4];
 761             jnrlistF         = jjnr[jidx+5];
 762             jnrlistG         = jjnr[jidx+6];
 763             jnrlistH         = jjnr[jidx+7];
 764             /* Sign of each element will be negative for non-real atoms.
 765              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 766              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 767              */
 768             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
 769                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
 770
 771             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 772             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 773             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 774             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 775             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
 776             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
 777             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
 778             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
 779             j_coord_offsetA  = DIM*jnrA;
 780             j_coord_offsetB  = DIM*jnrB;
 781             j_coord_offsetC  = DIM*jnrC;
 782             j_coord_offsetD  = DIM*jnrD;
 783             j_coord_offsetE  = DIM*jnrE;
 784             j_coord_offsetF  = DIM*jnrF;
 785             j_coord_offsetG  = DIM*jnrG;
 786             j_coord_offsetH  = DIM*jnrH;
 787
 788             /* load j atom coordinates */
 789             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 790                                                  x+j_coord_offsetC,x+j_coord_offsetD,
 791                                                  x+j_coord_offsetE,x+j_coord_offsetF,
 792                                                  x+j_coord_offsetG,x+j_coord_offsetH,
 793                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 794                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
 795
 796             /* Calculate displacement vector */
 797             dx00             = _mm256_sub_ps(ix0,jx0);
 798             dy00             = _mm256_sub_ps(iy0,jy0);
 799             dz00             = _mm256_sub_ps(iz0,jz0);
 800             dx11             = _mm256_sub_ps(ix1,jx1);
 801             dy11             = _mm256_sub_ps(iy1,jy1);
 802             dz11             = _mm256_sub_ps(iz1,jz1);
 803             dx12             = _mm256_sub_ps(ix1,jx2);
 804             dy12             = _mm256_sub_ps(iy1,jy2);
 805             dz12             = _mm256_sub_ps(iz1,jz2);
 806             dx13             = _mm256_sub_ps(ix1,jx3);
 807             dy13             = _mm256_sub_ps(iy1,jy3);
 808             dz13             = _mm256_sub_ps(iz1,jz3);
 809             dx21             = _mm256_sub_ps(ix2,jx1);
 810             dy21             = _mm256_sub_ps(iy2,jy1);
 811             dz21             = _mm256_sub_ps(iz2,jz1);
 812             dx22             = _mm256_sub_ps(ix2,jx2);
 813             dy22             = _mm256_sub_ps(iy2,jy2);
 814             dz22             = _mm256_sub_ps(iz2,jz2);
 815             dx23             = _mm256_sub_ps(ix2,jx3);
 816             dy23             = _mm256_sub_ps(iy2,jy3);
 817             dz23             = _mm256_sub_ps(iz2,jz3);
 818             dx31             = _mm256_sub_ps(ix3,jx1);
 819             dy31             = _mm256_sub_ps(iy3,jy1);
 820             dz31             = _mm256_sub_ps(iz3,jz1);
 821             dx32             = _mm256_sub_ps(ix3,jx2);
 822             dy32             = _mm256_sub_ps(iy3,jy2);
 823             dz32             = _mm256_sub_ps(iz3,jz2);
 824             dx33             = _mm256_sub_ps(ix3,jx3);
 825             dy33             = _mm256_sub_ps(iy3,jy3);
 826             dz33             = _mm256_sub_ps(iz3,jz3);
 827
 828             /* Calculate squared distance and things based on it */
 829             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
 830             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
 831             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
 832             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
 833             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
 834             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
 835             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
 836             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
 837             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
 838             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
 839
 840             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
 841             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
 842             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
 843             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
 844             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
 845             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
 846             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
 847             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
 848             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
 849             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
 850
 851             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
 852             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
 853             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
 854             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
 855             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
 856             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
 857             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
 858             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
 859             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
 860             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
 861
 862             fjx0             = _mm256_setzero_ps();
 863             fjy0             = _mm256_setzero_ps();
 864             fjz0             = _mm256_setzero_ps();
 865             fjx1             = _mm256_setzero_ps();
 866             fjy1             = _mm256_setzero_ps();
 867             fjz1             = _mm256_setzero_ps();
 868             fjx2             = _mm256_setzero_ps();
 869             fjy2             = _mm256_setzero_ps();
 870             fjz2             = _mm256_setzero_ps();
 871             fjx3             = _mm256_setzero_ps();
 872             fjy3             = _mm256_setzero_ps();
 873             fjz3             = _mm256_setzero_ps();
 874
 875             /**************************
 876              * CALCULATE INTERACTIONS *
 877              **************************/
 878
 879             if (gmx_mm256_any_lt(rsq00,rcutoff2))
 880             {
 881
 882             r00              = _mm256_mul_ps(rsq00,rinv00);
 883             r00              = _mm256_andnot_ps(dummy_mask,r00);
 884
 885             /* LENNARD-JONES DISPERSION/REPULSION */
 886
 887             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 888             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
 889             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
 890             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
 891             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
 892
 893             d                = _mm256_sub_ps(r00,rswitch);
 894             d                = _mm256_max_ps(d,_mm256_setzero_ps());
 895             d2               = _mm256_mul_ps(d,d);
 896             sw               = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
 897
 898             dsw              = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
 899
 900             /* Evaluate switch function */
 901             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 902             fvdw             = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
 903             vvdw             = _mm256_mul_ps(vvdw,sw);
 904             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
 905
 906             /* Update potential sum for this i atom from the interaction with this j atom. */
 907             vvdw             = _mm256_and_ps(vvdw,cutoff_mask);
 908             vvdw             = _mm256_andnot_ps(dummy_mask,vvdw);
 909             vvdwsum          = _mm256_add_ps(vvdwsum,vvdw);
 910
 911             fscal            = fvdw;
 912
 913             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 914
 915             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 916
 917             /* Calculate temporary vectorial force */
 918             tx               = _mm256_mul_ps(fscal,dx00);
 919             ty               = _mm256_mul_ps(fscal,dy00);
 920             tz               = _mm256_mul_ps(fscal,dz00);
 921
 922             /* Update vectorial force */
 923             fix0             = _mm256_add_ps(fix0,tx);
 924             fiy0             = _mm256_add_ps(fiy0,ty);
 925             fiz0             = _mm256_add_ps(fiz0,tz);
 926
 927             fjx0             = _mm256_add_ps(fjx0,tx);
 928             fjy0             = _mm256_add_ps(fjy0,ty);
 929             fjz0             = _mm256_add_ps(fjz0,tz);
 930
 931             }
 932
 933             /**************************
 934              * CALCULATE INTERACTIONS *
 935              **************************/
 936
 937             if (gmx_mm256_any_lt(rsq11,rcutoff2))
 938             {
 939
 940             /* REACTION-FIELD ELECTROSTATICS */
 941             velec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_add_ps(rinv11,_mm256_mul_ps(krf,rsq11)),crf));
 942             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
 943
 944             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
 945
 946             /* Update potential sum for this i atom from the interaction with this j atom. */
 947             velec            = _mm256_and_ps(velec,cutoff_mask);
 948             velec            = _mm256_andnot_ps(dummy_mask,velec);
 949             velecsum         = _mm256_add_ps(velecsum,velec);
 950
 951             fscal            = felec;
 952
 953             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 954
 955             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 956
 957             /* Calculate temporary vectorial force */
 958             tx               = _mm256_mul_ps(fscal,dx11);
 959             ty               = _mm256_mul_ps(fscal,dy11);
 960             tz               = _mm256_mul_ps(fscal,dz11);
 961
 962             /* Update vectorial force */
 963             fix1             = _mm256_add_ps(fix1,tx);
 964             fiy1             = _mm256_add_ps(fiy1,ty);
 965             fiz1             = _mm256_add_ps(fiz1,tz);
 966
 967             fjx1             = _mm256_add_ps(fjx1,tx);
 968             fjy1             = _mm256_add_ps(fjy1,ty);
 969             fjz1             = _mm256_add_ps(fjz1,tz);
 970
 971             }
 972
 973             /**************************
 974              * CALCULATE INTERACTIONS *
 975              **************************/
 976
 977             if (gmx_mm256_any_lt(rsq12,rcutoff2))
 978             {
 979
 980             /* REACTION-FIELD ELECTROSTATICS */
 981             velec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_add_ps(rinv12,_mm256_mul_ps(krf,rsq12)),crf));
 982             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
 983
 984             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
 985
 986             /* Update potential sum for this i atom from the interaction with this j atom. */
 987             velec            = _mm256_and_ps(velec,cutoff_mask);
 988             velec            = _mm256_andnot_ps(dummy_mask,velec);
 989             velecsum         = _mm256_add_ps(velecsum,velec);
 990
 991             fscal            = felec;
 992
 993             fscal            = _mm256_and_ps(fscal,cutoff_mask);
 994
 995             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
 996
 997             /* Calculate temporary vectorial force */
 998             tx               = _mm256_mul_ps(fscal,dx12);
 999             ty               = _mm256_mul_ps(fscal,dy12);
1000             tz               = _mm256_mul_ps(fscal,dz12);
1001
1002             /* Update vectorial force */
1003             fix1             = _mm256_add_ps(fix1,tx);
1004             fiy1             = _mm256_add_ps(fiy1,ty);
1005             fiz1             = _mm256_add_ps(fiz1,tz);
1006
1007             fjx2             = _mm256_add_ps(fjx2,tx);
1008             fjy2             = _mm256_add_ps(fjy2,ty);
1009             fjz2             = _mm256_add_ps(fjz2,tz);
1010
1011             }
1012
1013             /**************************
1014              * CALCULATE INTERACTIONS *
1015              **************************/
1016
1017             if (gmx_mm256_any_lt(rsq13,rcutoff2))
1018             {
1019
1020             /* REACTION-FIELD ELECTROSTATICS */
1021             velec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_add_ps(rinv13,_mm256_mul_ps(krf,rsq13)),crf));
1022             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1023
1024             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
1025
1026             /* Update potential sum for this i atom from the interaction with this j atom. */
1027             velec            = _mm256_and_ps(velec,cutoff_mask);
1028             velec            = _mm256_andnot_ps(dummy_mask,velec);
1029             velecsum         = _mm256_add_ps(velecsum,velec);
1030
1031             fscal            = felec;
1032
1033             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1034
1035             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1036
1037             /* Calculate temporary vectorial force */
1038             tx               = _mm256_mul_ps(fscal,dx13);
1039             ty               = _mm256_mul_ps(fscal,dy13);
1040             tz               = _mm256_mul_ps(fscal,dz13);
1041
1042             /* Update vectorial force */
1043             fix1             = _mm256_add_ps(fix1,tx);
1044             fiy1             = _mm256_add_ps(fiy1,ty);
1045             fiz1             = _mm256_add_ps(fiz1,tz);
1046
1047             fjx3             = _mm256_add_ps(fjx3,tx);
1048             fjy3             = _mm256_add_ps(fjy3,ty);
1049             fjz3             = _mm256_add_ps(fjz3,tz);
1050
1051             }
1052
1053             /**************************
1054              * CALCULATE INTERACTIONS *
1055              **************************/
1056
1057             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1058             {
1059
1060             /* REACTION-FIELD ELECTROSTATICS */
1061             velec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_add_ps(rinv21,_mm256_mul_ps(krf,rsq21)),crf));
1062             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1063
1064             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1065
1066             /* Update potential sum for this i atom from the interaction with this j atom. */
1067             velec            = _mm256_and_ps(velec,cutoff_mask);
1068             velec            = _mm256_andnot_ps(dummy_mask,velec);
1069             velecsum         = _mm256_add_ps(velecsum,velec);
1070
1071             fscal            = felec;
1072
1073             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1074
1075             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1076
1077             /* Calculate temporary vectorial force */
1078             tx               = _mm256_mul_ps(fscal,dx21);
1079             ty               = _mm256_mul_ps(fscal,dy21);
1080             tz               = _mm256_mul_ps(fscal,dz21);
1081
1082             /* Update vectorial force */
1083             fix2             = _mm256_add_ps(fix2,tx);
1084             fiy2             = _mm256_add_ps(fiy2,ty);
1085             fiz2             = _mm256_add_ps(fiz2,tz);
1086
1087             fjx1             = _mm256_add_ps(fjx1,tx);
1088             fjy1             = _mm256_add_ps(fjy1,ty);
1089             fjz1             = _mm256_add_ps(fjz1,tz);
1090
1091             }
1092
1093             /**************************
1094              * CALCULATE INTERACTIONS *
1095              **************************/
1096
1097             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1098             {
1099
1100             /* REACTION-FIELD ELECTROSTATICS */
1101             velec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_add_ps(rinv22,_mm256_mul_ps(krf,rsq22)),crf));
1102             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1103
1104             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1105
1106             /* Update potential sum for this i atom from the interaction with this j atom. */
1107             velec            = _mm256_and_ps(velec,cutoff_mask);
1108             velec            = _mm256_andnot_ps(dummy_mask,velec);
1109             velecsum         = _mm256_add_ps(velecsum,velec);
1110
1111             fscal            = felec;
1112
1113             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1114
1115             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1116
1117             /* Calculate temporary vectorial force */
1118             tx               = _mm256_mul_ps(fscal,dx22);
1119             ty               = _mm256_mul_ps(fscal,dy22);
1120             tz               = _mm256_mul_ps(fscal,dz22);
1121
1122             /* Update vectorial force */
1123             fix2             = _mm256_add_ps(fix2,tx);
1124             fiy2             = _mm256_add_ps(fiy2,ty);
1125             fiz2             = _mm256_add_ps(fiz2,tz);
1126
1127             fjx2             = _mm256_add_ps(fjx2,tx);
1128             fjy2             = _mm256_add_ps(fjy2,ty);
1129             fjz2             = _mm256_add_ps(fjz2,tz);
1130
1131             }
1132
1133             /**************************
1134              * CALCULATE INTERACTIONS *
1135              **************************/
1136
1137             if (gmx_mm256_any_lt(rsq23,rcutoff2))
1138             {
1139
1140             /* REACTION-FIELD ELECTROSTATICS */
1141             velec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_add_ps(rinv23,_mm256_mul_ps(krf,rsq23)),crf));
1142             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1143
1144             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
1145
1146             /* Update potential sum for this i atom from the interaction with this j atom. */
1147             velec            = _mm256_and_ps(velec,cutoff_mask);
1148             velec            = _mm256_andnot_ps(dummy_mask,velec);
1149             velecsum         = _mm256_add_ps(velecsum,velec);
1150
1151             fscal            = felec;
1152
1153             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1154
1155             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1156
1157             /* Calculate temporary vectorial force */
1158             tx               = _mm256_mul_ps(fscal,dx23);
1159             ty               = _mm256_mul_ps(fscal,dy23);
1160             tz               = _mm256_mul_ps(fscal,dz23);
1161
1162             /* Update vectorial force */
1163             fix2             = _mm256_add_ps(fix2,tx);
1164             fiy2             = _mm256_add_ps(fiy2,ty);
1165             fiz2             = _mm256_add_ps(fiz2,tz);
1166
1167             fjx3             = _mm256_add_ps(fjx3,tx);
1168             fjy3             = _mm256_add_ps(fjy3,ty);
1169             fjz3             = _mm256_add_ps(fjz3,tz);
1170
1171             }
1172
1173             /**************************
1174              * CALCULATE INTERACTIONS *
1175              **************************/
1176
1177             if (gmx_mm256_any_lt(rsq31,rcutoff2))
1178             {
1179
1180             /* REACTION-FIELD ELECTROSTATICS */
1181             velec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_add_ps(rinv31,_mm256_mul_ps(krf,rsq31)),crf));
1182             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1183
1184             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
1185
1186             /* Update potential sum for this i atom from the interaction with this j atom. */
1187             velec            = _mm256_and_ps(velec,cutoff_mask);
1188             velec            = _mm256_andnot_ps(dummy_mask,velec);
1189             velecsum         = _mm256_add_ps(velecsum,velec);
1190
1191             fscal            = felec;
1192
1193             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1194
1195             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1196
1197             /* Calculate temporary vectorial force */
1198             tx               = _mm256_mul_ps(fscal,dx31);
1199             ty               = _mm256_mul_ps(fscal,dy31);
1200             tz               = _mm256_mul_ps(fscal,dz31);
1201
1202             /* Update vectorial force */
1203             fix3             = _mm256_add_ps(fix3,tx);
1204             fiy3             = _mm256_add_ps(fiy3,ty);
1205             fiz3             = _mm256_add_ps(fiz3,tz);
1206
1207             fjx1             = _mm256_add_ps(fjx1,tx);
1208             fjy1             = _mm256_add_ps(fjy1,ty);
1209             fjz1             = _mm256_add_ps(fjz1,tz);
1210
1211             }
1212
1213             /**************************
1214              * CALCULATE INTERACTIONS *
1215              **************************/
1216
1217             if (gmx_mm256_any_lt(rsq32,rcutoff2))
1218             {
1219
1220             /* REACTION-FIELD ELECTROSTATICS */
1221             velec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_add_ps(rinv32,_mm256_mul_ps(krf,rsq32)),crf));
1222             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1223
1224             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
1225
1226             /* Update potential sum for this i atom from the interaction with this j atom. */
1227             velec            = _mm256_and_ps(velec,cutoff_mask);
1228             velec            = _mm256_andnot_ps(dummy_mask,velec);
1229             velecsum         = _mm256_add_ps(velecsum,velec);
1230
1231             fscal            = felec;
1232
1233             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1234
1235             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1236
1237             /* Calculate temporary vectorial force */
1238             tx               = _mm256_mul_ps(fscal,dx32);
1239             ty               = _mm256_mul_ps(fscal,dy32);
1240             tz               = _mm256_mul_ps(fscal,dz32);
1241
1242             /* Update vectorial force */
1243             fix3             = _mm256_add_ps(fix3,tx);
1244             fiy3             = _mm256_add_ps(fiy3,ty);
1245             fiz3             = _mm256_add_ps(fiz3,tz);
1246
1247             fjx2             = _mm256_add_ps(fjx2,tx);
1248             fjy2             = _mm256_add_ps(fjy2,ty);
1249             fjz2             = _mm256_add_ps(fjz2,tz);
1250
1251             }
1252
1253             /**************************
1254              * CALCULATE INTERACTIONS *
1255              **************************/
1256
1257             if (gmx_mm256_any_lt(rsq33,rcutoff2))
1258             {
1259
1260             /* REACTION-FIELD ELECTROSTATICS */
1261             velec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_add_ps(rinv33,_mm256_mul_ps(krf,rsq33)),crf));
1262             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1263
1264             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
1265
1266             /* Update potential sum for this i atom from the interaction with this j atom. */
1267             velec            = _mm256_and_ps(velec,cutoff_mask);
1268             velec            = _mm256_andnot_ps(dummy_mask,velec);
1269             velecsum         = _mm256_add_ps(velecsum,velec);
1270
1271             fscal            = felec;
1272
1273             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1274
1275             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
1276
1277             /* Calculate temporary vectorial force */
1278             tx               = _mm256_mul_ps(fscal,dx33);
1279             ty               = _mm256_mul_ps(fscal,dy33);
1280             tz               = _mm256_mul_ps(fscal,dz33);
1281
1282             /* Update vectorial force */
1283             fix3             = _mm256_add_ps(fix3,tx);
1284             fiy3             = _mm256_add_ps(fiy3,ty);
1285             fiz3             = _mm256_add_ps(fiz3,tz);
1286
1287             fjx3             = _mm256_add_ps(fjx3,tx);
1288             fjy3             = _mm256_add_ps(fjy3,ty);
1289             fjz3             = _mm256_add_ps(fjz3,tz);
1290
1291             }
1292
1293             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1294             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1295             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1296             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1297             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
1298             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
1299             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
1300             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
1301
1302             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1303                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1304                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1305
1306             /* Inner loop uses 387 flops */
1307         }
1308
1309         /* End of innermost loop */
1310
1311         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1312                                                  f+i_coord_offset,fshift+i_shift_offset);
1313
1314         ggid                        = gid[iidx];
1315         /* Update potential energies */
1316         gmx_mm256_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1317         gmx_mm256_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1318
1319         /* Increment number of inner iterations */
1320         inneriter                  += j_index_end - j_index_start;
1321
1322         /* Outer loop uses 26 flops */
1323     }
1324
1325     /* Increment number of outer iterations */
1326     outeriter        += nri;
1327
1328     /* Update outer/inner flops */
1329
1330     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*387);
1331 }
1332 /*
1333  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomW4W4_F_avx_256_single
1334  * Electrostatics interaction: ReactionField
1335  * VdW interaction:            LennardJones
1336  * Geometry:                   Water4-Water4
1337  * Calculate force/pot:        Force
1338  */
1339 void
1340 nb_kernel_ElecRFCut_VdwLJSw_GeomW4W4_F_avx_256_single
1341                     (t_nblist                    * gmx_restrict       nlist,
1342                      rvec                        * gmx_restrict          xx,
1343                      rvec                        * gmx_restrict          ff,
1344                      t_forcerec                  * gmx_restrict          fr,
1345                      t_mdatoms                   * gmx_restrict     mdatoms,
1346                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1347                      t_nrnb                      * gmx_restrict        nrnb)
1348 {
1349     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1350      * just 0 for non-waters.
1351      * Suffixes A,B,C,D,E,F,G,H refer to j loop unrolling done with AVX, e.g. for the eight different
1352      * jnr indices corresponding to data put in the four positions in the SIMD register.
1353      */
1354     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1355     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1356     int              jnrA,jnrB,jnrC,jnrD;
1357     int              jnrE,jnrF,jnrG,jnrH;
1358     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1359     int              jnrlistE,jnrlistF,jnrlistG,jnrlistH;
1360     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1361     int              j_coord_offsetE,j_coord_offsetF,j_coord_offsetG,j_coord_offsetH;
1362     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1363     real             rcutoff_scalar;
1364     real             *shiftvec,*fshift,*x,*f;
1365     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD,*fjptrE,*fjptrF,*fjptrG,*fjptrH;
1366     real             scratch[4*DIM];
1367     __m256           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1368     real *           vdwioffsetptr0;
1369     __m256           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1370     real *           vdwioffsetptr1;
1371     __m256           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1372     real *           vdwioffsetptr2;
1373     __m256           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1374     real *           vdwioffsetptr3;
1375     __m256           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1376     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D,vdwjidx0E,vdwjidx0F,vdwjidx0G,vdwjidx0H;
1377     __m256           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1378     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D,vdwjidx1E,vdwjidx1F,vdwjidx1G,vdwjidx1H;
1379     __m256           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1380     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D,vdwjidx2E,vdwjidx2F,vdwjidx2G,vdwjidx2H;
1381     __m256           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1382     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D,vdwjidx3E,vdwjidx3F,vdwjidx3G,vdwjidx3H;
1383     __m256           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1384     __m256           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1385     __m256           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1386     __m256           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1387     __m256           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1388     __m256           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1389     __m256           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1390     __m256           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1391     __m256           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1392     __m256           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1393     __m256           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1394     __m256           velec,felec,velecsum,facel,crf,krf,krf2;
1395     real             *charge;
1396     int              nvdwtype;
1397     __m256           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1398     int              *vdwtype;
1399     real             *vdwparam;
1400     __m256           one_sixth   = _mm256_set1_ps(1.0/6.0);
1401     __m256           one_twelfth = _mm256_set1_ps(1.0/12.0);
1402     __m256           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
1403     real             rswitch_scalar,d_scalar;
1404     __m256           dummy_mask,cutoff_mask;
1405     __m256           signbit = _mm256_castsi256_ps( _mm256_set1_epi32(0x80000000) );
1406     __m256           one     = _mm256_set1_ps(1.0);
1407     __m256           two     = _mm256_set1_ps(2.0);
1408     x                = xx[0];
1409     f                = ff[0];
1410
1411     nri              = nlist->nri;
1412     iinr             = nlist->iinr;
1413     jindex           = nlist->jindex;
1414     jjnr             = nlist->jjnr;
1415     shiftidx         = nlist->shift;
1416     gid              = nlist->gid;
1417     shiftvec         = fr->shift_vec[0];
1418     fshift           = fr->fshift[0];
1419     facel            = _mm256_set1_ps(fr->epsfac);
1420     charge           = mdatoms->chargeA;
1421     krf              = _mm256_set1_ps(fr->ic->k_rf);
1422     krf2             = _mm256_set1_ps(fr->ic->k_rf*2.0);
1423     crf              = _mm256_set1_ps(fr->ic->c_rf);
1424     nvdwtype         = fr->ntype;
1425     vdwparam         = fr->nbfp;
1426     vdwtype          = mdatoms->typeA;
1427
1428     /* Setup water-specific parameters */
1429     inr              = nlist->iinr[0];
1430     iq1              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+1]));
1431     iq2              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+2]));
1432     iq3              = _mm256_mul_ps(facel,_mm256_set1_ps(charge[inr+3]));
1433     vdwioffsetptr0   = vdwparam+2*nvdwtype*vdwtype[inr+0];
1434
1435     jq1              = _mm256_set1_ps(charge[inr+1]);
1436     jq2              = _mm256_set1_ps(charge[inr+2]);
1437     jq3              = _mm256_set1_ps(charge[inr+3]);
1438     vdwjidx0A        = 2*vdwtype[inr+0];
1439     c6_00            = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A]);
1440     c12_00           = _mm256_set1_ps(vdwioffsetptr0[vdwjidx0A+1]);
1441     qq11             = _mm256_mul_ps(iq1,jq1);
1442     qq12             = _mm256_mul_ps(iq1,jq2);
1443     qq13             = _mm256_mul_ps(iq1,jq3);
1444     qq21             = _mm256_mul_ps(iq2,jq1);
1445     qq22             = _mm256_mul_ps(iq2,jq2);
1446     qq23             = _mm256_mul_ps(iq2,jq3);
1447     qq31             = _mm256_mul_ps(iq3,jq1);
1448     qq32             = _mm256_mul_ps(iq3,jq2);
1449     qq33             = _mm256_mul_ps(iq3,jq3);
1450
1451     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1452     rcutoff_scalar   = fr->rcoulomb;
1453     rcutoff          = _mm256_set1_ps(rcutoff_scalar);
1454     rcutoff2         = _mm256_mul_ps(rcutoff,rcutoff);
1455
1456     rswitch_scalar   = fr->rvdw_switch;
1457     rswitch          = _mm256_set1_ps(rswitch_scalar);
1458     /* Setup switch parameters */
1459     d_scalar         = rcutoff_scalar-rswitch_scalar;
1460     d                = _mm256_set1_ps(d_scalar);
1461     swV3             = _mm256_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
1462     swV4             = _mm256_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1463     swV5             = _mm256_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1464     swF2             = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
1465     swF3             = _mm256_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
1466     swF4             = _mm256_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
1467
1468     /* Avoid stupid compiler warnings */
1469     jnrA = jnrB = jnrC = jnrD = jnrE = jnrF = jnrG = jnrH = 0;
1470     j_coord_offsetA = 0;
1471     j_coord_offsetB = 0;
1472     j_coord_offsetC = 0;
1473     j_coord_offsetD = 0;
1474     j_coord_offsetE = 0;
1475     j_coord_offsetF = 0;
1476     j_coord_offsetG = 0;
1477     j_coord_offsetH = 0;
1478
1479     outeriter        = 0;
1480     inneriter        = 0;
1481
1482     for(iidx=0;iidx<4*DIM;iidx++)
1483     {
1484         scratch[iidx] = 0.0;
1485     }
1486
1487     /* Start outer loop over neighborlists */
1488     for(iidx=0; iidx<nri; iidx++)
1489     {
1490         /* Load shift vector for this list */
1491         i_shift_offset   = DIM*shiftidx[iidx];
1492
1493         /* Load limits for loop over neighbors */
1494         j_index_start    = jindex[iidx];
1495         j_index_end      = jindex[iidx+1];
1496
1497         /* Get outer coordinate index */
1498         inr              = iinr[iidx];
1499         i_coord_offset   = DIM*inr;
1500
1501         /* Load i particle coords and add shift vector */
1502         gmx_mm256_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1503                                                     &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1504
1505         fix0             = _mm256_setzero_ps();
1506         fiy0             = _mm256_setzero_ps();
1507         fiz0             = _mm256_setzero_ps();
1508         fix1             = _mm256_setzero_ps();
1509         fiy1             = _mm256_setzero_ps();
1510         fiz1             = _mm256_setzero_ps();
1511         fix2             = _mm256_setzero_ps();
1512         fiy2             = _mm256_setzero_ps();
1513         fiz2             = _mm256_setzero_ps();
1514         fix3             = _mm256_setzero_ps();
1515         fiy3             = _mm256_setzero_ps();
1516         fiz3             = _mm256_setzero_ps();
1517
1518         /* Start inner kernel loop */
1519         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+7]>=0; jidx+=8)
1520         {
1521
1522             /* Get j neighbor index, and coordinate index */
1523             jnrA             = jjnr[jidx];
1524             jnrB             = jjnr[jidx+1];
1525             jnrC             = jjnr[jidx+2];
1526             jnrD             = jjnr[jidx+3];
1527             jnrE             = jjnr[jidx+4];
1528             jnrF             = jjnr[jidx+5];
1529             jnrG             = jjnr[jidx+6];
1530             jnrH             = jjnr[jidx+7];
1531             j_coord_offsetA  = DIM*jnrA;
1532             j_coord_offsetB  = DIM*jnrB;
1533             j_coord_offsetC  = DIM*jnrC;
1534             j_coord_offsetD  = DIM*jnrD;
1535             j_coord_offsetE  = DIM*jnrE;
1536             j_coord_offsetF  = DIM*jnrF;
1537             j_coord_offsetG  = DIM*jnrG;
1538             j_coord_offsetH  = DIM*jnrH;
1539
1540             /* load j atom coordinates */
1541             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1542                                                  x+j_coord_offsetC,x+j_coord_offsetD,
1543                                                  x+j_coord_offsetE,x+j_coord_offsetF,
1544                                                  x+j_coord_offsetG,x+j_coord_offsetH,
1545                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1546                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
1547
1548             /* Calculate displacement vector */
1549             dx00             = _mm256_sub_ps(ix0,jx0);
1550             dy00             = _mm256_sub_ps(iy0,jy0);
1551             dz00             = _mm256_sub_ps(iz0,jz0);
1552             dx11             = _mm256_sub_ps(ix1,jx1);
1553             dy11             = _mm256_sub_ps(iy1,jy1);
1554             dz11             = _mm256_sub_ps(iz1,jz1);
1555             dx12             = _mm256_sub_ps(ix1,jx2);
1556             dy12             = _mm256_sub_ps(iy1,jy2);
1557             dz12             = _mm256_sub_ps(iz1,jz2);
1558             dx13             = _mm256_sub_ps(ix1,jx3);
1559             dy13             = _mm256_sub_ps(iy1,jy3);
1560             dz13             = _mm256_sub_ps(iz1,jz3);
1561             dx21             = _mm256_sub_ps(ix2,jx1);
1562             dy21             = _mm256_sub_ps(iy2,jy1);
1563             dz21             = _mm256_sub_ps(iz2,jz1);
1564             dx22             = _mm256_sub_ps(ix2,jx2);
1565             dy22             = _mm256_sub_ps(iy2,jy2);
1566             dz22             = _mm256_sub_ps(iz2,jz2);
1567             dx23             = _mm256_sub_ps(ix2,jx3);
1568             dy23             = _mm256_sub_ps(iy2,jy3);
1569             dz23             = _mm256_sub_ps(iz2,jz3);
1570             dx31             = _mm256_sub_ps(ix3,jx1);
1571             dy31             = _mm256_sub_ps(iy3,jy1);
1572             dz31             = _mm256_sub_ps(iz3,jz1);
1573             dx32             = _mm256_sub_ps(ix3,jx2);
1574             dy32             = _mm256_sub_ps(iy3,jy2);
1575             dz32             = _mm256_sub_ps(iz3,jz2);
1576             dx33             = _mm256_sub_ps(ix3,jx3);
1577             dy33             = _mm256_sub_ps(iy3,jy3);
1578             dz33             = _mm256_sub_ps(iz3,jz3);
1579
1580             /* Calculate squared distance and things based on it */
1581             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
1582             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
1583             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
1584             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
1585             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
1586             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
1587             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
1588             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
1589             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
1590             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
1591
1592             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
1593             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
1594             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
1595             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
1596             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
1597             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
1598             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
1599             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
1600             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
1601             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
1602
1603             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
1604             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
1605             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
1606             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
1607             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
1608             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
1609             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
1610             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
1611             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
1612             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
1613
1614             fjx0             = _mm256_setzero_ps();
1615             fjy0             = _mm256_setzero_ps();
1616             fjz0             = _mm256_setzero_ps();
1617             fjx1             = _mm256_setzero_ps();
1618             fjy1             = _mm256_setzero_ps();
1619             fjz1             = _mm256_setzero_ps();
1620             fjx2             = _mm256_setzero_ps();
1621             fjy2             = _mm256_setzero_ps();
1622             fjz2             = _mm256_setzero_ps();
1623             fjx3             = _mm256_setzero_ps();
1624             fjy3             = _mm256_setzero_ps();
1625             fjz3             = _mm256_setzero_ps();
1626
1627             /**************************
1628              * CALCULATE INTERACTIONS *
1629              **************************/
1630
1631             if (gmx_mm256_any_lt(rsq00,rcutoff2))
1632             {
1633
1634             r00              = _mm256_mul_ps(rsq00,rinv00);
1635
1636             /* LENNARD-JONES DISPERSION/REPULSION */
1637
1638             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1639             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
1640             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
1641             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
1642             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
1643
1644             d                = _mm256_sub_ps(r00,rswitch);
1645             d                = _mm256_max_ps(d,_mm256_setzero_ps());
1646             d2               = _mm256_mul_ps(d,d);
1647             sw               = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
1648
1649             dsw              = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
1650
1651             /* Evaluate switch function */
1652             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1653             fvdw             = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
1654             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
1655
1656             fscal            = fvdw;
1657
1658             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1659
1660             /* Calculate temporary vectorial force */
1661             tx               = _mm256_mul_ps(fscal,dx00);
1662             ty               = _mm256_mul_ps(fscal,dy00);
1663             tz               = _mm256_mul_ps(fscal,dz00);
1664
1665             /* Update vectorial force */
1666             fix0             = _mm256_add_ps(fix0,tx);
1667             fiy0             = _mm256_add_ps(fiy0,ty);
1668             fiz0             = _mm256_add_ps(fiz0,tz);
1669
1670             fjx0             = _mm256_add_ps(fjx0,tx);
1671             fjy0             = _mm256_add_ps(fjy0,ty);
1672             fjz0             = _mm256_add_ps(fjz0,tz);
1673
1674             }
1675
1676             /**************************
1677              * CALCULATE INTERACTIONS *
1678              **************************/
1679
1680             if (gmx_mm256_any_lt(rsq11,rcutoff2))
1681             {
1682
1683             /* REACTION-FIELD ELECTROSTATICS */
1684             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
1685
1686             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
1687
1688             fscal            = felec;
1689
1690             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1691
1692             /* Calculate temporary vectorial force */
1693             tx               = _mm256_mul_ps(fscal,dx11);
1694             ty               = _mm256_mul_ps(fscal,dy11);
1695             tz               = _mm256_mul_ps(fscal,dz11);
1696
1697             /* Update vectorial force */
1698             fix1             = _mm256_add_ps(fix1,tx);
1699             fiy1             = _mm256_add_ps(fiy1,ty);
1700             fiz1             = _mm256_add_ps(fiz1,tz);
1701
1702             fjx1             = _mm256_add_ps(fjx1,tx);
1703             fjy1             = _mm256_add_ps(fjy1,ty);
1704             fjz1             = _mm256_add_ps(fjz1,tz);
1705
1706             }
1707
1708             /**************************
1709              * CALCULATE INTERACTIONS *
1710              **************************/
1711
1712             if (gmx_mm256_any_lt(rsq12,rcutoff2))
1713             {
1714
1715             /* REACTION-FIELD ELECTROSTATICS */
1716             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
1717
1718             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
1719
1720             fscal            = felec;
1721
1722             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1723
1724             /* Calculate temporary vectorial force */
1725             tx               = _mm256_mul_ps(fscal,dx12);
1726             ty               = _mm256_mul_ps(fscal,dy12);
1727             tz               = _mm256_mul_ps(fscal,dz12);
1728
1729             /* Update vectorial force */
1730             fix1             = _mm256_add_ps(fix1,tx);
1731             fiy1             = _mm256_add_ps(fiy1,ty);
1732             fiz1             = _mm256_add_ps(fiz1,tz);
1733
1734             fjx2             = _mm256_add_ps(fjx2,tx);
1735             fjy2             = _mm256_add_ps(fjy2,ty);
1736             fjz2             = _mm256_add_ps(fjz2,tz);
1737
1738             }
1739
1740             /**************************
1741              * CALCULATE INTERACTIONS *
1742              **************************/
1743
1744             if (gmx_mm256_any_lt(rsq13,rcutoff2))
1745             {
1746
1747             /* REACTION-FIELD ELECTROSTATICS */
1748             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
1749
1750             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
1751
1752             fscal            = felec;
1753
1754             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1755
1756             /* Calculate temporary vectorial force */
1757             tx               = _mm256_mul_ps(fscal,dx13);
1758             ty               = _mm256_mul_ps(fscal,dy13);
1759             tz               = _mm256_mul_ps(fscal,dz13);
1760
1761             /* Update vectorial force */
1762             fix1             = _mm256_add_ps(fix1,tx);
1763             fiy1             = _mm256_add_ps(fiy1,ty);
1764             fiz1             = _mm256_add_ps(fiz1,tz);
1765
1766             fjx3             = _mm256_add_ps(fjx3,tx);
1767             fjy3             = _mm256_add_ps(fjy3,ty);
1768             fjz3             = _mm256_add_ps(fjz3,tz);
1769
1770             }
1771
1772             /**************************
1773              * CALCULATE INTERACTIONS *
1774              **************************/
1775
1776             if (gmx_mm256_any_lt(rsq21,rcutoff2))
1777             {
1778
1779             /* REACTION-FIELD ELECTROSTATICS */
1780             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
1781
1782             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
1783
1784             fscal            = felec;
1785
1786             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1787
1788             /* Calculate temporary vectorial force */
1789             tx               = _mm256_mul_ps(fscal,dx21);
1790             ty               = _mm256_mul_ps(fscal,dy21);
1791             tz               = _mm256_mul_ps(fscal,dz21);
1792
1793             /* Update vectorial force */
1794             fix2             = _mm256_add_ps(fix2,tx);
1795             fiy2             = _mm256_add_ps(fiy2,ty);
1796             fiz2             = _mm256_add_ps(fiz2,tz);
1797
1798             fjx1             = _mm256_add_ps(fjx1,tx);
1799             fjy1             = _mm256_add_ps(fjy1,ty);
1800             fjz1             = _mm256_add_ps(fjz1,tz);
1801
1802             }
1803
1804             /**************************
1805              * CALCULATE INTERACTIONS *
1806              **************************/
1807
1808             if (gmx_mm256_any_lt(rsq22,rcutoff2))
1809             {
1810
1811             /* REACTION-FIELD ELECTROSTATICS */
1812             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
1813
1814             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
1815
1816             fscal            = felec;
1817
1818             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1819
1820             /* Calculate temporary vectorial force */
1821             tx               = _mm256_mul_ps(fscal,dx22);
1822             ty               = _mm256_mul_ps(fscal,dy22);
1823             tz               = _mm256_mul_ps(fscal,dz22);
1824
1825             /* Update vectorial force */
1826             fix2             = _mm256_add_ps(fix2,tx);
1827             fiy2             = _mm256_add_ps(fiy2,ty);
1828             fiz2             = _mm256_add_ps(fiz2,tz);
1829
1830             fjx2             = _mm256_add_ps(fjx2,tx);
1831             fjy2             = _mm256_add_ps(fjy2,ty);
1832             fjz2             = _mm256_add_ps(fjz2,tz);
1833
1834             }
1835
1836             /**************************
1837              * CALCULATE INTERACTIONS *
1838              **************************/
1839
1840             if (gmx_mm256_any_lt(rsq23,rcutoff2))
1841             {
1842
1843             /* REACTION-FIELD ELECTROSTATICS */
1844             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
1845
1846             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
1847
1848             fscal            = felec;
1849
1850             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1851
1852             /* Calculate temporary vectorial force */
1853             tx               = _mm256_mul_ps(fscal,dx23);
1854             ty               = _mm256_mul_ps(fscal,dy23);
1855             tz               = _mm256_mul_ps(fscal,dz23);
1856
1857             /* Update vectorial force */
1858             fix2             = _mm256_add_ps(fix2,tx);
1859             fiy2             = _mm256_add_ps(fiy2,ty);
1860             fiz2             = _mm256_add_ps(fiz2,tz);
1861
1862             fjx3             = _mm256_add_ps(fjx3,tx);
1863             fjy3             = _mm256_add_ps(fjy3,ty);
1864             fjz3             = _mm256_add_ps(fjz3,tz);
1865
1866             }
1867
1868             /**************************
1869              * CALCULATE INTERACTIONS *
1870              **************************/
1871
1872             if (gmx_mm256_any_lt(rsq31,rcutoff2))
1873             {
1874
1875             /* REACTION-FIELD ELECTROSTATICS */
1876             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
1877
1878             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
1879
1880             fscal            = felec;
1881
1882             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1883
1884             /* Calculate temporary vectorial force */
1885             tx               = _mm256_mul_ps(fscal,dx31);
1886             ty               = _mm256_mul_ps(fscal,dy31);
1887             tz               = _mm256_mul_ps(fscal,dz31);
1888
1889             /* Update vectorial force */
1890             fix3             = _mm256_add_ps(fix3,tx);
1891             fiy3             = _mm256_add_ps(fiy3,ty);
1892             fiz3             = _mm256_add_ps(fiz3,tz);
1893
1894             fjx1             = _mm256_add_ps(fjx1,tx);
1895             fjy1             = _mm256_add_ps(fjy1,ty);
1896             fjz1             = _mm256_add_ps(fjz1,tz);
1897
1898             }
1899
1900             /**************************
1901              * CALCULATE INTERACTIONS *
1902              **************************/
1903
1904             if (gmx_mm256_any_lt(rsq32,rcutoff2))
1905             {
1906
1907             /* REACTION-FIELD ELECTROSTATICS */
1908             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
1909
1910             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
1911
1912             fscal            = felec;
1913
1914             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1915
1916             /* Calculate temporary vectorial force */
1917             tx               = _mm256_mul_ps(fscal,dx32);
1918             ty               = _mm256_mul_ps(fscal,dy32);
1919             tz               = _mm256_mul_ps(fscal,dz32);
1920
1921             /* Update vectorial force */
1922             fix3             = _mm256_add_ps(fix3,tx);
1923             fiy3             = _mm256_add_ps(fiy3,ty);
1924             fiz3             = _mm256_add_ps(fiz3,tz);
1925
1926             fjx2             = _mm256_add_ps(fjx2,tx);
1927             fjy2             = _mm256_add_ps(fjy2,ty);
1928             fjz2             = _mm256_add_ps(fjz2,tz);
1929
1930             }
1931
1932             /**************************
1933              * CALCULATE INTERACTIONS *
1934              **************************/
1935
1936             if (gmx_mm256_any_lt(rsq33,rcutoff2))
1937             {
1938
1939             /* REACTION-FIELD ELECTROSTATICS */
1940             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
1941
1942             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
1943
1944             fscal            = felec;
1945
1946             fscal            = _mm256_and_ps(fscal,cutoff_mask);
1947
1948             /* Calculate temporary vectorial force */
1949             tx               = _mm256_mul_ps(fscal,dx33);
1950             ty               = _mm256_mul_ps(fscal,dy33);
1951             tz               = _mm256_mul_ps(fscal,dz33);
1952
1953             /* Update vectorial force */
1954             fix3             = _mm256_add_ps(fix3,tx);
1955             fiy3             = _mm256_add_ps(fiy3,ty);
1956             fiz3             = _mm256_add_ps(fiz3,tz);
1957
1958             fjx3             = _mm256_add_ps(fjx3,tx);
1959             fjy3             = _mm256_add_ps(fjy3,ty);
1960             fjz3             = _mm256_add_ps(fjz3,tz);
1961
1962             }
1963
1964             fjptrA             = f+j_coord_offsetA;
1965             fjptrB             = f+j_coord_offsetB;
1966             fjptrC             = f+j_coord_offsetC;
1967             fjptrD             = f+j_coord_offsetD;
1968             fjptrE             = f+j_coord_offsetE;
1969             fjptrF             = f+j_coord_offsetF;
1970             fjptrG             = f+j_coord_offsetG;
1971             fjptrH             = f+j_coord_offsetH;
1972
1973             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
1974                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1975                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1976
1977             /* Inner loop uses 329 flops */
1978         }
1979
1980         if(jidx<j_index_end)
1981         {
1982
1983             /* Get j neighbor index, and coordinate index */
1984             jnrlistA         = jjnr[jidx];
1985             jnrlistB         = jjnr[jidx+1];
1986             jnrlistC         = jjnr[jidx+2];
1987             jnrlistD         = jjnr[jidx+3];
1988             jnrlistE         = jjnr[jidx+4];
1989             jnrlistF         = jjnr[jidx+5];
1990             jnrlistG         = jjnr[jidx+6];
1991             jnrlistH         = jjnr[jidx+7];
1992             /* Sign of each element will be negative for non-real atoms.
1993              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1994              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1995              */
1996             dummy_mask = gmx_mm256_set_m128(gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx+4)),_mm_setzero_si128())),
1997                                             gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128())));
1998
1999             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2000             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2001             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2002             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2003             jnrE       = (jnrlistE>=0) ? jnrlistE : 0;
2004             jnrF       = (jnrlistF>=0) ? jnrlistF : 0;
2005             jnrG       = (jnrlistG>=0) ? jnrlistG : 0;
2006             jnrH       = (jnrlistH>=0) ? jnrlistH : 0;
2007             j_coord_offsetA  = DIM*jnrA;
2008             j_coord_offsetB  = DIM*jnrB;
2009             j_coord_offsetC  = DIM*jnrC;
2010             j_coord_offsetD  = DIM*jnrD;
2011             j_coord_offsetE  = DIM*jnrE;
2012             j_coord_offsetF  = DIM*jnrF;
2013             j_coord_offsetG  = DIM*jnrG;
2014             j_coord_offsetH  = DIM*jnrH;
2015
2016             /* load j atom coordinates */
2017             gmx_mm256_load_4rvec_8ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2018                                                  x+j_coord_offsetC,x+j_coord_offsetD,
2019                                                  x+j_coord_offsetE,x+j_coord_offsetF,
2020                                                  x+j_coord_offsetG,x+j_coord_offsetH,
2021                                                  &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2022                                                  &jy2,&jz2,&jx3,&jy3,&jz3);
2023
2024             /* Calculate displacement vector */
2025             dx00             = _mm256_sub_ps(ix0,jx0);
2026             dy00             = _mm256_sub_ps(iy0,jy0);
2027             dz00             = _mm256_sub_ps(iz0,jz0);
2028             dx11             = _mm256_sub_ps(ix1,jx1);
2029             dy11             = _mm256_sub_ps(iy1,jy1);
2030             dz11             = _mm256_sub_ps(iz1,jz1);
2031             dx12             = _mm256_sub_ps(ix1,jx2);
2032             dy12             = _mm256_sub_ps(iy1,jy2);
2033             dz12             = _mm256_sub_ps(iz1,jz2);
2034             dx13             = _mm256_sub_ps(ix1,jx3);
2035             dy13             = _mm256_sub_ps(iy1,jy3);
2036             dz13             = _mm256_sub_ps(iz1,jz3);
2037             dx21             = _mm256_sub_ps(ix2,jx1);
2038             dy21             = _mm256_sub_ps(iy2,jy1);
2039             dz21             = _mm256_sub_ps(iz2,jz1);
2040             dx22             = _mm256_sub_ps(ix2,jx2);
2041             dy22             = _mm256_sub_ps(iy2,jy2);
2042             dz22             = _mm256_sub_ps(iz2,jz2);
2043             dx23             = _mm256_sub_ps(ix2,jx3);
2044             dy23             = _mm256_sub_ps(iy2,jy3);
2045             dz23             = _mm256_sub_ps(iz2,jz3);
2046             dx31             = _mm256_sub_ps(ix3,jx1);
2047             dy31             = _mm256_sub_ps(iy3,jy1);
2048             dz31             = _mm256_sub_ps(iz3,jz1);
2049             dx32             = _mm256_sub_ps(ix3,jx2);
2050             dy32             = _mm256_sub_ps(iy3,jy2);
2051             dz32             = _mm256_sub_ps(iz3,jz2);
2052             dx33             = _mm256_sub_ps(ix3,jx3);
2053             dy33             = _mm256_sub_ps(iy3,jy3);
2054             dz33             = _mm256_sub_ps(iz3,jz3);
2055
2056             /* Calculate squared distance and things based on it */
2057             rsq00            = gmx_mm256_calc_rsq_ps(dx00,dy00,dz00);
2058             rsq11            = gmx_mm256_calc_rsq_ps(dx11,dy11,dz11);
2059             rsq12            = gmx_mm256_calc_rsq_ps(dx12,dy12,dz12);
2060             rsq13            = gmx_mm256_calc_rsq_ps(dx13,dy13,dz13);
2061             rsq21            = gmx_mm256_calc_rsq_ps(dx21,dy21,dz21);
2062             rsq22            = gmx_mm256_calc_rsq_ps(dx22,dy22,dz22);
2063             rsq23            = gmx_mm256_calc_rsq_ps(dx23,dy23,dz23);
2064             rsq31            = gmx_mm256_calc_rsq_ps(dx31,dy31,dz31);
2065             rsq32            = gmx_mm256_calc_rsq_ps(dx32,dy32,dz32);
2066             rsq33            = gmx_mm256_calc_rsq_ps(dx33,dy33,dz33);
2067
2068             rinv00           = gmx_mm256_invsqrt_ps(rsq00);
2069             rinv11           = gmx_mm256_invsqrt_ps(rsq11);
2070             rinv12           = gmx_mm256_invsqrt_ps(rsq12);
2071             rinv13           = gmx_mm256_invsqrt_ps(rsq13);
2072             rinv21           = gmx_mm256_invsqrt_ps(rsq21);
2073             rinv22           = gmx_mm256_invsqrt_ps(rsq22);
2074             rinv23           = gmx_mm256_invsqrt_ps(rsq23);
2075             rinv31           = gmx_mm256_invsqrt_ps(rsq31);
2076             rinv32           = gmx_mm256_invsqrt_ps(rsq32);
2077             rinv33           = gmx_mm256_invsqrt_ps(rsq33);
2078
2079             rinvsq00         = _mm256_mul_ps(rinv00,rinv00);
2080             rinvsq11         = _mm256_mul_ps(rinv11,rinv11);
2081             rinvsq12         = _mm256_mul_ps(rinv12,rinv12);
2082             rinvsq13         = _mm256_mul_ps(rinv13,rinv13);
2083             rinvsq21         = _mm256_mul_ps(rinv21,rinv21);
2084             rinvsq22         = _mm256_mul_ps(rinv22,rinv22);
2085             rinvsq23         = _mm256_mul_ps(rinv23,rinv23);
2086             rinvsq31         = _mm256_mul_ps(rinv31,rinv31);
2087             rinvsq32         = _mm256_mul_ps(rinv32,rinv32);
2088             rinvsq33         = _mm256_mul_ps(rinv33,rinv33);
2089
2090             fjx0             = _mm256_setzero_ps();
2091             fjy0             = _mm256_setzero_ps();
2092             fjz0             = _mm256_setzero_ps();
2093             fjx1             = _mm256_setzero_ps();
2094             fjy1             = _mm256_setzero_ps();
2095             fjz1             = _mm256_setzero_ps();
2096             fjx2             = _mm256_setzero_ps();
2097             fjy2             = _mm256_setzero_ps();
2098             fjz2             = _mm256_setzero_ps();
2099             fjx3             = _mm256_setzero_ps();
2100             fjy3             = _mm256_setzero_ps();
2101             fjz3             = _mm256_setzero_ps();
2102
2103             /**************************
2104              * CALCULATE INTERACTIONS *
2105              **************************/
2106
2107             if (gmx_mm256_any_lt(rsq00,rcutoff2))
2108             {
2109
2110             r00              = _mm256_mul_ps(rsq00,rinv00);
2111             r00              = _mm256_andnot_ps(dummy_mask,r00);
2112
2113             /* LENNARD-JONES DISPERSION/REPULSION */
2114
2115             rinvsix          = _mm256_mul_ps(_mm256_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2116             vvdw6            = _mm256_mul_ps(c6_00,rinvsix);
2117             vvdw12           = _mm256_mul_ps(c12_00,_mm256_mul_ps(rinvsix,rinvsix));
2118             vvdw             = _mm256_sub_ps( _mm256_mul_ps(vvdw12,one_twelfth) , _mm256_mul_ps(vvdw6,one_sixth) );
2119             fvdw             = _mm256_mul_ps(_mm256_sub_ps(vvdw12,vvdw6),rinvsq00);
2120
2121             d                = _mm256_sub_ps(r00,rswitch);
2122             d                = _mm256_max_ps(d,_mm256_setzero_ps());
2123             d2               = _mm256_mul_ps(d,d);
2124             sw               = _mm256_add_ps(one,_mm256_mul_ps(d2,_mm256_mul_ps(d,_mm256_add_ps(swV3,_mm256_mul_ps(d,_mm256_add_ps(swV4,_mm256_mul_ps(d,swV5)))))));
2125
2126             dsw              = _mm256_mul_ps(d2,_mm256_add_ps(swF2,_mm256_mul_ps(d,_mm256_add_ps(swF3,_mm256_mul_ps(d,swF4)))));
2127
2128             /* Evaluate switch function */
2129             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
2130             fvdw             = _mm256_sub_ps( _mm256_mul_ps(fvdw,sw) , _mm256_mul_ps(rinv00,_mm256_mul_ps(vvdw,dsw)) );
2131             cutoff_mask      = _mm256_cmp_ps(rsq00,rcutoff2,_CMP_LT_OQ);
2132
2133             fscal            = fvdw;
2134
2135             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2136
2137             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2138
2139             /* Calculate temporary vectorial force */
2140             tx               = _mm256_mul_ps(fscal,dx00);
2141             ty               = _mm256_mul_ps(fscal,dy00);
2142             tz               = _mm256_mul_ps(fscal,dz00);
2143
2144             /* Update vectorial force */
2145             fix0             = _mm256_add_ps(fix0,tx);
2146             fiy0             = _mm256_add_ps(fiy0,ty);
2147             fiz0             = _mm256_add_ps(fiz0,tz);
2148
2149             fjx0             = _mm256_add_ps(fjx0,tx);
2150             fjy0             = _mm256_add_ps(fjy0,ty);
2151             fjz0             = _mm256_add_ps(fjz0,tz);
2152
2153             }
2154
2155             /**************************
2156              * CALCULATE INTERACTIONS *
2157              **************************/
2158
2159             if (gmx_mm256_any_lt(rsq11,rcutoff2))
2160             {
2161
2162             /* REACTION-FIELD ELECTROSTATICS */
2163             felec            = _mm256_mul_ps(qq11,_mm256_sub_ps(_mm256_mul_ps(rinv11,rinvsq11),krf2));
2164
2165             cutoff_mask      = _mm256_cmp_ps(rsq11,rcutoff2,_CMP_LT_OQ);
2166
2167             fscal            = felec;
2168
2169             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2170
2171             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2172
2173             /* Calculate temporary vectorial force */
2174             tx               = _mm256_mul_ps(fscal,dx11);
2175             ty               = _mm256_mul_ps(fscal,dy11);
2176             tz               = _mm256_mul_ps(fscal,dz11);
2177
2178             /* Update vectorial force */
2179             fix1             = _mm256_add_ps(fix1,tx);
2180             fiy1             = _mm256_add_ps(fiy1,ty);
2181             fiz1             = _mm256_add_ps(fiz1,tz);
2182
2183             fjx1             = _mm256_add_ps(fjx1,tx);
2184             fjy1             = _mm256_add_ps(fjy1,ty);
2185             fjz1             = _mm256_add_ps(fjz1,tz);
2186
2187             }
2188
2189             /**************************
2190              * CALCULATE INTERACTIONS *
2191              **************************/
2192
2193             if (gmx_mm256_any_lt(rsq12,rcutoff2))
2194             {
2195
2196             /* REACTION-FIELD ELECTROSTATICS */
2197             felec            = _mm256_mul_ps(qq12,_mm256_sub_ps(_mm256_mul_ps(rinv12,rinvsq12),krf2));
2198
2199             cutoff_mask      = _mm256_cmp_ps(rsq12,rcutoff2,_CMP_LT_OQ);
2200
2201             fscal            = felec;
2202
2203             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2204
2205             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2206
2207             /* Calculate temporary vectorial force */
2208             tx               = _mm256_mul_ps(fscal,dx12);
2209             ty               = _mm256_mul_ps(fscal,dy12);
2210             tz               = _mm256_mul_ps(fscal,dz12);
2211
2212             /* Update vectorial force */
2213             fix1             = _mm256_add_ps(fix1,tx);
2214             fiy1             = _mm256_add_ps(fiy1,ty);
2215             fiz1             = _mm256_add_ps(fiz1,tz);
2216
2217             fjx2             = _mm256_add_ps(fjx2,tx);
2218             fjy2             = _mm256_add_ps(fjy2,ty);
2219             fjz2             = _mm256_add_ps(fjz2,tz);
2220
2221             }
2222
2223             /**************************
2224              * CALCULATE INTERACTIONS *
2225              **************************/
2226
2227             if (gmx_mm256_any_lt(rsq13,rcutoff2))
2228             {
2229
2230             /* REACTION-FIELD ELECTROSTATICS */
2231             felec            = _mm256_mul_ps(qq13,_mm256_sub_ps(_mm256_mul_ps(rinv13,rinvsq13),krf2));
2232
2233             cutoff_mask      = _mm256_cmp_ps(rsq13,rcutoff2,_CMP_LT_OQ);
2234
2235             fscal            = felec;
2236
2237             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2238
2239             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2240
2241             /* Calculate temporary vectorial force */
2242             tx               = _mm256_mul_ps(fscal,dx13);
2243             ty               = _mm256_mul_ps(fscal,dy13);
2244             tz               = _mm256_mul_ps(fscal,dz13);
2245
2246             /* Update vectorial force */
2247             fix1             = _mm256_add_ps(fix1,tx);
2248             fiy1             = _mm256_add_ps(fiy1,ty);
2249             fiz1             = _mm256_add_ps(fiz1,tz);
2250
2251             fjx3             = _mm256_add_ps(fjx3,tx);
2252             fjy3             = _mm256_add_ps(fjy3,ty);
2253             fjz3             = _mm256_add_ps(fjz3,tz);
2254
2255             }
2256
2257             /**************************
2258              * CALCULATE INTERACTIONS *
2259              **************************/
2260
2261             if (gmx_mm256_any_lt(rsq21,rcutoff2))
2262             {
2263
2264             /* REACTION-FIELD ELECTROSTATICS */
2265             felec            = _mm256_mul_ps(qq21,_mm256_sub_ps(_mm256_mul_ps(rinv21,rinvsq21),krf2));
2266
2267             cutoff_mask      = _mm256_cmp_ps(rsq21,rcutoff2,_CMP_LT_OQ);
2268
2269             fscal            = felec;
2270
2271             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2272
2273             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2274
2275             /* Calculate temporary vectorial force */
2276             tx               = _mm256_mul_ps(fscal,dx21);
2277             ty               = _mm256_mul_ps(fscal,dy21);
2278             tz               = _mm256_mul_ps(fscal,dz21);
2279
2280             /* Update vectorial force */
2281             fix2             = _mm256_add_ps(fix2,tx);
2282             fiy2             = _mm256_add_ps(fiy2,ty);
2283             fiz2             = _mm256_add_ps(fiz2,tz);
2284
2285             fjx1             = _mm256_add_ps(fjx1,tx);
2286             fjy1             = _mm256_add_ps(fjy1,ty);
2287             fjz1             = _mm256_add_ps(fjz1,tz);
2288
2289             }
2290
2291             /**************************
2292              * CALCULATE INTERACTIONS *
2293              **************************/
2294
2295             if (gmx_mm256_any_lt(rsq22,rcutoff2))
2296             {
2297
2298             /* REACTION-FIELD ELECTROSTATICS */
2299             felec            = _mm256_mul_ps(qq22,_mm256_sub_ps(_mm256_mul_ps(rinv22,rinvsq22),krf2));
2300
2301             cutoff_mask      = _mm256_cmp_ps(rsq22,rcutoff2,_CMP_LT_OQ);
2302
2303             fscal            = felec;
2304
2305             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2306
2307             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2308
2309             /* Calculate temporary vectorial force */
2310             tx               = _mm256_mul_ps(fscal,dx22);
2311             ty               = _mm256_mul_ps(fscal,dy22);
2312             tz               = _mm256_mul_ps(fscal,dz22);
2313
2314             /* Update vectorial force */
2315             fix2             = _mm256_add_ps(fix2,tx);
2316             fiy2             = _mm256_add_ps(fiy2,ty);
2317             fiz2             = _mm256_add_ps(fiz2,tz);
2318
2319             fjx2             = _mm256_add_ps(fjx2,tx);
2320             fjy2             = _mm256_add_ps(fjy2,ty);
2321             fjz2             = _mm256_add_ps(fjz2,tz);
2322
2323             }
2324
2325             /**************************
2326              * CALCULATE INTERACTIONS *
2327              **************************/
2328
2329             if (gmx_mm256_any_lt(rsq23,rcutoff2))
2330             {
2331
2332             /* REACTION-FIELD ELECTROSTATICS */
2333             felec            = _mm256_mul_ps(qq23,_mm256_sub_ps(_mm256_mul_ps(rinv23,rinvsq23),krf2));
2334
2335             cutoff_mask      = _mm256_cmp_ps(rsq23,rcutoff2,_CMP_LT_OQ);
2336
2337             fscal            = felec;
2338
2339             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2340
2341             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2342
2343             /* Calculate temporary vectorial force */
2344             tx               = _mm256_mul_ps(fscal,dx23);
2345             ty               = _mm256_mul_ps(fscal,dy23);
2346             tz               = _mm256_mul_ps(fscal,dz23);
2347
2348             /* Update vectorial force */
2349             fix2             = _mm256_add_ps(fix2,tx);
2350             fiy2             = _mm256_add_ps(fiy2,ty);
2351             fiz2             = _mm256_add_ps(fiz2,tz);
2352
2353             fjx3             = _mm256_add_ps(fjx3,tx);
2354             fjy3             = _mm256_add_ps(fjy3,ty);
2355             fjz3             = _mm256_add_ps(fjz3,tz);
2356
2357             }
2358
2359             /**************************
2360              * CALCULATE INTERACTIONS *
2361              **************************/
2362
2363             if (gmx_mm256_any_lt(rsq31,rcutoff2))
2364             {
2365
2366             /* REACTION-FIELD ELECTROSTATICS */
2367             felec            = _mm256_mul_ps(qq31,_mm256_sub_ps(_mm256_mul_ps(rinv31,rinvsq31),krf2));
2368
2369             cutoff_mask      = _mm256_cmp_ps(rsq31,rcutoff2,_CMP_LT_OQ);
2370
2371             fscal            = felec;
2372
2373             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2374
2375             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2376
2377             /* Calculate temporary vectorial force */
2378             tx               = _mm256_mul_ps(fscal,dx31);
2379             ty               = _mm256_mul_ps(fscal,dy31);
2380             tz               = _mm256_mul_ps(fscal,dz31);
2381
2382             /* Update vectorial force */
2383             fix3             = _mm256_add_ps(fix3,tx);
2384             fiy3             = _mm256_add_ps(fiy3,ty);
2385             fiz3             = _mm256_add_ps(fiz3,tz);
2386
2387             fjx1             = _mm256_add_ps(fjx1,tx);
2388             fjy1             = _mm256_add_ps(fjy1,ty);
2389             fjz1             = _mm256_add_ps(fjz1,tz);
2390
2391             }
2392
2393             /**************************
2394              * CALCULATE INTERACTIONS *
2395              **************************/
2396
2397             if (gmx_mm256_any_lt(rsq32,rcutoff2))
2398             {
2399
2400             /* REACTION-FIELD ELECTROSTATICS */
2401             felec            = _mm256_mul_ps(qq32,_mm256_sub_ps(_mm256_mul_ps(rinv32,rinvsq32),krf2));
2402
2403             cutoff_mask      = _mm256_cmp_ps(rsq32,rcutoff2,_CMP_LT_OQ);
2404
2405             fscal            = felec;
2406
2407             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2408
2409             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2410
2411             /* Calculate temporary vectorial force */
2412             tx               = _mm256_mul_ps(fscal,dx32);
2413             ty               = _mm256_mul_ps(fscal,dy32);
2414             tz               = _mm256_mul_ps(fscal,dz32);
2415
2416             /* Update vectorial force */
2417             fix3             = _mm256_add_ps(fix3,tx);
2418             fiy3             = _mm256_add_ps(fiy3,ty);
2419             fiz3             = _mm256_add_ps(fiz3,tz);
2420
2421             fjx2             = _mm256_add_ps(fjx2,tx);
2422             fjy2             = _mm256_add_ps(fjy2,ty);
2423             fjz2             = _mm256_add_ps(fjz2,tz);
2424
2425             }
2426
2427             /**************************
2428              * CALCULATE INTERACTIONS *
2429              **************************/
2430
2431             if (gmx_mm256_any_lt(rsq33,rcutoff2))
2432             {
2433
2434             /* REACTION-FIELD ELECTROSTATICS */
2435             felec            = _mm256_mul_ps(qq33,_mm256_sub_ps(_mm256_mul_ps(rinv33,rinvsq33),krf2));
2436
2437             cutoff_mask      = _mm256_cmp_ps(rsq33,rcutoff2,_CMP_LT_OQ);
2438
2439             fscal            = felec;
2440
2441             fscal            = _mm256_and_ps(fscal,cutoff_mask);
2442
2443             fscal            = _mm256_andnot_ps(dummy_mask,fscal);
2444
2445             /* Calculate temporary vectorial force */
2446             tx               = _mm256_mul_ps(fscal,dx33);
2447             ty               = _mm256_mul_ps(fscal,dy33);
2448             tz               = _mm256_mul_ps(fscal,dz33);
2449
2450             /* Update vectorial force */
2451             fix3             = _mm256_add_ps(fix3,tx);
2452             fiy3             = _mm256_add_ps(fiy3,ty);
2453             fiz3             = _mm256_add_ps(fiz3,tz);
2454
2455             fjx3             = _mm256_add_ps(fjx3,tx);
2456             fjy3             = _mm256_add_ps(fjy3,ty);
2457             fjz3             = _mm256_add_ps(fjz3,tz);
2458
2459             }
2460
2461             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2462             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2463             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2464             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2465             fjptrE             = (jnrlistE>=0) ? f+j_coord_offsetE : scratch;
2466             fjptrF             = (jnrlistF>=0) ? f+j_coord_offsetF : scratch;
2467             fjptrG             = (jnrlistG>=0) ? f+j_coord_offsetG : scratch;
2468             fjptrH             = (jnrlistH>=0) ? f+j_coord_offsetH : scratch;
2469
2470             gmx_mm256_decrement_4rvec_8ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjptrE,fjptrF,fjptrG,fjptrH,
2471                                                       fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2472                                                       fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2473
2474             /* Inner loop uses 330 flops */
2475         }
2476
2477         /* End of innermost loop */
2478
2479         gmx_mm256_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2480                                                  f+i_coord_offset,fshift+i_shift_offset);
2481
2482         /* Increment number of inner iterations */
2483         inneriter                  += j_index_end - j_index_start;
2484
2485         /* Outer loop uses 24 flops */
2486     }
2487
2488     /* Increment number of outer iterations */
2489     outeriter        += nri;
2490
2491     /* Update outer/inner flops */
2492
2493     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*330);
2494 }