src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecEw_VdwCSTab_GeomW4W4_sse4_1_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, 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 sse4_1_single kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse4_1_single.h"
  50 #include "kernelutil_x86_sse4_1_single.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_sse4_1_single
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            CubicSplineTable
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_sse4_1_single
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB,jnrC,jnrD;
  77     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  78     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  79     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  80     real             rcutoff_scalar;
  81     real             *shiftvec,*fshift,*x,*f;
  82     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  83     real             scratch[4*DIM];
  84     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  85     int              vdwioffset0;
  86     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  87     int              vdwioffset1;
  88     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  89     int              vdwioffset2;
  90     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  91     int              vdwioffset3;
  92     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  93     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  94     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  95     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  96     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  97     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  98     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  99     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
 100     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
 101     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 102     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 103     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 104     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 105     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 106     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 107     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 108     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 109     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 110     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 111     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 112     real             *charge;
 113     int              nvdwtype;
 114     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 115     int              *vdwtype;
 116     real             *vdwparam;
 117     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 118     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 119     __m128i          vfitab;
 120     __m128i          ifour       = _mm_set1_epi32(4);
 121     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 122     real             *vftab;
 123     __m128i          ewitab;
 124     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 125     real             *ewtab;
 126     __m128           dummy_mask,cutoff_mask;
 127     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 128     __m128           one     = _mm_set1_ps(1.0);
 129     __m128           two     = _mm_set1_ps(2.0);
 130     x                = xx[0];
 131     f                = ff[0];
 132
 133     nri              = nlist->nri;
 134     iinr             = nlist->iinr;
 135     jindex           = nlist->jindex;
 136     jjnr             = nlist->jjnr;
 137     shiftidx         = nlist->shift;
 138     gid              = nlist->gid;
 139     shiftvec         = fr->shift_vec[0];
 140     fshift           = fr->fshift[0];
 141     facel            = _mm_set1_ps(fr->epsfac);
 142     charge           = mdatoms->chargeA;
 143     nvdwtype         = fr->ntype;
 144     vdwparam         = fr->nbfp;
 145     vdwtype          = mdatoms->typeA;
 146
 147     vftab            = kernel_data->table_vdw->data;
 148     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 149
 150     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 151     ewtab            = fr->ic->tabq_coul_FDV0;
 152     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 153     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 154
 155     /* Setup water-specific parameters */
 156     inr              = nlist->iinr[0];
 157     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 158     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 159     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 160     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 161
 162     jq1              = _mm_set1_ps(charge[inr+1]);
 163     jq2              = _mm_set1_ps(charge[inr+2]);
 164     jq3              = _mm_set1_ps(charge[inr+3]);
 165     vdwjidx0A        = 2*vdwtype[inr+0];
 166     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 167     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 168     qq11             = _mm_mul_ps(iq1,jq1);
 169     qq12             = _mm_mul_ps(iq1,jq2);
 170     qq13             = _mm_mul_ps(iq1,jq3);
 171     qq21             = _mm_mul_ps(iq2,jq1);
 172     qq22             = _mm_mul_ps(iq2,jq2);
 173     qq23             = _mm_mul_ps(iq2,jq3);
 174     qq31             = _mm_mul_ps(iq3,jq1);
 175     qq32             = _mm_mul_ps(iq3,jq2);
 176     qq33             = _mm_mul_ps(iq3,jq3);
 177
 178     /* Avoid stupid compiler warnings */
 179     jnrA = jnrB = jnrC = jnrD = 0;
 180     j_coord_offsetA = 0;
 181     j_coord_offsetB = 0;
 182     j_coord_offsetC = 0;
 183     j_coord_offsetD = 0;
 184
 185     outeriter        = 0;
 186     inneriter        = 0;
 187
 188     for(iidx=0;iidx<4*DIM;iidx++)
 189     {
 190         scratch[iidx] = 0.0;
 191     }
 192
 193     /* Start outer loop over neighborlists */
 194     for(iidx=0; iidx<nri; iidx++)
 195     {
 196         /* Load shift vector for this list */
 197         i_shift_offset   = DIM*shiftidx[iidx];
 198
 199         /* Load limits for loop over neighbors */
 200         j_index_start    = jindex[iidx];
 201         j_index_end      = jindex[iidx+1];
 202
 203         /* Get outer coordinate index */
 204         inr              = iinr[iidx];
 205         i_coord_offset   = DIM*inr;
 206
 207         /* Load i particle coords and add shift vector */
 208         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 209                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 210
 211         fix0             = _mm_setzero_ps();
 212         fiy0             = _mm_setzero_ps();
 213         fiz0             = _mm_setzero_ps();
 214         fix1             = _mm_setzero_ps();
 215         fiy1             = _mm_setzero_ps();
 216         fiz1             = _mm_setzero_ps();
 217         fix2             = _mm_setzero_ps();
 218         fiy2             = _mm_setzero_ps();
 219         fiz2             = _mm_setzero_ps();
 220         fix3             = _mm_setzero_ps();
 221         fiy3             = _mm_setzero_ps();
 222         fiz3             = _mm_setzero_ps();
 223
 224         /* Reset potential sums */
 225         velecsum         = _mm_setzero_ps();
 226         vvdwsum          = _mm_setzero_ps();
 227
 228         /* Start inner kernel loop */
 229         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 230         {
 231
 232             /* Get j neighbor index, and coordinate index */
 233             jnrA             = jjnr[jidx];
 234             jnrB             = jjnr[jidx+1];
 235             jnrC             = jjnr[jidx+2];
 236             jnrD             = jjnr[jidx+3];
 237             j_coord_offsetA  = DIM*jnrA;
 238             j_coord_offsetB  = DIM*jnrB;
 239             j_coord_offsetC  = DIM*jnrC;
 240             j_coord_offsetD  = DIM*jnrD;
 241
 242             /* load j atom coordinates */
 243             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 244                                               x+j_coord_offsetC,x+j_coord_offsetD,
 245                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 246                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 247
 248             /* Calculate displacement vector */
 249             dx00             = _mm_sub_ps(ix0,jx0);
 250             dy00             = _mm_sub_ps(iy0,jy0);
 251             dz00             = _mm_sub_ps(iz0,jz0);
 252             dx11             = _mm_sub_ps(ix1,jx1);
 253             dy11             = _mm_sub_ps(iy1,jy1);
 254             dz11             = _mm_sub_ps(iz1,jz1);
 255             dx12             = _mm_sub_ps(ix1,jx2);
 256             dy12             = _mm_sub_ps(iy1,jy2);
 257             dz12             = _mm_sub_ps(iz1,jz2);
 258             dx13             = _mm_sub_ps(ix1,jx3);
 259             dy13             = _mm_sub_ps(iy1,jy3);
 260             dz13             = _mm_sub_ps(iz1,jz3);
 261             dx21             = _mm_sub_ps(ix2,jx1);
 262             dy21             = _mm_sub_ps(iy2,jy1);
 263             dz21             = _mm_sub_ps(iz2,jz1);
 264             dx22             = _mm_sub_ps(ix2,jx2);
 265             dy22             = _mm_sub_ps(iy2,jy2);
 266             dz22             = _mm_sub_ps(iz2,jz2);
 267             dx23             = _mm_sub_ps(ix2,jx3);
 268             dy23             = _mm_sub_ps(iy2,jy3);
 269             dz23             = _mm_sub_ps(iz2,jz3);
 270             dx31             = _mm_sub_ps(ix3,jx1);
 271             dy31             = _mm_sub_ps(iy3,jy1);
 272             dz31             = _mm_sub_ps(iz3,jz1);
 273             dx32             = _mm_sub_ps(ix3,jx2);
 274             dy32             = _mm_sub_ps(iy3,jy2);
 275             dz32             = _mm_sub_ps(iz3,jz2);
 276             dx33             = _mm_sub_ps(ix3,jx3);
 277             dy33             = _mm_sub_ps(iy3,jy3);
 278             dz33             = _mm_sub_ps(iz3,jz3);
 279
 280             /* Calculate squared distance and things based on it */
 281             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 282             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 283             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 284             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 285             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 286             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 287             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 288             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 289             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 290             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 291
 292             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 293             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 294             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 295             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 296             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 297             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 298             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 299             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 300             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 301             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 302
 303             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 304             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 305             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 306             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 307             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 308             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 309             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 310             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 311             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 312
 313             fjx0             = _mm_setzero_ps();
 314             fjy0             = _mm_setzero_ps();
 315             fjz0             = _mm_setzero_ps();
 316             fjx1             = _mm_setzero_ps();
 317             fjy1             = _mm_setzero_ps();
 318             fjz1             = _mm_setzero_ps();
 319             fjx2             = _mm_setzero_ps();
 320             fjy2             = _mm_setzero_ps();
 321             fjz2             = _mm_setzero_ps();
 322             fjx3             = _mm_setzero_ps();
 323             fjy3             = _mm_setzero_ps();
 324             fjz3             = _mm_setzero_ps();
 325
 326             /**************************
 327              * CALCULATE INTERACTIONS *
 328              **************************/
 329
 330             r00              = _mm_mul_ps(rsq00,rinv00);
 331
 332             /* Calculate table index by multiplying r with table scale and truncate to integer */
 333             rt               = _mm_mul_ps(r00,vftabscale);
 334             vfitab           = _mm_cvttps_epi32(rt);
 335             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 336             vfitab           = _mm_slli_epi32(vfitab,3);
 337
 338             /* CUBIC SPLINE TABLE DISPERSION */
 339             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 340             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 341             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 342             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 343             _MM_TRANSPOSE4_PS(Y,F,G,H);
 344             Heps             = _mm_mul_ps(vfeps,H);
 345             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 346             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 347             vvdw6            = _mm_mul_ps(c6_00,VV);
 348             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 349             fvdw6            = _mm_mul_ps(c6_00,FF);
 350
 351             /* CUBIC SPLINE TABLE REPULSION */
 352             vfitab           = _mm_add_epi32(vfitab,ifour);
 353             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 354             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 355             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 356             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 357             _MM_TRANSPOSE4_PS(Y,F,G,H);
 358             Heps             = _mm_mul_ps(vfeps,H);
 359             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 360             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 361             vvdw12           = _mm_mul_ps(c12_00,VV);
 362             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 363             fvdw12           = _mm_mul_ps(c12_00,FF);
 364             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 365             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 366
 367             /* Update potential sum for this i atom from the interaction with this j atom. */
 368             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 369
 370             fscal            = fvdw;
 371
 372             /* Calculate temporary vectorial force */
 373             tx               = _mm_mul_ps(fscal,dx00);
 374             ty               = _mm_mul_ps(fscal,dy00);
 375             tz               = _mm_mul_ps(fscal,dz00);
 376
 377             /* Update vectorial force */
 378             fix0             = _mm_add_ps(fix0,tx);
 379             fiy0             = _mm_add_ps(fiy0,ty);
 380             fiz0             = _mm_add_ps(fiz0,tz);
 381
 382             fjx0             = _mm_add_ps(fjx0,tx);
 383             fjy0             = _mm_add_ps(fjy0,ty);
 384             fjz0             = _mm_add_ps(fjz0,tz);
 385
 386             /**************************
 387              * CALCULATE INTERACTIONS *
 388              **************************/
 389
 390             r11              = _mm_mul_ps(rsq11,rinv11);
 391
 392             /* EWALD ELECTROSTATICS */
 393
 394             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 395             ewrt             = _mm_mul_ps(r11,ewtabscale);
 396             ewitab           = _mm_cvttps_epi32(ewrt);
 397             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 398             ewitab           = _mm_slli_epi32(ewitab,2);
 399             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 400             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 401             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 402             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 403             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 404             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 405             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 406             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 407             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 408
 409             /* Update potential sum for this i atom from the interaction with this j atom. */
 410             velecsum         = _mm_add_ps(velecsum,velec);
 411
 412             fscal            = felec;
 413
 414             /* Calculate temporary vectorial force */
 415             tx               = _mm_mul_ps(fscal,dx11);
 416             ty               = _mm_mul_ps(fscal,dy11);
 417             tz               = _mm_mul_ps(fscal,dz11);
 418
 419             /* Update vectorial force */
 420             fix1             = _mm_add_ps(fix1,tx);
 421             fiy1             = _mm_add_ps(fiy1,ty);
 422             fiz1             = _mm_add_ps(fiz1,tz);
 423
 424             fjx1             = _mm_add_ps(fjx1,tx);
 425             fjy1             = _mm_add_ps(fjy1,ty);
 426             fjz1             = _mm_add_ps(fjz1,tz);
 427
 428             /**************************
 429              * CALCULATE INTERACTIONS *
 430              **************************/
 431
 432             r12              = _mm_mul_ps(rsq12,rinv12);
 433
 434             /* EWALD ELECTROSTATICS */
 435
 436             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 437             ewrt             = _mm_mul_ps(r12,ewtabscale);
 438             ewitab           = _mm_cvttps_epi32(ewrt);
 439             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 440             ewitab           = _mm_slli_epi32(ewitab,2);
 441             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 442             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 443             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 444             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 445             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 446             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 447             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 448             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
 449             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
 450
 451             /* Update potential sum for this i atom from the interaction with this j atom. */
 452             velecsum         = _mm_add_ps(velecsum,velec);
 453
 454             fscal            = felec;
 455
 456             /* Calculate temporary vectorial force */
 457             tx               = _mm_mul_ps(fscal,dx12);
 458             ty               = _mm_mul_ps(fscal,dy12);
 459             tz               = _mm_mul_ps(fscal,dz12);
 460
 461             /* Update vectorial force */
 462             fix1             = _mm_add_ps(fix1,tx);
 463             fiy1             = _mm_add_ps(fiy1,ty);
 464             fiz1             = _mm_add_ps(fiz1,tz);
 465
 466             fjx2             = _mm_add_ps(fjx2,tx);
 467             fjy2             = _mm_add_ps(fjy2,ty);
 468             fjz2             = _mm_add_ps(fjz2,tz);
 469
 470             /**************************
 471              * CALCULATE INTERACTIONS *
 472              **************************/
 473
 474             r13              = _mm_mul_ps(rsq13,rinv13);
 475
 476             /* EWALD ELECTROSTATICS */
 477
 478             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 479             ewrt             = _mm_mul_ps(r13,ewtabscale);
 480             ewitab           = _mm_cvttps_epi32(ewrt);
 481             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 482             ewitab           = _mm_slli_epi32(ewitab,2);
 483             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 484             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 485             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 486             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 487             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 488             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 489             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 490             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
 491             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
 492
 493             /* Update potential sum for this i atom from the interaction with this j atom. */
 494             velecsum         = _mm_add_ps(velecsum,velec);
 495
 496             fscal            = felec;
 497
 498             /* Calculate temporary vectorial force */
 499             tx               = _mm_mul_ps(fscal,dx13);
 500             ty               = _mm_mul_ps(fscal,dy13);
 501             tz               = _mm_mul_ps(fscal,dz13);
 502
 503             /* Update vectorial force */
 504             fix1             = _mm_add_ps(fix1,tx);
 505             fiy1             = _mm_add_ps(fiy1,ty);
 506             fiz1             = _mm_add_ps(fiz1,tz);
 507
 508             fjx3             = _mm_add_ps(fjx3,tx);
 509             fjy3             = _mm_add_ps(fjy3,ty);
 510             fjz3             = _mm_add_ps(fjz3,tz);
 511
 512             /**************************
 513              * CALCULATE INTERACTIONS *
 514              **************************/
 515
 516             r21              = _mm_mul_ps(rsq21,rinv21);
 517
 518             /* EWALD ELECTROSTATICS */
 519
 520             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 521             ewrt             = _mm_mul_ps(r21,ewtabscale);
 522             ewitab           = _mm_cvttps_epi32(ewrt);
 523             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 524             ewitab           = _mm_slli_epi32(ewitab,2);
 525             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 526             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 527             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 528             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 529             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 530             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 531             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 532             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
 533             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
 534
 535             /* Update potential sum for this i atom from the interaction with this j atom. */
 536             velecsum         = _mm_add_ps(velecsum,velec);
 537
 538             fscal            = felec;
 539
 540             /* Calculate temporary vectorial force */
 541             tx               = _mm_mul_ps(fscal,dx21);
 542             ty               = _mm_mul_ps(fscal,dy21);
 543             tz               = _mm_mul_ps(fscal,dz21);
 544
 545             /* Update vectorial force */
 546             fix2             = _mm_add_ps(fix2,tx);
 547             fiy2             = _mm_add_ps(fiy2,ty);
 548             fiz2             = _mm_add_ps(fiz2,tz);
 549
 550             fjx1             = _mm_add_ps(fjx1,tx);
 551             fjy1             = _mm_add_ps(fjy1,ty);
 552             fjz1             = _mm_add_ps(fjz1,tz);
 553
 554             /**************************
 555              * CALCULATE INTERACTIONS *
 556              **************************/
 557
 558             r22              = _mm_mul_ps(rsq22,rinv22);
 559
 560             /* EWALD ELECTROSTATICS */
 561
 562             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 563             ewrt             = _mm_mul_ps(r22,ewtabscale);
 564             ewitab           = _mm_cvttps_epi32(ewrt);
 565             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 566             ewitab           = _mm_slli_epi32(ewitab,2);
 567             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 568             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 569             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 570             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 571             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 572             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 573             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 574             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
 575             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
 576
 577             /* Update potential sum for this i atom from the interaction with this j atom. */
 578             velecsum         = _mm_add_ps(velecsum,velec);
 579
 580             fscal            = felec;
 581
 582             /* Calculate temporary vectorial force */
 583             tx               = _mm_mul_ps(fscal,dx22);
 584             ty               = _mm_mul_ps(fscal,dy22);
 585             tz               = _mm_mul_ps(fscal,dz22);
 586
 587             /* Update vectorial force */
 588             fix2             = _mm_add_ps(fix2,tx);
 589             fiy2             = _mm_add_ps(fiy2,ty);
 590             fiz2             = _mm_add_ps(fiz2,tz);
 591
 592             fjx2             = _mm_add_ps(fjx2,tx);
 593             fjy2             = _mm_add_ps(fjy2,ty);
 594             fjz2             = _mm_add_ps(fjz2,tz);
 595
 596             /**************************
 597              * CALCULATE INTERACTIONS *
 598              **************************/
 599
 600             r23              = _mm_mul_ps(rsq23,rinv23);
 601
 602             /* EWALD ELECTROSTATICS */
 603
 604             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 605             ewrt             = _mm_mul_ps(r23,ewtabscale);
 606             ewitab           = _mm_cvttps_epi32(ewrt);
 607             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 608             ewitab           = _mm_slli_epi32(ewitab,2);
 609             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 610             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 611             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 612             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 613             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 614             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 615             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 616             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
 617             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
 618
 619             /* Update potential sum for this i atom from the interaction with this j atom. */
 620             velecsum         = _mm_add_ps(velecsum,velec);
 621
 622             fscal            = felec;
 623
 624             /* Calculate temporary vectorial force */
 625             tx               = _mm_mul_ps(fscal,dx23);
 626             ty               = _mm_mul_ps(fscal,dy23);
 627             tz               = _mm_mul_ps(fscal,dz23);
 628
 629             /* Update vectorial force */
 630             fix2             = _mm_add_ps(fix2,tx);
 631             fiy2             = _mm_add_ps(fiy2,ty);
 632             fiz2             = _mm_add_ps(fiz2,tz);
 633
 634             fjx3             = _mm_add_ps(fjx3,tx);
 635             fjy3             = _mm_add_ps(fjy3,ty);
 636             fjz3             = _mm_add_ps(fjz3,tz);
 637
 638             /**************************
 639              * CALCULATE INTERACTIONS *
 640              **************************/
 641
 642             r31              = _mm_mul_ps(rsq31,rinv31);
 643
 644             /* EWALD ELECTROSTATICS */
 645
 646             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 647             ewrt             = _mm_mul_ps(r31,ewtabscale);
 648             ewitab           = _mm_cvttps_epi32(ewrt);
 649             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 650             ewitab           = _mm_slli_epi32(ewitab,2);
 651             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 652             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 653             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 654             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 655             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 656             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 657             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 658             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
 659             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
 660
 661             /* Update potential sum for this i atom from the interaction with this j atom. */
 662             velecsum         = _mm_add_ps(velecsum,velec);
 663
 664             fscal            = felec;
 665
 666             /* Calculate temporary vectorial force */
 667             tx               = _mm_mul_ps(fscal,dx31);
 668             ty               = _mm_mul_ps(fscal,dy31);
 669             tz               = _mm_mul_ps(fscal,dz31);
 670
 671             /* Update vectorial force */
 672             fix3             = _mm_add_ps(fix3,tx);
 673             fiy3             = _mm_add_ps(fiy3,ty);
 674             fiz3             = _mm_add_ps(fiz3,tz);
 675
 676             fjx1             = _mm_add_ps(fjx1,tx);
 677             fjy1             = _mm_add_ps(fjy1,ty);
 678             fjz1             = _mm_add_ps(fjz1,tz);
 679
 680             /**************************
 681              * CALCULATE INTERACTIONS *
 682              **************************/
 683
 684             r32              = _mm_mul_ps(rsq32,rinv32);
 685
 686             /* EWALD ELECTROSTATICS */
 687
 688             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 689             ewrt             = _mm_mul_ps(r32,ewtabscale);
 690             ewitab           = _mm_cvttps_epi32(ewrt);
 691             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 692             ewitab           = _mm_slli_epi32(ewitab,2);
 693             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 694             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 695             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 696             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 697             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 698             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 699             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 700             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
 701             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
 702
 703             /* Update potential sum for this i atom from the interaction with this j atom. */
 704             velecsum         = _mm_add_ps(velecsum,velec);
 705
 706             fscal            = felec;
 707
 708             /* Calculate temporary vectorial force */
 709             tx               = _mm_mul_ps(fscal,dx32);
 710             ty               = _mm_mul_ps(fscal,dy32);
 711             tz               = _mm_mul_ps(fscal,dz32);
 712
 713             /* Update vectorial force */
 714             fix3             = _mm_add_ps(fix3,tx);
 715             fiy3             = _mm_add_ps(fiy3,ty);
 716             fiz3             = _mm_add_ps(fiz3,tz);
 717
 718             fjx2             = _mm_add_ps(fjx2,tx);
 719             fjy2             = _mm_add_ps(fjy2,ty);
 720             fjz2             = _mm_add_ps(fjz2,tz);
 721
 722             /**************************
 723              * CALCULATE INTERACTIONS *
 724              **************************/
 725
 726             r33              = _mm_mul_ps(rsq33,rinv33);
 727
 728             /* EWALD ELECTROSTATICS */
 729
 730             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 731             ewrt             = _mm_mul_ps(r33,ewtabscale);
 732             ewitab           = _mm_cvttps_epi32(ewrt);
 733             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 734             ewitab           = _mm_slli_epi32(ewitab,2);
 735             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 736             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 737             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 738             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 739             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 740             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 741             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 742             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
 743             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
 744
 745             /* Update potential sum for this i atom from the interaction with this j atom. */
 746             velecsum         = _mm_add_ps(velecsum,velec);
 747
 748             fscal            = felec;
 749
 750             /* Calculate temporary vectorial force */
 751             tx               = _mm_mul_ps(fscal,dx33);
 752             ty               = _mm_mul_ps(fscal,dy33);
 753             tz               = _mm_mul_ps(fscal,dz33);
 754
 755             /* Update vectorial force */
 756             fix3             = _mm_add_ps(fix3,tx);
 757             fiy3             = _mm_add_ps(fiy3,ty);
 758             fiz3             = _mm_add_ps(fiz3,tz);
 759
 760             fjx3             = _mm_add_ps(fjx3,tx);
 761             fjy3             = _mm_add_ps(fjy3,ty);
 762             fjz3             = _mm_add_ps(fjz3,tz);
 763
 764             fjptrA             = f+j_coord_offsetA;
 765             fjptrB             = f+j_coord_offsetB;
 766             fjptrC             = f+j_coord_offsetC;
 767             fjptrD             = f+j_coord_offsetD;
 768
 769             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 770                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 771                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 772
 773             /* Inner loop uses 428 flops */
 774         }
 775
 776         if(jidx<j_index_end)
 777         {
 778
 779             /* Get j neighbor index, and coordinate index */
 780             jnrlistA         = jjnr[jidx];
 781             jnrlistB         = jjnr[jidx+1];
 782             jnrlistC         = jjnr[jidx+2];
 783             jnrlistD         = jjnr[jidx+3];
 784             /* Sign of each element will be negative for non-real atoms.
 785              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 786              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 787              */
 788             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 789             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 790             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 791             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 792             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 793             j_coord_offsetA  = DIM*jnrA;
 794             j_coord_offsetB  = DIM*jnrB;
 795             j_coord_offsetC  = DIM*jnrC;
 796             j_coord_offsetD  = DIM*jnrD;
 797
 798             /* load j atom coordinates */
 799             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 800                                               x+j_coord_offsetC,x+j_coord_offsetD,
 801                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 802                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 803
 804             /* Calculate displacement vector */
 805             dx00             = _mm_sub_ps(ix0,jx0);
 806             dy00             = _mm_sub_ps(iy0,jy0);
 807             dz00             = _mm_sub_ps(iz0,jz0);
 808             dx11             = _mm_sub_ps(ix1,jx1);
 809             dy11             = _mm_sub_ps(iy1,jy1);
 810             dz11             = _mm_sub_ps(iz1,jz1);
 811             dx12             = _mm_sub_ps(ix1,jx2);
 812             dy12             = _mm_sub_ps(iy1,jy2);
 813             dz12             = _mm_sub_ps(iz1,jz2);
 814             dx13             = _mm_sub_ps(ix1,jx3);
 815             dy13             = _mm_sub_ps(iy1,jy3);
 816             dz13             = _mm_sub_ps(iz1,jz3);
 817             dx21             = _mm_sub_ps(ix2,jx1);
 818             dy21             = _mm_sub_ps(iy2,jy1);
 819             dz21             = _mm_sub_ps(iz2,jz1);
 820             dx22             = _mm_sub_ps(ix2,jx2);
 821             dy22             = _mm_sub_ps(iy2,jy2);
 822             dz22             = _mm_sub_ps(iz2,jz2);
 823             dx23             = _mm_sub_ps(ix2,jx3);
 824             dy23             = _mm_sub_ps(iy2,jy3);
 825             dz23             = _mm_sub_ps(iz2,jz3);
 826             dx31             = _mm_sub_ps(ix3,jx1);
 827             dy31             = _mm_sub_ps(iy3,jy1);
 828             dz31             = _mm_sub_ps(iz3,jz1);
 829             dx32             = _mm_sub_ps(ix3,jx2);
 830             dy32             = _mm_sub_ps(iy3,jy2);
 831             dz32             = _mm_sub_ps(iz3,jz2);
 832             dx33             = _mm_sub_ps(ix3,jx3);
 833             dy33             = _mm_sub_ps(iy3,jy3);
 834             dz33             = _mm_sub_ps(iz3,jz3);
 835
 836             /* Calculate squared distance and things based on it */
 837             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 838             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 839             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 840             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 841             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 842             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 843             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 844             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 845             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 846             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 847
 848             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 849             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 850             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 851             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 852             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 853             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 854             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 855             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 856             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 857             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 858
 859             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 860             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 861             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 862             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 863             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 864             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 865             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 866             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 867             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 868
 869             fjx0             = _mm_setzero_ps();
 870             fjy0             = _mm_setzero_ps();
 871             fjz0             = _mm_setzero_ps();
 872             fjx1             = _mm_setzero_ps();
 873             fjy1             = _mm_setzero_ps();
 874             fjz1             = _mm_setzero_ps();
 875             fjx2             = _mm_setzero_ps();
 876             fjy2             = _mm_setzero_ps();
 877             fjz2             = _mm_setzero_ps();
 878             fjx3             = _mm_setzero_ps();
 879             fjy3             = _mm_setzero_ps();
 880             fjz3             = _mm_setzero_ps();
 881
 882             /**************************
 883              * CALCULATE INTERACTIONS *
 884              **************************/
 885
 886             r00              = _mm_mul_ps(rsq00,rinv00);
 887             r00              = _mm_andnot_ps(dummy_mask,r00);
 888
 889             /* Calculate table index by multiplying r with table scale and truncate to integer */
 890             rt               = _mm_mul_ps(r00,vftabscale);
 891             vfitab           = _mm_cvttps_epi32(rt);
 892             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 893             vfitab           = _mm_slli_epi32(vfitab,3);
 894
 895             /* CUBIC SPLINE TABLE DISPERSION */
 896             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 897             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 898             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 899             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 900             _MM_TRANSPOSE4_PS(Y,F,G,H);
 901             Heps             = _mm_mul_ps(vfeps,H);
 902             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 903             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 904             vvdw6            = _mm_mul_ps(c6_00,VV);
 905             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 906             fvdw6            = _mm_mul_ps(c6_00,FF);
 907
 908             /* CUBIC SPLINE TABLE REPULSION */
 909             vfitab           = _mm_add_epi32(vfitab,ifour);
 910             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 911             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 912             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 913             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 914             _MM_TRANSPOSE4_PS(Y,F,G,H);
 915             Heps             = _mm_mul_ps(vfeps,H);
 916             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 917             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 918             vvdw12           = _mm_mul_ps(c12_00,VV);
 919             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 920             fvdw12           = _mm_mul_ps(c12_00,FF);
 921             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 922             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 923
 924             /* Update potential sum for this i atom from the interaction with this j atom. */
 925             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 926             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 927
 928             fscal            = fvdw;
 929
 930             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 931
 932             /* Calculate temporary vectorial force */
 933             tx               = _mm_mul_ps(fscal,dx00);
 934             ty               = _mm_mul_ps(fscal,dy00);
 935             tz               = _mm_mul_ps(fscal,dz00);
 936
 937             /* Update vectorial force */
 938             fix0             = _mm_add_ps(fix0,tx);
 939             fiy0             = _mm_add_ps(fiy0,ty);
 940             fiz0             = _mm_add_ps(fiz0,tz);
 941
 942             fjx0             = _mm_add_ps(fjx0,tx);
 943             fjy0             = _mm_add_ps(fjy0,ty);
 944             fjz0             = _mm_add_ps(fjz0,tz);
 945
 946             /**************************
 947              * CALCULATE INTERACTIONS *
 948              **************************/
 949
 950             r11              = _mm_mul_ps(rsq11,rinv11);
 951             r11              = _mm_andnot_ps(dummy_mask,r11);
 952
 953             /* EWALD ELECTROSTATICS */
 954
 955             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 956             ewrt             = _mm_mul_ps(r11,ewtabscale);
 957             ewitab           = _mm_cvttps_epi32(ewrt);
 958             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 959             ewitab           = _mm_slli_epi32(ewitab,2);
 960             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 961             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 962             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 963             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 964             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 965             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 966             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 967             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 968             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 969
 970             /* Update potential sum for this i atom from the interaction with this j atom. */
 971             velec            = _mm_andnot_ps(dummy_mask,velec);
 972             velecsum         = _mm_add_ps(velecsum,velec);
 973
 974             fscal            = felec;
 975
 976             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 977
 978             /* Calculate temporary vectorial force */
 979             tx               = _mm_mul_ps(fscal,dx11);
 980             ty               = _mm_mul_ps(fscal,dy11);
 981             tz               = _mm_mul_ps(fscal,dz11);
 982
 983             /* Update vectorial force */
 984             fix1             = _mm_add_ps(fix1,tx);
 985             fiy1             = _mm_add_ps(fiy1,ty);
 986             fiz1             = _mm_add_ps(fiz1,tz);
 987
 988             fjx1             = _mm_add_ps(fjx1,tx);
 989             fjy1             = _mm_add_ps(fjy1,ty);
 990             fjz1             = _mm_add_ps(fjz1,tz);
 991
 992             /**************************
 993              * CALCULATE INTERACTIONS *
 994              **************************/
 995
 996             r12              = _mm_mul_ps(rsq12,rinv12);
 997             r12              = _mm_andnot_ps(dummy_mask,r12);
 998
 999             /* EWALD ELECTROSTATICS */
1000
1001             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1002             ewrt             = _mm_mul_ps(r12,ewtabscale);
1003             ewitab           = _mm_cvttps_epi32(ewrt);
1004             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1005             ewitab           = _mm_slli_epi32(ewitab,2);
1006             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1007             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1008             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1009             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1010             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1011             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1012             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1013             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
1014             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1015
1016             /* Update potential sum for this i atom from the interaction with this j atom. */
1017             velec            = _mm_andnot_ps(dummy_mask,velec);
1018             velecsum         = _mm_add_ps(velecsum,velec);
1019
1020             fscal            = felec;
1021
1022             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1023
1024             /* Calculate temporary vectorial force */
1025             tx               = _mm_mul_ps(fscal,dx12);
1026             ty               = _mm_mul_ps(fscal,dy12);
1027             tz               = _mm_mul_ps(fscal,dz12);
1028
1029             /* Update vectorial force */
1030             fix1             = _mm_add_ps(fix1,tx);
1031             fiy1             = _mm_add_ps(fiy1,ty);
1032             fiz1             = _mm_add_ps(fiz1,tz);
1033
1034             fjx2             = _mm_add_ps(fjx2,tx);
1035             fjy2             = _mm_add_ps(fjy2,ty);
1036             fjz2             = _mm_add_ps(fjz2,tz);
1037
1038             /**************************
1039              * CALCULATE INTERACTIONS *
1040              **************************/
1041
1042             r13              = _mm_mul_ps(rsq13,rinv13);
1043             r13              = _mm_andnot_ps(dummy_mask,r13);
1044
1045             /* EWALD ELECTROSTATICS */
1046
1047             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1048             ewrt             = _mm_mul_ps(r13,ewtabscale);
1049             ewitab           = _mm_cvttps_epi32(ewrt);
1050             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1051             ewitab           = _mm_slli_epi32(ewitab,2);
1052             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1053             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1054             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1055             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1056             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1057             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1058             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1059             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
1060             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1061
1062             /* Update potential sum for this i atom from the interaction with this j atom. */
1063             velec            = _mm_andnot_ps(dummy_mask,velec);
1064             velecsum         = _mm_add_ps(velecsum,velec);
1065
1066             fscal            = felec;
1067
1068             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1069
1070             /* Calculate temporary vectorial force */
1071             tx               = _mm_mul_ps(fscal,dx13);
1072             ty               = _mm_mul_ps(fscal,dy13);
1073             tz               = _mm_mul_ps(fscal,dz13);
1074
1075             /* Update vectorial force */
1076             fix1             = _mm_add_ps(fix1,tx);
1077             fiy1             = _mm_add_ps(fiy1,ty);
1078             fiz1             = _mm_add_ps(fiz1,tz);
1079
1080             fjx3             = _mm_add_ps(fjx3,tx);
1081             fjy3             = _mm_add_ps(fjy3,ty);
1082             fjz3             = _mm_add_ps(fjz3,tz);
1083
1084             /**************************
1085              * CALCULATE INTERACTIONS *
1086              **************************/
1087
1088             r21              = _mm_mul_ps(rsq21,rinv21);
1089             r21              = _mm_andnot_ps(dummy_mask,r21);
1090
1091             /* EWALD ELECTROSTATICS */
1092
1093             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1094             ewrt             = _mm_mul_ps(r21,ewtabscale);
1095             ewitab           = _mm_cvttps_epi32(ewrt);
1096             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1097             ewitab           = _mm_slli_epi32(ewitab,2);
1098             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1099             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1100             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1101             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1102             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1103             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1104             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1105             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
1106             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1107
1108             /* Update potential sum for this i atom from the interaction with this j atom. */
1109             velec            = _mm_andnot_ps(dummy_mask,velec);
1110             velecsum         = _mm_add_ps(velecsum,velec);
1111
1112             fscal            = felec;
1113
1114             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1115
1116             /* Calculate temporary vectorial force */
1117             tx               = _mm_mul_ps(fscal,dx21);
1118             ty               = _mm_mul_ps(fscal,dy21);
1119             tz               = _mm_mul_ps(fscal,dz21);
1120
1121             /* Update vectorial force */
1122             fix2             = _mm_add_ps(fix2,tx);
1123             fiy2             = _mm_add_ps(fiy2,ty);
1124             fiz2             = _mm_add_ps(fiz2,tz);
1125
1126             fjx1             = _mm_add_ps(fjx1,tx);
1127             fjy1             = _mm_add_ps(fjy1,ty);
1128             fjz1             = _mm_add_ps(fjz1,tz);
1129
1130             /**************************
1131              * CALCULATE INTERACTIONS *
1132              **************************/
1133
1134             r22              = _mm_mul_ps(rsq22,rinv22);
1135             r22              = _mm_andnot_ps(dummy_mask,r22);
1136
1137             /* EWALD ELECTROSTATICS */
1138
1139             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1140             ewrt             = _mm_mul_ps(r22,ewtabscale);
1141             ewitab           = _mm_cvttps_epi32(ewrt);
1142             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1143             ewitab           = _mm_slli_epi32(ewitab,2);
1144             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1145             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1146             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1147             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1148             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1149             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1150             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1151             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
1152             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1153
1154             /* Update potential sum for this i atom from the interaction with this j atom. */
1155             velec            = _mm_andnot_ps(dummy_mask,velec);
1156             velecsum         = _mm_add_ps(velecsum,velec);
1157
1158             fscal            = felec;
1159
1160             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1161
1162             /* Calculate temporary vectorial force */
1163             tx               = _mm_mul_ps(fscal,dx22);
1164             ty               = _mm_mul_ps(fscal,dy22);
1165             tz               = _mm_mul_ps(fscal,dz22);
1166
1167             /* Update vectorial force */
1168             fix2             = _mm_add_ps(fix2,tx);
1169             fiy2             = _mm_add_ps(fiy2,ty);
1170             fiz2             = _mm_add_ps(fiz2,tz);
1171
1172             fjx2             = _mm_add_ps(fjx2,tx);
1173             fjy2             = _mm_add_ps(fjy2,ty);
1174             fjz2             = _mm_add_ps(fjz2,tz);
1175
1176             /**************************
1177              * CALCULATE INTERACTIONS *
1178              **************************/
1179
1180             r23              = _mm_mul_ps(rsq23,rinv23);
1181             r23              = _mm_andnot_ps(dummy_mask,r23);
1182
1183             /* EWALD ELECTROSTATICS */
1184
1185             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1186             ewrt             = _mm_mul_ps(r23,ewtabscale);
1187             ewitab           = _mm_cvttps_epi32(ewrt);
1188             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1189             ewitab           = _mm_slli_epi32(ewitab,2);
1190             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1191             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1192             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1193             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1194             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1195             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1196             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1197             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
1198             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1199
1200             /* Update potential sum for this i atom from the interaction with this j atom. */
1201             velec            = _mm_andnot_ps(dummy_mask,velec);
1202             velecsum         = _mm_add_ps(velecsum,velec);
1203
1204             fscal            = felec;
1205
1206             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1207
1208             /* Calculate temporary vectorial force */
1209             tx               = _mm_mul_ps(fscal,dx23);
1210             ty               = _mm_mul_ps(fscal,dy23);
1211             tz               = _mm_mul_ps(fscal,dz23);
1212
1213             /* Update vectorial force */
1214             fix2             = _mm_add_ps(fix2,tx);
1215             fiy2             = _mm_add_ps(fiy2,ty);
1216             fiz2             = _mm_add_ps(fiz2,tz);
1217
1218             fjx3             = _mm_add_ps(fjx3,tx);
1219             fjy3             = _mm_add_ps(fjy3,ty);
1220             fjz3             = _mm_add_ps(fjz3,tz);
1221
1222             /**************************
1223              * CALCULATE INTERACTIONS *
1224              **************************/
1225
1226             r31              = _mm_mul_ps(rsq31,rinv31);
1227             r31              = _mm_andnot_ps(dummy_mask,r31);
1228
1229             /* EWALD ELECTROSTATICS */
1230
1231             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1232             ewrt             = _mm_mul_ps(r31,ewtabscale);
1233             ewitab           = _mm_cvttps_epi32(ewrt);
1234             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1235             ewitab           = _mm_slli_epi32(ewitab,2);
1236             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1237             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1238             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1239             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1240             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1241             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1242             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1243             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
1244             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1245
1246             /* Update potential sum for this i atom from the interaction with this j atom. */
1247             velec            = _mm_andnot_ps(dummy_mask,velec);
1248             velecsum         = _mm_add_ps(velecsum,velec);
1249
1250             fscal            = felec;
1251
1252             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1253
1254             /* Calculate temporary vectorial force */
1255             tx               = _mm_mul_ps(fscal,dx31);
1256             ty               = _mm_mul_ps(fscal,dy31);
1257             tz               = _mm_mul_ps(fscal,dz31);
1258
1259             /* Update vectorial force */
1260             fix3             = _mm_add_ps(fix3,tx);
1261             fiy3             = _mm_add_ps(fiy3,ty);
1262             fiz3             = _mm_add_ps(fiz3,tz);
1263
1264             fjx1             = _mm_add_ps(fjx1,tx);
1265             fjy1             = _mm_add_ps(fjy1,ty);
1266             fjz1             = _mm_add_ps(fjz1,tz);
1267
1268             /**************************
1269              * CALCULATE INTERACTIONS *
1270              **************************/
1271
1272             r32              = _mm_mul_ps(rsq32,rinv32);
1273             r32              = _mm_andnot_ps(dummy_mask,r32);
1274
1275             /* EWALD ELECTROSTATICS */
1276
1277             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1278             ewrt             = _mm_mul_ps(r32,ewtabscale);
1279             ewitab           = _mm_cvttps_epi32(ewrt);
1280             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1281             ewitab           = _mm_slli_epi32(ewitab,2);
1282             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1283             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1284             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1285             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1286             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1287             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1288             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1289             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
1290             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1291
1292             /* Update potential sum for this i atom from the interaction with this j atom. */
1293             velec            = _mm_andnot_ps(dummy_mask,velec);
1294             velecsum         = _mm_add_ps(velecsum,velec);
1295
1296             fscal            = felec;
1297
1298             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1299
1300             /* Calculate temporary vectorial force */
1301             tx               = _mm_mul_ps(fscal,dx32);
1302             ty               = _mm_mul_ps(fscal,dy32);
1303             tz               = _mm_mul_ps(fscal,dz32);
1304
1305             /* Update vectorial force */
1306             fix3             = _mm_add_ps(fix3,tx);
1307             fiy3             = _mm_add_ps(fiy3,ty);
1308             fiz3             = _mm_add_ps(fiz3,tz);
1309
1310             fjx2             = _mm_add_ps(fjx2,tx);
1311             fjy2             = _mm_add_ps(fjy2,ty);
1312             fjz2             = _mm_add_ps(fjz2,tz);
1313
1314             /**************************
1315              * CALCULATE INTERACTIONS *
1316              **************************/
1317
1318             r33              = _mm_mul_ps(rsq33,rinv33);
1319             r33              = _mm_andnot_ps(dummy_mask,r33);
1320
1321             /* EWALD ELECTROSTATICS */
1322
1323             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1324             ewrt             = _mm_mul_ps(r33,ewtabscale);
1325             ewitab           = _mm_cvttps_epi32(ewrt);
1326             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1327             ewitab           = _mm_slli_epi32(ewitab,2);
1328             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1329             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1330             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1331             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1332             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1333             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1334             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1335             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
1336             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1337
1338             /* Update potential sum for this i atom from the interaction with this j atom. */
1339             velec            = _mm_andnot_ps(dummy_mask,velec);
1340             velecsum         = _mm_add_ps(velecsum,velec);
1341
1342             fscal            = felec;
1343
1344             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1345
1346             /* Calculate temporary vectorial force */
1347             tx               = _mm_mul_ps(fscal,dx33);
1348             ty               = _mm_mul_ps(fscal,dy33);
1349             tz               = _mm_mul_ps(fscal,dz33);
1350
1351             /* Update vectorial force */
1352             fix3             = _mm_add_ps(fix3,tx);
1353             fiy3             = _mm_add_ps(fiy3,ty);
1354             fiz3             = _mm_add_ps(fiz3,tz);
1355
1356             fjx3             = _mm_add_ps(fjx3,tx);
1357             fjy3             = _mm_add_ps(fjy3,ty);
1358             fjz3             = _mm_add_ps(fjz3,tz);
1359
1360             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1361             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1362             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1363             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1364
1365             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1366                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1367                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1368
1369             /* Inner loop uses 438 flops */
1370         }
1371
1372         /* End of innermost loop */
1373
1374         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1375                                               f+i_coord_offset,fshift+i_shift_offset);
1376
1377         ggid                        = gid[iidx];
1378         /* Update potential energies */
1379         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1380         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1381
1382         /* Increment number of inner iterations */
1383         inneriter                  += j_index_end - j_index_start;
1384
1385         /* Outer loop uses 26 flops */
1386     }
1387
1388     /* Increment number of outer iterations */
1389     outeriter        += nri;
1390
1391     /* Update outer/inner flops */
1392
1393     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*438);
1394 }
1395 /*
1396  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse4_1_single
1397  * Electrostatics interaction: Ewald
1398  * VdW interaction:            CubicSplineTable
1399  * Geometry:                   Water4-Water4
1400  * Calculate force/pot:        Force
1401  */
1402 void
1403 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse4_1_single
1404                     (t_nblist                    * gmx_restrict       nlist,
1405                      rvec                        * gmx_restrict          xx,
1406                      rvec                        * gmx_restrict          ff,
1407                      t_forcerec                  * gmx_restrict          fr,
1408                      t_mdatoms                   * gmx_restrict     mdatoms,
1409                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1410                      t_nrnb                      * gmx_restrict        nrnb)
1411 {
1412     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1413      * just 0 for non-waters.
1414      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1415      * jnr indices corresponding to data put in the four positions in the SIMD register.
1416      */
1417     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1418     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1419     int              jnrA,jnrB,jnrC,jnrD;
1420     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1421     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1422     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1423     real             rcutoff_scalar;
1424     real             *shiftvec,*fshift,*x,*f;
1425     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1426     real             scratch[4*DIM];
1427     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1428     int              vdwioffset0;
1429     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1430     int              vdwioffset1;
1431     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1432     int              vdwioffset2;
1433     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1434     int              vdwioffset3;
1435     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1436     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1437     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1438     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1439     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1440     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1441     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1442     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1443     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1444     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1445     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1446     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1447     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1448     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1449     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1450     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1451     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1452     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1453     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1454     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1455     real             *charge;
1456     int              nvdwtype;
1457     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1458     int              *vdwtype;
1459     real             *vdwparam;
1460     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1461     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1462     __m128i          vfitab;
1463     __m128i          ifour       = _mm_set1_epi32(4);
1464     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1465     real             *vftab;
1466     __m128i          ewitab;
1467     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1468     real             *ewtab;
1469     __m128           dummy_mask,cutoff_mask;
1470     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1471     __m128           one     = _mm_set1_ps(1.0);
1472     __m128           two     = _mm_set1_ps(2.0);
1473     x                = xx[0];
1474     f                = ff[0];
1475
1476     nri              = nlist->nri;
1477     iinr             = nlist->iinr;
1478     jindex           = nlist->jindex;
1479     jjnr             = nlist->jjnr;
1480     shiftidx         = nlist->shift;
1481     gid              = nlist->gid;
1482     shiftvec         = fr->shift_vec[0];
1483     fshift           = fr->fshift[0];
1484     facel            = _mm_set1_ps(fr->epsfac);
1485     charge           = mdatoms->chargeA;
1486     nvdwtype         = fr->ntype;
1487     vdwparam         = fr->nbfp;
1488     vdwtype          = mdatoms->typeA;
1489
1490     vftab            = kernel_data->table_vdw->data;
1491     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
1492
1493     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
1494     ewtab            = fr->ic->tabq_coul_F;
1495     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
1496     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1497
1498     /* Setup water-specific parameters */
1499     inr              = nlist->iinr[0];
1500     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1501     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1502     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1503     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1504
1505     jq1              = _mm_set1_ps(charge[inr+1]);
1506     jq2              = _mm_set1_ps(charge[inr+2]);
1507     jq3              = _mm_set1_ps(charge[inr+3]);
1508     vdwjidx0A        = 2*vdwtype[inr+0];
1509     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1510     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1511     qq11             = _mm_mul_ps(iq1,jq1);
1512     qq12             = _mm_mul_ps(iq1,jq2);
1513     qq13             = _mm_mul_ps(iq1,jq3);
1514     qq21             = _mm_mul_ps(iq2,jq1);
1515     qq22             = _mm_mul_ps(iq2,jq2);
1516     qq23             = _mm_mul_ps(iq2,jq3);
1517     qq31             = _mm_mul_ps(iq3,jq1);
1518     qq32             = _mm_mul_ps(iq3,jq2);
1519     qq33             = _mm_mul_ps(iq3,jq3);
1520
1521     /* Avoid stupid compiler warnings */
1522     jnrA = jnrB = jnrC = jnrD = 0;
1523     j_coord_offsetA = 0;
1524     j_coord_offsetB = 0;
1525     j_coord_offsetC = 0;
1526     j_coord_offsetD = 0;
1527
1528     outeriter        = 0;
1529     inneriter        = 0;
1530
1531     for(iidx=0;iidx<4*DIM;iidx++)
1532     {
1533         scratch[iidx] = 0.0;
1534     }
1535
1536     /* Start outer loop over neighborlists */
1537     for(iidx=0; iidx<nri; iidx++)
1538     {
1539         /* Load shift vector for this list */
1540         i_shift_offset   = DIM*shiftidx[iidx];
1541
1542         /* Load limits for loop over neighbors */
1543         j_index_start    = jindex[iidx];
1544         j_index_end      = jindex[iidx+1];
1545
1546         /* Get outer coordinate index */
1547         inr              = iinr[iidx];
1548         i_coord_offset   = DIM*inr;
1549
1550         /* Load i particle coords and add shift vector */
1551         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1552                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1553
1554         fix0             = _mm_setzero_ps();
1555         fiy0             = _mm_setzero_ps();
1556         fiz0             = _mm_setzero_ps();
1557         fix1             = _mm_setzero_ps();
1558         fiy1             = _mm_setzero_ps();
1559         fiz1             = _mm_setzero_ps();
1560         fix2             = _mm_setzero_ps();
1561         fiy2             = _mm_setzero_ps();
1562         fiz2             = _mm_setzero_ps();
1563         fix3             = _mm_setzero_ps();
1564         fiy3             = _mm_setzero_ps();
1565         fiz3             = _mm_setzero_ps();
1566
1567         /* Start inner kernel loop */
1568         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1569         {
1570
1571             /* Get j neighbor index, and coordinate index */
1572             jnrA             = jjnr[jidx];
1573             jnrB             = jjnr[jidx+1];
1574             jnrC             = jjnr[jidx+2];
1575             jnrD             = jjnr[jidx+3];
1576             j_coord_offsetA  = DIM*jnrA;
1577             j_coord_offsetB  = DIM*jnrB;
1578             j_coord_offsetC  = DIM*jnrC;
1579             j_coord_offsetD  = DIM*jnrD;
1580
1581             /* load j atom coordinates */
1582             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1583                                               x+j_coord_offsetC,x+j_coord_offsetD,
1584                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1585                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1586
1587             /* Calculate displacement vector */
1588             dx00             = _mm_sub_ps(ix0,jx0);
1589             dy00             = _mm_sub_ps(iy0,jy0);
1590             dz00             = _mm_sub_ps(iz0,jz0);
1591             dx11             = _mm_sub_ps(ix1,jx1);
1592             dy11             = _mm_sub_ps(iy1,jy1);
1593             dz11             = _mm_sub_ps(iz1,jz1);
1594             dx12             = _mm_sub_ps(ix1,jx2);
1595             dy12             = _mm_sub_ps(iy1,jy2);
1596             dz12             = _mm_sub_ps(iz1,jz2);
1597             dx13             = _mm_sub_ps(ix1,jx3);
1598             dy13             = _mm_sub_ps(iy1,jy3);
1599             dz13             = _mm_sub_ps(iz1,jz3);
1600             dx21             = _mm_sub_ps(ix2,jx1);
1601             dy21             = _mm_sub_ps(iy2,jy1);
1602             dz21             = _mm_sub_ps(iz2,jz1);
1603             dx22             = _mm_sub_ps(ix2,jx2);
1604             dy22             = _mm_sub_ps(iy2,jy2);
1605             dz22             = _mm_sub_ps(iz2,jz2);
1606             dx23             = _mm_sub_ps(ix2,jx3);
1607             dy23             = _mm_sub_ps(iy2,jy3);
1608             dz23             = _mm_sub_ps(iz2,jz3);
1609             dx31             = _mm_sub_ps(ix3,jx1);
1610             dy31             = _mm_sub_ps(iy3,jy1);
1611             dz31             = _mm_sub_ps(iz3,jz1);
1612             dx32             = _mm_sub_ps(ix3,jx2);
1613             dy32             = _mm_sub_ps(iy3,jy2);
1614             dz32             = _mm_sub_ps(iz3,jz2);
1615             dx33             = _mm_sub_ps(ix3,jx3);
1616             dy33             = _mm_sub_ps(iy3,jy3);
1617             dz33             = _mm_sub_ps(iz3,jz3);
1618
1619             /* Calculate squared distance and things based on it */
1620             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1621             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1622             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1623             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1624             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1625             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1626             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1627             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1628             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1629             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1630
1631             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1632             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1633             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1634             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1635             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1636             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1637             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1638             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1639             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1640             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1641
1642             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1643             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1644             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1645             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1646             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1647             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1648             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1649             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1650             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1651
1652             fjx0             = _mm_setzero_ps();
1653             fjy0             = _mm_setzero_ps();
1654             fjz0             = _mm_setzero_ps();
1655             fjx1             = _mm_setzero_ps();
1656             fjy1             = _mm_setzero_ps();
1657             fjz1             = _mm_setzero_ps();
1658             fjx2             = _mm_setzero_ps();
1659             fjy2             = _mm_setzero_ps();
1660             fjz2             = _mm_setzero_ps();
1661             fjx3             = _mm_setzero_ps();
1662             fjy3             = _mm_setzero_ps();
1663             fjz3             = _mm_setzero_ps();
1664
1665             /**************************
1666              * CALCULATE INTERACTIONS *
1667              **************************/
1668
1669             r00              = _mm_mul_ps(rsq00,rinv00);
1670
1671             /* Calculate table index by multiplying r with table scale and truncate to integer */
1672             rt               = _mm_mul_ps(r00,vftabscale);
1673             vfitab           = _mm_cvttps_epi32(rt);
1674             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1675             vfitab           = _mm_slli_epi32(vfitab,3);
1676
1677             /* CUBIC SPLINE TABLE DISPERSION */
1678             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1679             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1680             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1681             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1682             _MM_TRANSPOSE4_PS(Y,F,G,H);
1683             Heps             = _mm_mul_ps(vfeps,H);
1684             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1685             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1686             fvdw6            = _mm_mul_ps(c6_00,FF);
1687
1688             /* CUBIC SPLINE TABLE REPULSION */
1689             vfitab           = _mm_add_epi32(vfitab,ifour);
1690             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1691             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1692             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1693             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1694             _MM_TRANSPOSE4_PS(Y,F,G,H);
1695             Heps             = _mm_mul_ps(vfeps,H);
1696             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1697             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1698             fvdw12           = _mm_mul_ps(c12_00,FF);
1699             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1700
1701             fscal            = fvdw;
1702
1703             /* Calculate temporary vectorial force */
1704             tx               = _mm_mul_ps(fscal,dx00);
1705             ty               = _mm_mul_ps(fscal,dy00);
1706             tz               = _mm_mul_ps(fscal,dz00);
1707
1708             /* Update vectorial force */
1709             fix0             = _mm_add_ps(fix0,tx);
1710             fiy0             = _mm_add_ps(fiy0,ty);
1711             fiz0             = _mm_add_ps(fiz0,tz);
1712
1713             fjx0             = _mm_add_ps(fjx0,tx);
1714             fjy0             = _mm_add_ps(fjy0,ty);
1715             fjz0             = _mm_add_ps(fjz0,tz);
1716
1717             /**************************
1718              * CALCULATE INTERACTIONS *
1719              **************************/
1720
1721             r11              = _mm_mul_ps(rsq11,rinv11);
1722
1723             /* EWALD ELECTROSTATICS */
1724
1725             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1726             ewrt             = _mm_mul_ps(r11,ewtabscale);
1727             ewitab           = _mm_cvttps_epi32(ewrt);
1728             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1729             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1730                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1731                                          &ewtabF,&ewtabFn);
1732             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1733             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1734
1735             fscal            = felec;
1736
1737             /* Calculate temporary vectorial force */
1738             tx               = _mm_mul_ps(fscal,dx11);
1739             ty               = _mm_mul_ps(fscal,dy11);
1740             tz               = _mm_mul_ps(fscal,dz11);
1741
1742             /* Update vectorial force */
1743             fix1             = _mm_add_ps(fix1,tx);
1744             fiy1             = _mm_add_ps(fiy1,ty);
1745             fiz1             = _mm_add_ps(fiz1,tz);
1746
1747             fjx1             = _mm_add_ps(fjx1,tx);
1748             fjy1             = _mm_add_ps(fjy1,ty);
1749             fjz1             = _mm_add_ps(fjz1,tz);
1750
1751             /**************************
1752              * CALCULATE INTERACTIONS *
1753              **************************/
1754
1755             r12              = _mm_mul_ps(rsq12,rinv12);
1756
1757             /* EWALD ELECTROSTATICS */
1758
1759             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1760             ewrt             = _mm_mul_ps(r12,ewtabscale);
1761             ewitab           = _mm_cvttps_epi32(ewrt);
1762             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1763             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1764                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1765                                          &ewtabF,&ewtabFn);
1766             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1767             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1768
1769             fscal            = felec;
1770
1771             /* Calculate temporary vectorial force */
1772             tx               = _mm_mul_ps(fscal,dx12);
1773             ty               = _mm_mul_ps(fscal,dy12);
1774             tz               = _mm_mul_ps(fscal,dz12);
1775
1776             /* Update vectorial force */
1777             fix1             = _mm_add_ps(fix1,tx);
1778             fiy1             = _mm_add_ps(fiy1,ty);
1779             fiz1             = _mm_add_ps(fiz1,tz);
1780
1781             fjx2             = _mm_add_ps(fjx2,tx);
1782             fjy2             = _mm_add_ps(fjy2,ty);
1783             fjz2             = _mm_add_ps(fjz2,tz);
1784
1785             /**************************
1786              * CALCULATE INTERACTIONS *
1787              **************************/
1788
1789             r13              = _mm_mul_ps(rsq13,rinv13);
1790
1791             /* EWALD ELECTROSTATICS */
1792
1793             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1794             ewrt             = _mm_mul_ps(r13,ewtabscale);
1795             ewitab           = _mm_cvttps_epi32(ewrt);
1796             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1797             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1798                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1799                                          &ewtabF,&ewtabFn);
1800             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1801             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1802
1803             fscal            = felec;
1804
1805             /* Calculate temporary vectorial force */
1806             tx               = _mm_mul_ps(fscal,dx13);
1807             ty               = _mm_mul_ps(fscal,dy13);
1808             tz               = _mm_mul_ps(fscal,dz13);
1809
1810             /* Update vectorial force */
1811             fix1             = _mm_add_ps(fix1,tx);
1812             fiy1             = _mm_add_ps(fiy1,ty);
1813             fiz1             = _mm_add_ps(fiz1,tz);
1814
1815             fjx3             = _mm_add_ps(fjx3,tx);
1816             fjy3             = _mm_add_ps(fjy3,ty);
1817             fjz3             = _mm_add_ps(fjz3,tz);
1818
1819             /**************************
1820              * CALCULATE INTERACTIONS *
1821              **************************/
1822
1823             r21              = _mm_mul_ps(rsq21,rinv21);
1824
1825             /* EWALD ELECTROSTATICS */
1826
1827             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1828             ewrt             = _mm_mul_ps(r21,ewtabscale);
1829             ewitab           = _mm_cvttps_epi32(ewrt);
1830             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1831             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1832                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1833                                          &ewtabF,&ewtabFn);
1834             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1835             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1836
1837             fscal            = felec;
1838
1839             /* Calculate temporary vectorial force */
1840             tx               = _mm_mul_ps(fscal,dx21);
1841             ty               = _mm_mul_ps(fscal,dy21);
1842             tz               = _mm_mul_ps(fscal,dz21);
1843
1844             /* Update vectorial force */
1845             fix2             = _mm_add_ps(fix2,tx);
1846             fiy2             = _mm_add_ps(fiy2,ty);
1847             fiz2             = _mm_add_ps(fiz2,tz);
1848
1849             fjx1             = _mm_add_ps(fjx1,tx);
1850             fjy1             = _mm_add_ps(fjy1,ty);
1851             fjz1             = _mm_add_ps(fjz1,tz);
1852
1853             /**************************
1854              * CALCULATE INTERACTIONS *
1855              **************************/
1856
1857             r22              = _mm_mul_ps(rsq22,rinv22);
1858
1859             /* EWALD ELECTROSTATICS */
1860
1861             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1862             ewrt             = _mm_mul_ps(r22,ewtabscale);
1863             ewitab           = _mm_cvttps_epi32(ewrt);
1864             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1865             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1866                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1867                                          &ewtabF,&ewtabFn);
1868             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1869             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1870
1871             fscal            = felec;
1872
1873             /* Calculate temporary vectorial force */
1874             tx               = _mm_mul_ps(fscal,dx22);
1875             ty               = _mm_mul_ps(fscal,dy22);
1876             tz               = _mm_mul_ps(fscal,dz22);
1877
1878             /* Update vectorial force */
1879             fix2             = _mm_add_ps(fix2,tx);
1880             fiy2             = _mm_add_ps(fiy2,ty);
1881             fiz2             = _mm_add_ps(fiz2,tz);
1882
1883             fjx2             = _mm_add_ps(fjx2,tx);
1884             fjy2             = _mm_add_ps(fjy2,ty);
1885             fjz2             = _mm_add_ps(fjz2,tz);
1886
1887             /**************************
1888              * CALCULATE INTERACTIONS *
1889              **************************/
1890
1891             r23              = _mm_mul_ps(rsq23,rinv23);
1892
1893             /* EWALD ELECTROSTATICS */
1894
1895             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1896             ewrt             = _mm_mul_ps(r23,ewtabscale);
1897             ewitab           = _mm_cvttps_epi32(ewrt);
1898             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1899             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1900                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1901                                          &ewtabF,&ewtabFn);
1902             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1903             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1904
1905             fscal            = felec;
1906
1907             /* Calculate temporary vectorial force */
1908             tx               = _mm_mul_ps(fscal,dx23);
1909             ty               = _mm_mul_ps(fscal,dy23);
1910             tz               = _mm_mul_ps(fscal,dz23);
1911
1912             /* Update vectorial force */
1913             fix2             = _mm_add_ps(fix2,tx);
1914             fiy2             = _mm_add_ps(fiy2,ty);
1915             fiz2             = _mm_add_ps(fiz2,tz);
1916
1917             fjx3             = _mm_add_ps(fjx3,tx);
1918             fjy3             = _mm_add_ps(fjy3,ty);
1919             fjz3             = _mm_add_ps(fjz3,tz);
1920
1921             /**************************
1922              * CALCULATE INTERACTIONS *
1923              **************************/
1924
1925             r31              = _mm_mul_ps(rsq31,rinv31);
1926
1927             /* EWALD ELECTROSTATICS */
1928
1929             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1930             ewrt             = _mm_mul_ps(r31,ewtabscale);
1931             ewitab           = _mm_cvttps_epi32(ewrt);
1932             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1933             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1934                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1935                                          &ewtabF,&ewtabFn);
1936             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1937             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1938
1939             fscal            = felec;
1940
1941             /* Calculate temporary vectorial force */
1942             tx               = _mm_mul_ps(fscal,dx31);
1943             ty               = _mm_mul_ps(fscal,dy31);
1944             tz               = _mm_mul_ps(fscal,dz31);
1945
1946             /* Update vectorial force */
1947             fix3             = _mm_add_ps(fix3,tx);
1948             fiy3             = _mm_add_ps(fiy3,ty);
1949             fiz3             = _mm_add_ps(fiz3,tz);
1950
1951             fjx1             = _mm_add_ps(fjx1,tx);
1952             fjy1             = _mm_add_ps(fjy1,ty);
1953             fjz1             = _mm_add_ps(fjz1,tz);
1954
1955             /**************************
1956              * CALCULATE INTERACTIONS *
1957              **************************/
1958
1959             r32              = _mm_mul_ps(rsq32,rinv32);
1960
1961             /* EWALD ELECTROSTATICS */
1962
1963             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1964             ewrt             = _mm_mul_ps(r32,ewtabscale);
1965             ewitab           = _mm_cvttps_epi32(ewrt);
1966             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1967             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1968                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1969                                          &ewtabF,&ewtabFn);
1970             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1971             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1972
1973             fscal            = felec;
1974
1975             /* Calculate temporary vectorial force */
1976             tx               = _mm_mul_ps(fscal,dx32);
1977             ty               = _mm_mul_ps(fscal,dy32);
1978             tz               = _mm_mul_ps(fscal,dz32);
1979
1980             /* Update vectorial force */
1981             fix3             = _mm_add_ps(fix3,tx);
1982             fiy3             = _mm_add_ps(fiy3,ty);
1983             fiz3             = _mm_add_ps(fiz3,tz);
1984
1985             fjx2             = _mm_add_ps(fjx2,tx);
1986             fjy2             = _mm_add_ps(fjy2,ty);
1987             fjz2             = _mm_add_ps(fjz2,tz);
1988
1989             /**************************
1990              * CALCULATE INTERACTIONS *
1991              **************************/
1992
1993             r33              = _mm_mul_ps(rsq33,rinv33);
1994
1995             /* EWALD ELECTROSTATICS */
1996
1997             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1998             ewrt             = _mm_mul_ps(r33,ewtabscale);
1999             ewitab           = _mm_cvttps_epi32(ewrt);
2000             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2001             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2002                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2003                                          &ewtabF,&ewtabFn);
2004             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2005             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2006
2007             fscal            = felec;
2008
2009             /* Calculate temporary vectorial force */
2010             tx               = _mm_mul_ps(fscal,dx33);
2011             ty               = _mm_mul_ps(fscal,dy33);
2012             tz               = _mm_mul_ps(fscal,dz33);
2013
2014             /* Update vectorial force */
2015             fix3             = _mm_add_ps(fix3,tx);
2016             fiy3             = _mm_add_ps(fiy3,ty);
2017             fiz3             = _mm_add_ps(fiz3,tz);
2018
2019             fjx3             = _mm_add_ps(fjx3,tx);
2020             fjy3             = _mm_add_ps(fjy3,ty);
2021             fjz3             = _mm_add_ps(fjz3,tz);
2022
2023             fjptrA             = f+j_coord_offsetA;
2024             fjptrB             = f+j_coord_offsetB;
2025             fjptrC             = f+j_coord_offsetC;
2026             fjptrD             = f+j_coord_offsetD;
2027
2028             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2029                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2030                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2031
2032             /* Inner loop uses 375 flops */
2033         }
2034
2035         if(jidx<j_index_end)
2036         {
2037
2038             /* Get j neighbor index, and coordinate index */
2039             jnrlistA         = jjnr[jidx];
2040             jnrlistB         = jjnr[jidx+1];
2041             jnrlistC         = jjnr[jidx+2];
2042             jnrlistD         = jjnr[jidx+3];
2043             /* Sign of each element will be negative for non-real atoms.
2044              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2045              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2046              */
2047             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2048             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2049             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2050             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2051             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2052             j_coord_offsetA  = DIM*jnrA;
2053             j_coord_offsetB  = DIM*jnrB;
2054             j_coord_offsetC  = DIM*jnrC;
2055             j_coord_offsetD  = DIM*jnrD;
2056
2057             /* load j atom coordinates */
2058             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2059                                               x+j_coord_offsetC,x+j_coord_offsetD,
2060                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2061                                               &jy2,&jz2,&jx3,&jy3,&jz3);
2062
2063             /* Calculate displacement vector */
2064             dx00             = _mm_sub_ps(ix0,jx0);
2065             dy00             = _mm_sub_ps(iy0,jy0);
2066             dz00             = _mm_sub_ps(iz0,jz0);
2067             dx11             = _mm_sub_ps(ix1,jx1);
2068             dy11             = _mm_sub_ps(iy1,jy1);
2069             dz11             = _mm_sub_ps(iz1,jz1);
2070             dx12             = _mm_sub_ps(ix1,jx2);
2071             dy12             = _mm_sub_ps(iy1,jy2);
2072             dz12             = _mm_sub_ps(iz1,jz2);
2073             dx13             = _mm_sub_ps(ix1,jx3);
2074             dy13             = _mm_sub_ps(iy1,jy3);
2075             dz13             = _mm_sub_ps(iz1,jz3);
2076             dx21             = _mm_sub_ps(ix2,jx1);
2077             dy21             = _mm_sub_ps(iy2,jy1);
2078             dz21             = _mm_sub_ps(iz2,jz1);
2079             dx22             = _mm_sub_ps(ix2,jx2);
2080             dy22             = _mm_sub_ps(iy2,jy2);
2081             dz22             = _mm_sub_ps(iz2,jz2);
2082             dx23             = _mm_sub_ps(ix2,jx3);
2083             dy23             = _mm_sub_ps(iy2,jy3);
2084             dz23             = _mm_sub_ps(iz2,jz3);
2085             dx31             = _mm_sub_ps(ix3,jx1);
2086             dy31             = _mm_sub_ps(iy3,jy1);
2087             dz31             = _mm_sub_ps(iz3,jz1);
2088             dx32             = _mm_sub_ps(ix3,jx2);
2089             dy32             = _mm_sub_ps(iy3,jy2);
2090             dz32             = _mm_sub_ps(iz3,jz2);
2091             dx33             = _mm_sub_ps(ix3,jx3);
2092             dy33             = _mm_sub_ps(iy3,jy3);
2093             dz33             = _mm_sub_ps(iz3,jz3);
2094
2095             /* Calculate squared distance and things based on it */
2096             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2097             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2098             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2099             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2100             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2101             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2102             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2103             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2104             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2105             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2106
2107             rinv00           = gmx_mm_invsqrt_ps(rsq00);
2108             rinv11           = gmx_mm_invsqrt_ps(rsq11);
2109             rinv12           = gmx_mm_invsqrt_ps(rsq12);
2110             rinv13           = gmx_mm_invsqrt_ps(rsq13);
2111             rinv21           = gmx_mm_invsqrt_ps(rsq21);
2112             rinv22           = gmx_mm_invsqrt_ps(rsq22);
2113             rinv23           = gmx_mm_invsqrt_ps(rsq23);
2114             rinv31           = gmx_mm_invsqrt_ps(rsq31);
2115             rinv32           = gmx_mm_invsqrt_ps(rsq32);
2116             rinv33           = gmx_mm_invsqrt_ps(rsq33);
2117
2118             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
2119             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
2120             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
2121             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
2122             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
2123             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
2124             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
2125             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
2126             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
2127
2128             fjx0             = _mm_setzero_ps();
2129             fjy0             = _mm_setzero_ps();
2130             fjz0             = _mm_setzero_ps();
2131             fjx1             = _mm_setzero_ps();
2132             fjy1             = _mm_setzero_ps();
2133             fjz1             = _mm_setzero_ps();
2134             fjx2             = _mm_setzero_ps();
2135             fjy2             = _mm_setzero_ps();
2136             fjz2             = _mm_setzero_ps();
2137             fjx3             = _mm_setzero_ps();
2138             fjy3             = _mm_setzero_ps();
2139             fjz3             = _mm_setzero_ps();
2140
2141             /**************************
2142              * CALCULATE INTERACTIONS *
2143              **************************/
2144
2145             r00              = _mm_mul_ps(rsq00,rinv00);
2146             r00              = _mm_andnot_ps(dummy_mask,r00);
2147
2148             /* Calculate table index by multiplying r with table scale and truncate to integer */
2149             rt               = _mm_mul_ps(r00,vftabscale);
2150             vfitab           = _mm_cvttps_epi32(rt);
2151             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2152             vfitab           = _mm_slli_epi32(vfitab,3);
2153
2154             /* CUBIC SPLINE TABLE DISPERSION */
2155             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2156             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2157             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2158             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2159             _MM_TRANSPOSE4_PS(Y,F,G,H);
2160             Heps             = _mm_mul_ps(vfeps,H);
2161             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2162             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2163             fvdw6            = _mm_mul_ps(c6_00,FF);
2164
2165             /* CUBIC SPLINE TABLE REPULSION */
2166             vfitab           = _mm_add_epi32(vfitab,ifour);
2167             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2168             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2169             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2170             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2171             _MM_TRANSPOSE4_PS(Y,F,G,H);
2172             Heps             = _mm_mul_ps(vfeps,H);
2173             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2174             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2175             fvdw12           = _mm_mul_ps(c12_00,FF);
2176             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
2177
2178             fscal            = fvdw;
2179
2180             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2181
2182             /* Calculate temporary vectorial force */
2183             tx               = _mm_mul_ps(fscal,dx00);
2184             ty               = _mm_mul_ps(fscal,dy00);
2185             tz               = _mm_mul_ps(fscal,dz00);
2186
2187             /* Update vectorial force */
2188             fix0             = _mm_add_ps(fix0,tx);
2189             fiy0             = _mm_add_ps(fiy0,ty);
2190             fiz0             = _mm_add_ps(fiz0,tz);
2191
2192             fjx0             = _mm_add_ps(fjx0,tx);
2193             fjy0             = _mm_add_ps(fjy0,ty);
2194             fjz0             = _mm_add_ps(fjz0,tz);
2195
2196             /**************************
2197              * CALCULATE INTERACTIONS *
2198              **************************/
2199
2200             r11              = _mm_mul_ps(rsq11,rinv11);
2201             r11              = _mm_andnot_ps(dummy_mask,r11);
2202
2203             /* EWALD ELECTROSTATICS */
2204
2205             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2206             ewrt             = _mm_mul_ps(r11,ewtabscale);
2207             ewitab           = _mm_cvttps_epi32(ewrt);
2208             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2209             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2210                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2211                                          &ewtabF,&ewtabFn);
2212             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2213             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
2214
2215             fscal            = felec;
2216
2217             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2218
2219             /* Calculate temporary vectorial force */
2220             tx               = _mm_mul_ps(fscal,dx11);
2221             ty               = _mm_mul_ps(fscal,dy11);
2222             tz               = _mm_mul_ps(fscal,dz11);
2223
2224             /* Update vectorial force */
2225             fix1             = _mm_add_ps(fix1,tx);
2226             fiy1             = _mm_add_ps(fiy1,ty);
2227             fiz1             = _mm_add_ps(fiz1,tz);
2228
2229             fjx1             = _mm_add_ps(fjx1,tx);
2230             fjy1             = _mm_add_ps(fjy1,ty);
2231             fjz1             = _mm_add_ps(fjz1,tz);
2232
2233             /**************************
2234              * CALCULATE INTERACTIONS *
2235              **************************/
2236
2237             r12              = _mm_mul_ps(rsq12,rinv12);
2238             r12              = _mm_andnot_ps(dummy_mask,r12);
2239
2240             /* EWALD ELECTROSTATICS */
2241
2242             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2243             ewrt             = _mm_mul_ps(r12,ewtabscale);
2244             ewitab           = _mm_cvttps_epi32(ewrt);
2245             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2246             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2247                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2248                                          &ewtabF,&ewtabFn);
2249             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2250             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
2251
2252             fscal            = felec;
2253
2254             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2255
2256             /* Calculate temporary vectorial force */
2257             tx               = _mm_mul_ps(fscal,dx12);
2258             ty               = _mm_mul_ps(fscal,dy12);
2259             tz               = _mm_mul_ps(fscal,dz12);
2260
2261             /* Update vectorial force */
2262             fix1             = _mm_add_ps(fix1,tx);
2263             fiy1             = _mm_add_ps(fiy1,ty);
2264             fiz1             = _mm_add_ps(fiz1,tz);
2265
2266             fjx2             = _mm_add_ps(fjx2,tx);
2267             fjy2             = _mm_add_ps(fjy2,ty);
2268             fjz2             = _mm_add_ps(fjz2,tz);
2269
2270             /**************************
2271              * CALCULATE INTERACTIONS *
2272              **************************/
2273
2274             r13              = _mm_mul_ps(rsq13,rinv13);
2275             r13              = _mm_andnot_ps(dummy_mask,r13);
2276
2277             /* EWALD ELECTROSTATICS */
2278
2279             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2280             ewrt             = _mm_mul_ps(r13,ewtabscale);
2281             ewitab           = _mm_cvttps_epi32(ewrt);
2282             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2283             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2284                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2285                                          &ewtabF,&ewtabFn);
2286             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2287             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
2288
2289             fscal            = felec;
2290
2291             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2292
2293             /* Calculate temporary vectorial force */
2294             tx               = _mm_mul_ps(fscal,dx13);
2295             ty               = _mm_mul_ps(fscal,dy13);
2296             tz               = _mm_mul_ps(fscal,dz13);
2297
2298             /* Update vectorial force */
2299             fix1             = _mm_add_ps(fix1,tx);
2300             fiy1             = _mm_add_ps(fiy1,ty);
2301             fiz1             = _mm_add_ps(fiz1,tz);
2302
2303             fjx3             = _mm_add_ps(fjx3,tx);
2304             fjy3             = _mm_add_ps(fjy3,ty);
2305             fjz3             = _mm_add_ps(fjz3,tz);
2306
2307             /**************************
2308              * CALCULATE INTERACTIONS *
2309              **************************/
2310
2311             r21              = _mm_mul_ps(rsq21,rinv21);
2312             r21              = _mm_andnot_ps(dummy_mask,r21);
2313
2314             /* EWALD ELECTROSTATICS */
2315
2316             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2317             ewrt             = _mm_mul_ps(r21,ewtabscale);
2318             ewitab           = _mm_cvttps_epi32(ewrt);
2319             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2320             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2321                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2322                                          &ewtabF,&ewtabFn);
2323             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2324             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
2325
2326             fscal            = felec;
2327
2328             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2329
2330             /* Calculate temporary vectorial force */
2331             tx               = _mm_mul_ps(fscal,dx21);
2332             ty               = _mm_mul_ps(fscal,dy21);
2333             tz               = _mm_mul_ps(fscal,dz21);
2334
2335             /* Update vectorial force */
2336             fix2             = _mm_add_ps(fix2,tx);
2337             fiy2             = _mm_add_ps(fiy2,ty);
2338             fiz2             = _mm_add_ps(fiz2,tz);
2339
2340             fjx1             = _mm_add_ps(fjx1,tx);
2341             fjy1             = _mm_add_ps(fjy1,ty);
2342             fjz1             = _mm_add_ps(fjz1,tz);
2343
2344             /**************************
2345              * CALCULATE INTERACTIONS *
2346              **************************/
2347
2348             r22              = _mm_mul_ps(rsq22,rinv22);
2349             r22              = _mm_andnot_ps(dummy_mask,r22);
2350
2351             /* EWALD ELECTROSTATICS */
2352
2353             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2354             ewrt             = _mm_mul_ps(r22,ewtabscale);
2355             ewitab           = _mm_cvttps_epi32(ewrt);
2356             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2357             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2358                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2359                                          &ewtabF,&ewtabFn);
2360             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2361             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
2362
2363             fscal            = felec;
2364
2365             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2366
2367             /* Calculate temporary vectorial force */
2368             tx               = _mm_mul_ps(fscal,dx22);
2369             ty               = _mm_mul_ps(fscal,dy22);
2370             tz               = _mm_mul_ps(fscal,dz22);
2371
2372             /* Update vectorial force */
2373             fix2             = _mm_add_ps(fix2,tx);
2374             fiy2             = _mm_add_ps(fiy2,ty);
2375             fiz2             = _mm_add_ps(fiz2,tz);
2376
2377             fjx2             = _mm_add_ps(fjx2,tx);
2378             fjy2             = _mm_add_ps(fjy2,ty);
2379             fjz2             = _mm_add_ps(fjz2,tz);
2380
2381             /**************************
2382              * CALCULATE INTERACTIONS *
2383              **************************/
2384
2385             r23              = _mm_mul_ps(rsq23,rinv23);
2386             r23              = _mm_andnot_ps(dummy_mask,r23);
2387
2388             /* EWALD ELECTROSTATICS */
2389
2390             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2391             ewrt             = _mm_mul_ps(r23,ewtabscale);
2392             ewitab           = _mm_cvttps_epi32(ewrt);
2393             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2394             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2395                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2396                                          &ewtabF,&ewtabFn);
2397             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2398             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
2399
2400             fscal            = felec;
2401
2402             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2403
2404             /* Calculate temporary vectorial force */
2405             tx               = _mm_mul_ps(fscal,dx23);
2406             ty               = _mm_mul_ps(fscal,dy23);
2407             tz               = _mm_mul_ps(fscal,dz23);
2408
2409             /* Update vectorial force */
2410             fix2             = _mm_add_ps(fix2,tx);
2411             fiy2             = _mm_add_ps(fiy2,ty);
2412             fiz2             = _mm_add_ps(fiz2,tz);
2413
2414             fjx3             = _mm_add_ps(fjx3,tx);
2415             fjy3             = _mm_add_ps(fjy3,ty);
2416             fjz3             = _mm_add_ps(fjz3,tz);
2417
2418             /**************************
2419              * CALCULATE INTERACTIONS *
2420              **************************/
2421
2422             r31              = _mm_mul_ps(rsq31,rinv31);
2423             r31              = _mm_andnot_ps(dummy_mask,r31);
2424
2425             /* EWALD ELECTROSTATICS */
2426
2427             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2428             ewrt             = _mm_mul_ps(r31,ewtabscale);
2429             ewitab           = _mm_cvttps_epi32(ewrt);
2430             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2431             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2432                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2433                                          &ewtabF,&ewtabFn);
2434             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2435             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
2436
2437             fscal            = felec;
2438
2439             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2440
2441             /* Calculate temporary vectorial force */
2442             tx               = _mm_mul_ps(fscal,dx31);
2443             ty               = _mm_mul_ps(fscal,dy31);
2444             tz               = _mm_mul_ps(fscal,dz31);
2445
2446             /* Update vectorial force */
2447             fix3             = _mm_add_ps(fix3,tx);
2448             fiy3             = _mm_add_ps(fiy3,ty);
2449             fiz3             = _mm_add_ps(fiz3,tz);
2450
2451             fjx1             = _mm_add_ps(fjx1,tx);
2452             fjy1             = _mm_add_ps(fjy1,ty);
2453             fjz1             = _mm_add_ps(fjz1,tz);
2454
2455             /**************************
2456              * CALCULATE INTERACTIONS *
2457              **************************/
2458
2459             r32              = _mm_mul_ps(rsq32,rinv32);
2460             r32              = _mm_andnot_ps(dummy_mask,r32);
2461
2462             /* EWALD ELECTROSTATICS */
2463
2464             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2465             ewrt             = _mm_mul_ps(r32,ewtabscale);
2466             ewitab           = _mm_cvttps_epi32(ewrt);
2467             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2468             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2469                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2470                                          &ewtabF,&ewtabFn);
2471             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2472             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
2473
2474             fscal            = felec;
2475
2476             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2477
2478             /* Calculate temporary vectorial force */
2479             tx               = _mm_mul_ps(fscal,dx32);
2480             ty               = _mm_mul_ps(fscal,dy32);
2481             tz               = _mm_mul_ps(fscal,dz32);
2482
2483             /* Update vectorial force */
2484             fix3             = _mm_add_ps(fix3,tx);
2485             fiy3             = _mm_add_ps(fiy3,ty);
2486             fiz3             = _mm_add_ps(fiz3,tz);
2487
2488             fjx2             = _mm_add_ps(fjx2,tx);
2489             fjy2             = _mm_add_ps(fjy2,ty);
2490             fjz2             = _mm_add_ps(fjz2,tz);
2491
2492             /**************************
2493              * CALCULATE INTERACTIONS *
2494              **************************/
2495
2496             r33              = _mm_mul_ps(rsq33,rinv33);
2497             r33              = _mm_andnot_ps(dummy_mask,r33);
2498
2499             /* EWALD ELECTROSTATICS */
2500
2501             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2502             ewrt             = _mm_mul_ps(r33,ewtabscale);
2503             ewitab           = _mm_cvttps_epi32(ewrt);
2504             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2505             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2506                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2507                                          &ewtabF,&ewtabFn);
2508             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2509             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2510
2511             fscal            = felec;
2512
2513             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2514
2515             /* Calculate temporary vectorial force */
2516             tx               = _mm_mul_ps(fscal,dx33);
2517             ty               = _mm_mul_ps(fscal,dy33);
2518             tz               = _mm_mul_ps(fscal,dz33);
2519
2520             /* Update vectorial force */
2521             fix3             = _mm_add_ps(fix3,tx);
2522             fiy3             = _mm_add_ps(fiy3,ty);
2523             fiz3             = _mm_add_ps(fiz3,tz);
2524
2525             fjx3             = _mm_add_ps(fjx3,tx);
2526             fjy3             = _mm_add_ps(fjy3,ty);
2527             fjz3             = _mm_add_ps(fjz3,tz);
2528
2529             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2530             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2531             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2532             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2533
2534             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2535                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2536                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2537
2538             /* Inner loop uses 385 flops */
2539         }
2540
2541         /* End of innermost loop */
2542
2543         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2544                                               f+i_coord_offset,fshift+i_shift_offset);
2545
2546         /* Increment number of inner iterations */
2547         inneriter                  += j_index_end - j_index_start;
2548
2549         /* Outer loop uses 24 flops */
2550     }
2551
2552     /* Increment number of outer iterations */
2553     outeriter        += nri;
2554
2555     /* Update outer/inner flops */
2556
2557     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*385);
2558 }