src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_single/nb_kernel_ElecEw_VdwCSTab_GeomW4W4_sse2_single.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, 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 sse2_single kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_sse2_single.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_sse2_single
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water4-Water4
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_sse2_single
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  75     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  80     real             scratch[4*DIM];
  81     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwioffset3;
  89     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  90     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  91     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  93     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  94     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  95     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  96     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  97     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  98     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 100     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 101     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 102     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 103     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 104     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 105     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 106     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 107     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 108     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 109     real             *charge;
 110     int              nvdwtype;
 111     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 112     int              *vdwtype;
 113     real             *vdwparam;
 114     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 115     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 116     __m128i          vfitab;
 117     __m128i          ifour       = _mm_set1_epi32(4);
 118     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 119     real             *vftab;
 120     __m128i          ewitab;
 121     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 122     real             *ewtab;
 123     __m128           dummy_mask,cutoff_mask;
 124     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 125     __m128           one     = _mm_set1_ps(1.0);
 126     __m128           two     = _mm_set1_ps(2.0);
 127     x                = xx[0];
 128     f                = ff[0];
 129
 130     nri              = nlist->nri;
 131     iinr             = nlist->iinr;
 132     jindex           = nlist->jindex;
 133     jjnr             = nlist->jjnr;
 134     shiftidx         = nlist->shift;
 135     gid              = nlist->gid;
 136     shiftvec         = fr->shift_vec[0];
 137     fshift           = fr->fshift[0];
 138     facel            = _mm_set1_ps(fr->ic->epsfac);
 139     charge           = mdatoms->chargeA;
 140     nvdwtype         = fr->ntype;
 141     vdwparam         = fr->nbfp;
 142     vdwtype          = mdatoms->typeA;
 143
 144     vftab            = kernel_data->table_vdw->data;
 145     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 146
 147     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 148     ewtab            = fr->ic->tabq_coul_FDV0;
 149     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 150     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 151
 152     /* Setup water-specific parameters */
 153     inr              = nlist->iinr[0];
 154     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 155     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 156     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 157     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 158
 159     jq1              = _mm_set1_ps(charge[inr+1]);
 160     jq2              = _mm_set1_ps(charge[inr+2]);
 161     jq3              = _mm_set1_ps(charge[inr+3]);
 162     vdwjidx0A        = 2*vdwtype[inr+0];
 163     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 164     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 165     qq11             = _mm_mul_ps(iq1,jq1);
 166     qq12             = _mm_mul_ps(iq1,jq2);
 167     qq13             = _mm_mul_ps(iq1,jq3);
 168     qq21             = _mm_mul_ps(iq2,jq1);
 169     qq22             = _mm_mul_ps(iq2,jq2);
 170     qq23             = _mm_mul_ps(iq2,jq3);
 171     qq31             = _mm_mul_ps(iq3,jq1);
 172     qq32             = _mm_mul_ps(iq3,jq2);
 173     qq33             = _mm_mul_ps(iq3,jq3);
 174
 175     /* Avoid stupid compiler warnings */
 176     jnrA = jnrB = jnrC = jnrD = 0;
 177     j_coord_offsetA = 0;
 178     j_coord_offsetB = 0;
 179     j_coord_offsetC = 0;
 180     j_coord_offsetD = 0;
 181
 182     outeriter        = 0;
 183     inneriter        = 0;
 184
 185     for(iidx=0;iidx<4*DIM;iidx++)
 186     {
 187         scratch[iidx] = 0.0;
 188     }
 189
 190     /* Start outer loop over neighborlists */
 191     for(iidx=0; iidx<nri; iidx++)
 192     {
 193         /* Load shift vector for this list */
 194         i_shift_offset   = DIM*shiftidx[iidx];
 195
 196         /* Load limits for loop over neighbors */
 197         j_index_start    = jindex[iidx];
 198         j_index_end      = jindex[iidx+1];
 199
 200         /* Get outer coordinate index */
 201         inr              = iinr[iidx];
 202         i_coord_offset   = DIM*inr;
 203
 204         /* Load i particle coords and add shift vector */
 205         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 206                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 207
 208         fix0             = _mm_setzero_ps();
 209         fiy0             = _mm_setzero_ps();
 210         fiz0             = _mm_setzero_ps();
 211         fix1             = _mm_setzero_ps();
 212         fiy1             = _mm_setzero_ps();
 213         fiz1             = _mm_setzero_ps();
 214         fix2             = _mm_setzero_ps();
 215         fiy2             = _mm_setzero_ps();
 216         fiz2             = _mm_setzero_ps();
 217         fix3             = _mm_setzero_ps();
 218         fiy3             = _mm_setzero_ps();
 219         fiz3             = _mm_setzero_ps();
 220
 221         /* Reset potential sums */
 222         velecsum         = _mm_setzero_ps();
 223         vvdwsum          = _mm_setzero_ps();
 224
 225         /* Start inner kernel loop */
 226         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 227         {
 228
 229             /* Get j neighbor index, and coordinate index */
 230             jnrA             = jjnr[jidx];
 231             jnrB             = jjnr[jidx+1];
 232             jnrC             = jjnr[jidx+2];
 233             jnrD             = jjnr[jidx+3];
 234             j_coord_offsetA  = DIM*jnrA;
 235             j_coord_offsetB  = DIM*jnrB;
 236             j_coord_offsetC  = DIM*jnrC;
 237             j_coord_offsetD  = DIM*jnrD;
 238
 239             /* load j atom coordinates */
 240             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 241                                               x+j_coord_offsetC,x+j_coord_offsetD,
 242                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 243                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 244
 245             /* Calculate displacement vector */
 246             dx00             = _mm_sub_ps(ix0,jx0);
 247             dy00             = _mm_sub_ps(iy0,jy0);
 248             dz00             = _mm_sub_ps(iz0,jz0);
 249             dx11             = _mm_sub_ps(ix1,jx1);
 250             dy11             = _mm_sub_ps(iy1,jy1);
 251             dz11             = _mm_sub_ps(iz1,jz1);
 252             dx12             = _mm_sub_ps(ix1,jx2);
 253             dy12             = _mm_sub_ps(iy1,jy2);
 254             dz12             = _mm_sub_ps(iz1,jz2);
 255             dx13             = _mm_sub_ps(ix1,jx3);
 256             dy13             = _mm_sub_ps(iy1,jy3);
 257             dz13             = _mm_sub_ps(iz1,jz3);
 258             dx21             = _mm_sub_ps(ix2,jx1);
 259             dy21             = _mm_sub_ps(iy2,jy1);
 260             dz21             = _mm_sub_ps(iz2,jz1);
 261             dx22             = _mm_sub_ps(ix2,jx2);
 262             dy22             = _mm_sub_ps(iy2,jy2);
 263             dz22             = _mm_sub_ps(iz2,jz2);
 264             dx23             = _mm_sub_ps(ix2,jx3);
 265             dy23             = _mm_sub_ps(iy2,jy3);
 266             dz23             = _mm_sub_ps(iz2,jz3);
 267             dx31             = _mm_sub_ps(ix3,jx1);
 268             dy31             = _mm_sub_ps(iy3,jy1);
 269             dz31             = _mm_sub_ps(iz3,jz1);
 270             dx32             = _mm_sub_ps(ix3,jx2);
 271             dy32             = _mm_sub_ps(iy3,jy2);
 272             dz32             = _mm_sub_ps(iz3,jz2);
 273             dx33             = _mm_sub_ps(ix3,jx3);
 274             dy33             = _mm_sub_ps(iy3,jy3);
 275             dz33             = _mm_sub_ps(iz3,jz3);
 276
 277             /* Calculate squared distance and things based on it */
 278             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 279             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 280             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 281             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 282             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 283             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 284             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 285             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 286             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 287             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 288
 289             rinv00           = sse2_invsqrt_f(rsq00);
 290             rinv11           = sse2_invsqrt_f(rsq11);
 291             rinv12           = sse2_invsqrt_f(rsq12);
 292             rinv13           = sse2_invsqrt_f(rsq13);
 293             rinv21           = sse2_invsqrt_f(rsq21);
 294             rinv22           = sse2_invsqrt_f(rsq22);
 295             rinv23           = sse2_invsqrt_f(rsq23);
 296             rinv31           = sse2_invsqrt_f(rsq31);
 297             rinv32           = sse2_invsqrt_f(rsq32);
 298             rinv33           = sse2_invsqrt_f(rsq33);
 299
 300             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 301             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 302             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 303             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 304             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 305             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 306             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 307             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 308             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 309
 310             fjx0             = _mm_setzero_ps();
 311             fjy0             = _mm_setzero_ps();
 312             fjz0             = _mm_setzero_ps();
 313             fjx1             = _mm_setzero_ps();
 314             fjy1             = _mm_setzero_ps();
 315             fjz1             = _mm_setzero_ps();
 316             fjx2             = _mm_setzero_ps();
 317             fjy2             = _mm_setzero_ps();
 318             fjz2             = _mm_setzero_ps();
 319             fjx3             = _mm_setzero_ps();
 320             fjy3             = _mm_setzero_ps();
 321             fjz3             = _mm_setzero_ps();
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             r00              = _mm_mul_ps(rsq00,rinv00);
 328
 329             /* Calculate table index by multiplying r with table scale and truncate to integer */
 330             rt               = _mm_mul_ps(r00,vftabscale);
 331             vfitab           = _mm_cvttps_epi32(rt);
 332             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 333             vfitab           = _mm_slli_epi32(vfitab,3);
 334
 335             /* CUBIC SPLINE TABLE DISPERSION */
 336             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 337             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 338             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 339             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 340             _MM_TRANSPOSE4_PS(Y,F,G,H);
 341             Heps             = _mm_mul_ps(vfeps,H);
 342             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 343             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 344             vvdw6            = _mm_mul_ps(c6_00,VV);
 345             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 346             fvdw6            = _mm_mul_ps(c6_00,FF);
 347
 348             /* CUBIC SPLINE TABLE REPULSION */
 349             vfitab           = _mm_add_epi32(vfitab,ifour);
 350             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 351             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 352             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 353             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 354             _MM_TRANSPOSE4_PS(Y,F,G,H);
 355             Heps             = _mm_mul_ps(vfeps,H);
 356             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 357             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 358             vvdw12           = _mm_mul_ps(c12_00,VV);
 359             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 360             fvdw12           = _mm_mul_ps(c12_00,FF);
 361             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 362             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 363
 364             /* Update potential sum for this i atom from the interaction with this j atom. */
 365             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 366
 367             fscal            = fvdw;
 368
 369             /* Calculate temporary vectorial force */
 370             tx               = _mm_mul_ps(fscal,dx00);
 371             ty               = _mm_mul_ps(fscal,dy00);
 372             tz               = _mm_mul_ps(fscal,dz00);
 373
 374             /* Update vectorial force */
 375             fix0             = _mm_add_ps(fix0,tx);
 376             fiy0             = _mm_add_ps(fiy0,ty);
 377             fiz0             = _mm_add_ps(fiz0,tz);
 378
 379             fjx0             = _mm_add_ps(fjx0,tx);
 380             fjy0             = _mm_add_ps(fjy0,ty);
 381             fjz0             = _mm_add_ps(fjz0,tz);
 382
 383             /**************************
 384              * CALCULATE INTERACTIONS *
 385              **************************/
 386
 387             r11              = _mm_mul_ps(rsq11,rinv11);
 388
 389             /* EWALD ELECTROSTATICS */
 390
 391             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 392             ewrt             = _mm_mul_ps(r11,ewtabscale);
 393             ewitab           = _mm_cvttps_epi32(ewrt);
 394             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 395             ewitab           = _mm_slli_epi32(ewitab,2);
 396             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 397             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 398             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 399             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 400             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 401             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 402             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 403             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 404             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 405
 406             /* Update potential sum for this i atom from the interaction with this j atom. */
 407             velecsum         = _mm_add_ps(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             /* Calculate temporary vectorial force */
 412             tx               = _mm_mul_ps(fscal,dx11);
 413             ty               = _mm_mul_ps(fscal,dy11);
 414             tz               = _mm_mul_ps(fscal,dz11);
 415
 416             /* Update vectorial force */
 417             fix1             = _mm_add_ps(fix1,tx);
 418             fiy1             = _mm_add_ps(fiy1,ty);
 419             fiz1             = _mm_add_ps(fiz1,tz);
 420
 421             fjx1             = _mm_add_ps(fjx1,tx);
 422             fjy1             = _mm_add_ps(fjy1,ty);
 423             fjz1             = _mm_add_ps(fjz1,tz);
 424
 425             /**************************
 426              * CALCULATE INTERACTIONS *
 427              **************************/
 428
 429             r12              = _mm_mul_ps(rsq12,rinv12);
 430
 431             /* EWALD ELECTROSTATICS */
 432
 433             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 434             ewrt             = _mm_mul_ps(r12,ewtabscale);
 435             ewitab           = _mm_cvttps_epi32(ewrt);
 436             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 437             ewitab           = _mm_slli_epi32(ewitab,2);
 438             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 439             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 440             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 441             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 442             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 443             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 444             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 445             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
 446             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
 447
 448             /* Update potential sum for this i atom from the interaction with this j atom. */
 449             velecsum         = _mm_add_ps(velecsum,velec);
 450
 451             fscal            = felec;
 452
 453             /* Calculate temporary vectorial force */
 454             tx               = _mm_mul_ps(fscal,dx12);
 455             ty               = _mm_mul_ps(fscal,dy12);
 456             tz               = _mm_mul_ps(fscal,dz12);
 457
 458             /* Update vectorial force */
 459             fix1             = _mm_add_ps(fix1,tx);
 460             fiy1             = _mm_add_ps(fiy1,ty);
 461             fiz1             = _mm_add_ps(fiz1,tz);
 462
 463             fjx2             = _mm_add_ps(fjx2,tx);
 464             fjy2             = _mm_add_ps(fjy2,ty);
 465             fjz2             = _mm_add_ps(fjz2,tz);
 466
 467             /**************************
 468              * CALCULATE INTERACTIONS *
 469              **************************/
 470
 471             r13              = _mm_mul_ps(rsq13,rinv13);
 472
 473             /* EWALD ELECTROSTATICS */
 474
 475             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 476             ewrt             = _mm_mul_ps(r13,ewtabscale);
 477             ewitab           = _mm_cvttps_epi32(ewrt);
 478             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 479             ewitab           = _mm_slli_epi32(ewitab,2);
 480             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 481             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 482             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 483             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 484             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 485             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 486             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 487             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
 488             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
 489
 490             /* Update potential sum for this i atom from the interaction with this j atom. */
 491             velecsum         = _mm_add_ps(velecsum,velec);
 492
 493             fscal            = felec;
 494
 495             /* Calculate temporary vectorial force */
 496             tx               = _mm_mul_ps(fscal,dx13);
 497             ty               = _mm_mul_ps(fscal,dy13);
 498             tz               = _mm_mul_ps(fscal,dz13);
 499
 500             /* Update vectorial force */
 501             fix1             = _mm_add_ps(fix1,tx);
 502             fiy1             = _mm_add_ps(fiy1,ty);
 503             fiz1             = _mm_add_ps(fiz1,tz);
 504
 505             fjx3             = _mm_add_ps(fjx3,tx);
 506             fjy3             = _mm_add_ps(fjy3,ty);
 507             fjz3             = _mm_add_ps(fjz3,tz);
 508
 509             /**************************
 510              * CALCULATE INTERACTIONS *
 511              **************************/
 512
 513             r21              = _mm_mul_ps(rsq21,rinv21);
 514
 515             /* EWALD ELECTROSTATICS */
 516
 517             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 518             ewrt             = _mm_mul_ps(r21,ewtabscale);
 519             ewitab           = _mm_cvttps_epi32(ewrt);
 520             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 521             ewitab           = _mm_slli_epi32(ewitab,2);
 522             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 523             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 524             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 525             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 526             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 527             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 528             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 529             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
 530             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
 531
 532             /* Update potential sum for this i atom from the interaction with this j atom. */
 533             velecsum         = _mm_add_ps(velecsum,velec);
 534
 535             fscal            = felec;
 536
 537             /* Calculate temporary vectorial force */
 538             tx               = _mm_mul_ps(fscal,dx21);
 539             ty               = _mm_mul_ps(fscal,dy21);
 540             tz               = _mm_mul_ps(fscal,dz21);
 541
 542             /* Update vectorial force */
 543             fix2             = _mm_add_ps(fix2,tx);
 544             fiy2             = _mm_add_ps(fiy2,ty);
 545             fiz2             = _mm_add_ps(fiz2,tz);
 546
 547             fjx1             = _mm_add_ps(fjx1,tx);
 548             fjy1             = _mm_add_ps(fjy1,ty);
 549             fjz1             = _mm_add_ps(fjz1,tz);
 550
 551             /**************************
 552              * CALCULATE INTERACTIONS *
 553              **************************/
 554
 555             r22              = _mm_mul_ps(rsq22,rinv22);
 556
 557             /* EWALD ELECTROSTATICS */
 558
 559             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 560             ewrt             = _mm_mul_ps(r22,ewtabscale);
 561             ewitab           = _mm_cvttps_epi32(ewrt);
 562             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 563             ewitab           = _mm_slli_epi32(ewitab,2);
 564             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 565             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 566             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 567             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 568             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 569             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 570             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 571             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
 572             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
 573
 574             /* Update potential sum for this i atom from the interaction with this j atom. */
 575             velecsum         = _mm_add_ps(velecsum,velec);
 576
 577             fscal            = felec;
 578
 579             /* Calculate temporary vectorial force */
 580             tx               = _mm_mul_ps(fscal,dx22);
 581             ty               = _mm_mul_ps(fscal,dy22);
 582             tz               = _mm_mul_ps(fscal,dz22);
 583
 584             /* Update vectorial force */
 585             fix2             = _mm_add_ps(fix2,tx);
 586             fiy2             = _mm_add_ps(fiy2,ty);
 587             fiz2             = _mm_add_ps(fiz2,tz);
 588
 589             fjx2             = _mm_add_ps(fjx2,tx);
 590             fjy2             = _mm_add_ps(fjy2,ty);
 591             fjz2             = _mm_add_ps(fjz2,tz);
 592
 593             /**************************
 594              * CALCULATE INTERACTIONS *
 595              **************************/
 596
 597             r23              = _mm_mul_ps(rsq23,rinv23);
 598
 599             /* EWALD ELECTROSTATICS */
 600
 601             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 602             ewrt             = _mm_mul_ps(r23,ewtabscale);
 603             ewitab           = _mm_cvttps_epi32(ewrt);
 604             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 605             ewitab           = _mm_slli_epi32(ewitab,2);
 606             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 607             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 608             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 609             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 610             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 611             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 612             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 613             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
 614             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
 615
 616             /* Update potential sum for this i atom from the interaction with this j atom. */
 617             velecsum         = _mm_add_ps(velecsum,velec);
 618
 619             fscal            = felec;
 620
 621             /* Calculate temporary vectorial force */
 622             tx               = _mm_mul_ps(fscal,dx23);
 623             ty               = _mm_mul_ps(fscal,dy23);
 624             tz               = _mm_mul_ps(fscal,dz23);
 625
 626             /* Update vectorial force */
 627             fix2             = _mm_add_ps(fix2,tx);
 628             fiy2             = _mm_add_ps(fiy2,ty);
 629             fiz2             = _mm_add_ps(fiz2,tz);
 630
 631             fjx3             = _mm_add_ps(fjx3,tx);
 632             fjy3             = _mm_add_ps(fjy3,ty);
 633             fjz3             = _mm_add_ps(fjz3,tz);
 634
 635             /**************************
 636              * CALCULATE INTERACTIONS *
 637              **************************/
 638
 639             r31              = _mm_mul_ps(rsq31,rinv31);
 640
 641             /* EWALD ELECTROSTATICS */
 642
 643             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 644             ewrt             = _mm_mul_ps(r31,ewtabscale);
 645             ewitab           = _mm_cvttps_epi32(ewrt);
 646             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 647             ewitab           = _mm_slli_epi32(ewitab,2);
 648             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 649             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 650             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 651             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 652             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 653             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 654             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 655             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
 656             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
 657
 658             /* Update potential sum for this i atom from the interaction with this j atom. */
 659             velecsum         = _mm_add_ps(velecsum,velec);
 660
 661             fscal            = felec;
 662
 663             /* Calculate temporary vectorial force */
 664             tx               = _mm_mul_ps(fscal,dx31);
 665             ty               = _mm_mul_ps(fscal,dy31);
 666             tz               = _mm_mul_ps(fscal,dz31);
 667
 668             /* Update vectorial force */
 669             fix3             = _mm_add_ps(fix3,tx);
 670             fiy3             = _mm_add_ps(fiy3,ty);
 671             fiz3             = _mm_add_ps(fiz3,tz);
 672
 673             fjx1             = _mm_add_ps(fjx1,tx);
 674             fjy1             = _mm_add_ps(fjy1,ty);
 675             fjz1             = _mm_add_ps(fjz1,tz);
 676
 677             /**************************
 678              * CALCULATE INTERACTIONS *
 679              **************************/
 680
 681             r32              = _mm_mul_ps(rsq32,rinv32);
 682
 683             /* EWALD ELECTROSTATICS */
 684
 685             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 686             ewrt             = _mm_mul_ps(r32,ewtabscale);
 687             ewitab           = _mm_cvttps_epi32(ewrt);
 688             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 689             ewitab           = _mm_slli_epi32(ewitab,2);
 690             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 691             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 692             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 693             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 694             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 695             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 696             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 697             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
 698             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
 699
 700             /* Update potential sum for this i atom from the interaction with this j atom. */
 701             velecsum         = _mm_add_ps(velecsum,velec);
 702
 703             fscal            = felec;
 704
 705             /* Calculate temporary vectorial force */
 706             tx               = _mm_mul_ps(fscal,dx32);
 707             ty               = _mm_mul_ps(fscal,dy32);
 708             tz               = _mm_mul_ps(fscal,dz32);
 709
 710             /* Update vectorial force */
 711             fix3             = _mm_add_ps(fix3,tx);
 712             fiy3             = _mm_add_ps(fiy3,ty);
 713             fiz3             = _mm_add_ps(fiz3,tz);
 714
 715             fjx2             = _mm_add_ps(fjx2,tx);
 716             fjy2             = _mm_add_ps(fjy2,ty);
 717             fjz2             = _mm_add_ps(fjz2,tz);
 718
 719             /**************************
 720              * CALCULATE INTERACTIONS *
 721              **************************/
 722
 723             r33              = _mm_mul_ps(rsq33,rinv33);
 724
 725             /* EWALD ELECTROSTATICS */
 726
 727             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 728             ewrt             = _mm_mul_ps(r33,ewtabscale);
 729             ewitab           = _mm_cvttps_epi32(ewrt);
 730             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 731             ewitab           = _mm_slli_epi32(ewitab,2);
 732             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 733             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 734             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 735             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 736             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 737             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 738             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 739             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
 740             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
 741
 742             /* Update potential sum for this i atom from the interaction with this j atom. */
 743             velecsum         = _mm_add_ps(velecsum,velec);
 744
 745             fscal            = felec;
 746
 747             /* Calculate temporary vectorial force */
 748             tx               = _mm_mul_ps(fscal,dx33);
 749             ty               = _mm_mul_ps(fscal,dy33);
 750             tz               = _mm_mul_ps(fscal,dz33);
 751
 752             /* Update vectorial force */
 753             fix3             = _mm_add_ps(fix3,tx);
 754             fiy3             = _mm_add_ps(fiy3,ty);
 755             fiz3             = _mm_add_ps(fiz3,tz);
 756
 757             fjx3             = _mm_add_ps(fjx3,tx);
 758             fjy3             = _mm_add_ps(fjy3,ty);
 759             fjz3             = _mm_add_ps(fjz3,tz);
 760
 761             fjptrA             = f+j_coord_offsetA;
 762             fjptrB             = f+j_coord_offsetB;
 763             fjptrC             = f+j_coord_offsetC;
 764             fjptrD             = f+j_coord_offsetD;
 765
 766             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 767                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 768                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 769
 770             /* Inner loop uses 428 flops */
 771         }
 772
 773         if(jidx<j_index_end)
 774         {
 775
 776             /* Get j neighbor index, and coordinate index */
 777             jnrlistA         = jjnr[jidx];
 778             jnrlistB         = jjnr[jidx+1];
 779             jnrlistC         = jjnr[jidx+2];
 780             jnrlistD         = jjnr[jidx+3];
 781             /* Sign of each element will be negative for non-real atoms.
 782              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 783              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 784              */
 785             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 786             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 787             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 788             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 789             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 790             j_coord_offsetA  = DIM*jnrA;
 791             j_coord_offsetB  = DIM*jnrB;
 792             j_coord_offsetC  = DIM*jnrC;
 793             j_coord_offsetD  = DIM*jnrD;
 794
 795             /* load j atom coordinates */
 796             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 797                                               x+j_coord_offsetC,x+j_coord_offsetD,
 798                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 799                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 800
 801             /* Calculate displacement vector */
 802             dx00             = _mm_sub_ps(ix0,jx0);
 803             dy00             = _mm_sub_ps(iy0,jy0);
 804             dz00             = _mm_sub_ps(iz0,jz0);
 805             dx11             = _mm_sub_ps(ix1,jx1);
 806             dy11             = _mm_sub_ps(iy1,jy1);
 807             dz11             = _mm_sub_ps(iz1,jz1);
 808             dx12             = _mm_sub_ps(ix1,jx2);
 809             dy12             = _mm_sub_ps(iy1,jy2);
 810             dz12             = _mm_sub_ps(iz1,jz2);
 811             dx13             = _mm_sub_ps(ix1,jx3);
 812             dy13             = _mm_sub_ps(iy1,jy3);
 813             dz13             = _mm_sub_ps(iz1,jz3);
 814             dx21             = _mm_sub_ps(ix2,jx1);
 815             dy21             = _mm_sub_ps(iy2,jy1);
 816             dz21             = _mm_sub_ps(iz2,jz1);
 817             dx22             = _mm_sub_ps(ix2,jx2);
 818             dy22             = _mm_sub_ps(iy2,jy2);
 819             dz22             = _mm_sub_ps(iz2,jz2);
 820             dx23             = _mm_sub_ps(ix2,jx3);
 821             dy23             = _mm_sub_ps(iy2,jy3);
 822             dz23             = _mm_sub_ps(iz2,jz3);
 823             dx31             = _mm_sub_ps(ix3,jx1);
 824             dy31             = _mm_sub_ps(iy3,jy1);
 825             dz31             = _mm_sub_ps(iz3,jz1);
 826             dx32             = _mm_sub_ps(ix3,jx2);
 827             dy32             = _mm_sub_ps(iy3,jy2);
 828             dz32             = _mm_sub_ps(iz3,jz2);
 829             dx33             = _mm_sub_ps(ix3,jx3);
 830             dy33             = _mm_sub_ps(iy3,jy3);
 831             dz33             = _mm_sub_ps(iz3,jz3);
 832
 833             /* Calculate squared distance and things based on it */
 834             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 835             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 836             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 837             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 838             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 839             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 840             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 841             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 842             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 843             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 844
 845             rinv00           = sse2_invsqrt_f(rsq00);
 846             rinv11           = sse2_invsqrt_f(rsq11);
 847             rinv12           = sse2_invsqrt_f(rsq12);
 848             rinv13           = sse2_invsqrt_f(rsq13);
 849             rinv21           = sse2_invsqrt_f(rsq21);
 850             rinv22           = sse2_invsqrt_f(rsq22);
 851             rinv23           = sse2_invsqrt_f(rsq23);
 852             rinv31           = sse2_invsqrt_f(rsq31);
 853             rinv32           = sse2_invsqrt_f(rsq32);
 854             rinv33           = sse2_invsqrt_f(rsq33);
 855
 856             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 857             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 858             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 859             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 860             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 861             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 862             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 863             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 864             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 865
 866             fjx0             = _mm_setzero_ps();
 867             fjy0             = _mm_setzero_ps();
 868             fjz0             = _mm_setzero_ps();
 869             fjx1             = _mm_setzero_ps();
 870             fjy1             = _mm_setzero_ps();
 871             fjz1             = _mm_setzero_ps();
 872             fjx2             = _mm_setzero_ps();
 873             fjy2             = _mm_setzero_ps();
 874             fjz2             = _mm_setzero_ps();
 875             fjx3             = _mm_setzero_ps();
 876             fjy3             = _mm_setzero_ps();
 877             fjz3             = _mm_setzero_ps();
 878
 879             /**************************
 880              * CALCULATE INTERACTIONS *
 881              **************************/
 882
 883             r00              = _mm_mul_ps(rsq00,rinv00);
 884             r00              = _mm_andnot_ps(dummy_mask,r00);
 885
 886             /* Calculate table index by multiplying r with table scale and truncate to integer */
 887             rt               = _mm_mul_ps(r00,vftabscale);
 888             vfitab           = _mm_cvttps_epi32(rt);
 889             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 890             vfitab           = _mm_slli_epi32(vfitab,3);
 891
 892             /* CUBIC SPLINE TABLE DISPERSION */
 893             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 894             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 895             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 896             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 897             _MM_TRANSPOSE4_PS(Y,F,G,H);
 898             Heps             = _mm_mul_ps(vfeps,H);
 899             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 900             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 901             vvdw6            = _mm_mul_ps(c6_00,VV);
 902             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 903             fvdw6            = _mm_mul_ps(c6_00,FF);
 904
 905             /* CUBIC SPLINE TABLE REPULSION */
 906             vfitab           = _mm_add_epi32(vfitab,ifour);
 907             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 908             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 909             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 910             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 911             _MM_TRANSPOSE4_PS(Y,F,G,H);
 912             Heps             = _mm_mul_ps(vfeps,H);
 913             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 914             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 915             vvdw12           = _mm_mul_ps(c12_00,VV);
 916             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 917             fvdw12           = _mm_mul_ps(c12_00,FF);
 918             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 919             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 920
 921             /* Update potential sum for this i atom from the interaction with this j atom. */
 922             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 923             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 924
 925             fscal            = fvdw;
 926
 927             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 928
 929             /* Calculate temporary vectorial force */
 930             tx               = _mm_mul_ps(fscal,dx00);
 931             ty               = _mm_mul_ps(fscal,dy00);
 932             tz               = _mm_mul_ps(fscal,dz00);
 933
 934             /* Update vectorial force */
 935             fix0             = _mm_add_ps(fix0,tx);
 936             fiy0             = _mm_add_ps(fiy0,ty);
 937             fiz0             = _mm_add_ps(fiz0,tz);
 938
 939             fjx0             = _mm_add_ps(fjx0,tx);
 940             fjy0             = _mm_add_ps(fjy0,ty);
 941             fjz0             = _mm_add_ps(fjz0,tz);
 942
 943             /**************************
 944              * CALCULATE INTERACTIONS *
 945              **************************/
 946
 947             r11              = _mm_mul_ps(rsq11,rinv11);
 948             r11              = _mm_andnot_ps(dummy_mask,r11);
 949
 950             /* EWALD ELECTROSTATICS */
 951
 952             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 953             ewrt             = _mm_mul_ps(r11,ewtabscale);
 954             ewitab           = _mm_cvttps_epi32(ewrt);
 955             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 956             ewitab           = _mm_slli_epi32(ewitab,2);
 957             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 958             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 959             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 960             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 961             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 962             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 963             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 964             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 965             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 966
 967             /* Update potential sum for this i atom from the interaction with this j atom. */
 968             velec            = _mm_andnot_ps(dummy_mask,velec);
 969             velecsum         = _mm_add_ps(velecsum,velec);
 970
 971             fscal            = felec;
 972
 973             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 974
 975             /* Calculate temporary vectorial force */
 976             tx               = _mm_mul_ps(fscal,dx11);
 977             ty               = _mm_mul_ps(fscal,dy11);
 978             tz               = _mm_mul_ps(fscal,dz11);
 979
 980             /* Update vectorial force */
 981             fix1             = _mm_add_ps(fix1,tx);
 982             fiy1             = _mm_add_ps(fiy1,ty);
 983             fiz1             = _mm_add_ps(fiz1,tz);
 984
 985             fjx1             = _mm_add_ps(fjx1,tx);
 986             fjy1             = _mm_add_ps(fjy1,ty);
 987             fjz1             = _mm_add_ps(fjz1,tz);
 988
 989             /**************************
 990              * CALCULATE INTERACTIONS *
 991              **************************/
 992
 993             r12              = _mm_mul_ps(rsq12,rinv12);
 994             r12              = _mm_andnot_ps(dummy_mask,r12);
 995
 996             /* EWALD ELECTROSTATICS */
 997
 998             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 999             ewrt             = _mm_mul_ps(r12,ewtabscale);
1000             ewitab           = _mm_cvttps_epi32(ewrt);
1001             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1002             ewitab           = _mm_slli_epi32(ewitab,2);
1003             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1004             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1005             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1006             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1007             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1008             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1009             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1010             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
1011             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1012
1013             /* Update potential sum for this i atom from the interaction with this j atom. */
1014             velec            = _mm_andnot_ps(dummy_mask,velec);
1015             velecsum         = _mm_add_ps(velecsum,velec);
1016
1017             fscal            = felec;
1018
1019             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1020
1021             /* Calculate temporary vectorial force */
1022             tx               = _mm_mul_ps(fscal,dx12);
1023             ty               = _mm_mul_ps(fscal,dy12);
1024             tz               = _mm_mul_ps(fscal,dz12);
1025
1026             /* Update vectorial force */
1027             fix1             = _mm_add_ps(fix1,tx);
1028             fiy1             = _mm_add_ps(fiy1,ty);
1029             fiz1             = _mm_add_ps(fiz1,tz);
1030
1031             fjx2             = _mm_add_ps(fjx2,tx);
1032             fjy2             = _mm_add_ps(fjy2,ty);
1033             fjz2             = _mm_add_ps(fjz2,tz);
1034
1035             /**************************
1036              * CALCULATE INTERACTIONS *
1037              **************************/
1038
1039             r13              = _mm_mul_ps(rsq13,rinv13);
1040             r13              = _mm_andnot_ps(dummy_mask,r13);
1041
1042             /* EWALD ELECTROSTATICS */
1043
1044             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1045             ewrt             = _mm_mul_ps(r13,ewtabscale);
1046             ewitab           = _mm_cvttps_epi32(ewrt);
1047             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1048             ewitab           = _mm_slli_epi32(ewitab,2);
1049             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1050             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1051             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1052             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1053             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1054             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1055             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1056             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
1057             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1058
1059             /* Update potential sum for this i atom from the interaction with this j atom. */
1060             velec            = _mm_andnot_ps(dummy_mask,velec);
1061             velecsum         = _mm_add_ps(velecsum,velec);
1062
1063             fscal            = felec;
1064
1065             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1066
1067             /* Calculate temporary vectorial force */
1068             tx               = _mm_mul_ps(fscal,dx13);
1069             ty               = _mm_mul_ps(fscal,dy13);
1070             tz               = _mm_mul_ps(fscal,dz13);
1071
1072             /* Update vectorial force */
1073             fix1             = _mm_add_ps(fix1,tx);
1074             fiy1             = _mm_add_ps(fiy1,ty);
1075             fiz1             = _mm_add_ps(fiz1,tz);
1076
1077             fjx3             = _mm_add_ps(fjx3,tx);
1078             fjy3             = _mm_add_ps(fjy3,ty);
1079             fjz3             = _mm_add_ps(fjz3,tz);
1080
1081             /**************************
1082              * CALCULATE INTERACTIONS *
1083              **************************/
1084
1085             r21              = _mm_mul_ps(rsq21,rinv21);
1086             r21              = _mm_andnot_ps(dummy_mask,r21);
1087
1088             /* EWALD ELECTROSTATICS */
1089
1090             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1091             ewrt             = _mm_mul_ps(r21,ewtabscale);
1092             ewitab           = _mm_cvttps_epi32(ewrt);
1093             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1094             ewitab           = _mm_slli_epi32(ewitab,2);
1095             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1096             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1097             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1098             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1099             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1100             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1101             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1102             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
1103             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1104
1105             /* Update potential sum for this i atom from the interaction with this j atom. */
1106             velec            = _mm_andnot_ps(dummy_mask,velec);
1107             velecsum         = _mm_add_ps(velecsum,velec);
1108
1109             fscal            = felec;
1110
1111             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1112
1113             /* Calculate temporary vectorial force */
1114             tx               = _mm_mul_ps(fscal,dx21);
1115             ty               = _mm_mul_ps(fscal,dy21);
1116             tz               = _mm_mul_ps(fscal,dz21);
1117
1118             /* Update vectorial force */
1119             fix2             = _mm_add_ps(fix2,tx);
1120             fiy2             = _mm_add_ps(fiy2,ty);
1121             fiz2             = _mm_add_ps(fiz2,tz);
1122
1123             fjx1             = _mm_add_ps(fjx1,tx);
1124             fjy1             = _mm_add_ps(fjy1,ty);
1125             fjz1             = _mm_add_ps(fjz1,tz);
1126
1127             /**************************
1128              * CALCULATE INTERACTIONS *
1129              **************************/
1130
1131             r22              = _mm_mul_ps(rsq22,rinv22);
1132             r22              = _mm_andnot_ps(dummy_mask,r22);
1133
1134             /* EWALD ELECTROSTATICS */
1135
1136             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1137             ewrt             = _mm_mul_ps(r22,ewtabscale);
1138             ewitab           = _mm_cvttps_epi32(ewrt);
1139             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1140             ewitab           = _mm_slli_epi32(ewitab,2);
1141             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1142             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1143             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1144             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1145             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1146             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1147             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1148             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
1149             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1150
1151             /* Update potential sum for this i atom from the interaction with this j atom. */
1152             velec            = _mm_andnot_ps(dummy_mask,velec);
1153             velecsum         = _mm_add_ps(velecsum,velec);
1154
1155             fscal            = felec;
1156
1157             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1158
1159             /* Calculate temporary vectorial force */
1160             tx               = _mm_mul_ps(fscal,dx22);
1161             ty               = _mm_mul_ps(fscal,dy22);
1162             tz               = _mm_mul_ps(fscal,dz22);
1163
1164             /* Update vectorial force */
1165             fix2             = _mm_add_ps(fix2,tx);
1166             fiy2             = _mm_add_ps(fiy2,ty);
1167             fiz2             = _mm_add_ps(fiz2,tz);
1168
1169             fjx2             = _mm_add_ps(fjx2,tx);
1170             fjy2             = _mm_add_ps(fjy2,ty);
1171             fjz2             = _mm_add_ps(fjz2,tz);
1172
1173             /**************************
1174              * CALCULATE INTERACTIONS *
1175              **************************/
1176
1177             r23              = _mm_mul_ps(rsq23,rinv23);
1178             r23              = _mm_andnot_ps(dummy_mask,r23);
1179
1180             /* EWALD ELECTROSTATICS */
1181
1182             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1183             ewrt             = _mm_mul_ps(r23,ewtabscale);
1184             ewitab           = _mm_cvttps_epi32(ewrt);
1185             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1186             ewitab           = _mm_slli_epi32(ewitab,2);
1187             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1188             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1189             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1190             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1191             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1192             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1193             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1194             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
1195             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1196
1197             /* Update potential sum for this i atom from the interaction with this j atom. */
1198             velec            = _mm_andnot_ps(dummy_mask,velec);
1199             velecsum         = _mm_add_ps(velecsum,velec);
1200
1201             fscal            = felec;
1202
1203             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1204
1205             /* Calculate temporary vectorial force */
1206             tx               = _mm_mul_ps(fscal,dx23);
1207             ty               = _mm_mul_ps(fscal,dy23);
1208             tz               = _mm_mul_ps(fscal,dz23);
1209
1210             /* Update vectorial force */
1211             fix2             = _mm_add_ps(fix2,tx);
1212             fiy2             = _mm_add_ps(fiy2,ty);
1213             fiz2             = _mm_add_ps(fiz2,tz);
1214
1215             fjx3             = _mm_add_ps(fjx3,tx);
1216             fjy3             = _mm_add_ps(fjy3,ty);
1217             fjz3             = _mm_add_ps(fjz3,tz);
1218
1219             /**************************
1220              * CALCULATE INTERACTIONS *
1221              **************************/
1222
1223             r31              = _mm_mul_ps(rsq31,rinv31);
1224             r31              = _mm_andnot_ps(dummy_mask,r31);
1225
1226             /* EWALD ELECTROSTATICS */
1227
1228             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1229             ewrt             = _mm_mul_ps(r31,ewtabscale);
1230             ewitab           = _mm_cvttps_epi32(ewrt);
1231             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1232             ewitab           = _mm_slli_epi32(ewitab,2);
1233             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1234             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1235             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1236             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1237             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1238             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1239             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1240             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
1241             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1242
1243             /* Update potential sum for this i atom from the interaction with this j atom. */
1244             velec            = _mm_andnot_ps(dummy_mask,velec);
1245             velecsum         = _mm_add_ps(velecsum,velec);
1246
1247             fscal            = felec;
1248
1249             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1250
1251             /* Calculate temporary vectorial force */
1252             tx               = _mm_mul_ps(fscal,dx31);
1253             ty               = _mm_mul_ps(fscal,dy31);
1254             tz               = _mm_mul_ps(fscal,dz31);
1255
1256             /* Update vectorial force */
1257             fix3             = _mm_add_ps(fix3,tx);
1258             fiy3             = _mm_add_ps(fiy3,ty);
1259             fiz3             = _mm_add_ps(fiz3,tz);
1260
1261             fjx1             = _mm_add_ps(fjx1,tx);
1262             fjy1             = _mm_add_ps(fjy1,ty);
1263             fjz1             = _mm_add_ps(fjz1,tz);
1264
1265             /**************************
1266              * CALCULATE INTERACTIONS *
1267              **************************/
1268
1269             r32              = _mm_mul_ps(rsq32,rinv32);
1270             r32              = _mm_andnot_ps(dummy_mask,r32);
1271
1272             /* EWALD ELECTROSTATICS */
1273
1274             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1275             ewrt             = _mm_mul_ps(r32,ewtabscale);
1276             ewitab           = _mm_cvttps_epi32(ewrt);
1277             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1278             ewitab           = _mm_slli_epi32(ewitab,2);
1279             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1280             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1281             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1282             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1283             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1284             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1285             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1286             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
1287             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1288
1289             /* Update potential sum for this i atom from the interaction with this j atom. */
1290             velec            = _mm_andnot_ps(dummy_mask,velec);
1291             velecsum         = _mm_add_ps(velecsum,velec);
1292
1293             fscal            = felec;
1294
1295             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1296
1297             /* Calculate temporary vectorial force */
1298             tx               = _mm_mul_ps(fscal,dx32);
1299             ty               = _mm_mul_ps(fscal,dy32);
1300             tz               = _mm_mul_ps(fscal,dz32);
1301
1302             /* Update vectorial force */
1303             fix3             = _mm_add_ps(fix3,tx);
1304             fiy3             = _mm_add_ps(fiy3,ty);
1305             fiz3             = _mm_add_ps(fiz3,tz);
1306
1307             fjx2             = _mm_add_ps(fjx2,tx);
1308             fjy2             = _mm_add_ps(fjy2,ty);
1309             fjz2             = _mm_add_ps(fjz2,tz);
1310
1311             /**************************
1312              * CALCULATE INTERACTIONS *
1313              **************************/
1314
1315             r33              = _mm_mul_ps(rsq33,rinv33);
1316             r33              = _mm_andnot_ps(dummy_mask,r33);
1317
1318             /* EWALD ELECTROSTATICS */
1319
1320             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1321             ewrt             = _mm_mul_ps(r33,ewtabscale);
1322             ewitab           = _mm_cvttps_epi32(ewrt);
1323             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1324             ewitab           = _mm_slli_epi32(ewitab,2);
1325             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1326             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1327             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1328             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1329             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1330             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1331             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1332             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
1333             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1334
1335             /* Update potential sum for this i atom from the interaction with this j atom. */
1336             velec            = _mm_andnot_ps(dummy_mask,velec);
1337             velecsum         = _mm_add_ps(velecsum,velec);
1338
1339             fscal            = felec;
1340
1341             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1342
1343             /* Calculate temporary vectorial force */
1344             tx               = _mm_mul_ps(fscal,dx33);
1345             ty               = _mm_mul_ps(fscal,dy33);
1346             tz               = _mm_mul_ps(fscal,dz33);
1347
1348             /* Update vectorial force */
1349             fix3             = _mm_add_ps(fix3,tx);
1350             fiy3             = _mm_add_ps(fiy3,ty);
1351             fiz3             = _mm_add_ps(fiz3,tz);
1352
1353             fjx3             = _mm_add_ps(fjx3,tx);
1354             fjy3             = _mm_add_ps(fjy3,ty);
1355             fjz3             = _mm_add_ps(fjz3,tz);
1356
1357             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1358             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1359             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1360             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1361
1362             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1363                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1364                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1365
1366             /* Inner loop uses 438 flops */
1367         }
1368
1369         /* End of innermost loop */
1370
1371         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1372                                               f+i_coord_offset,fshift+i_shift_offset);
1373
1374         ggid                        = gid[iidx];
1375         /* Update potential energies */
1376         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1377         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1378
1379         /* Increment number of inner iterations */
1380         inneriter                  += j_index_end - j_index_start;
1381
1382         /* Outer loop uses 26 flops */
1383     }
1384
1385     /* Increment number of outer iterations */
1386     outeriter        += nri;
1387
1388     /* Update outer/inner flops */
1389
1390     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*438);
1391 }
1392 /*
1393  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse2_single
1394  * Electrostatics interaction: Ewald
1395  * VdW interaction:            CubicSplineTable
1396  * Geometry:                   Water4-Water4
1397  * Calculate force/pot:        Force
1398  */
1399 void
1400 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse2_single
1401                     (t_nblist                    * gmx_restrict       nlist,
1402                      rvec                        * gmx_restrict          xx,
1403                      rvec                        * gmx_restrict          ff,
1404                      struct t_forcerec           * gmx_restrict          fr,
1405                      t_mdatoms                   * gmx_restrict     mdatoms,
1406                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1407                      t_nrnb                      * gmx_restrict        nrnb)
1408 {
1409     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1410      * just 0 for non-waters.
1411      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1412      * jnr indices corresponding to data put in the four positions in the SIMD register.
1413      */
1414     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1415     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1416     int              jnrA,jnrB,jnrC,jnrD;
1417     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1418     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1419     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1420     real             rcutoff_scalar;
1421     real             *shiftvec,*fshift,*x,*f;
1422     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1423     real             scratch[4*DIM];
1424     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1425     int              vdwioffset0;
1426     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1427     int              vdwioffset1;
1428     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1429     int              vdwioffset2;
1430     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1431     int              vdwioffset3;
1432     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1433     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1434     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1435     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1436     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1437     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1438     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1439     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1440     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1441     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1442     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1443     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1444     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1445     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1446     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1447     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1448     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1449     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1450     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1451     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1452     real             *charge;
1453     int              nvdwtype;
1454     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1455     int              *vdwtype;
1456     real             *vdwparam;
1457     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1458     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1459     __m128i          vfitab;
1460     __m128i          ifour       = _mm_set1_epi32(4);
1461     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1462     real             *vftab;
1463     __m128i          ewitab;
1464     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1465     real             *ewtab;
1466     __m128           dummy_mask,cutoff_mask;
1467     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1468     __m128           one     = _mm_set1_ps(1.0);
1469     __m128           two     = _mm_set1_ps(2.0);
1470     x                = xx[0];
1471     f                = ff[0];
1472
1473     nri              = nlist->nri;
1474     iinr             = nlist->iinr;
1475     jindex           = nlist->jindex;
1476     jjnr             = nlist->jjnr;
1477     shiftidx         = nlist->shift;
1478     gid              = nlist->gid;
1479     shiftvec         = fr->shift_vec[0];
1480     fshift           = fr->fshift[0];
1481     facel            = _mm_set1_ps(fr->ic->epsfac);
1482     charge           = mdatoms->chargeA;
1483     nvdwtype         = fr->ntype;
1484     vdwparam         = fr->nbfp;
1485     vdwtype          = mdatoms->typeA;
1486
1487     vftab            = kernel_data->table_vdw->data;
1488     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
1489
1490     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
1491     ewtab            = fr->ic->tabq_coul_F;
1492     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
1493     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1494
1495     /* Setup water-specific parameters */
1496     inr              = nlist->iinr[0];
1497     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1498     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1499     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1500     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1501
1502     jq1              = _mm_set1_ps(charge[inr+1]);
1503     jq2              = _mm_set1_ps(charge[inr+2]);
1504     jq3              = _mm_set1_ps(charge[inr+3]);
1505     vdwjidx0A        = 2*vdwtype[inr+0];
1506     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1507     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1508     qq11             = _mm_mul_ps(iq1,jq1);
1509     qq12             = _mm_mul_ps(iq1,jq2);
1510     qq13             = _mm_mul_ps(iq1,jq3);
1511     qq21             = _mm_mul_ps(iq2,jq1);
1512     qq22             = _mm_mul_ps(iq2,jq2);
1513     qq23             = _mm_mul_ps(iq2,jq3);
1514     qq31             = _mm_mul_ps(iq3,jq1);
1515     qq32             = _mm_mul_ps(iq3,jq2);
1516     qq33             = _mm_mul_ps(iq3,jq3);
1517
1518     /* Avoid stupid compiler warnings */
1519     jnrA = jnrB = jnrC = jnrD = 0;
1520     j_coord_offsetA = 0;
1521     j_coord_offsetB = 0;
1522     j_coord_offsetC = 0;
1523     j_coord_offsetD = 0;
1524
1525     outeriter        = 0;
1526     inneriter        = 0;
1527
1528     for(iidx=0;iidx<4*DIM;iidx++)
1529     {
1530         scratch[iidx] = 0.0;
1531     }
1532
1533     /* Start outer loop over neighborlists */
1534     for(iidx=0; iidx<nri; iidx++)
1535     {
1536         /* Load shift vector for this list */
1537         i_shift_offset   = DIM*shiftidx[iidx];
1538
1539         /* Load limits for loop over neighbors */
1540         j_index_start    = jindex[iidx];
1541         j_index_end      = jindex[iidx+1];
1542
1543         /* Get outer coordinate index */
1544         inr              = iinr[iidx];
1545         i_coord_offset   = DIM*inr;
1546
1547         /* Load i particle coords and add shift vector */
1548         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1549                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1550
1551         fix0             = _mm_setzero_ps();
1552         fiy0             = _mm_setzero_ps();
1553         fiz0             = _mm_setzero_ps();
1554         fix1             = _mm_setzero_ps();
1555         fiy1             = _mm_setzero_ps();
1556         fiz1             = _mm_setzero_ps();
1557         fix2             = _mm_setzero_ps();
1558         fiy2             = _mm_setzero_ps();
1559         fiz2             = _mm_setzero_ps();
1560         fix3             = _mm_setzero_ps();
1561         fiy3             = _mm_setzero_ps();
1562         fiz3             = _mm_setzero_ps();
1563
1564         /* Start inner kernel loop */
1565         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1566         {
1567
1568             /* Get j neighbor index, and coordinate index */
1569             jnrA             = jjnr[jidx];
1570             jnrB             = jjnr[jidx+1];
1571             jnrC             = jjnr[jidx+2];
1572             jnrD             = jjnr[jidx+3];
1573             j_coord_offsetA  = DIM*jnrA;
1574             j_coord_offsetB  = DIM*jnrB;
1575             j_coord_offsetC  = DIM*jnrC;
1576             j_coord_offsetD  = DIM*jnrD;
1577
1578             /* load j atom coordinates */
1579             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1580                                               x+j_coord_offsetC,x+j_coord_offsetD,
1581                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1582                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1583
1584             /* Calculate displacement vector */
1585             dx00             = _mm_sub_ps(ix0,jx0);
1586             dy00             = _mm_sub_ps(iy0,jy0);
1587             dz00             = _mm_sub_ps(iz0,jz0);
1588             dx11             = _mm_sub_ps(ix1,jx1);
1589             dy11             = _mm_sub_ps(iy1,jy1);
1590             dz11             = _mm_sub_ps(iz1,jz1);
1591             dx12             = _mm_sub_ps(ix1,jx2);
1592             dy12             = _mm_sub_ps(iy1,jy2);
1593             dz12             = _mm_sub_ps(iz1,jz2);
1594             dx13             = _mm_sub_ps(ix1,jx3);
1595             dy13             = _mm_sub_ps(iy1,jy3);
1596             dz13             = _mm_sub_ps(iz1,jz3);
1597             dx21             = _mm_sub_ps(ix2,jx1);
1598             dy21             = _mm_sub_ps(iy2,jy1);
1599             dz21             = _mm_sub_ps(iz2,jz1);
1600             dx22             = _mm_sub_ps(ix2,jx2);
1601             dy22             = _mm_sub_ps(iy2,jy2);
1602             dz22             = _mm_sub_ps(iz2,jz2);
1603             dx23             = _mm_sub_ps(ix2,jx3);
1604             dy23             = _mm_sub_ps(iy2,jy3);
1605             dz23             = _mm_sub_ps(iz2,jz3);
1606             dx31             = _mm_sub_ps(ix3,jx1);
1607             dy31             = _mm_sub_ps(iy3,jy1);
1608             dz31             = _mm_sub_ps(iz3,jz1);
1609             dx32             = _mm_sub_ps(ix3,jx2);
1610             dy32             = _mm_sub_ps(iy3,jy2);
1611             dz32             = _mm_sub_ps(iz3,jz2);
1612             dx33             = _mm_sub_ps(ix3,jx3);
1613             dy33             = _mm_sub_ps(iy3,jy3);
1614             dz33             = _mm_sub_ps(iz3,jz3);
1615
1616             /* Calculate squared distance and things based on it */
1617             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1618             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1619             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1620             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1621             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1622             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1623             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1624             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1625             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1626             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1627
1628             rinv00           = sse2_invsqrt_f(rsq00);
1629             rinv11           = sse2_invsqrt_f(rsq11);
1630             rinv12           = sse2_invsqrt_f(rsq12);
1631             rinv13           = sse2_invsqrt_f(rsq13);
1632             rinv21           = sse2_invsqrt_f(rsq21);
1633             rinv22           = sse2_invsqrt_f(rsq22);
1634             rinv23           = sse2_invsqrt_f(rsq23);
1635             rinv31           = sse2_invsqrt_f(rsq31);
1636             rinv32           = sse2_invsqrt_f(rsq32);
1637             rinv33           = sse2_invsqrt_f(rsq33);
1638
1639             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1640             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1641             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1642             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1643             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1644             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1645             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1646             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1647             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1648
1649             fjx0             = _mm_setzero_ps();
1650             fjy0             = _mm_setzero_ps();
1651             fjz0             = _mm_setzero_ps();
1652             fjx1             = _mm_setzero_ps();
1653             fjy1             = _mm_setzero_ps();
1654             fjz1             = _mm_setzero_ps();
1655             fjx2             = _mm_setzero_ps();
1656             fjy2             = _mm_setzero_ps();
1657             fjz2             = _mm_setzero_ps();
1658             fjx3             = _mm_setzero_ps();
1659             fjy3             = _mm_setzero_ps();
1660             fjz3             = _mm_setzero_ps();
1661
1662             /**************************
1663              * CALCULATE INTERACTIONS *
1664              **************************/
1665
1666             r00              = _mm_mul_ps(rsq00,rinv00);
1667
1668             /* Calculate table index by multiplying r with table scale and truncate to integer */
1669             rt               = _mm_mul_ps(r00,vftabscale);
1670             vfitab           = _mm_cvttps_epi32(rt);
1671             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1672             vfitab           = _mm_slli_epi32(vfitab,3);
1673
1674             /* CUBIC SPLINE TABLE DISPERSION */
1675             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1676             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1677             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1678             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1679             _MM_TRANSPOSE4_PS(Y,F,G,H);
1680             Heps             = _mm_mul_ps(vfeps,H);
1681             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1682             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1683             fvdw6            = _mm_mul_ps(c6_00,FF);
1684
1685             /* CUBIC SPLINE TABLE REPULSION */
1686             vfitab           = _mm_add_epi32(vfitab,ifour);
1687             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1688             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1689             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1690             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1691             _MM_TRANSPOSE4_PS(Y,F,G,H);
1692             Heps             = _mm_mul_ps(vfeps,H);
1693             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1694             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1695             fvdw12           = _mm_mul_ps(c12_00,FF);
1696             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1697
1698             fscal            = fvdw;
1699
1700             /* Calculate temporary vectorial force */
1701             tx               = _mm_mul_ps(fscal,dx00);
1702             ty               = _mm_mul_ps(fscal,dy00);
1703             tz               = _mm_mul_ps(fscal,dz00);
1704
1705             /* Update vectorial force */
1706             fix0             = _mm_add_ps(fix0,tx);
1707             fiy0             = _mm_add_ps(fiy0,ty);
1708             fiz0             = _mm_add_ps(fiz0,tz);
1709
1710             fjx0             = _mm_add_ps(fjx0,tx);
1711             fjy0             = _mm_add_ps(fjy0,ty);
1712             fjz0             = _mm_add_ps(fjz0,tz);
1713
1714             /**************************
1715              * CALCULATE INTERACTIONS *
1716              **************************/
1717
1718             r11              = _mm_mul_ps(rsq11,rinv11);
1719
1720             /* EWALD ELECTROSTATICS */
1721
1722             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1723             ewrt             = _mm_mul_ps(r11,ewtabscale);
1724             ewitab           = _mm_cvttps_epi32(ewrt);
1725             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1726             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1727                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1728                                          &ewtabF,&ewtabFn);
1729             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1730             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1731
1732             fscal            = felec;
1733
1734             /* Calculate temporary vectorial force */
1735             tx               = _mm_mul_ps(fscal,dx11);
1736             ty               = _mm_mul_ps(fscal,dy11);
1737             tz               = _mm_mul_ps(fscal,dz11);
1738
1739             /* Update vectorial force */
1740             fix1             = _mm_add_ps(fix1,tx);
1741             fiy1             = _mm_add_ps(fiy1,ty);
1742             fiz1             = _mm_add_ps(fiz1,tz);
1743
1744             fjx1             = _mm_add_ps(fjx1,tx);
1745             fjy1             = _mm_add_ps(fjy1,ty);
1746             fjz1             = _mm_add_ps(fjz1,tz);
1747
1748             /**************************
1749              * CALCULATE INTERACTIONS *
1750              **************************/
1751
1752             r12              = _mm_mul_ps(rsq12,rinv12);
1753
1754             /* EWALD ELECTROSTATICS */
1755
1756             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1757             ewrt             = _mm_mul_ps(r12,ewtabscale);
1758             ewitab           = _mm_cvttps_epi32(ewrt);
1759             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1760             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1761                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1762                                          &ewtabF,&ewtabFn);
1763             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1764             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1765
1766             fscal            = felec;
1767
1768             /* Calculate temporary vectorial force */
1769             tx               = _mm_mul_ps(fscal,dx12);
1770             ty               = _mm_mul_ps(fscal,dy12);
1771             tz               = _mm_mul_ps(fscal,dz12);
1772
1773             /* Update vectorial force */
1774             fix1             = _mm_add_ps(fix1,tx);
1775             fiy1             = _mm_add_ps(fiy1,ty);
1776             fiz1             = _mm_add_ps(fiz1,tz);
1777
1778             fjx2             = _mm_add_ps(fjx2,tx);
1779             fjy2             = _mm_add_ps(fjy2,ty);
1780             fjz2             = _mm_add_ps(fjz2,tz);
1781
1782             /**************************
1783              * CALCULATE INTERACTIONS *
1784              **************************/
1785
1786             r13              = _mm_mul_ps(rsq13,rinv13);
1787
1788             /* EWALD ELECTROSTATICS */
1789
1790             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1791             ewrt             = _mm_mul_ps(r13,ewtabscale);
1792             ewitab           = _mm_cvttps_epi32(ewrt);
1793             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1794             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1795                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1796                                          &ewtabF,&ewtabFn);
1797             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1798             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1799
1800             fscal            = felec;
1801
1802             /* Calculate temporary vectorial force */
1803             tx               = _mm_mul_ps(fscal,dx13);
1804             ty               = _mm_mul_ps(fscal,dy13);
1805             tz               = _mm_mul_ps(fscal,dz13);
1806
1807             /* Update vectorial force */
1808             fix1             = _mm_add_ps(fix1,tx);
1809             fiy1             = _mm_add_ps(fiy1,ty);
1810             fiz1             = _mm_add_ps(fiz1,tz);
1811
1812             fjx3             = _mm_add_ps(fjx3,tx);
1813             fjy3             = _mm_add_ps(fjy3,ty);
1814             fjz3             = _mm_add_ps(fjz3,tz);
1815
1816             /**************************
1817              * CALCULATE INTERACTIONS *
1818              **************************/
1819
1820             r21              = _mm_mul_ps(rsq21,rinv21);
1821
1822             /* EWALD ELECTROSTATICS */
1823
1824             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1825             ewrt             = _mm_mul_ps(r21,ewtabscale);
1826             ewitab           = _mm_cvttps_epi32(ewrt);
1827             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1828             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1829                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1830                                          &ewtabF,&ewtabFn);
1831             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1832             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1833
1834             fscal            = felec;
1835
1836             /* Calculate temporary vectorial force */
1837             tx               = _mm_mul_ps(fscal,dx21);
1838             ty               = _mm_mul_ps(fscal,dy21);
1839             tz               = _mm_mul_ps(fscal,dz21);
1840
1841             /* Update vectorial force */
1842             fix2             = _mm_add_ps(fix2,tx);
1843             fiy2             = _mm_add_ps(fiy2,ty);
1844             fiz2             = _mm_add_ps(fiz2,tz);
1845
1846             fjx1             = _mm_add_ps(fjx1,tx);
1847             fjy1             = _mm_add_ps(fjy1,ty);
1848             fjz1             = _mm_add_ps(fjz1,tz);
1849
1850             /**************************
1851              * CALCULATE INTERACTIONS *
1852              **************************/
1853
1854             r22              = _mm_mul_ps(rsq22,rinv22);
1855
1856             /* EWALD ELECTROSTATICS */
1857
1858             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1859             ewrt             = _mm_mul_ps(r22,ewtabscale);
1860             ewitab           = _mm_cvttps_epi32(ewrt);
1861             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1862             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1863                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1864                                          &ewtabF,&ewtabFn);
1865             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1866             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1867
1868             fscal            = felec;
1869
1870             /* Calculate temporary vectorial force */
1871             tx               = _mm_mul_ps(fscal,dx22);
1872             ty               = _mm_mul_ps(fscal,dy22);
1873             tz               = _mm_mul_ps(fscal,dz22);
1874
1875             /* Update vectorial force */
1876             fix2             = _mm_add_ps(fix2,tx);
1877             fiy2             = _mm_add_ps(fiy2,ty);
1878             fiz2             = _mm_add_ps(fiz2,tz);
1879
1880             fjx2             = _mm_add_ps(fjx2,tx);
1881             fjy2             = _mm_add_ps(fjy2,ty);
1882             fjz2             = _mm_add_ps(fjz2,tz);
1883
1884             /**************************
1885              * CALCULATE INTERACTIONS *
1886              **************************/
1887
1888             r23              = _mm_mul_ps(rsq23,rinv23);
1889
1890             /* EWALD ELECTROSTATICS */
1891
1892             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1893             ewrt             = _mm_mul_ps(r23,ewtabscale);
1894             ewitab           = _mm_cvttps_epi32(ewrt);
1895             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1896             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1897                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1898                                          &ewtabF,&ewtabFn);
1899             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1900             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1901
1902             fscal            = felec;
1903
1904             /* Calculate temporary vectorial force */
1905             tx               = _mm_mul_ps(fscal,dx23);
1906             ty               = _mm_mul_ps(fscal,dy23);
1907             tz               = _mm_mul_ps(fscal,dz23);
1908
1909             /* Update vectorial force */
1910             fix2             = _mm_add_ps(fix2,tx);
1911             fiy2             = _mm_add_ps(fiy2,ty);
1912             fiz2             = _mm_add_ps(fiz2,tz);
1913
1914             fjx3             = _mm_add_ps(fjx3,tx);
1915             fjy3             = _mm_add_ps(fjy3,ty);
1916             fjz3             = _mm_add_ps(fjz3,tz);
1917
1918             /**************************
1919              * CALCULATE INTERACTIONS *
1920              **************************/
1921
1922             r31              = _mm_mul_ps(rsq31,rinv31);
1923
1924             /* EWALD ELECTROSTATICS */
1925
1926             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1927             ewrt             = _mm_mul_ps(r31,ewtabscale);
1928             ewitab           = _mm_cvttps_epi32(ewrt);
1929             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1930             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1931                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1932                                          &ewtabF,&ewtabFn);
1933             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1934             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1935
1936             fscal            = felec;
1937
1938             /* Calculate temporary vectorial force */
1939             tx               = _mm_mul_ps(fscal,dx31);
1940             ty               = _mm_mul_ps(fscal,dy31);
1941             tz               = _mm_mul_ps(fscal,dz31);
1942
1943             /* Update vectorial force */
1944             fix3             = _mm_add_ps(fix3,tx);
1945             fiy3             = _mm_add_ps(fiy3,ty);
1946             fiz3             = _mm_add_ps(fiz3,tz);
1947
1948             fjx1             = _mm_add_ps(fjx1,tx);
1949             fjy1             = _mm_add_ps(fjy1,ty);
1950             fjz1             = _mm_add_ps(fjz1,tz);
1951
1952             /**************************
1953              * CALCULATE INTERACTIONS *
1954              **************************/
1955
1956             r32              = _mm_mul_ps(rsq32,rinv32);
1957
1958             /* EWALD ELECTROSTATICS */
1959
1960             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1961             ewrt             = _mm_mul_ps(r32,ewtabscale);
1962             ewitab           = _mm_cvttps_epi32(ewrt);
1963             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1964             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1965                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1966                                          &ewtabF,&ewtabFn);
1967             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1968             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1969
1970             fscal            = felec;
1971
1972             /* Calculate temporary vectorial force */
1973             tx               = _mm_mul_ps(fscal,dx32);
1974             ty               = _mm_mul_ps(fscal,dy32);
1975             tz               = _mm_mul_ps(fscal,dz32);
1976
1977             /* Update vectorial force */
1978             fix3             = _mm_add_ps(fix3,tx);
1979             fiy3             = _mm_add_ps(fiy3,ty);
1980             fiz3             = _mm_add_ps(fiz3,tz);
1981
1982             fjx2             = _mm_add_ps(fjx2,tx);
1983             fjy2             = _mm_add_ps(fjy2,ty);
1984             fjz2             = _mm_add_ps(fjz2,tz);
1985
1986             /**************************
1987              * CALCULATE INTERACTIONS *
1988              **************************/
1989
1990             r33              = _mm_mul_ps(rsq33,rinv33);
1991
1992             /* EWALD ELECTROSTATICS */
1993
1994             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1995             ewrt             = _mm_mul_ps(r33,ewtabscale);
1996             ewitab           = _mm_cvttps_epi32(ewrt);
1997             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1998             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1999                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2000                                          &ewtabF,&ewtabFn);
2001             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2002             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2003
2004             fscal            = felec;
2005
2006             /* Calculate temporary vectorial force */
2007             tx               = _mm_mul_ps(fscal,dx33);
2008             ty               = _mm_mul_ps(fscal,dy33);
2009             tz               = _mm_mul_ps(fscal,dz33);
2010
2011             /* Update vectorial force */
2012             fix3             = _mm_add_ps(fix3,tx);
2013             fiy3             = _mm_add_ps(fiy3,ty);
2014             fiz3             = _mm_add_ps(fiz3,tz);
2015
2016             fjx3             = _mm_add_ps(fjx3,tx);
2017             fjy3             = _mm_add_ps(fjy3,ty);
2018             fjz3             = _mm_add_ps(fjz3,tz);
2019
2020             fjptrA             = f+j_coord_offsetA;
2021             fjptrB             = f+j_coord_offsetB;
2022             fjptrC             = f+j_coord_offsetC;
2023             fjptrD             = f+j_coord_offsetD;
2024
2025             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2026                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2027                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2028
2029             /* Inner loop uses 375 flops */
2030         }
2031
2032         if(jidx<j_index_end)
2033         {
2034
2035             /* Get j neighbor index, and coordinate index */
2036             jnrlistA         = jjnr[jidx];
2037             jnrlistB         = jjnr[jidx+1];
2038             jnrlistC         = jjnr[jidx+2];
2039             jnrlistD         = jjnr[jidx+3];
2040             /* Sign of each element will be negative for non-real atoms.
2041              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2042              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2043              */
2044             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2045             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2046             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2047             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2048             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2049             j_coord_offsetA  = DIM*jnrA;
2050             j_coord_offsetB  = DIM*jnrB;
2051             j_coord_offsetC  = DIM*jnrC;
2052             j_coord_offsetD  = DIM*jnrD;
2053
2054             /* load j atom coordinates */
2055             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2056                                               x+j_coord_offsetC,x+j_coord_offsetD,
2057                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2058                                               &jy2,&jz2,&jx3,&jy3,&jz3);
2059
2060             /* Calculate displacement vector */
2061             dx00             = _mm_sub_ps(ix0,jx0);
2062             dy00             = _mm_sub_ps(iy0,jy0);
2063             dz00             = _mm_sub_ps(iz0,jz0);
2064             dx11             = _mm_sub_ps(ix1,jx1);
2065             dy11             = _mm_sub_ps(iy1,jy1);
2066             dz11             = _mm_sub_ps(iz1,jz1);
2067             dx12             = _mm_sub_ps(ix1,jx2);
2068             dy12             = _mm_sub_ps(iy1,jy2);
2069             dz12             = _mm_sub_ps(iz1,jz2);
2070             dx13             = _mm_sub_ps(ix1,jx3);
2071             dy13             = _mm_sub_ps(iy1,jy3);
2072             dz13             = _mm_sub_ps(iz1,jz3);
2073             dx21             = _mm_sub_ps(ix2,jx1);
2074             dy21             = _mm_sub_ps(iy2,jy1);
2075             dz21             = _mm_sub_ps(iz2,jz1);
2076             dx22             = _mm_sub_ps(ix2,jx2);
2077             dy22             = _mm_sub_ps(iy2,jy2);
2078             dz22             = _mm_sub_ps(iz2,jz2);
2079             dx23             = _mm_sub_ps(ix2,jx3);
2080             dy23             = _mm_sub_ps(iy2,jy3);
2081             dz23             = _mm_sub_ps(iz2,jz3);
2082             dx31             = _mm_sub_ps(ix3,jx1);
2083             dy31             = _mm_sub_ps(iy3,jy1);
2084             dz31             = _mm_sub_ps(iz3,jz1);
2085             dx32             = _mm_sub_ps(ix3,jx2);
2086             dy32             = _mm_sub_ps(iy3,jy2);
2087             dz32             = _mm_sub_ps(iz3,jz2);
2088             dx33             = _mm_sub_ps(ix3,jx3);
2089             dy33             = _mm_sub_ps(iy3,jy3);
2090             dz33             = _mm_sub_ps(iz3,jz3);
2091
2092             /* Calculate squared distance and things based on it */
2093             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2094             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2095             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2096             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2097             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2098             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2099             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2100             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2101             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2102             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2103
2104             rinv00           = sse2_invsqrt_f(rsq00);
2105             rinv11           = sse2_invsqrt_f(rsq11);
2106             rinv12           = sse2_invsqrt_f(rsq12);
2107             rinv13           = sse2_invsqrt_f(rsq13);
2108             rinv21           = sse2_invsqrt_f(rsq21);
2109             rinv22           = sse2_invsqrt_f(rsq22);
2110             rinv23           = sse2_invsqrt_f(rsq23);
2111             rinv31           = sse2_invsqrt_f(rsq31);
2112             rinv32           = sse2_invsqrt_f(rsq32);
2113             rinv33           = sse2_invsqrt_f(rsq33);
2114
2115             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
2116             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
2117             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
2118             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
2119             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
2120             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
2121             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
2122             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
2123             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
2124
2125             fjx0             = _mm_setzero_ps();
2126             fjy0             = _mm_setzero_ps();
2127             fjz0             = _mm_setzero_ps();
2128             fjx1             = _mm_setzero_ps();
2129             fjy1             = _mm_setzero_ps();
2130             fjz1             = _mm_setzero_ps();
2131             fjx2             = _mm_setzero_ps();
2132             fjy2             = _mm_setzero_ps();
2133             fjz2             = _mm_setzero_ps();
2134             fjx3             = _mm_setzero_ps();
2135             fjy3             = _mm_setzero_ps();
2136             fjz3             = _mm_setzero_ps();
2137
2138             /**************************
2139              * CALCULATE INTERACTIONS *
2140              **************************/
2141
2142             r00              = _mm_mul_ps(rsq00,rinv00);
2143             r00              = _mm_andnot_ps(dummy_mask,r00);
2144
2145             /* Calculate table index by multiplying r with table scale and truncate to integer */
2146             rt               = _mm_mul_ps(r00,vftabscale);
2147             vfitab           = _mm_cvttps_epi32(rt);
2148             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2149             vfitab           = _mm_slli_epi32(vfitab,3);
2150
2151             /* CUBIC SPLINE TABLE DISPERSION */
2152             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2153             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2154             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2155             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2156             _MM_TRANSPOSE4_PS(Y,F,G,H);
2157             Heps             = _mm_mul_ps(vfeps,H);
2158             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2159             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2160             fvdw6            = _mm_mul_ps(c6_00,FF);
2161
2162             /* CUBIC SPLINE TABLE REPULSION */
2163             vfitab           = _mm_add_epi32(vfitab,ifour);
2164             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2165             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2166             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2167             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2168             _MM_TRANSPOSE4_PS(Y,F,G,H);
2169             Heps             = _mm_mul_ps(vfeps,H);
2170             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2171             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2172             fvdw12           = _mm_mul_ps(c12_00,FF);
2173             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
2174
2175             fscal            = fvdw;
2176
2177             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2178
2179             /* Calculate temporary vectorial force */
2180             tx               = _mm_mul_ps(fscal,dx00);
2181             ty               = _mm_mul_ps(fscal,dy00);
2182             tz               = _mm_mul_ps(fscal,dz00);
2183
2184             /* Update vectorial force */
2185             fix0             = _mm_add_ps(fix0,tx);
2186             fiy0             = _mm_add_ps(fiy0,ty);
2187             fiz0             = _mm_add_ps(fiz0,tz);
2188
2189             fjx0             = _mm_add_ps(fjx0,tx);
2190             fjy0             = _mm_add_ps(fjy0,ty);
2191             fjz0             = _mm_add_ps(fjz0,tz);
2192
2193             /**************************
2194              * CALCULATE INTERACTIONS *
2195              **************************/
2196
2197             r11              = _mm_mul_ps(rsq11,rinv11);
2198             r11              = _mm_andnot_ps(dummy_mask,r11);
2199
2200             /* EWALD ELECTROSTATICS */
2201
2202             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2203             ewrt             = _mm_mul_ps(r11,ewtabscale);
2204             ewitab           = _mm_cvttps_epi32(ewrt);
2205             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2206             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2207                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2208                                          &ewtabF,&ewtabFn);
2209             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2210             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
2211
2212             fscal            = felec;
2213
2214             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2215
2216             /* Calculate temporary vectorial force */
2217             tx               = _mm_mul_ps(fscal,dx11);
2218             ty               = _mm_mul_ps(fscal,dy11);
2219             tz               = _mm_mul_ps(fscal,dz11);
2220
2221             /* Update vectorial force */
2222             fix1             = _mm_add_ps(fix1,tx);
2223             fiy1             = _mm_add_ps(fiy1,ty);
2224             fiz1             = _mm_add_ps(fiz1,tz);
2225
2226             fjx1             = _mm_add_ps(fjx1,tx);
2227             fjy1             = _mm_add_ps(fjy1,ty);
2228             fjz1             = _mm_add_ps(fjz1,tz);
2229
2230             /**************************
2231              * CALCULATE INTERACTIONS *
2232              **************************/
2233
2234             r12              = _mm_mul_ps(rsq12,rinv12);
2235             r12              = _mm_andnot_ps(dummy_mask,r12);
2236
2237             /* EWALD ELECTROSTATICS */
2238
2239             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2240             ewrt             = _mm_mul_ps(r12,ewtabscale);
2241             ewitab           = _mm_cvttps_epi32(ewrt);
2242             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2243             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2244                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2245                                          &ewtabF,&ewtabFn);
2246             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2247             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
2248
2249             fscal            = felec;
2250
2251             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2252
2253             /* Calculate temporary vectorial force */
2254             tx               = _mm_mul_ps(fscal,dx12);
2255             ty               = _mm_mul_ps(fscal,dy12);
2256             tz               = _mm_mul_ps(fscal,dz12);
2257
2258             /* Update vectorial force */
2259             fix1             = _mm_add_ps(fix1,tx);
2260             fiy1             = _mm_add_ps(fiy1,ty);
2261             fiz1             = _mm_add_ps(fiz1,tz);
2262
2263             fjx2             = _mm_add_ps(fjx2,tx);
2264             fjy2             = _mm_add_ps(fjy2,ty);
2265             fjz2             = _mm_add_ps(fjz2,tz);
2266
2267             /**************************
2268              * CALCULATE INTERACTIONS *
2269              **************************/
2270
2271             r13              = _mm_mul_ps(rsq13,rinv13);
2272             r13              = _mm_andnot_ps(dummy_mask,r13);
2273
2274             /* EWALD ELECTROSTATICS */
2275
2276             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2277             ewrt             = _mm_mul_ps(r13,ewtabscale);
2278             ewitab           = _mm_cvttps_epi32(ewrt);
2279             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2280             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2281                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2282                                          &ewtabF,&ewtabFn);
2283             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2284             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
2285
2286             fscal            = felec;
2287
2288             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2289
2290             /* Calculate temporary vectorial force */
2291             tx               = _mm_mul_ps(fscal,dx13);
2292             ty               = _mm_mul_ps(fscal,dy13);
2293             tz               = _mm_mul_ps(fscal,dz13);
2294
2295             /* Update vectorial force */
2296             fix1             = _mm_add_ps(fix1,tx);
2297             fiy1             = _mm_add_ps(fiy1,ty);
2298             fiz1             = _mm_add_ps(fiz1,tz);
2299
2300             fjx3             = _mm_add_ps(fjx3,tx);
2301             fjy3             = _mm_add_ps(fjy3,ty);
2302             fjz3             = _mm_add_ps(fjz3,tz);
2303
2304             /**************************
2305              * CALCULATE INTERACTIONS *
2306              **************************/
2307
2308             r21              = _mm_mul_ps(rsq21,rinv21);
2309             r21              = _mm_andnot_ps(dummy_mask,r21);
2310
2311             /* EWALD ELECTROSTATICS */
2312
2313             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2314             ewrt             = _mm_mul_ps(r21,ewtabscale);
2315             ewitab           = _mm_cvttps_epi32(ewrt);
2316             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2317             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2318                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2319                                          &ewtabF,&ewtabFn);
2320             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2321             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
2322
2323             fscal            = felec;
2324
2325             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2326
2327             /* Calculate temporary vectorial force */
2328             tx               = _mm_mul_ps(fscal,dx21);
2329             ty               = _mm_mul_ps(fscal,dy21);
2330             tz               = _mm_mul_ps(fscal,dz21);
2331
2332             /* Update vectorial force */
2333             fix2             = _mm_add_ps(fix2,tx);
2334             fiy2             = _mm_add_ps(fiy2,ty);
2335             fiz2             = _mm_add_ps(fiz2,tz);
2336
2337             fjx1             = _mm_add_ps(fjx1,tx);
2338             fjy1             = _mm_add_ps(fjy1,ty);
2339             fjz1             = _mm_add_ps(fjz1,tz);
2340
2341             /**************************
2342              * CALCULATE INTERACTIONS *
2343              **************************/
2344
2345             r22              = _mm_mul_ps(rsq22,rinv22);
2346             r22              = _mm_andnot_ps(dummy_mask,r22);
2347
2348             /* EWALD ELECTROSTATICS */
2349
2350             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2351             ewrt             = _mm_mul_ps(r22,ewtabscale);
2352             ewitab           = _mm_cvttps_epi32(ewrt);
2353             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2354             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2355                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2356                                          &ewtabF,&ewtabFn);
2357             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2358             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
2359
2360             fscal            = felec;
2361
2362             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2363
2364             /* Calculate temporary vectorial force */
2365             tx               = _mm_mul_ps(fscal,dx22);
2366             ty               = _mm_mul_ps(fscal,dy22);
2367             tz               = _mm_mul_ps(fscal,dz22);
2368
2369             /* Update vectorial force */
2370             fix2             = _mm_add_ps(fix2,tx);
2371             fiy2             = _mm_add_ps(fiy2,ty);
2372             fiz2             = _mm_add_ps(fiz2,tz);
2373
2374             fjx2             = _mm_add_ps(fjx2,tx);
2375             fjy2             = _mm_add_ps(fjy2,ty);
2376             fjz2             = _mm_add_ps(fjz2,tz);
2377
2378             /**************************
2379              * CALCULATE INTERACTIONS *
2380              **************************/
2381
2382             r23              = _mm_mul_ps(rsq23,rinv23);
2383             r23              = _mm_andnot_ps(dummy_mask,r23);
2384
2385             /* EWALD ELECTROSTATICS */
2386
2387             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2388             ewrt             = _mm_mul_ps(r23,ewtabscale);
2389             ewitab           = _mm_cvttps_epi32(ewrt);
2390             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2391             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2392                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2393                                          &ewtabF,&ewtabFn);
2394             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2395             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
2396
2397             fscal            = felec;
2398
2399             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2400
2401             /* Calculate temporary vectorial force */
2402             tx               = _mm_mul_ps(fscal,dx23);
2403             ty               = _mm_mul_ps(fscal,dy23);
2404             tz               = _mm_mul_ps(fscal,dz23);
2405
2406             /* Update vectorial force */
2407             fix2             = _mm_add_ps(fix2,tx);
2408             fiy2             = _mm_add_ps(fiy2,ty);
2409             fiz2             = _mm_add_ps(fiz2,tz);
2410
2411             fjx3             = _mm_add_ps(fjx3,tx);
2412             fjy3             = _mm_add_ps(fjy3,ty);
2413             fjz3             = _mm_add_ps(fjz3,tz);
2414
2415             /**************************
2416              * CALCULATE INTERACTIONS *
2417              **************************/
2418
2419             r31              = _mm_mul_ps(rsq31,rinv31);
2420             r31              = _mm_andnot_ps(dummy_mask,r31);
2421
2422             /* EWALD ELECTROSTATICS */
2423
2424             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2425             ewrt             = _mm_mul_ps(r31,ewtabscale);
2426             ewitab           = _mm_cvttps_epi32(ewrt);
2427             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2428             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2429                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2430                                          &ewtabF,&ewtabFn);
2431             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2432             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
2433
2434             fscal            = felec;
2435
2436             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2437
2438             /* Calculate temporary vectorial force */
2439             tx               = _mm_mul_ps(fscal,dx31);
2440             ty               = _mm_mul_ps(fscal,dy31);
2441             tz               = _mm_mul_ps(fscal,dz31);
2442
2443             /* Update vectorial force */
2444             fix3             = _mm_add_ps(fix3,tx);
2445             fiy3             = _mm_add_ps(fiy3,ty);
2446             fiz3             = _mm_add_ps(fiz3,tz);
2447
2448             fjx1             = _mm_add_ps(fjx1,tx);
2449             fjy1             = _mm_add_ps(fjy1,ty);
2450             fjz1             = _mm_add_ps(fjz1,tz);
2451
2452             /**************************
2453              * CALCULATE INTERACTIONS *
2454              **************************/
2455
2456             r32              = _mm_mul_ps(rsq32,rinv32);
2457             r32              = _mm_andnot_ps(dummy_mask,r32);
2458
2459             /* EWALD ELECTROSTATICS */
2460
2461             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2462             ewrt             = _mm_mul_ps(r32,ewtabscale);
2463             ewitab           = _mm_cvttps_epi32(ewrt);
2464             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2465             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2466                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2467                                          &ewtabF,&ewtabFn);
2468             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2469             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
2470
2471             fscal            = felec;
2472
2473             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2474
2475             /* Calculate temporary vectorial force */
2476             tx               = _mm_mul_ps(fscal,dx32);
2477             ty               = _mm_mul_ps(fscal,dy32);
2478             tz               = _mm_mul_ps(fscal,dz32);
2479
2480             /* Update vectorial force */
2481             fix3             = _mm_add_ps(fix3,tx);
2482             fiy3             = _mm_add_ps(fiy3,ty);
2483             fiz3             = _mm_add_ps(fiz3,tz);
2484
2485             fjx2             = _mm_add_ps(fjx2,tx);
2486             fjy2             = _mm_add_ps(fjy2,ty);
2487             fjz2             = _mm_add_ps(fjz2,tz);
2488
2489             /**************************
2490              * CALCULATE INTERACTIONS *
2491              **************************/
2492
2493             r33              = _mm_mul_ps(rsq33,rinv33);
2494             r33              = _mm_andnot_ps(dummy_mask,r33);
2495
2496             /* EWALD ELECTROSTATICS */
2497
2498             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2499             ewrt             = _mm_mul_ps(r33,ewtabscale);
2500             ewitab           = _mm_cvttps_epi32(ewrt);
2501             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2502             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2503                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2504                                          &ewtabF,&ewtabFn);
2505             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2506             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2507
2508             fscal            = felec;
2509
2510             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2511
2512             /* Calculate temporary vectorial force */
2513             tx               = _mm_mul_ps(fscal,dx33);
2514             ty               = _mm_mul_ps(fscal,dy33);
2515             tz               = _mm_mul_ps(fscal,dz33);
2516
2517             /* Update vectorial force */
2518             fix3             = _mm_add_ps(fix3,tx);
2519             fiy3             = _mm_add_ps(fiy3,ty);
2520             fiz3             = _mm_add_ps(fiz3,tz);
2521
2522             fjx3             = _mm_add_ps(fjx3,tx);
2523             fjy3             = _mm_add_ps(fjy3,ty);
2524             fjz3             = _mm_add_ps(fjz3,tz);
2525
2526             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2527             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2528             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2529             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2530
2531             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2532                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2533                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2534
2535             /* Inner loop uses 385 flops */
2536         }
2537
2538         /* End of innermost loop */
2539
2540         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2541                                               f+i_coord_offset,fshift+i_shift_offset);
2542
2543         /* Increment number of inner iterations */
2544         inneriter                  += j_index_end - j_index_start;
2545
2546         /* Outer loop uses 24 flops */
2547     }
2548
2549     /* Increment number of outer iterations */
2550     outeriter        += nri;
2551
2552     /* Update outer/inner flops */
2553
2554     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*385);
2555 }