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

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse4_1_single.h"
  48 #include "kernelutil_x86_sse4_1_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_sse4_1_single
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_sse4_1_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  83     int              vdwioffset0;
  84     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  85     int              vdwioffset1;
  86     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  87     int              vdwioffset2;
  88     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  89     int              vdwioffset3;
  90     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  92     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  94     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  95     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  96     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  97     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  98     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  99     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 100     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 101     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 102     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 103     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 104     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 105     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 106     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 107     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 108     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 109     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 110     real             *charge;
 111     int              nvdwtype;
 112     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 113     int              *vdwtype;
 114     real             *vdwparam;
 115     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 116     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 117     __m128i          vfitab;
 118     __m128i          ifour       = _mm_set1_epi32(4);
 119     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 120     real             *vftab;
 121     __m128i          ewitab;
 122     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 123     real             *ewtab;
 124     __m128           dummy_mask,cutoff_mask;
 125     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 126     __m128           one     = _mm_set1_ps(1.0);
 127     __m128           two     = _mm_set1_ps(2.0);
 128     x                = xx[0];
 129     f                = ff[0];
 130
 131     nri              = nlist->nri;
 132     iinr             = nlist->iinr;
 133     jindex           = nlist->jindex;
 134     jjnr             = nlist->jjnr;
 135     shiftidx         = nlist->shift;
 136     gid              = nlist->gid;
 137     shiftvec         = fr->shift_vec[0];
 138     fshift           = fr->fshift[0];
 139     facel            = _mm_set1_ps(fr->epsfac);
 140     charge           = mdatoms->chargeA;
 141     nvdwtype         = fr->ntype;
 142     vdwparam         = fr->nbfp;
 143     vdwtype          = mdatoms->typeA;
 144
 145     vftab            = kernel_data->table_vdw->data;
 146     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 147
 148     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 149     ewtab            = fr->ic->tabq_coul_FDV0;
 150     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 151     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 152
 153     /* Setup water-specific parameters */
 154     inr              = nlist->iinr[0];
 155     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 156     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 157     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 158     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 159
 160     jq1              = _mm_set1_ps(charge[inr+1]);
 161     jq2              = _mm_set1_ps(charge[inr+2]);
 162     jq3              = _mm_set1_ps(charge[inr+3]);
 163     vdwjidx0A        = 2*vdwtype[inr+0];
 164     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 165     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 166     qq11             = _mm_mul_ps(iq1,jq1);
 167     qq12             = _mm_mul_ps(iq1,jq2);
 168     qq13             = _mm_mul_ps(iq1,jq3);
 169     qq21             = _mm_mul_ps(iq2,jq1);
 170     qq22             = _mm_mul_ps(iq2,jq2);
 171     qq23             = _mm_mul_ps(iq2,jq3);
 172     qq31             = _mm_mul_ps(iq3,jq1);
 173     qq32             = _mm_mul_ps(iq3,jq2);
 174     qq33             = _mm_mul_ps(iq3,jq3);
 175
 176     /* Avoid stupid compiler warnings */
 177     jnrA = jnrB = jnrC = jnrD = 0;
 178     j_coord_offsetA = 0;
 179     j_coord_offsetB = 0;
 180     j_coord_offsetC = 0;
 181     j_coord_offsetD = 0;
 182
 183     outeriter        = 0;
 184     inneriter        = 0;
 185
 186     for(iidx=0;iidx<4*DIM;iidx++)
 187     {
 188         scratch[iidx] = 0.0;
 189     }
 190
 191     /* Start outer loop over neighborlists */
 192     for(iidx=0; iidx<nri; iidx++)
 193     {
 194         /* Load shift vector for this list */
 195         i_shift_offset   = DIM*shiftidx[iidx];
 196
 197         /* Load limits for loop over neighbors */
 198         j_index_start    = jindex[iidx];
 199         j_index_end      = jindex[iidx+1];
 200
 201         /* Get outer coordinate index */
 202         inr              = iinr[iidx];
 203         i_coord_offset   = DIM*inr;
 204
 205         /* Load i particle coords and add shift vector */
 206         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 207                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 208
 209         fix0             = _mm_setzero_ps();
 210         fiy0             = _mm_setzero_ps();
 211         fiz0             = _mm_setzero_ps();
 212         fix1             = _mm_setzero_ps();
 213         fiy1             = _mm_setzero_ps();
 214         fiz1             = _mm_setzero_ps();
 215         fix2             = _mm_setzero_ps();
 216         fiy2             = _mm_setzero_ps();
 217         fiz2             = _mm_setzero_ps();
 218         fix3             = _mm_setzero_ps();
 219         fiy3             = _mm_setzero_ps();
 220         fiz3             = _mm_setzero_ps();
 221
 222         /* Reset potential sums */
 223         velecsum         = _mm_setzero_ps();
 224         vvdwsum          = _mm_setzero_ps();
 225
 226         /* Start inner kernel loop */
 227         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 228         {
 229
 230             /* Get j neighbor index, and coordinate index */
 231             jnrA             = jjnr[jidx];
 232             jnrB             = jjnr[jidx+1];
 233             jnrC             = jjnr[jidx+2];
 234             jnrD             = jjnr[jidx+3];
 235             j_coord_offsetA  = DIM*jnrA;
 236             j_coord_offsetB  = DIM*jnrB;
 237             j_coord_offsetC  = DIM*jnrC;
 238             j_coord_offsetD  = DIM*jnrD;
 239
 240             /* load j atom coordinates */
 241             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 242                                               x+j_coord_offsetC,x+j_coord_offsetD,
 243                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 244                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 245
 246             /* Calculate displacement vector */
 247             dx00             = _mm_sub_ps(ix0,jx0);
 248             dy00             = _mm_sub_ps(iy0,jy0);
 249             dz00             = _mm_sub_ps(iz0,jz0);
 250             dx11             = _mm_sub_ps(ix1,jx1);
 251             dy11             = _mm_sub_ps(iy1,jy1);
 252             dz11             = _mm_sub_ps(iz1,jz1);
 253             dx12             = _mm_sub_ps(ix1,jx2);
 254             dy12             = _mm_sub_ps(iy1,jy2);
 255             dz12             = _mm_sub_ps(iz1,jz2);
 256             dx13             = _mm_sub_ps(ix1,jx3);
 257             dy13             = _mm_sub_ps(iy1,jy3);
 258             dz13             = _mm_sub_ps(iz1,jz3);
 259             dx21             = _mm_sub_ps(ix2,jx1);
 260             dy21             = _mm_sub_ps(iy2,jy1);
 261             dz21             = _mm_sub_ps(iz2,jz1);
 262             dx22             = _mm_sub_ps(ix2,jx2);
 263             dy22             = _mm_sub_ps(iy2,jy2);
 264             dz22             = _mm_sub_ps(iz2,jz2);
 265             dx23             = _mm_sub_ps(ix2,jx3);
 266             dy23             = _mm_sub_ps(iy2,jy3);
 267             dz23             = _mm_sub_ps(iz2,jz3);
 268             dx31             = _mm_sub_ps(ix3,jx1);
 269             dy31             = _mm_sub_ps(iy3,jy1);
 270             dz31             = _mm_sub_ps(iz3,jz1);
 271             dx32             = _mm_sub_ps(ix3,jx2);
 272             dy32             = _mm_sub_ps(iy3,jy2);
 273             dz32             = _mm_sub_ps(iz3,jz2);
 274             dx33             = _mm_sub_ps(ix3,jx3);
 275             dy33             = _mm_sub_ps(iy3,jy3);
 276             dz33             = _mm_sub_ps(iz3,jz3);
 277
 278             /* Calculate squared distance and things based on it */
 279             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 280             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 281             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 282             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 283             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 284             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 285             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 286             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 287             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 288             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 289
 290             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 291             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 292             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 293             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 294             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 295             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 296             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 297             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 298             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 299             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 300
 301             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 302             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 303             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 304             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 305             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 306             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 307             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 308             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 309             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 310
 311             fjx0             = _mm_setzero_ps();
 312             fjy0             = _mm_setzero_ps();
 313             fjz0             = _mm_setzero_ps();
 314             fjx1             = _mm_setzero_ps();
 315             fjy1             = _mm_setzero_ps();
 316             fjz1             = _mm_setzero_ps();
 317             fjx2             = _mm_setzero_ps();
 318             fjy2             = _mm_setzero_ps();
 319             fjz2             = _mm_setzero_ps();
 320             fjx3             = _mm_setzero_ps();
 321             fjy3             = _mm_setzero_ps();
 322             fjz3             = _mm_setzero_ps();
 323
 324             /**************************
 325              * CALCULATE INTERACTIONS *
 326              **************************/
 327
 328             r00              = _mm_mul_ps(rsq00,rinv00);
 329
 330             /* Calculate table index by multiplying r with table scale and truncate to integer */
 331             rt               = _mm_mul_ps(r00,vftabscale);
 332             vfitab           = _mm_cvttps_epi32(rt);
 333             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 334             vfitab           = _mm_slli_epi32(vfitab,3);
 335
 336             /* CUBIC SPLINE TABLE DISPERSION */
 337             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 338             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 339             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 340             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 341             _MM_TRANSPOSE4_PS(Y,F,G,H);
 342             Heps             = _mm_mul_ps(vfeps,H);
 343             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 344             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 345             vvdw6            = _mm_mul_ps(c6_00,VV);
 346             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 347             fvdw6            = _mm_mul_ps(c6_00,FF);
 348
 349             /* CUBIC SPLINE TABLE REPULSION */
 350             vfitab           = _mm_add_epi32(vfitab,ifour);
 351             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 352             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 353             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 354             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 355             _MM_TRANSPOSE4_PS(Y,F,G,H);
 356             Heps             = _mm_mul_ps(vfeps,H);
 357             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 358             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 359             vvdw12           = _mm_mul_ps(c12_00,VV);
 360             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 361             fvdw12           = _mm_mul_ps(c12_00,FF);
 362             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 363             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 364
 365             /* Update potential sum for this i atom from the interaction with this j atom. */
 366             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 367
 368             fscal            = fvdw;
 369
 370             /* Calculate temporary vectorial force */
 371             tx               = _mm_mul_ps(fscal,dx00);
 372             ty               = _mm_mul_ps(fscal,dy00);
 373             tz               = _mm_mul_ps(fscal,dz00);
 374
 375             /* Update vectorial force */
 376             fix0             = _mm_add_ps(fix0,tx);
 377             fiy0             = _mm_add_ps(fiy0,ty);
 378             fiz0             = _mm_add_ps(fiz0,tz);
 379
 380             fjx0             = _mm_add_ps(fjx0,tx);
 381             fjy0             = _mm_add_ps(fjy0,ty);
 382             fjz0             = _mm_add_ps(fjz0,tz);
 383
 384             /**************************
 385              * CALCULATE INTERACTIONS *
 386              **************************/
 387
 388             r11              = _mm_mul_ps(rsq11,rinv11);
 389
 390             /* EWALD ELECTROSTATICS */
 391
 392             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 393             ewrt             = _mm_mul_ps(r11,ewtabscale);
 394             ewitab           = _mm_cvttps_epi32(ewrt);
 395             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 396             ewitab           = _mm_slli_epi32(ewitab,2);
 397             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 398             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 399             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 400             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 401             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 402             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 403             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 404             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 405             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 406
 407             /* Update potential sum for this i atom from the interaction with this j atom. */
 408             velecsum         = _mm_add_ps(velecsum,velec);
 409
 410             fscal            = felec;
 411
 412             /* Calculate temporary vectorial force */
 413             tx               = _mm_mul_ps(fscal,dx11);
 414             ty               = _mm_mul_ps(fscal,dy11);
 415             tz               = _mm_mul_ps(fscal,dz11);
 416
 417             /* Update vectorial force */
 418             fix1             = _mm_add_ps(fix1,tx);
 419             fiy1             = _mm_add_ps(fiy1,ty);
 420             fiz1             = _mm_add_ps(fiz1,tz);
 421
 422             fjx1             = _mm_add_ps(fjx1,tx);
 423             fjy1             = _mm_add_ps(fjy1,ty);
 424             fjz1             = _mm_add_ps(fjz1,tz);
 425
 426             /**************************
 427              * CALCULATE INTERACTIONS *
 428              **************************/
 429
 430             r12              = _mm_mul_ps(rsq12,rinv12);
 431
 432             /* EWALD ELECTROSTATICS */
 433
 434             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 435             ewrt             = _mm_mul_ps(r12,ewtabscale);
 436             ewitab           = _mm_cvttps_epi32(ewrt);
 437             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 438             ewitab           = _mm_slli_epi32(ewitab,2);
 439             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 440             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 441             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 442             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 443             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 444             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 445             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 446             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
 447             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
 448
 449             /* Update potential sum for this i atom from the interaction with this j atom. */
 450             velecsum         = _mm_add_ps(velecsum,velec);
 451
 452             fscal            = felec;
 453
 454             /* Calculate temporary vectorial force */
 455             tx               = _mm_mul_ps(fscal,dx12);
 456             ty               = _mm_mul_ps(fscal,dy12);
 457             tz               = _mm_mul_ps(fscal,dz12);
 458
 459             /* Update vectorial force */
 460             fix1             = _mm_add_ps(fix1,tx);
 461             fiy1             = _mm_add_ps(fiy1,ty);
 462             fiz1             = _mm_add_ps(fiz1,tz);
 463
 464             fjx2             = _mm_add_ps(fjx2,tx);
 465             fjy2             = _mm_add_ps(fjy2,ty);
 466             fjz2             = _mm_add_ps(fjz2,tz);
 467
 468             /**************************
 469              * CALCULATE INTERACTIONS *
 470              **************************/
 471
 472             r13              = _mm_mul_ps(rsq13,rinv13);
 473
 474             /* EWALD ELECTROSTATICS */
 475
 476             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 477             ewrt             = _mm_mul_ps(r13,ewtabscale);
 478             ewitab           = _mm_cvttps_epi32(ewrt);
 479             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 480             ewitab           = _mm_slli_epi32(ewitab,2);
 481             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 482             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 483             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 484             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 485             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 486             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 487             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 488             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
 489             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
 490
 491             /* Update potential sum for this i atom from the interaction with this j atom. */
 492             velecsum         = _mm_add_ps(velecsum,velec);
 493
 494             fscal            = felec;
 495
 496             /* Calculate temporary vectorial force */
 497             tx               = _mm_mul_ps(fscal,dx13);
 498             ty               = _mm_mul_ps(fscal,dy13);
 499             tz               = _mm_mul_ps(fscal,dz13);
 500
 501             /* Update vectorial force */
 502             fix1             = _mm_add_ps(fix1,tx);
 503             fiy1             = _mm_add_ps(fiy1,ty);
 504             fiz1             = _mm_add_ps(fiz1,tz);
 505
 506             fjx3             = _mm_add_ps(fjx3,tx);
 507             fjy3             = _mm_add_ps(fjy3,ty);
 508             fjz3             = _mm_add_ps(fjz3,tz);
 509
 510             /**************************
 511              * CALCULATE INTERACTIONS *
 512              **************************/
 513
 514             r21              = _mm_mul_ps(rsq21,rinv21);
 515
 516             /* EWALD ELECTROSTATICS */
 517
 518             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 519             ewrt             = _mm_mul_ps(r21,ewtabscale);
 520             ewitab           = _mm_cvttps_epi32(ewrt);
 521             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 522             ewitab           = _mm_slli_epi32(ewitab,2);
 523             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 524             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 525             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 526             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 527             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 528             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 529             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 530             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
 531             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
 532
 533             /* Update potential sum for this i atom from the interaction with this j atom. */
 534             velecsum         = _mm_add_ps(velecsum,velec);
 535
 536             fscal            = felec;
 537
 538             /* Calculate temporary vectorial force */
 539             tx               = _mm_mul_ps(fscal,dx21);
 540             ty               = _mm_mul_ps(fscal,dy21);
 541             tz               = _mm_mul_ps(fscal,dz21);
 542
 543             /* Update vectorial force */
 544             fix2             = _mm_add_ps(fix2,tx);
 545             fiy2             = _mm_add_ps(fiy2,ty);
 546             fiz2             = _mm_add_ps(fiz2,tz);
 547
 548             fjx1             = _mm_add_ps(fjx1,tx);
 549             fjy1             = _mm_add_ps(fjy1,ty);
 550             fjz1             = _mm_add_ps(fjz1,tz);
 551
 552             /**************************
 553              * CALCULATE INTERACTIONS *
 554              **************************/
 555
 556             r22              = _mm_mul_ps(rsq22,rinv22);
 557
 558             /* EWALD ELECTROSTATICS */
 559
 560             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 561             ewrt             = _mm_mul_ps(r22,ewtabscale);
 562             ewitab           = _mm_cvttps_epi32(ewrt);
 563             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 564             ewitab           = _mm_slli_epi32(ewitab,2);
 565             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 566             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 567             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 568             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 569             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 570             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 571             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 572             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
 573             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
 574
 575             /* Update potential sum for this i atom from the interaction with this j atom. */
 576             velecsum         = _mm_add_ps(velecsum,velec);
 577
 578             fscal            = felec;
 579
 580             /* Calculate temporary vectorial force */
 581             tx               = _mm_mul_ps(fscal,dx22);
 582             ty               = _mm_mul_ps(fscal,dy22);
 583             tz               = _mm_mul_ps(fscal,dz22);
 584
 585             /* Update vectorial force */
 586             fix2             = _mm_add_ps(fix2,tx);
 587             fiy2             = _mm_add_ps(fiy2,ty);
 588             fiz2             = _mm_add_ps(fiz2,tz);
 589
 590             fjx2             = _mm_add_ps(fjx2,tx);
 591             fjy2             = _mm_add_ps(fjy2,ty);
 592             fjz2             = _mm_add_ps(fjz2,tz);
 593
 594             /**************************
 595              * CALCULATE INTERACTIONS *
 596              **************************/
 597
 598             r23              = _mm_mul_ps(rsq23,rinv23);
 599
 600             /* EWALD ELECTROSTATICS */
 601
 602             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 603             ewrt             = _mm_mul_ps(r23,ewtabscale);
 604             ewitab           = _mm_cvttps_epi32(ewrt);
 605             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 606             ewitab           = _mm_slli_epi32(ewitab,2);
 607             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 608             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 609             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 610             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 611             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 612             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 613             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 614             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
 615             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
 616
 617             /* Update potential sum for this i atom from the interaction with this j atom. */
 618             velecsum         = _mm_add_ps(velecsum,velec);
 619
 620             fscal            = felec;
 621
 622             /* Calculate temporary vectorial force */
 623             tx               = _mm_mul_ps(fscal,dx23);
 624             ty               = _mm_mul_ps(fscal,dy23);
 625             tz               = _mm_mul_ps(fscal,dz23);
 626
 627             /* Update vectorial force */
 628             fix2             = _mm_add_ps(fix2,tx);
 629             fiy2             = _mm_add_ps(fiy2,ty);
 630             fiz2             = _mm_add_ps(fiz2,tz);
 631
 632             fjx3             = _mm_add_ps(fjx3,tx);
 633             fjy3             = _mm_add_ps(fjy3,ty);
 634             fjz3             = _mm_add_ps(fjz3,tz);
 635
 636             /**************************
 637              * CALCULATE INTERACTIONS *
 638              **************************/
 639
 640             r31              = _mm_mul_ps(rsq31,rinv31);
 641
 642             /* EWALD ELECTROSTATICS */
 643
 644             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 645             ewrt             = _mm_mul_ps(r31,ewtabscale);
 646             ewitab           = _mm_cvttps_epi32(ewrt);
 647             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 648             ewitab           = _mm_slli_epi32(ewitab,2);
 649             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 650             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 651             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 652             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 653             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 654             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 655             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 656             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
 657             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
 658
 659             /* Update potential sum for this i atom from the interaction with this j atom. */
 660             velecsum         = _mm_add_ps(velecsum,velec);
 661
 662             fscal            = felec;
 663
 664             /* Calculate temporary vectorial force */
 665             tx               = _mm_mul_ps(fscal,dx31);
 666             ty               = _mm_mul_ps(fscal,dy31);
 667             tz               = _mm_mul_ps(fscal,dz31);
 668
 669             /* Update vectorial force */
 670             fix3             = _mm_add_ps(fix3,tx);
 671             fiy3             = _mm_add_ps(fiy3,ty);
 672             fiz3             = _mm_add_ps(fiz3,tz);
 673
 674             fjx1             = _mm_add_ps(fjx1,tx);
 675             fjy1             = _mm_add_ps(fjy1,ty);
 676             fjz1             = _mm_add_ps(fjz1,tz);
 677
 678             /**************************
 679              * CALCULATE INTERACTIONS *
 680              **************************/
 681
 682             r32              = _mm_mul_ps(rsq32,rinv32);
 683
 684             /* EWALD ELECTROSTATICS */
 685
 686             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 687             ewrt             = _mm_mul_ps(r32,ewtabscale);
 688             ewitab           = _mm_cvttps_epi32(ewrt);
 689             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 690             ewitab           = _mm_slli_epi32(ewitab,2);
 691             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 692             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 693             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 694             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 695             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 696             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 697             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 698             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
 699             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
 700
 701             /* Update potential sum for this i atom from the interaction with this j atom. */
 702             velecsum         = _mm_add_ps(velecsum,velec);
 703
 704             fscal            = felec;
 705
 706             /* Calculate temporary vectorial force */
 707             tx               = _mm_mul_ps(fscal,dx32);
 708             ty               = _mm_mul_ps(fscal,dy32);
 709             tz               = _mm_mul_ps(fscal,dz32);
 710
 711             /* Update vectorial force */
 712             fix3             = _mm_add_ps(fix3,tx);
 713             fiy3             = _mm_add_ps(fiy3,ty);
 714             fiz3             = _mm_add_ps(fiz3,tz);
 715
 716             fjx2             = _mm_add_ps(fjx2,tx);
 717             fjy2             = _mm_add_ps(fjy2,ty);
 718             fjz2             = _mm_add_ps(fjz2,tz);
 719
 720             /**************************
 721              * CALCULATE INTERACTIONS *
 722              **************************/
 723
 724             r33              = _mm_mul_ps(rsq33,rinv33);
 725
 726             /* EWALD ELECTROSTATICS */
 727
 728             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 729             ewrt             = _mm_mul_ps(r33,ewtabscale);
 730             ewitab           = _mm_cvttps_epi32(ewrt);
 731             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 732             ewitab           = _mm_slli_epi32(ewitab,2);
 733             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 734             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 735             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 736             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 737             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 738             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 739             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 740             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
 741             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
 742
 743             /* Update potential sum for this i atom from the interaction with this j atom. */
 744             velecsum         = _mm_add_ps(velecsum,velec);
 745
 746             fscal            = felec;
 747
 748             /* Calculate temporary vectorial force */
 749             tx               = _mm_mul_ps(fscal,dx33);
 750             ty               = _mm_mul_ps(fscal,dy33);
 751             tz               = _mm_mul_ps(fscal,dz33);
 752
 753             /* Update vectorial force */
 754             fix3             = _mm_add_ps(fix3,tx);
 755             fiy3             = _mm_add_ps(fiy3,ty);
 756             fiz3             = _mm_add_ps(fiz3,tz);
 757
 758             fjx3             = _mm_add_ps(fjx3,tx);
 759             fjy3             = _mm_add_ps(fjy3,ty);
 760             fjz3             = _mm_add_ps(fjz3,tz);
 761
 762             fjptrA             = f+j_coord_offsetA;
 763             fjptrB             = f+j_coord_offsetB;
 764             fjptrC             = f+j_coord_offsetC;
 765             fjptrD             = f+j_coord_offsetD;
 766
 767             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 768                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 769                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 770
 771             /* Inner loop uses 428 flops */
 772         }
 773
 774         if(jidx<j_index_end)
 775         {
 776
 777             /* Get j neighbor index, and coordinate index */
 778             jnrlistA         = jjnr[jidx];
 779             jnrlistB         = jjnr[jidx+1];
 780             jnrlistC         = jjnr[jidx+2];
 781             jnrlistD         = jjnr[jidx+3];
 782             /* Sign of each element will be negative for non-real atoms.
 783              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 784              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 785              */
 786             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 787             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 788             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 789             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 790             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 791             j_coord_offsetA  = DIM*jnrA;
 792             j_coord_offsetB  = DIM*jnrB;
 793             j_coord_offsetC  = DIM*jnrC;
 794             j_coord_offsetD  = DIM*jnrD;
 795
 796             /* load j atom coordinates */
 797             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 798                                               x+j_coord_offsetC,x+j_coord_offsetD,
 799                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 800                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 801
 802             /* Calculate displacement vector */
 803             dx00             = _mm_sub_ps(ix0,jx0);
 804             dy00             = _mm_sub_ps(iy0,jy0);
 805             dz00             = _mm_sub_ps(iz0,jz0);
 806             dx11             = _mm_sub_ps(ix1,jx1);
 807             dy11             = _mm_sub_ps(iy1,jy1);
 808             dz11             = _mm_sub_ps(iz1,jz1);
 809             dx12             = _mm_sub_ps(ix1,jx2);
 810             dy12             = _mm_sub_ps(iy1,jy2);
 811             dz12             = _mm_sub_ps(iz1,jz2);
 812             dx13             = _mm_sub_ps(ix1,jx3);
 813             dy13             = _mm_sub_ps(iy1,jy3);
 814             dz13             = _mm_sub_ps(iz1,jz3);
 815             dx21             = _mm_sub_ps(ix2,jx1);
 816             dy21             = _mm_sub_ps(iy2,jy1);
 817             dz21             = _mm_sub_ps(iz2,jz1);
 818             dx22             = _mm_sub_ps(ix2,jx2);
 819             dy22             = _mm_sub_ps(iy2,jy2);
 820             dz22             = _mm_sub_ps(iz2,jz2);
 821             dx23             = _mm_sub_ps(ix2,jx3);
 822             dy23             = _mm_sub_ps(iy2,jy3);
 823             dz23             = _mm_sub_ps(iz2,jz3);
 824             dx31             = _mm_sub_ps(ix3,jx1);
 825             dy31             = _mm_sub_ps(iy3,jy1);
 826             dz31             = _mm_sub_ps(iz3,jz1);
 827             dx32             = _mm_sub_ps(ix3,jx2);
 828             dy32             = _mm_sub_ps(iy3,jy2);
 829             dz32             = _mm_sub_ps(iz3,jz2);
 830             dx33             = _mm_sub_ps(ix3,jx3);
 831             dy33             = _mm_sub_ps(iy3,jy3);
 832             dz33             = _mm_sub_ps(iz3,jz3);
 833
 834             /* Calculate squared distance and things based on it */
 835             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 836             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 837             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 838             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 839             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 840             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 841             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 842             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 843             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 844             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 845
 846             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 847             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 848             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 849             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 850             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 851             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 852             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 853             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 854             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 855             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 856
 857             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 858             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 859             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 860             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 861             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 862             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 863             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 864             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 865             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 866
 867             fjx0             = _mm_setzero_ps();
 868             fjy0             = _mm_setzero_ps();
 869             fjz0             = _mm_setzero_ps();
 870             fjx1             = _mm_setzero_ps();
 871             fjy1             = _mm_setzero_ps();
 872             fjz1             = _mm_setzero_ps();
 873             fjx2             = _mm_setzero_ps();
 874             fjy2             = _mm_setzero_ps();
 875             fjz2             = _mm_setzero_ps();
 876             fjx3             = _mm_setzero_ps();
 877             fjy3             = _mm_setzero_ps();
 878             fjz3             = _mm_setzero_ps();
 879
 880             /**************************
 881              * CALCULATE INTERACTIONS *
 882              **************************/
 883
 884             r00              = _mm_mul_ps(rsq00,rinv00);
 885             r00              = _mm_andnot_ps(dummy_mask,r00);
 886
 887             /* Calculate table index by multiplying r with table scale and truncate to integer */
 888             rt               = _mm_mul_ps(r00,vftabscale);
 889             vfitab           = _mm_cvttps_epi32(rt);
 890             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 891             vfitab           = _mm_slli_epi32(vfitab,3);
 892
 893             /* CUBIC SPLINE TABLE DISPERSION */
 894             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 895             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 896             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 897             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 898             _MM_TRANSPOSE4_PS(Y,F,G,H);
 899             Heps             = _mm_mul_ps(vfeps,H);
 900             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 901             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 902             vvdw6            = _mm_mul_ps(c6_00,VV);
 903             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 904             fvdw6            = _mm_mul_ps(c6_00,FF);
 905
 906             /* CUBIC SPLINE TABLE REPULSION */
 907             vfitab           = _mm_add_epi32(vfitab,ifour);
 908             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 909             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 910             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 911             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 912             _MM_TRANSPOSE4_PS(Y,F,G,H);
 913             Heps             = _mm_mul_ps(vfeps,H);
 914             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 915             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 916             vvdw12           = _mm_mul_ps(c12_00,VV);
 917             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 918             fvdw12           = _mm_mul_ps(c12_00,FF);
 919             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 920             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 921
 922             /* Update potential sum for this i atom from the interaction with this j atom. */
 923             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 924             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 925
 926             fscal            = fvdw;
 927
 928             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 929
 930             /* Calculate temporary vectorial force */
 931             tx               = _mm_mul_ps(fscal,dx00);
 932             ty               = _mm_mul_ps(fscal,dy00);
 933             tz               = _mm_mul_ps(fscal,dz00);
 934
 935             /* Update vectorial force */
 936             fix0             = _mm_add_ps(fix0,tx);
 937             fiy0             = _mm_add_ps(fiy0,ty);
 938             fiz0             = _mm_add_ps(fiz0,tz);
 939
 940             fjx0             = _mm_add_ps(fjx0,tx);
 941             fjy0             = _mm_add_ps(fjy0,ty);
 942             fjz0             = _mm_add_ps(fjz0,tz);
 943
 944             /**************************
 945              * CALCULATE INTERACTIONS *
 946              **************************/
 947
 948             r11              = _mm_mul_ps(rsq11,rinv11);
 949             r11              = _mm_andnot_ps(dummy_mask,r11);
 950
 951             /* EWALD ELECTROSTATICS */
 952
 953             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 954             ewrt             = _mm_mul_ps(r11,ewtabscale);
 955             ewitab           = _mm_cvttps_epi32(ewrt);
 956             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
 957             ewitab           = _mm_slli_epi32(ewitab,2);
 958             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 959             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 960             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 961             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 962             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 963             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 964             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 965             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 966             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 967
 968             /* Update potential sum for this i atom from the interaction with this j atom. */
 969             velec            = _mm_andnot_ps(dummy_mask,velec);
 970             velecsum         = _mm_add_ps(velecsum,velec);
 971
 972             fscal            = felec;
 973
 974             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 975
 976             /* Calculate temporary vectorial force */
 977             tx               = _mm_mul_ps(fscal,dx11);
 978             ty               = _mm_mul_ps(fscal,dy11);
 979             tz               = _mm_mul_ps(fscal,dz11);
 980
 981             /* Update vectorial force */
 982             fix1             = _mm_add_ps(fix1,tx);
 983             fiy1             = _mm_add_ps(fiy1,ty);
 984             fiz1             = _mm_add_ps(fiz1,tz);
 985
 986             fjx1             = _mm_add_ps(fjx1,tx);
 987             fjy1             = _mm_add_ps(fjy1,ty);
 988             fjz1             = _mm_add_ps(fjz1,tz);
 989
 990             /**************************
 991              * CALCULATE INTERACTIONS *
 992              **************************/
 993
 994             r12              = _mm_mul_ps(rsq12,rinv12);
 995             r12              = _mm_andnot_ps(dummy_mask,r12);
 996
 997             /* EWALD ELECTROSTATICS */
 998
 999             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1000             ewrt             = _mm_mul_ps(r12,ewtabscale);
1001             ewitab           = _mm_cvttps_epi32(ewrt);
1002             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1003             ewitab           = _mm_slli_epi32(ewitab,2);
1004             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1005             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1006             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1007             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1008             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1009             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1010             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1011             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
1012             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1013
1014             /* Update potential sum for this i atom from the interaction with this j atom. */
1015             velec            = _mm_andnot_ps(dummy_mask,velec);
1016             velecsum         = _mm_add_ps(velecsum,velec);
1017
1018             fscal            = felec;
1019
1020             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1021
1022             /* Calculate temporary vectorial force */
1023             tx               = _mm_mul_ps(fscal,dx12);
1024             ty               = _mm_mul_ps(fscal,dy12);
1025             tz               = _mm_mul_ps(fscal,dz12);
1026
1027             /* Update vectorial force */
1028             fix1             = _mm_add_ps(fix1,tx);
1029             fiy1             = _mm_add_ps(fiy1,ty);
1030             fiz1             = _mm_add_ps(fiz1,tz);
1031
1032             fjx2             = _mm_add_ps(fjx2,tx);
1033             fjy2             = _mm_add_ps(fjy2,ty);
1034             fjz2             = _mm_add_ps(fjz2,tz);
1035
1036             /**************************
1037              * CALCULATE INTERACTIONS *
1038              **************************/
1039
1040             r13              = _mm_mul_ps(rsq13,rinv13);
1041             r13              = _mm_andnot_ps(dummy_mask,r13);
1042
1043             /* EWALD ELECTROSTATICS */
1044
1045             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1046             ewrt             = _mm_mul_ps(r13,ewtabscale);
1047             ewitab           = _mm_cvttps_epi32(ewrt);
1048             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1049             ewitab           = _mm_slli_epi32(ewitab,2);
1050             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1051             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1052             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1053             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1054             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1055             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1056             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1057             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
1058             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1059
1060             /* Update potential sum for this i atom from the interaction with this j atom. */
1061             velec            = _mm_andnot_ps(dummy_mask,velec);
1062             velecsum         = _mm_add_ps(velecsum,velec);
1063
1064             fscal            = felec;
1065
1066             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1067
1068             /* Calculate temporary vectorial force */
1069             tx               = _mm_mul_ps(fscal,dx13);
1070             ty               = _mm_mul_ps(fscal,dy13);
1071             tz               = _mm_mul_ps(fscal,dz13);
1072
1073             /* Update vectorial force */
1074             fix1             = _mm_add_ps(fix1,tx);
1075             fiy1             = _mm_add_ps(fiy1,ty);
1076             fiz1             = _mm_add_ps(fiz1,tz);
1077
1078             fjx3             = _mm_add_ps(fjx3,tx);
1079             fjy3             = _mm_add_ps(fjy3,ty);
1080             fjz3             = _mm_add_ps(fjz3,tz);
1081
1082             /**************************
1083              * CALCULATE INTERACTIONS *
1084              **************************/
1085
1086             r21              = _mm_mul_ps(rsq21,rinv21);
1087             r21              = _mm_andnot_ps(dummy_mask,r21);
1088
1089             /* EWALD ELECTROSTATICS */
1090
1091             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1092             ewrt             = _mm_mul_ps(r21,ewtabscale);
1093             ewitab           = _mm_cvttps_epi32(ewrt);
1094             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1095             ewitab           = _mm_slli_epi32(ewitab,2);
1096             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1097             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1098             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1099             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1100             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1101             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1102             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1103             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
1104             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1105
1106             /* Update potential sum for this i atom from the interaction with this j atom. */
1107             velec            = _mm_andnot_ps(dummy_mask,velec);
1108             velecsum         = _mm_add_ps(velecsum,velec);
1109
1110             fscal            = felec;
1111
1112             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1113
1114             /* Calculate temporary vectorial force */
1115             tx               = _mm_mul_ps(fscal,dx21);
1116             ty               = _mm_mul_ps(fscal,dy21);
1117             tz               = _mm_mul_ps(fscal,dz21);
1118
1119             /* Update vectorial force */
1120             fix2             = _mm_add_ps(fix2,tx);
1121             fiy2             = _mm_add_ps(fiy2,ty);
1122             fiz2             = _mm_add_ps(fiz2,tz);
1123
1124             fjx1             = _mm_add_ps(fjx1,tx);
1125             fjy1             = _mm_add_ps(fjy1,ty);
1126             fjz1             = _mm_add_ps(fjz1,tz);
1127
1128             /**************************
1129              * CALCULATE INTERACTIONS *
1130              **************************/
1131
1132             r22              = _mm_mul_ps(rsq22,rinv22);
1133             r22              = _mm_andnot_ps(dummy_mask,r22);
1134
1135             /* EWALD ELECTROSTATICS */
1136
1137             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1138             ewrt             = _mm_mul_ps(r22,ewtabscale);
1139             ewitab           = _mm_cvttps_epi32(ewrt);
1140             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1141             ewitab           = _mm_slli_epi32(ewitab,2);
1142             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1143             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1144             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1145             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1146             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1147             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1148             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1149             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
1150             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1151
1152             /* Update potential sum for this i atom from the interaction with this j atom. */
1153             velec            = _mm_andnot_ps(dummy_mask,velec);
1154             velecsum         = _mm_add_ps(velecsum,velec);
1155
1156             fscal            = felec;
1157
1158             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1159
1160             /* Calculate temporary vectorial force */
1161             tx               = _mm_mul_ps(fscal,dx22);
1162             ty               = _mm_mul_ps(fscal,dy22);
1163             tz               = _mm_mul_ps(fscal,dz22);
1164
1165             /* Update vectorial force */
1166             fix2             = _mm_add_ps(fix2,tx);
1167             fiy2             = _mm_add_ps(fiy2,ty);
1168             fiz2             = _mm_add_ps(fiz2,tz);
1169
1170             fjx2             = _mm_add_ps(fjx2,tx);
1171             fjy2             = _mm_add_ps(fjy2,ty);
1172             fjz2             = _mm_add_ps(fjz2,tz);
1173
1174             /**************************
1175              * CALCULATE INTERACTIONS *
1176              **************************/
1177
1178             r23              = _mm_mul_ps(rsq23,rinv23);
1179             r23              = _mm_andnot_ps(dummy_mask,r23);
1180
1181             /* EWALD ELECTROSTATICS */
1182
1183             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1184             ewrt             = _mm_mul_ps(r23,ewtabscale);
1185             ewitab           = _mm_cvttps_epi32(ewrt);
1186             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1187             ewitab           = _mm_slli_epi32(ewitab,2);
1188             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1189             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1190             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1191             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1192             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1193             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1194             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1195             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
1196             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1197
1198             /* Update potential sum for this i atom from the interaction with this j atom. */
1199             velec            = _mm_andnot_ps(dummy_mask,velec);
1200             velecsum         = _mm_add_ps(velecsum,velec);
1201
1202             fscal            = felec;
1203
1204             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1205
1206             /* Calculate temporary vectorial force */
1207             tx               = _mm_mul_ps(fscal,dx23);
1208             ty               = _mm_mul_ps(fscal,dy23);
1209             tz               = _mm_mul_ps(fscal,dz23);
1210
1211             /* Update vectorial force */
1212             fix2             = _mm_add_ps(fix2,tx);
1213             fiy2             = _mm_add_ps(fiy2,ty);
1214             fiz2             = _mm_add_ps(fiz2,tz);
1215
1216             fjx3             = _mm_add_ps(fjx3,tx);
1217             fjy3             = _mm_add_ps(fjy3,ty);
1218             fjz3             = _mm_add_ps(fjz3,tz);
1219
1220             /**************************
1221              * CALCULATE INTERACTIONS *
1222              **************************/
1223
1224             r31              = _mm_mul_ps(rsq31,rinv31);
1225             r31              = _mm_andnot_ps(dummy_mask,r31);
1226
1227             /* EWALD ELECTROSTATICS */
1228
1229             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1230             ewrt             = _mm_mul_ps(r31,ewtabscale);
1231             ewitab           = _mm_cvttps_epi32(ewrt);
1232             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1233             ewitab           = _mm_slli_epi32(ewitab,2);
1234             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1235             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1236             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1237             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1238             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1239             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1240             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1241             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
1242             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1243
1244             /* Update potential sum for this i atom from the interaction with this j atom. */
1245             velec            = _mm_andnot_ps(dummy_mask,velec);
1246             velecsum         = _mm_add_ps(velecsum,velec);
1247
1248             fscal            = felec;
1249
1250             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1251
1252             /* Calculate temporary vectorial force */
1253             tx               = _mm_mul_ps(fscal,dx31);
1254             ty               = _mm_mul_ps(fscal,dy31);
1255             tz               = _mm_mul_ps(fscal,dz31);
1256
1257             /* Update vectorial force */
1258             fix3             = _mm_add_ps(fix3,tx);
1259             fiy3             = _mm_add_ps(fiy3,ty);
1260             fiz3             = _mm_add_ps(fiz3,tz);
1261
1262             fjx1             = _mm_add_ps(fjx1,tx);
1263             fjy1             = _mm_add_ps(fjy1,ty);
1264             fjz1             = _mm_add_ps(fjz1,tz);
1265
1266             /**************************
1267              * CALCULATE INTERACTIONS *
1268              **************************/
1269
1270             r32              = _mm_mul_ps(rsq32,rinv32);
1271             r32              = _mm_andnot_ps(dummy_mask,r32);
1272
1273             /* EWALD ELECTROSTATICS */
1274
1275             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1276             ewrt             = _mm_mul_ps(r32,ewtabscale);
1277             ewitab           = _mm_cvttps_epi32(ewrt);
1278             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1279             ewitab           = _mm_slli_epi32(ewitab,2);
1280             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1281             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1282             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1283             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1284             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1285             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1286             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1287             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
1288             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1289
1290             /* Update potential sum for this i atom from the interaction with this j atom. */
1291             velec            = _mm_andnot_ps(dummy_mask,velec);
1292             velecsum         = _mm_add_ps(velecsum,velec);
1293
1294             fscal            = felec;
1295
1296             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1297
1298             /* Calculate temporary vectorial force */
1299             tx               = _mm_mul_ps(fscal,dx32);
1300             ty               = _mm_mul_ps(fscal,dy32);
1301             tz               = _mm_mul_ps(fscal,dz32);
1302
1303             /* Update vectorial force */
1304             fix3             = _mm_add_ps(fix3,tx);
1305             fiy3             = _mm_add_ps(fiy3,ty);
1306             fiz3             = _mm_add_ps(fiz3,tz);
1307
1308             fjx2             = _mm_add_ps(fjx2,tx);
1309             fjy2             = _mm_add_ps(fjy2,ty);
1310             fjz2             = _mm_add_ps(fjz2,tz);
1311
1312             /**************************
1313              * CALCULATE INTERACTIONS *
1314              **************************/
1315
1316             r33              = _mm_mul_ps(rsq33,rinv33);
1317             r33              = _mm_andnot_ps(dummy_mask,r33);
1318
1319             /* EWALD ELECTROSTATICS */
1320
1321             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1322             ewrt             = _mm_mul_ps(r33,ewtabscale);
1323             ewitab           = _mm_cvttps_epi32(ewrt);
1324             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1325             ewitab           = _mm_slli_epi32(ewitab,2);
1326             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1327             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1328             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1329             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1330             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1331             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1332             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1333             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
1334             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1335
1336             /* Update potential sum for this i atom from the interaction with this j atom. */
1337             velec            = _mm_andnot_ps(dummy_mask,velec);
1338             velecsum         = _mm_add_ps(velecsum,velec);
1339
1340             fscal            = felec;
1341
1342             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1343
1344             /* Calculate temporary vectorial force */
1345             tx               = _mm_mul_ps(fscal,dx33);
1346             ty               = _mm_mul_ps(fscal,dy33);
1347             tz               = _mm_mul_ps(fscal,dz33);
1348
1349             /* Update vectorial force */
1350             fix3             = _mm_add_ps(fix3,tx);
1351             fiy3             = _mm_add_ps(fiy3,ty);
1352             fiz3             = _mm_add_ps(fiz3,tz);
1353
1354             fjx3             = _mm_add_ps(fjx3,tx);
1355             fjy3             = _mm_add_ps(fjy3,ty);
1356             fjz3             = _mm_add_ps(fjz3,tz);
1357
1358             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1359             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1360             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1361             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1362
1363             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1364                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1365                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1366
1367             /* Inner loop uses 438 flops */
1368         }
1369
1370         /* End of innermost loop */
1371
1372         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1373                                               f+i_coord_offset,fshift+i_shift_offset);
1374
1375         ggid                        = gid[iidx];
1376         /* Update potential energies */
1377         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1378         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1379
1380         /* Increment number of inner iterations */
1381         inneriter                  += j_index_end - j_index_start;
1382
1383         /* Outer loop uses 26 flops */
1384     }
1385
1386     /* Increment number of outer iterations */
1387     outeriter        += nri;
1388
1389     /* Update outer/inner flops */
1390
1391     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*438);
1392 }
1393 /*
1394  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse4_1_single
1395  * Electrostatics interaction: Ewald
1396  * VdW interaction:            CubicSplineTable
1397  * Geometry:                   Water4-Water4
1398  * Calculate force/pot:        Force
1399  */
1400 void
1401 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse4_1_single
1402                     (t_nblist                    * gmx_restrict       nlist,
1403                      rvec                        * gmx_restrict          xx,
1404                      rvec                        * gmx_restrict          ff,
1405                      t_forcerec                  * gmx_restrict          fr,
1406                      t_mdatoms                   * gmx_restrict     mdatoms,
1407                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1408                      t_nrnb                      * gmx_restrict        nrnb)
1409 {
1410     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1411      * just 0 for non-waters.
1412      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1413      * jnr indices corresponding to data put in the four positions in the SIMD register.
1414      */
1415     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1416     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1417     int              jnrA,jnrB,jnrC,jnrD;
1418     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1419     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1420     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1421     real             rcutoff_scalar;
1422     real             *shiftvec,*fshift,*x,*f;
1423     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1424     real             scratch[4*DIM];
1425     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1426     int              vdwioffset0;
1427     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1428     int              vdwioffset1;
1429     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1430     int              vdwioffset2;
1431     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1432     int              vdwioffset3;
1433     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1434     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1435     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1436     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1437     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1438     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1439     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1440     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1441     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1442     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1443     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1444     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1445     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1446     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1447     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1448     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1449     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1450     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1451     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1452     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1453     real             *charge;
1454     int              nvdwtype;
1455     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1456     int              *vdwtype;
1457     real             *vdwparam;
1458     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1459     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1460     __m128i          vfitab;
1461     __m128i          ifour       = _mm_set1_epi32(4);
1462     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1463     real             *vftab;
1464     __m128i          ewitab;
1465     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1466     real             *ewtab;
1467     __m128           dummy_mask,cutoff_mask;
1468     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1469     __m128           one     = _mm_set1_ps(1.0);
1470     __m128           two     = _mm_set1_ps(2.0);
1471     x                = xx[0];
1472     f                = ff[0];
1473
1474     nri              = nlist->nri;
1475     iinr             = nlist->iinr;
1476     jindex           = nlist->jindex;
1477     jjnr             = nlist->jjnr;
1478     shiftidx         = nlist->shift;
1479     gid              = nlist->gid;
1480     shiftvec         = fr->shift_vec[0];
1481     fshift           = fr->fshift[0];
1482     facel            = _mm_set1_ps(fr->epsfac);
1483     charge           = mdatoms->chargeA;
1484     nvdwtype         = fr->ntype;
1485     vdwparam         = fr->nbfp;
1486     vdwtype          = mdatoms->typeA;
1487
1488     vftab            = kernel_data->table_vdw->data;
1489     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
1490
1491     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
1492     ewtab            = fr->ic->tabq_coul_F;
1493     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
1494     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1495
1496     /* Setup water-specific parameters */
1497     inr              = nlist->iinr[0];
1498     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1499     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1500     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1501     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1502
1503     jq1              = _mm_set1_ps(charge[inr+1]);
1504     jq2              = _mm_set1_ps(charge[inr+2]);
1505     jq3              = _mm_set1_ps(charge[inr+3]);
1506     vdwjidx0A        = 2*vdwtype[inr+0];
1507     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1508     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1509     qq11             = _mm_mul_ps(iq1,jq1);
1510     qq12             = _mm_mul_ps(iq1,jq2);
1511     qq13             = _mm_mul_ps(iq1,jq3);
1512     qq21             = _mm_mul_ps(iq2,jq1);
1513     qq22             = _mm_mul_ps(iq2,jq2);
1514     qq23             = _mm_mul_ps(iq2,jq3);
1515     qq31             = _mm_mul_ps(iq3,jq1);
1516     qq32             = _mm_mul_ps(iq3,jq2);
1517     qq33             = _mm_mul_ps(iq3,jq3);
1518
1519     /* Avoid stupid compiler warnings */
1520     jnrA = jnrB = jnrC = jnrD = 0;
1521     j_coord_offsetA = 0;
1522     j_coord_offsetB = 0;
1523     j_coord_offsetC = 0;
1524     j_coord_offsetD = 0;
1525
1526     outeriter        = 0;
1527     inneriter        = 0;
1528
1529     for(iidx=0;iidx<4*DIM;iidx++)
1530     {
1531         scratch[iidx] = 0.0;
1532     }
1533
1534     /* Start outer loop over neighborlists */
1535     for(iidx=0; iidx<nri; iidx++)
1536     {
1537         /* Load shift vector for this list */
1538         i_shift_offset   = DIM*shiftidx[iidx];
1539
1540         /* Load limits for loop over neighbors */
1541         j_index_start    = jindex[iidx];
1542         j_index_end      = jindex[iidx+1];
1543
1544         /* Get outer coordinate index */
1545         inr              = iinr[iidx];
1546         i_coord_offset   = DIM*inr;
1547
1548         /* Load i particle coords and add shift vector */
1549         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1550                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1551
1552         fix0             = _mm_setzero_ps();
1553         fiy0             = _mm_setzero_ps();
1554         fiz0             = _mm_setzero_ps();
1555         fix1             = _mm_setzero_ps();
1556         fiy1             = _mm_setzero_ps();
1557         fiz1             = _mm_setzero_ps();
1558         fix2             = _mm_setzero_ps();
1559         fiy2             = _mm_setzero_ps();
1560         fiz2             = _mm_setzero_ps();
1561         fix3             = _mm_setzero_ps();
1562         fiy3             = _mm_setzero_ps();
1563         fiz3             = _mm_setzero_ps();
1564
1565         /* Start inner kernel loop */
1566         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1567         {
1568
1569             /* Get j neighbor index, and coordinate index */
1570             jnrA             = jjnr[jidx];
1571             jnrB             = jjnr[jidx+1];
1572             jnrC             = jjnr[jidx+2];
1573             jnrD             = jjnr[jidx+3];
1574             j_coord_offsetA  = DIM*jnrA;
1575             j_coord_offsetB  = DIM*jnrB;
1576             j_coord_offsetC  = DIM*jnrC;
1577             j_coord_offsetD  = DIM*jnrD;
1578
1579             /* load j atom coordinates */
1580             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1581                                               x+j_coord_offsetC,x+j_coord_offsetD,
1582                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1583                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1584
1585             /* Calculate displacement vector */
1586             dx00             = _mm_sub_ps(ix0,jx0);
1587             dy00             = _mm_sub_ps(iy0,jy0);
1588             dz00             = _mm_sub_ps(iz0,jz0);
1589             dx11             = _mm_sub_ps(ix1,jx1);
1590             dy11             = _mm_sub_ps(iy1,jy1);
1591             dz11             = _mm_sub_ps(iz1,jz1);
1592             dx12             = _mm_sub_ps(ix1,jx2);
1593             dy12             = _mm_sub_ps(iy1,jy2);
1594             dz12             = _mm_sub_ps(iz1,jz2);
1595             dx13             = _mm_sub_ps(ix1,jx3);
1596             dy13             = _mm_sub_ps(iy1,jy3);
1597             dz13             = _mm_sub_ps(iz1,jz3);
1598             dx21             = _mm_sub_ps(ix2,jx1);
1599             dy21             = _mm_sub_ps(iy2,jy1);
1600             dz21             = _mm_sub_ps(iz2,jz1);
1601             dx22             = _mm_sub_ps(ix2,jx2);
1602             dy22             = _mm_sub_ps(iy2,jy2);
1603             dz22             = _mm_sub_ps(iz2,jz2);
1604             dx23             = _mm_sub_ps(ix2,jx3);
1605             dy23             = _mm_sub_ps(iy2,jy3);
1606             dz23             = _mm_sub_ps(iz2,jz3);
1607             dx31             = _mm_sub_ps(ix3,jx1);
1608             dy31             = _mm_sub_ps(iy3,jy1);
1609             dz31             = _mm_sub_ps(iz3,jz1);
1610             dx32             = _mm_sub_ps(ix3,jx2);
1611             dy32             = _mm_sub_ps(iy3,jy2);
1612             dz32             = _mm_sub_ps(iz3,jz2);
1613             dx33             = _mm_sub_ps(ix3,jx3);
1614             dy33             = _mm_sub_ps(iy3,jy3);
1615             dz33             = _mm_sub_ps(iz3,jz3);
1616
1617             /* Calculate squared distance and things based on it */
1618             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1619             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1620             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1621             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1622             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1623             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1624             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1625             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1626             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1627             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1628
1629             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1630             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1631             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1632             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1633             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1634             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1635             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1636             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1637             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1638             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1639
1640             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1641             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1642             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1643             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1644             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1645             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1646             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1647             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1648             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1649
1650             fjx0             = _mm_setzero_ps();
1651             fjy0             = _mm_setzero_ps();
1652             fjz0             = _mm_setzero_ps();
1653             fjx1             = _mm_setzero_ps();
1654             fjy1             = _mm_setzero_ps();
1655             fjz1             = _mm_setzero_ps();
1656             fjx2             = _mm_setzero_ps();
1657             fjy2             = _mm_setzero_ps();
1658             fjz2             = _mm_setzero_ps();
1659             fjx3             = _mm_setzero_ps();
1660             fjy3             = _mm_setzero_ps();
1661             fjz3             = _mm_setzero_ps();
1662
1663             /**************************
1664              * CALCULATE INTERACTIONS *
1665              **************************/
1666
1667             r00              = _mm_mul_ps(rsq00,rinv00);
1668
1669             /* Calculate table index by multiplying r with table scale and truncate to integer */
1670             rt               = _mm_mul_ps(r00,vftabscale);
1671             vfitab           = _mm_cvttps_epi32(rt);
1672             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1673             vfitab           = _mm_slli_epi32(vfitab,3);
1674
1675             /* CUBIC SPLINE TABLE DISPERSION */
1676             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1677             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1678             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1679             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1680             _MM_TRANSPOSE4_PS(Y,F,G,H);
1681             Heps             = _mm_mul_ps(vfeps,H);
1682             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1683             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1684             fvdw6            = _mm_mul_ps(c6_00,FF);
1685
1686             /* CUBIC SPLINE TABLE REPULSION */
1687             vfitab           = _mm_add_epi32(vfitab,ifour);
1688             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1689             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1690             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1691             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1692             _MM_TRANSPOSE4_PS(Y,F,G,H);
1693             Heps             = _mm_mul_ps(vfeps,H);
1694             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1695             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1696             fvdw12           = _mm_mul_ps(c12_00,FF);
1697             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1698
1699             fscal            = fvdw;
1700
1701             /* Calculate temporary vectorial force */
1702             tx               = _mm_mul_ps(fscal,dx00);
1703             ty               = _mm_mul_ps(fscal,dy00);
1704             tz               = _mm_mul_ps(fscal,dz00);
1705
1706             /* Update vectorial force */
1707             fix0             = _mm_add_ps(fix0,tx);
1708             fiy0             = _mm_add_ps(fiy0,ty);
1709             fiz0             = _mm_add_ps(fiz0,tz);
1710
1711             fjx0             = _mm_add_ps(fjx0,tx);
1712             fjy0             = _mm_add_ps(fjy0,ty);
1713             fjz0             = _mm_add_ps(fjz0,tz);
1714
1715             /**************************
1716              * CALCULATE INTERACTIONS *
1717              **************************/
1718
1719             r11              = _mm_mul_ps(rsq11,rinv11);
1720
1721             /* EWALD ELECTROSTATICS */
1722
1723             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1724             ewrt             = _mm_mul_ps(r11,ewtabscale);
1725             ewitab           = _mm_cvttps_epi32(ewrt);
1726             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1727             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1728                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1729                                          &ewtabF,&ewtabFn);
1730             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1731             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1732
1733             fscal            = felec;
1734
1735             /* Calculate temporary vectorial force */
1736             tx               = _mm_mul_ps(fscal,dx11);
1737             ty               = _mm_mul_ps(fscal,dy11);
1738             tz               = _mm_mul_ps(fscal,dz11);
1739
1740             /* Update vectorial force */
1741             fix1             = _mm_add_ps(fix1,tx);
1742             fiy1             = _mm_add_ps(fiy1,ty);
1743             fiz1             = _mm_add_ps(fiz1,tz);
1744
1745             fjx1             = _mm_add_ps(fjx1,tx);
1746             fjy1             = _mm_add_ps(fjy1,ty);
1747             fjz1             = _mm_add_ps(fjz1,tz);
1748
1749             /**************************
1750              * CALCULATE INTERACTIONS *
1751              **************************/
1752
1753             r12              = _mm_mul_ps(rsq12,rinv12);
1754
1755             /* EWALD ELECTROSTATICS */
1756
1757             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1758             ewrt             = _mm_mul_ps(r12,ewtabscale);
1759             ewitab           = _mm_cvttps_epi32(ewrt);
1760             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1761             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1762                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1763                                          &ewtabF,&ewtabFn);
1764             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1765             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1766
1767             fscal            = felec;
1768
1769             /* Calculate temporary vectorial force */
1770             tx               = _mm_mul_ps(fscal,dx12);
1771             ty               = _mm_mul_ps(fscal,dy12);
1772             tz               = _mm_mul_ps(fscal,dz12);
1773
1774             /* Update vectorial force */
1775             fix1             = _mm_add_ps(fix1,tx);
1776             fiy1             = _mm_add_ps(fiy1,ty);
1777             fiz1             = _mm_add_ps(fiz1,tz);
1778
1779             fjx2             = _mm_add_ps(fjx2,tx);
1780             fjy2             = _mm_add_ps(fjy2,ty);
1781             fjz2             = _mm_add_ps(fjz2,tz);
1782
1783             /**************************
1784              * CALCULATE INTERACTIONS *
1785              **************************/
1786
1787             r13              = _mm_mul_ps(rsq13,rinv13);
1788
1789             /* EWALD ELECTROSTATICS */
1790
1791             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1792             ewrt             = _mm_mul_ps(r13,ewtabscale);
1793             ewitab           = _mm_cvttps_epi32(ewrt);
1794             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1795             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1796                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1797                                          &ewtabF,&ewtabFn);
1798             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1799             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1800
1801             fscal            = felec;
1802
1803             /* Calculate temporary vectorial force */
1804             tx               = _mm_mul_ps(fscal,dx13);
1805             ty               = _mm_mul_ps(fscal,dy13);
1806             tz               = _mm_mul_ps(fscal,dz13);
1807
1808             /* Update vectorial force */
1809             fix1             = _mm_add_ps(fix1,tx);
1810             fiy1             = _mm_add_ps(fiy1,ty);
1811             fiz1             = _mm_add_ps(fiz1,tz);
1812
1813             fjx3             = _mm_add_ps(fjx3,tx);
1814             fjy3             = _mm_add_ps(fjy3,ty);
1815             fjz3             = _mm_add_ps(fjz3,tz);
1816
1817             /**************************
1818              * CALCULATE INTERACTIONS *
1819              **************************/
1820
1821             r21              = _mm_mul_ps(rsq21,rinv21);
1822
1823             /* EWALD ELECTROSTATICS */
1824
1825             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1826             ewrt             = _mm_mul_ps(r21,ewtabscale);
1827             ewitab           = _mm_cvttps_epi32(ewrt);
1828             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1829             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1830                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1831                                          &ewtabF,&ewtabFn);
1832             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1833             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1834
1835             fscal            = felec;
1836
1837             /* Calculate temporary vectorial force */
1838             tx               = _mm_mul_ps(fscal,dx21);
1839             ty               = _mm_mul_ps(fscal,dy21);
1840             tz               = _mm_mul_ps(fscal,dz21);
1841
1842             /* Update vectorial force */
1843             fix2             = _mm_add_ps(fix2,tx);
1844             fiy2             = _mm_add_ps(fiy2,ty);
1845             fiz2             = _mm_add_ps(fiz2,tz);
1846
1847             fjx1             = _mm_add_ps(fjx1,tx);
1848             fjy1             = _mm_add_ps(fjy1,ty);
1849             fjz1             = _mm_add_ps(fjz1,tz);
1850
1851             /**************************
1852              * CALCULATE INTERACTIONS *
1853              **************************/
1854
1855             r22              = _mm_mul_ps(rsq22,rinv22);
1856
1857             /* EWALD ELECTROSTATICS */
1858
1859             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1860             ewrt             = _mm_mul_ps(r22,ewtabscale);
1861             ewitab           = _mm_cvttps_epi32(ewrt);
1862             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1863             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1864                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1865                                          &ewtabF,&ewtabFn);
1866             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1867             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1868
1869             fscal            = felec;
1870
1871             /* Calculate temporary vectorial force */
1872             tx               = _mm_mul_ps(fscal,dx22);
1873             ty               = _mm_mul_ps(fscal,dy22);
1874             tz               = _mm_mul_ps(fscal,dz22);
1875
1876             /* Update vectorial force */
1877             fix2             = _mm_add_ps(fix2,tx);
1878             fiy2             = _mm_add_ps(fiy2,ty);
1879             fiz2             = _mm_add_ps(fiz2,tz);
1880
1881             fjx2             = _mm_add_ps(fjx2,tx);
1882             fjy2             = _mm_add_ps(fjy2,ty);
1883             fjz2             = _mm_add_ps(fjz2,tz);
1884
1885             /**************************
1886              * CALCULATE INTERACTIONS *
1887              **************************/
1888
1889             r23              = _mm_mul_ps(rsq23,rinv23);
1890
1891             /* EWALD ELECTROSTATICS */
1892
1893             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1894             ewrt             = _mm_mul_ps(r23,ewtabscale);
1895             ewitab           = _mm_cvttps_epi32(ewrt);
1896             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1897             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1898                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1899                                          &ewtabF,&ewtabFn);
1900             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1901             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1902
1903             fscal            = felec;
1904
1905             /* Calculate temporary vectorial force */
1906             tx               = _mm_mul_ps(fscal,dx23);
1907             ty               = _mm_mul_ps(fscal,dy23);
1908             tz               = _mm_mul_ps(fscal,dz23);
1909
1910             /* Update vectorial force */
1911             fix2             = _mm_add_ps(fix2,tx);
1912             fiy2             = _mm_add_ps(fiy2,ty);
1913             fiz2             = _mm_add_ps(fiz2,tz);
1914
1915             fjx3             = _mm_add_ps(fjx3,tx);
1916             fjy3             = _mm_add_ps(fjy3,ty);
1917             fjz3             = _mm_add_ps(fjz3,tz);
1918
1919             /**************************
1920              * CALCULATE INTERACTIONS *
1921              **************************/
1922
1923             r31              = _mm_mul_ps(rsq31,rinv31);
1924
1925             /* EWALD ELECTROSTATICS */
1926
1927             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1928             ewrt             = _mm_mul_ps(r31,ewtabscale);
1929             ewitab           = _mm_cvttps_epi32(ewrt);
1930             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1931             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1932                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1933                                          &ewtabF,&ewtabFn);
1934             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1935             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1936
1937             fscal            = felec;
1938
1939             /* Calculate temporary vectorial force */
1940             tx               = _mm_mul_ps(fscal,dx31);
1941             ty               = _mm_mul_ps(fscal,dy31);
1942             tz               = _mm_mul_ps(fscal,dz31);
1943
1944             /* Update vectorial force */
1945             fix3             = _mm_add_ps(fix3,tx);
1946             fiy3             = _mm_add_ps(fiy3,ty);
1947             fiz3             = _mm_add_ps(fiz3,tz);
1948
1949             fjx1             = _mm_add_ps(fjx1,tx);
1950             fjy1             = _mm_add_ps(fjy1,ty);
1951             fjz1             = _mm_add_ps(fjz1,tz);
1952
1953             /**************************
1954              * CALCULATE INTERACTIONS *
1955              **************************/
1956
1957             r32              = _mm_mul_ps(rsq32,rinv32);
1958
1959             /* EWALD ELECTROSTATICS */
1960
1961             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1962             ewrt             = _mm_mul_ps(r32,ewtabscale);
1963             ewitab           = _mm_cvttps_epi32(ewrt);
1964             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1965             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
1966                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
1967                                          &ewtabF,&ewtabFn);
1968             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1969             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1970
1971             fscal            = felec;
1972
1973             /* Calculate temporary vectorial force */
1974             tx               = _mm_mul_ps(fscal,dx32);
1975             ty               = _mm_mul_ps(fscal,dy32);
1976             tz               = _mm_mul_ps(fscal,dz32);
1977
1978             /* Update vectorial force */
1979             fix3             = _mm_add_ps(fix3,tx);
1980             fiy3             = _mm_add_ps(fiy3,ty);
1981             fiz3             = _mm_add_ps(fiz3,tz);
1982
1983             fjx2             = _mm_add_ps(fjx2,tx);
1984             fjy2             = _mm_add_ps(fjy2,ty);
1985             fjz2             = _mm_add_ps(fjz2,tz);
1986
1987             /**************************
1988              * CALCULATE INTERACTIONS *
1989              **************************/
1990
1991             r33              = _mm_mul_ps(rsq33,rinv33);
1992
1993             /* EWALD ELECTROSTATICS */
1994
1995             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1996             ewrt             = _mm_mul_ps(r33,ewtabscale);
1997             ewitab           = _mm_cvttps_epi32(ewrt);
1998             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
1999             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2000                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2001                                          &ewtabF,&ewtabFn);
2002             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2003             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2004
2005             fscal            = felec;
2006
2007             /* Calculate temporary vectorial force */
2008             tx               = _mm_mul_ps(fscal,dx33);
2009             ty               = _mm_mul_ps(fscal,dy33);
2010             tz               = _mm_mul_ps(fscal,dz33);
2011
2012             /* Update vectorial force */
2013             fix3             = _mm_add_ps(fix3,tx);
2014             fiy3             = _mm_add_ps(fiy3,ty);
2015             fiz3             = _mm_add_ps(fiz3,tz);
2016
2017             fjx3             = _mm_add_ps(fjx3,tx);
2018             fjy3             = _mm_add_ps(fjy3,ty);
2019             fjz3             = _mm_add_ps(fjz3,tz);
2020
2021             fjptrA             = f+j_coord_offsetA;
2022             fjptrB             = f+j_coord_offsetB;
2023             fjptrC             = f+j_coord_offsetC;
2024             fjptrD             = f+j_coord_offsetD;
2025
2026             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2027                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2028                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2029
2030             /* Inner loop uses 375 flops */
2031         }
2032
2033         if(jidx<j_index_end)
2034         {
2035
2036             /* Get j neighbor index, and coordinate index */
2037             jnrlistA         = jjnr[jidx];
2038             jnrlistB         = jjnr[jidx+1];
2039             jnrlistC         = jjnr[jidx+2];
2040             jnrlistD         = jjnr[jidx+3];
2041             /* Sign of each element will be negative for non-real atoms.
2042              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2043              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2044              */
2045             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2046             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2047             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2048             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2049             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2050             j_coord_offsetA  = DIM*jnrA;
2051             j_coord_offsetB  = DIM*jnrB;
2052             j_coord_offsetC  = DIM*jnrC;
2053             j_coord_offsetD  = DIM*jnrD;
2054
2055             /* load j atom coordinates */
2056             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2057                                               x+j_coord_offsetC,x+j_coord_offsetD,
2058                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2059                                               &jy2,&jz2,&jx3,&jy3,&jz3);
2060
2061             /* Calculate displacement vector */
2062             dx00             = _mm_sub_ps(ix0,jx0);
2063             dy00             = _mm_sub_ps(iy0,jy0);
2064             dz00             = _mm_sub_ps(iz0,jz0);
2065             dx11             = _mm_sub_ps(ix1,jx1);
2066             dy11             = _mm_sub_ps(iy1,jy1);
2067             dz11             = _mm_sub_ps(iz1,jz1);
2068             dx12             = _mm_sub_ps(ix1,jx2);
2069             dy12             = _mm_sub_ps(iy1,jy2);
2070             dz12             = _mm_sub_ps(iz1,jz2);
2071             dx13             = _mm_sub_ps(ix1,jx3);
2072             dy13             = _mm_sub_ps(iy1,jy3);
2073             dz13             = _mm_sub_ps(iz1,jz3);
2074             dx21             = _mm_sub_ps(ix2,jx1);
2075             dy21             = _mm_sub_ps(iy2,jy1);
2076             dz21             = _mm_sub_ps(iz2,jz1);
2077             dx22             = _mm_sub_ps(ix2,jx2);
2078             dy22             = _mm_sub_ps(iy2,jy2);
2079             dz22             = _mm_sub_ps(iz2,jz2);
2080             dx23             = _mm_sub_ps(ix2,jx3);
2081             dy23             = _mm_sub_ps(iy2,jy3);
2082             dz23             = _mm_sub_ps(iz2,jz3);
2083             dx31             = _mm_sub_ps(ix3,jx1);
2084             dy31             = _mm_sub_ps(iy3,jy1);
2085             dz31             = _mm_sub_ps(iz3,jz1);
2086             dx32             = _mm_sub_ps(ix3,jx2);
2087             dy32             = _mm_sub_ps(iy3,jy2);
2088             dz32             = _mm_sub_ps(iz3,jz2);
2089             dx33             = _mm_sub_ps(ix3,jx3);
2090             dy33             = _mm_sub_ps(iy3,jy3);
2091             dz33             = _mm_sub_ps(iz3,jz3);
2092
2093             /* Calculate squared distance and things based on it */
2094             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2095             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2096             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2097             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2098             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2099             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2100             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2101             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2102             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2103             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2104
2105             rinv00           = gmx_mm_invsqrt_ps(rsq00);
2106             rinv11           = gmx_mm_invsqrt_ps(rsq11);
2107             rinv12           = gmx_mm_invsqrt_ps(rsq12);
2108             rinv13           = gmx_mm_invsqrt_ps(rsq13);
2109             rinv21           = gmx_mm_invsqrt_ps(rsq21);
2110             rinv22           = gmx_mm_invsqrt_ps(rsq22);
2111             rinv23           = gmx_mm_invsqrt_ps(rsq23);
2112             rinv31           = gmx_mm_invsqrt_ps(rsq31);
2113             rinv32           = gmx_mm_invsqrt_ps(rsq32);
2114             rinv33           = gmx_mm_invsqrt_ps(rsq33);
2115
2116             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
2117             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
2118             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
2119             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
2120             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
2121             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
2122             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
2123             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
2124             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
2125
2126             fjx0             = _mm_setzero_ps();
2127             fjy0             = _mm_setzero_ps();
2128             fjz0             = _mm_setzero_ps();
2129             fjx1             = _mm_setzero_ps();
2130             fjy1             = _mm_setzero_ps();
2131             fjz1             = _mm_setzero_ps();
2132             fjx2             = _mm_setzero_ps();
2133             fjy2             = _mm_setzero_ps();
2134             fjz2             = _mm_setzero_ps();
2135             fjx3             = _mm_setzero_ps();
2136             fjy3             = _mm_setzero_ps();
2137             fjz3             = _mm_setzero_ps();
2138
2139             /**************************
2140              * CALCULATE INTERACTIONS *
2141              **************************/
2142
2143             r00              = _mm_mul_ps(rsq00,rinv00);
2144             r00              = _mm_andnot_ps(dummy_mask,r00);
2145
2146             /* Calculate table index by multiplying r with table scale and truncate to integer */
2147             rt               = _mm_mul_ps(r00,vftabscale);
2148             vfitab           = _mm_cvttps_epi32(rt);
2149             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2150             vfitab           = _mm_slli_epi32(vfitab,3);
2151
2152             /* CUBIC SPLINE TABLE DISPERSION */
2153             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2154             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2155             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2156             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2157             _MM_TRANSPOSE4_PS(Y,F,G,H);
2158             Heps             = _mm_mul_ps(vfeps,H);
2159             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2160             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2161             fvdw6            = _mm_mul_ps(c6_00,FF);
2162
2163             /* CUBIC SPLINE TABLE REPULSION */
2164             vfitab           = _mm_add_epi32(vfitab,ifour);
2165             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2166             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2167             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2168             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2169             _MM_TRANSPOSE4_PS(Y,F,G,H);
2170             Heps             = _mm_mul_ps(vfeps,H);
2171             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2172             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2173             fvdw12           = _mm_mul_ps(c12_00,FF);
2174             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
2175
2176             fscal            = fvdw;
2177
2178             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2179
2180             /* Calculate temporary vectorial force */
2181             tx               = _mm_mul_ps(fscal,dx00);
2182             ty               = _mm_mul_ps(fscal,dy00);
2183             tz               = _mm_mul_ps(fscal,dz00);
2184
2185             /* Update vectorial force */
2186             fix0             = _mm_add_ps(fix0,tx);
2187             fiy0             = _mm_add_ps(fiy0,ty);
2188             fiz0             = _mm_add_ps(fiz0,tz);
2189
2190             fjx0             = _mm_add_ps(fjx0,tx);
2191             fjy0             = _mm_add_ps(fjy0,ty);
2192             fjz0             = _mm_add_ps(fjz0,tz);
2193
2194             /**************************
2195              * CALCULATE INTERACTIONS *
2196              **************************/
2197
2198             r11              = _mm_mul_ps(rsq11,rinv11);
2199             r11              = _mm_andnot_ps(dummy_mask,r11);
2200
2201             /* EWALD ELECTROSTATICS */
2202
2203             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2204             ewrt             = _mm_mul_ps(r11,ewtabscale);
2205             ewitab           = _mm_cvttps_epi32(ewrt);
2206             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2207             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2208                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2209                                          &ewtabF,&ewtabFn);
2210             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2211             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
2212
2213             fscal            = felec;
2214
2215             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2216
2217             /* Calculate temporary vectorial force */
2218             tx               = _mm_mul_ps(fscal,dx11);
2219             ty               = _mm_mul_ps(fscal,dy11);
2220             tz               = _mm_mul_ps(fscal,dz11);
2221
2222             /* Update vectorial force */
2223             fix1             = _mm_add_ps(fix1,tx);
2224             fiy1             = _mm_add_ps(fiy1,ty);
2225             fiz1             = _mm_add_ps(fiz1,tz);
2226
2227             fjx1             = _mm_add_ps(fjx1,tx);
2228             fjy1             = _mm_add_ps(fjy1,ty);
2229             fjz1             = _mm_add_ps(fjz1,tz);
2230
2231             /**************************
2232              * CALCULATE INTERACTIONS *
2233              **************************/
2234
2235             r12              = _mm_mul_ps(rsq12,rinv12);
2236             r12              = _mm_andnot_ps(dummy_mask,r12);
2237
2238             /* EWALD ELECTROSTATICS */
2239
2240             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2241             ewrt             = _mm_mul_ps(r12,ewtabscale);
2242             ewitab           = _mm_cvttps_epi32(ewrt);
2243             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2244             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2245                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2246                                          &ewtabF,&ewtabFn);
2247             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2248             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
2249
2250             fscal            = felec;
2251
2252             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2253
2254             /* Calculate temporary vectorial force */
2255             tx               = _mm_mul_ps(fscal,dx12);
2256             ty               = _mm_mul_ps(fscal,dy12);
2257             tz               = _mm_mul_ps(fscal,dz12);
2258
2259             /* Update vectorial force */
2260             fix1             = _mm_add_ps(fix1,tx);
2261             fiy1             = _mm_add_ps(fiy1,ty);
2262             fiz1             = _mm_add_ps(fiz1,tz);
2263
2264             fjx2             = _mm_add_ps(fjx2,tx);
2265             fjy2             = _mm_add_ps(fjy2,ty);
2266             fjz2             = _mm_add_ps(fjz2,tz);
2267
2268             /**************************
2269              * CALCULATE INTERACTIONS *
2270              **************************/
2271
2272             r13              = _mm_mul_ps(rsq13,rinv13);
2273             r13              = _mm_andnot_ps(dummy_mask,r13);
2274
2275             /* EWALD ELECTROSTATICS */
2276
2277             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2278             ewrt             = _mm_mul_ps(r13,ewtabscale);
2279             ewitab           = _mm_cvttps_epi32(ewrt);
2280             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2281             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2282                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2283                                          &ewtabF,&ewtabFn);
2284             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2285             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
2286
2287             fscal            = felec;
2288
2289             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2290
2291             /* Calculate temporary vectorial force */
2292             tx               = _mm_mul_ps(fscal,dx13);
2293             ty               = _mm_mul_ps(fscal,dy13);
2294             tz               = _mm_mul_ps(fscal,dz13);
2295
2296             /* Update vectorial force */
2297             fix1             = _mm_add_ps(fix1,tx);
2298             fiy1             = _mm_add_ps(fiy1,ty);
2299             fiz1             = _mm_add_ps(fiz1,tz);
2300
2301             fjx3             = _mm_add_ps(fjx3,tx);
2302             fjy3             = _mm_add_ps(fjy3,ty);
2303             fjz3             = _mm_add_ps(fjz3,tz);
2304
2305             /**************************
2306              * CALCULATE INTERACTIONS *
2307              **************************/
2308
2309             r21              = _mm_mul_ps(rsq21,rinv21);
2310             r21              = _mm_andnot_ps(dummy_mask,r21);
2311
2312             /* EWALD ELECTROSTATICS */
2313
2314             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2315             ewrt             = _mm_mul_ps(r21,ewtabscale);
2316             ewitab           = _mm_cvttps_epi32(ewrt);
2317             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2318             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2319                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2320                                          &ewtabF,&ewtabFn);
2321             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2322             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
2323
2324             fscal            = felec;
2325
2326             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2327
2328             /* Calculate temporary vectorial force */
2329             tx               = _mm_mul_ps(fscal,dx21);
2330             ty               = _mm_mul_ps(fscal,dy21);
2331             tz               = _mm_mul_ps(fscal,dz21);
2332
2333             /* Update vectorial force */
2334             fix2             = _mm_add_ps(fix2,tx);
2335             fiy2             = _mm_add_ps(fiy2,ty);
2336             fiz2             = _mm_add_ps(fiz2,tz);
2337
2338             fjx1             = _mm_add_ps(fjx1,tx);
2339             fjy1             = _mm_add_ps(fjy1,ty);
2340             fjz1             = _mm_add_ps(fjz1,tz);
2341
2342             /**************************
2343              * CALCULATE INTERACTIONS *
2344              **************************/
2345
2346             r22              = _mm_mul_ps(rsq22,rinv22);
2347             r22              = _mm_andnot_ps(dummy_mask,r22);
2348
2349             /* EWALD ELECTROSTATICS */
2350
2351             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2352             ewrt             = _mm_mul_ps(r22,ewtabscale);
2353             ewitab           = _mm_cvttps_epi32(ewrt);
2354             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2355             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2356                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2357                                          &ewtabF,&ewtabFn);
2358             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2359             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
2360
2361             fscal            = felec;
2362
2363             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2364
2365             /* Calculate temporary vectorial force */
2366             tx               = _mm_mul_ps(fscal,dx22);
2367             ty               = _mm_mul_ps(fscal,dy22);
2368             tz               = _mm_mul_ps(fscal,dz22);
2369
2370             /* Update vectorial force */
2371             fix2             = _mm_add_ps(fix2,tx);
2372             fiy2             = _mm_add_ps(fiy2,ty);
2373             fiz2             = _mm_add_ps(fiz2,tz);
2374
2375             fjx2             = _mm_add_ps(fjx2,tx);
2376             fjy2             = _mm_add_ps(fjy2,ty);
2377             fjz2             = _mm_add_ps(fjz2,tz);
2378
2379             /**************************
2380              * CALCULATE INTERACTIONS *
2381              **************************/
2382
2383             r23              = _mm_mul_ps(rsq23,rinv23);
2384             r23              = _mm_andnot_ps(dummy_mask,r23);
2385
2386             /* EWALD ELECTROSTATICS */
2387
2388             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2389             ewrt             = _mm_mul_ps(r23,ewtabscale);
2390             ewitab           = _mm_cvttps_epi32(ewrt);
2391             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2392             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2393                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2394                                          &ewtabF,&ewtabFn);
2395             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2396             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
2397
2398             fscal            = felec;
2399
2400             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2401
2402             /* Calculate temporary vectorial force */
2403             tx               = _mm_mul_ps(fscal,dx23);
2404             ty               = _mm_mul_ps(fscal,dy23);
2405             tz               = _mm_mul_ps(fscal,dz23);
2406
2407             /* Update vectorial force */
2408             fix2             = _mm_add_ps(fix2,tx);
2409             fiy2             = _mm_add_ps(fiy2,ty);
2410             fiz2             = _mm_add_ps(fiz2,tz);
2411
2412             fjx3             = _mm_add_ps(fjx3,tx);
2413             fjy3             = _mm_add_ps(fjy3,ty);
2414             fjz3             = _mm_add_ps(fjz3,tz);
2415
2416             /**************************
2417              * CALCULATE INTERACTIONS *
2418              **************************/
2419
2420             r31              = _mm_mul_ps(rsq31,rinv31);
2421             r31              = _mm_andnot_ps(dummy_mask,r31);
2422
2423             /* EWALD ELECTROSTATICS */
2424
2425             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2426             ewrt             = _mm_mul_ps(r31,ewtabscale);
2427             ewitab           = _mm_cvttps_epi32(ewrt);
2428             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2429             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2430                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2431                                          &ewtabF,&ewtabFn);
2432             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2433             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
2434
2435             fscal            = felec;
2436
2437             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2438
2439             /* Calculate temporary vectorial force */
2440             tx               = _mm_mul_ps(fscal,dx31);
2441             ty               = _mm_mul_ps(fscal,dy31);
2442             tz               = _mm_mul_ps(fscal,dz31);
2443
2444             /* Update vectorial force */
2445             fix3             = _mm_add_ps(fix3,tx);
2446             fiy3             = _mm_add_ps(fiy3,ty);
2447             fiz3             = _mm_add_ps(fiz3,tz);
2448
2449             fjx1             = _mm_add_ps(fjx1,tx);
2450             fjy1             = _mm_add_ps(fjy1,ty);
2451             fjz1             = _mm_add_ps(fjz1,tz);
2452
2453             /**************************
2454              * CALCULATE INTERACTIONS *
2455              **************************/
2456
2457             r32              = _mm_mul_ps(rsq32,rinv32);
2458             r32              = _mm_andnot_ps(dummy_mask,r32);
2459
2460             /* EWALD ELECTROSTATICS */
2461
2462             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2463             ewrt             = _mm_mul_ps(r32,ewtabscale);
2464             ewitab           = _mm_cvttps_epi32(ewrt);
2465             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2466             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2467                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2468                                          &ewtabF,&ewtabFn);
2469             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2470             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
2471
2472             fscal            = felec;
2473
2474             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2475
2476             /* Calculate temporary vectorial force */
2477             tx               = _mm_mul_ps(fscal,dx32);
2478             ty               = _mm_mul_ps(fscal,dy32);
2479             tz               = _mm_mul_ps(fscal,dz32);
2480
2481             /* Update vectorial force */
2482             fix3             = _mm_add_ps(fix3,tx);
2483             fiy3             = _mm_add_ps(fiy3,ty);
2484             fiz3             = _mm_add_ps(fiz3,tz);
2485
2486             fjx2             = _mm_add_ps(fjx2,tx);
2487             fjy2             = _mm_add_ps(fjy2,ty);
2488             fjz2             = _mm_add_ps(fjz2,tz);
2489
2490             /**************************
2491              * CALCULATE INTERACTIONS *
2492              **************************/
2493
2494             r33              = _mm_mul_ps(rsq33,rinv33);
2495             r33              = _mm_andnot_ps(dummy_mask,r33);
2496
2497             /* EWALD ELECTROSTATICS */
2498
2499             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2500             ewrt             = _mm_mul_ps(r33,ewtabscale);
2501             ewitab           = _mm_cvttps_epi32(ewrt);
2502             eweps            = _mm_sub_ps(ewrt,_mm_round_ps(ewrt, _MM_FROUND_FLOOR));
2503             gmx_mm_load_4pair_swizzle_ps(ewtab + gmx_mm_extract_epi32(ewitab,0),ewtab + gmx_mm_extract_epi32(ewitab,1),
2504                                          ewtab + gmx_mm_extract_epi32(ewitab,2),ewtab + gmx_mm_extract_epi32(ewitab,3),
2505                                          &ewtabF,&ewtabFn);
2506             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2507             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2508
2509             fscal            = felec;
2510
2511             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2512
2513             /* Calculate temporary vectorial force */
2514             tx               = _mm_mul_ps(fscal,dx33);
2515             ty               = _mm_mul_ps(fscal,dy33);
2516             tz               = _mm_mul_ps(fscal,dz33);
2517
2518             /* Update vectorial force */
2519             fix3             = _mm_add_ps(fix3,tx);
2520             fiy3             = _mm_add_ps(fiy3,ty);
2521             fiz3             = _mm_add_ps(fiz3,tz);
2522
2523             fjx3             = _mm_add_ps(fjx3,tx);
2524             fjy3             = _mm_add_ps(fjy3,ty);
2525             fjz3             = _mm_add_ps(fjz3,tz);
2526
2527             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2528             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2529             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2530             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2531
2532             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2533                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2534                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2535
2536             /* Inner loop uses 385 flops */
2537         }
2538
2539         /* End of innermost loop */
2540
2541         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2542                                               f+i_coord_offset,fshift+i_shift_offset);
2543
2544         /* Increment number of inner iterations */
2545         inneriter                  += j_index_end - j_index_start;
2546
2547         /* Outer loop uses 24 flops */
2548     }
2549
2550     /* Increment number of outer iterations */
2551     outeriter        += nri;
2552
2553     /* Update outer/inner flops */
2554
2555     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*385);
2556 }