src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_single/nb_kernel_ElecCSTab_VdwLJ_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 "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/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_ElecCSTab_VdwLJ_GeomW4W4_VF_sse4_1_single
  52  * Electrostatics interaction: CubicSplineTable
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCSTab_VdwLJ_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     __m128           dummy_mask,cutoff_mask;
 122     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 123     __m128           one     = _mm_set1_ps(1.0);
 124     __m128           two     = _mm_set1_ps(2.0);
 125     x                = xx[0];
 126     f                = ff[0];
 127
 128     nri              = nlist->nri;
 129     iinr             = nlist->iinr;
 130     jindex           = nlist->jindex;
 131     jjnr             = nlist->jjnr;
 132     shiftidx         = nlist->shift;
 133     gid              = nlist->gid;
 134     shiftvec         = fr->shift_vec[0];
 135     fshift           = fr->fshift[0];
 136     facel            = _mm_set1_ps(fr->epsfac);
 137     charge           = mdatoms->chargeA;
 138     nvdwtype         = fr->ntype;
 139     vdwparam         = fr->nbfp;
 140     vdwtype          = mdatoms->typeA;
 141
 142     vftab            = kernel_data->table_elec->data;
 143     vftabscale       = _mm_set1_ps(kernel_data->table_elec->scale);
 144
 145     /* Setup water-specific parameters */
 146     inr              = nlist->iinr[0];
 147     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 148     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 149     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 150     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 151
 152     jq1              = _mm_set1_ps(charge[inr+1]);
 153     jq2              = _mm_set1_ps(charge[inr+2]);
 154     jq3              = _mm_set1_ps(charge[inr+3]);
 155     vdwjidx0A        = 2*vdwtype[inr+0];
 156     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 157     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 158     qq11             = _mm_mul_ps(iq1,jq1);
 159     qq12             = _mm_mul_ps(iq1,jq2);
 160     qq13             = _mm_mul_ps(iq1,jq3);
 161     qq21             = _mm_mul_ps(iq2,jq1);
 162     qq22             = _mm_mul_ps(iq2,jq2);
 163     qq23             = _mm_mul_ps(iq2,jq3);
 164     qq31             = _mm_mul_ps(iq3,jq1);
 165     qq32             = _mm_mul_ps(iq3,jq2);
 166     qq33             = _mm_mul_ps(iq3,jq3);
 167
 168     /* Avoid stupid compiler warnings */
 169     jnrA = jnrB = jnrC = jnrD = 0;
 170     j_coord_offsetA = 0;
 171     j_coord_offsetB = 0;
 172     j_coord_offsetC = 0;
 173     j_coord_offsetD = 0;
 174
 175     outeriter        = 0;
 176     inneriter        = 0;
 177
 178     for(iidx=0;iidx<4*DIM;iidx++)
 179     {
 180         scratch[iidx] = 0.0;
 181     }
 182
 183     /* Start outer loop over neighborlists */
 184     for(iidx=0; iidx<nri; iidx++)
 185     {
 186         /* Load shift vector for this list */
 187         i_shift_offset   = DIM*shiftidx[iidx];
 188
 189         /* Load limits for loop over neighbors */
 190         j_index_start    = jindex[iidx];
 191         j_index_end      = jindex[iidx+1];
 192
 193         /* Get outer coordinate index */
 194         inr              = iinr[iidx];
 195         i_coord_offset   = DIM*inr;
 196
 197         /* Load i particle coords and add shift vector */
 198         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 199                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 200
 201         fix0             = _mm_setzero_ps();
 202         fiy0             = _mm_setzero_ps();
 203         fiz0             = _mm_setzero_ps();
 204         fix1             = _mm_setzero_ps();
 205         fiy1             = _mm_setzero_ps();
 206         fiz1             = _mm_setzero_ps();
 207         fix2             = _mm_setzero_ps();
 208         fiy2             = _mm_setzero_ps();
 209         fiz2             = _mm_setzero_ps();
 210         fix3             = _mm_setzero_ps();
 211         fiy3             = _mm_setzero_ps();
 212         fiz3             = _mm_setzero_ps();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm_setzero_ps();
 216         vvdwsum          = _mm_setzero_ps();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 220         {
 221
 222             /* Get j neighbor index, and coordinate index */
 223             jnrA             = jjnr[jidx];
 224             jnrB             = jjnr[jidx+1];
 225             jnrC             = jjnr[jidx+2];
 226             jnrD             = jjnr[jidx+3];
 227             j_coord_offsetA  = DIM*jnrA;
 228             j_coord_offsetB  = DIM*jnrB;
 229             j_coord_offsetC  = DIM*jnrC;
 230             j_coord_offsetD  = DIM*jnrD;
 231
 232             /* load j atom coordinates */
 233             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 234                                               x+j_coord_offsetC,x+j_coord_offsetD,
 235                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 236                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 237
 238             /* Calculate displacement vector */
 239             dx00             = _mm_sub_ps(ix0,jx0);
 240             dy00             = _mm_sub_ps(iy0,jy0);
 241             dz00             = _mm_sub_ps(iz0,jz0);
 242             dx11             = _mm_sub_ps(ix1,jx1);
 243             dy11             = _mm_sub_ps(iy1,jy1);
 244             dz11             = _mm_sub_ps(iz1,jz1);
 245             dx12             = _mm_sub_ps(ix1,jx2);
 246             dy12             = _mm_sub_ps(iy1,jy2);
 247             dz12             = _mm_sub_ps(iz1,jz2);
 248             dx13             = _mm_sub_ps(ix1,jx3);
 249             dy13             = _mm_sub_ps(iy1,jy3);
 250             dz13             = _mm_sub_ps(iz1,jz3);
 251             dx21             = _mm_sub_ps(ix2,jx1);
 252             dy21             = _mm_sub_ps(iy2,jy1);
 253             dz21             = _mm_sub_ps(iz2,jz1);
 254             dx22             = _mm_sub_ps(ix2,jx2);
 255             dy22             = _mm_sub_ps(iy2,jy2);
 256             dz22             = _mm_sub_ps(iz2,jz2);
 257             dx23             = _mm_sub_ps(ix2,jx3);
 258             dy23             = _mm_sub_ps(iy2,jy3);
 259             dz23             = _mm_sub_ps(iz2,jz3);
 260             dx31             = _mm_sub_ps(ix3,jx1);
 261             dy31             = _mm_sub_ps(iy3,jy1);
 262             dz31             = _mm_sub_ps(iz3,jz1);
 263             dx32             = _mm_sub_ps(ix3,jx2);
 264             dy32             = _mm_sub_ps(iy3,jy2);
 265             dz32             = _mm_sub_ps(iz3,jz2);
 266             dx33             = _mm_sub_ps(ix3,jx3);
 267             dy33             = _mm_sub_ps(iy3,jy3);
 268             dz33             = _mm_sub_ps(iz3,jz3);
 269
 270             /* Calculate squared distance and things based on it */
 271             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 272             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 273             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 274             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 275             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 276             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 277             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 278             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 279             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 280             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 281
 282             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 283             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 284             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 285             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 286             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 287             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 288             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 289             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 290             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 291
 292             rinvsq00         = gmx_mm_inv_ps(rsq00);
 293
 294             fjx0             = _mm_setzero_ps();
 295             fjy0             = _mm_setzero_ps();
 296             fjz0             = _mm_setzero_ps();
 297             fjx1             = _mm_setzero_ps();
 298             fjy1             = _mm_setzero_ps();
 299             fjz1             = _mm_setzero_ps();
 300             fjx2             = _mm_setzero_ps();
 301             fjy2             = _mm_setzero_ps();
 302             fjz2             = _mm_setzero_ps();
 303             fjx3             = _mm_setzero_ps();
 304             fjy3             = _mm_setzero_ps();
 305             fjz3             = _mm_setzero_ps();
 306
 307             /**************************
 308              * CALCULATE INTERACTIONS *
 309              **************************/
 310
 311             /* LENNARD-JONES DISPERSION/REPULSION */
 312
 313             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 314             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 315             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 316             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 317             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 318
 319             /* Update potential sum for this i atom from the interaction with this j atom. */
 320             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 321
 322             fscal            = fvdw;
 323
 324             /* Calculate temporary vectorial force */
 325             tx               = _mm_mul_ps(fscal,dx00);
 326             ty               = _mm_mul_ps(fscal,dy00);
 327             tz               = _mm_mul_ps(fscal,dz00);
 328
 329             /* Update vectorial force */
 330             fix0             = _mm_add_ps(fix0,tx);
 331             fiy0             = _mm_add_ps(fiy0,ty);
 332             fiz0             = _mm_add_ps(fiz0,tz);
 333
 334             fjx0             = _mm_add_ps(fjx0,tx);
 335             fjy0             = _mm_add_ps(fjy0,ty);
 336             fjz0             = _mm_add_ps(fjz0,tz);
 337
 338             /**************************
 339              * CALCULATE INTERACTIONS *
 340              **************************/
 341
 342             r11              = _mm_mul_ps(rsq11,rinv11);
 343
 344             /* Calculate table index by multiplying r with table scale and truncate to integer */
 345             rt               = _mm_mul_ps(r11,vftabscale);
 346             vfitab           = _mm_cvttps_epi32(rt);
 347             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 348             vfitab           = _mm_slli_epi32(vfitab,2);
 349
 350             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 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             velec            = _mm_mul_ps(qq11,VV);
 360             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 361             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 362
 363             /* Update potential sum for this i atom from the interaction with this j atom. */
 364             velecsum         = _mm_add_ps(velecsum,velec);
 365
 366             fscal            = felec;
 367
 368             /* Calculate temporary vectorial force */
 369             tx               = _mm_mul_ps(fscal,dx11);
 370             ty               = _mm_mul_ps(fscal,dy11);
 371             tz               = _mm_mul_ps(fscal,dz11);
 372
 373             /* Update vectorial force */
 374             fix1             = _mm_add_ps(fix1,tx);
 375             fiy1             = _mm_add_ps(fiy1,ty);
 376             fiz1             = _mm_add_ps(fiz1,tz);
 377
 378             fjx1             = _mm_add_ps(fjx1,tx);
 379             fjy1             = _mm_add_ps(fjy1,ty);
 380             fjz1             = _mm_add_ps(fjz1,tz);
 381
 382             /**************************
 383              * CALCULATE INTERACTIONS *
 384              **************************/
 385
 386             r12              = _mm_mul_ps(rsq12,rinv12);
 387
 388             /* Calculate table index by multiplying r with table scale and truncate to integer */
 389             rt               = _mm_mul_ps(r12,vftabscale);
 390             vfitab           = _mm_cvttps_epi32(rt);
 391             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 392             vfitab           = _mm_slli_epi32(vfitab,2);
 393
 394             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 395             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 396             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 397             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 398             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 399             _MM_TRANSPOSE4_PS(Y,F,G,H);
 400             Heps             = _mm_mul_ps(vfeps,H);
 401             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 402             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 403             velec            = _mm_mul_ps(qq12,VV);
 404             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 405             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 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,dx12);
 414             ty               = _mm_mul_ps(fscal,dy12);
 415             tz               = _mm_mul_ps(fscal,dz12);
 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             fjx2             = _mm_add_ps(fjx2,tx);
 423             fjy2             = _mm_add_ps(fjy2,ty);
 424             fjz2             = _mm_add_ps(fjz2,tz);
 425
 426             /**************************
 427              * CALCULATE INTERACTIONS *
 428              **************************/
 429
 430             r13              = _mm_mul_ps(rsq13,rinv13);
 431
 432             /* Calculate table index by multiplying r with table scale and truncate to integer */
 433             rt               = _mm_mul_ps(r13,vftabscale);
 434             vfitab           = _mm_cvttps_epi32(rt);
 435             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 436             vfitab           = _mm_slli_epi32(vfitab,2);
 437
 438             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 439             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 440             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 441             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 442             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 443             _MM_TRANSPOSE4_PS(Y,F,G,H);
 444             Heps             = _mm_mul_ps(vfeps,H);
 445             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 446             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 447             velec            = _mm_mul_ps(qq13,VV);
 448             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 449             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
 450
 451             /* Update potential sum for this i atom from the interaction with this j atom. */
 452             velecsum         = _mm_add_ps(velecsum,velec);
 453
 454             fscal            = felec;
 455
 456             /* Calculate temporary vectorial force */
 457             tx               = _mm_mul_ps(fscal,dx13);
 458             ty               = _mm_mul_ps(fscal,dy13);
 459             tz               = _mm_mul_ps(fscal,dz13);
 460
 461             /* Update vectorial force */
 462             fix1             = _mm_add_ps(fix1,tx);
 463             fiy1             = _mm_add_ps(fiy1,ty);
 464             fiz1             = _mm_add_ps(fiz1,tz);
 465
 466             fjx3             = _mm_add_ps(fjx3,tx);
 467             fjy3             = _mm_add_ps(fjy3,ty);
 468             fjz3             = _mm_add_ps(fjz3,tz);
 469
 470             /**************************
 471              * CALCULATE INTERACTIONS *
 472              **************************/
 473
 474             r21              = _mm_mul_ps(rsq21,rinv21);
 475
 476             /* Calculate table index by multiplying r with table scale and truncate to integer */
 477             rt               = _mm_mul_ps(r21,vftabscale);
 478             vfitab           = _mm_cvttps_epi32(rt);
 479             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 480             vfitab           = _mm_slli_epi32(vfitab,2);
 481
 482             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 483             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 484             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 485             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 486             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 487             _MM_TRANSPOSE4_PS(Y,F,G,H);
 488             Heps             = _mm_mul_ps(vfeps,H);
 489             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 490             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 491             velec            = _mm_mul_ps(qq21,VV);
 492             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 493             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
 494
 495             /* Update potential sum for this i atom from the interaction with this j atom. */
 496             velecsum         = _mm_add_ps(velecsum,velec);
 497
 498             fscal            = felec;
 499
 500             /* Calculate temporary vectorial force */
 501             tx               = _mm_mul_ps(fscal,dx21);
 502             ty               = _mm_mul_ps(fscal,dy21);
 503             tz               = _mm_mul_ps(fscal,dz21);
 504
 505             /* Update vectorial force */
 506             fix2             = _mm_add_ps(fix2,tx);
 507             fiy2             = _mm_add_ps(fiy2,ty);
 508             fiz2             = _mm_add_ps(fiz2,tz);
 509
 510             fjx1             = _mm_add_ps(fjx1,tx);
 511             fjy1             = _mm_add_ps(fjy1,ty);
 512             fjz1             = _mm_add_ps(fjz1,tz);
 513
 514             /**************************
 515              * CALCULATE INTERACTIONS *
 516              **************************/
 517
 518             r22              = _mm_mul_ps(rsq22,rinv22);
 519
 520             /* Calculate table index by multiplying r with table scale and truncate to integer */
 521             rt               = _mm_mul_ps(r22,vftabscale);
 522             vfitab           = _mm_cvttps_epi32(rt);
 523             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 524             vfitab           = _mm_slli_epi32(vfitab,2);
 525
 526             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 527             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 528             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 529             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 530             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 531             _MM_TRANSPOSE4_PS(Y,F,G,H);
 532             Heps             = _mm_mul_ps(vfeps,H);
 533             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 534             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 535             velec            = _mm_mul_ps(qq22,VV);
 536             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 537             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
 538
 539             /* Update potential sum for this i atom from the interaction with this j atom. */
 540             velecsum         = _mm_add_ps(velecsum,velec);
 541
 542             fscal            = felec;
 543
 544             /* Calculate temporary vectorial force */
 545             tx               = _mm_mul_ps(fscal,dx22);
 546             ty               = _mm_mul_ps(fscal,dy22);
 547             tz               = _mm_mul_ps(fscal,dz22);
 548
 549             /* Update vectorial force */
 550             fix2             = _mm_add_ps(fix2,tx);
 551             fiy2             = _mm_add_ps(fiy2,ty);
 552             fiz2             = _mm_add_ps(fiz2,tz);
 553
 554             fjx2             = _mm_add_ps(fjx2,tx);
 555             fjy2             = _mm_add_ps(fjy2,ty);
 556             fjz2             = _mm_add_ps(fjz2,tz);
 557
 558             /**************************
 559              * CALCULATE INTERACTIONS *
 560              **************************/
 561
 562             r23              = _mm_mul_ps(rsq23,rinv23);
 563
 564             /* Calculate table index by multiplying r with table scale and truncate to integer */
 565             rt               = _mm_mul_ps(r23,vftabscale);
 566             vfitab           = _mm_cvttps_epi32(rt);
 567             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 568             vfitab           = _mm_slli_epi32(vfitab,2);
 569
 570             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 571             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 572             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 573             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 574             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 575             _MM_TRANSPOSE4_PS(Y,F,G,H);
 576             Heps             = _mm_mul_ps(vfeps,H);
 577             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 578             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 579             velec            = _mm_mul_ps(qq23,VV);
 580             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 581             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
 582
 583             /* Update potential sum for this i atom from the interaction with this j atom. */
 584             velecsum         = _mm_add_ps(velecsum,velec);
 585
 586             fscal            = felec;
 587
 588             /* Calculate temporary vectorial force */
 589             tx               = _mm_mul_ps(fscal,dx23);
 590             ty               = _mm_mul_ps(fscal,dy23);
 591             tz               = _mm_mul_ps(fscal,dz23);
 592
 593             /* Update vectorial force */
 594             fix2             = _mm_add_ps(fix2,tx);
 595             fiy2             = _mm_add_ps(fiy2,ty);
 596             fiz2             = _mm_add_ps(fiz2,tz);
 597
 598             fjx3             = _mm_add_ps(fjx3,tx);
 599             fjy3             = _mm_add_ps(fjy3,ty);
 600             fjz3             = _mm_add_ps(fjz3,tz);
 601
 602             /**************************
 603              * CALCULATE INTERACTIONS *
 604              **************************/
 605
 606             r31              = _mm_mul_ps(rsq31,rinv31);
 607
 608             /* Calculate table index by multiplying r with table scale and truncate to integer */
 609             rt               = _mm_mul_ps(r31,vftabscale);
 610             vfitab           = _mm_cvttps_epi32(rt);
 611             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 612             vfitab           = _mm_slli_epi32(vfitab,2);
 613
 614             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 615             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 616             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 617             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 618             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 619             _MM_TRANSPOSE4_PS(Y,F,G,H);
 620             Heps             = _mm_mul_ps(vfeps,H);
 621             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 622             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 623             velec            = _mm_mul_ps(qq31,VV);
 624             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 625             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
 626
 627             /* Update potential sum for this i atom from the interaction with this j atom. */
 628             velecsum         = _mm_add_ps(velecsum,velec);
 629
 630             fscal            = felec;
 631
 632             /* Calculate temporary vectorial force */
 633             tx               = _mm_mul_ps(fscal,dx31);
 634             ty               = _mm_mul_ps(fscal,dy31);
 635             tz               = _mm_mul_ps(fscal,dz31);
 636
 637             /* Update vectorial force */
 638             fix3             = _mm_add_ps(fix3,tx);
 639             fiy3             = _mm_add_ps(fiy3,ty);
 640             fiz3             = _mm_add_ps(fiz3,tz);
 641
 642             fjx1             = _mm_add_ps(fjx1,tx);
 643             fjy1             = _mm_add_ps(fjy1,ty);
 644             fjz1             = _mm_add_ps(fjz1,tz);
 645
 646             /**************************
 647              * CALCULATE INTERACTIONS *
 648              **************************/
 649
 650             r32              = _mm_mul_ps(rsq32,rinv32);
 651
 652             /* Calculate table index by multiplying r with table scale and truncate to integer */
 653             rt               = _mm_mul_ps(r32,vftabscale);
 654             vfitab           = _mm_cvttps_epi32(rt);
 655             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 656             vfitab           = _mm_slli_epi32(vfitab,2);
 657
 658             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 659             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 660             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 661             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 662             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 663             _MM_TRANSPOSE4_PS(Y,F,G,H);
 664             Heps             = _mm_mul_ps(vfeps,H);
 665             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 666             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 667             velec            = _mm_mul_ps(qq32,VV);
 668             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 669             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
 670
 671             /* Update potential sum for this i atom from the interaction with this j atom. */
 672             velecsum         = _mm_add_ps(velecsum,velec);
 673
 674             fscal            = felec;
 675
 676             /* Calculate temporary vectorial force */
 677             tx               = _mm_mul_ps(fscal,dx32);
 678             ty               = _mm_mul_ps(fscal,dy32);
 679             tz               = _mm_mul_ps(fscal,dz32);
 680
 681             /* Update vectorial force */
 682             fix3             = _mm_add_ps(fix3,tx);
 683             fiy3             = _mm_add_ps(fiy3,ty);
 684             fiz3             = _mm_add_ps(fiz3,tz);
 685
 686             fjx2             = _mm_add_ps(fjx2,tx);
 687             fjy2             = _mm_add_ps(fjy2,ty);
 688             fjz2             = _mm_add_ps(fjz2,tz);
 689
 690             /**************************
 691              * CALCULATE INTERACTIONS *
 692              **************************/
 693
 694             r33              = _mm_mul_ps(rsq33,rinv33);
 695
 696             /* Calculate table index by multiplying r with table scale and truncate to integer */
 697             rt               = _mm_mul_ps(r33,vftabscale);
 698             vfitab           = _mm_cvttps_epi32(rt);
 699             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 700             vfitab           = _mm_slli_epi32(vfitab,2);
 701
 702             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 703             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 704             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 705             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 706             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 707             _MM_TRANSPOSE4_PS(Y,F,G,H);
 708             Heps             = _mm_mul_ps(vfeps,H);
 709             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 710             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 711             velec            = _mm_mul_ps(qq33,VV);
 712             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 713             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
 714
 715             /* Update potential sum for this i atom from the interaction with this j atom. */
 716             velecsum         = _mm_add_ps(velecsum,velec);
 717
 718             fscal            = felec;
 719
 720             /* Calculate temporary vectorial force */
 721             tx               = _mm_mul_ps(fscal,dx33);
 722             ty               = _mm_mul_ps(fscal,dy33);
 723             tz               = _mm_mul_ps(fscal,dz33);
 724
 725             /* Update vectorial force */
 726             fix3             = _mm_add_ps(fix3,tx);
 727             fiy3             = _mm_add_ps(fiy3,ty);
 728             fiz3             = _mm_add_ps(fiz3,tz);
 729
 730             fjx3             = _mm_add_ps(fjx3,tx);
 731             fjy3             = _mm_add_ps(fjy3,ty);
 732             fjz3             = _mm_add_ps(fjz3,tz);
 733
 734             fjptrA             = f+j_coord_offsetA;
 735             fjptrB             = f+j_coord_offsetB;
 736             fjptrC             = f+j_coord_offsetC;
 737             fjptrD             = f+j_coord_offsetD;
 738
 739             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 740                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 741                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 742
 743             /* Inner loop uses 422 flops */
 744         }
 745
 746         if(jidx<j_index_end)
 747         {
 748
 749             /* Get j neighbor index, and coordinate index */
 750             jnrlistA         = jjnr[jidx];
 751             jnrlistB         = jjnr[jidx+1];
 752             jnrlistC         = jjnr[jidx+2];
 753             jnrlistD         = jjnr[jidx+3];
 754             /* Sign of each element will be negative for non-real atoms.
 755              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 756              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 757              */
 758             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 759             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 760             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 761             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 762             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 763             j_coord_offsetA  = DIM*jnrA;
 764             j_coord_offsetB  = DIM*jnrB;
 765             j_coord_offsetC  = DIM*jnrC;
 766             j_coord_offsetD  = DIM*jnrD;
 767
 768             /* load j atom coordinates */
 769             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 770                                               x+j_coord_offsetC,x+j_coord_offsetD,
 771                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 772                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 773
 774             /* Calculate displacement vector */
 775             dx00             = _mm_sub_ps(ix0,jx0);
 776             dy00             = _mm_sub_ps(iy0,jy0);
 777             dz00             = _mm_sub_ps(iz0,jz0);
 778             dx11             = _mm_sub_ps(ix1,jx1);
 779             dy11             = _mm_sub_ps(iy1,jy1);
 780             dz11             = _mm_sub_ps(iz1,jz1);
 781             dx12             = _mm_sub_ps(ix1,jx2);
 782             dy12             = _mm_sub_ps(iy1,jy2);
 783             dz12             = _mm_sub_ps(iz1,jz2);
 784             dx13             = _mm_sub_ps(ix1,jx3);
 785             dy13             = _mm_sub_ps(iy1,jy3);
 786             dz13             = _mm_sub_ps(iz1,jz3);
 787             dx21             = _mm_sub_ps(ix2,jx1);
 788             dy21             = _mm_sub_ps(iy2,jy1);
 789             dz21             = _mm_sub_ps(iz2,jz1);
 790             dx22             = _mm_sub_ps(ix2,jx2);
 791             dy22             = _mm_sub_ps(iy2,jy2);
 792             dz22             = _mm_sub_ps(iz2,jz2);
 793             dx23             = _mm_sub_ps(ix2,jx3);
 794             dy23             = _mm_sub_ps(iy2,jy3);
 795             dz23             = _mm_sub_ps(iz2,jz3);
 796             dx31             = _mm_sub_ps(ix3,jx1);
 797             dy31             = _mm_sub_ps(iy3,jy1);
 798             dz31             = _mm_sub_ps(iz3,jz1);
 799             dx32             = _mm_sub_ps(ix3,jx2);
 800             dy32             = _mm_sub_ps(iy3,jy2);
 801             dz32             = _mm_sub_ps(iz3,jz2);
 802             dx33             = _mm_sub_ps(ix3,jx3);
 803             dy33             = _mm_sub_ps(iy3,jy3);
 804             dz33             = _mm_sub_ps(iz3,jz3);
 805
 806             /* Calculate squared distance and things based on it */
 807             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 808             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 809             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 810             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 811             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 812             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 813             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 814             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 815             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 816             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 817
 818             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 819             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 820             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 821             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 822             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 823             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 824             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 825             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 826             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 827
 828             rinvsq00         = gmx_mm_inv_ps(rsq00);
 829
 830             fjx0             = _mm_setzero_ps();
 831             fjy0             = _mm_setzero_ps();
 832             fjz0             = _mm_setzero_ps();
 833             fjx1             = _mm_setzero_ps();
 834             fjy1             = _mm_setzero_ps();
 835             fjz1             = _mm_setzero_ps();
 836             fjx2             = _mm_setzero_ps();
 837             fjy2             = _mm_setzero_ps();
 838             fjz2             = _mm_setzero_ps();
 839             fjx3             = _mm_setzero_ps();
 840             fjy3             = _mm_setzero_ps();
 841             fjz3             = _mm_setzero_ps();
 842
 843             /**************************
 844              * CALCULATE INTERACTIONS *
 845              **************************/
 846
 847             /* LENNARD-JONES DISPERSION/REPULSION */
 848
 849             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 850             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 851             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 852             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 853             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 854
 855             /* Update potential sum for this i atom from the interaction with this j atom. */
 856             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 857             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 858
 859             fscal            = fvdw;
 860
 861             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 862
 863             /* Calculate temporary vectorial force */
 864             tx               = _mm_mul_ps(fscal,dx00);
 865             ty               = _mm_mul_ps(fscal,dy00);
 866             tz               = _mm_mul_ps(fscal,dz00);
 867
 868             /* Update vectorial force */
 869             fix0             = _mm_add_ps(fix0,tx);
 870             fiy0             = _mm_add_ps(fiy0,ty);
 871             fiz0             = _mm_add_ps(fiz0,tz);
 872
 873             fjx0             = _mm_add_ps(fjx0,tx);
 874             fjy0             = _mm_add_ps(fjy0,ty);
 875             fjz0             = _mm_add_ps(fjz0,tz);
 876
 877             /**************************
 878              * CALCULATE INTERACTIONS *
 879              **************************/
 880
 881             r11              = _mm_mul_ps(rsq11,rinv11);
 882             r11              = _mm_andnot_ps(dummy_mask,r11);
 883
 884             /* Calculate table index by multiplying r with table scale and truncate to integer */
 885             rt               = _mm_mul_ps(r11,vftabscale);
 886             vfitab           = _mm_cvttps_epi32(rt);
 887             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 888             vfitab           = _mm_slli_epi32(vfitab,2);
 889
 890             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 891             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 892             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 893             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 894             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 895             _MM_TRANSPOSE4_PS(Y,F,G,H);
 896             Heps             = _mm_mul_ps(vfeps,H);
 897             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 898             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 899             velec            = _mm_mul_ps(qq11,VV);
 900             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 901             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 902
 903             /* Update potential sum for this i atom from the interaction with this j atom. */
 904             velec            = _mm_andnot_ps(dummy_mask,velec);
 905             velecsum         = _mm_add_ps(velecsum,velec);
 906
 907             fscal            = felec;
 908
 909             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 910
 911             /* Calculate temporary vectorial force */
 912             tx               = _mm_mul_ps(fscal,dx11);
 913             ty               = _mm_mul_ps(fscal,dy11);
 914             tz               = _mm_mul_ps(fscal,dz11);
 915
 916             /* Update vectorial force */
 917             fix1             = _mm_add_ps(fix1,tx);
 918             fiy1             = _mm_add_ps(fiy1,ty);
 919             fiz1             = _mm_add_ps(fiz1,tz);
 920
 921             fjx1             = _mm_add_ps(fjx1,tx);
 922             fjy1             = _mm_add_ps(fjy1,ty);
 923             fjz1             = _mm_add_ps(fjz1,tz);
 924
 925             /**************************
 926              * CALCULATE INTERACTIONS *
 927              **************************/
 928
 929             r12              = _mm_mul_ps(rsq12,rinv12);
 930             r12              = _mm_andnot_ps(dummy_mask,r12);
 931
 932             /* Calculate table index by multiplying r with table scale and truncate to integer */
 933             rt               = _mm_mul_ps(r12,vftabscale);
 934             vfitab           = _mm_cvttps_epi32(rt);
 935             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 936             vfitab           = _mm_slli_epi32(vfitab,2);
 937
 938             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 939             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 940             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 941             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 942             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 943             _MM_TRANSPOSE4_PS(Y,F,G,H);
 944             Heps             = _mm_mul_ps(vfeps,H);
 945             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 946             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 947             velec            = _mm_mul_ps(qq12,VV);
 948             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 949             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 950
 951             /* Update potential sum for this i atom from the interaction with this j atom. */
 952             velec            = _mm_andnot_ps(dummy_mask,velec);
 953             velecsum         = _mm_add_ps(velecsum,velec);
 954
 955             fscal            = felec;
 956
 957             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 958
 959             /* Calculate temporary vectorial force */
 960             tx               = _mm_mul_ps(fscal,dx12);
 961             ty               = _mm_mul_ps(fscal,dy12);
 962             tz               = _mm_mul_ps(fscal,dz12);
 963
 964             /* Update vectorial force */
 965             fix1             = _mm_add_ps(fix1,tx);
 966             fiy1             = _mm_add_ps(fiy1,ty);
 967             fiz1             = _mm_add_ps(fiz1,tz);
 968
 969             fjx2             = _mm_add_ps(fjx2,tx);
 970             fjy2             = _mm_add_ps(fjy2,ty);
 971             fjz2             = _mm_add_ps(fjz2,tz);
 972
 973             /**************************
 974              * CALCULATE INTERACTIONS *
 975              **************************/
 976
 977             r13              = _mm_mul_ps(rsq13,rinv13);
 978             r13              = _mm_andnot_ps(dummy_mask,r13);
 979
 980             /* Calculate table index by multiplying r with table scale and truncate to integer */
 981             rt               = _mm_mul_ps(r13,vftabscale);
 982             vfitab           = _mm_cvttps_epi32(rt);
 983             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 984             vfitab           = _mm_slli_epi32(vfitab,2);
 985
 986             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 987             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 988             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 989             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 990             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 991             _MM_TRANSPOSE4_PS(Y,F,G,H);
 992             Heps             = _mm_mul_ps(vfeps,H);
 993             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 994             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 995             velec            = _mm_mul_ps(qq13,VV);
 996             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 997             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
 998
 999             /* Update potential sum for this i atom from the interaction with this j atom. */
1000             velec            = _mm_andnot_ps(dummy_mask,velec);
1001             velecsum         = _mm_add_ps(velecsum,velec);
1002
1003             fscal            = felec;
1004
1005             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1006
1007             /* Calculate temporary vectorial force */
1008             tx               = _mm_mul_ps(fscal,dx13);
1009             ty               = _mm_mul_ps(fscal,dy13);
1010             tz               = _mm_mul_ps(fscal,dz13);
1011
1012             /* Update vectorial force */
1013             fix1             = _mm_add_ps(fix1,tx);
1014             fiy1             = _mm_add_ps(fiy1,ty);
1015             fiz1             = _mm_add_ps(fiz1,tz);
1016
1017             fjx3             = _mm_add_ps(fjx3,tx);
1018             fjy3             = _mm_add_ps(fjy3,ty);
1019             fjz3             = _mm_add_ps(fjz3,tz);
1020
1021             /**************************
1022              * CALCULATE INTERACTIONS *
1023              **************************/
1024
1025             r21              = _mm_mul_ps(rsq21,rinv21);
1026             r21              = _mm_andnot_ps(dummy_mask,r21);
1027
1028             /* Calculate table index by multiplying r with table scale and truncate to integer */
1029             rt               = _mm_mul_ps(r21,vftabscale);
1030             vfitab           = _mm_cvttps_epi32(rt);
1031             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1032             vfitab           = _mm_slli_epi32(vfitab,2);
1033
1034             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1035             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1036             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1037             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1038             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1039             _MM_TRANSPOSE4_PS(Y,F,G,H);
1040             Heps             = _mm_mul_ps(vfeps,H);
1041             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1042             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1043             velec            = _mm_mul_ps(qq21,VV);
1044             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1045             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1046
1047             /* Update potential sum for this i atom from the interaction with this j atom. */
1048             velec            = _mm_andnot_ps(dummy_mask,velec);
1049             velecsum         = _mm_add_ps(velecsum,velec);
1050
1051             fscal            = felec;
1052
1053             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1054
1055             /* Calculate temporary vectorial force */
1056             tx               = _mm_mul_ps(fscal,dx21);
1057             ty               = _mm_mul_ps(fscal,dy21);
1058             tz               = _mm_mul_ps(fscal,dz21);
1059
1060             /* Update vectorial force */
1061             fix2             = _mm_add_ps(fix2,tx);
1062             fiy2             = _mm_add_ps(fiy2,ty);
1063             fiz2             = _mm_add_ps(fiz2,tz);
1064
1065             fjx1             = _mm_add_ps(fjx1,tx);
1066             fjy1             = _mm_add_ps(fjy1,ty);
1067             fjz1             = _mm_add_ps(fjz1,tz);
1068
1069             /**************************
1070              * CALCULATE INTERACTIONS *
1071              **************************/
1072
1073             r22              = _mm_mul_ps(rsq22,rinv22);
1074             r22              = _mm_andnot_ps(dummy_mask,r22);
1075
1076             /* Calculate table index by multiplying r with table scale and truncate to integer */
1077             rt               = _mm_mul_ps(r22,vftabscale);
1078             vfitab           = _mm_cvttps_epi32(rt);
1079             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1080             vfitab           = _mm_slli_epi32(vfitab,2);
1081
1082             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1083             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1084             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1085             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1086             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1087             _MM_TRANSPOSE4_PS(Y,F,G,H);
1088             Heps             = _mm_mul_ps(vfeps,H);
1089             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1090             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1091             velec            = _mm_mul_ps(qq22,VV);
1092             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1093             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1094
1095             /* Update potential sum for this i atom from the interaction with this j atom. */
1096             velec            = _mm_andnot_ps(dummy_mask,velec);
1097             velecsum         = _mm_add_ps(velecsum,velec);
1098
1099             fscal            = felec;
1100
1101             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1102
1103             /* Calculate temporary vectorial force */
1104             tx               = _mm_mul_ps(fscal,dx22);
1105             ty               = _mm_mul_ps(fscal,dy22);
1106             tz               = _mm_mul_ps(fscal,dz22);
1107
1108             /* Update vectorial force */
1109             fix2             = _mm_add_ps(fix2,tx);
1110             fiy2             = _mm_add_ps(fiy2,ty);
1111             fiz2             = _mm_add_ps(fiz2,tz);
1112
1113             fjx2             = _mm_add_ps(fjx2,tx);
1114             fjy2             = _mm_add_ps(fjy2,ty);
1115             fjz2             = _mm_add_ps(fjz2,tz);
1116
1117             /**************************
1118              * CALCULATE INTERACTIONS *
1119              **************************/
1120
1121             r23              = _mm_mul_ps(rsq23,rinv23);
1122             r23              = _mm_andnot_ps(dummy_mask,r23);
1123
1124             /* Calculate table index by multiplying r with table scale and truncate to integer */
1125             rt               = _mm_mul_ps(r23,vftabscale);
1126             vfitab           = _mm_cvttps_epi32(rt);
1127             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1128             vfitab           = _mm_slli_epi32(vfitab,2);
1129
1130             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1131             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1132             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1133             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1134             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1135             _MM_TRANSPOSE4_PS(Y,F,G,H);
1136             Heps             = _mm_mul_ps(vfeps,H);
1137             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1138             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1139             velec            = _mm_mul_ps(qq23,VV);
1140             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1141             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1142
1143             /* Update potential sum for this i atom from the interaction with this j atom. */
1144             velec            = _mm_andnot_ps(dummy_mask,velec);
1145             velecsum         = _mm_add_ps(velecsum,velec);
1146
1147             fscal            = felec;
1148
1149             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1150
1151             /* Calculate temporary vectorial force */
1152             tx               = _mm_mul_ps(fscal,dx23);
1153             ty               = _mm_mul_ps(fscal,dy23);
1154             tz               = _mm_mul_ps(fscal,dz23);
1155
1156             /* Update vectorial force */
1157             fix2             = _mm_add_ps(fix2,tx);
1158             fiy2             = _mm_add_ps(fiy2,ty);
1159             fiz2             = _mm_add_ps(fiz2,tz);
1160
1161             fjx3             = _mm_add_ps(fjx3,tx);
1162             fjy3             = _mm_add_ps(fjy3,ty);
1163             fjz3             = _mm_add_ps(fjz3,tz);
1164
1165             /**************************
1166              * CALCULATE INTERACTIONS *
1167              **************************/
1168
1169             r31              = _mm_mul_ps(rsq31,rinv31);
1170             r31              = _mm_andnot_ps(dummy_mask,r31);
1171
1172             /* Calculate table index by multiplying r with table scale and truncate to integer */
1173             rt               = _mm_mul_ps(r31,vftabscale);
1174             vfitab           = _mm_cvttps_epi32(rt);
1175             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1176             vfitab           = _mm_slli_epi32(vfitab,2);
1177
1178             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1179             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1180             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1181             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1182             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1183             _MM_TRANSPOSE4_PS(Y,F,G,H);
1184             Heps             = _mm_mul_ps(vfeps,H);
1185             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1186             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1187             velec            = _mm_mul_ps(qq31,VV);
1188             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1189             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1190
1191             /* Update potential sum for this i atom from the interaction with this j atom. */
1192             velec            = _mm_andnot_ps(dummy_mask,velec);
1193             velecsum         = _mm_add_ps(velecsum,velec);
1194
1195             fscal            = felec;
1196
1197             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1198
1199             /* Calculate temporary vectorial force */
1200             tx               = _mm_mul_ps(fscal,dx31);
1201             ty               = _mm_mul_ps(fscal,dy31);
1202             tz               = _mm_mul_ps(fscal,dz31);
1203
1204             /* Update vectorial force */
1205             fix3             = _mm_add_ps(fix3,tx);
1206             fiy3             = _mm_add_ps(fiy3,ty);
1207             fiz3             = _mm_add_ps(fiz3,tz);
1208
1209             fjx1             = _mm_add_ps(fjx1,tx);
1210             fjy1             = _mm_add_ps(fjy1,ty);
1211             fjz1             = _mm_add_ps(fjz1,tz);
1212
1213             /**************************
1214              * CALCULATE INTERACTIONS *
1215              **************************/
1216
1217             r32              = _mm_mul_ps(rsq32,rinv32);
1218             r32              = _mm_andnot_ps(dummy_mask,r32);
1219
1220             /* Calculate table index by multiplying r with table scale and truncate to integer */
1221             rt               = _mm_mul_ps(r32,vftabscale);
1222             vfitab           = _mm_cvttps_epi32(rt);
1223             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1224             vfitab           = _mm_slli_epi32(vfitab,2);
1225
1226             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1227             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1228             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1229             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1230             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1231             _MM_TRANSPOSE4_PS(Y,F,G,H);
1232             Heps             = _mm_mul_ps(vfeps,H);
1233             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1234             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1235             velec            = _mm_mul_ps(qq32,VV);
1236             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1237             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1238
1239             /* Update potential sum for this i atom from the interaction with this j atom. */
1240             velec            = _mm_andnot_ps(dummy_mask,velec);
1241             velecsum         = _mm_add_ps(velecsum,velec);
1242
1243             fscal            = felec;
1244
1245             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1246
1247             /* Calculate temporary vectorial force */
1248             tx               = _mm_mul_ps(fscal,dx32);
1249             ty               = _mm_mul_ps(fscal,dy32);
1250             tz               = _mm_mul_ps(fscal,dz32);
1251
1252             /* Update vectorial force */
1253             fix3             = _mm_add_ps(fix3,tx);
1254             fiy3             = _mm_add_ps(fiy3,ty);
1255             fiz3             = _mm_add_ps(fiz3,tz);
1256
1257             fjx2             = _mm_add_ps(fjx2,tx);
1258             fjy2             = _mm_add_ps(fjy2,ty);
1259             fjz2             = _mm_add_ps(fjz2,tz);
1260
1261             /**************************
1262              * CALCULATE INTERACTIONS *
1263              **************************/
1264
1265             r33              = _mm_mul_ps(rsq33,rinv33);
1266             r33              = _mm_andnot_ps(dummy_mask,r33);
1267
1268             /* Calculate table index by multiplying r with table scale and truncate to integer */
1269             rt               = _mm_mul_ps(r33,vftabscale);
1270             vfitab           = _mm_cvttps_epi32(rt);
1271             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1272             vfitab           = _mm_slli_epi32(vfitab,2);
1273
1274             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1275             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1276             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1277             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1278             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1279             _MM_TRANSPOSE4_PS(Y,F,G,H);
1280             Heps             = _mm_mul_ps(vfeps,H);
1281             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1282             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1283             velec            = _mm_mul_ps(qq33,VV);
1284             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1285             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
1286
1287             /* Update potential sum for this i atom from the interaction with this j atom. */
1288             velec            = _mm_andnot_ps(dummy_mask,velec);
1289             velecsum         = _mm_add_ps(velecsum,velec);
1290
1291             fscal            = felec;
1292
1293             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1294
1295             /* Calculate temporary vectorial force */
1296             tx               = _mm_mul_ps(fscal,dx33);
1297             ty               = _mm_mul_ps(fscal,dy33);
1298             tz               = _mm_mul_ps(fscal,dz33);
1299
1300             /* Update vectorial force */
1301             fix3             = _mm_add_ps(fix3,tx);
1302             fiy3             = _mm_add_ps(fiy3,ty);
1303             fiz3             = _mm_add_ps(fiz3,tz);
1304
1305             fjx3             = _mm_add_ps(fjx3,tx);
1306             fjy3             = _mm_add_ps(fjy3,ty);
1307             fjz3             = _mm_add_ps(fjz3,tz);
1308
1309             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1310             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1311             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1312             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1313
1314             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1315                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1316                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1317
1318             /* Inner loop uses 431 flops */
1319         }
1320
1321         /* End of innermost loop */
1322
1323         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1324                                               f+i_coord_offset,fshift+i_shift_offset);
1325
1326         ggid                        = gid[iidx];
1327         /* Update potential energies */
1328         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1329         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1330
1331         /* Increment number of inner iterations */
1332         inneriter                  += j_index_end - j_index_start;
1333
1334         /* Outer loop uses 26 flops */
1335     }
1336
1337     /* Increment number of outer iterations */
1338     outeriter        += nri;
1339
1340     /* Update outer/inner flops */
1341
1342     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*431);
1343 }
1344 /*
1345  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse4_1_single
1346  * Electrostatics interaction: CubicSplineTable
1347  * VdW interaction:            LennardJones
1348  * Geometry:                   Water4-Water4
1349  * Calculate force/pot:        Force
1350  */
1351 void
1352 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse4_1_single
1353                     (t_nblist                    * gmx_restrict       nlist,
1354                      rvec                        * gmx_restrict          xx,
1355                      rvec                        * gmx_restrict          ff,
1356                      t_forcerec                  * gmx_restrict          fr,
1357                      t_mdatoms                   * gmx_restrict     mdatoms,
1358                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1359                      t_nrnb                      * gmx_restrict        nrnb)
1360 {
1361     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1362      * just 0 for non-waters.
1363      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1364      * jnr indices corresponding to data put in the four positions in the SIMD register.
1365      */
1366     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1367     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1368     int              jnrA,jnrB,jnrC,jnrD;
1369     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1370     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1371     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1372     real             rcutoff_scalar;
1373     real             *shiftvec,*fshift,*x,*f;
1374     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1375     real             scratch[4*DIM];
1376     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1377     int              vdwioffset0;
1378     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1379     int              vdwioffset1;
1380     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1381     int              vdwioffset2;
1382     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1383     int              vdwioffset3;
1384     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1385     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1386     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1387     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1388     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1389     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1390     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1391     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1392     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1393     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1394     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1395     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1396     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1397     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1398     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1399     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1400     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1401     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1402     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1403     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1404     real             *charge;
1405     int              nvdwtype;
1406     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1407     int              *vdwtype;
1408     real             *vdwparam;
1409     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1410     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1411     __m128i          vfitab;
1412     __m128i          ifour       = _mm_set1_epi32(4);
1413     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1414     real             *vftab;
1415     __m128           dummy_mask,cutoff_mask;
1416     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1417     __m128           one     = _mm_set1_ps(1.0);
1418     __m128           two     = _mm_set1_ps(2.0);
1419     x                = xx[0];
1420     f                = ff[0];
1421
1422     nri              = nlist->nri;
1423     iinr             = nlist->iinr;
1424     jindex           = nlist->jindex;
1425     jjnr             = nlist->jjnr;
1426     shiftidx         = nlist->shift;
1427     gid              = nlist->gid;
1428     shiftvec         = fr->shift_vec[0];
1429     fshift           = fr->fshift[0];
1430     facel            = _mm_set1_ps(fr->epsfac);
1431     charge           = mdatoms->chargeA;
1432     nvdwtype         = fr->ntype;
1433     vdwparam         = fr->nbfp;
1434     vdwtype          = mdatoms->typeA;
1435
1436     vftab            = kernel_data->table_elec->data;
1437     vftabscale       = _mm_set1_ps(kernel_data->table_elec->scale);
1438
1439     /* Setup water-specific parameters */
1440     inr              = nlist->iinr[0];
1441     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1442     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1443     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1444     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1445
1446     jq1              = _mm_set1_ps(charge[inr+1]);
1447     jq2              = _mm_set1_ps(charge[inr+2]);
1448     jq3              = _mm_set1_ps(charge[inr+3]);
1449     vdwjidx0A        = 2*vdwtype[inr+0];
1450     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1451     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1452     qq11             = _mm_mul_ps(iq1,jq1);
1453     qq12             = _mm_mul_ps(iq1,jq2);
1454     qq13             = _mm_mul_ps(iq1,jq3);
1455     qq21             = _mm_mul_ps(iq2,jq1);
1456     qq22             = _mm_mul_ps(iq2,jq2);
1457     qq23             = _mm_mul_ps(iq2,jq3);
1458     qq31             = _mm_mul_ps(iq3,jq1);
1459     qq32             = _mm_mul_ps(iq3,jq2);
1460     qq33             = _mm_mul_ps(iq3,jq3);
1461
1462     /* Avoid stupid compiler warnings */
1463     jnrA = jnrB = jnrC = jnrD = 0;
1464     j_coord_offsetA = 0;
1465     j_coord_offsetB = 0;
1466     j_coord_offsetC = 0;
1467     j_coord_offsetD = 0;
1468
1469     outeriter        = 0;
1470     inneriter        = 0;
1471
1472     for(iidx=0;iidx<4*DIM;iidx++)
1473     {
1474         scratch[iidx] = 0.0;
1475     }
1476
1477     /* Start outer loop over neighborlists */
1478     for(iidx=0; iidx<nri; iidx++)
1479     {
1480         /* Load shift vector for this list */
1481         i_shift_offset   = DIM*shiftidx[iidx];
1482
1483         /* Load limits for loop over neighbors */
1484         j_index_start    = jindex[iidx];
1485         j_index_end      = jindex[iidx+1];
1486
1487         /* Get outer coordinate index */
1488         inr              = iinr[iidx];
1489         i_coord_offset   = DIM*inr;
1490
1491         /* Load i particle coords and add shift vector */
1492         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1493                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1494
1495         fix0             = _mm_setzero_ps();
1496         fiy0             = _mm_setzero_ps();
1497         fiz0             = _mm_setzero_ps();
1498         fix1             = _mm_setzero_ps();
1499         fiy1             = _mm_setzero_ps();
1500         fiz1             = _mm_setzero_ps();
1501         fix2             = _mm_setzero_ps();
1502         fiy2             = _mm_setzero_ps();
1503         fiz2             = _mm_setzero_ps();
1504         fix3             = _mm_setzero_ps();
1505         fiy3             = _mm_setzero_ps();
1506         fiz3             = _mm_setzero_ps();
1507
1508         /* Start inner kernel loop */
1509         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1510         {
1511
1512             /* Get j neighbor index, and coordinate index */
1513             jnrA             = jjnr[jidx];
1514             jnrB             = jjnr[jidx+1];
1515             jnrC             = jjnr[jidx+2];
1516             jnrD             = jjnr[jidx+3];
1517             j_coord_offsetA  = DIM*jnrA;
1518             j_coord_offsetB  = DIM*jnrB;
1519             j_coord_offsetC  = DIM*jnrC;
1520             j_coord_offsetD  = DIM*jnrD;
1521
1522             /* load j atom coordinates */
1523             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1524                                               x+j_coord_offsetC,x+j_coord_offsetD,
1525                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1526                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1527
1528             /* Calculate displacement vector */
1529             dx00             = _mm_sub_ps(ix0,jx0);
1530             dy00             = _mm_sub_ps(iy0,jy0);
1531             dz00             = _mm_sub_ps(iz0,jz0);
1532             dx11             = _mm_sub_ps(ix1,jx1);
1533             dy11             = _mm_sub_ps(iy1,jy1);
1534             dz11             = _mm_sub_ps(iz1,jz1);
1535             dx12             = _mm_sub_ps(ix1,jx2);
1536             dy12             = _mm_sub_ps(iy1,jy2);
1537             dz12             = _mm_sub_ps(iz1,jz2);
1538             dx13             = _mm_sub_ps(ix1,jx3);
1539             dy13             = _mm_sub_ps(iy1,jy3);
1540             dz13             = _mm_sub_ps(iz1,jz3);
1541             dx21             = _mm_sub_ps(ix2,jx1);
1542             dy21             = _mm_sub_ps(iy2,jy1);
1543             dz21             = _mm_sub_ps(iz2,jz1);
1544             dx22             = _mm_sub_ps(ix2,jx2);
1545             dy22             = _mm_sub_ps(iy2,jy2);
1546             dz22             = _mm_sub_ps(iz2,jz2);
1547             dx23             = _mm_sub_ps(ix2,jx3);
1548             dy23             = _mm_sub_ps(iy2,jy3);
1549             dz23             = _mm_sub_ps(iz2,jz3);
1550             dx31             = _mm_sub_ps(ix3,jx1);
1551             dy31             = _mm_sub_ps(iy3,jy1);
1552             dz31             = _mm_sub_ps(iz3,jz1);
1553             dx32             = _mm_sub_ps(ix3,jx2);
1554             dy32             = _mm_sub_ps(iy3,jy2);
1555             dz32             = _mm_sub_ps(iz3,jz2);
1556             dx33             = _mm_sub_ps(ix3,jx3);
1557             dy33             = _mm_sub_ps(iy3,jy3);
1558             dz33             = _mm_sub_ps(iz3,jz3);
1559
1560             /* Calculate squared distance and things based on it */
1561             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1562             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1563             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1564             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1565             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1566             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1567             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1568             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1569             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1570             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1571
1572             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1573             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1574             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1575             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1576             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1577             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1578             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1579             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1580             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1581
1582             rinvsq00         = gmx_mm_inv_ps(rsq00);
1583
1584             fjx0             = _mm_setzero_ps();
1585             fjy0             = _mm_setzero_ps();
1586             fjz0             = _mm_setzero_ps();
1587             fjx1             = _mm_setzero_ps();
1588             fjy1             = _mm_setzero_ps();
1589             fjz1             = _mm_setzero_ps();
1590             fjx2             = _mm_setzero_ps();
1591             fjy2             = _mm_setzero_ps();
1592             fjz2             = _mm_setzero_ps();
1593             fjx3             = _mm_setzero_ps();
1594             fjy3             = _mm_setzero_ps();
1595             fjz3             = _mm_setzero_ps();
1596
1597             /**************************
1598              * CALCULATE INTERACTIONS *
1599              **************************/
1600
1601             /* LENNARD-JONES DISPERSION/REPULSION */
1602
1603             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1604             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
1605
1606             fscal            = fvdw;
1607
1608             /* Calculate temporary vectorial force */
1609             tx               = _mm_mul_ps(fscal,dx00);
1610             ty               = _mm_mul_ps(fscal,dy00);
1611             tz               = _mm_mul_ps(fscal,dz00);
1612
1613             /* Update vectorial force */
1614             fix0             = _mm_add_ps(fix0,tx);
1615             fiy0             = _mm_add_ps(fiy0,ty);
1616             fiz0             = _mm_add_ps(fiz0,tz);
1617
1618             fjx0             = _mm_add_ps(fjx0,tx);
1619             fjy0             = _mm_add_ps(fjy0,ty);
1620             fjz0             = _mm_add_ps(fjz0,tz);
1621
1622             /**************************
1623              * CALCULATE INTERACTIONS *
1624              **************************/
1625
1626             r11              = _mm_mul_ps(rsq11,rinv11);
1627
1628             /* Calculate table index by multiplying r with table scale and truncate to integer */
1629             rt               = _mm_mul_ps(r11,vftabscale);
1630             vfitab           = _mm_cvttps_epi32(rt);
1631             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1632             vfitab           = _mm_slli_epi32(vfitab,2);
1633
1634             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1635             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1636             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1637             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1638             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1639             _MM_TRANSPOSE4_PS(Y,F,G,H);
1640             Heps             = _mm_mul_ps(vfeps,H);
1641             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1642             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1643             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
1644
1645             fscal            = felec;
1646
1647             /* Calculate temporary vectorial force */
1648             tx               = _mm_mul_ps(fscal,dx11);
1649             ty               = _mm_mul_ps(fscal,dy11);
1650             tz               = _mm_mul_ps(fscal,dz11);
1651
1652             /* Update vectorial force */
1653             fix1             = _mm_add_ps(fix1,tx);
1654             fiy1             = _mm_add_ps(fiy1,ty);
1655             fiz1             = _mm_add_ps(fiz1,tz);
1656
1657             fjx1             = _mm_add_ps(fjx1,tx);
1658             fjy1             = _mm_add_ps(fjy1,ty);
1659             fjz1             = _mm_add_ps(fjz1,tz);
1660
1661             /**************************
1662              * CALCULATE INTERACTIONS *
1663              **************************/
1664
1665             r12              = _mm_mul_ps(rsq12,rinv12);
1666
1667             /* Calculate table index by multiplying r with table scale and truncate to integer */
1668             rt               = _mm_mul_ps(r12,vftabscale);
1669             vfitab           = _mm_cvttps_epi32(rt);
1670             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1671             vfitab           = _mm_slli_epi32(vfitab,2);
1672
1673             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1674             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1675             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1676             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1677             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1678             _MM_TRANSPOSE4_PS(Y,F,G,H);
1679             Heps             = _mm_mul_ps(vfeps,H);
1680             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1681             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1682             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
1683
1684             fscal            = felec;
1685
1686             /* Calculate temporary vectorial force */
1687             tx               = _mm_mul_ps(fscal,dx12);
1688             ty               = _mm_mul_ps(fscal,dy12);
1689             tz               = _mm_mul_ps(fscal,dz12);
1690
1691             /* Update vectorial force */
1692             fix1             = _mm_add_ps(fix1,tx);
1693             fiy1             = _mm_add_ps(fiy1,ty);
1694             fiz1             = _mm_add_ps(fiz1,tz);
1695
1696             fjx2             = _mm_add_ps(fjx2,tx);
1697             fjy2             = _mm_add_ps(fjy2,ty);
1698             fjz2             = _mm_add_ps(fjz2,tz);
1699
1700             /**************************
1701              * CALCULATE INTERACTIONS *
1702              **************************/
1703
1704             r13              = _mm_mul_ps(rsq13,rinv13);
1705
1706             /* Calculate table index by multiplying r with table scale and truncate to integer */
1707             rt               = _mm_mul_ps(r13,vftabscale);
1708             vfitab           = _mm_cvttps_epi32(rt);
1709             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1710             vfitab           = _mm_slli_epi32(vfitab,2);
1711
1712             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1713             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1714             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1715             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1716             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1717             _MM_TRANSPOSE4_PS(Y,F,G,H);
1718             Heps             = _mm_mul_ps(vfeps,H);
1719             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1720             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1721             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1722
1723             fscal            = felec;
1724
1725             /* Calculate temporary vectorial force */
1726             tx               = _mm_mul_ps(fscal,dx13);
1727             ty               = _mm_mul_ps(fscal,dy13);
1728             tz               = _mm_mul_ps(fscal,dz13);
1729
1730             /* Update vectorial force */
1731             fix1             = _mm_add_ps(fix1,tx);
1732             fiy1             = _mm_add_ps(fiy1,ty);
1733             fiz1             = _mm_add_ps(fiz1,tz);
1734
1735             fjx3             = _mm_add_ps(fjx3,tx);
1736             fjy3             = _mm_add_ps(fjy3,ty);
1737             fjz3             = _mm_add_ps(fjz3,tz);
1738
1739             /**************************
1740              * CALCULATE INTERACTIONS *
1741              **************************/
1742
1743             r21              = _mm_mul_ps(rsq21,rinv21);
1744
1745             /* Calculate table index by multiplying r with table scale and truncate to integer */
1746             rt               = _mm_mul_ps(r21,vftabscale);
1747             vfitab           = _mm_cvttps_epi32(rt);
1748             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1749             vfitab           = _mm_slli_epi32(vfitab,2);
1750
1751             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1752             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1753             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1754             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1755             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1756             _MM_TRANSPOSE4_PS(Y,F,G,H);
1757             Heps             = _mm_mul_ps(vfeps,H);
1758             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1759             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1760             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1761
1762             fscal            = felec;
1763
1764             /* Calculate temporary vectorial force */
1765             tx               = _mm_mul_ps(fscal,dx21);
1766             ty               = _mm_mul_ps(fscal,dy21);
1767             tz               = _mm_mul_ps(fscal,dz21);
1768
1769             /* Update vectorial force */
1770             fix2             = _mm_add_ps(fix2,tx);
1771             fiy2             = _mm_add_ps(fiy2,ty);
1772             fiz2             = _mm_add_ps(fiz2,tz);
1773
1774             fjx1             = _mm_add_ps(fjx1,tx);
1775             fjy1             = _mm_add_ps(fjy1,ty);
1776             fjz1             = _mm_add_ps(fjz1,tz);
1777
1778             /**************************
1779              * CALCULATE INTERACTIONS *
1780              **************************/
1781
1782             r22              = _mm_mul_ps(rsq22,rinv22);
1783
1784             /* Calculate table index by multiplying r with table scale and truncate to integer */
1785             rt               = _mm_mul_ps(r22,vftabscale);
1786             vfitab           = _mm_cvttps_epi32(rt);
1787             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1788             vfitab           = _mm_slli_epi32(vfitab,2);
1789
1790             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1791             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1792             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1793             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1794             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1795             _MM_TRANSPOSE4_PS(Y,F,G,H);
1796             Heps             = _mm_mul_ps(vfeps,H);
1797             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1798             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1799             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1800
1801             fscal            = felec;
1802
1803             /* Calculate temporary vectorial force */
1804             tx               = _mm_mul_ps(fscal,dx22);
1805             ty               = _mm_mul_ps(fscal,dy22);
1806             tz               = _mm_mul_ps(fscal,dz22);
1807
1808             /* Update vectorial force */
1809             fix2             = _mm_add_ps(fix2,tx);
1810             fiy2             = _mm_add_ps(fiy2,ty);
1811             fiz2             = _mm_add_ps(fiz2,tz);
1812
1813             fjx2             = _mm_add_ps(fjx2,tx);
1814             fjy2             = _mm_add_ps(fjy2,ty);
1815             fjz2             = _mm_add_ps(fjz2,tz);
1816
1817             /**************************
1818              * CALCULATE INTERACTIONS *
1819              **************************/
1820
1821             r23              = _mm_mul_ps(rsq23,rinv23);
1822
1823             /* Calculate table index by multiplying r with table scale and truncate to integer */
1824             rt               = _mm_mul_ps(r23,vftabscale);
1825             vfitab           = _mm_cvttps_epi32(rt);
1826             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1827             vfitab           = _mm_slli_epi32(vfitab,2);
1828
1829             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1830             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1831             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1832             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1833             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1834             _MM_TRANSPOSE4_PS(Y,F,G,H);
1835             Heps             = _mm_mul_ps(vfeps,H);
1836             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1837             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1838             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1839
1840             fscal            = felec;
1841
1842             /* Calculate temporary vectorial force */
1843             tx               = _mm_mul_ps(fscal,dx23);
1844             ty               = _mm_mul_ps(fscal,dy23);
1845             tz               = _mm_mul_ps(fscal,dz23);
1846
1847             /* Update vectorial force */
1848             fix2             = _mm_add_ps(fix2,tx);
1849             fiy2             = _mm_add_ps(fiy2,ty);
1850             fiz2             = _mm_add_ps(fiz2,tz);
1851
1852             fjx3             = _mm_add_ps(fjx3,tx);
1853             fjy3             = _mm_add_ps(fjy3,ty);
1854             fjz3             = _mm_add_ps(fjz3,tz);
1855
1856             /**************************
1857              * CALCULATE INTERACTIONS *
1858              **************************/
1859
1860             r31              = _mm_mul_ps(rsq31,rinv31);
1861
1862             /* Calculate table index by multiplying r with table scale and truncate to integer */
1863             rt               = _mm_mul_ps(r31,vftabscale);
1864             vfitab           = _mm_cvttps_epi32(rt);
1865             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1866             vfitab           = _mm_slli_epi32(vfitab,2);
1867
1868             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1869             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1870             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1871             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1872             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1873             _MM_TRANSPOSE4_PS(Y,F,G,H);
1874             Heps             = _mm_mul_ps(vfeps,H);
1875             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1876             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1877             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1878
1879             fscal            = felec;
1880
1881             /* Calculate temporary vectorial force */
1882             tx               = _mm_mul_ps(fscal,dx31);
1883             ty               = _mm_mul_ps(fscal,dy31);
1884             tz               = _mm_mul_ps(fscal,dz31);
1885
1886             /* Update vectorial force */
1887             fix3             = _mm_add_ps(fix3,tx);
1888             fiy3             = _mm_add_ps(fiy3,ty);
1889             fiz3             = _mm_add_ps(fiz3,tz);
1890
1891             fjx1             = _mm_add_ps(fjx1,tx);
1892             fjy1             = _mm_add_ps(fjy1,ty);
1893             fjz1             = _mm_add_ps(fjz1,tz);
1894
1895             /**************************
1896              * CALCULATE INTERACTIONS *
1897              **************************/
1898
1899             r32              = _mm_mul_ps(rsq32,rinv32);
1900
1901             /* Calculate table index by multiplying r with table scale and truncate to integer */
1902             rt               = _mm_mul_ps(r32,vftabscale);
1903             vfitab           = _mm_cvttps_epi32(rt);
1904             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1905             vfitab           = _mm_slli_epi32(vfitab,2);
1906
1907             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1908             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1909             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1910             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1911             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1912             _MM_TRANSPOSE4_PS(Y,F,G,H);
1913             Heps             = _mm_mul_ps(vfeps,H);
1914             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1915             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1916             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1917
1918             fscal            = felec;
1919
1920             /* Calculate temporary vectorial force */
1921             tx               = _mm_mul_ps(fscal,dx32);
1922             ty               = _mm_mul_ps(fscal,dy32);
1923             tz               = _mm_mul_ps(fscal,dz32);
1924
1925             /* Update vectorial force */
1926             fix3             = _mm_add_ps(fix3,tx);
1927             fiy3             = _mm_add_ps(fiy3,ty);
1928             fiz3             = _mm_add_ps(fiz3,tz);
1929
1930             fjx2             = _mm_add_ps(fjx2,tx);
1931             fjy2             = _mm_add_ps(fjy2,ty);
1932             fjz2             = _mm_add_ps(fjz2,tz);
1933
1934             /**************************
1935              * CALCULATE INTERACTIONS *
1936              **************************/
1937
1938             r33              = _mm_mul_ps(rsq33,rinv33);
1939
1940             /* Calculate table index by multiplying r with table scale and truncate to integer */
1941             rt               = _mm_mul_ps(r33,vftabscale);
1942             vfitab           = _mm_cvttps_epi32(rt);
1943             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1944             vfitab           = _mm_slli_epi32(vfitab,2);
1945
1946             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1947             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1948             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1949             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1950             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1951             _MM_TRANSPOSE4_PS(Y,F,G,H);
1952             Heps             = _mm_mul_ps(vfeps,H);
1953             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1954             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1955             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
1956
1957             fscal            = felec;
1958
1959             /* Calculate temporary vectorial force */
1960             tx               = _mm_mul_ps(fscal,dx33);
1961             ty               = _mm_mul_ps(fscal,dy33);
1962             tz               = _mm_mul_ps(fscal,dz33);
1963
1964             /* Update vectorial force */
1965             fix3             = _mm_add_ps(fix3,tx);
1966             fiy3             = _mm_add_ps(fiy3,ty);
1967             fiz3             = _mm_add_ps(fiz3,tz);
1968
1969             fjx3             = _mm_add_ps(fjx3,tx);
1970             fjy3             = _mm_add_ps(fjy3,ty);
1971             fjz3             = _mm_add_ps(fjz3,tz);
1972
1973             fjptrA             = f+j_coord_offsetA;
1974             fjptrB             = f+j_coord_offsetB;
1975             fjptrC             = f+j_coord_offsetC;
1976             fjptrD             = f+j_coord_offsetD;
1977
1978             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1979                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1980                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1981
1982             /* Inner loop uses 381 flops */
1983         }
1984
1985         if(jidx<j_index_end)
1986         {
1987
1988             /* Get j neighbor index, and coordinate index */
1989             jnrlistA         = jjnr[jidx];
1990             jnrlistB         = jjnr[jidx+1];
1991             jnrlistC         = jjnr[jidx+2];
1992             jnrlistD         = jjnr[jidx+3];
1993             /* Sign of each element will be negative for non-real atoms.
1994              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1995              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1996              */
1997             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1998             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1999             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2000             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2001             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2002             j_coord_offsetA  = DIM*jnrA;
2003             j_coord_offsetB  = DIM*jnrB;
2004             j_coord_offsetC  = DIM*jnrC;
2005             j_coord_offsetD  = DIM*jnrD;
2006
2007             /* load j atom coordinates */
2008             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2009                                               x+j_coord_offsetC,x+j_coord_offsetD,
2010                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2011                                               &jy2,&jz2,&jx3,&jy3,&jz3);
2012
2013             /* Calculate displacement vector */
2014             dx00             = _mm_sub_ps(ix0,jx0);
2015             dy00             = _mm_sub_ps(iy0,jy0);
2016             dz00             = _mm_sub_ps(iz0,jz0);
2017             dx11             = _mm_sub_ps(ix1,jx1);
2018             dy11             = _mm_sub_ps(iy1,jy1);
2019             dz11             = _mm_sub_ps(iz1,jz1);
2020             dx12             = _mm_sub_ps(ix1,jx2);
2021             dy12             = _mm_sub_ps(iy1,jy2);
2022             dz12             = _mm_sub_ps(iz1,jz2);
2023             dx13             = _mm_sub_ps(ix1,jx3);
2024             dy13             = _mm_sub_ps(iy1,jy3);
2025             dz13             = _mm_sub_ps(iz1,jz3);
2026             dx21             = _mm_sub_ps(ix2,jx1);
2027             dy21             = _mm_sub_ps(iy2,jy1);
2028             dz21             = _mm_sub_ps(iz2,jz1);
2029             dx22             = _mm_sub_ps(ix2,jx2);
2030             dy22             = _mm_sub_ps(iy2,jy2);
2031             dz22             = _mm_sub_ps(iz2,jz2);
2032             dx23             = _mm_sub_ps(ix2,jx3);
2033             dy23             = _mm_sub_ps(iy2,jy3);
2034             dz23             = _mm_sub_ps(iz2,jz3);
2035             dx31             = _mm_sub_ps(ix3,jx1);
2036             dy31             = _mm_sub_ps(iy3,jy1);
2037             dz31             = _mm_sub_ps(iz3,jz1);
2038             dx32             = _mm_sub_ps(ix3,jx2);
2039             dy32             = _mm_sub_ps(iy3,jy2);
2040             dz32             = _mm_sub_ps(iz3,jz2);
2041             dx33             = _mm_sub_ps(ix3,jx3);
2042             dy33             = _mm_sub_ps(iy3,jy3);
2043             dz33             = _mm_sub_ps(iz3,jz3);
2044
2045             /* Calculate squared distance and things based on it */
2046             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2047             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2048             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2049             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2050             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2051             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2052             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2053             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2054             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2055             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2056
2057             rinv11           = gmx_mm_invsqrt_ps(rsq11);
2058             rinv12           = gmx_mm_invsqrt_ps(rsq12);
2059             rinv13           = gmx_mm_invsqrt_ps(rsq13);
2060             rinv21           = gmx_mm_invsqrt_ps(rsq21);
2061             rinv22           = gmx_mm_invsqrt_ps(rsq22);
2062             rinv23           = gmx_mm_invsqrt_ps(rsq23);
2063             rinv31           = gmx_mm_invsqrt_ps(rsq31);
2064             rinv32           = gmx_mm_invsqrt_ps(rsq32);
2065             rinv33           = gmx_mm_invsqrt_ps(rsq33);
2066
2067             rinvsq00         = gmx_mm_inv_ps(rsq00);
2068
2069             fjx0             = _mm_setzero_ps();
2070             fjy0             = _mm_setzero_ps();
2071             fjz0             = _mm_setzero_ps();
2072             fjx1             = _mm_setzero_ps();
2073             fjy1             = _mm_setzero_ps();
2074             fjz1             = _mm_setzero_ps();
2075             fjx2             = _mm_setzero_ps();
2076             fjy2             = _mm_setzero_ps();
2077             fjz2             = _mm_setzero_ps();
2078             fjx3             = _mm_setzero_ps();
2079             fjy3             = _mm_setzero_ps();
2080             fjz3             = _mm_setzero_ps();
2081
2082             /**************************
2083              * CALCULATE INTERACTIONS *
2084              **************************/
2085
2086             /* LENNARD-JONES DISPERSION/REPULSION */
2087
2088             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2089             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
2090
2091             fscal            = fvdw;
2092
2093             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2094
2095             /* Calculate temporary vectorial force */
2096             tx               = _mm_mul_ps(fscal,dx00);
2097             ty               = _mm_mul_ps(fscal,dy00);
2098             tz               = _mm_mul_ps(fscal,dz00);
2099
2100             /* Update vectorial force */
2101             fix0             = _mm_add_ps(fix0,tx);
2102             fiy0             = _mm_add_ps(fiy0,ty);
2103             fiz0             = _mm_add_ps(fiz0,tz);
2104
2105             fjx0             = _mm_add_ps(fjx0,tx);
2106             fjy0             = _mm_add_ps(fjy0,ty);
2107             fjz0             = _mm_add_ps(fjz0,tz);
2108
2109             /**************************
2110              * CALCULATE INTERACTIONS *
2111              **************************/
2112
2113             r11              = _mm_mul_ps(rsq11,rinv11);
2114             r11              = _mm_andnot_ps(dummy_mask,r11);
2115
2116             /* Calculate table index by multiplying r with table scale and truncate to integer */
2117             rt               = _mm_mul_ps(r11,vftabscale);
2118             vfitab           = _mm_cvttps_epi32(rt);
2119             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2120             vfitab           = _mm_slli_epi32(vfitab,2);
2121
2122             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2123             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2124             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2125             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2126             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2127             _MM_TRANSPOSE4_PS(Y,F,G,H);
2128             Heps             = _mm_mul_ps(vfeps,H);
2129             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2130             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2131             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
2132
2133             fscal            = felec;
2134
2135             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2136
2137             /* Calculate temporary vectorial force */
2138             tx               = _mm_mul_ps(fscal,dx11);
2139             ty               = _mm_mul_ps(fscal,dy11);
2140             tz               = _mm_mul_ps(fscal,dz11);
2141
2142             /* Update vectorial force */
2143             fix1             = _mm_add_ps(fix1,tx);
2144             fiy1             = _mm_add_ps(fiy1,ty);
2145             fiz1             = _mm_add_ps(fiz1,tz);
2146
2147             fjx1             = _mm_add_ps(fjx1,tx);
2148             fjy1             = _mm_add_ps(fjy1,ty);
2149             fjz1             = _mm_add_ps(fjz1,tz);
2150
2151             /**************************
2152              * CALCULATE INTERACTIONS *
2153              **************************/
2154
2155             r12              = _mm_mul_ps(rsq12,rinv12);
2156             r12              = _mm_andnot_ps(dummy_mask,r12);
2157
2158             /* Calculate table index by multiplying r with table scale and truncate to integer */
2159             rt               = _mm_mul_ps(r12,vftabscale);
2160             vfitab           = _mm_cvttps_epi32(rt);
2161             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2162             vfitab           = _mm_slli_epi32(vfitab,2);
2163
2164             /* CUBIC SPLINE TABLE ELECTROSTATICS */
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             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
2174
2175             fscal            = felec;
2176
2177             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2178
2179             /* Calculate temporary vectorial force */
2180             tx               = _mm_mul_ps(fscal,dx12);
2181             ty               = _mm_mul_ps(fscal,dy12);
2182             tz               = _mm_mul_ps(fscal,dz12);
2183
2184             /* Update vectorial force */
2185             fix1             = _mm_add_ps(fix1,tx);
2186             fiy1             = _mm_add_ps(fiy1,ty);
2187             fiz1             = _mm_add_ps(fiz1,tz);
2188
2189             fjx2             = _mm_add_ps(fjx2,tx);
2190             fjy2             = _mm_add_ps(fjy2,ty);
2191             fjz2             = _mm_add_ps(fjz2,tz);
2192
2193             /**************************
2194              * CALCULATE INTERACTIONS *
2195              **************************/
2196
2197             r13              = _mm_mul_ps(rsq13,rinv13);
2198             r13              = _mm_andnot_ps(dummy_mask,r13);
2199
2200             /* Calculate table index by multiplying r with table scale and truncate to integer */
2201             rt               = _mm_mul_ps(r13,vftabscale);
2202             vfitab           = _mm_cvttps_epi32(rt);
2203             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2204             vfitab           = _mm_slli_epi32(vfitab,2);
2205
2206             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2207             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2208             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2209             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2210             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2211             _MM_TRANSPOSE4_PS(Y,F,G,H);
2212             Heps             = _mm_mul_ps(vfeps,H);
2213             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2214             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2215             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
2216
2217             fscal            = felec;
2218
2219             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2220
2221             /* Calculate temporary vectorial force */
2222             tx               = _mm_mul_ps(fscal,dx13);
2223             ty               = _mm_mul_ps(fscal,dy13);
2224             tz               = _mm_mul_ps(fscal,dz13);
2225
2226             /* Update vectorial force */
2227             fix1             = _mm_add_ps(fix1,tx);
2228             fiy1             = _mm_add_ps(fiy1,ty);
2229             fiz1             = _mm_add_ps(fiz1,tz);
2230
2231             fjx3             = _mm_add_ps(fjx3,tx);
2232             fjy3             = _mm_add_ps(fjy3,ty);
2233             fjz3             = _mm_add_ps(fjz3,tz);
2234
2235             /**************************
2236              * CALCULATE INTERACTIONS *
2237              **************************/
2238
2239             r21              = _mm_mul_ps(rsq21,rinv21);
2240             r21              = _mm_andnot_ps(dummy_mask,r21);
2241
2242             /* Calculate table index by multiplying r with table scale and truncate to integer */
2243             rt               = _mm_mul_ps(r21,vftabscale);
2244             vfitab           = _mm_cvttps_epi32(rt);
2245             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2246             vfitab           = _mm_slli_epi32(vfitab,2);
2247
2248             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2249             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2250             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2251             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2252             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2253             _MM_TRANSPOSE4_PS(Y,F,G,H);
2254             Heps             = _mm_mul_ps(vfeps,H);
2255             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2256             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2257             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
2258
2259             fscal            = felec;
2260
2261             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2262
2263             /* Calculate temporary vectorial force */
2264             tx               = _mm_mul_ps(fscal,dx21);
2265             ty               = _mm_mul_ps(fscal,dy21);
2266             tz               = _mm_mul_ps(fscal,dz21);
2267
2268             /* Update vectorial force */
2269             fix2             = _mm_add_ps(fix2,tx);
2270             fiy2             = _mm_add_ps(fiy2,ty);
2271             fiz2             = _mm_add_ps(fiz2,tz);
2272
2273             fjx1             = _mm_add_ps(fjx1,tx);
2274             fjy1             = _mm_add_ps(fjy1,ty);
2275             fjz1             = _mm_add_ps(fjz1,tz);
2276
2277             /**************************
2278              * CALCULATE INTERACTIONS *
2279              **************************/
2280
2281             r22              = _mm_mul_ps(rsq22,rinv22);
2282             r22              = _mm_andnot_ps(dummy_mask,r22);
2283
2284             /* Calculate table index by multiplying r with table scale and truncate to integer */
2285             rt               = _mm_mul_ps(r22,vftabscale);
2286             vfitab           = _mm_cvttps_epi32(rt);
2287             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2288             vfitab           = _mm_slli_epi32(vfitab,2);
2289
2290             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2291             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2292             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2293             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2294             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2295             _MM_TRANSPOSE4_PS(Y,F,G,H);
2296             Heps             = _mm_mul_ps(vfeps,H);
2297             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2298             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2299             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
2300
2301             fscal            = felec;
2302
2303             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2304
2305             /* Calculate temporary vectorial force */
2306             tx               = _mm_mul_ps(fscal,dx22);
2307             ty               = _mm_mul_ps(fscal,dy22);
2308             tz               = _mm_mul_ps(fscal,dz22);
2309
2310             /* Update vectorial force */
2311             fix2             = _mm_add_ps(fix2,tx);
2312             fiy2             = _mm_add_ps(fiy2,ty);
2313             fiz2             = _mm_add_ps(fiz2,tz);
2314
2315             fjx2             = _mm_add_ps(fjx2,tx);
2316             fjy2             = _mm_add_ps(fjy2,ty);
2317             fjz2             = _mm_add_ps(fjz2,tz);
2318
2319             /**************************
2320              * CALCULATE INTERACTIONS *
2321              **************************/
2322
2323             r23              = _mm_mul_ps(rsq23,rinv23);
2324             r23              = _mm_andnot_ps(dummy_mask,r23);
2325
2326             /* Calculate table index by multiplying r with table scale and truncate to integer */
2327             rt               = _mm_mul_ps(r23,vftabscale);
2328             vfitab           = _mm_cvttps_epi32(rt);
2329             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2330             vfitab           = _mm_slli_epi32(vfitab,2);
2331
2332             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2333             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2334             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2335             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2336             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2337             _MM_TRANSPOSE4_PS(Y,F,G,H);
2338             Heps             = _mm_mul_ps(vfeps,H);
2339             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2340             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2341             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
2342
2343             fscal            = felec;
2344
2345             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2346
2347             /* Calculate temporary vectorial force */
2348             tx               = _mm_mul_ps(fscal,dx23);
2349             ty               = _mm_mul_ps(fscal,dy23);
2350             tz               = _mm_mul_ps(fscal,dz23);
2351
2352             /* Update vectorial force */
2353             fix2             = _mm_add_ps(fix2,tx);
2354             fiy2             = _mm_add_ps(fiy2,ty);
2355             fiz2             = _mm_add_ps(fiz2,tz);
2356
2357             fjx3             = _mm_add_ps(fjx3,tx);
2358             fjy3             = _mm_add_ps(fjy3,ty);
2359             fjz3             = _mm_add_ps(fjz3,tz);
2360
2361             /**************************
2362              * CALCULATE INTERACTIONS *
2363              **************************/
2364
2365             r31              = _mm_mul_ps(rsq31,rinv31);
2366             r31              = _mm_andnot_ps(dummy_mask,r31);
2367
2368             /* Calculate table index by multiplying r with table scale and truncate to integer */
2369             rt               = _mm_mul_ps(r31,vftabscale);
2370             vfitab           = _mm_cvttps_epi32(rt);
2371             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2372             vfitab           = _mm_slli_epi32(vfitab,2);
2373
2374             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2375             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2376             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2377             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2378             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2379             _MM_TRANSPOSE4_PS(Y,F,G,H);
2380             Heps             = _mm_mul_ps(vfeps,H);
2381             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2382             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2383             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
2384
2385             fscal            = felec;
2386
2387             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2388
2389             /* Calculate temporary vectorial force */
2390             tx               = _mm_mul_ps(fscal,dx31);
2391             ty               = _mm_mul_ps(fscal,dy31);
2392             tz               = _mm_mul_ps(fscal,dz31);
2393
2394             /* Update vectorial force */
2395             fix3             = _mm_add_ps(fix3,tx);
2396             fiy3             = _mm_add_ps(fiy3,ty);
2397             fiz3             = _mm_add_ps(fiz3,tz);
2398
2399             fjx1             = _mm_add_ps(fjx1,tx);
2400             fjy1             = _mm_add_ps(fjy1,ty);
2401             fjz1             = _mm_add_ps(fjz1,tz);
2402
2403             /**************************
2404              * CALCULATE INTERACTIONS *
2405              **************************/
2406
2407             r32              = _mm_mul_ps(rsq32,rinv32);
2408             r32              = _mm_andnot_ps(dummy_mask,r32);
2409
2410             /* Calculate table index by multiplying r with table scale and truncate to integer */
2411             rt               = _mm_mul_ps(r32,vftabscale);
2412             vfitab           = _mm_cvttps_epi32(rt);
2413             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2414             vfitab           = _mm_slli_epi32(vfitab,2);
2415
2416             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2417             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2418             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2419             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2420             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2421             _MM_TRANSPOSE4_PS(Y,F,G,H);
2422             Heps             = _mm_mul_ps(vfeps,H);
2423             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2424             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2425             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
2426
2427             fscal            = felec;
2428
2429             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2430
2431             /* Calculate temporary vectorial force */
2432             tx               = _mm_mul_ps(fscal,dx32);
2433             ty               = _mm_mul_ps(fscal,dy32);
2434             tz               = _mm_mul_ps(fscal,dz32);
2435
2436             /* Update vectorial force */
2437             fix3             = _mm_add_ps(fix3,tx);
2438             fiy3             = _mm_add_ps(fiy3,ty);
2439             fiz3             = _mm_add_ps(fiz3,tz);
2440
2441             fjx2             = _mm_add_ps(fjx2,tx);
2442             fjy2             = _mm_add_ps(fjy2,ty);
2443             fjz2             = _mm_add_ps(fjz2,tz);
2444
2445             /**************************
2446              * CALCULATE INTERACTIONS *
2447              **************************/
2448
2449             r33              = _mm_mul_ps(rsq33,rinv33);
2450             r33              = _mm_andnot_ps(dummy_mask,r33);
2451
2452             /* Calculate table index by multiplying r with table scale and truncate to integer */
2453             rt               = _mm_mul_ps(r33,vftabscale);
2454             vfitab           = _mm_cvttps_epi32(rt);
2455             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
2456             vfitab           = _mm_slli_epi32(vfitab,2);
2457
2458             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2459             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2460             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2461             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2462             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2463             _MM_TRANSPOSE4_PS(Y,F,G,H);
2464             Heps             = _mm_mul_ps(vfeps,H);
2465             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2466             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2467             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2468
2469             fscal            = felec;
2470
2471             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2472
2473             /* Calculate temporary vectorial force */
2474             tx               = _mm_mul_ps(fscal,dx33);
2475             ty               = _mm_mul_ps(fscal,dy33);
2476             tz               = _mm_mul_ps(fscal,dz33);
2477
2478             /* Update vectorial force */
2479             fix3             = _mm_add_ps(fix3,tx);
2480             fiy3             = _mm_add_ps(fiy3,ty);
2481             fiz3             = _mm_add_ps(fiz3,tz);
2482
2483             fjx3             = _mm_add_ps(fjx3,tx);
2484             fjy3             = _mm_add_ps(fjy3,ty);
2485             fjz3             = _mm_add_ps(fjz3,tz);
2486
2487             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2488             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2489             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2490             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2491
2492             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2493                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2494                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2495
2496             /* Inner loop uses 390 flops */
2497         }
2498
2499         /* End of innermost loop */
2500
2501         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2502                                               f+i_coord_offset,fshift+i_shift_offset);
2503
2504         /* Increment number of inner iterations */
2505         inneriter                  += j_index_end - j_index_start;
2506
2507         /* Outer loop uses 24 flops */
2508     }
2509
2510     /* Increment number of outer iterations */
2511     outeriter        += nri;
2512
2513     /* Update outer/inner flops */
2514
2515     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*390);
2516 }