src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_sse4_1_double.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_double 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_double.h"
  48 #include "kernelutil_x86_sse4_1_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_VF_sse4_1_double
  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_double
  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 refer to j loop unrolling done with SSE double precision, e.g. for the two 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;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwioffset3;
  87     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  88     int              vdwjidx0A,vdwjidx0B;
  89     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  90     int              vdwjidx1A,vdwjidx1B;
  91     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  92     int              vdwjidx2A,vdwjidx2B;
  93     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  94     int              vdwjidx3A,vdwjidx3B;
  95     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  96     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  97     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  98     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  99     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 100     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 101     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 102     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 103     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 104     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 105     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 106     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     int              nvdwtype;
 109     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 110     int              *vdwtype;
 111     real             *vdwparam;
 112     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 113     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 114     __m128i          vfitab;
 115     __m128i          ifour       = _mm_set1_epi32(4);
 116     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 117     real             *vftab;
 118     __m128d          dummy_mask,cutoff_mask;
 119     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 120     __m128d          one     = _mm_set1_pd(1.0);
 121     __m128d          two     = _mm_set1_pd(2.0);
 122     x                = xx[0];
 123     f                = ff[0];
 124
 125     nri              = nlist->nri;
 126     iinr             = nlist->iinr;
 127     jindex           = nlist->jindex;
 128     jjnr             = nlist->jjnr;
 129     shiftidx         = nlist->shift;
 130     gid              = nlist->gid;
 131     shiftvec         = fr->shift_vec[0];
 132     fshift           = fr->fshift[0];
 133     facel            = _mm_set1_pd(fr->epsfac);
 134     charge           = mdatoms->chargeA;
 135     nvdwtype         = fr->ntype;
 136     vdwparam         = fr->nbfp;
 137     vdwtype          = mdatoms->typeA;
 138
 139     vftab            = kernel_data->table_elec->data;
 140     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
 141
 142     /* Setup water-specific parameters */
 143     inr              = nlist->iinr[0];
 144     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 145     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 146     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 147     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 148
 149     jq1              = _mm_set1_pd(charge[inr+1]);
 150     jq2              = _mm_set1_pd(charge[inr+2]);
 151     jq3              = _mm_set1_pd(charge[inr+3]);
 152     vdwjidx0A        = 2*vdwtype[inr+0];
 153     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 154     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 155     qq11             = _mm_mul_pd(iq1,jq1);
 156     qq12             = _mm_mul_pd(iq1,jq2);
 157     qq13             = _mm_mul_pd(iq1,jq3);
 158     qq21             = _mm_mul_pd(iq2,jq1);
 159     qq22             = _mm_mul_pd(iq2,jq2);
 160     qq23             = _mm_mul_pd(iq2,jq3);
 161     qq31             = _mm_mul_pd(iq3,jq1);
 162     qq32             = _mm_mul_pd(iq3,jq2);
 163     qq33             = _mm_mul_pd(iq3,jq3);
 164
 165     /* Avoid stupid compiler warnings */
 166     jnrA = jnrB = 0;
 167     j_coord_offsetA = 0;
 168     j_coord_offsetB = 0;
 169
 170     outeriter        = 0;
 171     inneriter        = 0;
 172
 173     /* Start outer loop over neighborlists */
 174     for(iidx=0; iidx<nri; iidx++)
 175     {
 176         /* Load shift vector for this list */
 177         i_shift_offset   = DIM*shiftidx[iidx];
 178
 179         /* Load limits for loop over neighbors */
 180         j_index_start    = jindex[iidx];
 181         j_index_end      = jindex[iidx+1];
 182
 183         /* Get outer coordinate index */
 184         inr              = iinr[iidx];
 185         i_coord_offset   = DIM*inr;
 186
 187         /* Load i particle coords and add shift vector */
 188         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 189                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 190
 191         fix0             = _mm_setzero_pd();
 192         fiy0             = _mm_setzero_pd();
 193         fiz0             = _mm_setzero_pd();
 194         fix1             = _mm_setzero_pd();
 195         fiy1             = _mm_setzero_pd();
 196         fiz1             = _mm_setzero_pd();
 197         fix2             = _mm_setzero_pd();
 198         fiy2             = _mm_setzero_pd();
 199         fiz2             = _mm_setzero_pd();
 200         fix3             = _mm_setzero_pd();
 201         fiy3             = _mm_setzero_pd();
 202         fiz3             = _mm_setzero_pd();
 203
 204         /* Reset potential sums */
 205         velecsum         = _mm_setzero_pd();
 206         vvdwsum          = _mm_setzero_pd();
 207
 208         /* Start inner kernel loop */
 209         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 210         {
 211
 212             /* Get j neighbor index, and coordinate index */
 213             jnrA             = jjnr[jidx];
 214             jnrB             = jjnr[jidx+1];
 215             j_coord_offsetA  = DIM*jnrA;
 216             j_coord_offsetB  = DIM*jnrB;
 217
 218             /* load j atom coordinates */
 219             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 220                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 221                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 222
 223             /* Calculate displacement vector */
 224             dx00             = _mm_sub_pd(ix0,jx0);
 225             dy00             = _mm_sub_pd(iy0,jy0);
 226             dz00             = _mm_sub_pd(iz0,jz0);
 227             dx11             = _mm_sub_pd(ix1,jx1);
 228             dy11             = _mm_sub_pd(iy1,jy1);
 229             dz11             = _mm_sub_pd(iz1,jz1);
 230             dx12             = _mm_sub_pd(ix1,jx2);
 231             dy12             = _mm_sub_pd(iy1,jy2);
 232             dz12             = _mm_sub_pd(iz1,jz2);
 233             dx13             = _mm_sub_pd(ix1,jx3);
 234             dy13             = _mm_sub_pd(iy1,jy3);
 235             dz13             = _mm_sub_pd(iz1,jz3);
 236             dx21             = _mm_sub_pd(ix2,jx1);
 237             dy21             = _mm_sub_pd(iy2,jy1);
 238             dz21             = _mm_sub_pd(iz2,jz1);
 239             dx22             = _mm_sub_pd(ix2,jx2);
 240             dy22             = _mm_sub_pd(iy2,jy2);
 241             dz22             = _mm_sub_pd(iz2,jz2);
 242             dx23             = _mm_sub_pd(ix2,jx3);
 243             dy23             = _mm_sub_pd(iy2,jy3);
 244             dz23             = _mm_sub_pd(iz2,jz3);
 245             dx31             = _mm_sub_pd(ix3,jx1);
 246             dy31             = _mm_sub_pd(iy3,jy1);
 247             dz31             = _mm_sub_pd(iz3,jz1);
 248             dx32             = _mm_sub_pd(ix3,jx2);
 249             dy32             = _mm_sub_pd(iy3,jy2);
 250             dz32             = _mm_sub_pd(iz3,jz2);
 251             dx33             = _mm_sub_pd(ix3,jx3);
 252             dy33             = _mm_sub_pd(iy3,jy3);
 253             dz33             = _mm_sub_pd(iz3,jz3);
 254
 255             /* Calculate squared distance and things based on it */
 256             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 257             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 258             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 259             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 260             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 261             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 262             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 263             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 264             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 265             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 266
 267             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 268             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 269             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 270             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 271             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 272             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 273             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 274             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 275             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 276
 277             rinvsq00         = gmx_mm_inv_pd(rsq00);
 278
 279             fjx0             = _mm_setzero_pd();
 280             fjy0             = _mm_setzero_pd();
 281             fjz0             = _mm_setzero_pd();
 282             fjx1             = _mm_setzero_pd();
 283             fjy1             = _mm_setzero_pd();
 284             fjz1             = _mm_setzero_pd();
 285             fjx2             = _mm_setzero_pd();
 286             fjy2             = _mm_setzero_pd();
 287             fjz2             = _mm_setzero_pd();
 288             fjx3             = _mm_setzero_pd();
 289             fjy3             = _mm_setzero_pd();
 290             fjz3             = _mm_setzero_pd();
 291
 292             /**************************
 293              * CALCULATE INTERACTIONS *
 294              **************************/
 295
 296             /* LENNARD-JONES DISPERSION/REPULSION */
 297
 298             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 299             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 300             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 301             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 302             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 303
 304             /* Update potential sum for this i atom from the interaction with this j atom. */
 305             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 306
 307             fscal            = fvdw;
 308
 309             /* Calculate temporary vectorial force */
 310             tx               = _mm_mul_pd(fscal,dx00);
 311             ty               = _mm_mul_pd(fscal,dy00);
 312             tz               = _mm_mul_pd(fscal,dz00);
 313
 314             /* Update vectorial force */
 315             fix0             = _mm_add_pd(fix0,tx);
 316             fiy0             = _mm_add_pd(fiy0,ty);
 317             fiz0             = _mm_add_pd(fiz0,tz);
 318
 319             fjx0             = _mm_add_pd(fjx0,tx);
 320             fjy0             = _mm_add_pd(fjy0,ty);
 321             fjz0             = _mm_add_pd(fjz0,tz);
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             r11              = _mm_mul_pd(rsq11,rinv11);
 328
 329             /* Calculate table index by multiplying r with table scale and truncate to integer */
 330             rt               = _mm_mul_pd(r11,vftabscale);
 331             vfitab           = _mm_cvttpd_epi32(rt);
 332             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 333             vfitab           = _mm_slli_epi32(vfitab,2);
 334
 335             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 336             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 337             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 338             GMX_MM_TRANSPOSE2_PD(Y,F);
 339             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 340             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 341             GMX_MM_TRANSPOSE2_PD(G,H);
 342             Heps             = _mm_mul_pd(vfeps,H);
 343             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 344             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 345             velec            = _mm_mul_pd(qq11,VV);
 346             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 347             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 348
 349             /* Update potential sum for this i atom from the interaction with this j atom. */
 350             velecsum         = _mm_add_pd(velecsum,velec);
 351
 352             fscal            = felec;
 353
 354             /* Calculate temporary vectorial force */
 355             tx               = _mm_mul_pd(fscal,dx11);
 356             ty               = _mm_mul_pd(fscal,dy11);
 357             tz               = _mm_mul_pd(fscal,dz11);
 358
 359             /* Update vectorial force */
 360             fix1             = _mm_add_pd(fix1,tx);
 361             fiy1             = _mm_add_pd(fiy1,ty);
 362             fiz1             = _mm_add_pd(fiz1,tz);
 363
 364             fjx1             = _mm_add_pd(fjx1,tx);
 365             fjy1             = _mm_add_pd(fjy1,ty);
 366             fjz1             = _mm_add_pd(fjz1,tz);
 367
 368             /**************************
 369              * CALCULATE INTERACTIONS *
 370              **************************/
 371
 372             r12              = _mm_mul_pd(rsq12,rinv12);
 373
 374             /* Calculate table index by multiplying r with table scale and truncate to integer */
 375             rt               = _mm_mul_pd(r12,vftabscale);
 376             vfitab           = _mm_cvttpd_epi32(rt);
 377             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 378             vfitab           = _mm_slli_epi32(vfitab,2);
 379
 380             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 381             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 382             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 383             GMX_MM_TRANSPOSE2_PD(Y,F);
 384             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 385             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 386             GMX_MM_TRANSPOSE2_PD(G,H);
 387             Heps             = _mm_mul_pd(vfeps,H);
 388             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 389             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 390             velec            = _mm_mul_pd(qq12,VV);
 391             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 392             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 393
 394             /* Update potential sum for this i atom from the interaction with this j atom. */
 395             velecsum         = _mm_add_pd(velecsum,velec);
 396
 397             fscal            = felec;
 398
 399             /* Calculate temporary vectorial force */
 400             tx               = _mm_mul_pd(fscal,dx12);
 401             ty               = _mm_mul_pd(fscal,dy12);
 402             tz               = _mm_mul_pd(fscal,dz12);
 403
 404             /* Update vectorial force */
 405             fix1             = _mm_add_pd(fix1,tx);
 406             fiy1             = _mm_add_pd(fiy1,ty);
 407             fiz1             = _mm_add_pd(fiz1,tz);
 408
 409             fjx2             = _mm_add_pd(fjx2,tx);
 410             fjy2             = _mm_add_pd(fjy2,ty);
 411             fjz2             = _mm_add_pd(fjz2,tz);
 412
 413             /**************************
 414              * CALCULATE INTERACTIONS *
 415              **************************/
 416
 417             r13              = _mm_mul_pd(rsq13,rinv13);
 418
 419             /* Calculate table index by multiplying r with table scale and truncate to integer */
 420             rt               = _mm_mul_pd(r13,vftabscale);
 421             vfitab           = _mm_cvttpd_epi32(rt);
 422             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 423             vfitab           = _mm_slli_epi32(vfitab,2);
 424
 425             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 426             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 427             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 428             GMX_MM_TRANSPOSE2_PD(Y,F);
 429             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 430             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 431             GMX_MM_TRANSPOSE2_PD(G,H);
 432             Heps             = _mm_mul_pd(vfeps,H);
 433             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 434             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 435             velec            = _mm_mul_pd(qq13,VV);
 436             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 437             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
 438
 439             /* Update potential sum for this i atom from the interaction with this j atom. */
 440             velecsum         = _mm_add_pd(velecsum,velec);
 441
 442             fscal            = felec;
 443
 444             /* Calculate temporary vectorial force */
 445             tx               = _mm_mul_pd(fscal,dx13);
 446             ty               = _mm_mul_pd(fscal,dy13);
 447             tz               = _mm_mul_pd(fscal,dz13);
 448
 449             /* Update vectorial force */
 450             fix1             = _mm_add_pd(fix1,tx);
 451             fiy1             = _mm_add_pd(fiy1,ty);
 452             fiz1             = _mm_add_pd(fiz1,tz);
 453
 454             fjx3             = _mm_add_pd(fjx3,tx);
 455             fjy3             = _mm_add_pd(fjy3,ty);
 456             fjz3             = _mm_add_pd(fjz3,tz);
 457
 458             /**************************
 459              * CALCULATE INTERACTIONS *
 460              **************************/
 461
 462             r21              = _mm_mul_pd(rsq21,rinv21);
 463
 464             /* Calculate table index by multiplying r with table scale and truncate to integer */
 465             rt               = _mm_mul_pd(r21,vftabscale);
 466             vfitab           = _mm_cvttpd_epi32(rt);
 467             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 468             vfitab           = _mm_slli_epi32(vfitab,2);
 469
 470             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 471             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 472             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 473             GMX_MM_TRANSPOSE2_PD(Y,F);
 474             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 475             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 476             GMX_MM_TRANSPOSE2_PD(G,H);
 477             Heps             = _mm_mul_pd(vfeps,H);
 478             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 479             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 480             velec            = _mm_mul_pd(qq21,VV);
 481             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 482             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 483
 484             /* Update potential sum for this i atom from the interaction with this j atom. */
 485             velecsum         = _mm_add_pd(velecsum,velec);
 486
 487             fscal            = felec;
 488
 489             /* Calculate temporary vectorial force */
 490             tx               = _mm_mul_pd(fscal,dx21);
 491             ty               = _mm_mul_pd(fscal,dy21);
 492             tz               = _mm_mul_pd(fscal,dz21);
 493
 494             /* Update vectorial force */
 495             fix2             = _mm_add_pd(fix2,tx);
 496             fiy2             = _mm_add_pd(fiy2,ty);
 497             fiz2             = _mm_add_pd(fiz2,tz);
 498
 499             fjx1             = _mm_add_pd(fjx1,tx);
 500             fjy1             = _mm_add_pd(fjy1,ty);
 501             fjz1             = _mm_add_pd(fjz1,tz);
 502
 503             /**************************
 504              * CALCULATE INTERACTIONS *
 505              **************************/
 506
 507             r22              = _mm_mul_pd(rsq22,rinv22);
 508
 509             /* Calculate table index by multiplying r with table scale and truncate to integer */
 510             rt               = _mm_mul_pd(r22,vftabscale);
 511             vfitab           = _mm_cvttpd_epi32(rt);
 512             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 513             vfitab           = _mm_slli_epi32(vfitab,2);
 514
 515             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 516             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 517             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 518             GMX_MM_TRANSPOSE2_PD(Y,F);
 519             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 520             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 521             GMX_MM_TRANSPOSE2_PD(G,H);
 522             Heps             = _mm_mul_pd(vfeps,H);
 523             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 524             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 525             velec            = _mm_mul_pd(qq22,VV);
 526             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 527             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 528
 529             /* Update potential sum for this i atom from the interaction with this j atom. */
 530             velecsum         = _mm_add_pd(velecsum,velec);
 531
 532             fscal            = felec;
 533
 534             /* Calculate temporary vectorial force */
 535             tx               = _mm_mul_pd(fscal,dx22);
 536             ty               = _mm_mul_pd(fscal,dy22);
 537             tz               = _mm_mul_pd(fscal,dz22);
 538
 539             /* Update vectorial force */
 540             fix2             = _mm_add_pd(fix2,tx);
 541             fiy2             = _mm_add_pd(fiy2,ty);
 542             fiz2             = _mm_add_pd(fiz2,tz);
 543
 544             fjx2             = _mm_add_pd(fjx2,tx);
 545             fjy2             = _mm_add_pd(fjy2,ty);
 546             fjz2             = _mm_add_pd(fjz2,tz);
 547
 548             /**************************
 549              * CALCULATE INTERACTIONS *
 550              **************************/
 551
 552             r23              = _mm_mul_pd(rsq23,rinv23);
 553
 554             /* Calculate table index by multiplying r with table scale and truncate to integer */
 555             rt               = _mm_mul_pd(r23,vftabscale);
 556             vfitab           = _mm_cvttpd_epi32(rt);
 557             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 558             vfitab           = _mm_slli_epi32(vfitab,2);
 559
 560             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 561             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 562             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 563             GMX_MM_TRANSPOSE2_PD(Y,F);
 564             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 565             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 566             GMX_MM_TRANSPOSE2_PD(G,H);
 567             Heps             = _mm_mul_pd(vfeps,H);
 568             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 569             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 570             velec            = _mm_mul_pd(qq23,VV);
 571             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 572             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
 573
 574             /* Update potential sum for this i atom from the interaction with this j atom. */
 575             velecsum         = _mm_add_pd(velecsum,velec);
 576
 577             fscal            = felec;
 578
 579             /* Calculate temporary vectorial force */
 580             tx               = _mm_mul_pd(fscal,dx23);
 581             ty               = _mm_mul_pd(fscal,dy23);
 582             tz               = _mm_mul_pd(fscal,dz23);
 583
 584             /* Update vectorial force */
 585             fix2             = _mm_add_pd(fix2,tx);
 586             fiy2             = _mm_add_pd(fiy2,ty);
 587             fiz2             = _mm_add_pd(fiz2,tz);
 588
 589             fjx3             = _mm_add_pd(fjx3,tx);
 590             fjy3             = _mm_add_pd(fjy3,ty);
 591             fjz3             = _mm_add_pd(fjz3,tz);
 592
 593             /**************************
 594              * CALCULATE INTERACTIONS *
 595              **************************/
 596
 597             r31              = _mm_mul_pd(rsq31,rinv31);
 598
 599             /* Calculate table index by multiplying r with table scale and truncate to integer */
 600             rt               = _mm_mul_pd(r31,vftabscale);
 601             vfitab           = _mm_cvttpd_epi32(rt);
 602             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 603             vfitab           = _mm_slli_epi32(vfitab,2);
 604
 605             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 606             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 607             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 608             GMX_MM_TRANSPOSE2_PD(Y,F);
 609             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 610             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 611             GMX_MM_TRANSPOSE2_PD(G,H);
 612             Heps             = _mm_mul_pd(vfeps,H);
 613             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 614             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 615             velec            = _mm_mul_pd(qq31,VV);
 616             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 617             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
 618
 619             /* Update potential sum for this i atom from the interaction with this j atom. */
 620             velecsum         = _mm_add_pd(velecsum,velec);
 621
 622             fscal            = felec;
 623
 624             /* Calculate temporary vectorial force */
 625             tx               = _mm_mul_pd(fscal,dx31);
 626             ty               = _mm_mul_pd(fscal,dy31);
 627             tz               = _mm_mul_pd(fscal,dz31);
 628
 629             /* Update vectorial force */
 630             fix3             = _mm_add_pd(fix3,tx);
 631             fiy3             = _mm_add_pd(fiy3,ty);
 632             fiz3             = _mm_add_pd(fiz3,tz);
 633
 634             fjx1             = _mm_add_pd(fjx1,tx);
 635             fjy1             = _mm_add_pd(fjy1,ty);
 636             fjz1             = _mm_add_pd(fjz1,tz);
 637
 638             /**************************
 639              * CALCULATE INTERACTIONS *
 640              **************************/
 641
 642             r32              = _mm_mul_pd(rsq32,rinv32);
 643
 644             /* Calculate table index by multiplying r with table scale and truncate to integer */
 645             rt               = _mm_mul_pd(r32,vftabscale);
 646             vfitab           = _mm_cvttpd_epi32(rt);
 647             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 648             vfitab           = _mm_slli_epi32(vfitab,2);
 649
 650             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 651             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 652             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 653             GMX_MM_TRANSPOSE2_PD(Y,F);
 654             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 655             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 656             GMX_MM_TRANSPOSE2_PD(G,H);
 657             Heps             = _mm_mul_pd(vfeps,H);
 658             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 659             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 660             velec            = _mm_mul_pd(qq32,VV);
 661             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 662             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
 663
 664             /* Update potential sum for this i atom from the interaction with this j atom. */
 665             velecsum         = _mm_add_pd(velecsum,velec);
 666
 667             fscal            = felec;
 668
 669             /* Calculate temporary vectorial force */
 670             tx               = _mm_mul_pd(fscal,dx32);
 671             ty               = _mm_mul_pd(fscal,dy32);
 672             tz               = _mm_mul_pd(fscal,dz32);
 673
 674             /* Update vectorial force */
 675             fix3             = _mm_add_pd(fix3,tx);
 676             fiy3             = _mm_add_pd(fiy3,ty);
 677             fiz3             = _mm_add_pd(fiz3,tz);
 678
 679             fjx2             = _mm_add_pd(fjx2,tx);
 680             fjy2             = _mm_add_pd(fjy2,ty);
 681             fjz2             = _mm_add_pd(fjz2,tz);
 682
 683             /**************************
 684              * CALCULATE INTERACTIONS *
 685              **************************/
 686
 687             r33              = _mm_mul_pd(rsq33,rinv33);
 688
 689             /* Calculate table index by multiplying r with table scale and truncate to integer */
 690             rt               = _mm_mul_pd(r33,vftabscale);
 691             vfitab           = _mm_cvttpd_epi32(rt);
 692             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 693             vfitab           = _mm_slli_epi32(vfitab,2);
 694
 695             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 696             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 697             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 698             GMX_MM_TRANSPOSE2_PD(Y,F);
 699             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 700             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 701             GMX_MM_TRANSPOSE2_PD(G,H);
 702             Heps             = _mm_mul_pd(vfeps,H);
 703             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 704             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 705             velec            = _mm_mul_pd(qq33,VV);
 706             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 707             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
 708
 709             /* Update potential sum for this i atom from the interaction with this j atom. */
 710             velecsum         = _mm_add_pd(velecsum,velec);
 711
 712             fscal            = felec;
 713
 714             /* Calculate temporary vectorial force */
 715             tx               = _mm_mul_pd(fscal,dx33);
 716             ty               = _mm_mul_pd(fscal,dy33);
 717             tz               = _mm_mul_pd(fscal,dz33);
 718
 719             /* Update vectorial force */
 720             fix3             = _mm_add_pd(fix3,tx);
 721             fiy3             = _mm_add_pd(fiy3,ty);
 722             fiz3             = _mm_add_pd(fiz3,tz);
 723
 724             fjx3             = _mm_add_pd(fjx3,tx);
 725             fjy3             = _mm_add_pd(fjy3,ty);
 726             fjz3             = _mm_add_pd(fjz3,tz);
 727
 728             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 729
 730             /* Inner loop uses 422 flops */
 731         }
 732
 733         if(jidx<j_index_end)
 734         {
 735
 736             jnrA             = jjnr[jidx];
 737             j_coord_offsetA  = DIM*jnrA;
 738
 739             /* load j atom coordinates */
 740             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 741                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 742                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 743
 744             /* Calculate displacement vector */
 745             dx00             = _mm_sub_pd(ix0,jx0);
 746             dy00             = _mm_sub_pd(iy0,jy0);
 747             dz00             = _mm_sub_pd(iz0,jz0);
 748             dx11             = _mm_sub_pd(ix1,jx1);
 749             dy11             = _mm_sub_pd(iy1,jy1);
 750             dz11             = _mm_sub_pd(iz1,jz1);
 751             dx12             = _mm_sub_pd(ix1,jx2);
 752             dy12             = _mm_sub_pd(iy1,jy2);
 753             dz12             = _mm_sub_pd(iz1,jz2);
 754             dx13             = _mm_sub_pd(ix1,jx3);
 755             dy13             = _mm_sub_pd(iy1,jy3);
 756             dz13             = _mm_sub_pd(iz1,jz3);
 757             dx21             = _mm_sub_pd(ix2,jx1);
 758             dy21             = _mm_sub_pd(iy2,jy1);
 759             dz21             = _mm_sub_pd(iz2,jz1);
 760             dx22             = _mm_sub_pd(ix2,jx2);
 761             dy22             = _mm_sub_pd(iy2,jy2);
 762             dz22             = _mm_sub_pd(iz2,jz2);
 763             dx23             = _mm_sub_pd(ix2,jx3);
 764             dy23             = _mm_sub_pd(iy2,jy3);
 765             dz23             = _mm_sub_pd(iz2,jz3);
 766             dx31             = _mm_sub_pd(ix3,jx1);
 767             dy31             = _mm_sub_pd(iy3,jy1);
 768             dz31             = _mm_sub_pd(iz3,jz1);
 769             dx32             = _mm_sub_pd(ix3,jx2);
 770             dy32             = _mm_sub_pd(iy3,jy2);
 771             dz32             = _mm_sub_pd(iz3,jz2);
 772             dx33             = _mm_sub_pd(ix3,jx3);
 773             dy33             = _mm_sub_pd(iy3,jy3);
 774             dz33             = _mm_sub_pd(iz3,jz3);
 775
 776             /* Calculate squared distance and things based on it */
 777             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 778             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 779             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 780             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 781             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 782             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 783             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 784             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 785             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 786             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 787
 788             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 789             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 790             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 791             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 792             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 793             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 794             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 795             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 796             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 797
 798             rinvsq00         = gmx_mm_inv_pd(rsq00);
 799
 800             fjx0             = _mm_setzero_pd();
 801             fjy0             = _mm_setzero_pd();
 802             fjz0             = _mm_setzero_pd();
 803             fjx1             = _mm_setzero_pd();
 804             fjy1             = _mm_setzero_pd();
 805             fjz1             = _mm_setzero_pd();
 806             fjx2             = _mm_setzero_pd();
 807             fjy2             = _mm_setzero_pd();
 808             fjz2             = _mm_setzero_pd();
 809             fjx3             = _mm_setzero_pd();
 810             fjy3             = _mm_setzero_pd();
 811             fjz3             = _mm_setzero_pd();
 812
 813             /**************************
 814              * CALCULATE INTERACTIONS *
 815              **************************/
 816
 817             /* LENNARD-JONES DISPERSION/REPULSION */
 818
 819             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 820             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 821             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 822             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 823             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 824
 825             /* Update potential sum for this i atom from the interaction with this j atom. */
 826             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 827             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 828
 829             fscal            = fvdw;
 830
 831             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 832
 833             /* Calculate temporary vectorial force */
 834             tx               = _mm_mul_pd(fscal,dx00);
 835             ty               = _mm_mul_pd(fscal,dy00);
 836             tz               = _mm_mul_pd(fscal,dz00);
 837
 838             /* Update vectorial force */
 839             fix0             = _mm_add_pd(fix0,tx);
 840             fiy0             = _mm_add_pd(fiy0,ty);
 841             fiz0             = _mm_add_pd(fiz0,tz);
 842
 843             fjx0             = _mm_add_pd(fjx0,tx);
 844             fjy0             = _mm_add_pd(fjy0,ty);
 845             fjz0             = _mm_add_pd(fjz0,tz);
 846
 847             /**************************
 848              * CALCULATE INTERACTIONS *
 849              **************************/
 850
 851             r11              = _mm_mul_pd(rsq11,rinv11);
 852
 853             /* Calculate table index by multiplying r with table scale and truncate to integer */
 854             rt               = _mm_mul_pd(r11,vftabscale);
 855             vfitab           = _mm_cvttpd_epi32(rt);
 856             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 857             vfitab           = _mm_slli_epi32(vfitab,2);
 858
 859             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 860             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 861             F                = _mm_setzero_pd();
 862             GMX_MM_TRANSPOSE2_PD(Y,F);
 863             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 864             H                = _mm_setzero_pd();
 865             GMX_MM_TRANSPOSE2_PD(G,H);
 866             Heps             = _mm_mul_pd(vfeps,H);
 867             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 868             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 869             velec            = _mm_mul_pd(qq11,VV);
 870             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 871             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 872
 873             /* Update potential sum for this i atom from the interaction with this j atom. */
 874             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 875             velecsum         = _mm_add_pd(velecsum,velec);
 876
 877             fscal            = felec;
 878
 879             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 880
 881             /* Calculate temporary vectorial force */
 882             tx               = _mm_mul_pd(fscal,dx11);
 883             ty               = _mm_mul_pd(fscal,dy11);
 884             tz               = _mm_mul_pd(fscal,dz11);
 885
 886             /* Update vectorial force */
 887             fix1             = _mm_add_pd(fix1,tx);
 888             fiy1             = _mm_add_pd(fiy1,ty);
 889             fiz1             = _mm_add_pd(fiz1,tz);
 890
 891             fjx1             = _mm_add_pd(fjx1,tx);
 892             fjy1             = _mm_add_pd(fjy1,ty);
 893             fjz1             = _mm_add_pd(fjz1,tz);
 894
 895             /**************************
 896              * CALCULATE INTERACTIONS *
 897              **************************/
 898
 899             r12              = _mm_mul_pd(rsq12,rinv12);
 900
 901             /* Calculate table index by multiplying r with table scale and truncate to integer */
 902             rt               = _mm_mul_pd(r12,vftabscale);
 903             vfitab           = _mm_cvttpd_epi32(rt);
 904             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 905             vfitab           = _mm_slli_epi32(vfitab,2);
 906
 907             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 908             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 909             F                = _mm_setzero_pd();
 910             GMX_MM_TRANSPOSE2_PD(Y,F);
 911             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 912             H                = _mm_setzero_pd();
 913             GMX_MM_TRANSPOSE2_PD(G,H);
 914             Heps             = _mm_mul_pd(vfeps,H);
 915             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 916             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 917             velec            = _mm_mul_pd(qq12,VV);
 918             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 919             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 920
 921             /* Update potential sum for this i atom from the interaction with this j atom. */
 922             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 923             velecsum         = _mm_add_pd(velecsum,velec);
 924
 925             fscal            = felec;
 926
 927             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 928
 929             /* Calculate temporary vectorial force */
 930             tx               = _mm_mul_pd(fscal,dx12);
 931             ty               = _mm_mul_pd(fscal,dy12);
 932             tz               = _mm_mul_pd(fscal,dz12);
 933
 934             /* Update vectorial force */
 935             fix1             = _mm_add_pd(fix1,tx);
 936             fiy1             = _mm_add_pd(fiy1,ty);
 937             fiz1             = _mm_add_pd(fiz1,tz);
 938
 939             fjx2             = _mm_add_pd(fjx2,tx);
 940             fjy2             = _mm_add_pd(fjy2,ty);
 941             fjz2             = _mm_add_pd(fjz2,tz);
 942
 943             /**************************
 944              * CALCULATE INTERACTIONS *
 945              **************************/
 946
 947             r13              = _mm_mul_pd(rsq13,rinv13);
 948
 949             /* Calculate table index by multiplying r with table scale and truncate to integer */
 950             rt               = _mm_mul_pd(r13,vftabscale);
 951             vfitab           = _mm_cvttpd_epi32(rt);
 952             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 953             vfitab           = _mm_slli_epi32(vfitab,2);
 954
 955             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 956             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 957             F                = _mm_setzero_pd();
 958             GMX_MM_TRANSPOSE2_PD(Y,F);
 959             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 960             H                = _mm_setzero_pd();
 961             GMX_MM_TRANSPOSE2_PD(G,H);
 962             Heps             = _mm_mul_pd(vfeps,H);
 963             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 964             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 965             velec            = _mm_mul_pd(qq13,VV);
 966             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 967             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
 968
 969             /* Update potential sum for this i atom from the interaction with this j atom. */
 970             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 971             velecsum         = _mm_add_pd(velecsum,velec);
 972
 973             fscal            = felec;
 974
 975             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 976
 977             /* Calculate temporary vectorial force */
 978             tx               = _mm_mul_pd(fscal,dx13);
 979             ty               = _mm_mul_pd(fscal,dy13);
 980             tz               = _mm_mul_pd(fscal,dz13);
 981
 982             /* Update vectorial force */
 983             fix1             = _mm_add_pd(fix1,tx);
 984             fiy1             = _mm_add_pd(fiy1,ty);
 985             fiz1             = _mm_add_pd(fiz1,tz);
 986
 987             fjx3             = _mm_add_pd(fjx3,tx);
 988             fjy3             = _mm_add_pd(fjy3,ty);
 989             fjz3             = _mm_add_pd(fjz3,tz);
 990
 991             /**************************
 992              * CALCULATE INTERACTIONS *
 993              **************************/
 994
 995             r21              = _mm_mul_pd(rsq21,rinv21);
 996
 997             /* Calculate table index by multiplying r with table scale and truncate to integer */
 998             rt               = _mm_mul_pd(r21,vftabscale);
 999             vfitab           = _mm_cvttpd_epi32(rt);
1000             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1001             vfitab           = _mm_slli_epi32(vfitab,2);
1002
1003             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1004             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1005             F                = _mm_setzero_pd();
1006             GMX_MM_TRANSPOSE2_PD(Y,F);
1007             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1008             H                = _mm_setzero_pd();
1009             GMX_MM_TRANSPOSE2_PD(G,H);
1010             Heps             = _mm_mul_pd(vfeps,H);
1011             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1012             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1013             velec            = _mm_mul_pd(qq21,VV);
1014             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1015             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1016
1017             /* Update potential sum for this i atom from the interaction with this j atom. */
1018             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1019             velecsum         = _mm_add_pd(velecsum,velec);
1020
1021             fscal            = felec;
1022
1023             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1024
1025             /* Calculate temporary vectorial force */
1026             tx               = _mm_mul_pd(fscal,dx21);
1027             ty               = _mm_mul_pd(fscal,dy21);
1028             tz               = _mm_mul_pd(fscal,dz21);
1029
1030             /* Update vectorial force */
1031             fix2             = _mm_add_pd(fix2,tx);
1032             fiy2             = _mm_add_pd(fiy2,ty);
1033             fiz2             = _mm_add_pd(fiz2,tz);
1034
1035             fjx1             = _mm_add_pd(fjx1,tx);
1036             fjy1             = _mm_add_pd(fjy1,ty);
1037             fjz1             = _mm_add_pd(fjz1,tz);
1038
1039             /**************************
1040              * CALCULATE INTERACTIONS *
1041              **************************/
1042
1043             r22              = _mm_mul_pd(rsq22,rinv22);
1044
1045             /* Calculate table index by multiplying r with table scale and truncate to integer */
1046             rt               = _mm_mul_pd(r22,vftabscale);
1047             vfitab           = _mm_cvttpd_epi32(rt);
1048             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1049             vfitab           = _mm_slli_epi32(vfitab,2);
1050
1051             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1052             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1053             F                = _mm_setzero_pd();
1054             GMX_MM_TRANSPOSE2_PD(Y,F);
1055             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1056             H                = _mm_setzero_pd();
1057             GMX_MM_TRANSPOSE2_PD(G,H);
1058             Heps             = _mm_mul_pd(vfeps,H);
1059             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1060             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1061             velec            = _mm_mul_pd(qq22,VV);
1062             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1063             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1064
1065             /* Update potential sum for this i atom from the interaction with this j atom. */
1066             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1067             velecsum         = _mm_add_pd(velecsum,velec);
1068
1069             fscal            = felec;
1070
1071             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1072
1073             /* Calculate temporary vectorial force */
1074             tx               = _mm_mul_pd(fscal,dx22);
1075             ty               = _mm_mul_pd(fscal,dy22);
1076             tz               = _mm_mul_pd(fscal,dz22);
1077
1078             /* Update vectorial force */
1079             fix2             = _mm_add_pd(fix2,tx);
1080             fiy2             = _mm_add_pd(fiy2,ty);
1081             fiz2             = _mm_add_pd(fiz2,tz);
1082
1083             fjx2             = _mm_add_pd(fjx2,tx);
1084             fjy2             = _mm_add_pd(fjy2,ty);
1085             fjz2             = _mm_add_pd(fjz2,tz);
1086
1087             /**************************
1088              * CALCULATE INTERACTIONS *
1089              **************************/
1090
1091             r23              = _mm_mul_pd(rsq23,rinv23);
1092
1093             /* Calculate table index by multiplying r with table scale and truncate to integer */
1094             rt               = _mm_mul_pd(r23,vftabscale);
1095             vfitab           = _mm_cvttpd_epi32(rt);
1096             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1097             vfitab           = _mm_slli_epi32(vfitab,2);
1098
1099             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1100             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1101             F                = _mm_setzero_pd();
1102             GMX_MM_TRANSPOSE2_PD(Y,F);
1103             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1104             H                = _mm_setzero_pd();
1105             GMX_MM_TRANSPOSE2_PD(G,H);
1106             Heps             = _mm_mul_pd(vfeps,H);
1107             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1108             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1109             velec            = _mm_mul_pd(qq23,VV);
1110             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1111             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1112
1113             /* Update potential sum for this i atom from the interaction with this j atom. */
1114             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1115             velecsum         = _mm_add_pd(velecsum,velec);
1116
1117             fscal            = felec;
1118
1119             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1120
1121             /* Calculate temporary vectorial force */
1122             tx               = _mm_mul_pd(fscal,dx23);
1123             ty               = _mm_mul_pd(fscal,dy23);
1124             tz               = _mm_mul_pd(fscal,dz23);
1125
1126             /* Update vectorial force */
1127             fix2             = _mm_add_pd(fix2,tx);
1128             fiy2             = _mm_add_pd(fiy2,ty);
1129             fiz2             = _mm_add_pd(fiz2,tz);
1130
1131             fjx3             = _mm_add_pd(fjx3,tx);
1132             fjy3             = _mm_add_pd(fjy3,ty);
1133             fjz3             = _mm_add_pd(fjz3,tz);
1134
1135             /**************************
1136              * CALCULATE INTERACTIONS *
1137              **************************/
1138
1139             r31              = _mm_mul_pd(rsq31,rinv31);
1140
1141             /* Calculate table index by multiplying r with table scale and truncate to integer */
1142             rt               = _mm_mul_pd(r31,vftabscale);
1143             vfitab           = _mm_cvttpd_epi32(rt);
1144             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1145             vfitab           = _mm_slli_epi32(vfitab,2);
1146
1147             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1148             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1149             F                = _mm_setzero_pd();
1150             GMX_MM_TRANSPOSE2_PD(Y,F);
1151             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1152             H                = _mm_setzero_pd();
1153             GMX_MM_TRANSPOSE2_PD(G,H);
1154             Heps             = _mm_mul_pd(vfeps,H);
1155             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1156             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1157             velec            = _mm_mul_pd(qq31,VV);
1158             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1159             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1160
1161             /* Update potential sum for this i atom from the interaction with this j atom. */
1162             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1163             velecsum         = _mm_add_pd(velecsum,velec);
1164
1165             fscal            = felec;
1166
1167             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1168
1169             /* Calculate temporary vectorial force */
1170             tx               = _mm_mul_pd(fscal,dx31);
1171             ty               = _mm_mul_pd(fscal,dy31);
1172             tz               = _mm_mul_pd(fscal,dz31);
1173
1174             /* Update vectorial force */
1175             fix3             = _mm_add_pd(fix3,tx);
1176             fiy3             = _mm_add_pd(fiy3,ty);
1177             fiz3             = _mm_add_pd(fiz3,tz);
1178
1179             fjx1             = _mm_add_pd(fjx1,tx);
1180             fjy1             = _mm_add_pd(fjy1,ty);
1181             fjz1             = _mm_add_pd(fjz1,tz);
1182
1183             /**************************
1184              * CALCULATE INTERACTIONS *
1185              **************************/
1186
1187             r32              = _mm_mul_pd(rsq32,rinv32);
1188
1189             /* Calculate table index by multiplying r with table scale and truncate to integer */
1190             rt               = _mm_mul_pd(r32,vftabscale);
1191             vfitab           = _mm_cvttpd_epi32(rt);
1192             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1193             vfitab           = _mm_slli_epi32(vfitab,2);
1194
1195             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1196             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1197             F                = _mm_setzero_pd();
1198             GMX_MM_TRANSPOSE2_PD(Y,F);
1199             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1200             H                = _mm_setzero_pd();
1201             GMX_MM_TRANSPOSE2_PD(G,H);
1202             Heps             = _mm_mul_pd(vfeps,H);
1203             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1204             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1205             velec            = _mm_mul_pd(qq32,VV);
1206             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1207             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1208
1209             /* Update potential sum for this i atom from the interaction with this j atom. */
1210             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1211             velecsum         = _mm_add_pd(velecsum,velec);
1212
1213             fscal            = felec;
1214
1215             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1216
1217             /* Calculate temporary vectorial force */
1218             tx               = _mm_mul_pd(fscal,dx32);
1219             ty               = _mm_mul_pd(fscal,dy32);
1220             tz               = _mm_mul_pd(fscal,dz32);
1221
1222             /* Update vectorial force */
1223             fix3             = _mm_add_pd(fix3,tx);
1224             fiy3             = _mm_add_pd(fiy3,ty);
1225             fiz3             = _mm_add_pd(fiz3,tz);
1226
1227             fjx2             = _mm_add_pd(fjx2,tx);
1228             fjy2             = _mm_add_pd(fjy2,ty);
1229             fjz2             = _mm_add_pd(fjz2,tz);
1230
1231             /**************************
1232              * CALCULATE INTERACTIONS *
1233              **************************/
1234
1235             r33              = _mm_mul_pd(rsq33,rinv33);
1236
1237             /* Calculate table index by multiplying r with table scale and truncate to integer */
1238             rt               = _mm_mul_pd(r33,vftabscale);
1239             vfitab           = _mm_cvttpd_epi32(rt);
1240             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1241             vfitab           = _mm_slli_epi32(vfitab,2);
1242
1243             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1244             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1245             F                = _mm_setzero_pd();
1246             GMX_MM_TRANSPOSE2_PD(Y,F);
1247             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1248             H                = _mm_setzero_pd();
1249             GMX_MM_TRANSPOSE2_PD(G,H);
1250             Heps             = _mm_mul_pd(vfeps,H);
1251             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1252             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1253             velec            = _mm_mul_pd(qq33,VV);
1254             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1255             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1256
1257             /* Update potential sum for this i atom from the interaction with this j atom. */
1258             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1259             velecsum         = _mm_add_pd(velecsum,velec);
1260
1261             fscal            = felec;
1262
1263             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1264
1265             /* Calculate temporary vectorial force */
1266             tx               = _mm_mul_pd(fscal,dx33);
1267             ty               = _mm_mul_pd(fscal,dy33);
1268             tz               = _mm_mul_pd(fscal,dz33);
1269
1270             /* Update vectorial force */
1271             fix3             = _mm_add_pd(fix3,tx);
1272             fiy3             = _mm_add_pd(fiy3,ty);
1273             fiz3             = _mm_add_pd(fiz3,tz);
1274
1275             fjx3             = _mm_add_pd(fjx3,tx);
1276             fjy3             = _mm_add_pd(fjy3,ty);
1277             fjz3             = _mm_add_pd(fjz3,tz);
1278
1279             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1280
1281             /* Inner loop uses 422 flops */
1282         }
1283
1284         /* End of innermost loop */
1285
1286         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1287                                               f+i_coord_offset,fshift+i_shift_offset);
1288
1289         ggid                        = gid[iidx];
1290         /* Update potential energies */
1291         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1292         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1293
1294         /* Increment number of inner iterations */
1295         inneriter                  += j_index_end - j_index_start;
1296
1297         /* Outer loop uses 26 flops */
1298     }
1299
1300     /* Increment number of outer iterations */
1301     outeriter        += nri;
1302
1303     /* Update outer/inner flops */
1304
1305     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*422);
1306 }
1307 /*
1308  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse4_1_double
1309  * Electrostatics interaction: CubicSplineTable
1310  * VdW interaction:            LennardJones
1311  * Geometry:                   Water4-Water4
1312  * Calculate force/pot:        Force
1313  */
1314 void
1315 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse4_1_double
1316                     (t_nblist                    * gmx_restrict       nlist,
1317                      rvec                        * gmx_restrict          xx,
1318                      rvec                        * gmx_restrict          ff,
1319                      t_forcerec                  * gmx_restrict          fr,
1320                      t_mdatoms                   * gmx_restrict     mdatoms,
1321                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1322                      t_nrnb                      * gmx_restrict        nrnb)
1323 {
1324     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1325      * just 0 for non-waters.
1326      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1327      * jnr indices corresponding to data put in the four positions in the SIMD register.
1328      */
1329     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1330     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1331     int              jnrA,jnrB;
1332     int              j_coord_offsetA,j_coord_offsetB;
1333     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1334     real             rcutoff_scalar;
1335     real             *shiftvec,*fshift,*x,*f;
1336     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1337     int              vdwioffset0;
1338     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1339     int              vdwioffset1;
1340     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1341     int              vdwioffset2;
1342     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1343     int              vdwioffset3;
1344     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1345     int              vdwjidx0A,vdwjidx0B;
1346     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1347     int              vdwjidx1A,vdwjidx1B;
1348     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1349     int              vdwjidx2A,vdwjidx2B;
1350     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1351     int              vdwjidx3A,vdwjidx3B;
1352     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1353     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1354     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1355     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1356     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1357     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1358     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1359     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1360     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1361     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1362     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1363     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1364     real             *charge;
1365     int              nvdwtype;
1366     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1367     int              *vdwtype;
1368     real             *vdwparam;
1369     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1370     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1371     __m128i          vfitab;
1372     __m128i          ifour       = _mm_set1_epi32(4);
1373     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1374     real             *vftab;
1375     __m128d          dummy_mask,cutoff_mask;
1376     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1377     __m128d          one     = _mm_set1_pd(1.0);
1378     __m128d          two     = _mm_set1_pd(2.0);
1379     x                = xx[0];
1380     f                = ff[0];
1381
1382     nri              = nlist->nri;
1383     iinr             = nlist->iinr;
1384     jindex           = nlist->jindex;
1385     jjnr             = nlist->jjnr;
1386     shiftidx         = nlist->shift;
1387     gid              = nlist->gid;
1388     shiftvec         = fr->shift_vec[0];
1389     fshift           = fr->fshift[0];
1390     facel            = _mm_set1_pd(fr->epsfac);
1391     charge           = mdatoms->chargeA;
1392     nvdwtype         = fr->ntype;
1393     vdwparam         = fr->nbfp;
1394     vdwtype          = mdatoms->typeA;
1395
1396     vftab            = kernel_data->table_elec->data;
1397     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
1398
1399     /* Setup water-specific parameters */
1400     inr              = nlist->iinr[0];
1401     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1402     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1403     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1404     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1405
1406     jq1              = _mm_set1_pd(charge[inr+1]);
1407     jq2              = _mm_set1_pd(charge[inr+2]);
1408     jq3              = _mm_set1_pd(charge[inr+3]);
1409     vdwjidx0A        = 2*vdwtype[inr+0];
1410     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1411     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1412     qq11             = _mm_mul_pd(iq1,jq1);
1413     qq12             = _mm_mul_pd(iq1,jq2);
1414     qq13             = _mm_mul_pd(iq1,jq3);
1415     qq21             = _mm_mul_pd(iq2,jq1);
1416     qq22             = _mm_mul_pd(iq2,jq2);
1417     qq23             = _mm_mul_pd(iq2,jq3);
1418     qq31             = _mm_mul_pd(iq3,jq1);
1419     qq32             = _mm_mul_pd(iq3,jq2);
1420     qq33             = _mm_mul_pd(iq3,jq3);
1421
1422     /* Avoid stupid compiler warnings */
1423     jnrA = jnrB = 0;
1424     j_coord_offsetA = 0;
1425     j_coord_offsetB = 0;
1426
1427     outeriter        = 0;
1428     inneriter        = 0;
1429
1430     /* Start outer loop over neighborlists */
1431     for(iidx=0; iidx<nri; iidx++)
1432     {
1433         /* Load shift vector for this list */
1434         i_shift_offset   = DIM*shiftidx[iidx];
1435
1436         /* Load limits for loop over neighbors */
1437         j_index_start    = jindex[iidx];
1438         j_index_end      = jindex[iidx+1];
1439
1440         /* Get outer coordinate index */
1441         inr              = iinr[iidx];
1442         i_coord_offset   = DIM*inr;
1443
1444         /* Load i particle coords and add shift vector */
1445         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1446                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1447
1448         fix0             = _mm_setzero_pd();
1449         fiy0             = _mm_setzero_pd();
1450         fiz0             = _mm_setzero_pd();
1451         fix1             = _mm_setzero_pd();
1452         fiy1             = _mm_setzero_pd();
1453         fiz1             = _mm_setzero_pd();
1454         fix2             = _mm_setzero_pd();
1455         fiy2             = _mm_setzero_pd();
1456         fiz2             = _mm_setzero_pd();
1457         fix3             = _mm_setzero_pd();
1458         fiy3             = _mm_setzero_pd();
1459         fiz3             = _mm_setzero_pd();
1460
1461         /* Start inner kernel loop */
1462         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1463         {
1464
1465             /* Get j neighbor index, and coordinate index */
1466             jnrA             = jjnr[jidx];
1467             jnrB             = jjnr[jidx+1];
1468             j_coord_offsetA  = DIM*jnrA;
1469             j_coord_offsetB  = DIM*jnrB;
1470
1471             /* load j atom coordinates */
1472             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1473                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1474                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1475
1476             /* Calculate displacement vector */
1477             dx00             = _mm_sub_pd(ix0,jx0);
1478             dy00             = _mm_sub_pd(iy0,jy0);
1479             dz00             = _mm_sub_pd(iz0,jz0);
1480             dx11             = _mm_sub_pd(ix1,jx1);
1481             dy11             = _mm_sub_pd(iy1,jy1);
1482             dz11             = _mm_sub_pd(iz1,jz1);
1483             dx12             = _mm_sub_pd(ix1,jx2);
1484             dy12             = _mm_sub_pd(iy1,jy2);
1485             dz12             = _mm_sub_pd(iz1,jz2);
1486             dx13             = _mm_sub_pd(ix1,jx3);
1487             dy13             = _mm_sub_pd(iy1,jy3);
1488             dz13             = _mm_sub_pd(iz1,jz3);
1489             dx21             = _mm_sub_pd(ix2,jx1);
1490             dy21             = _mm_sub_pd(iy2,jy1);
1491             dz21             = _mm_sub_pd(iz2,jz1);
1492             dx22             = _mm_sub_pd(ix2,jx2);
1493             dy22             = _mm_sub_pd(iy2,jy2);
1494             dz22             = _mm_sub_pd(iz2,jz2);
1495             dx23             = _mm_sub_pd(ix2,jx3);
1496             dy23             = _mm_sub_pd(iy2,jy3);
1497             dz23             = _mm_sub_pd(iz2,jz3);
1498             dx31             = _mm_sub_pd(ix3,jx1);
1499             dy31             = _mm_sub_pd(iy3,jy1);
1500             dz31             = _mm_sub_pd(iz3,jz1);
1501             dx32             = _mm_sub_pd(ix3,jx2);
1502             dy32             = _mm_sub_pd(iy3,jy2);
1503             dz32             = _mm_sub_pd(iz3,jz2);
1504             dx33             = _mm_sub_pd(ix3,jx3);
1505             dy33             = _mm_sub_pd(iy3,jy3);
1506             dz33             = _mm_sub_pd(iz3,jz3);
1507
1508             /* Calculate squared distance and things based on it */
1509             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1510             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1511             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1512             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1513             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1514             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1515             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1516             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1517             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1518             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1519
1520             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1521             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1522             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1523             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1524             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1525             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1526             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1527             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1528             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1529
1530             rinvsq00         = gmx_mm_inv_pd(rsq00);
1531
1532             fjx0             = _mm_setzero_pd();
1533             fjy0             = _mm_setzero_pd();
1534             fjz0             = _mm_setzero_pd();
1535             fjx1             = _mm_setzero_pd();
1536             fjy1             = _mm_setzero_pd();
1537             fjz1             = _mm_setzero_pd();
1538             fjx2             = _mm_setzero_pd();
1539             fjy2             = _mm_setzero_pd();
1540             fjz2             = _mm_setzero_pd();
1541             fjx3             = _mm_setzero_pd();
1542             fjy3             = _mm_setzero_pd();
1543             fjz3             = _mm_setzero_pd();
1544
1545             /**************************
1546              * CALCULATE INTERACTIONS *
1547              **************************/
1548
1549             /* LENNARD-JONES DISPERSION/REPULSION */
1550
1551             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1552             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1553
1554             fscal            = fvdw;
1555
1556             /* Calculate temporary vectorial force */
1557             tx               = _mm_mul_pd(fscal,dx00);
1558             ty               = _mm_mul_pd(fscal,dy00);
1559             tz               = _mm_mul_pd(fscal,dz00);
1560
1561             /* Update vectorial force */
1562             fix0             = _mm_add_pd(fix0,tx);
1563             fiy0             = _mm_add_pd(fiy0,ty);
1564             fiz0             = _mm_add_pd(fiz0,tz);
1565
1566             fjx0             = _mm_add_pd(fjx0,tx);
1567             fjy0             = _mm_add_pd(fjy0,ty);
1568             fjz0             = _mm_add_pd(fjz0,tz);
1569
1570             /**************************
1571              * CALCULATE INTERACTIONS *
1572              **************************/
1573
1574             r11              = _mm_mul_pd(rsq11,rinv11);
1575
1576             /* Calculate table index by multiplying r with table scale and truncate to integer */
1577             rt               = _mm_mul_pd(r11,vftabscale);
1578             vfitab           = _mm_cvttpd_epi32(rt);
1579             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1580             vfitab           = _mm_slli_epi32(vfitab,2);
1581
1582             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1583             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1584             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1585             GMX_MM_TRANSPOSE2_PD(Y,F);
1586             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1587             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1588             GMX_MM_TRANSPOSE2_PD(G,H);
1589             Heps             = _mm_mul_pd(vfeps,H);
1590             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1591             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1592             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1593
1594             fscal            = felec;
1595
1596             /* Calculate temporary vectorial force */
1597             tx               = _mm_mul_pd(fscal,dx11);
1598             ty               = _mm_mul_pd(fscal,dy11);
1599             tz               = _mm_mul_pd(fscal,dz11);
1600
1601             /* Update vectorial force */
1602             fix1             = _mm_add_pd(fix1,tx);
1603             fiy1             = _mm_add_pd(fiy1,ty);
1604             fiz1             = _mm_add_pd(fiz1,tz);
1605
1606             fjx1             = _mm_add_pd(fjx1,tx);
1607             fjy1             = _mm_add_pd(fjy1,ty);
1608             fjz1             = _mm_add_pd(fjz1,tz);
1609
1610             /**************************
1611              * CALCULATE INTERACTIONS *
1612              **************************/
1613
1614             r12              = _mm_mul_pd(rsq12,rinv12);
1615
1616             /* Calculate table index by multiplying r with table scale and truncate to integer */
1617             rt               = _mm_mul_pd(r12,vftabscale);
1618             vfitab           = _mm_cvttpd_epi32(rt);
1619             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1620             vfitab           = _mm_slli_epi32(vfitab,2);
1621
1622             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1623             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1624             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1625             GMX_MM_TRANSPOSE2_PD(Y,F);
1626             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1627             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1628             GMX_MM_TRANSPOSE2_PD(G,H);
1629             Heps             = _mm_mul_pd(vfeps,H);
1630             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1631             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1632             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1633
1634             fscal            = felec;
1635
1636             /* Calculate temporary vectorial force */
1637             tx               = _mm_mul_pd(fscal,dx12);
1638             ty               = _mm_mul_pd(fscal,dy12);
1639             tz               = _mm_mul_pd(fscal,dz12);
1640
1641             /* Update vectorial force */
1642             fix1             = _mm_add_pd(fix1,tx);
1643             fiy1             = _mm_add_pd(fiy1,ty);
1644             fiz1             = _mm_add_pd(fiz1,tz);
1645
1646             fjx2             = _mm_add_pd(fjx2,tx);
1647             fjy2             = _mm_add_pd(fjy2,ty);
1648             fjz2             = _mm_add_pd(fjz2,tz);
1649
1650             /**************************
1651              * CALCULATE INTERACTIONS *
1652              **************************/
1653
1654             r13              = _mm_mul_pd(rsq13,rinv13);
1655
1656             /* Calculate table index by multiplying r with table scale and truncate to integer */
1657             rt               = _mm_mul_pd(r13,vftabscale);
1658             vfitab           = _mm_cvttpd_epi32(rt);
1659             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1660             vfitab           = _mm_slli_epi32(vfitab,2);
1661
1662             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1663             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1664             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1665             GMX_MM_TRANSPOSE2_PD(Y,F);
1666             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1667             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1668             GMX_MM_TRANSPOSE2_PD(G,H);
1669             Heps             = _mm_mul_pd(vfeps,H);
1670             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1671             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1672             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1673
1674             fscal            = felec;
1675
1676             /* Calculate temporary vectorial force */
1677             tx               = _mm_mul_pd(fscal,dx13);
1678             ty               = _mm_mul_pd(fscal,dy13);
1679             tz               = _mm_mul_pd(fscal,dz13);
1680
1681             /* Update vectorial force */
1682             fix1             = _mm_add_pd(fix1,tx);
1683             fiy1             = _mm_add_pd(fiy1,ty);
1684             fiz1             = _mm_add_pd(fiz1,tz);
1685
1686             fjx3             = _mm_add_pd(fjx3,tx);
1687             fjy3             = _mm_add_pd(fjy3,ty);
1688             fjz3             = _mm_add_pd(fjz3,tz);
1689
1690             /**************************
1691              * CALCULATE INTERACTIONS *
1692              **************************/
1693
1694             r21              = _mm_mul_pd(rsq21,rinv21);
1695
1696             /* Calculate table index by multiplying r with table scale and truncate to integer */
1697             rt               = _mm_mul_pd(r21,vftabscale);
1698             vfitab           = _mm_cvttpd_epi32(rt);
1699             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1700             vfitab           = _mm_slli_epi32(vfitab,2);
1701
1702             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1703             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1704             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1705             GMX_MM_TRANSPOSE2_PD(Y,F);
1706             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1707             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1708             GMX_MM_TRANSPOSE2_PD(G,H);
1709             Heps             = _mm_mul_pd(vfeps,H);
1710             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1711             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1712             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1713
1714             fscal            = felec;
1715
1716             /* Calculate temporary vectorial force */
1717             tx               = _mm_mul_pd(fscal,dx21);
1718             ty               = _mm_mul_pd(fscal,dy21);
1719             tz               = _mm_mul_pd(fscal,dz21);
1720
1721             /* Update vectorial force */
1722             fix2             = _mm_add_pd(fix2,tx);
1723             fiy2             = _mm_add_pd(fiy2,ty);
1724             fiz2             = _mm_add_pd(fiz2,tz);
1725
1726             fjx1             = _mm_add_pd(fjx1,tx);
1727             fjy1             = _mm_add_pd(fjy1,ty);
1728             fjz1             = _mm_add_pd(fjz1,tz);
1729
1730             /**************************
1731              * CALCULATE INTERACTIONS *
1732              **************************/
1733
1734             r22              = _mm_mul_pd(rsq22,rinv22);
1735
1736             /* Calculate table index by multiplying r with table scale and truncate to integer */
1737             rt               = _mm_mul_pd(r22,vftabscale);
1738             vfitab           = _mm_cvttpd_epi32(rt);
1739             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1740             vfitab           = _mm_slli_epi32(vfitab,2);
1741
1742             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1743             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1744             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1745             GMX_MM_TRANSPOSE2_PD(Y,F);
1746             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1747             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1748             GMX_MM_TRANSPOSE2_PD(G,H);
1749             Heps             = _mm_mul_pd(vfeps,H);
1750             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1751             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1752             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1753
1754             fscal            = felec;
1755
1756             /* Calculate temporary vectorial force */
1757             tx               = _mm_mul_pd(fscal,dx22);
1758             ty               = _mm_mul_pd(fscal,dy22);
1759             tz               = _mm_mul_pd(fscal,dz22);
1760
1761             /* Update vectorial force */
1762             fix2             = _mm_add_pd(fix2,tx);
1763             fiy2             = _mm_add_pd(fiy2,ty);
1764             fiz2             = _mm_add_pd(fiz2,tz);
1765
1766             fjx2             = _mm_add_pd(fjx2,tx);
1767             fjy2             = _mm_add_pd(fjy2,ty);
1768             fjz2             = _mm_add_pd(fjz2,tz);
1769
1770             /**************************
1771              * CALCULATE INTERACTIONS *
1772              **************************/
1773
1774             r23              = _mm_mul_pd(rsq23,rinv23);
1775
1776             /* Calculate table index by multiplying r with table scale and truncate to integer */
1777             rt               = _mm_mul_pd(r23,vftabscale);
1778             vfitab           = _mm_cvttpd_epi32(rt);
1779             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1780             vfitab           = _mm_slli_epi32(vfitab,2);
1781
1782             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1783             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1784             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1785             GMX_MM_TRANSPOSE2_PD(Y,F);
1786             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1787             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1788             GMX_MM_TRANSPOSE2_PD(G,H);
1789             Heps             = _mm_mul_pd(vfeps,H);
1790             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1791             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1792             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1793
1794             fscal            = felec;
1795
1796             /* Calculate temporary vectorial force */
1797             tx               = _mm_mul_pd(fscal,dx23);
1798             ty               = _mm_mul_pd(fscal,dy23);
1799             tz               = _mm_mul_pd(fscal,dz23);
1800
1801             /* Update vectorial force */
1802             fix2             = _mm_add_pd(fix2,tx);
1803             fiy2             = _mm_add_pd(fiy2,ty);
1804             fiz2             = _mm_add_pd(fiz2,tz);
1805
1806             fjx3             = _mm_add_pd(fjx3,tx);
1807             fjy3             = _mm_add_pd(fjy3,ty);
1808             fjz3             = _mm_add_pd(fjz3,tz);
1809
1810             /**************************
1811              * CALCULATE INTERACTIONS *
1812              **************************/
1813
1814             r31              = _mm_mul_pd(rsq31,rinv31);
1815
1816             /* Calculate table index by multiplying r with table scale and truncate to integer */
1817             rt               = _mm_mul_pd(r31,vftabscale);
1818             vfitab           = _mm_cvttpd_epi32(rt);
1819             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1820             vfitab           = _mm_slli_epi32(vfitab,2);
1821
1822             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1823             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1824             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1825             GMX_MM_TRANSPOSE2_PD(Y,F);
1826             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1827             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1828             GMX_MM_TRANSPOSE2_PD(G,H);
1829             Heps             = _mm_mul_pd(vfeps,H);
1830             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1831             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1832             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1833
1834             fscal            = felec;
1835
1836             /* Calculate temporary vectorial force */
1837             tx               = _mm_mul_pd(fscal,dx31);
1838             ty               = _mm_mul_pd(fscal,dy31);
1839             tz               = _mm_mul_pd(fscal,dz31);
1840
1841             /* Update vectorial force */
1842             fix3             = _mm_add_pd(fix3,tx);
1843             fiy3             = _mm_add_pd(fiy3,ty);
1844             fiz3             = _mm_add_pd(fiz3,tz);
1845
1846             fjx1             = _mm_add_pd(fjx1,tx);
1847             fjy1             = _mm_add_pd(fjy1,ty);
1848             fjz1             = _mm_add_pd(fjz1,tz);
1849
1850             /**************************
1851              * CALCULATE INTERACTIONS *
1852              **************************/
1853
1854             r32              = _mm_mul_pd(rsq32,rinv32);
1855
1856             /* Calculate table index by multiplying r with table scale and truncate to integer */
1857             rt               = _mm_mul_pd(r32,vftabscale);
1858             vfitab           = _mm_cvttpd_epi32(rt);
1859             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1860             vfitab           = _mm_slli_epi32(vfitab,2);
1861
1862             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1863             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1864             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1865             GMX_MM_TRANSPOSE2_PD(Y,F);
1866             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1867             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1868             GMX_MM_TRANSPOSE2_PD(G,H);
1869             Heps             = _mm_mul_pd(vfeps,H);
1870             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1871             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1872             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1873
1874             fscal            = felec;
1875
1876             /* Calculate temporary vectorial force */
1877             tx               = _mm_mul_pd(fscal,dx32);
1878             ty               = _mm_mul_pd(fscal,dy32);
1879             tz               = _mm_mul_pd(fscal,dz32);
1880
1881             /* Update vectorial force */
1882             fix3             = _mm_add_pd(fix3,tx);
1883             fiy3             = _mm_add_pd(fiy3,ty);
1884             fiz3             = _mm_add_pd(fiz3,tz);
1885
1886             fjx2             = _mm_add_pd(fjx2,tx);
1887             fjy2             = _mm_add_pd(fjy2,ty);
1888             fjz2             = _mm_add_pd(fjz2,tz);
1889
1890             /**************************
1891              * CALCULATE INTERACTIONS *
1892              **************************/
1893
1894             r33              = _mm_mul_pd(rsq33,rinv33);
1895
1896             /* Calculate table index by multiplying r with table scale and truncate to integer */
1897             rt               = _mm_mul_pd(r33,vftabscale);
1898             vfitab           = _mm_cvttpd_epi32(rt);
1899             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1900             vfitab           = _mm_slli_epi32(vfitab,2);
1901
1902             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1903             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1904             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1905             GMX_MM_TRANSPOSE2_PD(Y,F);
1906             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1907             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1908             GMX_MM_TRANSPOSE2_PD(G,H);
1909             Heps             = _mm_mul_pd(vfeps,H);
1910             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1911             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1912             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1913
1914             fscal            = felec;
1915
1916             /* Calculate temporary vectorial force */
1917             tx               = _mm_mul_pd(fscal,dx33);
1918             ty               = _mm_mul_pd(fscal,dy33);
1919             tz               = _mm_mul_pd(fscal,dz33);
1920
1921             /* Update vectorial force */
1922             fix3             = _mm_add_pd(fix3,tx);
1923             fiy3             = _mm_add_pd(fiy3,ty);
1924             fiz3             = _mm_add_pd(fiz3,tz);
1925
1926             fjx3             = _mm_add_pd(fjx3,tx);
1927             fjy3             = _mm_add_pd(fjy3,ty);
1928             fjz3             = _mm_add_pd(fjz3,tz);
1929
1930             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1931
1932             /* Inner loop uses 381 flops */
1933         }
1934
1935         if(jidx<j_index_end)
1936         {
1937
1938             jnrA             = jjnr[jidx];
1939             j_coord_offsetA  = DIM*jnrA;
1940
1941             /* load j atom coordinates */
1942             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1943                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1944                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1945
1946             /* Calculate displacement vector */
1947             dx00             = _mm_sub_pd(ix0,jx0);
1948             dy00             = _mm_sub_pd(iy0,jy0);
1949             dz00             = _mm_sub_pd(iz0,jz0);
1950             dx11             = _mm_sub_pd(ix1,jx1);
1951             dy11             = _mm_sub_pd(iy1,jy1);
1952             dz11             = _mm_sub_pd(iz1,jz1);
1953             dx12             = _mm_sub_pd(ix1,jx2);
1954             dy12             = _mm_sub_pd(iy1,jy2);
1955             dz12             = _mm_sub_pd(iz1,jz2);
1956             dx13             = _mm_sub_pd(ix1,jx3);
1957             dy13             = _mm_sub_pd(iy1,jy3);
1958             dz13             = _mm_sub_pd(iz1,jz3);
1959             dx21             = _mm_sub_pd(ix2,jx1);
1960             dy21             = _mm_sub_pd(iy2,jy1);
1961             dz21             = _mm_sub_pd(iz2,jz1);
1962             dx22             = _mm_sub_pd(ix2,jx2);
1963             dy22             = _mm_sub_pd(iy2,jy2);
1964             dz22             = _mm_sub_pd(iz2,jz2);
1965             dx23             = _mm_sub_pd(ix2,jx3);
1966             dy23             = _mm_sub_pd(iy2,jy3);
1967             dz23             = _mm_sub_pd(iz2,jz3);
1968             dx31             = _mm_sub_pd(ix3,jx1);
1969             dy31             = _mm_sub_pd(iy3,jy1);
1970             dz31             = _mm_sub_pd(iz3,jz1);
1971             dx32             = _mm_sub_pd(ix3,jx2);
1972             dy32             = _mm_sub_pd(iy3,jy2);
1973             dz32             = _mm_sub_pd(iz3,jz2);
1974             dx33             = _mm_sub_pd(ix3,jx3);
1975             dy33             = _mm_sub_pd(iy3,jy3);
1976             dz33             = _mm_sub_pd(iz3,jz3);
1977
1978             /* Calculate squared distance and things based on it */
1979             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1980             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1981             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1982             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1983             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1984             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1985             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1986             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1987             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1988             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1989
1990             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1991             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1992             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1993             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1994             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1995             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1996             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1997             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1998             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1999
2000             rinvsq00         = gmx_mm_inv_pd(rsq00);
2001
2002             fjx0             = _mm_setzero_pd();
2003             fjy0             = _mm_setzero_pd();
2004             fjz0             = _mm_setzero_pd();
2005             fjx1             = _mm_setzero_pd();
2006             fjy1             = _mm_setzero_pd();
2007             fjz1             = _mm_setzero_pd();
2008             fjx2             = _mm_setzero_pd();
2009             fjy2             = _mm_setzero_pd();
2010             fjz2             = _mm_setzero_pd();
2011             fjx3             = _mm_setzero_pd();
2012             fjy3             = _mm_setzero_pd();
2013             fjz3             = _mm_setzero_pd();
2014
2015             /**************************
2016              * CALCULATE INTERACTIONS *
2017              **************************/
2018
2019             /* LENNARD-JONES DISPERSION/REPULSION */
2020
2021             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2022             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
2023
2024             fscal            = fvdw;
2025
2026             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2027
2028             /* Calculate temporary vectorial force */
2029             tx               = _mm_mul_pd(fscal,dx00);
2030             ty               = _mm_mul_pd(fscal,dy00);
2031             tz               = _mm_mul_pd(fscal,dz00);
2032
2033             /* Update vectorial force */
2034             fix0             = _mm_add_pd(fix0,tx);
2035             fiy0             = _mm_add_pd(fiy0,ty);
2036             fiz0             = _mm_add_pd(fiz0,tz);
2037
2038             fjx0             = _mm_add_pd(fjx0,tx);
2039             fjy0             = _mm_add_pd(fjy0,ty);
2040             fjz0             = _mm_add_pd(fjz0,tz);
2041
2042             /**************************
2043              * CALCULATE INTERACTIONS *
2044              **************************/
2045
2046             r11              = _mm_mul_pd(rsq11,rinv11);
2047
2048             /* Calculate table index by multiplying r with table scale and truncate to integer */
2049             rt               = _mm_mul_pd(r11,vftabscale);
2050             vfitab           = _mm_cvttpd_epi32(rt);
2051             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2052             vfitab           = _mm_slli_epi32(vfitab,2);
2053
2054             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2055             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2056             F                = _mm_setzero_pd();
2057             GMX_MM_TRANSPOSE2_PD(Y,F);
2058             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2059             H                = _mm_setzero_pd();
2060             GMX_MM_TRANSPOSE2_PD(G,H);
2061             Heps             = _mm_mul_pd(vfeps,H);
2062             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2063             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2064             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2065
2066             fscal            = felec;
2067
2068             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2069
2070             /* Calculate temporary vectorial force */
2071             tx               = _mm_mul_pd(fscal,dx11);
2072             ty               = _mm_mul_pd(fscal,dy11);
2073             tz               = _mm_mul_pd(fscal,dz11);
2074
2075             /* Update vectorial force */
2076             fix1             = _mm_add_pd(fix1,tx);
2077             fiy1             = _mm_add_pd(fiy1,ty);
2078             fiz1             = _mm_add_pd(fiz1,tz);
2079
2080             fjx1             = _mm_add_pd(fjx1,tx);
2081             fjy1             = _mm_add_pd(fjy1,ty);
2082             fjz1             = _mm_add_pd(fjz1,tz);
2083
2084             /**************************
2085              * CALCULATE INTERACTIONS *
2086              **************************/
2087
2088             r12              = _mm_mul_pd(rsq12,rinv12);
2089
2090             /* Calculate table index by multiplying r with table scale and truncate to integer */
2091             rt               = _mm_mul_pd(r12,vftabscale);
2092             vfitab           = _mm_cvttpd_epi32(rt);
2093             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2094             vfitab           = _mm_slli_epi32(vfitab,2);
2095
2096             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2097             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2098             F                = _mm_setzero_pd();
2099             GMX_MM_TRANSPOSE2_PD(Y,F);
2100             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2101             H                = _mm_setzero_pd();
2102             GMX_MM_TRANSPOSE2_PD(G,H);
2103             Heps             = _mm_mul_pd(vfeps,H);
2104             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2105             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2106             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2107
2108             fscal            = felec;
2109
2110             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2111
2112             /* Calculate temporary vectorial force */
2113             tx               = _mm_mul_pd(fscal,dx12);
2114             ty               = _mm_mul_pd(fscal,dy12);
2115             tz               = _mm_mul_pd(fscal,dz12);
2116
2117             /* Update vectorial force */
2118             fix1             = _mm_add_pd(fix1,tx);
2119             fiy1             = _mm_add_pd(fiy1,ty);
2120             fiz1             = _mm_add_pd(fiz1,tz);
2121
2122             fjx2             = _mm_add_pd(fjx2,tx);
2123             fjy2             = _mm_add_pd(fjy2,ty);
2124             fjz2             = _mm_add_pd(fjz2,tz);
2125
2126             /**************************
2127              * CALCULATE INTERACTIONS *
2128              **************************/
2129
2130             r13              = _mm_mul_pd(rsq13,rinv13);
2131
2132             /* Calculate table index by multiplying r with table scale and truncate to integer */
2133             rt               = _mm_mul_pd(r13,vftabscale);
2134             vfitab           = _mm_cvttpd_epi32(rt);
2135             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2136             vfitab           = _mm_slli_epi32(vfitab,2);
2137
2138             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2139             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2140             F                = _mm_setzero_pd();
2141             GMX_MM_TRANSPOSE2_PD(Y,F);
2142             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2143             H                = _mm_setzero_pd();
2144             GMX_MM_TRANSPOSE2_PD(G,H);
2145             Heps             = _mm_mul_pd(vfeps,H);
2146             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2147             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2148             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
2149
2150             fscal            = felec;
2151
2152             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2153
2154             /* Calculate temporary vectorial force */
2155             tx               = _mm_mul_pd(fscal,dx13);
2156             ty               = _mm_mul_pd(fscal,dy13);
2157             tz               = _mm_mul_pd(fscal,dz13);
2158
2159             /* Update vectorial force */
2160             fix1             = _mm_add_pd(fix1,tx);
2161             fiy1             = _mm_add_pd(fiy1,ty);
2162             fiz1             = _mm_add_pd(fiz1,tz);
2163
2164             fjx3             = _mm_add_pd(fjx3,tx);
2165             fjy3             = _mm_add_pd(fjy3,ty);
2166             fjz3             = _mm_add_pd(fjz3,tz);
2167
2168             /**************************
2169              * CALCULATE INTERACTIONS *
2170              **************************/
2171
2172             r21              = _mm_mul_pd(rsq21,rinv21);
2173
2174             /* Calculate table index by multiplying r with table scale and truncate to integer */
2175             rt               = _mm_mul_pd(r21,vftabscale);
2176             vfitab           = _mm_cvttpd_epi32(rt);
2177             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2178             vfitab           = _mm_slli_epi32(vfitab,2);
2179
2180             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2181             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2182             F                = _mm_setzero_pd();
2183             GMX_MM_TRANSPOSE2_PD(Y,F);
2184             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2185             H                = _mm_setzero_pd();
2186             GMX_MM_TRANSPOSE2_PD(G,H);
2187             Heps             = _mm_mul_pd(vfeps,H);
2188             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2189             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2190             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2191
2192             fscal            = felec;
2193
2194             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2195
2196             /* Calculate temporary vectorial force */
2197             tx               = _mm_mul_pd(fscal,dx21);
2198             ty               = _mm_mul_pd(fscal,dy21);
2199             tz               = _mm_mul_pd(fscal,dz21);
2200
2201             /* Update vectorial force */
2202             fix2             = _mm_add_pd(fix2,tx);
2203             fiy2             = _mm_add_pd(fiy2,ty);
2204             fiz2             = _mm_add_pd(fiz2,tz);
2205
2206             fjx1             = _mm_add_pd(fjx1,tx);
2207             fjy1             = _mm_add_pd(fjy1,ty);
2208             fjz1             = _mm_add_pd(fjz1,tz);
2209
2210             /**************************
2211              * CALCULATE INTERACTIONS *
2212              **************************/
2213
2214             r22              = _mm_mul_pd(rsq22,rinv22);
2215
2216             /* Calculate table index by multiplying r with table scale and truncate to integer */
2217             rt               = _mm_mul_pd(r22,vftabscale);
2218             vfitab           = _mm_cvttpd_epi32(rt);
2219             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2220             vfitab           = _mm_slli_epi32(vfitab,2);
2221
2222             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2223             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2224             F                = _mm_setzero_pd();
2225             GMX_MM_TRANSPOSE2_PD(Y,F);
2226             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2227             H                = _mm_setzero_pd();
2228             GMX_MM_TRANSPOSE2_PD(G,H);
2229             Heps             = _mm_mul_pd(vfeps,H);
2230             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2231             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2232             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2233
2234             fscal            = felec;
2235
2236             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2237
2238             /* Calculate temporary vectorial force */
2239             tx               = _mm_mul_pd(fscal,dx22);
2240             ty               = _mm_mul_pd(fscal,dy22);
2241             tz               = _mm_mul_pd(fscal,dz22);
2242
2243             /* Update vectorial force */
2244             fix2             = _mm_add_pd(fix2,tx);
2245             fiy2             = _mm_add_pd(fiy2,ty);
2246             fiz2             = _mm_add_pd(fiz2,tz);
2247
2248             fjx2             = _mm_add_pd(fjx2,tx);
2249             fjy2             = _mm_add_pd(fjy2,ty);
2250             fjz2             = _mm_add_pd(fjz2,tz);
2251
2252             /**************************
2253              * CALCULATE INTERACTIONS *
2254              **************************/
2255
2256             r23              = _mm_mul_pd(rsq23,rinv23);
2257
2258             /* Calculate table index by multiplying r with table scale and truncate to integer */
2259             rt               = _mm_mul_pd(r23,vftabscale);
2260             vfitab           = _mm_cvttpd_epi32(rt);
2261             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2262             vfitab           = _mm_slli_epi32(vfitab,2);
2263
2264             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2265             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2266             F                = _mm_setzero_pd();
2267             GMX_MM_TRANSPOSE2_PD(Y,F);
2268             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2269             H                = _mm_setzero_pd();
2270             GMX_MM_TRANSPOSE2_PD(G,H);
2271             Heps             = _mm_mul_pd(vfeps,H);
2272             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2273             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2274             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
2275
2276             fscal            = felec;
2277
2278             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2279
2280             /* Calculate temporary vectorial force */
2281             tx               = _mm_mul_pd(fscal,dx23);
2282             ty               = _mm_mul_pd(fscal,dy23);
2283             tz               = _mm_mul_pd(fscal,dz23);
2284
2285             /* Update vectorial force */
2286             fix2             = _mm_add_pd(fix2,tx);
2287             fiy2             = _mm_add_pd(fiy2,ty);
2288             fiz2             = _mm_add_pd(fiz2,tz);
2289
2290             fjx3             = _mm_add_pd(fjx3,tx);
2291             fjy3             = _mm_add_pd(fjy3,ty);
2292             fjz3             = _mm_add_pd(fjz3,tz);
2293
2294             /**************************
2295              * CALCULATE INTERACTIONS *
2296              **************************/
2297
2298             r31              = _mm_mul_pd(rsq31,rinv31);
2299
2300             /* Calculate table index by multiplying r with table scale and truncate to integer */
2301             rt               = _mm_mul_pd(r31,vftabscale);
2302             vfitab           = _mm_cvttpd_epi32(rt);
2303             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2304             vfitab           = _mm_slli_epi32(vfitab,2);
2305
2306             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2307             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2308             F                = _mm_setzero_pd();
2309             GMX_MM_TRANSPOSE2_PD(Y,F);
2310             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2311             H                = _mm_setzero_pd();
2312             GMX_MM_TRANSPOSE2_PD(G,H);
2313             Heps             = _mm_mul_pd(vfeps,H);
2314             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2315             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2316             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
2317
2318             fscal            = felec;
2319
2320             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2321
2322             /* Calculate temporary vectorial force */
2323             tx               = _mm_mul_pd(fscal,dx31);
2324             ty               = _mm_mul_pd(fscal,dy31);
2325             tz               = _mm_mul_pd(fscal,dz31);
2326
2327             /* Update vectorial force */
2328             fix3             = _mm_add_pd(fix3,tx);
2329             fiy3             = _mm_add_pd(fiy3,ty);
2330             fiz3             = _mm_add_pd(fiz3,tz);
2331
2332             fjx1             = _mm_add_pd(fjx1,tx);
2333             fjy1             = _mm_add_pd(fjy1,ty);
2334             fjz1             = _mm_add_pd(fjz1,tz);
2335
2336             /**************************
2337              * CALCULATE INTERACTIONS *
2338              **************************/
2339
2340             r32              = _mm_mul_pd(rsq32,rinv32);
2341
2342             /* Calculate table index by multiplying r with table scale and truncate to integer */
2343             rt               = _mm_mul_pd(r32,vftabscale);
2344             vfitab           = _mm_cvttpd_epi32(rt);
2345             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2346             vfitab           = _mm_slli_epi32(vfitab,2);
2347
2348             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2349             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2350             F                = _mm_setzero_pd();
2351             GMX_MM_TRANSPOSE2_PD(Y,F);
2352             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2353             H                = _mm_setzero_pd();
2354             GMX_MM_TRANSPOSE2_PD(G,H);
2355             Heps             = _mm_mul_pd(vfeps,H);
2356             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2357             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2358             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
2359
2360             fscal            = felec;
2361
2362             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2363
2364             /* Calculate temporary vectorial force */
2365             tx               = _mm_mul_pd(fscal,dx32);
2366             ty               = _mm_mul_pd(fscal,dy32);
2367             tz               = _mm_mul_pd(fscal,dz32);
2368
2369             /* Update vectorial force */
2370             fix3             = _mm_add_pd(fix3,tx);
2371             fiy3             = _mm_add_pd(fiy3,ty);
2372             fiz3             = _mm_add_pd(fiz3,tz);
2373
2374             fjx2             = _mm_add_pd(fjx2,tx);
2375             fjy2             = _mm_add_pd(fjy2,ty);
2376             fjz2             = _mm_add_pd(fjz2,tz);
2377
2378             /**************************
2379              * CALCULATE INTERACTIONS *
2380              **************************/
2381
2382             r33              = _mm_mul_pd(rsq33,rinv33);
2383
2384             /* Calculate table index by multiplying r with table scale and truncate to integer */
2385             rt               = _mm_mul_pd(r33,vftabscale);
2386             vfitab           = _mm_cvttpd_epi32(rt);
2387             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2388             vfitab           = _mm_slli_epi32(vfitab,2);
2389
2390             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2391             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2392             F                = _mm_setzero_pd();
2393             GMX_MM_TRANSPOSE2_PD(Y,F);
2394             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2395             H                = _mm_setzero_pd();
2396             GMX_MM_TRANSPOSE2_PD(G,H);
2397             Heps             = _mm_mul_pd(vfeps,H);
2398             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2399             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2400             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2401
2402             fscal            = felec;
2403
2404             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2405
2406             /* Calculate temporary vectorial force */
2407             tx               = _mm_mul_pd(fscal,dx33);
2408             ty               = _mm_mul_pd(fscal,dy33);
2409             tz               = _mm_mul_pd(fscal,dz33);
2410
2411             /* Update vectorial force */
2412             fix3             = _mm_add_pd(fix3,tx);
2413             fiy3             = _mm_add_pd(fiy3,ty);
2414             fiz3             = _mm_add_pd(fiz3,tz);
2415
2416             fjx3             = _mm_add_pd(fjx3,tx);
2417             fjy3             = _mm_add_pd(fjy3,ty);
2418             fjz3             = _mm_add_pd(fjz3,tz);
2419
2420             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2421
2422             /* Inner loop uses 381 flops */
2423         }
2424
2425         /* End of innermost loop */
2426
2427         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2428                                               f+i_coord_offset,fshift+i_shift_offset);
2429
2430         /* Increment number of inner iterations */
2431         inneriter                  += j_index_end - j_index_start;
2432
2433         /* Outer loop uses 24 flops */
2434     }
2435
2436     /* Increment number of outer iterations */
2437     outeriter        += nri;
2438
2439     /* Update outer/inner flops */
2440
2441     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*381);
2442 }