src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_sse2_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse2_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_sse2_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_VF_sse2_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Water3
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_VF_sse2_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     int              vdwjidx1A,vdwjidx1B;
  88     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  89     int              vdwjidx2A,vdwjidx2B;
  90     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  91     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  93     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  94     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  95     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  96     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  97     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  98     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  99     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 100     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 101     real             *charge;
 102     int              nvdwtype;
 103     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 104     int              *vdwtype;
 105     real             *vdwparam;
 106     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 107     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 108     __m128i          vfitab;
 109     __m128i          ifour       = _mm_set1_epi32(4);
 110     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 111     real             *vftab;
 112     __m128d          dummy_mask,cutoff_mask;
 113     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 114     __m128d          one     = _mm_set1_pd(1.0);
 115     __m128d          two     = _mm_set1_pd(2.0);
 116     x                = xx[0];
 117     f                = ff[0];
 118
 119     nri              = nlist->nri;
 120     iinr             = nlist->iinr;
 121     jindex           = nlist->jindex;
 122     jjnr             = nlist->jjnr;
 123     shiftidx         = nlist->shift;
 124     gid              = nlist->gid;
 125     shiftvec         = fr->shift_vec[0];
 126     fshift           = fr->fshift[0];
 127     facel            = _mm_set1_pd(fr->ic->epsfac);
 128     charge           = mdatoms->chargeA;
 129     nvdwtype         = fr->ntype;
 130     vdwparam         = fr->nbfp;
 131     vdwtype          = mdatoms->typeA;
 132
 133     vftab            = kernel_data->table_elec_vdw->data;
 134     vftabscale       = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
 135
 136     /* Setup water-specific parameters */
 137     inr              = nlist->iinr[0];
 138     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 139     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 140     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 141     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 142
 143     jq0              = _mm_set1_pd(charge[inr+0]);
 144     jq1              = _mm_set1_pd(charge[inr+1]);
 145     jq2              = _mm_set1_pd(charge[inr+2]);
 146     vdwjidx0A        = 2*vdwtype[inr+0];
 147     qq00             = _mm_mul_pd(iq0,jq0);
 148     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 149     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 150     qq01             = _mm_mul_pd(iq0,jq1);
 151     qq02             = _mm_mul_pd(iq0,jq2);
 152     qq10             = _mm_mul_pd(iq1,jq0);
 153     qq11             = _mm_mul_pd(iq1,jq1);
 154     qq12             = _mm_mul_pd(iq1,jq2);
 155     qq20             = _mm_mul_pd(iq2,jq0);
 156     qq21             = _mm_mul_pd(iq2,jq1);
 157     qq22             = _mm_mul_pd(iq2,jq2);
 158
 159     /* Avoid stupid compiler warnings */
 160     jnrA = jnrB = 0;
 161     j_coord_offsetA = 0;
 162     j_coord_offsetB = 0;
 163
 164     outeriter        = 0;
 165     inneriter        = 0;
 166
 167     /* Start outer loop over neighborlists */
 168     for(iidx=0; iidx<nri; iidx++)
 169     {
 170         /* Load shift vector for this list */
 171         i_shift_offset   = DIM*shiftidx[iidx];
 172
 173         /* Load limits for loop over neighbors */
 174         j_index_start    = jindex[iidx];
 175         j_index_end      = jindex[iidx+1];
 176
 177         /* Get outer coordinate index */
 178         inr              = iinr[iidx];
 179         i_coord_offset   = DIM*inr;
 180
 181         /* Load i particle coords and add shift vector */
 182         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 183                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 184
 185         fix0             = _mm_setzero_pd();
 186         fiy0             = _mm_setzero_pd();
 187         fiz0             = _mm_setzero_pd();
 188         fix1             = _mm_setzero_pd();
 189         fiy1             = _mm_setzero_pd();
 190         fiz1             = _mm_setzero_pd();
 191         fix2             = _mm_setzero_pd();
 192         fiy2             = _mm_setzero_pd();
 193         fiz2             = _mm_setzero_pd();
 194
 195         /* Reset potential sums */
 196         velecsum         = _mm_setzero_pd();
 197         vvdwsum          = _mm_setzero_pd();
 198
 199         /* Start inner kernel loop */
 200         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 201         {
 202
 203             /* Get j neighbor index, and coordinate index */
 204             jnrA             = jjnr[jidx];
 205             jnrB             = jjnr[jidx+1];
 206             j_coord_offsetA  = DIM*jnrA;
 207             j_coord_offsetB  = DIM*jnrB;
 208
 209             /* load j atom coordinates */
 210             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 211                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 212
 213             /* Calculate displacement vector */
 214             dx00             = _mm_sub_pd(ix0,jx0);
 215             dy00             = _mm_sub_pd(iy0,jy0);
 216             dz00             = _mm_sub_pd(iz0,jz0);
 217             dx01             = _mm_sub_pd(ix0,jx1);
 218             dy01             = _mm_sub_pd(iy0,jy1);
 219             dz01             = _mm_sub_pd(iz0,jz1);
 220             dx02             = _mm_sub_pd(ix0,jx2);
 221             dy02             = _mm_sub_pd(iy0,jy2);
 222             dz02             = _mm_sub_pd(iz0,jz2);
 223             dx10             = _mm_sub_pd(ix1,jx0);
 224             dy10             = _mm_sub_pd(iy1,jy0);
 225             dz10             = _mm_sub_pd(iz1,jz0);
 226             dx11             = _mm_sub_pd(ix1,jx1);
 227             dy11             = _mm_sub_pd(iy1,jy1);
 228             dz11             = _mm_sub_pd(iz1,jz1);
 229             dx12             = _mm_sub_pd(ix1,jx2);
 230             dy12             = _mm_sub_pd(iy1,jy2);
 231             dz12             = _mm_sub_pd(iz1,jz2);
 232             dx20             = _mm_sub_pd(ix2,jx0);
 233             dy20             = _mm_sub_pd(iy2,jy0);
 234             dz20             = _mm_sub_pd(iz2,jz0);
 235             dx21             = _mm_sub_pd(ix2,jx1);
 236             dy21             = _mm_sub_pd(iy2,jy1);
 237             dz21             = _mm_sub_pd(iz2,jz1);
 238             dx22             = _mm_sub_pd(ix2,jx2);
 239             dy22             = _mm_sub_pd(iy2,jy2);
 240             dz22             = _mm_sub_pd(iz2,jz2);
 241
 242             /* Calculate squared distance and things based on it */
 243             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 244             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 245             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 246             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 247             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 248             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 249             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 250             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 251             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 252
 253             rinv00           = sse2_invsqrt_d(rsq00);
 254             rinv01           = sse2_invsqrt_d(rsq01);
 255             rinv02           = sse2_invsqrt_d(rsq02);
 256             rinv10           = sse2_invsqrt_d(rsq10);
 257             rinv11           = sse2_invsqrt_d(rsq11);
 258             rinv12           = sse2_invsqrt_d(rsq12);
 259             rinv20           = sse2_invsqrt_d(rsq20);
 260             rinv21           = sse2_invsqrt_d(rsq21);
 261             rinv22           = sse2_invsqrt_d(rsq22);
 262
 263             fjx0             = _mm_setzero_pd();
 264             fjy0             = _mm_setzero_pd();
 265             fjz0             = _mm_setzero_pd();
 266             fjx1             = _mm_setzero_pd();
 267             fjy1             = _mm_setzero_pd();
 268             fjz1             = _mm_setzero_pd();
 269             fjx2             = _mm_setzero_pd();
 270             fjy2             = _mm_setzero_pd();
 271             fjz2             = _mm_setzero_pd();
 272
 273             /**************************
 274              * CALCULATE INTERACTIONS *
 275              **************************/
 276
 277             r00              = _mm_mul_pd(rsq00,rinv00);
 278
 279             /* Calculate table index by multiplying r with table scale and truncate to integer */
 280             rt               = _mm_mul_pd(r00,vftabscale);
 281             vfitab           = _mm_cvttpd_epi32(rt);
 282             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 283             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 284
 285             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 286             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 287             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 288             GMX_MM_TRANSPOSE2_PD(Y,F);
 289             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 290             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 291             GMX_MM_TRANSPOSE2_PD(G,H);
 292             Heps             = _mm_mul_pd(vfeps,H);
 293             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 294             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 295             velec            = _mm_mul_pd(qq00,VV);
 296             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 297             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 298
 299             /* CUBIC SPLINE TABLE DISPERSION */
 300             vfitab           = _mm_add_epi32(vfitab,ifour);
 301             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 302             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 303             GMX_MM_TRANSPOSE2_PD(Y,F);
 304             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 305             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 306             GMX_MM_TRANSPOSE2_PD(G,H);
 307             Heps             = _mm_mul_pd(vfeps,H);
 308             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 309             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 310             vvdw6            = _mm_mul_pd(c6_00,VV);
 311             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 312             fvdw6            = _mm_mul_pd(c6_00,FF);
 313
 314             /* CUBIC SPLINE TABLE REPULSION */
 315             vfitab           = _mm_add_epi32(vfitab,ifour);
 316             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 317             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 318             GMX_MM_TRANSPOSE2_PD(Y,F);
 319             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 320             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 321             GMX_MM_TRANSPOSE2_PD(G,H);
 322             Heps             = _mm_mul_pd(vfeps,H);
 323             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 324             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 325             vvdw12           = _mm_mul_pd(c12_00,VV);
 326             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 327             fvdw12           = _mm_mul_pd(c12_00,FF);
 328             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 329             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 330
 331             /* Update potential sum for this i atom from the interaction with this j atom. */
 332             velecsum         = _mm_add_pd(velecsum,velec);
 333             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 334
 335             fscal            = _mm_add_pd(felec,fvdw);
 336
 337             /* Calculate temporary vectorial force */
 338             tx               = _mm_mul_pd(fscal,dx00);
 339             ty               = _mm_mul_pd(fscal,dy00);
 340             tz               = _mm_mul_pd(fscal,dz00);
 341
 342             /* Update vectorial force */
 343             fix0             = _mm_add_pd(fix0,tx);
 344             fiy0             = _mm_add_pd(fiy0,ty);
 345             fiz0             = _mm_add_pd(fiz0,tz);
 346
 347             fjx0             = _mm_add_pd(fjx0,tx);
 348             fjy0             = _mm_add_pd(fjy0,ty);
 349             fjz0             = _mm_add_pd(fjz0,tz);
 350
 351             /**************************
 352              * CALCULATE INTERACTIONS *
 353              **************************/
 354
 355             r01              = _mm_mul_pd(rsq01,rinv01);
 356
 357             /* Calculate table index by multiplying r with table scale and truncate to integer */
 358             rt               = _mm_mul_pd(r01,vftabscale);
 359             vfitab           = _mm_cvttpd_epi32(rt);
 360             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 361             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 362
 363             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 364             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 365             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 366             GMX_MM_TRANSPOSE2_PD(Y,F);
 367             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 368             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 369             GMX_MM_TRANSPOSE2_PD(G,H);
 370             Heps             = _mm_mul_pd(vfeps,H);
 371             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 372             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 373             velec            = _mm_mul_pd(qq01,VV);
 374             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 375             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 376
 377             /* Update potential sum for this i atom from the interaction with this j atom. */
 378             velecsum         = _mm_add_pd(velecsum,velec);
 379
 380             fscal            = felec;
 381
 382             /* Calculate temporary vectorial force */
 383             tx               = _mm_mul_pd(fscal,dx01);
 384             ty               = _mm_mul_pd(fscal,dy01);
 385             tz               = _mm_mul_pd(fscal,dz01);
 386
 387             /* Update vectorial force */
 388             fix0             = _mm_add_pd(fix0,tx);
 389             fiy0             = _mm_add_pd(fiy0,ty);
 390             fiz0             = _mm_add_pd(fiz0,tz);
 391
 392             fjx1             = _mm_add_pd(fjx1,tx);
 393             fjy1             = _mm_add_pd(fjy1,ty);
 394             fjz1             = _mm_add_pd(fjz1,tz);
 395
 396             /**************************
 397              * CALCULATE INTERACTIONS *
 398              **************************/
 399
 400             r02              = _mm_mul_pd(rsq02,rinv02);
 401
 402             /* Calculate table index by multiplying r with table scale and truncate to integer */
 403             rt               = _mm_mul_pd(r02,vftabscale);
 404             vfitab           = _mm_cvttpd_epi32(rt);
 405             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 406             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 407
 408             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 409             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 410             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 411             GMX_MM_TRANSPOSE2_PD(Y,F);
 412             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 413             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 414             GMX_MM_TRANSPOSE2_PD(G,H);
 415             Heps             = _mm_mul_pd(vfeps,H);
 416             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 417             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 418             velec            = _mm_mul_pd(qq02,VV);
 419             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 420             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 421
 422             /* Update potential sum for this i atom from the interaction with this j atom. */
 423             velecsum         = _mm_add_pd(velecsum,velec);
 424
 425             fscal            = felec;
 426
 427             /* Calculate temporary vectorial force */
 428             tx               = _mm_mul_pd(fscal,dx02);
 429             ty               = _mm_mul_pd(fscal,dy02);
 430             tz               = _mm_mul_pd(fscal,dz02);
 431
 432             /* Update vectorial force */
 433             fix0             = _mm_add_pd(fix0,tx);
 434             fiy0             = _mm_add_pd(fiy0,ty);
 435             fiz0             = _mm_add_pd(fiz0,tz);
 436
 437             fjx2             = _mm_add_pd(fjx2,tx);
 438             fjy2             = _mm_add_pd(fjy2,ty);
 439             fjz2             = _mm_add_pd(fjz2,tz);
 440
 441             /**************************
 442              * CALCULATE INTERACTIONS *
 443              **************************/
 444
 445             r10              = _mm_mul_pd(rsq10,rinv10);
 446
 447             /* Calculate table index by multiplying r with table scale and truncate to integer */
 448             rt               = _mm_mul_pd(r10,vftabscale);
 449             vfitab           = _mm_cvttpd_epi32(rt);
 450             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 451             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 452
 453             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 454             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 455             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 456             GMX_MM_TRANSPOSE2_PD(Y,F);
 457             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 458             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 459             GMX_MM_TRANSPOSE2_PD(G,H);
 460             Heps             = _mm_mul_pd(vfeps,H);
 461             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 462             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 463             velec            = _mm_mul_pd(qq10,VV);
 464             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 465             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 466
 467             /* Update potential sum for this i atom from the interaction with this j atom. */
 468             velecsum         = _mm_add_pd(velecsum,velec);
 469
 470             fscal            = felec;
 471
 472             /* Calculate temporary vectorial force */
 473             tx               = _mm_mul_pd(fscal,dx10);
 474             ty               = _mm_mul_pd(fscal,dy10);
 475             tz               = _mm_mul_pd(fscal,dz10);
 476
 477             /* Update vectorial force */
 478             fix1             = _mm_add_pd(fix1,tx);
 479             fiy1             = _mm_add_pd(fiy1,ty);
 480             fiz1             = _mm_add_pd(fiz1,tz);
 481
 482             fjx0             = _mm_add_pd(fjx0,tx);
 483             fjy0             = _mm_add_pd(fjy0,ty);
 484             fjz0             = _mm_add_pd(fjz0,tz);
 485
 486             /**************************
 487              * CALCULATE INTERACTIONS *
 488              **************************/
 489
 490             r11              = _mm_mul_pd(rsq11,rinv11);
 491
 492             /* Calculate table index by multiplying r with table scale and truncate to integer */
 493             rt               = _mm_mul_pd(r11,vftabscale);
 494             vfitab           = _mm_cvttpd_epi32(rt);
 495             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 496             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 497
 498             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 499             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 500             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 501             GMX_MM_TRANSPOSE2_PD(Y,F);
 502             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 503             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 504             GMX_MM_TRANSPOSE2_PD(G,H);
 505             Heps             = _mm_mul_pd(vfeps,H);
 506             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 507             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 508             velec            = _mm_mul_pd(qq11,VV);
 509             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 510             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 511
 512             /* Update potential sum for this i atom from the interaction with this j atom. */
 513             velecsum         = _mm_add_pd(velecsum,velec);
 514
 515             fscal            = felec;
 516
 517             /* Calculate temporary vectorial force */
 518             tx               = _mm_mul_pd(fscal,dx11);
 519             ty               = _mm_mul_pd(fscal,dy11);
 520             tz               = _mm_mul_pd(fscal,dz11);
 521
 522             /* Update vectorial force */
 523             fix1             = _mm_add_pd(fix1,tx);
 524             fiy1             = _mm_add_pd(fiy1,ty);
 525             fiz1             = _mm_add_pd(fiz1,tz);
 526
 527             fjx1             = _mm_add_pd(fjx1,tx);
 528             fjy1             = _mm_add_pd(fjy1,ty);
 529             fjz1             = _mm_add_pd(fjz1,tz);
 530
 531             /**************************
 532              * CALCULATE INTERACTIONS *
 533              **************************/
 534
 535             r12              = _mm_mul_pd(rsq12,rinv12);
 536
 537             /* Calculate table index by multiplying r with table scale and truncate to integer */
 538             rt               = _mm_mul_pd(r12,vftabscale);
 539             vfitab           = _mm_cvttpd_epi32(rt);
 540             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 541             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 542
 543             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 544             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 545             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 546             GMX_MM_TRANSPOSE2_PD(Y,F);
 547             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 548             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 549             GMX_MM_TRANSPOSE2_PD(G,H);
 550             Heps             = _mm_mul_pd(vfeps,H);
 551             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 552             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 553             velec            = _mm_mul_pd(qq12,VV);
 554             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 555             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 556
 557             /* Update potential sum for this i atom from the interaction with this j atom. */
 558             velecsum         = _mm_add_pd(velecsum,velec);
 559
 560             fscal            = felec;
 561
 562             /* Calculate temporary vectorial force */
 563             tx               = _mm_mul_pd(fscal,dx12);
 564             ty               = _mm_mul_pd(fscal,dy12);
 565             tz               = _mm_mul_pd(fscal,dz12);
 566
 567             /* Update vectorial force */
 568             fix1             = _mm_add_pd(fix1,tx);
 569             fiy1             = _mm_add_pd(fiy1,ty);
 570             fiz1             = _mm_add_pd(fiz1,tz);
 571
 572             fjx2             = _mm_add_pd(fjx2,tx);
 573             fjy2             = _mm_add_pd(fjy2,ty);
 574             fjz2             = _mm_add_pd(fjz2,tz);
 575
 576             /**************************
 577              * CALCULATE INTERACTIONS *
 578              **************************/
 579
 580             r20              = _mm_mul_pd(rsq20,rinv20);
 581
 582             /* Calculate table index by multiplying r with table scale and truncate to integer */
 583             rt               = _mm_mul_pd(r20,vftabscale);
 584             vfitab           = _mm_cvttpd_epi32(rt);
 585             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 586             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 587
 588             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 589             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 590             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 591             GMX_MM_TRANSPOSE2_PD(Y,F);
 592             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 593             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 594             GMX_MM_TRANSPOSE2_PD(G,H);
 595             Heps             = _mm_mul_pd(vfeps,H);
 596             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 597             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 598             velec            = _mm_mul_pd(qq20,VV);
 599             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 600             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
 601
 602             /* Update potential sum for this i atom from the interaction with this j atom. */
 603             velecsum         = _mm_add_pd(velecsum,velec);
 604
 605             fscal            = felec;
 606
 607             /* Calculate temporary vectorial force */
 608             tx               = _mm_mul_pd(fscal,dx20);
 609             ty               = _mm_mul_pd(fscal,dy20);
 610             tz               = _mm_mul_pd(fscal,dz20);
 611
 612             /* Update vectorial force */
 613             fix2             = _mm_add_pd(fix2,tx);
 614             fiy2             = _mm_add_pd(fiy2,ty);
 615             fiz2             = _mm_add_pd(fiz2,tz);
 616
 617             fjx0             = _mm_add_pd(fjx0,tx);
 618             fjy0             = _mm_add_pd(fjy0,ty);
 619             fjz0             = _mm_add_pd(fjz0,tz);
 620
 621             /**************************
 622              * CALCULATE INTERACTIONS *
 623              **************************/
 624
 625             r21              = _mm_mul_pd(rsq21,rinv21);
 626
 627             /* Calculate table index by multiplying r with table scale and truncate to integer */
 628             rt               = _mm_mul_pd(r21,vftabscale);
 629             vfitab           = _mm_cvttpd_epi32(rt);
 630             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 631             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 632
 633             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 634             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 635             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 636             GMX_MM_TRANSPOSE2_PD(Y,F);
 637             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 638             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 639             GMX_MM_TRANSPOSE2_PD(G,H);
 640             Heps             = _mm_mul_pd(vfeps,H);
 641             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 642             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 643             velec            = _mm_mul_pd(qq21,VV);
 644             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 645             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 646
 647             /* Update potential sum for this i atom from the interaction with this j atom. */
 648             velecsum         = _mm_add_pd(velecsum,velec);
 649
 650             fscal            = felec;
 651
 652             /* Calculate temporary vectorial force */
 653             tx               = _mm_mul_pd(fscal,dx21);
 654             ty               = _mm_mul_pd(fscal,dy21);
 655             tz               = _mm_mul_pd(fscal,dz21);
 656
 657             /* Update vectorial force */
 658             fix2             = _mm_add_pd(fix2,tx);
 659             fiy2             = _mm_add_pd(fiy2,ty);
 660             fiz2             = _mm_add_pd(fiz2,tz);
 661
 662             fjx1             = _mm_add_pd(fjx1,tx);
 663             fjy1             = _mm_add_pd(fjy1,ty);
 664             fjz1             = _mm_add_pd(fjz1,tz);
 665
 666             /**************************
 667              * CALCULATE INTERACTIONS *
 668              **************************/
 669
 670             r22              = _mm_mul_pd(rsq22,rinv22);
 671
 672             /* Calculate table index by multiplying r with table scale and truncate to integer */
 673             rt               = _mm_mul_pd(r22,vftabscale);
 674             vfitab           = _mm_cvttpd_epi32(rt);
 675             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 676             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 677
 678             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 679             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 680             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 681             GMX_MM_TRANSPOSE2_PD(Y,F);
 682             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 683             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 684             GMX_MM_TRANSPOSE2_PD(G,H);
 685             Heps             = _mm_mul_pd(vfeps,H);
 686             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 687             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 688             velec            = _mm_mul_pd(qq22,VV);
 689             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 690             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 691
 692             /* Update potential sum for this i atom from the interaction with this j atom. */
 693             velecsum         = _mm_add_pd(velecsum,velec);
 694
 695             fscal            = felec;
 696
 697             /* Calculate temporary vectorial force */
 698             tx               = _mm_mul_pd(fscal,dx22);
 699             ty               = _mm_mul_pd(fscal,dy22);
 700             tz               = _mm_mul_pd(fscal,dz22);
 701
 702             /* Update vectorial force */
 703             fix2             = _mm_add_pd(fix2,tx);
 704             fiy2             = _mm_add_pd(fiy2,ty);
 705             fiz2             = _mm_add_pd(fiz2,tz);
 706
 707             fjx2             = _mm_add_pd(fjx2,tx);
 708             fjy2             = _mm_add_pd(fjy2,ty);
 709             fjz2             = _mm_add_pd(fjz2,tz);
 710
 711             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 712
 713             /* Inner loop uses 417 flops */
 714         }
 715
 716         if(jidx<j_index_end)
 717         {
 718
 719             jnrA             = jjnr[jidx];
 720             j_coord_offsetA  = DIM*jnrA;
 721
 722             /* load j atom coordinates */
 723             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 724                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 725
 726             /* Calculate displacement vector */
 727             dx00             = _mm_sub_pd(ix0,jx0);
 728             dy00             = _mm_sub_pd(iy0,jy0);
 729             dz00             = _mm_sub_pd(iz0,jz0);
 730             dx01             = _mm_sub_pd(ix0,jx1);
 731             dy01             = _mm_sub_pd(iy0,jy1);
 732             dz01             = _mm_sub_pd(iz0,jz1);
 733             dx02             = _mm_sub_pd(ix0,jx2);
 734             dy02             = _mm_sub_pd(iy0,jy2);
 735             dz02             = _mm_sub_pd(iz0,jz2);
 736             dx10             = _mm_sub_pd(ix1,jx0);
 737             dy10             = _mm_sub_pd(iy1,jy0);
 738             dz10             = _mm_sub_pd(iz1,jz0);
 739             dx11             = _mm_sub_pd(ix1,jx1);
 740             dy11             = _mm_sub_pd(iy1,jy1);
 741             dz11             = _mm_sub_pd(iz1,jz1);
 742             dx12             = _mm_sub_pd(ix1,jx2);
 743             dy12             = _mm_sub_pd(iy1,jy2);
 744             dz12             = _mm_sub_pd(iz1,jz2);
 745             dx20             = _mm_sub_pd(ix2,jx0);
 746             dy20             = _mm_sub_pd(iy2,jy0);
 747             dz20             = _mm_sub_pd(iz2,jz0);
 748             dx21             = _mm_sub_pd(ix2,jx1);
 749             dy21             = _mm_sub_pd(iy2,jy1);
 750             dz21             = _mm_sub_pd(iz2,jz1);
 751             dx22             = _mm_sub_pd(ix2,jx2);
 752             dy22             = _mm_sub_pd(iy2,jy2);
 753             dz22             = _mm_sub_pd(iz2,jz2);
 754
 755             /* Calculate squared distance and things based on it */
 756             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 757             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 758             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 759             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 760             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 761             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 762             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 763             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 764             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 765
 766             rinv00           = sse2_invsqrt_d(rsq00);
 767             rinv01           = sse2_invsqrt_d(rsq01);
 768             rinv02           = sse2_invsqrt_d(rsq02);
 769             rinv10           = sse2_invsqrt_d(rsq10);
 770             rinv11           = sse2_invsqrt_d(rsq11);
 771             rinv12           = sse2_invsqrt_d(rsq12);
 772             rinv20           = sse2_invsqrt_d(rsq20);
 773             rinv21           = sse2_invsqrt_d(rsq21);
 774             rinv22           = sse2_invsqrt_d(rsq22);
 775
 776             fjx0             = _mm_setzero_pd();
 777             fjy0             = _mm_setzero_pd();
 778             fjz0             = _mm_setzero_pd();
 779             fjx1             = _mm_setzero_pd();
 780             fjy1             = _mm_setzero_pd();
 781             fjz1             = _mm_setzero_pd();
 782             fjx2             = _mm_setzero_pd();
 783             fjy2             = _mm_setzero_pd();
 784             fjz2             = _mm_setzero_pd();
 785
 786             /**************************
 787              * CALCULATE INTERACTIONS *
 788              **************************/
 789
 790             r00              = _mm_mul_pd(rsq00,rinv00);
 791
 792             /* Calculate table index by multiplying r with table scale and truncate to integer */
 793             rt               = _mm_mul_pd(r00,vftabscale);
 794             vfitab           = _mm_cvttpd_epi32(rt);
 795             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 796             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 797
 798             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 799             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 800             F                = _mm_setzero_pd();
 801             GMX_MM_TRANSPOSE2_PD(Y,F);
 802             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 803             H                = _mm_setzero_pd();
 804             GMX_MM_TRANSPOSE2_PD(G,H);
 805             Heps             = _mm_mul_pd(vfeps,H);
 806             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 807             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 808             velec            = _mm_mul_pd(qq00,VV);
 809             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 810             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 811
 812             /* CUBIC SPLINE TABLE DISPERSION */
 813             vfitab           = _mm_add_epi32(vfitab,ifour);
 814             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 815             F                = _mm_setzero_pd();
 816             GMX_MM_TRANSPOSE2_PD(Y,F);
 817             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 818             H                = _mm_setzero_pd();
 819             GMX_MM_TRANSPOSE2_PD(G,H);
 820             Heps             = _mm_mul_pd(vfeps,H);
 821             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 822             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 823             vvdw6            = _mm_mul_pd(c6_00,VV);
 824             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 825             fvdw6            = _mm_mul_pd(c6_00,FF);
 826
 827             /* CUBIC SPLINE TABLE REPULSION */
 828             vfitab           = _mm_add_epi32(vfitab,ifour);
 829             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 830             F                = _mm_setzero_pd();
 831             GMX_MM_TRANSPOSE2_PD(Y,F);
 832             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 833             H                = _mm_setzero_pd();
 834             GMX_MM_TRANSPOSE2_PD(G,H);
 835             Heps             = _mm_mul_pd(vfeps,H);
 836             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 837             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 838             vvdw12           = _mm_mul_pd(c12_00,VV);
 839             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 840             fvdw12           = _mm_mul_pd(c12_00,FF);
 841             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 842             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 843
 844             /* Update potential sum for this i atom from the interaction with this j atom. */
 845             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 846             velecsum         = _mm_add_pd(velecsum,velec);
 847             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 848             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 849
 850             fscal            = _mm_add_pd(felec,fvdw);
 851
 852             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 853
 854             /* Calculate temporary vectorial force */
 855             tx               = _mm_mul_pd(fscal,dx00);
 856             ty               = _mm_mul_pd(fscal,dy00);
 857             tz               = _mm_mul_pd(fscal,dz00);
 858
 859             /* Update vectorial force */
 860             fix0             = _mm_add_pd(fix0,tx);
 861             fiy0             = _mm_add_pd(fiy0,ty);
 862             fiz0             = _mm_add_pd(fiz0,tz);
 863
 864             fjx0             = _mm_add_pd(fjx0,tx);
 865             fjy0             = _mm_add_pd(fjy0,ty);
 866             fjz0             = _mm_add_pd(fjz0,tz);
 867
 868             /**************************
 869              * CALCULATE INTERACTIONS *
 870              **************************/
 871
 872             r01              = _mm_mul_pd(rsq01,rinv01);
 873
 874             /* Calculate table index by multiplying r with table scale and truncate to integer */
 875             rt               = _mm_mul_pd(r01,vftabscale);
 876             vfitab           = _mm_cvttpd_epi32(rt);
 877             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 878             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 879
 880             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 881             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 882             F                = _mm_setzero_pd();
 883             GMX_MM_TRANSPOSE2_PD(Y,F);
 884             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 885             H                = _mm_setzero_pd();
 886             GMX_MM_TRANSPOSE2_PD(G,H);
 887             Heps             = _mm_mul_pd(vfeps,H);
 888             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 889             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 890             velec            = _mm_mul_pd(qq01,VV);
 891             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 892             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 893
 894             /* Update potential sum for this i atom from the interaction with this j atom. */
 895             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 896             velecsum         = _mm_add_pd(velecsum,velec);
 897
 898             fscal            = felec;
 899
 900             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 901
 902             /* Calculate temporary vectorial force */
 903             tx               = _mm_mul_pd(fscal,dx01);
 904             ty               = _mm_mul_pd(fscal,dy01);
 905             tz               = _mm_mul_pd(fscal,dz01);
 906
 907             /* Update vectorial force */
 908             fix0             = _mm_add_pd(fix0,tx);
 909             fiy0             = _mm_add_pd(fiy0,ty);
 910             fiz0             = _mm_add_pd(fiz0,tz);
 911
 912             fjx1             = _mm_add_pd(fjx1,tx);
 913             fjy1             = _mm_add_pd(fjy1,ty);
 914             fjz1             = _mm_add_pd(fjz1,tz);
 915
 916             /**************************
 917              * CALCULATE INTERACTIONS *
 918              **************************/
 919
 920             r02              = _mm_mul_pd(rsq02,rinv02);
 921
 922             /* Calculate table index by multiplying r with table scale and truncate to integer */
 923             rt               = _mm_mul_pd(r02,vftabscale);
 924             vfitab           = _mm_cvttpd_epi32(rt);
 925             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 926             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 927
 928             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 929             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 930             F                = _mm_setzero_pd();
 931             GMX_MM_TRANSPOSE2_PD(Y,F);
 932             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 933             H                = _mm_setzero_pd();
 934             GMX_MM_TRANSPOSE2_PD(G,H);
 935             Heps             = _mm_mul_pd(vfeps,H);
 936             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 937             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 938             velec            = _mm_mul_pd(qq02,VV);
 939             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 940             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 941
 942             /* Update potential sum for this i atom from the interaction with this j atom. */
 943             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 944             velecsum         = _mm_add_pd(velecsum,velec);
 945
 946             fscal            = felec;
 947
 948             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 949
 950             /* Calculate temporary vectorial force */
 951             tx               = _mm_mul_pd(fscal,dx02);
 952             ty               = _mm_mul_pd(fscal,dy02);
 953             tz               = _mm_mul_pd(fscal,dz02);
 954
 955             /* Update vectorial force */
 956             fix0             = _mm_add_pd(fix0,tx);
 957             fiy0             = _mm_add_pd(fiy0,ty);
 958             fiz0             = _mm_add_pd(fiz0,tz);
 959
 960             fjx2             = _mm_add_pd(fjx2,tx);
 961             fjy2             = _mm_add_pd(fjy2,ty);
 962             fjz2             = _mm_add_pd(fjz2,tz);
 963
 964             /**************************
 965              * CALCULATE INTERACTIONS *
 966              **************************/
 967
 968             r10              = _mm_mul_pd(rsq10,rinv10);
 969
 970             /* Calculate table index by multiplying r with table scale and truncate to integer */
 971             rt               = _mm_mul_pd(r10,vftabscale);
 972             vfitab           = _mm_cvttpd_epi32(rt);
 973             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 974             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 975
 976             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 977             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 978             F                = _mm_setzero_pd();
 979             GMX_MM_TRANSPOSE2_PD(Y,F);
 980             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 981             H                = _mm_setzero_pd();
 982             GMX_MM_TRANSPOSE2_PD(G,H);
 983             Heps             = _mm_mul_pd(vfeps,H);
 984             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 985             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 986             velec            = _mm_mul_pd(qq10,VV);
 987             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 988             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 989
 990             /* Update potential sum for this i atom from the interaction with this j atom. */
 991             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 992             velecsum         = _mm_add_pd(velecsum,velec);
 993
 994             fscal            = felec;
 995
 996             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 997
 998             /* Calculate temporary vectorial force */
 999             tx               = _mm_mul_pd(fscal,dx10);
1000             ty               = _mm_mul_pd(fscal,dy10);
1001             tz               = _mm_mul_pd(fscal,dz10);
1002
1003             /* Update vectorial force */
1004             fix1             = _mm_add_pd(fix1,tx);
1005             fiy1             = _mm_add_pd(fiy1,ty);
1006             fiz1             = _mm_add_pd(fiz1,tz);
1007
1008             fjx0             = _mm_add_pd(fjx0,tx);
1009             fjy0             = _mm_add_pd(fjy0,ty);
1010             fjz0             = _mm_add_pd(fjz0,tz);
1011
1012             /**************************
1013              * CALCULATE INTERACTIONS *
1014              **************************/
1015
1016             r11              = _mm_mul_pd(rsq11,rinv11);
1017
1018             /* Calculate table index by multiplying r with table scale and truncate to integer */
1019             rt               = _mm_mul_pd(r11,vftabscale);
1020             vfitab           = _mm_cvttpd_epi32(rt);
1021             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1022             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1023
1024             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1025             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1026             F                = _mm_setzero_pd();
1027             GMX_MM_TRANSPOSE2_PD(Y,F);
1028             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1029             H                = _mm_setzero_pd();
1030             GMX_MM_TRANSPOSE2_PD(G,H);
1031             Heps             = _mm_mul_pd(vfeps,H);
1032             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1033             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1034             velec            = _mm_mul_pd(qq11,VV);
1035             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1036             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1037
1038             /* Update potential sum for this i atom from the interaction with this j atom. */
1039             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1040             velecsum         = _mm_add_pd(velecsum,velec);
1041
1042             fscal            = felec;
1043
1044             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1045
1046             /* Calculate temporary vectorial force */
1047             tx               = _mm_mul_pd(fscal,dx11);
1048             ty               = _mm_mul_pd(fscal,dy11);
1049             tz               = _mm_mul_pd(fscal,dz11);
1050
1051             /* Update vectorial force */
1052             fix1             = _mm_add_pd(fix1,tx);
1053             fiy1             = _mm_add_pd(fiy1,ty);
1054             fiz1             = _mm_add_pd(fiz1,tz);
1055
1056             fjx1             = _mm_add_pd(fjx1,tx);
1057             fjy1             = _mm_add_pd(fjy1,ty);
1058             fjz1             = _mm_add_pd(fjz1,tz);
1059
1060             /**************************
1061              * CALCULATE INTERACTIONS *
1062              **************************/
1063
1064             r12              = _mm_mul_pd(rsq12,rinv12);
1065
1066             /* Calculate table index by multiplying r with table scale and truncate to integer */
1067             rt               = _mm_mul_pd(r12,vftabscale);
1068             vfitab           = _mm_cvttpd_epi32(rt);
1069             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1070             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1071
1072             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1073             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1074             F                = _mm_setzero_pd();
1075             GMX_MM_TRANSPOSE2_PD(Y,F);
1076             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1077             H                = _mm_setzero_pd();
1078             GMX_MM_TRANSPOSE2_PD(G,H);
1079             Heps             = _mm_mul_pd(vfeps,H);
1080             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1081             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1082             velec            = _mm_mul_pd(qq12,VV);
1083             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1084             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1085
1086             /* Update potential sum for this i atom from the interaction with this j atom. */
1087             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1088             velecsum         = _mm_add_pd(velecsum,velec);
1089
1090             fscal            = felec;
1091
1092             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1093
1094             /* Calculate temporary vectorial force */
1095             tx               = _mm_mul_pd(fscal,dx12);
1096             ty               = _mm_mul_pd(fscal,dy12);
1097             tz               = _mm_mul_pd(fscal,dz12);
1098
1099             /* Update vectorial force */
1100             fix1             = _mm_add_pd(fix1,tx);
1101             fiy1             = _mm_add_pd(fiy1,ty);
1102             fiz1             = _mm_add_pd(fiz1,tz);
1103
1104             fjx2             = _mm_add_pd(fjx2,tx);
1105             fjy2             = _mm_add_pd(fjy2,ty);
1106             fjz2             = _mm_add_pd(fjz2,tz);
1107
1108             /**************************
1109              * CALCULATE INTERACTIONS *
1110              **************************/
1111
1112             r20              = _mm_mul_pd(rsq20,rinv20);
1113
1114             /* Calculate table index by multiplying r with table scale and truncate to integer */
1115             rt               = _mm_mul_pd(r20,vftabscale);
1116             vfitab           = _mm_cvttpd_epi32(rt);
1117             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1118             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1119
1120             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1121             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1122             F                = _mm_setzero_pd();
1123             GMX_MM_TRANSPOSE2_PD(Y,F);
1124             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1125             H                = _mm_setzero_pd();
1126             GMX_MM_TRANSPOSE2_PD(G,H);
1127             Heps             = _mm_mul_pd(vfeps,H);
1128             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1129             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1130             velec            = _mm_mul_pd(qq20,VV);
1131             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1132             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1133
1134             /* Update potential sum for this i atom from the interaction with this j atom. */
1135             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1136             velecsum         = _mm_add_pd(velecsum,velec);
1137
1138             fscal            = felec;
1139
1140             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1141
1142             /* Calculate temporary vectorial force */
1143             tx               = _mm_mul_pd(fscal,dx20);
1144             ty               = _mm_mul_pd(fscal,dy20);
1145             tz               = _mm_mul_pd(fscal,dz20);
1146
1147             /* Update vectorial force */
1148             fix2             = _mm_add_pd(fix2,tx);
1149             fiy2             = _mm_add_pd(fiy2,ty);
1150             fiz2             = _mm_add_pd(fiz2,tz);
1151
1152             fjx0             = _mm_add_pd(fjx0,tx);
1153             fjy0             = _mm_add_pd(fjy0,ty);
1154             fjz0             = _mm_add_pd(fjz0,tz);
1155
1156             /**************************
1157              * CALCULATE INTERACTIONS *
1158              **************************/
1159
1160             r21              = _mm_mul_pd(rsq21,rinv21);
1161
1162             /* Calculate table index by multiplying r with table scale and truncate to integer */
1163             rt               = _mm_mul_pd(r21,vftabscale);
1164             vfitab           = _mm_cvttpd_epi32(rt);
1165             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1166             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1167
1168             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1169             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1170             F                = _mm_setzero_pd();
1171             GMX_MM_TRANSPOSE2_PD(Y,F);
1172             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1173             H                = _mm_setzero_pd();
1174             GMX_MM_TRANSPOSE2_PD(G,H);
1175             Heps             = _mm_mul_pd(vfeps,H);
1176             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1177             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1178             velec            = _mm_mul_pd(qq21,VV);
1179             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1180             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1181
1182             /* Update potential sum for this i atom from the interaction with this j atom. */
1183             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1184             velecsum         = _mm_add_pd(velecsum,velec);
1185
1186             fscal            = felec;
1187
1188             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1189
1190             /* Calculate temporary vectorial force */
1191             tx               = _mm_mul_pd(fscal,dx21);
1192             ty               = _mm_mul_pd(fscal,dy21);
1193             tz               = _mm_mul_pd(fscal,dz21);
1194
1195             /* Update vectorial force */
1196             fix2             = _mm_add_pd(fix2,tx);
1197             fiy2             = _mm_add_pd(fiy2,ty);
1198             fiz2             = _mm_add_pd(fiz2,tz);
1199
1200             fjx1             = _mm_add_pd(fjx1,tx);
1201             fjy1             = _mm_add_pd(fjy1,ty);
1202             fjz1             = _mm_add_pd(fjz1,tz);
1203
1204             /**************************
1205              * CALCULATE INTERACTIONS *
1206              **************************/
1207
1208             r22              = _mm_mul_pd(rsq22,rinv22);
1209
1210             /* Calculate table index by multiplying r with table scale and truncate to integer */
1211             rt               = _mm_mul_pd(r22,vftabscale);
1212             vfitab           = _mm_cvttpd_epi32(rt);
1213             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1214             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1215
1216             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1217             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1218             F                = _mm_setzero_pd();
1219             GMX_MM_TRANSPOSE2_PD(Y,F);
1220             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1221             H                = _mm_setzero_pd();
1222             GMX_MM_TRANSPOSE2_PD(G,H);
1223             Heps             = _mm_mul_pd(vfeps,H);
1224             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1225             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1226             velec            = _mm_mul_pd(qq22,VV);
1227             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1228             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1229
1230             /* Update potential sum for this i atom from the interaction with this j atom. */
1231             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1232             velecsum         = _mm_add_pd(velecsum,velec);
1233
1234             fscal            = felec;
1235
1236             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1237
1238             /* Calculate temporary vectorial force */
1239             tx               = _mm_mul_pd(fscal,dx22);
1240             ty               = _mm_mul_pd(fscal,dy22);
1241             tz               = _mm_mul_pd(fscal,dz22);
1242
1243             /* Update vectorial force */
1244             fix2             = _mm_add_pd(fix2,tx);
1245             fiy2             = _mm_add_pd(fiy2,ty);
1246             fiz2             = _mm_add_pd(fiz2,tz);
1247
1248             fjx2             = _mm_add_pd(fjx2,tx);
1249             fjy2             = _mm_add_pd(fjy2,ty);
1250             fjz2             = _mm_add_pd(fjz2,tz);
1251
1252             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1253
1254             /* Inner loop uses 417 flops */
1255         }
1256
1257         /* End of innermost loop */
1258
1259         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1260                                               f+i_coord_offset,fshift+i_shift_offset);
1261
1262         ggid                        = gid[iidx];
1263         /* Update potential energies */
1264         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1265         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1266
1267         /* Increment number of inner iterations */
1268         inneriter                  += j_index_end - j_index_start;
1269
1270         /* Outer loop uses 20 flops */
1271     }
1272
1273     /* Increment number of outer iterations */
1274     outeriter        += nri;
1275
1276     /* Update outer/inner flops */
1277
1278     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*417);
1279 }
1280 /*
1281  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_F_sse2_double
1282  * Electrostatics interaction: CubicSplineTable
1283  * VdW interaction:            CubicSplineTable
1284  * Geometry:                   Water3-Water3
1285  * Calculate force/pot:        Force
1286  */
1287 void
1288 nb_kernel_ElecCSTab_VdwCSTab_GeomW3W3_F_sse2_double
1289                     (t_nblist                    * gmx_restrict       nlist,
1290                      rvec                        * gmx_restrict          xx,
1291                      rvec                        * gmx_restrict          ff,
1292                      struct t_forcerec           * gmx_restrict          fr,
1293                      t_mdatoms                   * gmx_restrict     mdatoms,
1294                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1295                      t_nrnb                      * gmx_restrict        nrnb)
1296 {
1297     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1298      * just 0 for non-waters.
1299      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1300      * jnr indices corresponding to data put in the four positions in the SIMD register.
1301      */
1302     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1303     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1304     int              jnrA,jnrB;
1305     int              j_coord_offsetA,j_coord_offsetB;
1306     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1307     real             rcutoff_scalar;
1308     real             *shiftvec,*fshift,*x,*f;
1309     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1310     int              vdwioffset0;
1311     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1312     int              vdwioffset1;
1313     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1314     int              vdwioffset2;
1315     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1316     int              vdwjidx0A,vdwjidx0B;
1317     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1318     int              vdwjidx1A,vdwjidx1B;
1319     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1320     int              vdwjidx2A,vdwjidx2B;
1321     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1322     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1323     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1324     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1325     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1326     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1327     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1328     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1329     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1330     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1331     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1332     real             *charge;
1333     int              nvdwtype;
1334     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1335     int              *vdwtype;
1336     real             *vdwparam;
1337     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1338     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1339     __m128i          vfitab;
1340     __m128i          ifour       = _mm_set1_epi32(4);
1341     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1342     real             *vftab;
1343     __m128d          dummy_mask,cutoff_mask;
1344     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1345     __m128d          one     = _mm_set1_pd(1.0);
1346     __m128d          two     = _mm_set1_pd(2.0);
1347     x                = xx[0];
1348     f                = ff[0];
1349
1350     nri              = nlist->nri;
1351     iinr             = nlist->iinr;
1352     jindex           = nlist->jindex;
1353     jjnr             = nlist->jjnr;
1354     shiftidx         = nlist->shift;
1355     gid              = nlist->gid;
1356     shiftvec         = fr->shift_vec[0];
1357     fshift           = fr->fshift[0];
1358     facel            = _mm_set1_pd(fr->ic->epsfac);
1359     charge           = mdatoms->chargeA;
1360     nvdwtype         = fr->ntype;
1361     vdwparam         = fr->nbfp;
1362     vdwtype          = mdatoms->typeA;
1363
1364     vftab            = kernel_data->table_elec_vdw->data;
1365     vftabscale       = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
1366
1367     /* Setup water-specific parameters */
1368     inr              = nlist->iinr[0];
1369     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1370     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1371     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1372     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1373
1374     jq0              = _mm_set1_pd(charge[inr+0]);
1375     jq1              = _mm_set1_pd(charge[inr+1]);
1376     jq2              = _mm_set1_pd(charge[inr+2]);
1377     vdwjidx0A        = 2*vdwtype[inr+0];
1378     qq00             = _mm_mul_pd(iq0,jq0);
1379     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1380     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1381     qq01             = _mm_mul_pd(iq0,jq1);
1382     qq02             = _mm_mul_pd(iq0,jq2);
1383     qq10             = _mm_mul_pd(iq1,jq0);
1384     qq11             = _mm_mul_pd(iq1,jq1);
1385     qq12             = _mm_mul_pd(iq1,jq2);
1386     qq20             = _mm_mul_pd(iq2,jq0);
1387     qq21             = _mm_mul_pd(iq2,jq1);
1388     qq22             = _mm_mul_pd(iq2,jq2);
1389
1390     /* Avoid stupid compiler warnings */
1391     jnrA = jnrB = 0;
1392     j_coord_offsetA = 0;
1393     j_coord_offsetB = 0;
1394
1395     outeriter        = 0;
1396     inneriter        = 0;
1397
1398     /* Start outer loop over neighborlists */
1399     for(iidx=0; iidx<nri; iidx++)
1400     {
1401         /* Load shift vector for this list */
1402         i_shift_offset   = DIM*shiftidx[iidx];
1403
1404         /* Load limits for loop over neighbors */
1405         j_index_start    = jindex[iidx];
1406         j_index_end      = jindex[iidx+1];
1407
1408         /* Get outer coordinate index */
1409         inr              = iinr[iidx];
1410         i_coord_offset   = DIM*inr;
1411
1412         /* Load i particle coords and add shift vector */
1413         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1414                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1415
1416         fix0             = _mm_setzero_pd();
1417         fiy0             = _mm_setzero_pd();
1418         fiz0             = _mm_setzero_pd();
1419         fix1             = _mm_setzero_pd();
1420         fiy1             = _mm_setzero_pd();
1421         fiz1             = _mm_setzero_pd();
1422         fix2             = _mm_setzero_pd();
1423         fiy2             = _mm_setzero_pd();
1424         fiz2             = _mm_setzero_pd();
1425
1426         /* Start inner kernel loop */
1427         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1428         {
1429
1430             /* Get j neighbor index, and coordinate index */
1431             jnrA             = jjnr[jidx];
1432             jnrB             = jjnr[jidx+1];
1433             j_coord_offsetA  = DIM*jnrA;
1434             j_coord_offsetB  = DIM*jnrB;
1435
1436             /* load j atom coordinates */
1437             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1438                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1439
1440             /* Calculate displacement vector */
1441             dx00             = _mm_sub_pd(ix0,jx0);
1442             dy00             = _mm_sub_pd(iy0,jy0);
1443             dz00             = _mm_sub_pd(iz0,jz0);
1444             dx01             = _mm_sub_pd(ix0,jx1);
1445             dy01             = _mm_sub_pd(iy0,jy1);
1446             dz01             = _mm_sub_pd(iz0,jz1);
1447             dx02             = _mm_sub_pd(ix0,jx2);
1448             dy02             = _mm_sub_pd(iy0,jy2);
1449             dz02             = _mm_sub_pd(iz0,jz2);
1450             dx10             = _mm_sub_pd(ix1,jx0);
1451             dy10             = _mm_sub_pd(iy1,jy0);
1452             dz10             = _mm_sub_pd(iz1,jz0);
1453             dx11             = _mm_sub_pd(ix1,jx1);
1454             dy11             = _mm_sub_pd(iy1,jy1);
1455             dz11             = _mm_sub_pd(iz1,jz1);
1456             dx12             = _mm_sub_pd(ix1,jx2);
1457             dy12             = _mm_sub_pd(iy1,jy2);
1458             dz12             = _mm_sub_pd(iz1,jz2);
1459             dx20             = _mm_sub_pd(ix2,jx0);
1460             dy20             = _mm_sub_pd(iy2,jy0);
1461             dz20             = _mm_sub_pd(iz2,jz0);
1462             dx21             = _mm_sub_pd(ix2,jx1);
1463             dy21             = _mm_sub_pd(iy2,jy1);
1464             dz21             = _mm_sub_pd(iz2,jz1);
1465             dx22             = _mm_sub_pd(ix2,jx2);
1466             dy22             = _mm_sub_pd(iy2,jy2);
1467             dz22             = _mm_sub_pd(iz2,jz2);
1468
1469             /* Calculate squared distance and things based on it */
1470             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1471             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1472             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1473             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1474             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1475             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1476             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1477             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1478             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1479
1480             rinv00           = sse2_invsqrt_d(rsq00);
1481             rinv01           = sse2_invsqrt_d(rsq01);
1482             rinv02           = sse2_invsqrt_d(rsq02);
1483             rinv10           = sse2_invsqrt_d(rsq10);
1484             rinv11           = sse2_invsqrt_d(rsq11);
1485             rinv12           = sse2_invsqrt_d(rsq12);
1486             rinv20           = sse2_invsqrt_d(rsq20);
1487             rinv21           = sse2_invsqrt_d(rsq21);
1488             rinv22           = sse2_invsqrt_d(rsq22);
1489
1490             fjx0             = _mm_setzero_pd();
1491             fjy0             = _mm_setzero_pd();
1492             fjz0             = _mm_setzero_pd();
1493             fjx1             = _mm_setzero_pd();
1494             fjy1             = _mm_setzero_pd();
1495             fjz1             = _mm_setzero_pd();
1496             fjx2             = _mm_setzero_pd();
1497             fjy2             = _mm_setzero_pd();
1498             fjz2             = _mm_setzero_pd();
1499
1500             /**************************
1501              * CALCULATE INTERACTIONS *
1502              **************************/
1503
1504             r00              = _mm_mul_pd(rsq00,rinv00);
1505
1506             /* Calculate table index by multiplying r with table scale and truncate to integer */
1507             rt               = _mm_mul_pd(r00,vftabscale);
1508             vfitab           = _mm_cvttpd_epi32(rt);
1509             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1510             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1511
1512             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1513             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1514             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1515             GMX_MM_TRANSPOSE2_PD(Y,F);
1516             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1517             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1518             GMX_MM_TRANSPOSE2_PD(G,H);
1519             Heps             = _mm_mul_pd(vfeps,H);
1520             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1521             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1522             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1523
1524             /* CUBIC SPLINE TABLE DISPERSION */
1525             vfitab           = _mm_add_epi32(vfitab,ifour);
1526             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1527             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1528             GMX_MM_TRANSPOSE2_PD(Y,F);
1529             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1530             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1531             GMX_MM_TRANSPOSE2_PD(G,H);
1532             Heps             = _mm_mul_pd(vfeps,H);
1533             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1534             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1535             fvdw6            = _mm_mul_pd(c6_00,FF);
1536
1537             /* CUBIC SPLINE TABLE REPULSION */
1538             vfitab           = _mm_add_epi32(vfitab,ifour);
1539             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1540             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1541             GMX_MM_TRANSPOSE2_PD(Y,F);
1542             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1543             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1544             GMX_MM_TRANSPOSE2_PD(G,H);
1545             Heps             = _mm_mul_pd(vfeps,H);
1546             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1547             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1548             fvdw12           = _mm_mul_pd(c12_00,FF);
1549             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1550
1551             fscal            = _mm_add_pd(felec,fvdw);
1552
1553             /* Calculate temporary vectorial force */
1554             tx               = _mm_mul_pd(fscal,dx00);
1555             ty               = _mm_mul_pd(fscal,dy00);
1556             tz               = _mm_mul_pd(fscal,dz00);
1557
1558             /* Update vectorial force */
1559             fix0             = _mm_add_pd(fix0,tx);
1560             fiy0             = _mm_add_pd(fiy0,ty);
1561             fiz0             = _mm_add_pd(fiz0,tz);
1562
1563             fjx0             = _mm_add_pd(fjx0,tx);
1564             fjy0             = _mm_add_pd(fjy0,ty);
1565             fjz0             = _mm_add_pd(fjz0,tz);
1566
1567             /**************************
1568              * CALCULATE INTERACTIONS *
1569              **************************/
1570
1571             r01              = _mm_mul_pd(rsq01,rinv01);
1572
1573             /* Calculate table index by multiplying r with table scale and truncate to integer */
1574             rt               = _mm_mul_pd(r01,vftabscale);
1575             vfitab           = _mm_cvttpd_epi32(rt);
1576             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1577             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1578
1579             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1580             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1581             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1582             GMX_MM_TRANSPOSE2_PD(Y,F);
1583             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1584             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1585             GMX_MM_TRANSPOSE2_PD(G,H);
1586             Heps             = _mm_mul_pd(vfeps,H);
1587             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1588             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1589             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1590
1591             fscal            = felec;
1592
1593             /* Calculate temporary vectorial force */
1594             tx               = _mm_mul_pd(fscal,dx01);
1595             ty               = _mm_mul_pd(fscal,dy01);
1596             tz               = _mm_mul_pd(fscal,dz01);
1597
1598             /* Update vectorial force */
1599             fix0             = _mm_add_pd(fix0,tx);
1600             fiy0             = _mm_add_pd(fiy0,ty);
1601             fiz0             = _mm_add_pd(fiz0,tz);
1602
1603             fjx1             = _mm_add_pd(fjx1,tx);
1604             fjy1             = _mm_add_pd(fjy1,ty);
1605             fjz1             = _mm_add_pd(fjz1,tz);
1606
1607             /**************************
1608              * CALCULATE INTERACTIONS *
1609              **************************/
1610
1611             r02              = _mm_mul_pd(rsq02,rinv02);
1612
1613             /* Calculate table index by multiplying r with table scale and truncate to integer */
1614             rt               = _mm_mul_pd(r02,vftabscale);
1615             vfitab           = _mm_cvttpd_epi32(rt);
1616             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1617             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1618
1619             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1620             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1621             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1622             GMX_MM_TRANSPOSE2_PD(Y,F);
1623             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1624             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1625             GMX_MM_TRANSPOSE2_PD(G,H);
1626             Heps             = _mm_mul_pd(vfeps,H);
1627             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1628             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1629             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1630
1631             fscal            = felec;
1632
1633             /* Calculate temporary vectorial force */
1634             tx               = _mm_mul_pd(fscal,dx02);
1635             ty               = _mm_mul_pd(fscal,dy02);
1636             tz               = _mm_mul_pd(fscal,dz02);
1637
1638             /* Update vectorial force */
1639             fix0             = _mm_add_pd(fix0,tx);
1640             fiy0             = _mm_add_pd(fiy0,ty);
1641             fiz0             = _mm_add_pd(fiz0,tz);
1642
1643             fjx2             = _mm_add_pd(fjx2,tx);
1644             fjy2             = _mm_add_pd(fjy2,ty);
1645             fjz2             = _mm_add_pd(fjz2,tz);
1646
1647             /**************************
1648              * CALCULATE INTERACTIONS *
1649              **************************/
1650
1651             r10              = _mm_mul_pd(rsq10,rinv10);
1652
1653             /* Calculate table index by multiplying r with table scale and truncate to integer */
1654             rt               = _mm_mul_pd(r10,vftabscale);
1655             vfitab           = _mm_cvttpd_epi32(rt);
1656             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1657             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1658
1659             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1660             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1661             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1662             GMX_MM_TRANSPOSE2_PD(Y,F);
1663             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1664             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1665             GMX_MM_TRANSPOSE2_PD(G,H);
1666             Heps             = _mm_mul_pd(vfeps,H);
1667             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1668             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1669             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1670
1671             fscal            = felec;
1672
1673             /* Calculate temporary vectorial force */
1674             tx               = _mm_mul_pd(fscal,dx10);
1675             ty               = _mm_mul_pd(fscal,dy10);
1676             tz               = _mm_mul_pd(fscal,dz10);
1677
1678             /* Update vectorial force */
1679             fix1             = _mm_add_pd(fix1,tx);
1680             fiy1             = _mm_add_pd(fiy1,ty);
1681             fiz1             = _mm_add_pd(fiz1,tz);
1682
1683             fjx0             = _mm_add_pd(fjx0,tx);
1684             fjy0             = _mm_add_pd(fjy0,ty);
1685             fjz0             = _mm_add_pd(fjz0,tz);
1686
1687             /**************************
1688              * CALCULATE INTERACTIONS *
1689              **************************/
1690
1691             r11              = _mm_mul_pd(rsq11,rinv11);
1692
1693             /* Calculate table index by multiplying r with table scale and truncate to integer */
1694             rt               = _mm_mul_pd(r11,vftabscale);
1695             vfitab           = _mm_cvttpd_epi32(rt);
1696             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1697             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1698
1699             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1700             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1701             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1702             GMX_MM_TRANSPOSE2_PD(Y,F);
1703             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1704             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1705             GMX_MM_TRANSPOSE2_PD(G,H);
1706             Heps             = _mm_mul_pd(vfeps,H);
1707             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1708             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1709             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1710
1711             fscal            = felec;
1712
1713             /* Calculate temporary vectorial force */
1714             tx               = _mm_mul_pd(fscal,dx11);
1715             ty               = _mm_mul_pd(fscal,dy11);
1716             tz               = _mm_mul_pd(fscal,dz11);
1717
1718             /* Update vectorial force */
1719             fix1             = _mm_add_pd(fix1,tx);
1720             fiy1             = _mm_add_pd(fiy1,ty);
1721             fiz1             = _mm_add_pd(fiz1,tz);
1722
1723             fjx1             = _mm_add_pd(fjx1,tx);
1724             fjy1             = _mm_add_pd(fjy1,ty);
1725             fjz1             = _mm_add_pd(fjz1,tz);
1726
1727             /**************************
1728              * CALCULATE INTERACTIONS *
1729              **************************/
1730
1731             r12              = _mm_mul_pd(rsq12,rinv12);
1732
1733             /* Calculate table index by multiplying r with table scale and truncate to integer */
1734             rt               = _mm_mul_pd(r12,vftabscale);
1735             vfitab           = _mm_cvttpd_epi32(rt);
1736             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1737             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1738
1739             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1740             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1741             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1742             GMX_MM_TRANSPOSE2_PD(Y,F);
1743             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1744             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1745             GMX_MM_TRANSPOSE2_PD(G,H);
1746             Heps             = _mm_mul_pd(vfeps,H);
1747             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1748             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1749             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1750
1751             fscal            = felec;
1752
1753             /* Calculate temporary vectorial force */
1754             tx               = _mm_mul_pd(fscal,dx12);
1755             ty               = _mm_mul_pd(fscal,dy12);
1756             tz               = _mm_mul_pd(fscal,dz12);
1757
1758             /* Update vectorial force */
1759             fix1             = _mm_add_pd(fix1,tx);
1760             fiy1             = _mm_add_pd(fiy1,ty);
1761             fiz1             = _mm_add_pd(fiz1,tz);
1762
1763             fjx2             = _mm_add_pd(fjx2,tx);
1764             fjy2             = _mm_add_pd(fjy2,ty);
1765             fjz2             = _mm_add_pd(fjz2,tz);
1766
1767             /**************************
1768              * CALCULATE INTERACTIONS *
1769              **************************/
1770
1771             r20              = _mm_mul_pd(rsq20,rinv20);
1772
1773             /* Calculate table index by multiplying r with table scale and truncate to integer */
1774             rt               = _mm_mul_pd(r20,vftabscale);
1775             vfitab           = _mm_cvttpd_epi32(rt);
1776             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1777             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1778
1779             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1780             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1781             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1782             GMX_MM_TRANSPOSE2_PD(Y,F);
1783             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1784             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1785             GMX_MM_TRANSPOSE2_PD(G,H);
1786             Heps             = _mm_mul_pd(vfeps,H);
1787             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1788             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1789             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1790
1791             fscal            = felec;
1792
1793             /* Calculate temporary vectorial force */
1794             tx               = _mm_mul_pd(fscal,dx20);
1795             ty               = _mm_mul_pd(fscal,dy20);
1796             tz               = _mm_mul_pd(fscal,dz20);
1797
1798             /* Update vectorial force */
1799             fix2             = _mm_add_pd(fix2,tx);
1800             fiy2             = _mm_add_pd(fiy2,ty);
1801             fiz2             = _mm_add_pd(fiz2,tz);
1802
1803             fjx0             = _mm_add_pd(fjx0,tx);
1804             fjy0             = _mm_add_pd(fjy0,ty);
1805             fjz0             = _mm_add_pd(fjz0,tz);
1806
1807             /**************************
1808              * CALCULATE INTERACTIONS *
1809              **************************/
1810
1811             r21              = _mm_mul_pd(rsq21,rinv21);
1812
1813             /* Calculate table index by multiplying r with table scale and truncate to integer */
1814             rt               = _mm_mul_pd(r21,vftabscale);
1815             vfitab           = _mm_cvttpd_epi32(rt);
1816             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1817             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1818
1819             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1820             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1821             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1822             GMX_MM_TRANSPOSE2_PD(Y,F);
1823             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1824             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1825             GMX_MM_TRANSPOSE2_PD(G,H);
1826             Heps             = _mm_mul_pd(vfeps,H);
1827             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1828             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1829             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1830
1831             fscal            = felec;
1832
1833             /* Calculate temporary vectorial force */
1834             tx               = _mm_mul_pd(fscal,dx21);
1835             ty               = _mm_mul_pd(fscal,dy21);
1836             tz               = _mm_mul_pd(fscal,dz21);
1837
1838             /* Update vectorial force */
1839             fix2             = _mm_add_pd(fix2,tx);
1840             fiy2             = _mm_add_pd(fiy2,ty);
1841             fiz2             = _mm_add_pd(fiz2,tz);
1842
1843             fjx1             = _mm_add_pd(fjx1,tx);
1844             fjy1             = _mm_add_pd(fjy1,ty);
1845             fjz1             = _mm_add_pd(fjz1,tz);
1846
1847             /**************************
1848              * CALCULATE INTERACTIONS *
1849              **************************/
1850
1851             r22              = _mm_mul_pd(rsq22,rinv22);
1852
1853             /* Calculate table index by multiplying r with table scale and truncate to integer */
1854             rt               = _mm_mul_pd(r22,vftabscale);
1855             vfitab           = _mm_cvttpd_epi32(rt);
1856             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1857             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1858
1859             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1860             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1861             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1862             GMX_MM_TRANSPOSE2_PD(Y,F);
1863             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1864             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1865             GMX_MM_TRANSPOSE2_PD(G,H);
1866             Heps             = _mm_mul_pd(vfeps,H);
1867             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1868             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1869             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1870
1871             fscal            = felec;
1872
1873             /* Calculate temporary vectorial force */
1874             tx               = _mm_mul_pd(fscal,dx22);
1875             ty               = _mm_mul_pd(fscal,dy22);
1876             tz               = _mm_mul_pd(fscal,dz22);
1877
1878             /* Update vectorial force */
1879             fix2             = _mm_add_pd(fix2,tx);
1880             fiy2             = _mm_add_pd(fiy2,ty);
1881             fiz2             = _mm_add_pd(fiz2,tz);
1882
1883             fjx2             = _mm_add_pd(fjx2,tx);
1884             fjy2             = _mm_add_pd(fjy2,ty);
1885             fjz2             = _mm_add_pd(fjz2,tz);
1886
1887             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1888
1889             /* Inner loop uses 373 flops */
1890         }
1891
1892         if(jidx<j_index_end)
1893         {
1894
1895             jnrA             = jjnr[jidx];
1896             j_coord_offsetA  = DIM*jnrA;
1897
1898             /* load j atom coordinates */
1899             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1900                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1901
1902             /* Calculate displacement vector */
1903             dx00             = _mm_sub_pd(ix0,jx0);
1904             dy00             = _mm_sub_pd(iy0,jy0);
1905             dz00             = _mm_sub_pd(iz0,jz0);
1906             dx01             = _mm_sub_pd(ix0,jx1);
1907             dy01             = _mm_sub_pd(iy0,jy1);
1908             dz01             = _mm_sub_pd(iz0,jz1);
1909             dx02             = _mm_sub_pd(ix0,jx2);
1910             dy02             = _mm_sub_pd(iy0,jy2);
1911             dz02             = _mm_sub_pd(iz0,jz2);
1912             dx10             = _mm_sub_pd(ix1,jx0);
1913             dy10             = _mm_sub_pd(iy1,jy0);
1914             dz10             = _mm_sub_pd(iz1,jz0);
1915             dx11             = _mm_sub_pd(ix1,jx1);
1916             dy11             = _mm_sub_pd(iy1,jy1);
1917             dz11             = _mm_sub_pd(iz1,jz1);
1918             dx12             = _mm_sub_pd(ix1,jx2);
1919             dy12             = _mm_sub_pd(iy1,jy2);
1920             dz12             = _mm_sub_pd(iz1,jz2);
1921             dx20             = _mm_sub_pd(ix2,jx0);
1922             dy20             = _mm_sub_pd(iy2,jy0);
1923             dz20             = _mm_sub_pd(iz2,jz0);
1924             dx21             = _mm_sub_pd(ix2,jx1);
1925             dy21             = _mm_sub_pd(iy2,jy1);
1926             dz21             = _mm_sub_pd(iz2,jz1);
1927             dx22             = _mm_sub_pd(ix2,jx2);
1928             dy22             = _mm_sub_pd(iy2,jy2);
1929             dz22             = _mm_sub_pd(iz2,jz2);
1930
1931             /* Calculate squared distance and things based on it */
1932             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1933             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1934             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1935             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1936             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1937             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1938             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1939             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1940             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1941
1942             rinv00           = sse2_invsqrt_d(rsq00);
1943             rinv01           = sse2_invsqrt_d(rsq01);
1944             rinv02           = sse2_invsqrt_d(rsq02);
1945             rinv10           = sse2_invsqrt_d(rsq10);
1946             rinv11           = sse2_invsqrt_d(rsq11);
1947             rinv12           = sse2_invsqrt_d(rsq12);
1948             rinv20           = sse2_invsqrt_d(rsq20);
1949             rinv21           = sse2_invsqrt_d(rsq21);
1950             rinv22           = sse2_invsqrt_d(rsq22);
1951
1952             fjx0             = _mm_setzero_pd();
1953             fjy0             = _mm_setzero_pd();
1954             fjz0             = _mm_setzero_pd();
1955             fjx1             = _mm_setzero_pd();
1956             fjy1             = _mm_setzero_pd();
1957             fjz1             = _mm_setzero_pd();
1958             fjx2             = _mm_setzero_pd();
1959             fjy2             = _mm_setzero_pd();
1960             fjz2             = _mm_setzero_pd();
1961
1962             /**************************
1963              * CALCULATE INTERACTIONS *
1964              **************************/
1965
1966             r00              = _mm_mul_pd(rsq00,rinv00);
1967
1968             /* Calculate table index by multiplying r with table scale and truncate to integer */
1969             rt               = _mm_mul_pd(r00,vftabscale);
1970             vfitab           = _mm_cvttpd_epi32(rt);
1971             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1972             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1973
1974             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1975             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1976             F                = _mm_setzero_pd();
1977             GMX_MM_TRANSPOSE2_PD(Y,F);
1978             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1979             H                = _mm_setzero_pd();
1980             GMX_MM_TRANSPOSE2_PD(G,H);
1981             Heps             = _mm_mul_pd(vfeps,H);
1982             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1983             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1984             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1985
1986             /* CUBIC SPLINE TABLE DISPERSION */
1987             vfitab           = _mm_add_epi32(vfitab,ifour);
1988             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1989             F                = _mm_setzero_pd();
1990             GMX_MM_TRANSPOSE2_PD(Y,F);
1991             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1992             H                = _mm_setzero_pd();
1993             GMX_MM_TRANSPOSE2_PD(G,H);
1994             Heps             = _mm_mul_pd(vfeps,H);
1995             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1996             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1997             fvdw6            = _mm_mul_pd(c6_00,FF);
1998
1999             /* CUBIC SPLINE TABLE REPULSION */
2000             vfitab           = _mm_add_epi32(vfitab,ifour);
2001             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2002             F                = _mm_setzero_pd();
2003             GMX_MM_TRANSPOSE2_PD(Y,F);
2004             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2005             H                = _mm_setzero_pd();
2006             GMX_MM_TRANSPOSE2_PD(G,H);
2007             Heps             = _mm_mul_pd(vfeps,H);
2008             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2009             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2010             fvdw12           = _mm_mul_pd(c12_00,FF);
2011             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
2012
2013             fscal            = _mm_add_pd(felec,fvdw);
2014
2015             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2016
2017             /* Calculate temporary vectorial force */
2018             tx               = _mm_mul_pd(fscal,dx00);
2019             ty               = _mm_mul_pd(fscal,dy00);
2020             tz               = _mm_mul_pd(fscal,dz00);
2021
2022             /* Update vectorial force */
2023             fix0             = _mm_add_pd(fix0,tx);
2024             fiy0             = _mm_add_pd(fiy0,ty);
2025             fiz0             = _mm_add_pd(fiz0,tz);
2026
2027             fjx0             = _mm_add_pd(fjx0,tx);
2028             fjy0             = _mm_add_pd(fjy0,ty);
2029             fjz0             = _mm_add_pd(fjz0,tz);
2030
2031             /**************************
2032              * CALCULATE INTERACTIONS *
2033              **************************/
2034
2035             r01              = _mm_mul_pd(rsq01,rinv01);
2036
2037             /* Calculate table index by multiplying r with table scale and truncate to integer */
2038             rt               = _mm_mul_pd(r01,vftabscale);
2039             vfitab           = _mm_cvttpd_epi32(rt);
2040             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2041             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2042
2043             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2044             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2045             F                = _mm_setzero_pd();
2046             GMX_MM_TRANSPOSE2_PD(Y,F);
2047             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2048             H                = _mm_setzero_pd();
2049             GMX_MM_TRANSPOSE2_PD(G,H);
2050             Heps             = _mm_mul_pd(vfeps,H);
2051             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2052             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2053             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
2054
2055             fscal            = felec;
2056
2057             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2058
2059             /* Calculate temporary vectorial force */
2060             tx               = _mm_mul_pd(fscal,dx01);
2061             ty               = _mm_mul_pd(fscal,dy01);
2062             tz               = _mm_mul_pd(fscal,dz01);
2063
2064             /* Update vectorial force */
2065             fix0             = _mm_add_pd(fix0,tx);
2066             fiy0             = _mm_add_pd(fiy0,ty);
2067             fiz0             = _mm_add_pd(fiz0,tz);
2068
2069             fjx1             = _mm_add_pd(fjx1,tx);
2070             fjy1             = _mm_add_pd(fjy1,ty);
2071             fjz1             = _mm_add_pd(fjz1,tz);
2072
2073             /**************************
2074              * CALCULATE INTERACTIONS *
2075              **************************/
2076
2077             r02              = _mm_mul_pd(rsq02,rinv02);
2078
2079             /* Calculate table index by multiplying r with table scale and truncate to integer */
2080             rt               = _mm_mul_pd(r02,vftabscale);
2081             vfitab           = _mm_cvttpd_epi32(rt);
2082             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2083             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2084
2085             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2086             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2087             F                = _mm_setzero_pd();
2088             GMX_MM_TRANSPOSE2_PD(Y,F);
2089             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2090             H                = _mm_setzero_pd();
2091             GMX_MM_TRANSPOSE2_PD(G,H);
2092             Heps             = _mm_mul_pd(vfeps,H);
2093             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2094             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2095             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
2096
2097             fscal            = felec;
2098
2099             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2100
2101             /* Calculate temporary vectorial force */
2102             tx               = _mm_mul_pd(fscal,dx02);
2103             ty               = _mm_mul_pd(fscal,dy02);
2104             tz               = _mm_mul_pd(fscal,dz02);
2105
2106             /* Update vectorial force */
2107             fix0             = _mm_add_pd(fix0,tx);
2108             fiy0             = _mm_add_pd(fiy0,ty);
2109             fiz0             = _mm_add_pd(fiz0,tz);
2110
2111             fjx2             = _mm_add_pd(fjx2,tx);
2112             fjy2             = _mm_add_pd(fjy2,ty);
2113             fjz2             = _mm_add_pd(fjz2,tz);
2114
2115             /**************************
2116              * CALCULATE INTERACTIONS *
2117              **************************/
2118
2119             r10              = _mm_mul_pd(rsq10,rinv10);
2120
2121             /* Calculate table index by multiplying r with table scale and truncate to integer */
2122             rt               = _mm_mul_pd(r10,vftabscale);
2123             vfitab           = _mm_cvttpd_epi32(rt);
2124             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2125             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2126
2127             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2128             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2129             F                = _mm_setzero_pd();
2130             GMX_MM_TRANSPOSE2_PD(Y,F);
2131             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2132             H                = _mm_setzero_pd();
2133             GMX_MM_TRANSPOSE2_PD(G,H);
2134             Heps             = _mm_mul_pd(vfeps,H);
2135             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2136             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2137             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
2138
2139             fscal            = felec;
2140
2141             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2142
2143             /* Calculate temporary vectorial force */
2144             tx               = _mm_mul_pd(fscal,dx10);
2145             ty               = _mm_mul_pd(fscal,dy10);
2146             tz               = _mm_mul_pd(fscal,dz10);
2147
2148             /* Update vectorial force */
2149             fix1             = _mm_add_pd(fix1,tx);
2150             fiy1             = _mm_add_pd(fiy1,ty);
2151             fiz1             = _mm_add_pd(fiz1,tz);
2152
2153             fjx0             = _mm_add_pd(fjx0,tx);
2154             fjy0             = _mm_add_pd(fjy0,ty);
2155             fjz0             = _mm_add_pd(fjz0,tz);
2156
2157             /**************************
2158              * CALCULATE INTERACTIONS *
2159              **************************/
2160
2161             r11              = _mm_mul_pd(rsq11,rinv11);
2162
2163             /* Calculate table index by multiplying r with table scale and truncate to integer */
2164             rt               = _mm_mul_pd(r11,vftabscale);
2165             vfitab           = _mm_cvttpd_epi32(rt);
2166             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2167             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2168
2169             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2170             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2171             F                = _mm_setzero_pd();
2172             GMX_MM_TRANSPOSE2_PD(Y,F);
2173             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2174             H                = _mm_setzero_pd();
2175             GMX_MM_TRANSPOSE2_PD(G,H);
2176             Heps             = _mm_mul_pd(vfeps,H);
2177             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2178             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2179             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2180
2181             fscal            = felec;
2182
2183             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2184
2185             /* Calculate temporary vectorial force */
2186             tx               = _mm_mul_pd(fscal,dx11);
2187             ty               = _mm_mul_pd(fscal,dy11);
2188             tz               = _mm_mul_pd(fscal,dz11);
2189
2190             /* Update vectorial force */
2191             fix1             = _mm_add_pd(fix1,tx);
2192             fiy1             = _mm_add_pd(fiy1,ty);
2193             fiz1             = _mm_add_pd(fiz1,tz);
2194
2195             fjx1             = _mm_add_pd(fjx1,tx);
2196             fjy1             = _mm_add_pd(fjy1,ty);
2197             fjz1             = _mm_add_pd(fjz1,tz);
2198
2199             /**************************
2200              * CALCULATE INTERACTIONS *
2201              **************************/
2202
2203             r12              = _mm_mul_pd(rsq12,rinv12);
2204
2205             /* Calculate table index by multiplying r with table scale and truncate to integer */
2206             rt               = _mm_mul_pd(r12,vftabscale);
2207             vfitab           = _mm_cvttpd_epi32(rt);
2208             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2209             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2210
2211             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2212             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2213             F                = _mm_setzero_pd();
2214             GMX_MM_TRANSPOSE2_PD(Y,F);
2215             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2216             H                = _mm_setzero_pd();
2217             GMX_MM_TRANSPOSE2_PD(G,H);
2218             Heps             = _mm_mul_pd(vfeps,H);
2219             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2220             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2221             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2222
2223             fscal            = felec;
2224
2225             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2226
2227             /* Calculate temporary vectorial force */
2228             tx               = _mm_mul_pd(fscal,dx12);
2229             ty               = _mm_mul_pd(fscal,dy12);
2230             tz               = _mm_mul_pd(fscal,dz12);
2231
2232             /* Update vectorial force */
2233             fix1             = _mm_add_pd(fix1,tx);
2234             fiy1             = _mm_add_pd(fiy1,ty);
2235             fiz1             = _mm_add_pd(fiz1,tz);
2236
2237             fjx2             = _mm_add_pd(fjx2,tx);
2238             fjy2             = _mm_add_pd(fjy2,ty);
2239             fjz2             = _mm_add_pd(fjz2,tz);
2240
2241             /**************************
2242              * CALCULATE INTERACTIONS *
2243              **************************/
2244
2245             r20              = _mm_mul_pd(rsq20,rinv20);
2246
2247             /* Calculate table index by multiplying r with table scale and truncate to integer */
2248             rt               = _mm_mul_pd(r20,vftabscale);
2249             vfitab           = _mm_cvttpd_epi32(rt);
2250             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2251             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2252
2253             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2254             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2255             F                = _mm_setzero_pd();
2256             GMX_MM_TRANSPOSE2_PD(Y,F);
2257             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2258             H                = _mm_setzero_pd();
2259             GMX_MM_TRANSPOSE2_PD(G,H);
2260             Heps             = _mm_mul_pd(vfeps,H);
2261             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2262             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2263             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
2264
2265             fscal            = felec;
2266
2267             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2268
2269             /* Calculate temporary vectorial force */
2270             tx               = _mm_mul_pd(fscal,dx20);
2271             ty               = _mm_mul_pd(fscal,dy20);
2272             tz               = _mm_mul_pd(fscal,dz20);
2273
2274             /* Update vectorial force */
2275             fix2             = _mm_add_pd(fix2,tx);
2276             fiy2             = _mm_add_pd(fiy2,ty);
2277             fiz2             = _mm_add_pd(fiz2,tz);
2278
2279             fjx0             = _mm_add_pd(fjx0,tx);
2280             fjy0             = _mm_add_pd(fjy0,ty);
2281             fjz0             = _mm_add_pd(fjz0,tz);
2282
2283             /**************************
2284              * CALCULATE INTERACTIONS *
2285              **************************/
2286
2287             r21              = _mm_mul_pd(rsq21,rinv21);
2288
2289             /* Calculate table index by multiplying r with table scale and truncate to integer */
2290             rt               = _mm_mul_pd(r21,vftabscale);
2291             vfitab           = _mm_cvttpd_epi32(rt);
2292             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2293             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2294
2295             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2296             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2297             F                = _mm_setzero_pd();
2298             GMX_MM_TRANSPOSE2_PD(Y,F);
2299             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2300             H                = _mm_setzero_pd();
2301             GMX_MM_TRANSPOSE2_PD(G,H);
2302             Heps             = _mm_mul_pd(vfeps,H);
2303             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2304             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2305             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2306
2307             fscal            = felec;
2308
2309             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2310
2311             /* Calculate temporary vectorial force */
2312             tx               = _mm_mul_pd(fscal,dx21);
2313             ty               = _mm_mul_pd(fscal,dy21);
2314             tz               = _mm_mul_pd(fscal,dz21);
2315
2316             /* Update vectorial force */
2317             fix2             = _mm_add_pd(fix2,tx);
2318             fiy2             = _mm_add_pd(fiy2,ty);
2319             fiz2             = _mm_add_pd(fiz2,tz);
2320
2321             fjx1             = _mm_add_pd(fjx1,tx);
2322             fjy1             = _mm_add_pd(fjy1,ty);
2323             fjz1             = _mm_add_pd(fjz1,tz);
2324
2325             /**************************
2326              * CALCULATE INTERACTIONS *
2327              **************************/
2328
2329             r22              = _mm_mul_pd(rsq22,rinv22);
2330
2331             /* Calculate table index by multiplying r with table scale and truncate to integer */
2332             rt               = _mm_mul_pd(r22,vftabscale);
2333             vfitab           = _mm_cvttpd_epi32(rt);
2334             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2335             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2336
2337             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2338             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2339             F                = _mm_setzero_pd();
2340             GMX_MM_TRANSPOSE2_PD(Y,F);
2341             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2342             H                = _mm_setzero_pd();
2343             GMX_MM_TRANSPOSE2_PD(G,H);
2344             Heps             = _mm_mul_pd(vfeps,H);
2345             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2346             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2347             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2348
2349             fscal            = felec;
2350
2351             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2352
2353             /* Calculate temporary vectorial force */
2354             tx               = _mm_mul_pd(fscal,dx22);
2355             ty               = _mm_mul_pd(fscal,dy22);
2356             tz               = _mm_mul_pd(fscal,dz22);
2357
2358             /* Update vectorial force */
2359             fix2             = _mm_add_pd(fix2,tx);
2360             fiy2             = _mm_add_pd(fiy2,ty);
2361             fiz2             = _mm_add_pd(fiz2,tz);
2362
2363             fjx2             = _mm_add_pd(fjx2,tx);
2364             fjy2             = _mm_add_pd(fjy2,ty);
2365             fjz2             = _mm_add_pd(fjz2,tz);
2366
2367             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2368
2369             /* Inner loop uses 373 flops */
2370         }
2371
2372         /* End of innermost loop */
2373
2374         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2375                                               f+i_coord_offset,fshift+i_shift_offset);
2376
2377         /* Increment number of inner iterations */
2378         inneriter                  += j_index_end - j_index_start;
2379
2380         /* Outer loop uses 18 flops */
2381     }
2382
2383     /* Increment number of outer iterations */
2384     outeriter        += nri;
2385
2386     /* Update outer/inner flops */
2387
2388     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*373);
2389 }