src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecCSTab_VdwLJ_GeomW3W3_sse2_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse2_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse2_double.h"
  48 #include "kernelutil_x86_sse2_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3W3_VF_sse2_double
  52  * Electrostatics interaction: CubicSplineTable
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water3-Water3
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCSTab_VdwLJ_GeomW3W3_VF_sse2_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwjidx0A,vdwjidx0B;
  87     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  88     int              vdwjidx1A,vdwjidx1B;
  89     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  90     int              vdwjidx2A,vdwjidx2B;
  91     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  92     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  93     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  94     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  95     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  96     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  97     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  98     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  99     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 100     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 101     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 102     real             *charge;
 103     int              nvdwtype;
 104     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 105     int              *vdwtype;
 106     real             *vdwparam;
 107     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 108     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 109     __m128i          vfitab;
 110     __m128i          ifour       = _mm_set1_epi32(4);
 111     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 112     real             *vftab;
 113     __m128d          dummy_mask,cutoff_mask;
 114     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 115     __m128d          one     = _mm_set1_pd(1.0);
 116     __m128d          two     = _mm_set1_pd(2.0);
 117     x                = xx[0];
 118     f                = ff[0];
 119
 120     nri              = nlist->nri;
 121     iinr             = nlist->iinr;
 122     jindex           = nlist->jindex;
 123     jjnr             = nlist->jjnr;
 124     shiftidx         = nlist->shift;
 125     gid              = nlist->gid;
 126     shiftvec         = fr->shift_vec[0];
 127     fshift           = fr->fshift[0];
 128     facel            = _mm_set1_pd(fr->epsfac);
 129     charge           = mdatoms->chargeA;
 130     nvdwtype         = fr->ntype;
 131     vdwparam         = fr->nbfp;
 132     vdwtype          = mdatoms->typeA;
 133
 134     vftab            = kernel_data->table_elec->data;
 135     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
 136
 137     /* Setup water-specific parameters */
 138     inr              = nlist->iinr[0];
 139     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 140     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 141     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 142     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 143
 144     jq0              = _mm_set1_pd(charge[inr+0]);
 145     jq1              = _mm_set1_pd(charge[inr+1]);
 146     jq2              = _mm_set1_pd(charge[inr+2]);
 147     vdwjidx0A        = 2*vdwtype[inr+0];
 148     qq00             = _mm_mul_pd(iq0,jq0);
 149     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 150     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 151     qq01             = _mm_mul_pd(iq0,jq1);
 152     qq02             = _mm_mul_pd(iq0,jq2);
 153     qq10             = _mm_mul_pd(iq1,jq0);
 154     qq11             = _mm_mul_pd(iq1,jq1);
 155     qq12             = _mm_mul_pd(iq1,jq2);
 156     qq20             = _mm_mul_pd(iq2,jq0);
 157     qq21             = _mm_mul_pd(iq2,jq1);
 158     qq22             = _mm_mul_pd(iq2,jq2);
 159
 160     /* Avoid stupid compiler warnings */
 161     jnrA = jnrB = 0;
 162     j_coord_offsetA = 0;
 163     j_coord_offsetB = 0;
 164
 165     outeriter        = 0;
 166     inneriter        = 0;
 167
 168     /* Start outer loop over neighborlists */
 169     for(iidx=0; iidx<nri; iidx++)
 170     {
 171         /* Load shift vector for this list */
 172         i_shift_offset   = DIM*shiftidx[iidx];
 173
 174         /* Load limits for loop over neighbors */
 175         j_index_start    = jindex[iidx];
 176         j_index_end      = jindex[iidx+1];
 177
 178         /* Get outer coordinate index */
 179         inr              = iinr[iidx];
 180         i_coord_offset   = DIM*inr;
 181
 182         /* Load i particle coords and add shift vector */
 183         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 184                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 185
 186         fix0             = _mm_setzero_pd();
 187         fiy0             = _mm_setzero_pd();
 188         fiz0             = _mm_setzero_pd();
 189         fix1             = _mm_setzero_pd();
 190         fiy1             = _mm_setzero_pd();
 191         fiz1             = _mm_setzero_pd();
 192         fix2             = _mm_setzero_pd();
 193         fiy2             = _mm_setzero_pd();
 194         fiz2             = _mm_setzero_pd();
 195
 196         /* Reset potential sums */
 197         velecsum         = _mm_setzero_pd();
 198         vvdwsum          = _mm_setzero_pd();
 199
 200         /* Start inner kernel loop */
 201         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 202         {
 203
 204             /* Get j neighbor index, and coordinate index */
 205             jnrA             = jjnr[jidx];
 206             jnrB             = jjnr[jidx+1];
 207             j_coord_offsetA  = DIM*jnrA;
 208             j_coord_offsetB  = DIM*jnrB;
 209
 210             /* load j atom coordinates */
 211             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 212                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 213
 214             /* Calculate displacement vector */
 215             dx00             = _mm_sub_pd(ix0,jx0);
 216             dy00             = _mm_sub_pd(iy0,jy0);
 217             dz00             = _mm_sub_pd(iz0,jz0);
 218             dx01             = _mm_sub_pd(ix0,jx1);
 219             dy01             = _mm_sub_pd(iy0,jy1);
 220             dz01             = _mm_sub_pd(iz0,jz1);
 221             dx02             = _mm_sub_pd(ix0,jx2);
 222             dy02             = _mm_sub_pd(iy0,jy2);
 223             dz02             = _mm_sub_pd(iz0,jz2);
 224             dx10             = _mm_sub_pd(ix1,jx0);
 225             dy10             = _mm_sub_pd(iy1,jy0);
 226             dz10             = _mm_sub_pd(iz1,jz0);
 227             dx11             = _mm_sub_pd(ix1,jx1);
 228             dy11             = _mm_sub_pd(iy1,jy1);
 229             dz11             = _mm_sub_pd(iz1,jz1);
 230             dx12             = _mm_sub_pd(ix1,jx2);
 231             dy12             = _mm_sub_pd(iy1,jy2);
 232             dz12             = _mm_sub_pd(iz1,jz2);
 233             dx20             = _mm_sub_pd(ix2,jx0);
 234             dy20             = _mm_sub_pd(iy2,jy0);
 235             dz20             = _mm_sub_pd(iz2,jz0);
 236             dx21             = _mm_sub_pd(ix2,jx1);
 237             dy21             = _mm_sub_pd(iy2,jy1);
 238             dz21             = _mm_sub_pd(iz2,jz1);
 239             dx22             = _mm_sub_pd(ix2,jx2);
 240             dy22             = _mm_sub_pd(iy2,jy2);
 241             dz22             = _mm_sub_pd(iz2,jz2);
 242
 243             /* Calculate squared distance and things based on it */
 244             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 245             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 246             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 247             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 248             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 249             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 250             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 251             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 252             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 253
 254             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 255             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 256             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 257             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 258             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 259             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 260             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 261             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 262             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 263
 264             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 265
 266             fjx0             = _mm_setzero_pd();
 267             fjy0             = _mm_setzero_pd();
 268             fjz0             = _mm_setzero_pd();
 269             fjx1             = _mm_setzero_pd();
 270             fjy1             = _mm_setzero_pd();
 271             fjz1             = _mm_setzero_pd();
 272             fjx2             = _mm_setzero_pd();
 273             fjy2             = _mm_setzero_pd();
 274             fjz2             = _mm_setzero_pd();
 275
 276             /**************************
 277              * CALCULATE INTERACTIONS *
 278              **************************/
 279
 280             r00              = _mm_mul_pd(rsq00,rinv00);
 281
 282             /* Calculate table index by multiplying r with table scale and truncate to integer */
 283             rt               = _mm_mul_pd(r00,vftabscale);
 284             vfitab           = _mm_cvttpd_epi32(rt);
 285             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 286             vfitab           = _mm_slli_epi32(vfitab,2);
 287
 288             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 289             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 290             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 291             GMX_MM_TRANSPOSE2_PD(Y,F);
 292             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 293             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 294             GMX_MM_TRANSPOSE2_PD(G,H);
 295             Heps             = _mm_mul_pd(vfeps,H);
 296             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 297             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 298             velec            = _mm_mul_pd(qq00,VV);
 299             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 300             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 301
 302             /* LENNARD-JONES DISPERSION/REPULSION */
 303
 304             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 305             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 306             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 307             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 308             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 309
 310             /* Update potential sum for this i atom from the interaction with this j atom. */
 311             velecsum         = _mm_add_pd(velecsum,velec);
 312             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 313
 314             fscal            = _mm_add_pd(felec,fvdw);
 315
 316             /* Calculate temporary vectorial force */
 317             tx               = _mm_mul_pd(fscal,dx00);
 318             ty               = _mm_mul_pd(fscal,dy00);
 319             tz               = _mm_mul_pd(fscal,dz00);
 320
 321             /* Update vectorial force */
 322             fix0             = _mm_add_pd(fix0,tx);
 323             fiy0             = _mm_add_pd(fiy0,ty);
 324             fiz0             = _mm_add_pd(fiz0,tz);
 325
 326             fjx0             = _mm_add_pd(fjx0,tx);
 327             fjy0             = _mm_add_pd(fjy0,ty);
 328             fjz0             = _mm_add_pd(fjz0,tz);
 329
 330             /**************************
 331              * CALCULATE INTERACTIONS *
 332              **************************/
 333
 334             r01              = _mm_mul_pd(rsq01,rinv01);
 335
 336             /* Calculate table index by multiplying r with table scale and truncate to integer */
 337             rt               = _mm_mul_pd(r01,vftabscale);
 338             vfitab           = _mm_cvttpd_epi32(rt);
 339             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 340             vfitab           = _mm_slli_epi32(vfitab,2);
 341
 342             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 343             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 344             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 345             GMX_MM_TRANSPOSE2_PD(Y,F);
 346             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 347             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 348             GMX_MM_TRANSPOSE2_PD(G,H);
 349             Heps             = _mm_mul_pd(vfeps,H);
 350             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 351             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 352             velec            = _mm_mul_pd(qq01,VV);
 353             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 354             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 355
 356             /* Update potential sum for this i atom from the interaction with this j atom. */
 357             velecsum         = _mm_add_pd(velecsum,velec);
 358
 359             fscal            = felec;
 360
 361             /* Calculate temporary vectorial force */
 362             tx               = _mm_mul_pd(fscal,dx01);
 363             ty               = _mm_mul_pd(fscal,dy01);
 364             tz               = _mm_mul_pd(fscal,dz01);
 365
 366             /* Update vectorial force */
 367             fix0             = _mm_add_pd(fix0,tx);
 368             fiy0             = _mm_add_pd(fiy0,ty);
 369             fiz0             = _mm_add_pd(fiz0,tz);
 370
 371             fjx1             = _mm_add_pd(fjx1,tx);
 372             fjy1             = _mm_add_pd(fjy1,ty);
 373             fjz1             = _mm_add_pd(fjz1,tz);
 374
 375             /**************************
 376              * CALCULATE INTERACTIONS *
 377              **************************/
 378
 379             r02              = _mm_mul_pd(rsq02,rinv02);
 380
 381             /* Calculate table index by multiplying r with table scale and truncate to integer */
 382             rt               = _mm_mul_pd(r02,vftabscale);
 383             vfitab           = _mm_cvttpd_epi32(rt);
 384             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 385             vfitab           = _mm_slli_epi32(vfitab,2);
 386
 387             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 388             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 389             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 390             GMX_MM_TRANSPOSE2_PD(Y,F);
 391             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 392             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 393             GMX_MM_TRANSPOSE2_PD(G,H);
 394             Heps             = _mm_mul_pd(vfeps,H);
 395             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 396             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 397             velec            = _mm_mul_pd(qq02,VV);
 398             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 399             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 400
 401             /* Update potential sum for this i atom from the interaction with this j atom. */
 402             velecsum         = _mm_add_pd(velecsum,velec);
 403
 404             fscal            = felec;
 405
 406             /* Calculate temporary vectorial force */
 407             tx               = _mm_mul_pd(fscal,dx02);
 408             ty               = _mm_mul_pd(fscal,dy02);
 409             tz               = _mm_mul_pd(fscal,dz02);
 410
 411             /* Update vectorial force */
 412             fix0             = _mm_add_pd(fix0,tx);
 413             fiy0             = _mm_add_pd(fiy0,ty);
 414             fiz0             = _mm_add_pd(fiz0,tz);
 415
 416             fjx2             = _mm_add_pd(fjx2,tx);
 417             fjy2             = _mm_add_pd(fjy2,ty);
 418             fjz2             = _mm_add_pd(fjz2,tz);
 419
 420             /**************************
 421              * CALCULATE INTERACTIONS *
 422              **************************/
 423
 424             r10              = _mm_mul_pd(rsq10,rinv10);
 425
 426             /* Calculate table index by multiplying r with table scale and truncate to integer */
 427             rt               = _mm_mul_pd(r10,vftabscale);
 428             vfitab           = _mm_cvttpd_epi32(rt);
 429             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 430             vfitab           = _mm_slli_epi32(vfitab,2);
 431
 432             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 433             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 434             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 435             GMX_MM_TRANSPOSE2_PD(Y,F);
 436             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 437             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 438             GMX_MM_TRANSPOSE2_PD(G,H);
 439             Heps             = _mm_mul_pd(vfeps,H);
 440             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 441             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 442             velec            = _mm_mul_pd(qq10,VV);
 443             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 444             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 445
 446             /* Update potential sum for this i atom from the interaction with this j atom. */
 447             velecsum         = _mm_add_pd(velecsum,velec);
 448
 449             fscal            = felec;
 450
 451             /* Calculate temporary vectorial force */
 452             tx               = _mm_mul_pd(fscal,dx10);
 453             ty               = _mm_mul_pd(fscal,dy10);
 454             tz               = _mm_mul_pd(fscal,dz10);
 455
 456             /* Update vectorial force */
 457             fix1             = _mm_add_pd(fix1,tx);
 458             fiy1             = _mm_add_pd(fiy1,ty);
 459             fiz1             = _mm_add_pd(fiz1,tz);
 460
 461             fjx0             = _mm_add_pd(fjx0,tx);
 462             fjy0             = _mm_add_pd(fjy0,ty);
 463             fjz0             = _mm_add_pd(fjz0,tz);
 464
 465             /**************************
 466              * CALCULATE INTERACTIONS *
 467              **************************/
 468
 469             r11              = _mm_mul_pd(rsq11,rinv11);
 470
 471             /* Calculate table index by multiplying r with table scale and truncate to integer */
 472             rt               = _mm_mul_pd(r11,vftabscale);
 473             vfitab           = _mm_cvttpd_epi32(rt);
 474             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 475             vfitab           = _mm_slli_epi32(vfitab,2);
 476
 477             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 478             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 479             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 480             GMX_MM_TRANSPOSE2_PD(Y,F);
 481             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 482             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 483             GMX_MM_TRANSPOSE2_PD(G,H);
 484             Heps             = _mm_mul_pd(vfeps,H);
 485             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 486             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 487             velec            = _mm_mul_pd(qq11,VV);
 488             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 489             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 490
 491             /* Update potential sum for this i atom from the interaction with this j atom. */
 492             velecsum         = _mm_add_pd(velecsum,velec);
 493
 494             fscal            = felec;
 495
 496             /* Calculate temporary vectorial force */
 497             tx               = _mm_mul_pd(fscal,dx11);
 498             ty               = _mm_mul_pd(fscal,dy11);
 499             tz               = _mm_mul_pd(fscal,dz11);
 500
 501             /* Update vectorial force */
 502             fix1             = _mm_add_pd(fix1,tx);
 503             fiy1             = _mm_add_pd(fiy1,ty);
 504             fiz1             = _mm_add_pd(fiz1,tz);
 505
 506             fjx1             = _mm_add_pd(fjx1,tx);
 507             fjy1             = _mm_add_pd(fjy1,ty);
 508             fjz1             = _mm_add_pd(fjz1,tz);
 509
 510             /**************************
 511              * CALCULATE INTERACTIONS *
 512              **************************/
 513
 514             r12              = _mm_mul_pd(rsq12,rinv12);
 515
 516             /* Calculate table index by multiplying r with table scale and truncate to integer */
 517             rt               = _mm_mul_pd(r12,vftabscale);
 518             vfitab           = _mm_cvttpd_epi32(rt);
 519             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 520             vfitab           = _mm_slli_epi32(vfitab,2);
 521
 522             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 523             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 524             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 525             GMX_MM_TRANSPOSE2_PD(Y,F);
 526             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 527             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 528             GMX_MM_TRANSPOSE2_PD(G,H);
 529             Heps             = _mm_mul_pd(vfeps,H);
 530             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 531             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 532             velec            = _mm_mul_pd(qq12,VV);
 533             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 534             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 535
 536             /* Update potential sum for this i atom from the interaction with this j atom. */
 537             velecsum         = _mm_add_pd(velecsum,velec);
 538
 539             fscal            = felec;
 540
 541             /* Calculate temporary vectorial force */
 542             tx               = _mm_mul_pd(fscal,dx12);
 543             ty               = _mm_mul_pd(fscal,dy12);
 544             tz               = _mm_mul_pd(fscal,dz12);
 545
 546             /* Update vectorial force */
 547             fix1             = _mm_add_pd(fix1,tx);
 548             fiy1             = _mm_add_pd(fiy1,ty);
 549             fiz1             = _mm_add_pd(fiz1,tz);
 550
 551             fjx2             = _mm_add_pd(fjx2,tx);
 552             fjy2             = _mm_add_pd(fjy2,ty);
 553             fjz2             = _mm_add_pd(fjz2,tz);
 554
 555             /**************************
 556              * CALCULATE INTERACTIONS *
 557              **************************/
 558
 559             r20              = _mm_mul_pd(rsq20,rinv20);
 560
 561             /* Calculate table index by multiplying r with table scale and truncate to integer */
 562             rt               = _mm_mul_pd(r20,vftabscale);
 563             vfitab           = _mm_cvttpd_epi32(rt);
 564             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 565             vfitab           = _mm_slli_epi32(vfitab,2);
 566
 567             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 568             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 569             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 570             GMX_MM_TRANSPOSE2_PD(Y,F);
 571             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 572             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 573             GMX_MM_TRANSPOSE2_PD(G,H);
 574             Heps             = _mm_mul_pd(vfeps,H);
 575             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 576             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 577             velec            = _mm_mul_pd(qq20,VV);
 578             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 579             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
 580
 581             /* Update potential sum for this i atom from the interaction with this j atom. */
 582             velecsum         = _mm_add_pd(velecsum,velec);
 583
 584             fscal            = felec;
 585
 586             /* Calculate temporary vectorial force */
 587             tx               = _mm_mul_pd(fscal,dx20);
 588             ty               = _mm_mul_pd(fscal,dy20);
 589             tz               = _mm_mul_pd(fscal,dz20);
 590
 591             /* Update vectorial force */
 592             fix2             = _mm_add_pd(fix2,tx);
 593             fiy2             = _mm_add_pd(fiy2,ty);
 594             fiz2             = _mm_add_pd(fiz2,tz);
 595
 596             fjx0             = _mm_add_pd(fjx0,tx);
 597             fjy0             = _mm_add_pd(fjy0,ty);
 598             fjz0             = _mm_add_pd(fjz0,tz);
 599
 600             /**************************
 601              * CALCULATE INTERACTIONS *
 602              **************************/
 603
 604             r21              = _mm_mul_pd(rsq21,rinv21);
 605
 606             /* Calculate table index by multiplying r with table scale and truncate to integer */
 607             rt               = _mm_mul_pd(r21,vftabscale);
 608             vfitab           = _mm_cvttpd_epi32(rt);
 609             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 610             vfitab           = _mm_slli_epi32(vfitab,2);
 611
 612             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 613             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 614             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 615             GMX_MM_TRANSPOSE2_PD(Y,F);
 616             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 617             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 618             GMX_MM_TRANSPOSE2_PD(G,H);
 619             Heps             = _mm_mul_pd(vfeps,H);
 620             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 621             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 622             velec            = _mm_mul_pd(qq21,VV);
 623             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 624             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 625
 626             /* Update potential sum for this i atom from the interaction with this j atom. */
 627             velecsum         = _mm_add_pd(velecsum,velec);
 628
 629             fscal            = felec;
 630
 631             /* Calculate temporary vectorial force */
 632             tx               = _mm_mul_pd(fscal,dx21);
 633             ty               = _mm_mul_pd(fscal,dy21);
 634             tz               = _mm_mul_pd(fscal,dz21);
 635
 636             /* Update vectorial force */
 637             fix2             = _mm_add_pd(fix2,tx);
 638             fiy2             = _mm_add_pd(fiy2,ty);
 639             fiz2             = _mm_add_pd(fiz2,tz);
 640
 641             fjx1             = _mm_add_pd(fjx1,tx);
 642             fjy1             = _mm_add_pd(fjy1,ty);
 643             fjz1             = _mm_add_pd(fjz1,tz);
 644
 645             /**************************
 646              * CALCULATE INTERACTIONS *
 647              **************************/
 648
 649             r22              = _mm_mul_pd(rsq22,rinv22);
 650
 651             /* Calculate table index by multiplying r with table scale and truncate to integer */
 652             rt               = _mm_mul_pd(r22,vftabscale);
 653             vfitab           = _mm_cvttpd_epi32(rt);
 654             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 655             vfitab           = _mm_slli_epi32(vfitab,2);
 656
 657             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 658             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 659             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 660             GMX_MM_TRANSPOSE2_PD(Y,F);
 661             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 662             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 663             GMX_MM_TRANSPOSE2_PD(G,H);
 664             Heps             = _mm_mul_pd(vfeps,H);
 665             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 666             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 667             velec            = _mm_mul_pd(qq22,VV);
 668             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 669             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 670
 671             /* Update potential sum for this i atom from the interaction with this j atom. */
 672             velecsum         = _mm_add_pd(velecsum,velec);
 673
 674             fscal            = felec;
 675
 676             /* Calculate temporary vectorial force */
 677             tx               = _mm_mul_pd(fscal,dx22);
 678             ty               = _mm_mul_pd(fscal,dy22);
 679             tz               = _mm_mul_pd(fscal,dz22);
 680
 681             /* Update vectorial force */
 682             fix2             = _mm_add_pd(fix2,tx);
 683             fiy2             = _mm_add_pd(fiy2,ty);
 684             fiz2             = _mm_add_pd(fiz2,tz);
 685
 686             fjx2             = _mm_add_pd(fjx2,tx);
 687             fjy2             = _mm_add_pd(fjy2,ty);
 688             fjz2             = _mm_add_pd(fjz2,tz);
 689
 690             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 691
 692             /* Inner loop uses 400 flops */
 693         }
 694
 695         if(jidx<j_index_end)
 696         {
 697
 698             jnrA             = jjnr[jidx];
 699             j_coord_offsetA  = DIM*jnrA;
 700
 701             /* load j atom coordinates */
 702             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 703                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 704
 705             /* Calculate displacement vector */
 706             dx00             = _mm_sub_pd(ix0,jx0);
 707             dy00             = _mm_sub_pd(iy0,jy0);
 708             dz00             = _mm_sub_pd(iz0,jz0);
 709             dx01             = _mm_sub_pd(ix0,jx1);
 710             dy01             = _mm_sub_pd(iy0,jy1);
 711             dz01             = _mm_sub_pd(iz0,jz1);
 712             dx02             = _mm_sub_pd(ix0,jx2);
 713             dy02             = _mm_sub_pd(iy0,jy2);
 714             dz02             = _mm_sub_pd(iz0,jz2);
 715             dx10             = _mm_sub_pd(ix1,jx0);
 716             dy10             = _mm_sub_pd(iy1,jy0);
 717             dz10             = _mm_sub_pd(iz1,jz0);
 718             dx11             = _mm_sub_pd(ix1,jx1);
 719             dy11             = _mm_sub_pd(iy1,jy1);
 720             dz11             = _mm_sub_pd(iz1,jz1);
 721             dx12             = _mm_sub_pd(ix1,jx2);
 722             dy12             = _mm_sub_pd(iy1,jy2);
 723             dz12             = _mm_sub_pd(iz1,jz2);
 724             dx20             = _mm_sub_pd(ix2,jx0);
 725             dy20             = _mm_sub_pd(iy2,jy0);
 726             dz20             = _mm_sub_pd(iz2,jz0);
 727             dx21             = _mm_sub_pd(ix2,jx1);
 728             dy21             = _mm_sub_pd(iy2,jy1);
 729             dz21             = _mm_sub_pd(iz2,jz1);
 730             dx22             = _mm_sub_pd(ix2,jx2);
 731             dy22             = _mm_sub_pd(iy2,jy2);
 732             dz22             = _mm_sub_pd(iz2,jz2);
 733
 734             /* Calculate squared distance and things based on it */
 735             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 736             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 737             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 738             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 739             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 740             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 741             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 742             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 743             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 744
 745             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 746             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 747             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 748             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 749             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 750             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 751             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 752             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 753             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 754
 755             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 756
 757             fjx0             = _mm_setzero_pd();
 758             fjy0             = _mm_setzero_pd();
 759             fjz0             = _mm_setzero_pd();
 760             fjx1             = _mm_setzero_pd();
 761             fjy1             = _mm_setzero_pd();
 762             fjz1             = _mm_setzero_pd();
 763             fjx2             = _mm_setzero_pd();
 764             fjy2             = _mm_setzero_pd();
 765             fjz2             = _mm_setzero_pd();
 766
 767             /**************************
 768              * CALCULATE INTERACTIONS *
 769              **************************/
 770
 771             r00              = _mm_mul_pd(rsq00,rinv00);
 772
 773             /* Calculate table index by multiplying r with table scale and truncate to integer */
 774             rt               = _mm_mul_pd(r00,vftabscale);
 775             vfitab           = _mm_cvttpd_epi32(rt);
 776             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 777             vfitab           = _mm_slli_epi32(vfitab,2);
 778
 779             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 780             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 781             F                = _mm_setzero_pd();
 782             GMX_MM_TRANSPOSE2_PD(Y,F);
 783             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 784             H                = _mm_setzero_pd();
 785             GMX_MM_TRANSPOSE2_PD(G,H);
 786             Heps             = _mm_mul_pd(vfeps,H);
 787             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 788             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 789             velec            = _mm_mul_pd(qq00,VV);
 790             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 791             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
 792
 793             /* LENNARD-JONES DISPERSION/REPULSION */
 794
 795             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 796             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 797             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 798             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 799             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 800
 801             /* Update potential sum for this i atom from the interaction with this j atom. */
 802             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 803             velecsum         = _mm_add_pd(velecsum,velec);
 804             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 805             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 806
 807             fscal            = _mm_add_pd(felec,fvdw);
 808
 809             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 810
 811             /* Calculate temporary vectorial force */
 812             tx               = _mm_mul_pd(fscal,dx00);
 813             ty               = _mm_mul_pd(fscal,dy00);
 814             tz               = _mm_mul_pd(fscal,dz00);
 815
 816             /* Update vectorial force */
 817             fix0             = _mm_add_pd(fix0,tx);
 818             fiy0             = _mm_add_pd(fiy0,ty);
 819             fiz0             = _mm_add_pd(fiz0,tz);
 820
 821             fjx0             = _mm_add_pd(fjx0,tx);
 822             fjy0             = _mm_add_pd(fjy0,ty);
 823             fjz0             = _mm_add_pd(fjz0,tz);
 824
 825             /**************************
 826              * CALCULATE INTERACTIONS *
 827              **************************/
 828
 829             r01              = _mm_mul_pd(rsq01,rinv01);
 830
 831             /* Calculate table index by multiplying r with table scale and truncate to integer */
 832             rt               = _mm_mul_pd(r01,vftabscale);
 833             vfitab           = _mm_cvttpd_epi32(rt);
 834             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 835             vfitab           = _mm_slli_epi32(vfitab,2);
 836
 837             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 838             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 839             F                = _mm_setzero_pd();
 840             GMX_MM_TRANSPOSE2_PD(Y,F);
 841             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 842             H                = _mm_setzero_pd();
 843             GMX_MM_TRANSPOSE2_PD(G,H);
 844             Heps             = _mm_mul_pd(vfeps,H);
 845             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 846             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 847             velec            = _mm_mul_pd(qq01,VV);
 848             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 849             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
 850
 851             /* Update potential sum for this i atom from the interaction with this j atom. */
 852             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 853             velecsum         = _mm_add_pd(velecsum,velec);
 854
 855             fscal            = felec;
 856
 857             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 858
 859             /* Calculate temporary vectorial force */
 860             tx               = _mm_mul_pd(fscal,dx01);
 861             ty               = _mm_mul_pd(fscal,dy01);
 862             tz               = _mm_mul_pd(fscal,dz01);
 863
 864             /* Update vectorial force */
 865             fix0             = _mm_add_pd(fix0,tx);
 866             fiy0             = _mm_add_pd(fiy0,ty);
 867             fiz0             = _mm_add_pd(fiz0,tz);
 868
 869             fjx1             = _mm_add_pd(fjx1,tx);
 870             fjy1             = _mm_add_pd(fjy1,ty);
 871             fjz1             = _mm_add_pd(fjz1,tz);
 872
 873             /**************************
 874              * CALCULATE INTERACTIONS *
 875              **************************/
 876
 877             r02              = _mm_mul_pd(rsq02,rinv02);
 878
 879             /* Calculate table index by multiplying r with table scale and truncate to integer */
 880             rt               = _mm_mul_pd(r02,vftabscale);
 881             vfitab           = _mm_cvttpd_epi32(rt);
 882             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 883             vfitab           = _mm_slli_epi32(vfitab,2);
 884
 885             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 886             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 887             F                = _mm_setzero_pd();
 888             GMX_MM_TRANSPOSE2_PD(Y,F);
 889             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 890             H                = _mm_setzero_pd();
 891             GMX_MM_TRANSPOSE2_PD(G,H);
 892             Heps             = _mm_mul_pd(vfeps,H);
 893             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 894             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 895             velec            = _mm_mul_pd(qq02,VV);
 896             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 897             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
 898
 899             /* Update potential sum for this i atom from the interaction with this j atom. */
 900             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 901             velecsum         = _mm_add_pd(velecsum,velec);
 902
 903             fscal            = felec;
 904
 905             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 906
 907             /* Calculate temporary vectorial force */
 908             tx               = _mm_mul_pd(fscal,dx02);
 909             ty               = _mm_mul_pd(fscal,dy02);
 910             tz               = _mm_mul_pd(fscal,dz02);
 911
 912             /* Update vectorial force */
 913             fix0             = _mm_add_pd(fix0,tx);
 914             fiy0             = _mm_add_pd(fiy0,ty);
 915             fiz0             = _mm_add_pd(fiz0,tz);
 916
 917             fjx2             = _mm_add_pd(fjx2,tx);
 918             fjy2             = _mm_add_pd(fjy2,ty);
 919             fjz2             = _mm_add_pd(fjz2,tz);
 920
 921             /**************************
 922              * CALCULATE INTERACTIONS *
 923              **************************/
 924
 925             r10              = _mm_mul_pd(rsq10,rinv10);
 926
 927             /* Calculate table index by multiplying r with table scale and truncate to integer */
 928             rt               = _mm_mul_pd(r10,vftabscale);
 929             vfitab           = _mm_cvttpd_epi32(rt);
 930             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 931             vfitab           = _mm_slli_epi32(vfitab,2);
 932
 933             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 934             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 935             F                = _mm_setzero_pd();
 936             GMX_MM_TRANSPOSE2_PD(Y,F);
 937             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 938             H                = _mm_setzero_pd();
 939             GMX_MM_TRANSPOSE2_PD(G,H);
 940             Heps             = _mm_mul_pd(vfeps,H);
 941             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 942             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 943             velec            = _mm_mul_pd(qq10,VV);
 944             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 945             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
 946
 947             /* Update potential sum for this i atom from the interaction with this j atom. */
 948             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 949             velecsum         = _mm_add_pd(velecsum,velec);
 950
 951             fscal            = felec;
 952
 953             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 954
 955             /* Calculate temporary vectorial force */
 956             tx               = _mm_mul_pd(fscal,dx10);
 957             ty               = _mm_mul_pd(fscal,dy10);
 958             tz               = _mm_mul_pd(fscal,dz10);
 959
 960             /* Update vectorial force */
 961             fix1             = _mm_add_pd(fix1,tx);
 962             fiy1             = _mm_add_pd(fiy1,ty);
 963             fiz1             = _mm_add_pd(fiz1,tz);
 964
 965             fjx0             = _mm_add_pd(fjx0,tx);
 966             fjy0             = _mm_add_pd(fjy0,ty);
 967             fjz0             = _mm_add_pd(fjz0,tz);
 968
 969             /**************************
 970              * CALCULATE INTERACTIONS *
 971              **************************/
 972
 973             r11              = _mm_mul_pd(rsq11,rinv11);
 974
 975             /* Calculate table index by multiplying r with table scale and truncate to integer */
 976             rt               = _mm_mul_pd(r11,vftabscale);
 977             vfitab           = _mm_cvttpd_epi32(rt);
 978             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 979             vfitab           = _mm_slli_epi32(vfitab,2);
 980
 981             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 982             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 983             F                = _mm_setzero_pd();
 984             GMX_MM_TRANSPOSE2_PD(Y,F);
 985             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 986             H                = _mm_setzero_pd();
 987             GMX_MM_TRANSPOSE2_PD(G,H);
 988             Heps             = _mm_mul_pd(vfeps,H);
 989             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 990             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 991             velec            = _mm_mul_pd(qq11,VV);
 992             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 993             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 994
 995             /* Update potential sum for this i atom from the interaction with this j atom. */
 996             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 997             velecsum         = _mm_add_pd(velecsum,velec);
 998
 999             fscal            = felec;
1000
1001             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1002
1003             /* Calculate temporary vectorial force */
1004             tx               = _mm_mul_pd(fscal,dx11);
1005             ty               = _mm_mul_pd(fscal,dy11);
1006             tz               = _mm_mul_pd(fscal,dz11);
1007
1008             /* Update vectorial force */
1009             fix1             = _mm_add_pd(fix1,tx);
1010             fiy1             = _mm_add_pd(fiy1,ty);
1011             fiz1             = _mm_add_pd(fiz1,tz);
1012
1013             fjx1             = _mm_add_pd(fjx1,tx);
1014             fjy1             = _mm_add_pd(fjy1,ty);
1015             fjz1             = _mm_add_pd(fjz1,tz);
1016
1017             /**************************
1018              * CALCULATE INTERACTIONS *
1019              **************************/
1020
1021             r12              = _mm_mul_pd(rsq12,rinv12);
1022
1023             /* Calculate table index by multiplying r with table scale and truncate to integer */
1024             rt               = _mm_mul_pd(r12,vftabscale);
1025             vfitab           = _mm_cvttpd_epi32(rt);
1026             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1027             vfitab           = _mm_slli_epi32(vfitab,2);
1028
1029             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1030             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1031             F                = _mm_setzero_pd();
1032             GMX_MM_TRANSPOSE2_PD(Y,F);
1033             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1034             H                = _mm_setzero_pd();
1035             GMX_MM_TRANSPOSE2_PD(G,H);
1036             Heps             = _mm_mul_pd(vfeps,H);
1037             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1038             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1039             velec            = _mm_mul_pd(qq12,VV);
1040             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1041             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1042
1043             /* Update potential sum for this i atom from the interaction with this j atom. */
1044             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1045             velecsum         = _mm_add_pd(velecsum,velec);
1046
1047             fscal            = felec;
1048
1049             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1050
1051             /* Calculate temporary vectorial force */
1052             tx               = _mm_mul_pd(fscal,dx12);
1053             ty               = _mm_mul_pd(fscal,dy12);
1054             tz               = _mm_mul_pd(fscal,dz12);
1055
1056             /* Update vectorial force */
1057             fix1             = _mm_add_pd(fix1,tx);
1058             fiy1             = _mm_add_pd(fiy1,ty);
1059             fiz1             = _mm_add_pd(fiz1,tz);
1060
1061             fjx2             = _mm_add_pd(fjx2,tx);
1062             fjy2             = _mm_add_pd(fjy2,ty);
1063             fjz2             = _mm_add_pd(fjz2,tz);
1064
1065             /**************************
1066              * CALCULATE INTERACTIONS *
1067              **************************/
1068
1069             r20              = _mm_mul_pd(rsq20,rinv20);
1070
1071             /* Calculate table index by multiplying r with table scale and truncate to integer */
1072             rt               = _mm_mul_pd(r20,vftabscale);
1073             vfitab           = _mm_cvttpd_epi32(rt);
1074             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1075             vfitab           = _mm_slli_epi32(vfitab,2);
1076
1077             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1078             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1079             F                = _mm_setzero_pd();
1080             GMX_MM_TRANSPOSE2_PD(Y,F);
1081             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1082             H                = _mm_setzero_pd();
1083             GMX_MM_TRANSPOSE2_PD(G,H);
1084             Heps             = _mm_mul_pd(vfeps,H);
1085             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1086             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1087             velec            = _mm_mul_pd(qq20,VV);
1088             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1089             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1090
1091             /* Update potential sum for this i atom from the interaction with this j atom. */
1092             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1093             velecsum         = _mm_add_pd(velecsum,velec);
1094
1095             fscal            = felec;
1096
1097             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1098
1099             /* Calculate temporary vectorial force */
1100             tx               = _mm_mul_pd(fscal,dx20);
1101             ty               = _mm_mul_pd(fscal,dy20);
1102             tz               = _mm_mul_pd(fscal,dz20);
1103
1104             /* Update vectorial force */
1105             fix2             = _mm_add_pd(fix2,tx);
1106             fiy2             = _mm_add_pd(fiy2,ty);
1107             fiz2             = _mm_add_pd(fiz2,tz);
1108
1109             fjx0             = _mm_add_pd(fjx0,tx);
1110             fjy0             = _mm_add_pd(fjy0,ty);
1111             fjz0             = _mm_add_pd(fjz0,tz);
1112
1113             /**************************
1114              * CALCULATE INTERACTIONS *
1115              **************************/
1116
1117             r21              = _mm_mul_pd(rsq21,rinv21);
1118
1119             /* Calculate table index by multiplying r with table scale and truncate to integer */
1120             rt               = _mm_mul_pd(r21,vftabscale);
1121             vfitab           = _mm_cvttpd_epi32(rt);
1122             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1123             vfitab           = _mm_slli_epi32(vfitab,2);
1124
1125             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1126             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1127             F                = _mm_setzero_pd();
1128             GMX_MM_TRANSPOSE2_PD(Y,F);
1129             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1130             H                = _mm_setzero_pd();
1131             GMX_MM_TRANSPOSE2_PD(G,H);
1132             Heps             = _mm_mul_pd(vfeps,H);
1133             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1134             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1135             velec            = _mm_mul_pd(qq21,VV);
1136             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1137             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1138
1139             /* Update potential sum for this i atom from the interaction with this j atom. */
1140             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1141             velecsum         = _mm_add_pd(velecsum,velec);
1142
1143             fscal            = felec;
1144
1145             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1146
1147             /* Calculate temporary vectorial force */
1148             tx               = _mm_mul_pd(fscal,dx21);
1149             ty               = _mm_mul_pd(fscal,dy21);
1150             tz               = _mm_mul_pd(fscal,dz21);
1151
1152             /* Update vectorial force */
1153             fix2             = _mm_add_pd(fix2,tx);
1154             fiy2             = _mm_add_pd(fiy2,ty);
1155             fiz2             = _mm_add_pd(fiz2,tz);
1156
1157             fjx1             = _mm_add_pd(fjx1,tx);
1158             fjy1             = _mm_add_pd(fjy1,ty);
1159             fjz1             = _mm_add_pd(fjz1,tz);
1160
1161             /**************************
1162              * CALCULATE INTERACTIONS *
1163              **************************/
1164
1165             r22              = _mm_mul_pd(rsq22,rinv22);
1166
1167             /* Calculate table index by multiplying r with table scale and truncate to integer */
1168             rt               = _mm_mul_pd(r22,vftabscale);
1169             vfitab           = _mm_cvttpd_epi32(rt);
1170             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1171             vfitab           = _mm_slli_epi32(vfitab,2);
1172
1173             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1174             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1175             F                = _mm_setzero_pd();
1176             GMX_MM_TRANSPOSE2_PD(Y,F);
1177             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1178             H                = _mm_setzero_pd();
1179             GMX_MM_TRANSPOSE2_PD(G,H);
1180             Heps             = _mm_mul_pd(vfeps,H);
1181             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1182             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1183             velec            = _mm_mul_pd(qq22,VV);
1184             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1185             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1186
1187             /* Update potential sum for this i atom from the interaction with this j atom. */
1188             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1189             velecsum         = _mm_add_pd(velecsum,velec);
1190
1191             fscal            = felec;
1192
1193             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1194
1195             /* Calculate temporary vectorial force */
1196             tx               = _mm_mul_pd(fscal,dx22);
1197             ty               = _mm_mul_pd(fscal,dy22);
1198             tz               = _mm_mul_pd(fscal,dz22);
1199
1200             /* Update vectorial force */
1201             fix2             = _mm_add_pd(fix2,tx);
1202             fiy2             = _mm_add_pd(fiy2,ty);
1203             fiz2             = _mm_add_pd(fiz2,tz);
1204
1205             fjx2             = _mm_add_pd(fjx2,tx);
1206             fjy2             = _mm_add_pd(fjy2,ty);
1207             fjz2             = _mm_add_pd(fjz2,tz);
1208
1209             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1210
1211             /* Inner loop uses 400 flops */
1212         }
1213
1214         /* End of innermost loop */
1215
1216         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1217                                               f+i_coord_offset,fshift+i_shift_offset);
1218
1219         ggid                        = gid[iidx];
1220         /* Update potential energies */
1221         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1222         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1223
1224         /* Increment number of inner iterations */
1225         inneriter                  += j_index_end - j_index_start;
1226
1227         /* Outer loop uses 20 flops */
1228     }
1229
1230     /* Increment number of outer iterations */
1231     outeriter        += nri;
1232
1233     /* Update outer/inner flops */
1234
1235     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*400);
1236 }
1237 /*
1238  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3W3_F_sse2_double
1239  * Electrostatics interaction: CubicSplineTable
1240  * VdW interaction:            LennardJones
1241  * Geometry:                   Water3-Water3
1242  * Calculate force/pot:        Force
1243  */
1244 void
1245 nb_kernel_ElecCSTab_VdwLJ_GeomW3W3_F_sse2_double
1246                     (t_nblist                    * gmx_restrict       nlist,
1247                      rvec                        * gmx_restrict          xx,
1248                      rvec                        * gmx_restrict          ff,
1249                      t_forcerec                  * gmx_restrict          fr,
1250                      t_mdatoms                   * gmx_restrict     mdatoms,
1251                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1252                      t_nrnb                      * gmx_restrict        nrnb)
1253 {
1254     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1255      * just 0 for non-waters.
1256      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1257      * jnr indices corresponding to data put in the four positions in the SIMD register.
1258      */
1259     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1260     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1261     int              jnrA,jnrB;
1262     int              j_coord_offsetA,j_coord_offsetB;
1263     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1264     real             rcutoff_scalar;
1265     real             *shiftvec,*fshift,*x,*f;
1266     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1267     int              vdwioffset0;
1268     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1269     int              vdwioffset1;
1270     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1271     int              vdwioffset2;
1272     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1273     int              vdwjidx0A,vdwjidx0B;
1274     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1275     int              vdwjidx1A,vdwjidx1B;
1276     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1277     int              vdwjidx2A,vdwjidx2B;
1278     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1279     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1280     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1281     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1282     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1283     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1284     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1285     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1286     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1287     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1288     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1289     real             *charge;
1290     int              nvdwtype;
1291     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1292     int              *vdwtype;
1293     real             *vdwparam;
1294     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1295     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1296     __m128i          vfitab;
1297     __m128i          ifour       = _mm_set1_epi32(4);
1298     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1299     real             *vftab;
1300     __m128d          dummy_mask,cutoff_mask;
1301     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1302     __m128d          one     = _mm_set1_pd(1.0);
1303     __m128d          two     = _mm_set1_pd(2.0);
1304     x                = xx[0];
1305     f                = ff[0];
1306
1307     nri              = nlist->nri;
1308     iinr             = nlist->iinr;
1309     jindex           = nlist->jindex;
1310     jjnr             = nlist->jjnr;
1311     shiftidx         = nlist->shift;
1312     gid              = nlist->gid;
1313     shiftvec         = fr->shift_vec[0];
1314     fshift           = fr->fshift[0];
1315     facel            = _mm_set1_pd(fr->epsfac);
1316     charge           = mdatoms->chargeA;
1317     nvdwtype         = fr->ntype;
1318     vdwparam         = fr->nbfp;
1319     vdwtype          = mdatoms->typeA;
1320
1321     vftab            = kernel_data->table_elec->data;
1322     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
1323
1324     /* Setup water-specific parameters */
1325     inr              = nlist->iinr[0];
1326     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1327     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1328     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1329     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1330
1331     jq0              = _mm_set1_pd(charge[inr+0]);
1332     jq1              = _mm_set1_pd(charge[inr+1]);
1333     jq2              = _mm_set1_pd(charge[inr+2]);
1334     vdwjidx0A        = 2*vdwtype[inr+0];
1335     qq00             = _mm_mul_pd(iq0,jq0);
1336     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1337     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1338     qq01             = _mm_mul_pd(iq0,jq1);
1339     qq02             = _mm_mul_pd(iq0,jq2);
1340     qq10             = _mm_mul_pd(iq1,jq0);
1341     qq11             = _mm_mul_pd(iq1,jq1);
1342     qq12             = _mm_mul_pd(iq1,jq2);
1343     qq20             = _mm_mul_pd(iq2,jq0);
1344     qq21             = _mm_mul_pd(iq2,jq1);
1345     qq22             = _mm_mul_pd(iq2,jq2);
1346
1347     /* Avoid stupid compiler warnings */
1348     jnrA = jnrB = 0;
1349     j_coord_offsetA = 0;
1350     j_coord_offsetB = 0;
1351
1352     outeriter        = 0;
1353     inneriter        = 0;
1354
1355     /* Start outer loop over neighborlists */
1356     for(iidx=0; iidx<nri; iidx++)
1357     {
1358         /* Load shift vector for this list */
1359         i_shift_offset   = DIM*shiftidx[iidx];
1360
1361         /* Load limits for loop over neighbors */
1362         j_index_start    = jindex[iidx];
1363         j_index_end      = jindex[iidx+1];
1364
1365         /* Get outer coordinate index */
1366         inr              = iinr[iidx];
1367         i_coord_offset   = DIM*inr;
1368
1369         /* Load i particle coords and add shift vector */
1370         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1371                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1372
1373         fix0             = _mm_setzero_pd();
1374         fiy0             = _mm_setzero_pd();
1375         fiz0             = _mm_setzero_pd();
1376         fix1             = _mm_setzero_pd();
1377         fiy1             = _mm_setzero_pd();
1378         fiz1             = _mm_setzero_pd();
1379         fix2             = _mm_setzero_pd();
1380         fiy2             = _mm_setzero_pd();
1381         fiz2             = _mm_setzero_pd();
1382
1383         /* Start inner kernel loop */
1384         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1385         {
1386
1387             /* Get j neighbor index, and coordinate index */
1388             jnrA             = jjnr[jidx];
1389             jnrB             = jjnr[jidx+1];
1390             j_coord_offsetA  = DIM*jnrA;
1391             j_coord_offsetB  = DIM*jnrB;
1392
1393             /* load j atom coordinates */
1394             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1395                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1396
1397             /* Calculate displacement vector */
1398             dx00             = _mm_sub_pd(ix0,jx0);
1399             dy00             = _mm_sub_pd(iy0,jy0);
1400             dz00             = _mm_sub_pd(iz0,jz0);
1401             dx01             = _mm_sub_pd(ix0,jx1);
1402             dy01             = _mm_sub_pd(iy0,jy1);
1403             dz01             = _mm_sub_pd(iz0,jz1);
1404             dx02             = _mm_sub_pd(ix0,jx2);
1405             dy02             = _mm_sub_pd(iy0,jy2);
1406             dz02             = _mm_sub_pd(iz0,jz2);
1407             dx10             = _mm_sub_pd(ix1,jx0);
1408             dy10             = _mm_sub_pd(iy1,jy0);
1409             dz10             = _mm_sub_pd(iz1,jz0);
1410             dx11             = _mm_sub_pd(ix1,jx1);
1411             dy11             = _mm_sub_pd(iy1,jy1);
1412             dz11             = _mm_sub_pd(iz1,jz1);
1413             dx12             = _mm_sub_pd(ix1,jx2);
1414             dy12             = _mm_sub_pd(iy1,jy2);
1415             dz12             = _mm_sub_pd(iz1,jz2);
1416             dx20             = _mm_sub_pd(ix2,jx0);
1417             dy20             = _mm_sub_pd(iy2,jy0);
1418             dz20             = _mm_sub_pd(iz2,jz0);
1419             dx21             = _mm_sub_pd(ix2,jx1);
1420             dy21             = _mm_sub_pd(iy2,jy1);
1421             dz21             = _mm_sub_pd(iz2,jz1);
1422             dx22             = _mm_sub_pd(ix2,jx2);
1423             dy22             = _mm_sub_pd(iy2,jy2);
1424             dz22             = _mm_sub_pd(iz2,jz2);
1425
1426             /* Calculate squared distance and things based on it */
1427             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1428             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1429             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1430             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1431             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1432             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1433             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1434             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1435             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1436
1437             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1438             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1439             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1440             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1441             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1442             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1443             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1444             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1445             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1446
1447             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1448
1449             fjx0             = _mm_setzero_pd();
1450             fjy0             = _mm_setzero_pd();
1451             fjz0             = _mm_setzero_pd();
1452             fjx1             = _mm_setzero_pd();
1453             fjy1             = _mm_setzero_pd();
1454             fjz1             = _mm_setzero_pd();
1455             fjx2             = _mm_setzero_pd();
1456             fjy2             = _mm_setzero_pd();
1457             fjz2             = _mm_setzero_pd();
1458
1459             /**************************
1460              * CALCULATE INTERACTIONS *
1461              **************************/
1462
1463             r00              = _mm_mul_pd(rsq00,rinv00);
1464
1465             /* Calculate table index by multiplying r with table scale and truncate to integer */
1466             rt               = _mm_mul_pd(r00,vftabscale);
1467             vfitab           = _mm_cvttpd_epi32(rt);
1468             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1469             vfitab           = _mm_slli_epi32(vfitab,2);
1470
1471             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1472             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1473             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1474             GMX_MM_TRANSPOSE2_PD(Y,F);
1475             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1476             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1477             GMX_MM_TRANSPOSE2_PD(G,H);
1478             Heps             = _mm_mul_pd(vfeps,H);
1479             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1480             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1481             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1482
1483             /* LENNARD-JONES DISPERSION/REPULSION */
1484
1485             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1486             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1487
1488             fscal            = _mm_add_pd(felec,fvdw);
1489
1490             /* Calculate temporary vectorial force */
1491             tx               = _mm_mul_pd(fscal,dx00);
1492             ty               = _mm_mul_pd(fscal,dy00);
1493             tz               = _mm_mul_pd(fscal,dz00);
1494
1495             /* Update vectorial force */
1496             fix0             = _mm_add_pd(fix0,tx);
1497             fiy0             = _mm_add_pd(fiy0,ty);
1498             fiz0             = _mm_add_pd(fiz0,tz);
1499
1500             fjx0             = _mm_add_pd(fjx0,tx);
1501             fjy0             = _mm_add_pd(fjy0,ty);
1502             fjz0             = _mm_add_pd(fjz0,tz);
1503
1504             /**************************
1505              * CALCULATE INTERACTIONS *
1506              **************************/
1507
1508             r01              = _mm_mul_pd(rsq01,rinv01);
1509
1510             /* Calculate table index by multiplying r with table scale and truncate to integer */
1511             rt               = _mm_mul_pd(r01,vftabscale);
1512             vfitab           = _mm_cvttpd_epi32(rt);
1513             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1514             vfitab           = _mm_slli_epi32(vfitab,2);
1515
1516             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1517             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1518             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1519             GMX_MM_TRANSPOSE2_PD(Y,F);
1520             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1521             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1522             GMX_MM_TRANSPOSE2_PD(G,H);
1523             Heps             = _mm_mul_pd(vfeps,H);
1524             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1525             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1526             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1527
1528             fscal            = felec;
1529
1530             /* Calculate temporary vectorial force */
1531             tx               = _mm_mul_pd(fscal,dx01);
1532             ty               = _mm_mul_pd(fscal,dy01);
1533             tz               = _mm_mul_pd(fscal,dz01);
1534
1535             /* Update vectorial force */
1536             fix0             = _mm_add_pd(fix0,tx);
1537             fiy0             = _mm_add_pd(fiy0,ty);
1538             fiz0             = _mm_add_pd(fiz0,tz);
1539
1540             fjx1             = _mm_add_pd(fjx1,tx);
1541             fjy1             = _mm_add_pd(fjy1,ty);
1542             fjz1             = _mm_add_pd(fjz1,tz);
1543
1544             /**************************
1545              * CALCULATE INTERACTIONS *
1546              **************************/
1547
1548             r02              = _mm_mul_pd(rsq02,rinv02);
1549
1550             /* Calculate table index by multiplying r with table scale and truncate to integer */
1551             rt               = _mm_mul_pd(r02,vftabscale);
1552             vfitab           = _mm_cvttpd_epi32(rt);
1553             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1554             vfitab           = _mm_slli_epi32(vfitab,2);
1555
1556             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1557             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1558             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1559             GMX_MM_TRANSPOSE2_PD(Y,F);
1560             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1561             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1562             GMX_MM_TRANSPOSE2_PD(G,H);
1563             Heps             = _mm_mul_pd(vfeps,H);
1564             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1565             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1566             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
1567
1568             fscal            = felec;
1569
1570             /* Calculate temporary vectorial force */
1571             tx               = _mm_mul_pd(fscal,dx02);
1572             ty               = _mm_mul_pd(fscal,dy02);
1573             tz               = _mm_mul_pd(fscal,dz02);
1574
1575             /* Update vectorial force */
1576             fix0             = _mm_add_pd(fix0,tx);
1577             fiy0             = _mm_add_pd(fiy0,ty);
1578             fiz0             = _mm_add_pd(fiz0,tz);
1579
1580             fjx2             = _mm_add_pd(fjx2,tx);
1581             fjy2             = _mm_add_pd(fjy2,ty);
1582             fjz2             = _mm_add_pd(fjz2,tz);
1583
1584             /**************************
1585              * CALCULATE INTERACTIONS *
1586              **************************/
1587
1588             r10              = _mm_mul_pd(rsq10,rinv10);
1589
1590             /* Calculate table index by multiplying r with table scale and truncate to integer */
1591             rt               = _mm_mul_pd(r10,vftabscale);
1592             vfitab           = _mm_cvttpd_epi32(rt);
1593             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1594             vfitab           = _mm_slli_epi32(vfitab,2);
1595
1596             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1597             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1598             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1599             GMX_MM_TRANSPOSE2_PD(Y,F);
1600             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1601             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1602             GMX_MM_TRANSPOSE2_PD(G,H);
1603             Heps             = _mm_mul_pd(vfeps,H);
1604             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1605             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1606             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
1607
1608             fscal            = felec;
1609
1610             /* Calculate temporary vectorial force */
1611             tx               = _mm_mul_pd(fscal,dx10);
1612             ty               = _mm_mul_pd(fscal,dy10);
1613             tz               = _mm_mul_pd(fscal,dz10);
1614
1615             /* Update vectorial force */
1616             fix1             = _mm_add_pd(fix1,tx);
1617             fiy1             = _mm_add_pd(fiy1,ty);
1618             fiz1             = _mm_add_pd(fiz1,tz);
1619
1620             fjx0             = _mm_add_pd(fjx0,tx);
1621             fjy0             = _mm_add_pd(fjy0,ty);
1622             fjz0             = _mm_add_pd(fjz0,tz);
1623
1624             /**************************
1625              * CALCULATE INTERACTIONS *
1626              **************************/
1627
1628             r11              = _mm_mul_pd(rsq11,rinv11);
1629
1630             /* Calculate table index by multiplying r with table scale and truncate to integer */
1631             rt               = _mm_mul_pd(r11,vftabscale);
1632             vfitab           = _mm_cvttpd_epi32(rt);
1633             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1634             vfitab           = _mm_slli_epi32(vfitab,2);
1635
1636             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1637             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1638             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1639             GMX_MM_TRANSPOSE2_PD(Y,F);
1640             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1641             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1642             GMX_MM_TRANSPOSE2_PD(G,H);
1643             Heps             = _mm_mul_pd(vfeps,H);
1644             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1645             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1646             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1647
1648             fscal            = felec;
1649
1650             /* Calculate temporary vectorial force */
1651             tx               = _mm_mul_pd(fscal,dx11);
1652             ty               = _mm_mul_pd(fscal,dy11);
1653             tz               = _mm_mul_pd(fscal,dz11);
1654
1655             /* Update vectorial force */
1656             fix1             = _mm_add_pd(fix1,tx);
1657             fiy1             = _mm_add_pd(fiy1,ty);
1658             fiz1             = _mm_add_pd(fiz1,tz);
1659
1660             fjx1             = _mm_add_pd(fjx1,tx);
1661             fjy1             = _mm_add_pd(fjy1,ty);
1662             fjz1             = _mm_add_pd(fjz1,tz);
1663
1664             /**************************
1665              * CALCULATE INTERACTIONS *
1666              **************************/
1667
1668             r12              = _mm_mul_pd(rsq12,rinv12);
1669
1670             /* Calculate table index by multiplying r with table scale and truncate to integer */
1671             rt               = _mm_mul_pd(r12,vftabscale);
1672             vfitab           = _mm_cvttpd_epi32(rt);
1673             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1674             vfitab           = _mm_slli_epi32(vfitab,2);
1675
1676             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1677             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1678             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1679             GMX_MM_TRANSPOSE2_PD(Y,F);
1680             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1681             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1682             GMX_MM_TRANSPOSE2_PD(G,H);
1683             Heps             = _mm_mul_pd(vfeps,H);
1684             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1685             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1686             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1687
1688             fscal            = felec;
1689
1690             /* Calculate temporary vectorial force */
1691             tx               = _mm_mul_pd(fscal,dx12);
1692             ty               = _mm_mul_pd(fscal,dy12);
1693             tz               = _mm_mul_pd(fscal,dz12);
1694
1695             /* Update vectorial force */
1696             fix1             = _mm_add_pd(fix1,tx);
1697             fiy1             = _mm_add_pd(fiy1,ty);
1698             fiz1             = _mm_add_pd(fiz1,tz);
1699
1700             fjx2             = _mm_add_pd(fjx2,tx);
1701             fjy2             = _mm_add_pd(fjy2,ty);
1702             fjz2             = _mm_add_pd(fjz2,tz);
1703
1704             /**************************
1705              * CALCULATE INTERACTIONS *
1706              **************************/
1707
1708             r20              = _mm_mul_pd(rsq20,rinv20);
1709
1710             /* Calculate table index by multiplying r with table scale and truncate to integer */
1711             rt               = _mm_mul_pd(r20,vftabscale);
1712             vfitab           = _mm_cvttpd_epi32(rt);
1713             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1714             vfitab           = _mm_slli_epi32(vfitab,2);
1715
1716             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1717             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1718             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1719             GMX_MM_TRANSPOSE2_PD(Y,F);
1720             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1721             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1722             GMX_MM_TRANSPOSE2_PD(G,H);
1723             Heps             = _mm_mul_pd(vfeps,H);
1724             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1725             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1726             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
1727
1728             fscal            = felec;
1729
1730             /* Calculate temporary vectorial force */
1731             tx               = _mm_mul_pd(fscal,dx20);
1732             ty               = _mm_mul_pd(fscal,dy20);
1733             tz               = _mm_mul_pd(fscal,dz20);
1734
1735             /* Update vectorial force */
1736             fix2             = _mm_add_pd(fix2,tx);
1737             fiy2             = _mm_add_pd(fiy2,ty);
1738             fiz2             = _mm_add_pd(fiz2,tz);
1739
1740             fjx0             = _mm_add_pd(fjx0,tx);
1741             fjy0             = _mm_add_pd(fjy0,ty);
1742             fjz0             = _mm_add_pd(fjz0,tz);
1743
1744             /**************************
1745              * CALCULATE INTERACTIONS *
1746              **************************/
1747
1748             r21              = _mm_mul_pd(rsq21,rinv21);
1749
1750             /* Calculate table index by multiplying r with table scale and truncate to integer */
1751             rt               = _mm_mul_pd(r21,vftabscale);
1752             vfitab           = _mm_cvttpd_epi32(rt);
1753             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1754             vfitab           = _mm_slli_epi32(vfitab,2);
1755
1756             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1757             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1758             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1759             GMX_MM_TRANSPOSE2_PD(Y,F);
1760             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1761             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1762             GMX_MM_TRANSPOSE2_PD(G,H);
1763             Heps             = _mm_mul_pd(vfeps,H);
1764             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1765             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1766             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1767
1768             fscal            = felec;
1769
1770             /* Calculate temporary vectorial force */
1771             tx               = _mm_mul_pd(fscal,dx21);
1772             ty               = _mm_mul_pd(fscal,dy21);
1773             tz               = _mm_mul_pd(fscal,dz21);
1774
1775             /* Update vectorial force */
1776             fix2             = _mm_add_pd(fix2,tx);
1777             fiy2             = _mm_add_pd(fiy2,ty);
1778             fiz2             = _mm_add_pd(fiz2,tz);
1779
1780             fjx1             = _mm_add_pd(fjx1,tx);
1781             fjy1             = _mm_add_pd(fjy1,ty);
1782             fjz1             = _mm_add_pd(fjz1,tz);
1783
1784             /**************************
1785              * CALCULATE INTERACTIONS *
1786              **************************/
1787
1788             r22              = _mm_mul_pd(rsq22,rinv22);
1789
1790             /* Calculate table index by multiplying r with table scale and truncate to integer */
1791             rt               = _mm_mul_pd(r22,vftabscale);
1792             vfitab           = _mm_cvttpd_epi32(rt);
1793             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1794             vfitab           = _mm_slli_epi32(vfitab,2);
1795
1796             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1797             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1798             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1799             GMX_MM_TRANSPOSE2_PD(Y,F);
1800             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1801             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1802             GMX_MM_TRANSPOSE2_PD(G,H);
1803             Heps             = _mm_mul_pd(vfeps,H);
1804             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1805             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1806             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1807
1808             fscal            = felec;
1809
1810             /* Calculate temporary vectorial force */
1811             tx               = _mm_mul_pd(fscal,dx22);
1812             ty               = _mm_mul_pd(fscal,dy22);
1813             tz               = _mm_mul_pd(fscal,dz22);
1814
1815             /* Update vectorial force */
1816             fix2             = _mm_add_pd(fix2,tx);
1817             fiy2             = _mm_add_pd(fiy2,ty);
1818             fiz2             = _mm_add_pd(fiz2,tz);
1819
1820             fjx2             = _mm_add_pd(fjx2,tx);
1821             fjy2             = _mm_add_pd(fjy2,ty);
1822             fjz2             = _mm_add_pd(fjz2,tz);
1823
1824             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1825
1826             /* Inner loop uses 359 flops */
1827         }
1828
1829         if(jidx<j_index_end)
1830         {
1831
1832             jnrA             = jjnr[jidx];
1833             j_coord_offsetA  = DIM*jnrA;
1834
1835             /* load j atom coordinates */
1836             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1837                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1838
1839             /* Calculate displacement vector */
1840             dx00             = _mm_sub_pd(ix0,jx0);
1841             dy00             = _mm_sub_pd(iy0,jy0);
1842             dz00             = _mm_sub_pd(iz0,jz0);
1843             dx01             = _mm_sub_pd(ix0,jx1);
1844             dy01             = _mm_sub_pd(iy0,jy1);
1845             dz01             = _mm_sub_pd(iz0,jz1);
1846             dx02             = _mm_sub_pd(ix0,jx2);
1847             dy02             = _mm_sub_pd(iy0,jy2);
1848             dz02             = _mm_sub_pd(iz0,jz2);
1849             dx10             = _mm_sub_pd(ix1,jx0);
1850             dy10             = _mm_sub_pd(iy1,jy0);
1851             dz10             = _mm_sub_pd(iz1,jz0);
1852             dx11             = _mm_sub_pd(ix1,jx1);
1853             dy11             = _mm_sub_pd(iy1,jy1);
1854             dz11             = _mm_sub_pd(iz1,jz1);
1855             dx12             = _mm_sub_pd(ix1,jx2);
1856             dy12             = _mm_sub_pd(iy1,jy2);
1857             dz12             = _mm_sub_pd(iz1,jz2);
1858             dx20             = _mm_sub_pd(ix2,jx0);
1859             dy20             = _mm_sub_pd(iy2,jy0);
1860             dz20             = _mm_sub_pd(iz2,jz0);
1861             dx21             = _mm_sub_pd(ix2,jx1);
1862             dy21             = _mm_sub_pd(iy2,jy1);
1863             dz21             = _mm_sub_pd(iz2,jz1);
1864             dx22             = _mm_sub_pd(ix2,jx2);
1865             dy22             = _mm_sub_pd(iy2,jy2);
1866             dz22             = _mm_sub_pd(iz2,jz2);
1867
1868             /* Calculate squared distance and things based on it */
1869             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1870             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1871             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1872             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1873             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1874             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1875             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1876             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1877             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1878
1879             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1880             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1881             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1882             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1883             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1884             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1885             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1886             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1887             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1888
1889             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1890
1891             fjx0             = _mm_setzero_pd();
1892             fjy0             = _mm_setzero_pd();
1893             fjz0             = _mm_setzero_pd();
1894             fjx1             = _mm_setzero_pd();
1895             fjy1             = _mm_setzero_pd();
1896             fjz1             = _mm_setzero_pd();
1897             fjx2             = _mm_setzero_pd();
1898             fjy2             = _mm_setzero_pd();
1899             fjz2             = _mm_setzero_pd();
1900
1901             /**************************
1902              * CALCULATE INTERACTIONS *
1903              **************************/
1904
1905             r00              = _mm_mul_pd(rsq00,rinv00);
1906
1907             /* Calculate table index by multiplying r with table scale and truncate to integer */
1908             rt               = _mm_mul_pd(r00,vftabscale);
1909             vfitab           = _mm_cvttpd_epi32(rt);
1910             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1911             vfitab           = _mm_slli_epi32(vfitab,2);
1912
1913             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1914             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1915             F                = _mm_setzero_pd();
1916             GMX_MM_TRANSPOSE2_PD(Y,F);
1917             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1918             H                = _mm_setzero_pd();
1919             GMX_MM_TRANSPOSE2_PD(G,H);
1920             Heps             = _mm_mul_pd(vfeps,H);
1921             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1922             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1923             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq00,FF),_mm_mul_pd(vftabscale,rinv00)));
1924
1925             /* LENNARD-JONES DISPERSION/REPULSION */
1926
1927             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1928             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1929
1930             fscal            = _mm_add_pd(felec,fvdw);
1931
1932             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1933
1934             /* Calculate temporary vectorial force */
1935             tx               = _mm_mul_pd(fscal,dx00);
1936             ty               = _mm_mul_pd(fscal,dy00);
1937             tz               = _mm_mul_pd(fscal,dz00);
1938
1939             /* Update vectorial force */
1940             fix0             = _mm_add_pd(fix0,tx);
1941             fiy0             = _mm_add_pd(fiy0,ty);
1942             fiz0             = _mm_add_pd(fiz0,tz);
1943
1944             fjx0             = _mm_add_pd(fjx0,tx);
1945             fjy0             = _mm_add_pd(fjy0,ty);
1946             fjz0             = _mm_add_pd(fjz0,tz);
1947
1948             /**************************
1949              * CALCULATE INTERACTIONS *
1950              **************************/
1951
1952             r01              = _mm_mul_pd(rsq01,rinv01);
1953
1954             /* Calculate table index by multiplying r with table scale and truncate to integer */
1955             rt               = _mm_mul_pd(r01,vftabscale);
1956             vfitab           = _mm_cvttpd_epi32(rt);
1957             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1958             vfitab           = _mm_slli_epi32(vfitab,2);
1959
1960             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1961             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1962             F                = _mm_setzero_pd();
1963             GMX_MM_TRANSPOSE2_PD(Y,F);
1964             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1965             H                = _mm_setzero_pd();
1966             GMX_MM_TRANSPOSE2_PD(G,H);
1967             Heps             = _mm_mul_pd(vfeps,H);
1968             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1969             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1970             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq01,FF),_mm_mul_pd(vftabscale,rinv01)));
1971
1972             fscal            = felec;
1973
1974             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1975
1976             /* Calculate temporary vectorial force */
1977             tx               = _mm_mul_pd(fscal,dx01);
1978             ty               = _mm_mul_pd(fscal,dy01);
1979             tz               = _mm_mul_pd(fscal,dz01);
1980
1981             /* Update vectorial force */
1982             fix0             = _mm_add_pd(fix0,tx);
1983             fiy0             = _mm_add_pd(fiy0,ty);
1984             fiz0             = _mm_add_pd(fiz0,tz);
1985
1986             fjx1             = _mm_add_pd(fjx1,tx);
1987             fjy1             = _mm_add_pd(fjy1,ty);
1988             fjz1             = _mm_add_pd(fjz1,tz);
1989
1990             /**************************
1991              * CALCULATE INTERACTIONS *
1992              **************************/
1993
1994             r02              = _mm_mul_pd(rsq02,rinv02);
1995
1996             /* Calculate table index by multiplying r with table scale and truncate to integer */
1997             rt               = _mm_mul_pd(r02,vftabscale);
1998             vfitab           = _mm_cvttpd_epi32(rt);
1999             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2000             vfitab           = _mm_slli_epi32(vfitab,2);
2001
2002             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2003             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2004             F                = _mm_setzero_pd();
2005             GMX_MM_TRANSPOSE2_PD(Y,F);
2006             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2007             H                = _mm_setzero_pd();
2008             GMX_MM_TRANSPOSE2_PD(G,H);
2009             Heps             = _mm_mul_pd(vfeps,H);
2010             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2011             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2012             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq02,FF),_mm_mul_pd(vftabscale,rinv02)));
2013
2014             fscal            = felec;
2015
2016             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2017
2018             /* Calculate temporary vectorial force */
2019             tx               = _mm_mul_pd(fscal,dx02);
2020             ty               = _mm_mul_pd(fscal,dy02);
2021             tz               = _mm_mul_pd(fscal,dz02);
2022
2023             /* Update vectorial force */
2024             fix0             = _mm_add_pd(fix0,tx);
2025             fiy0             = _mm_add_pd(fiy0,ty);
2026             fiz0             = _mm_add_pd(fiz0,tz);
2027
2028             fjx2             = _mm_add_pd(fjx2,tx);
2029             fjy2             = _mm_add_pd(fjy2,ty);
2030             fjz2             = _mm_add_pd(fjz2,tz);
2031
2032             /**************************
2033              * CALCULATE INTERACTIONS *
2034              **************************/
2035
2036             r10              = _mm_mul_pd(rsq10,rinv10);
2037
2038             /* Calculate table index by multiplying r with table scale and truncate to integer */
2039             rt               = _mm_mul_pd(r10,vftabscale);
2040             vfitab           = _mm_cvttpd_epi32(rt);
2041             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2042             vfitab           = _mm_slli_epi32(vfitab,2);
2043
2044             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2045             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2046             F                = _mm_setzero_pd();
2047             GMX_MM_TRANSPOSE2_PD(Y,F);
2048             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2049             H                = _mm_setzero_pd();
2050             GMX_MM_TRANSPOSE2_PD(G,H);
2051             Heps             = _mm_mul_pd(vfeps,H);
2052             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2053             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2054             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq10,FF),_mm_mul_pd(vftabscale,rinv10)));
2055
2056             fscal            = felec;
2057
2058             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2059
2060             /* Calculate temporary vectorial force */
2061             tx               = _mm_mul_pd(fscal,dx10);
2062             ty               = _mm_mul_pd(fscal,dy10);
2063             tz               = _mm_mul_pd(fscal,dz10);
2064
2065             /* Update vectorial force */
2066             fix1             = _mm_add_pd(fix1,tx);
2067             fiy1             = _mm_add_pd(fiy1,ty);
2068             fiz1             = _mm_add_pd(fiz1,tz);
2069
2070             fjx0             = _mm_add_pd(fjx0,tx);
2071             fjy0             = _mm_add_pd(fjy0,ty);
2072             fjz0             = _mm_add_pd(fjz0,tz);
2073
2074             /**************************
2075              * CALCULATE INTERACTIONS *
2076              **************************/
2077
2078             r11              = _mm_mul_pd(rsq11,rinv11);
2079
2080             /* Calculate table index by multiplying r with table scale and truncate to integer */
2081             rt               = _mm_mul_pd(r11,vftabscale);
2082             vfitab           = _mm_cvttpd_epi32(rt);
2083             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2084             vfitab           = _mm_slli_epi32(vfitab,2);
2085
2086             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2087             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2088             F                = _mm_setzero_pd();
2089             GMX_MM_TRANSPOSE2_PD(Y,F);
2090             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2091             H                = _mm_setzero_pd();
2092             GMX_MM_TRANSPOSE2_PD(G,H);
2093             Heps             = _mm_mul_pd(vfeps,H);
2094             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2095             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2096             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2097
2098             fscal            = felec;
2099
2100             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2101
2102             /* Calculate temporary vectorial force */
2103             tx               = _mm_mul_pd(fscal,dx11);
2104             ty               = _mm_mul_pd(fscal,dy11);
2105             tz               = _mm_mul_pd(fscal,dz11);
2106
2107             /* Update vectorial force */
2108             fix1             = _mm_add_pd(fix1,tx);
2109             fiy1             = _mm_add_pd(fiy1,ty);
2110             fiz1             = _mm_add_pd(fiz1,tz);
2111
2112             fjx1             = _mm_add_pd(fjx1,tx);
2113             fjy1             = _mm_add_pd(fjy1,ty);
2114             fjz1             = _mm_add_pd(fjz1,tz);
2115
2116             /**************************
2117              * CALCULATE INTERACTIONS *
2118              **************************/
2119
2120             r12              = _mm_mul_pd(rsq12,rinv12);
2121
2122             /* Calculate table index by multiplying r with table scale and truncate to integer */
2123             rt               = _mm_mul_pd(r12,vftabscale);
2124             vfitab           = _mm_cvttpd_epi32(rt);
2125             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2126             vfitab           = _mm_slli_epi32(vfitab,2);
2127
2128             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2129             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2130             F                = _mm_setzero_pd();
2131             GMX_MM_TRANSPOSE2_PD(Y,F);
2132             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2133             H                = _mm_setzero_pd();
2134             GMX_MM_TRANSPOSE2_PD(G,H);
2135             Heps             = _mm_mul_pd(vfeps,H);
2136             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2137             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2138             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2139
2140             fscal            = felec;
2141
2142             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2143
2144             /* Calculate temporary vectorial force */
2145             tx               = _mm_mul_pd(fscal,dx12);
2146             ty               = _mm_mul_pd(fscal,dy12);
2147             tz               = _mm_mul_pd(fscal,dz12);
2148
2149             /* Update vectorial force */
2150             fix1             = _mm_add_pd(fix1,tx);
2151             fiy1             = _mm_add_pd(fiy1,ty);
2152             fiz1             = _mm_add_pd(fiz1,tz);
2153
2154             fjx2             = _mm_add_pd(fjx2,tx);
2155             fjy2             = _mm_add_pd(fjy2,ty);
2156             fjz2             = _mm_add_pd(fjz2,tz);
2157
2158             /**************************
2159              * CALCULATE INTERACTIONS *
2160              **************************/
2161
2162             r20              = _mm_mul_pd(rsq20,rinv20);
2163
2164             /* Calculate table index by multiplying r with table scale and truncate to integer */
2165             rt               = _mm_mul_pd(r20,vftabscale);
2166             vfitab           = _mm_cvttpd_epi32(rt);
2167             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2168             vfitab           = _mm_slli_epi32(vfitab,2);
2169
2170             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2171             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2172             F                = _mm_setzero_pd();
2173             GMX_MM_TRANSPOSE2_PD(Y,F);
2174             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2175             H                = _mm_setzero_pd();
2176             GMX_MM_TRANSPOSE2_PD(G,H);
2177             Heps             = _mm_mul_pd(vfeps,H);
2178             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2179             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2180             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq20,FF),_mm_mul_pd(vftabscale,rinv20)));
2181
2182             fscal            = felec;
2183
2184             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2185
2186             /* Calculate temporary vectorial force */
2187             tx               = _mm_mul_pd(fscal,dx20);
2188             ty               = _mm_mul_pd(fscal,dy20);
2189             tz               = _mm_mul_pd(fscal,dz20);
2190
2191             /* Update vectorial force */
2192             fix2             = _mm_add_pd(fix2,tx);
2193             fiy2             = _mm_add_pd(fiy2,ty);
2194             fiz2             = _mm_add_pd(fiz2,tz);
2195
2196             fjx0             = _mm_add_pd(fjx0,tx);
2197             fjy0             = _mm_add_pd(fjy0,ty);
2198             fjz0             = _mm_add_pd(fjz0,tz);
2199
2200             /**************************
2201              * CALCULATE INTERACTIONS *
2202              **************************/
2203
2204             r21              = _mm_mul_pd(rsq21,rinv21);
2205
2206             /* Calculate table index by multiplying r with table scale and truncate to integer */
2207             rt               = _mm_mul_pd(r21,vftabscale);
2208             vfitab           = _mm_cvttpd_epi32(rt);
2209             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2210             vfitab           = _mm_slli_epi32(vfitab,2);
2211
2212             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2213             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2214             F                = _mm_setzero_pd();
2215             GMX_MM_TRANSPOSE2_PD(Y,F);
2216             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2217             H                = _mm_setzero_pd();
2218             GMX_MM_TRANSPOSE2_PD(G,H);
2219             Heps             = _mm_mul_pd(vfeps,H);
2220             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2221             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2222             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2223
2224             fscal            = felec;
2225
2226             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2227
2228             /* Calculate temporary vectorial force */
2229             tx               = _mm_mul_pd(fscal,dx21);
2230             ty               = _mm_mul_pd(fscal,dy21);
2231             tz               = _mm_mul_pd(fscal,dz21);
2232
2233             /* Update vectorial force */
2234             fix2             = _mm_add_pd(fix2,tx);
2235             fiy2             = _mm_add_pd(fiy2,ty);
2236             fiz2             = _mm_add_pd(fiz2,tz);
2237
2238             fjx1             = _mm_add_pd(fjx1,tx);
2239             fjy1             = _mm_add_pd(fjy1,ty);
2240             fjz1             = _mm_add_pd(fjz1,tz);
2241
2242             /**************************
2243              * CALCULATE INTERACTIONS *
2244              **************************/
2245
2246             r22              = _mm_mul_pd(rsq22,rinv22);
2247
2248             /* Calculate table index by multiplying r with table scale and truncate to integer */
2249             rt               = _mm_mul_pd(r22,vftabscale);
2250             vfitab           = _mm_cvttpd_epi32(rt);
2251             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2252             vfitab           = _mm_slli_epi32(vfitab,2);
2253
2254             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2255             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2256             F                = _mm_setzero_pd();
2257             GMX_MM_TRANSPOSE2_PD(Y,F);
2258             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2259             H                = _mm_setzero_pd();
2260             GMX_MM_TRANSPOSE2_PD(G,H);
2261             Heps             = _mm_mul_pd(vfeps,H);
2262             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2263             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2264             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2265
2266             fscal            = felec;
2267
2268             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2269
2270             /* Calculate temporary vectorial force */
2271             tx               = _mm_mul_pd(fscal,dx22);
2272             ty               = _mm_mul_pd(fscal,dy22);
2273             tz               = _mm_mul_pd(fscal,dz22);
2274
2275             /* Update vectorial force */
2276             fix2             = _mm_add_pd(fix2,tx);
2277             fiy2             = _mm_add_pd(fiy2,ty);
2278             fiz2             = _mm_add_pd(fiz2,tz);
2279
2280             fjx2             = _mm_add_pd(fjx2,tx);
2281             fjy2             = _mm_add_pd(fjy2,ty);
2282             fjz2             = _mm_add_pd(fjz2,tz);
2283
2284             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2285
2286             /* Inner loop uses 359 flops */
2287         }
2288
2289         /* End of innermost loop */
2290
2291         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2292                                               f+i_coord_offset,fshift+i_shift_offset);
2293
2294         /* Increment number of inner iterations */
2295         inneriter                  += j_index_end - j_index_start;
2296
2297         /* Outer loop uses 18 flops */
2298     }
2299
2300     /* Increment number of outer iterations */
2301     outeriter        += nri;
2302
2303     /* Update outer/inner flops */
2304
2305     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*359);
2306 }