src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecEw_VdwCSTab_GeomW4W4_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_ElecEw_VdwCSTab_GeomW4W4_VF_sse2_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_VF_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              vdwioffset3;
  87     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  88     int              vdwjidx0A,vdwjidx0B;
  89     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  90     int              vdwjidx1A,vdwjidx1B;
  91     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  92     int              vdwjidx2A,vdwjidx2B;
  93     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  94     int              vdwjidx3A,vdwjidx3B;
  95     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  96     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  97     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  98     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  99     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 100     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 101     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 102     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 103     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 104     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 105     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 106     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     int              nvdwtype;
 109     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 110     int              *vdwtype;
 111     real             *vdwparam;
 112     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 113     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 114     __m128i          vfitab;
 115     __m128i          ifour       = _mm_set1_epi32(4);
 116     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 117     real             *vftab;
 118     __m128i          ewitab;
 119     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 120     real             *ewtab;
 121     __m128d          dummy_mask,cutoff_mask;
 122     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 123     __m128d          one     = _mm_set1_pd(1.0);
 124     __m128d          two     = _mm_set1_pd(2.0);
 125     x                = xx[0];
 126     f                = ff[0];
 127
 128     nri              = nlist->nri;
 129     iinr             = nlist->iinr;
 130     jindex           = nlist->jindex;
 131     jjnr             = nlist->jjnr;
 132     shiftidx         = nlist->shift;
 133     gid              = nlist->gid;
 134     shiftvec         = fr->shift_vec[0];
 135     fshift           = fr->fshift[0];
 136     facel            = _mm_set1_pd(fr->epsfac);
 137     charge           = mdatoms->chargeA;
 138     nvdwtype         = fr->ntype;
 139     vdwparam         = fr->nbfp;
 140     vdwtype          = mdatoms->typeA;
 141
 142     vftab            = kernel_data->table_vdw->data;
 143     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 144
 145     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 146     ewtab            = fr->ic->tabq_coul_FDV0;
 147     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 148     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 149
 150     /* Setup water-specific parameters */
 151     inr              = nlist->iinr[0];
 152     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 153     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 154     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 155     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 156
 157     jq1              = _mm_set1_pd(charge[inr+1]);
 158     jq2              = _mm_set1_pd(charge[inr+2]);
 159     jq3              = _mm_set1_pd(charge[inr+3]);
 160     vdwjidx0A        = 2*vdwtype[inr+0];
 161     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 162     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 163     qq11             = _mm_mul_pd(iq1,jq1);
 164     qq12             = _mm_mul_pd(iq1,jq2);
 165     qq13             = _mm_mul_pd(iq1,jq3);
 166     qq21             = _mm_mul_pd(iq2,jq1);
 167     qq22             = _mm_mul_pd(iq2,jq2);
 168     qq23             = _mm_mul_pd(iq2,jq3);
 169     qq31             = _mm_mul_pd(iq3,jq1);
 170     qq32             = _mm_mul_pd(iq3,jq2);
 171     qq33             = _mm_mul_pd(iq3,jq3);
 172
 173     /* Avoid stupid compiler warnings */
 174     jnrA = jnrB = 0;
 175     j_coord_offsetA = 0;
 176     j_coord_offsetB = 0;
 177
 178     outeriter        = 0;
 179     inneriter        = 0;
 180
 181     /* Start outer loop over neighborlists */
 182     for(iidx=0; iidx<nri; iidx++)
 183     {
 184         /* Load shift vector for this list */
 185         i_shift_offset   = DIM*shiftidx[iidx];
 186
 187         /* Load limits for loop over neighbors */
 188         j_index_start    = jindex[iidx];
 189         j_index_end      = jindex[iidx+1];
 190
 191         /* Get outer coordinate index */
 192         inr              = iinr[iidx];
 193         i_coord_offset   = DIM*inr;
 194
 195         /* Load i particle coords and add shift vector */
 196         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 197                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 198
 199         fix0             = _mm_setzero_pd();
 200         fiy0             = _mm_setzero_pd();
 201         fiz0             = _mm_setzero_pd();
 202         fix1             = _mm_setzero_pd();
 203         fiy1             = _mm_setzero_pd();
 204         fiz1             = _mm_setzero_pd();
 205         fix2             = _mm_setzero_pd();
 206         fiy2             = _mm_setzero_pd();
 207         fiz2             = _mm_setzero_pd();
 208         fix3             = _mm_setzero_pd();
 209         fiy3             = _mm_setzero_pd();
 210         fiz3             = _mm_setzero_pd();
 211
 212         /* Reset potential sums */
 213         velecsum         = _mm_setzero_pd();
 214         vvdwsum          = _mm_setzero_pd();
 215
 216         /* Start inner kernel loop */
 217         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 218         {
 219
 220             /* Get j neighbor index, and coordinate index */
 221             jnrA             = jjnr[jidx];
 222             jnrB             = jjnr[jidx+1];
 223             j_coord_offsetA  = DIM*jnrA;
 224             j_coord_offsetB  = DIM*jnrB;
 225
 226             /* load j atom coordinates */
 227             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 228                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 229                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 230
 231             /* Calculate displacement vector */
 232             dx00             = _mm_sub_pd(ix0,jx0);
 233             dy00             = _mm_sub_pd(iy0,jy0);
 234             dz00             = _mm_sub_pd(iz0,jz0);
 235             dx11             = _mm_sub_pd(ix1,jx1);
 236             dy11             = _mm_sub_pd(iy1,jy1);
 237             dz11             = _mm_sub_pd(iz1,jz1);
 238             dx12             = _mm_sub_pd(ix1,jx2);
 239             dy12             = _mm_sub_pd(iy1,jy2);
 240             dz12             = _mm_sub_pd(iz1,jz2);
 241             dx13             = _mm_sub_pd(ix1,jx3);
 242             dy13             = _mm_sub_pd(iy1,jy3);
 243             dz13             = _mm_sub_pd(iz1,jz3);
 244             dx21             = _mm_sub_pd(ix2,jx1);
 245             dy21             = _mm_sub_pd(iy2,jy1);
 246             dz21             = _mm_sub_pd(iz2,jz1);
 247             dx22             = _mm_sub_pd(ix2,jx2);
 248             dy22             = _mm_sub_pd(iy2,jy2);
 249             dz22             = _mm_sub_pd(iz2,jz2);
 250             dx23             = _mm_sub_pd(ix2,jx3);
 251             dy23             = _mm_sub_pd(iy2,jy3);
 252             dz23             = _mm_sub_pd(iz2,jz3);
 253             dx31             = _mm_sub_pd(ix3,jx1);
 254             dy31             = _mm_sub_pd(iy3,jy1);
 255             dz31             = _mm_sub_pd(iz3,jz1);
 256             dx32             = _mm_sub_pd(ix3,jx2);
 257             dy32             = _mm_sub_pd(iy3,jy2);
 258             dz32             = _mm_sub_pd(iz3,jz2);
 259             dx33             = _mm_sub_pd(ix3,jx3);
 260             dy33             = _mm_sub_pd(iy3,jy3);
 261             dz33             = _mm_sub_pd(iz3,jz3);
 262
 263             /* Calculate squared distance and things based on it */
 264             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 265             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 266             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 267             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 268             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 269             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 270             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 271             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 272             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 273             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 274
 275             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 276             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 277             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 278             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 279             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 280             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 281             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 282             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 283             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 284             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 285
 286             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 287             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 288             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 289             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 290             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 291             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 292             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 293             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 294             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 295
 296             fjx0             = _mm_setzero_pd();
 297             fjy0             = _mm_setzero_pd();
 298             fjz0             = _mm_setzero_pd();
 299             fjx1             = _mm_setzero_pd();
 300             fjy1             = _mm_setzero_pd();
 301             fjz1             = _mm_setzero_pd();
 302             fjx2             = _mm_setzero_pd();
 303             fjy2             = _mm_setzero_pd();
 304             fjz2             = _mm_setzero_pd();
 305             fjx3             = _mm_setzero_pd();
 306             fjy3             = _mm_setzero_pd();
 307             fjz3             = _mm_setzero_pd();
 308
 309             /**************************
 310              * CALCULATE INTERACTIONS *
 311              **************************/
 312
 313             r00              = _mm_mul_pd(rsq00,rinv00);
 314
 315             /* Calculate table index by multiplying r with table scale and truncate to integer */
 316             rt               = _mm_mul_pd(r00,vftabscale);
 317             vfitab           = _mm_cvttpd_epi32(rt);
 318             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 319             vfitab           = _mm_slli_epi32(vfitab,3);
 320
 321             /* CUBIC SPLINE TABLE DISPERSION */
 322             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 323             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 324             GMX_MM_TRANSPOSE2_PD(Y,F);
 325             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 326             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 327             GMX_MM_TRANSPOSE2_PD(G,H);
 328             Heps             = _mm_mul_pd(vfeps,H);
 329             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 330             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 331             vvdw6            = _mm_mul_pd(c6_00,VV);
 332             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 333             fvdw6            = _mm_mul_pd(c6_00,FF);
 334
 335             /* CUBIC SPLINE TABLE REPULSION */
 336             vfitab           = _mm_add_epi32(vfitab,ifour);
 337             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 338             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 339             GMX_MM_TRANSPOSE2_PD(Y,F);
 340             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 341             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 342             GMX_MM_TRANSPOSE2_PD(G,H);
 343             Heps             = _mm_mul_pd(vfeps,H);
 344             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 345             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 346             vvdw12           = _mm_mul_pd(c12_00,VV);
 347             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 348             fvdw12           = _mm_mul_pd(c12_00,FF);
 349             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 350             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 351
 352             /* Update potential sum for this i atom from the interaction with this j atom. */
 353             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 354
 355             fscal            = fvdw;
 356
 357             /* Calculate temporary vectorial force */
 358             tx               = _mm_mul_pd(fscal,dx00);
 359             ty               = _mm_mul_pd(fscal,dy00);
 360             tz               = _mm_mul_pd(fscal,dz00);
 361
 362             /* Update vectorial force */
 363             fix0             = _mm_add_pd(fix0,tx);
 364             fiy0             = _mm_add_pd(fiy0,ty);
 365             fiz0             = _mm_add_pd(fiz0,tz);
 366
 367             fjx0             = _mm_add_pd(fjx0,tx);
 368             fjy0             = _mm_add_pd(fjy0,ty);
 369             fjz0             = _mm_add_pd(fjz0,tz);
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             r11              = _mm_mul_pd(rsq11,rinv11);
 376
 377             /* EWALD ELECTROSTATICS */
 378
 379             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 380             ewrt             = _mm_mul_pd(r11,ewtabscale);
 381             ewitab           = _mm_cvttpd_epi32(ewrt);
 382             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 383             ewitab           = _mm_slli_epi32(ewitab,2);
 384             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 385             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 386             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 387             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 388             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 389             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 390             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 391             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 392             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 393             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 394
 395             /* Update potential sum for this i atom from the interaction with this j atom. */
 396             velecsum         = _mm_add_pd(velecsum,velec);
 397
 398             fscal            = felec;
 399
 400             /* Calculate temporary vectorial force */
 401             tx               = _mm_mul_pd(fscal,dx11);
 402             ty               = _mm_mul_pd(fscal,dy11);
 403             tz               = _mm_mul_pd(fscal,dz11);
 404
 405             /* Update vectorial force */
 406             fix1             = _mm_add_pd(fix1,tx);
 407             fiy1             = _mm_add_pd(fiy1,ty);
 408             fiz1             = _mm_add_pd(fiz1,tz);
 409
 410             fjx1             = _mm_add_pd(fjx1,tx);
 411             fjy1             = _mm_add_pd(fjy1,ty);
 412             fjz1             = _mm_add_pd(fjz1,tz);
 413
 414             /**************************
 415              * CALCULATE INTERACTIONS *
 416              **************************/
 417
 418             r12              = _mm_mul_pd(rsq12,rinv12);
 419
 420             /* EWALD ELECTROSTATICS */
 421
 422             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 423             ewrt             = _mm_mul_pd(r12,ewtabscale);
 424             ewitab           = _mm_cvttpd_epi32(ewrt);
 425             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 426             ewitab           = _mm_slli_epi32(ewitab,2);
 427             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 428             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 429             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 430             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 431             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 432             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 433             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 434             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 435             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 436             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 437
 438             /* Update potential sum for this i atom from the interaction with this j atom. */
 439             velecsum         = _mm_add_pd(velecsum,velec);
 440
 441             fscal            = felec;
 442
 443             /* Calculate temporary vectorial force */
 444             tx               = _mm_mul_pd(fscal,dx12);
 445             ty               = _mm_mul_pd(fscal,dy12);
 446             tz               = _mm_mul_pd(fscal,dz12);
 447
 448             /* Update vectorial force */
 449             fix1             = _mm_add_pd(fix1,tx);
 450             fiy1             = _mm_add_pd(fiy1,ty);
 451             fiz1             = _mm_add_pd(fiz1,tz);
 452
 453             fjx2             = _mm_add_pd(fjx2,tx);
 454             fjy2             = _mm_add_pd(fjy2,ty);
 455             fjz2             = _mm_add_pd(fjz2,tz);
 456
 457             /**************************
 458              * CALCULATE INTERACTIONS *
 459              **************************/
 460
 461             r13              = _mm_mul_pd(rsq13,rinv13);
 462
 463             /* EWALD ELECTROSTATICS */
 464
 465             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 466             ewrt             = _mm_mul_pd(r13,ewtabscale);
 467             ewitab           = _mm_cvttpd_epi32(ewrt);
 468             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 469             ewitab           = _mm_slli_epi32(ewitab,2);
 470             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 471             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 472             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 473             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 474             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 475             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 476             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 477             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 478             velec            = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
 479             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
 480
 481             /* Update potential sum for this i atom from the interaction with this j atom. */
 482             velecsum         = _mm_add_pd(velecsum,velec);
 483
 484             fscal            = felec;
 485
 486             /* Calculate temporary vectorial force */
 487             tx               = _mm_mul_pd(fscal,dx13);
 488             ty               = _mm_mul_pd(fscal,dy13);
 489             tz               = _mm_mul_pd(fscal,dz13);
 490
 491             /* Update vectorial force */
 492             fix1             = _mm_add_pd(fix1,tx);
 493             fiy1             = _mm_add_pd(fiy1,ty);
 494             fiz1             = _mm_add_pd(fiz1,tz);
 495
 496             fjx3             = _mm_add_pd(fjx3,tx);
 497             fjy3             = _mm_add_pd(fjy3,ty);
 498             fjz3             = _mm_add_pd(fjz3,tz);
 499
 500             /**************************
 501              * CALCULATE INTERACTIONS *
 502              **************************/
 503
 504             r21              = _mm_mul_pd(rsq21,rinv21);
 505
 506             /* EWALD ELECTROSTATICS */
 507
 508             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 509             ewrt             = _mm_mul_pd(r21,ewtabscale);
 510             ewitab           = _mm_cvttpd_epi32(ewrt);
 511             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 512             ewitab           = _mm_slli_epi32(ewitab,2);
 513             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 514             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 515             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 516             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 517             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 518             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 519             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 520             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 521             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 522             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 523
 524             /* Update potential sum for this i atom from the interaction with this j atom. */
 525             velecsum         = _mm_add_pd(velecsum,velec);
 526
 527             fscal            = felec;
 528
 529             /* Calculate temporary vectorial force */
 530             tx               = _mm_mul_pd(fscal,dx21);
 531             ty               = _mm_mul_pd(fscal,dy21);
 532             tz               = _mm_mul_pd(fscal,dz21);
 533
 534             /* Update vectorial force */
 535             fix2             = _mm_add_pd(fix2,tx);
 536             fiy2             = _mm_add_pd(fiy2,ty);
 537             fiz2             = _mm_add_pd(fiz2,tz);
 538
 539             fjx1             = _mm_add_pd(fjx1,tx);
 540             fjy1             = _mm_add_pd(fjy1,ty);
 541             fjz1             = _mm_add_pd(fjz1,tz);
 542
 543             /**************************
 544              * CALCULATE INTERACTIONS *
 545              **************************/
 546
 547             r22              = _mm_mul_pd(rsq22,rinv22);
 548
 549             /* EWALD ELECTROSTATICS */
 550
 551             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 552             ewrt             = _mm_mul_pd(r22,ewtabscale);
 553             ewitab           = _mm_cvttpd_epi32(ewrt);
 554             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 555             ewitab           = _mm_slli_epi32(ewitab,2);
 556             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 557             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 558             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 559             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 560             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 561             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 562             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 563             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 564             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 565             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 566
 567             /* Update potential sum for this i atom from the interaction with this j atom. */
 568             velecsum         = _mm_add_pd(velecsum,velec);
 569
 570             fscal            = felec;
 571
 572             /* Calculate temporary vectorial force */
 573             tx               = _mm_mul_pd(fscal,dx22);
 574             ty               = _mm_mul_pd(fscal,dy22);
 575             tz               = _mm_mul_pd(fscal,dz22);
 576
 577             /* Update vectorial force */
 578             fix2             = _mm_add_pd(fix2,tx);
 579             fiy2             = _mm_add_pd(fiy2,ty);
 580             fiz2             = _mm_add_pd(fiz2,tz);
 581
 582             fjx2             = _mm_add_pd(fjx2,tx);
 583             fjy2             = _mm_add_pd(fjy2,ty);
 584             fjz2             = _mm_add_pd(fjz2,tz);
 585
 586             /**************************
 587              * CALCULATE INTERACTIONS *
 588              **************************/
 589
 590             r23              = _mm_mul_pd(rsq23,rinv23);
 591
 592             /* EWALD ELECTROSTATICS */
 593
 594             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 595             ewrt             = _mm_mul_pd(r23,ewtabscale);
 596             ewitab           = _mm_cvttpd_epi32(ewrt);
 597             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 598             ewitab           = _mm_slli_epi32(ewitab,2);
 599             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 600             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 601             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 602             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 603             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 604             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 605             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 606             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 607             velec            = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
 608             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
 609
 610             /* Update potential sum for this i atom from the interaction with this j atom. */
 611             velecsum         = _mm_add_pd(velecsum,velec);
 612
 613             fscal            = felec;
 614
 615             /* Calculate temporary vectorial force */
 616             tx               = _mm_mul_pd(fscal,dx23);
 617             ty               = _mm_mul_pd(fscal,dy23);
 618             tz               = _mm_mul_pd(fscal,dz23);
 619
 620             /* Update vectorial force */
 621             fix2             = _mm_add_pd(fix2,tx);
 622             fiy2             = _mm_add_pd(fiy2,ty);
 623             fiz2             = _mm_add_pd(fiz2,tz);
 624
 625             fjx3             = _mm_add_pd(fjx3,tx);
 626             fjy3             = _mm_add_pd(fjy3,ty);
 627             fjz3             = _mm_add_pd(fjz3,tz);
 628
 629             /**************************
 630              * CALCULATE INTERACTIONS *
 631              **************************/
 632
 633             r31              = _mm_mul_pd(rsq31,rinv31);
 634
 635             /* EWALD ELECTROSTATICS */
 636
 637             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 638             ewrt             = _mm_mul_pd(r31,ewtabscale);
 639             ewitab           = _mm_cvttpd_epi32(ewrt);
 640             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 641             ewitab           = _mm_slli_epi32(ewitab,2);
 642             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 643             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 644             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 645             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 646             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 647             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 648             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 649             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 650             velec            = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
 651             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
 652
 653             /* Update potential sum for this i atom from the interaction with this j atom. */
 654             velecsum         = _mm_add_pd(velecsum,velec);
 655
 656             fscal            = felec;
 657
 658             /* Calculate temporary vectorial force */
 659             tx               = _mm_mul_pd(fscal,dx31);
 660             ty               = _mm_mul_pd(fscal,dy31);
 661             tz               = _mm_mul_pd(fscal,dz31);
 662
 663             /* Update vectorial force */
 664             fix3             = _mm_add_pd(fix3,tx);
 665             fiy3             = _mm_add_pd(fiy3,ty);
 666             fiz3             = _mm_add_pd(fiz3,tz);
 667
 668             fjx1             = _mm_add_pd(fjx1,tx);
 669             fjy1             = _mm_add_pd(fjy1,ty);
 670             fjz1             = _mm_add_pd(fjz1,tz);
 671
 672             /**************************
 673              * CALCULATE INTERACTIONS *
 674              **************************/
 675
 676             r32              = _mm_mul_pd(rsq32,rinv32);
 677
 678             /* EWALD ELECTROSTATICS */
 679
 680             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 681             ewrt             = _mm_mul_pd(r32,ewtabscale);
 682             ewitab           = _mm_cvttpd_epi32(ewrt);
 683             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 684             ewitab           = _mm_slli_epi32(ewitab,2);
 685             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 686             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 687             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 688             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 689             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 690             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 691             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 692             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 693             velec            = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
 694             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
 695
 696             /* Update potential sum for this i atom from the interaction with this j atom. */
 697             velecsum         = _mm_add_pd(velecsum,velec);
 698
 699             fscal            = felec;
 700
 701             /* Calculate temporary vectorial force */
 702             tx               = _mm_mul_pd(fscal,dx32);
 703             ty               = _mm_mul_pd(fscal,dy32);
 704             tz               = _mm_mul_pd(fscal,dz32);
 705
 706             /* Update vectorial force */
 707             fix3             = _mm_add_pd(fix3,tx);
 708             fiy3             = _mm_add_pd(fiy3,ty);
 709             fiz3             = _mm_add_pd(fiz3,tz);
 710
 711             fjx2             = _mm_add_pd(fjx2,tx);
 712             fjy2             = _mm_add_pd(fjy2,ty);
 713             fjz2             = _mm_add_pd(fjz2,tz);
 714
 715             /**************************
 716              * CALCULATE INTERACTIONS *
 717              **************************/
 718
 719             r33              = _mm_mul_pd(rsq33,rinv33);
 720
 721             /* EWALD ELECTROSTATICS */
 722
 723             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 724             ewrt             = _mm_mul_pd(r33,ewtabscale);
 725             ewitab           = _mm_cvttpd_epi32(ewrt);
 726             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 727             ewitab           = _mm_slli_epi32(ewitab,2);
 728             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 729             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 730             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 731             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 732             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 733             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 734             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 735             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 736             velec            = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
 737             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
 738
 739             /* Update potential sum for this i atom from the interaction with this j atom. */
 740             velecsum         = _mm_add_pd(velecsum,velec);
 741
 742             fscal            = felec;
 743
 744             /* Calculate temporary vectorial force */
 745             tx               = _mm_mul_pd(fscal,dx33);
 746             ty               = _mm_mul_pd(fscal,dy33);
 747             tz               = _mm_mul_pd(fscal,dz33);
 748
 749             /* Update vectorial force */
 750             fix3             = _mm_add_pd(fix3,tx);
 751             fiy3             = _mm_add_pd(fiy3,ty);
 752             fiz3             = _mm_add_pd(fiz3,tz);
 753
 754             fjx3             = _mm_add_pd(fjx3,tx);
 755             fjy3             = _mm_add_pd(fjy3,ty);
 756             fjz3             = _mm_add_pd(fjz3,tz);
 757
 758             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 759
 760             /* Inner loop uses 428 flops */
 761         }
 762
 763         if(jidx<j_index_end)
 764         {
 765
 766             jnrA             = jjnr[jidx];
 767             j_coord_offsetA  = DIM*jnrA;
 768
 769             /* load j atom coordinates */
 770             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 771                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 772                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 773
 774             /* Calculate displacement vector */
 775             dx00             = _mm_sub_pd(ix0,jx0);
 776             dy00             = _mm_sub_pd(iy0,jy0);
 777             dz00             = _mm_sub_pd(iz0,jz0);
 778             dx11             = _mm_sub_pd(ix1,jx1);
 779             dy11             = _mm_sub_pd(iy1,jy1);
 780             dz11             = _mm_sub_pd(iz1,jz1);
 781             dx12             = _mm_sub_pd(ix1,jx2);
 782             dy12             = _mm_sub_pd(iy1,jy2);
 783             dz12             = _mm_sub_pd(iz1,jz2);
 784             dx13             = _mm_sub_pd(ix1,jx3);
 785             dy13             = _mm_sub_pd(iy1,jy3);
 786             dz13             = _mm_sub_pd(iz1,jz3);
 787             dx21             = _mm_sub_pd(ix2,jx1);
 788             dy21             = _mm_sub_pd(iy2,jy1);
 789             dz21             = _mm_sub_pd(iz2,jz1);
 790             dx22             = _mm_sub_pd(ix2,jx2);
 791             dy22             = _mm_sub_pd(iy2,jy2);
 792             dz22             = _mm_sub_pd(iz2,jz2);
 793             dx23             = _mm_sub_pd(ix2,jx3);
 794             dy23             = _mm_sub_pd(iy2,jy3);
 795             dz23             = _mm_sub_pd(iz2,jz3);
 796             dx31             = _mm_sub_pd(ix3,jx1);
 797             dy31             = _mm_sub_pd(iy3,jy1);
 798             dz31             = _mm_sub_pd(iz3,jz1);
 799             dx32             = _mm_sub_pd(ix3,jx2);
 800             dy32             = _mm_sub_pd(iy3,jy2);
 801             dz32             = _mm_sub_pd(iz3,jz2);
 802             dx33             = _mm_sub_pd(ix3,jx3);
 803             dy33             = _mm_sub_pd(iy3,jy3);
 804             dz33             = _mm_sub_pd(iz3,jz3);
 805
 806             /* Calculate squared distance and things based on it */
 807             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 808             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 809             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 810             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 811             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 812             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 813             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 814             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 815             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 816             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 817
 818             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 819             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 820             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 821             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 822             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 823             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 824             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 825             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 826             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 827             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 828
 829             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 830             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 831             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 832             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 833             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 834             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 835             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 836             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 837             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 838
 839             fjx0             = _mm_setzero_pd();
 840             fjy0             = _mm_setzero_pd();
 841             fjz0             = _mm_setzero_pd();
 842             fjx1             = _mm_setzero_pd();
 843             fjy1             = _mm_setzero_pd();
 844             fjz1             = _mm_setzero_pd();
 845             fjx2             = _mm_setzero_pd();
 846             fjy2             = _mm_setzero_pd();
 847             fjz2             = _mm_setzero_pd();
 848             fjx3             = _mm_setzero_pd();
 849             fjy3             = _mm_setzero_pd();
 850             fjz3             = _mm_setzero_pd();
 851
 852             /**************************
 853              * CALCULATE INTERACTIONS *
 854              **************************/
 855
 856             r00              = _mm_mul_pd(rsq00,rinv00);
 857
 858             /* Calculate table index by multiplying r with table scale and truncate to integer */
 859             rt               = _mm_mul_pd(r00,vftabscale);
 860             vfitab           = _mm_cvttpd_epi32(rt);
 861             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 862             vfitab           = _mm_slli_epi32(vfitab,3);
 863
 864             /* CUBIC SPLINE TABLE DISPERSION */
 865             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 866             F                = _mm_setzero_pd();
 867             GMX_MM_TRANSPOSE2_PD(Y,F);
 868             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 869             H                = _mm_setzero_pd();
 870             GMX_MM_TRANSPOSE2_PD(G,H);
 871             Heps             = _mm_mul_pd(vfeps,H);
 872             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 873             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 874             vvdw6            = _mm_mul_pd(c6_00,VV);
 875             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 876             fvdw6            = _mm_mul_pd(c6_00,FF);
 877
 878             /* CUBIC SPLINE TABLE REPULSION */
 879             vfitab           = _mm_add_epi32(vfitab,ifour);
 880             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 881             F                = _mm_setzero_pd();
 882             GMX_MM_TRANSPOSE2_PD(Y,F);
 883             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 884             H                = _mm_setzero_pd();
 885             GMX_MM_TRANSPOSE2_PD(G,H);
 886             Heps             = _mm_mul_pd(vfeps,H);
 887             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 888             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 889             vvdw12           = _mm_mul_pd(c12_00,VV);
 890             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 891             fvdw12           = _mm_mul_pd(c12_00,FF);
 892             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 893             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 894
 895             /* Update potential sum for this i atom from the interaction with this j atom. */
 896             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 897             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 898
 899             fscal            = fvdw;
 900
 901             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 902
 903             /* Calculate temporary vectorial force */
 904             tx               = _mm_mul_pd(fscal,dx00);
 905             ty               = _mm_mul_pd(fscal,dy00);
 906             tz               = _mm_mul_pd(fscal,dz00);
 907
 908             /* Update vectorial force */
 909             fix0             = _mm_add_pd(fix0,tx);
 910             fiy0             = _mm_add_pd(fiy0,ty);
 911             fiz0             = _mm_add_pd(fiz0,tz);
 912
 913             fjx0             = _mm_add_pd(fjx0,tx);
 914             fjy0             = _mm_add_pd(fjy0,ty);
 915             fjz0             = _mm_add_pd(fjz0,tz);
 916
 917             /**************************
 918              * CALCULATE INTERACTIONS *
 919              **************************/
 920
 921             r11              = _mm_mul_pd(rsq11,rinv11);
 922
 923             /* EWALD ELECTROSTATICS */
 924
 925             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 926             ewrt             = _mm_mul_pd(r11,ewtabscale);
 927             ewitab           = _mm_cvttpd_epi32(ewrt);
 928             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 929             ewitab           = _mm_slli_epi32(ewitab,2);
 930             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 931             ewtabD           = _mm_setzero_pd();
 932             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 933             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 934             ewtabFn          = _mm_setzero_pd();
 935             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 936             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 937             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 938             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 939             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 940
 941             /* Update potential sum for this i atom from the interaction with this j atom. */
 942             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 943             velecsum         = _mm_add_pd(velecsum,velec);
 944
 945             fscal            = felec;
 946
 947             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 948
 949             /* Calculate temporary vectorial force */
 950             tx               = _mm_mul_pd(fscal,dx11);
 951             ty               = _mm_mul_pd(fscal,dy11);
 952             tz               = _mm_mul_pd(fscal,dz11);
 953
 954             /* Update vectorial force */
 955             fix1             = _mm_add_pd(fix1,tx);
 956             fiy1             = _mm_add_pd(fiy1,ty);
 957             fiz1             = _mm_add_pd(fiz1,tz);
 958
 959             fjx1             = _mm_add_pd(fjx1,tx);
 960             fjy1             = _mm_add_pd(fjy1,ty);
 961             fjz1             = _mm_add_pd(fjz1,tz);
 962
 963             /**************************
 964              * CALCULATE INTERACTIONS *
 965              **************************/
 966
 967             r12              = _mm_mul_pd(rsq12,rinv12);
 968
 969             /* EWALD ELECTROSTATICS */
 970
 971             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 972             ewrt             = _mm_mul_pd(r12,ewtabscale);
 973             ewitab           = _mm_cvttpd_epi32(ewrt);
 974             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 975             ewitab           = _mm_slli_epi32(ewitab,2);
 976             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 977             ewtabD           = _mm_setzero_pd();
 978             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 979             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 980             ewtabFn          = _mm_setzero_pd();
 981             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 982             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 983             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 984             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 985             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 986
 987             /* Update potential sum for this i atom from the interaction with this j atom. */
 988             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 989             velecsum         = _mm_add_pd(velecsum,velec);
 990
 991             fscal            = felec;
 992
 993             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 994
 995             /* Calculate temporary vectorial force */
 996             tx               = _mm_mul_pd(fscal,dx12);
 997             ty               = _mm_mul_pd(fscal,dy12);
 998             tz               = _mm_mul_pd(fscal,dz12);
 999
1000             /* Update vectorial force */
1001             fix1             = _mm_add_pd(fix1,tx);
1002             fiy1             = _mm_add_pd(fiy1,ty);
1003             fiz1             = _mm_add_pd(fiz1,tz);
1004
1005             fjx2             = _mm_add_pd(fjx2,tx);
1006             fjy2             = _mm_add_pd(fjy2,ty);
1007             fjz2             = _mm_add_pd(fjz2,tz);
1008
1009             /**************************
1010              * CALCULATE INTERACTIONS *
1011              **************************/
1012
1013             r13              = _mm_mul_pd(rsq13,rinv13);
1014
1015             /* EWALD ELECTROSTATICS */
1016
1017             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1018             ewrt             = _mm_mul_pd(r13,ewtabscale);
1019             ewitab           = _mm_cvttpd_epi32(ewrt);
1020             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1021             ewitab           = _mm_slli_epi32(ewitab,2);
1022             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1023             ewtabD           = _mm_setzero_pd();
1024             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1025             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1026             ewtabFn          = _mm_setzero_pd();
1027             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1028             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1029             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1030             velec            = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
1031             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1032
1033             /* Update potential sum for this i atom from the interaction with this j atom. */
1034             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1035             velecsum         = _mm_add_pd(velecsum,velec);
1036
1037             fscal            = felec;
1038
1039             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1040
1041             /* Calculate temporary vectorial force */
1042             tx               = _mm_mul_pd(fscal,dx13);
1043             ty               = _mm_mul_pd(fscal,dy13);
1044             tz               = _mm_mul_pd(fscal,dz13);
1045
1046             /* Update vectorial force */
1047             fix1             = _mm_add_pd(fix1,tx);
1048             fiy1             = _mm_add_pd(fiy1,ty);
1049             fiz1             = _mm_add_pd(fiz1,tz);
1050
1051             fjx3             = _mm_add_pd(fjx3,tx);
1052             fjy3             = _mm_add_pd(fjy3,ty);
1053             fjz3             = _mm_add_pd(fjz3,tz);
1054
1055             /**************************
1056              * CALCULATE INTERACTIONS *
1057              **************************/
1058
1059             r21              = _mm_mul_pd(rsq21,rinv21);
1060
1061             /* EWALD ELECTROSTATICS */
1062
1063             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1064             ewrt             = _mm_mul_pd(r21,ewtabscale);
1065             ewitab           = _mm_cvttpd_epi32(ewrt);
1066             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1067             ewitab           = _mm_slli_epi32(ewitab,2);
1068             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1069             ewtabD           = _mm_setzero_pd();
1070             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1071             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1072             ewtabFn          = _mm_setzero_pd();
1073             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1074             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1075             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1076             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1077             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1078
1079             /* Update potential sum for this i atom from the interaction with this j atom. */
1080             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1081             velecsum         = _mm_add_pd(velecsum,velec);
1082
1083             fscal            = felec;
1084
1085             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1086
1087             /* Calculate temporary vectorial force */
1088             tx               = _mm_mul_pd(fscal,dx21);
1089             ty               = _mm_mul_pd(fscal,dy21);
1090             tz               = _mm_mul_pd(fscal,dz21);
1091
1092             /* Update vectorial force */
1093             fix2             = _mm_add_pd(fix2,tx);
1094             fiy2             = _mm_add_pd(fiy2,ty);
1095             fiz2             = _mm_add_pd(fiz2,tz);
1096
1097             fjx1             = _mm_add_pd(fjx1,tx);
1098             fjy1             = _mm_add_pd(fjy1,ty);
1099             fjz1             = _mm_add_pd(fjz1,tz);
1100
1101             /**************************
1102              * CALCULATE INTERACTIONS *
1103              **************************/
1104
1105             r22              = _mm_mul_pd(rsq22,rinv22);
1106
1107             /* EWALD ELECTROSTATICS */
1108
1109             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1110             ewrt             = _mm_mul_pd(r22,ewtabscale);
1111             ewitab           = _mm_cvttpd_epi32(ewrt);
1112             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1113             ewitab           = _mm_slli_epi32(ewitab,2);
1114             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1115             ewtabD           = _mm_setzero_pd();
1116             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1117             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1118             ewtabFn          = _mm_setzero_pd();
1119             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1120             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1121             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1122             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1123             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1124
1125             /* Update potential sum for this i atom from the interaction with this j atom. */
1126             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1127             velecsum         = _mm_add_pd(velecsum,velec);
1128
1129             fscal            = felec;
1130
1131             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1132
1133             /* Calculate temporary vectorial force */
1134             tx               = _mm_mul_pd(fscal,dx22);
1135             ty               = _mm_mul_pd(fscal,dy22);
1136             tz               = _mm_mul_pd(fscal,dz22);
1137
1138             /* Update vectorial force */
1139             fix2             = _mm_add_pd(fix2,tx);
1140             fiy2             = _mm_add_pd(fiy2,ty);
1141             fiz2             = _mm_add_pd(fiz2,tz);
1142
1143             fjx2             = _mm_add_pd(fjx2,tx);
1144             fjy2             = _mm_add_pd(fjy2,ty);
1145             fjz2             = _mm_add_pd(fjz2,tz);
1146
1147             /**************************
1148              * CALCULATE INTERACTIONS *
1149              **************************/
1150
1151             r23              = _mm_mul_pd(rsq23,rinv23);
1152
1153             /* EWALD ELECTROSTATICS */
1154
1155             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1156             ewrt             = _mm_mul_pd(r23,ewtabscale);
1157             ewitab           = _mm_cvttpd_epi32(ewrt);
1158             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1159             ewitab           = _mm_slli_epi32(ewitab,2);
1160             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1161             ewtabD           = _mm_setzero_pd();
1162             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1163             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1164             ewtabFn          = _mm_setzero_pd();
1165             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1166             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1167             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1168             velec            = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
1169             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1170
1171             /* Update potential sum for this i atom from the interaction with this j atom. */
1172             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1173             velecsum         = _mm_add_pd(velecsum,velec);
1174
1175             fscal            = felec;
1176
1177             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1178
1179             /* Calculate temporary vectorial force */
1180             tx               = _mm_mul_pd(fscal,dx23);
1181             ty               = _mm_mul_pd(fscal,dy23);
1182             tz               = _mm_mul_pd(fscal,dz23);
1183
1184             /* Update vectorial force */
1185             fix2             = _mm_add_pd(fix2,tx);
1186             fiy2             = _mm_add_pd(fiy2,ty);
1187             fiz2             = _mm_add_pd(fiz2,tz);
1188
1189             fjx3             = _mm_add_pd(fjx3,tx);
1190             fjy3             = _mm_add_pd(fjy3,ty);
1191             fjz3             = _mm_add_pd(fjz3,tz);
1192
1193             /**************************
1194              * CALCULATE INTERACTIONS *
1195              **************************/
1196
1197             r31              = _mm_mul_pd(rsq31,rinv31);
1198
1199             /* EWALD ELECTROSTATICS */
1200
1201             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1202             ewrt             = _mm_mul_pd(r31,ewtabscale);
1203             ewitab           = _mm_cvttpd_epi32(ewrt);
1204             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1205             ewitab           = _mm_slli_epi32(ewitab,2);
1206             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1207             ewtabD           = _mm_setzero_pd();
1208             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1209             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1210             ewtabFn          = _mm_setzero_pd();
1211             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1212             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1213             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1214             velec            = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
1215             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1216
1217             /* Update potential sum for this i atom from the interaction with this j atom. */
1218             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1219             velecsum         = _mm_add_pd(velecsum,velec);
1220
1221             fscal            = felec;
1222
1223             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1224
1225             /* Calculate temporary vectorial force */
1226             tx               = _mm_mul_pd(fscal,dx31);
1227             ty               = _mm_mul_pd(fscal,dy31);
1228             tz               = _mm_mul_pd(fscal,dz31);
1229
1230             /* Update vectorial force */
1231             fix3             = _mm_add_pd(fix3,tx);
1232             fiy3             = _mm_add_pd(fiy3,ty);
1233             fiz3             = _mm_add_pd(fiz3,tz);
1234
1235             fjx1             = _mm_add_pd(fjx1,tx);
1236             fjy1             = _mm_add_pd(fjy1,ty);
1237             fjz1             = _mm_add_pd(fjz1,tz);
1238
1239             /**************************
1240              * CALCULATE INTERACTIONS *
1241              **************************/
1242
1243             r32              = _mm_mul_pd(rsq32,rinv32);
1244
1245             /* EWALD ELECTROSTATICS */
1246
1247             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1248             ewrt             = _mm_mul_pd(r32,ewtabscale);
1249             ewitab           = _mm_cvttpd_epi32(ewrt);
1250             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1251             ewitab           = _mm_slli_epi32(ewitab,2);
1252             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1253             ewtabD           = _mm_setzero_pd();
1254             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1255             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1256             ewtabFn          = _mm_setzero_pd();
1257             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1258             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1259             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1260             velec            = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
1261             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1262
1263             /* Update potential sum for this i atom from the interaction with this j atom. */
1264             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1265             velecsum         = _mm_add_pd(velecsum,velec);
1266
1267             fscal            = felec;
1268
1269             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1270
1271             /* Calculate temporary vectorial force */
1272             tx               = _mm_mul_pd(fscal,dx32);
1273             ty               = _mm_mul_pd(fscal,dy32);
1274             tz               = _mm_mul_pd(fscal,dz32);
1275
1276             /* Update vectorial force */
1277             fix3             = _mm_add_pd(fix3,tx);
1278             fiy3             = _mm_add_pd(fiy3,ty);
1279             fiz3             = _mm_add_pd(fiz3,tz);
1280
1281             fjx2             = _mm_add_pd(fjx2,tx);
1282             fjy2             = _mm_add_pd(fjy2,ty);
1283             fjz2             = _mm_add_pd(fjz2,tz);
1284
1285             /**************************
1286              * CALCULATE INTERACTIONS *
1287              **************************/
1288
1289             r33              = _mm_mul_pd(rsq33,rinv33);
1290
1291             /* EWALD ELECTROSTATICS */
1292
1293             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1294             ewrt             = _mm_mul_pd(r33,ewtabscale);
1295             ewitab           = _mm_cvttpd_epi32(ewrt);
1296             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1297             ewitab           = _mm_slli_epi32(ewitab,2);
1298             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1299             ewtabD           = _mm_setzero_pd();
1300             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1301             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1302             ewtabFn          = _mm_setzero_pd();
1303             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1304             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1305             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1306             velec            = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
1307             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1308
1309             /* Update potential sum for this i atom from the interaction with this j atom. */
1310             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1311             velecsum         = _mm_add_pd(velecsum,velec);
1312
1313             fscal            = felec;
1314
1315             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1316
1317             /* Calculate temporary vectorial force */
1318             tx               = _mm_mul_pd(fscal,dx33);
1319             ty               = _mm_mul_pd(fscal,dy33);
1320             tz               = _mm_mul_pd(fscal,dz33);
1321
1322             /* Update vectorial force */
1323             fix3             = _mm_add_pd(fix3,tx);
1324             fiy3             = _mm_add_pd(fiy3,ty);
1325             fiz3             = _mm_add_pd(fiz3,tz);
1326
1327             fjx3             = _mm_add_pd(fjx3,tx);
1328             fjy3             = _mm_add_pd(fjy3,ty);
1329             fjz3             = _mm_add_pd(fjz3,tz);
1330
1331             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1332
1333             /* Inner loop uses 428 flops */
1334         }
1335
1336         /* End of innermost loop */
1337
1338         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1339                                               f+i_coord_offset,fshift+i_shift_offset);
1340
1341         ggid                        = gid[iidx];
1342         /* Update potential energies */
1343         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1344         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1345
1346         /* Increment number of inner iterations */
1347         inneriter                  += j_index_end - j_index_start;
1348
1349         /* Outer loop uses 26 flops */
1350     }
1351
1352     /* Increment number of outer iterations */
1353     outeriter        += nri;
1354
1355     /* Update outer/inner flops */
1356
1357     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*428);
1358 }
1359 /*
1360  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse2_double
1361  * Electrostatics interaction: Ewald
1362  * VdW interaction:            CubicSplineTable
1363  * Geometry:                   Water4-Water4
1364  * Calculate force/pot:        Force
1365  */
1366 void
1367 nb_kernel_ElecEw_VdwCSTab_GeomW4W4_F_sse2_double
1368                     (t_nblist                    * gmx_restrict       nlist,
1369                      rvec                        * gmx_restrict          xx,
1370                      rvec                        * gmx_restrict          ff,
1371                      t_forcerec                  * gmx_restrict          fr,
1372                      t_mdatoms                   * gmx_restrict     mdatoms,
1373                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1374                      t_nrnb                      * gmx_restrict        nrnb)
1375 {
1376     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1377      * just 0 for non-waters.
1378      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1379      * jnr indices corresponding to data put in the four positions in the SIMD register.
1380      */
1381     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1382     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1383     int              jnrA,jnrB;
1384     int              j_coord_offsetA,j_coord_offsetB;
1385     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1386     real             rcutoff_scalar;
1387     real             *shiftvec,*fshift,*x,*f;
1388     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1389     int              vdwioffset0;
1390     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1391     int              vdwioffset1;
1392     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1393     int              vdwioffset2;
1394     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1395     int              vdwioffset3;
1396     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1397     int              vdwjidx0A,vdwjidx0B;
1398     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1399     int              vdwjidx1A,vdwjidx1B;
1400     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1401     int              vdwjidx2A,vdwjidx2B;
1402     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1403     int              vdwjidx3A,vdwjidx3B;
1404     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1405     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1406     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1407     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1408     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1409     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1410     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1411     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1412     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1413     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1414     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1415     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1416     real             *charge;
1417     int              nvdwtype;
1418     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1419     int              *vdwtype;
1420     real             *vdwparam;
1421     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1422     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1423     __m128i          vfitab;
1424     __m128i          ifour       = _mm_set1_epi32(4);
1425     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1426     real             *vftab;
1427     __m128i          ewitab;
1428     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1429     real             *ewtab;
1430     __m128d          dummy_mask,cutoff_mask;
1431     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1432     __m128d          one     = _mm_set1_pd(1.0);
1433     __m128d          two     = _mm_set1_pd(2.0);
1434     x                = xx[0];
1435     f                = ff[0];
1436
1437     nri              = nlist->nri;
1438     iinr             = nlist->iinr;
1439     jindex           = nlist->jindex;
1440     jjnr             = nlist->jjnr;
1441     shiftidx         = nlist->shift;
1442     gid              = nlist->gid;
1443     shiftvec         = fr->shift_vec[0];
1444     fshift           = fr->fshift[0];
1445     facel            = _mm_set1_pd(fr->epsfac);
1446     charge           = mdatoms->chargeA;
1447     nvdwtype         = fr->ntype;
1448     vdwparam         = fr->nbfp;
1449     vdwtype          = mdatoms->typeA;
1450
1451     vftab            = kernel_data->table_vdw->data;
1452     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
1453
1454     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1455     ewtab            = fr->ic->tabq_coul_F;
1456     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1457     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1458
1459     /* Setup water-specific parameters */
1460     inr              = nlist->iinr[0];
1461     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1462     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1463     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1464     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1465
1466     jq1              = _mm_set1_pd(charge[inr+1]);
1467     jq2              = _mm_set1_pd(charge[inr+2]);
1468     jq3              = _mm_set1_pd(charge[inr+3]);
1469     vdwjidx0A        = 2*vdwtype[inr+0];
1470     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1471     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1472     qq11             = _mm_mul_pd(iq1,jq1);
1473     qq12             = _mm_mul_pd(iq1,jq2);
1474     qq13             = _mm_mul_pd(iq1,jq3);
1475     qq21             = _mm_mul_pd(iq2,jq1);
1476     qq22             = _mm_mul_pd(iq2,jq2);
1477     qq23             = _mm_mul_pd(iq2,jq3);
1478     qq31             = _mm_mul_pd(iq3,jq1);
1479     qq32             = _mm_mul_pd(iq3,jq2);
1480     qq33             = _mm_mul_pd(iq3,jq3);
1481
1482     /* Avoid stupid compiler warnings */
1483     jnrA = jnrB = 0;
1484     j_coord_offsetA = 0;
1485     j_coord_offsetB = 0;
1486
1487     outeriter        = 0;
1488     inneriter        = 0;
1489
1490     /* Start outer loop over neighborlists */
1491     for(iidx=0; iidx<nri; iidx++)
1492     {
1493         /* Load shift vector for this list */
1494         i_shift_offset   = DIM*shiftidx[iidx];
1495
1496         /* Load limits for loop over neighbors */
1497         j_index_start    = jindex[iidx];
1498         j_index_end      = jindex[iidx+1];
1499
1500         /* Get outer coordinate index */
1501         inr              = iinr[iidx];
1502         i_coord_offset   = DIM*inr;
1503
1504         /* Load i particle coords and add shift vector */
1505         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1506                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1507
1508         fix0             = _mm_setzero_pd();
1509         fiy0             = _mm_setzero_pd();
1510         fiz0             = _mm_setzero_pd();
1511         fix1             = _mm_setzero_pd();
1512         fiy1             = _mm_setzero_pd();
1513         fiz1             = _mm_setzero_pd();
1514         fix2             = _mm_setzero_pd();
1515         fiy2             = _mm_setzero_pd();
1516         fiz2             = _mm_setzero_pd();
1517         fix3             = _mm_setzero_pd();
1518         fiy3             = _mm_setzero_pd();
1519         fiz3             = _mm_setzero_pd();
1520
1521         /* Start inner kernel loop */
1522         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1523         {
1524
1525             /* Get j neighbor index, and coordinate index */
1526             jnrA             = jjnr[jidx];
1527             jnrB             = jjnr[jidx+1];
1528             j_coord_offsetA  = DIM*jnrA;
1529             j_coord_offsetB  = DIM*jnrB;
1530
1531             /* load j atom coordinates */
1532             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1533                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1534                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1535
1536             /* Calculate displacement vector */
1537             dx00             = _mm_sub_pd(ix0,jx0);
1538             dy00             = _mm_sub_pd(iy0,jy0);
1539             dz00             = _mm_sub_pd(iz0,jz0);
1540             dx11             = _mm_sub_pd(ix1,jx1);
1541             dy11             = _mm_sub_pd(iy1,jy1);
1542             dz11             = _mm_sub_pd(iz1,jz1);
1543             dx12             = _mm_sub_pd(ix1,jx2);
1544             dy12             = _mm_sub_pd(iy1,jy2);
1545             dz12             = _mm_sub_pd(iz1,jz2);
1546             dx13             = _mm_sub_pd(ix1,jx3);
1547             dy13             = _mm_sub_pd(iy1,jy3);
1548             dz13             = _mm_sub_pd(iz1,jz3);
1549             dx21             = _mm_sub_pd(ix2,jx1);
1550             dy21             = _mm_sub_pd(iy2,jy1);
1551             dz21             = _mm_sub_pd(iz2,jz1);
1552             dx22             = _mm_sub_pd(ix2,jx2);
1553             dy22             = _mm_sub_pd(iy2,jy2);
1554             dz22             = _mm_sub_pd(iz2,jz2);
1555             dx23             = _mm_sub_pd(ix2,jx3);
1556             dy23             = _mm_sub_pd(iy2,jy3);
1557             dz23             = _mm_sub_pd(iz2,jz3);
1558             dx31             = _mm_sub_pd(ix3,jx1);
1559             dy31             = _mm_sub_pd(iy3,jy1);
1560             dz31             = _mm_sub_pd(iz3,jz1);
1561             dx32             = _mm_sub_pd(ix3,jx2);
1562             dy32             = _mm_sub_pd(iy3,jy2);
1563             dz32             = _mm_sub_pd(iz3,jz2);
1564             dx33             = _mm_sub_pd(ix3,jx3);
1565             dy33             = _mm_sub_pd(iy3,jy3);
1566             dz33             = _mm_sub_pd(iz3,jz3);
1567
1568             /* Calculate squared distance and things based on it */
1569             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1570             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1571             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1572             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1573             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1574             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1575             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1576             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1577             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1578             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1579
1580             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1581             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1582             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1583             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1584             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1585             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1586             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1587             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1588             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1589             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1590
1591             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1592             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1593             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1594             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1595             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1596             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1597             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1598             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1599             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1600
1601             fjx0             = _mm_setzero_pd();
1602             fjy0             = _mm_setzero_pd();
1603             fjz0             = _mm_setzero_pd();
1604             fjx1             = _mm_setzero_pd();
1605             fjy1             = _mm_setzero_pd();
1606             fjz1             = _mm_setzero_pd();
1607             fjx2             = _mm_setzero_pd();
1608             fjy2             = _mm_setzero_pd();
1609             fjz2             = _mm_setzero_pd();
1610             fjx3             = _mm_setzero_pd();
1611             fjy3             = _mm_setzero_pd();
1612             fjz3             = _mm_setzero_pd();
1613
1614             /**************************
1615              * CALCULATE INTERACTIONS *
1616              **************************/
1617
1618             r00              = _mm_mul_pd(rsq00,rinv00);
1619
1620             /* Calculate table index by multiplying r with table scale and truncate to integer */
1621             rt               = _mm_mul_pd(r00,vftabscale);
1622             vfitab           = _mm_cvttpd_epi32(rt);
1623             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1624             vfitab           = _mm_slli_epi32(vfitab,3);
1625
1626             /* CUBIC SPLINE TABLE DISPERSION */
1627             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1628             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1629             GMX_MM_TRANSPOSE2_PD(Y,F);
1630             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1631             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1632             GMX_MM_TRANSPOSE2_PD(G,H);
1633             Heps             = _mm_mul_pd(vfeps,H);
1634             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1635             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1636             fvdw6            = _mm_mul_pd(c6_00,FF);
1637
1638             /* CUBIC SPLINE TABLE REPULSION */
1639             vfitab           = _mm_add_epi32(vfitab,ifour);
1640             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1641             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1642             GMX_MM_TRANSPOSE2_PD(Y,F);
1643             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1644             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1645             GMX_MM_TRANSPOSE2_PD(G,H);
1646             Heps             = _mm_mul_pd(vfeps,H);
1647             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1648             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1649             fvdw12           = _mm_mul_pd(c12_00,FF);
1650             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1651
1652             fscal            = fvdw;
1653
1654             /* Calculate temporary vectorial force */
1655             tx               = _mm_mul_pd(fscal,dx00);
1656             ty               = _mm_mul_pd(fscal,dy00);
1657             tz               = _mm_mul_pd(fscal,dz00);
1658
1659             /* Update vectorial force */
1660             fix0             = _mm_add_pd(fix0,tx);
1661             fiy0             = _mm_add_pd(fiy0,ty);
1662             fiz0             = _mm_add_pd(fiz0,tz);
1663
1664             fjx0             = _mm_add_pd(fjx0,tx);
1665             fjy0             = _mm_add_pd(fjy0,ty);
1666             fjz0             = _mm_add_pd(fjz0,tz);
1667
1668             /**************************
1669              * CALCULATE INTERACTIONS *
1670              **************************/
1671
1672             r11              = _mm_mul_pd(rsq11,rinv11);
1673
1674             /* EWALD ELECTROSTATICS */
1675
1676             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1677             ewrt             = _mm_mul_pd(r11,ewtabscale);
1678             ewitab           = _mm_cvttpd_epi32(ewrt);
1679             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1680             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1681                                          &ewtabF,&ewtabFn);
1682             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1683             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1684
1685             fscal            = felec;
1686
1687             /* Calculate temporary vectorial force */
1688             tx               = _mm_mul_pd(fscal,dx11);
1689             ty               = _mm_mul_pd(fscal,dy11);
1690             tz               = _mm_mul_pd(fscal,dz11);
1691
1692             /* Update vectorial force */
1693             fix1             = _mm_add_pd(fix1,tx);
1694             fiy1             = _mm_add_pd(fiy1,ty);
1695             fiz1             = _mm_add_pd(fiz1,tz);
1696
1697             fjx1             = _mm_add_pd(fjx1,tx);
1698             fjy1             = _mm_add_pd(fjy1,ty);
1699             fjz1             = _mm_add_pd(fjz1,tz);
1700
1701             /**************************
1702              * CALCULATE INTERACTIONS *
1703              **************************/
1704
1705             r12              = _mm_mul_pd(rsq12,rinv12);
1706
1707             /* EWALD ELECTROSTATICS */
1708
1709             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1710             ewrt             = _mm_mul_pd(r12,ewtabscale);
1711             ewitab           = _mm_cvttpd_epi32(ewrt);
1712             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1713             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1714                                          &ewtabF,&ewtabFn);
1715             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1716             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1717
1718             fscal            = felec;
1719
1720             /* Calculate temporary vectorial force */
1721             tx               = _mm_mul_pd(fscal,dx12);
1722             ty               = _mm_mul_pd(fscal,dy12);
1723             tz               = _mm_mul_pd(fscal,dz12);
1724
1725             /* Update vectorial force */
1726             fix1             = _mm_add_pd(fix1,tx);
1727             fiy1             = _mm_add_pd(fiy1,ty);
1728             fiz1             = _mm_add_pd(fiz1,tz);
1729
1730             fjx2             = _mm_add_pd(fjx2,tx);
1731             fjy2             = _mm_add_pd(fjy2,ty);
1732             fjz2             = _mm_add_pd(fjz2,tz);
1733
1734             /**************************
1735              * CALCULATE INTERACTIONS *
1736              **************************/
1737
1738             r13              = _mm_mul_pd(rsq13,rinv13);
1739
1740             /* EWALD ELECTROSTATICS */
1741
1742             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1743             ewrt             = _mm_mul_pd(r13,ewtabscale);
1744             ewitab           = _mm_cvttpd_epi32(ewrt);
1745             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1746             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1747                                          &ewtabF,&ewtabFn);
1748             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1749             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1750
1751             fscal            = felec;
1752
1753             /* Calculate temporary vectorial force */
1754             tx               = _mm_mul_pd(fscal,dx13);
1755             ty               = _mm_mul_pd(fscal,dy13);
1756             tz               = _mm_mul_pd(fscal,dz13);
1757
1758             /* Update vectorial force */
1759             fix1             = _mm_add_pd(fix1,tx);
1760             fiy1             = _mm_add_pd(fiy1,ty);
1761             fiz1             = _mm_add_pd(fiz1,tz);
1762
1763             fjx3             = _mm_add_pd(fjx3,tx);
1764             fjy3             = _mm_add_pd(fjy3,ty);
1765             fjz3             = _mm_add_pd(fjz3,tz);
1766
1767             /**************************
1768              * CALCULATE INTERACTIONS *
1769              **************************/
1770
1771             r21              = _mm_mul_pd(rsq21,rinv21);
1772
1773             /* EWALD ELECTROSTATICS */
1774
1775             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1776             ewrt             = _mm_mul_pd(r21,ewtabscale);
1777             ewitab           = _mm_cvttpd_epi32(ewrt);
1778             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1779             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1780                                          &ewtabF,&ewtabFn);
1781             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1782             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1783
1784             fscal            = felec;
1785
1786             /* Calculate temporary vectorial force */
1787             tx               = _mm_mul_pd(fscal,dx21);
1788             ty               = _mm_mul_pd(fscal,dy21);
1789             tz               = _mm_mul_pd(fscal,dz21);
1790
1791             /* Update vectorial force */
1792             fix2             = _mm_add_pd(fix2,tx);
1793             fiy2             = _mm_add_pd(fiy2,ty);
1794             fiz2             = _mm_add_pd(fiz2,tz);
1795
1796             fjx1             = _mm_add_pd(fjx1,tx);
1797             fjy1             = _mm_add_pd(fjy1,ty);
1798             fjz1             = _mm_add_pd(fjz1,tz);
1799
1800             /**************************
1801              * CALCULATE INTERACTIONS *
1802              **************************/
1803
1804             r22              = _mm_mul_pd(rsq22,rinv22);
1805
1806             /* EWALD ELECTROSTATICS */
1807
1808             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1809             ewrt             = _mm_mul_pd(r22,ewtabscale);
1810             ewitab           = _mm_cvttpd_epi32(ewrt);
1811             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1812             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1813                                          &ewtabF,&ewtabFn);
1814             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1815             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1816
1817             fscal            = felec;
1818
1819             /* Calculate temporary vectorial force */
1820             tx               = _mm_mul_pd(fscal,dx22);
1821             ty               = _mm_mul_pd(fscal,dy22);
1822             tz               = _mm_mul_pd(fscal,dz22);
1823
1824             /* Update vectorial force */
1825             fix2             = _mm_add_pd(fix2,tx);
1826             fiy2             = _mm_add_pd(fiy2,ty);
1827             fiz2             = _mm_add_pd(fiz2,tz);
1828
1829             fjx2             = _mm_add_pd(fjx2,tx);
1830             fjy2             = _mm_add_pd(fjy2,ty);
1831             fjz2             = _mm_add_pd(fjz2,tz);
1832
1833             /**************************
1834              * CALCULATE INTERACTIONS *
1835              **************************/
1836
1837             r23              = _mm_mul_pd(rsq23,rinv23);
1838
1839             /* EWALD ELECTROSTATICS */
1840
1841             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1842             ewrt             = _mm_mul_pd(r23,ewtabscale);
1843             ewitab           = _mm_cvttpd_epi32(ewrt);
1844             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1845             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1846                                          &ewtabF,&ewtabFn);
1847             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1848             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1849
1850             fscal            = felec;
1851
1852             /* Calculate temporary vectorial force */
1853             tx               = _mm_mul_pd(fscal,dx23);
1854             ty               = _mm_mul_pd(fscal,dy23);
1855             tz               = _mm_mul_pd(fscal,dz23);
1856
1857             /* Update vectorial force */
1858             fix2             = _mm_add_pd(fix2,tx);
1859             fiy2             = _mm_add_pd(fiy2,ty);
1860             fiz2             = _mm_add_pd(fiz2,tz);
1861
1862             fjx3             = _mm_add_pd(fjx3,tx);
1863             fjy3             = _mm_add_pd(fjy3,ty);
1864             fjz3             = _mm_add_pd(fjz3,tz);
1865
1866             /**************************
1867              * CALCULATE INTERACTIONS *
1868              **************************/
1869
1870             r31              = _mm_mul_pd(rsq31,rinv31);
1871
1872             /* EWALD ELECTROSTATICS */
1873
1874             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1875             ewrt             = _mm_mul_pd(r31,ewtabscale);
1876             ewitab           = _mm_cvttpd_epi32(ewrt);
1877             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1878             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1879                                          &ewtabF,&ewtabFn);
1880             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1881             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1882
1883             fscal            = felec;
1884
1885             /* Calculate temporary vectorial force */
1886             tx               = _mm_mul_pd(fscal,dx31);
1887             ty               = _mm_mul_pd(fscal,dy31);
1888             tz               = _mm_mul_pd(fscal,dz31);
1889
1890             /* Update vectorial force */
1891             fix3             = _mm_add_pd(fix3,tx);
1892             fiy3             = _mm_add_pd(fiy3,ty);
1893             fiz3             = _mm_add_pd(fiz3,tz);
1894
1895             fjx1             = _mm_add_pd(fjx1,tx);
1896             fjy1             = _mm_add_pd(fjy1,ty);
1897             fjz1             = _mm_add_pd(fjz1,tz);
1898
1899             /**************************
1900              * CALCULATE INTERACTIONS *
1901              **************************/
1902
1903             r32              = _mm_mul_pd(rsq32,rinv32);
1904
1905             /* EWALD ELECTROSTATICS */
1906
1907             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1908             ewrt             = _mm_mul_pd(r32,ewtabscale);
1909             ewitab           = _mm_cvttpd_epi32(ewrt);
1910             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1911             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1912                                          &ewtabF,&ewtabFn);
1913             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1914             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1915
1916             fscal            = felec;
1917
1918             /* Calculate temporary vectorial force */
1919             tx               = _mm_mul_pd(fscal,dx32);
1920             ty               = _mm_mul_pd(fscal,dy32);
1921             tz               = _mm_mul_pd(fscal,dz32);
1922
1923             /* Update vectorial force */
1924             fix3             = _mm_add_pd(fix3,tx);
1925             fiy3             = _mm_add_pd(fiy3,ty);
1926             fiz3             = _mm_add_pd(fiz3,tz);
1927
1928             fjx2             = _mm_add_pd(fjx2,tx);
1929             fjy2             = _mm_add_pd(fjy2,ty);
1930             fjz2             = _mm_add_pd(fjz2,tz);
1931
1932             /**************************
1933              * CALCULATE INTERACTIONS *
1934              **************************/
1935
1936             r33              = _mm_mul_pd(rsq33,rinv33);
1937
1938             /* EWALD ELECTROSTATICS */
1939
1940             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1941             ewrt             = _mm_mul_pd(r33,ewtabscale);
1942             ewitab           = _mm_cvttpd_epi32(ewrt);
1943             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1944             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1945                                          &ewtabF,&ewtabFn);
1946             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1947             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1948
1949             fscal            = felec;
1950
1951             /* Calculate temporary vectorial force */
1952             tx               = _mm_mul_pd(fscal,dx33);
1953             ty               = _mm_mul_pd(fscal,dy33);
1954             tz               = _mm_mul_pd(fscal,dz33);
1955
1956             /* Update vectorial force */
1957             fix3             = _mm_add_pd(fix3,tx);
1958             fiy3             = _mm_add_pd(fiy3,ty);
1959             fiz3             = _mm_add_pd(fiz3,tz);
1960
1961             fjx3             = _mm_add_pd(fjx3,tx);
1962             fjy3             = _mm_add_pd(fjy3,ty);
1963             fjz3             = _mm_add_pd(fjz3,tz);
1964
1965             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1966
1967             /* Inner loop uses 375 flops */
1968         }
1969
1970         if(jidx<j_index_end)
1971         {
1972
1973             jnrA             = jjnr[jidx];
1974             j_coord_offsetA  = DIM*jnrA;
1975
1976             /* load j atom coordinates */
1977             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1978                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1979                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1980
1981             /* Calculate displacement vector */
1982             dx00             = _mm_sub_pd(ix0,jx0);
1983             dy00             = _mm_sub_pd(iy0,jy0);
1984             dz00             = _mm_sub_pd(iz0,jz0);
1985             dx11             = _mm_sub_pd(ix1,jx1);
1986             dy11             = _mm_sub_pd(iy1,jy1);
1987             dz11             = _mm_sub_pd(iz1,jz1);
1988             dx12             = _mm_sub_pd(ix1,jx2);
1989             dy12             = _mm_sub_pd(iy1,jy2);
1990             dz12             = _mm_sub_pd(iz1,jz2);
1991             dx13             = _mm_sub_pd(ix1,jx3);
1992             dy13             = _mm_sub_pd(iy1,jy3);
1993             dz13             = _mm_sub_pd(iz1,jz3);
1994             dx21             = _mm_sub_pd(ix2,jx1);
1995             dy21             = _mm_sub_pd(iy2,jy1);
1996             dz21             = _mm_sub_pd(iz2,jz1);
1997             dx22             = _mm_sub_pd(ix2,jx2);
1998             dy22             = _mm_sub_pd(iy2,jy2);
1999             dz22             = _mm_sub_pd(iz2,jz2);
2000             dx23             = _mm_sub_pd(ix2,jx3);
2001             dy23             = _mm_sub_pd(iy2,jy3);
2002             dz23             = _mm_sub_pd(iz2,jz3);
2003             dx31             = _mm_sub_pd(ix3,jx1);
2004             dy31             = _mm_sub_pd(iy3,jy1);
2005             dz31             = _mm_sub_pd(iz3,jz1);
2006             dx32             = _mm_sub_pd(ix3,jx2);
2007             dy32             = _mm_sub_pd(iy3,jy2);
2008             dz32             = _mm_sub_pd(iz3,jz2);
2009             dx33             = _mm_sub_pd(ix3,jx3);
2010             dy33             = _mm_sub_pd(iy3,jy3);
2011             dz33             = _mm_sub_pd(iz3,jz3);
2012
2013             /* Calculate squared distance and things based on it */
2014             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2015             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2016             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2017             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
2018             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2019             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2020             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
2021             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
2022             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
2023             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
2024
2025             rinv00           = gmx_mm_invsqrt_pd(rsq00);
2026             rinv11           = gmx_mm_invsqrt_pd(rsq11);
2027             rinv12           = gmx_mm_invsqrt_pd(rsq12);
2028             rinv13           = gmx_mm_invsqrt_pd(rsq13);
2029             rinv21           = gmx_mm_invsqrt_pd(rsq21);
2030             rinv22           = gmx_mm_invsqrt_pd(rsq22);
2031             rinv23           = gmx_mm_invsqrt_pd(rsq23);
2032             rinv31           = gmx_mm_invsqrt_pd(rsq31);
2033             rinv32           = gmx_mm_invsqrt_pd(rsq32);
2034             rinv33           = gmx_mm_invsqrt_pd(rsq33);
2035
2036             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
2037             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
2038             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
2039             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
2040             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
2041             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
2042             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
2043             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
2044             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
2045
2046             fjx0             = _mm_setzero_pd();
2047             fjy0             = _mm_setzero_pd();
2048             fjz0             = _mm_setzero_pd();
2049             fjx1             = _mm_setzero_pd();
2050             fjy1             = _mm_setzero_pd();
2051             fjz1             = _mm_setzero_pd();
2052             fjx2             = _mm_setzero_pd();
2053             fjy2             = _mm_setzero_pd();
2054             fjz2             = _mm_setzero_pd();
2055             fjx3             = _mm_setzero_pd();
2056             fjy3             = _mm_setzero_pd();
2057             fjz3             = _mm_setzero_pd();
2058
2059             /**************************
2060              * CALCULATE INTERACTIONS *
2061              **************************/
2062
2063             r00              = _mm_mul_pd(rsq00,rinv00);
2064
2065             /* Calculate table index by multiplying r with table scale and truncate to integer */
2066             rt               = _mm_mul_pd(r00,vftabscale);
2067             vfitab           = _mm_cvttpd_epi32(rt);
2068             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
2069             vfitab           = _mm_slli_epi32(vfitab,3);
2070
2071             /* CUBIC SPLINE TABLE DISPERSION */
2072             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2073             F                = _mm_setzero_pd();
2074             GMX_MM_TRANSPOSE2_PD(Y,F);
2075             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2076             H                = _mm_setzero_pd();
2077             GMX_MM_TRANSPOSE2_PD(G,H);
2078             Heps             = _mm_mul_pd(vfeps,H);
2079             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2080             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2081             fvdw6            = _mm_mul_pd(c6_00,FF);
2082
2083             /* CUBIC SPLINE TABLE REPULSION */
2084             vfitab           = _mm_add_epi32(vfitab,ifour);
2085             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2086             F                = _mm_setzero_pd();
2087             GMX_MM_TRANSPOSE2_PD(Y,F);
2088             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2089             H                = _mm_setzero_pd();
2090             GMX_MM_TRANSPOSE2_PD(G,H);
2091             Heps             = _mm_mul_pd(vfeps,H);
2092             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2093             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2094             fvdw12           = _mm_mul_pd(c12_00,FF);
2095             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
2096
2097             fscal            = fvdw;
2098
2099             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2100
2101             /* Calculate temporary vectorial force */
2102             tx               = _mm_mul_pd(fscal,dx00);
2103             ty               = _mm_mul_pd(fscal,dy00);
2104             tz               = _mm_mul_pd(fscal,dz00);
2105
2106             /* Update vectorial force */
2107             fix0             = _mm_add_pd(fix0,tx);
2108             fiy0             = _mm_add_pd(fiy0,ty);
2109             fiz0             = _mm_add_pd(fiz0,tz);
2110
2111             fjx0             = _mm_add_pd(fjx0,tx);
2112             fjy0             = _mm_add_pd(fjy0,ty);
2113             fjz0             = _mm_add_pd(fjz0,tz);
2114
2115             /**************************
2116              * CALCULATE INTERACTIONS *
2117              **************************/
2118
2119             r11              = _mm_mul_pd(rsq11,rinv11);
2120
2121             /* EWALD ELECTROSTATICS */
2122
2123             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2124             ewrt             = _mm_mul_pd(r11,ewtabscale);
2125             ewitab           = _mm_cvttpd_epi32(ewrt);
2126             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2127             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2128             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2129             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2130
2131             fscal            = felec;
2132
2133             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2134
2135             /* Calculate temporary vectorial force */
2136             tx               = _mm_mul_pd(fscal,dx11);
2137             ty               = _mm_mul_pd(fscal,dy11);
2138             tz               = _mm_mul_pd(fscal,dz11);
2139
2140             /* Update vectorial force */
2141             fix1             = _mm_add_pd(fix1,tx);
2142             fiy1             = _mm_add_pd(fiy1,ty);
2143             fiz1             = _mm_add_pd(fiz1,tz);
2144
2145             fjx1             = _mm_add_pd(fjx1,tx);
2146             fjy1             = _mm_add_pd(fjy1,ty);
2147             fjz1             = _mm_add_pd(fjz1,tz);
2148
2149             /**************************
2150              * CALCULATE INTERACTIONS *
2151              **************************/
2152
2153             r12              = _mm_mul_pd(rsq12,rinv12);
2154
2155             /* EWALD ELECTROSTATICS */
2156
2157             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2158             ewrt             = _mm_mul_pd(r12,ewtabscale);
2159             ewitab           = _mm_cvttpd_epi32(ewrt);
2160             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2161             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2162             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2163             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2164
2165             fscal            = felec;
2166
2167             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2168
2169             /* Calculate temporary vectorial force */
2170             tx               = _mm_mul_pd(fscal,dx12);
2171             ty               = _mm_mul_pd(fscal,dy12);
2172             tz               = _mm_mul_pd(fscal,dz12);
2173
2174             /* Update vectorial force */
2175             fix1             = _mm_add_pd(fix1,tx);
2176             fiy1             = _mm_add_pd(fiy1,ty);
2177             fiz1             = _mm_add_pd(fiz1,tz);
2178
2179             fjx2             = _mm_add_pd(fjx2,tx);
2180             fjy2             = _mm_add_pd(fjy2,ty);
2181             fjz2             = _mm_add_pd(fjz2,tz);
2182
2183             /**************************
2184              * CALCULATE INTERACTIONS *
2185              **************************/
2186
2187             r13              = _mm_mul_pd(rsq13,rinv13);
2188
2189             /* EWALD ELECTROSTATICS */
2190
2191             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2192             ewrt             = _mm_mul_pd(r13,ewtabscale);
2193             ewitab           = _mm_cvttpd_epi32(ewrt);
2194             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2195             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2196             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2197             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
2198
2199             fscal            = felec;
2200
2201             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2202
2203             /* Calculate temporary vectorial force */
2204             tx               = _mm_mul_pd(fscal,dx13);
2205             ty               = _mm_mul_pd(fscal,dy13);
2206             tz               = _mm_mul_pd(fscal,dz13);
2207
2208             /* Update vectorial force */
2209             fix1             = _mm_add_pd(fix1,tx);
2210             fiy1             = _mm_add_pd(fiy1,ty);
2211             fiz1             = _mm_add_pd(fiz1,tz);
2212
2213             fjx3             = _mm_add_pd(fjx3,tx);
2214             fjy3             = _mm_add_pd(fjy3,ty);
2215             fjz3             = _mm_add_pd(fjz3,tz);
2216
2217             /**************************
2218              * CALCULATE INTERACTIONS *
2219              **************************/
2220
2221             r21              = _mm_mul_pd(rsq21,rinv21);
2222
2223             /* EWALD ELECTROSTATICS */
2224
2225             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2226             ewrt             = _mm_mul_pd(r21,ewtabscale);
2227             ewitab           = _mm_cvttpd_epi32(ewrt);
2228             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2229             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2230             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2231             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2232
2233             fscal            = felec;
2234
2235             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2236
2237             /* Calculate temporary vectorial force */
2238             tx               = _mm_mul_pd(fscal,dx21);
2239             ty               = _mm_mul_pd(fscal,dy21);
2240             tz               = _mm_mul_pd(fscal,dz21);
2241
2242             /* Update vectorial force */
2243             fix2             = _mm_add_pd(fix2,tx);
2244             fiy2             = _mm_add_pd(fiy2,ty);
2245             fiz2             = _mm_add_pd(fiz2,tz);
2246
2247             fjx1             = _mm_add_pd(fjx1,tx);
2248             fjy1             = _mm_add_pd(fjy1,ty);
2249             fjz1             = _mm_add_pd(fjz1,tz);
2250
2251             /**************************
2252              * CALCULATE INTERACTIONS *
2253              **************************/
2254
2255             r22              = _mm_mul_pd(rsq22,rinv22);
2256
2257             /* EWALD ELECTROSTATICS */
2258
2259             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2260             ewrt             = _mm_mul_pd(r22,ewtabscale);
2261             ewitab           = _mm_cvttpd_epi32(ewrt);
2262             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2263             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2264             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2265             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2266
2267             fscal            = felec;
2268
2269             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2270
2271             /* Calculate temporary vectorial force */
2272             tx               = _mm_mul_pd(fscal,dx22);
2273             ty               = _mm_mul_pd(fscal,dy22);
2274             tz               = _mm_mul_pd(fscal,dz22);
2275
2276             /* Update vectorial force */
2277             fix2             = _mm_add_pd(fix2,tx);
2278             fiy2             = _mm_add_pd(fiy2,ty);
2279             fiz2             = _mm_add_pd(fiz2,tz);
2280
2281             fjx2             = _mm_add_pd(fjx2,tx);
2282             fjy2             = _mm_add_pd(fjy2,ty);
2283             fjz2             = _mm_add_pd(fjz2,tz);
2284
2285             /**************************
2286              * CALCULATE INTERACTIONS *
2287              **************************/
2288
2289             r23              = _mm_mul_pd(rsq23,rinv23);
2290
2291             /* EWALD ELECTROSTATICS */
2292
2293             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2294             ewrt             = _mm_mul_pd(r23,ewtabscale);
2295             ewitab           = _mm_cvttpd_epi32(ewrt);
2296             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2297             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2298             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2299             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2300
2301             fscal            = felec;
2302
2303             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2304
2305             /* Calculate temporary vectorial force */
2306             tx               = _mm_mul_pd(fscal,dx23);
2307             ty               = _mm_mul_pd(fscal,dy23);
2308             tz               = _mm_mul_pd(fscal,dz23);
2309
2310             /* Update vectorial force */
2311             fix2             = _mm_add_pd(fix2,tx);
2312             fiy2             = _mm_add_pd(fiy2,ty);
2313             fiz2             = _mm_add_pd(fiz2,tz);
2314
2315             fjx3             = _mm_add_pd(fjx3,tx);
2316             fjy3             = _mm_add_pd(fjy3,ty);
2317             fjz3             = _mm_add_pd(fjz3,tz);
2318
2319             /**************************
2320              * CALCULATE INTERACTIONS *
2321              **************************/
2322
2323             r31              = _mm_mul_pd(rsq31,rinv31);
2324
2325             /* EWALD ELECTROSTATICS */
2326
2327             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2328             ewrt             = _mm_mul_pd(r31,ewtabscale);
2329             ewitab           = _mm_cvttpd_epi32(ewrt);
2330             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2331             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2332             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2333             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2334
2335             fscal            = felec;
2336
2337             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2338
2339             /* Calculate temporary vectorial force */
2340             tx               = _mm_mul_pd(fscal,dx31);
2341             ty               = _mm_mul_pd(fscal,dy31);
2342             tz               = _mm_mul_pd(fscal,dz31);
2343
2344             /* Update vectorial force */
2345             fix3             = _mm_add_pd(fix3,tx);
2346             fiy3             = _mm_add_pd(fiy3,ty);
2347             fiz3             = _mm_add_pd(fiz3,tz);
2348
2349             fjx1             = _mm_add_pd(fjx1,tx);
2350             fjy1             = _mm_add_pd(fjy1,ty);
2351             fjz1             = _mm_add_pd(fjz1,tz);
2352
2353             /**************************
2354              * CALCULATE INTERACTIONS *
2355              **************************/
2356
2357             r32              = _mm_mul_pd(rsq32,rinv32);
2358
2359             /* EWALD ELECTROSTATICS */
2360
2361             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2362             ewrt             = _mm_mul_pd(r32,ewtabscale);
2363             ewitab           = _mm_cvttpd_epi32(ewrt);
2364             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2365             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2366             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2367             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2368
2369             fscal            = felec;
2370
2371             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2372
2373             /* Calculate temporary vectorial force */
2374             tx               = _mm_mul_pd(fscal,dx32);
2375             ty               = _mm_mul_pd(fscal,dy32);
2376             tz               = _mm_mul_pd(fscal,dz32);
2377
2378             /* Update vectorial force */
2379             fix3             = _mm_add_pd(fix3,tx);
2380             fiy3             = _mm_add_pd(fiy3,ty);
2381             fiz3             = _mm_add_pd(fiz3,tz);
2382
2383             fjx2             = _mm_add_pd(fjx2,tx);
2384             fjy2             = _mm_add_pd(fjy2,ty);
2385             fjz2             = _mm_add_pd(fjz2,tz);
2386
2387             /**************************
2388              * CALCULATE INTERACTIONS *
2389              **************************/
2390
2391             r33              = _mm_mul_pd(rsq33,rinv33);
2392
2393             /* EWALD ELECTROSTATICS */
2394
2395             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2396             ewrt             = _mm_mul_pd(r33,ewtabscale);
2397             ewitab           = _mm_cvttpd_epi32(ewrt);
2398             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2399             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2400             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2401             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2402
2403             fscal            = felec;
2404
2405             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2406
2407             /* Calculate temporary vectorial force */
2408             tx               = _mm_mul_pd(fscal,dx33);
2409             ty               = _mm_mul_pd(fscal,dy33);
2410             tz               = _mm_mul_pd(fscal,dz33);
2411
2412             /* Update vectorial force */
2413             fix3             = _mm_add_pd(fix3,tx);
2414             fiy3             = _mm_add_pd(fiy3,ty);
2415             fiz3             = _mm_add_pd(fiz3,tz);
2416
2417             fjx3             = _mm_add_pd(fjx3,tx);
2418             fjy3             = _mm_add_pd(fjy3,ty);
2419             fjz3             = _mm_add_pd(fjz3,tz);
2420
2421             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2422
2423             /* Inner loop uses 375 flops */
2424         }
2425
2426         /* End of innermost loop */
2427
2428         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2429                                               f+i_coord_offset,fshift+i_shift_offset);
2430
2431         /* Increment number of inner iterations */
2432         inneriter                  += j_index_end - j_index_start;
2433
2434         /* Outer loop uses 24 flops */
2435     }
2436
2437     /* Increment number of outer iterations */
2438     outeriter        += nri;
2439
2440     /* Update outer/inner flops */
2441
2442     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*375);
2443 }