src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecCSTab_VdwNone_GeomW4W4_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse4_1_double.h"
  48 #include "kernelutil_x86_sse4_1_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW4W4_VF_sse4_1_double
  52  * Electrostatics interaction: CubicSplineTable
  53  * VdW interaction:            None
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCSTab_VdwNone_GeomW4W4_VF_sse4_1_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset1;
  81     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  82     int              vdwioffset2;
  83     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  84     int              vdwioffset3;
  85     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  86     int              vdwjidx1A,vdwjidx1B;
  87     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  88     int              vdwjidx2A,vdwjidx2B;
  89     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  90     int              vdwjidx3A,vdwjidx3B;
  91     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  92     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  93     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  94     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  95     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  96     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  97     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  98     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  99     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 100     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 101     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 102     real             *charge;
 103     __m128i          vfitab;
 104     __m128i          ifour       = _mm_set1_epi32(4);
 105     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 106     real             *vftab;
 107     __m128d          dummy_mask,cutoff_mask;
 108     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 109     __m128d          one     = _mm_set1_pd(1.0);
 110     __m128d          two     = _mm_set1_pd(2.0);
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = _mm_set1_pd(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124
 125     vftab            = kernel_data->table_elec->data;
 126     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 131     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 132     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 133
 134     jq1              = _mm_set1_pd(charge[inr+1]);
 135     jq2              = _mm_set1_pd(charge[inr+2]);
 136     jq3              = _mm_set1_pd(charge[inr+3]);
 137     qq11             = _mm_mul_pd(iq1,jq1);
 138     qq12             = _mm_mul_pd(iq1,jq2);
 139     qq13             = _mm_mul_pd(iq1,jq3);
 140     qq21             = _mm_mul_pd(iq2,jq1);
 141     qq22             = _mm_mul_pd(iq2,jq2);
 142     qq23             = _mm_mul_pd(iq2,jq3);
 143     qq31             = _mm_mul_pd(iq3,jq1);
 144     qq32             = _mm_mul_pd(iq3,jq2);
 145     qq33             = _mm_mul_pd(iq3,jq3);
 146
 147     /* Avoid stupid compiler warnings */
 148     jnrA = jnrB = 0;
 149     j_coord_offsetA = 0;
 150     j_coord_offsetB = 0;
 151
 152     outeriter        = 0;
 153     inneriter        = 0;
 154
 155     /* Start outer loop over neighborlists */
 156     for(iidx=0; iidx<nri; iidx++)
 157     {
 158         /* Load shift vector for this list */
 159         i_shift_offset   = DIM*shiftidx[iidx];
 160
 161         /* Load limits for loop over neighbors */
 162         j_index_start    = jindex[iidx];
 163         j_index_end      = jindex[iidx+1];
 164
 165         /* Get outer coordinate index */
 166         inr              = iinr[iidx];
 167         i_coord_offset   = DIM*inr;
 168
 169         /* Load i particle coords and add shift vector */
 170         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 171                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 172
 173         fix1             = _mm_setzero_pd();
 174         fiy1             = _mm_setzero_pd();
 175         fiz1             = _mm_setzero_pd();
 176         fix2             = _mm_setzero_pd();
 177         fiy2             = _mm_setzero_pd();
 178         fiz2             = _mm_setzero_pd();
 179         fix3             = _mm_setzero_pd();
 180         fiy3             = _mm_setzero_pd();
 181         fiz3             = _mm_setzero_pd();
 182
 183         /* Reset potential sums */
 184         velecsum         = _mm_setzero_pd();
 185
 186         /* Start inner kernel loop */
 187         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 188         {
 189
 190             /* Get j neighbor index, and coordinate index */
 191             jnrA             = jjnr[jidx];
 192             jnrB             = jjnr[jidx+1];
 193             j_coord_offsetA  = DIM*jnrA;
 194             j_coord_offsetB  = DIM*jnrB;
 195
 196             /* load j atom coordinates */
 197             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
 198                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 199
 200             /* Calculate displacement vector */
 201             dx11             = _mm_sub_pd(ix1,jx1);
 202             dy11             = _mm_sub_pd(iy1,jy1);
 203             dz11             = _mm_sub_pd(iz1,jz1);
 204             dx12             = _mm_sub_pd(ix1,jx2);
 205             dy12             = _mm_sub_pd(iy1,jy2);
 206             dz12             = _mm_sub_pd(iz1,jz2);
 207             dx13             = _mm_sub_pd(ix1,jx3);
 208             dy13             = _mm_sub_pd(iy1,jy3);
 209             dz13             = _mm_sub_pd(iz1,jz3);
 210             dx21             = _mm_sub_pd(ix2,jx1);
 211             dy21             = _mm_sub_pd(iy2,jy1);
 212             dz21             = _mm_sub_pd(iz2,jz1);
 213             dx22             = _mm_sub_pd(ix2,jx2);
 214             dy22             = _mm_sub_pd(iy2,jy2);
 215             dz22             = _mm_sub_pd(iz2,jz2);
 216             dx23             = _mm_sub_pd(ix2,jx3);
 217             dy23             = _mm_sub_pd(iy2,jy3);
 218             dz23             = _mm_sub_pd(iz2,jz3);
 219             dx31             = _mm_sub_pd(ix3,jx1);
 220             dy31             = _mm_sub_pd(iy3,jy1);
 221             dz31             = _mm_sub_pd(iz3,jz1);
 222             dx32             = _mm_sub_pd(ix3,jx2);
 223             dy32             = _mm_sub_pd(iy3,jy2);
 224             dz32             = _mm_sub_pd(iz3,jz2);
 225             dx33             = _mm_sub_pd(ix3,jx3);
 226             dy33             = _mm_sub_pd(iy3,jy3);
 227             dz33             = _mm_sub_pd(iz3,jz3);
 228
 229             /* Calculate squared distance and things based on it */
 230             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 231             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 232             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 233             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 234             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 235             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 236             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 237             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 238             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 239
 240             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 241             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 242             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 243             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 244             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 245             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 246             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 247             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 248             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 249
 250             fjx1             = _mm_setzero_pd();
 251             fjy1             = _mm_setzero_pd();
 252             fjz1             = _mm_setzero_pd();
 253             fjx2             = _mm_setzero_pd();
 254             fjy2             = _mm_setzero_pd();
 255             fjz2             = _mm_setzero_pd();
 256             fjx3             = _mm_setzero_pd();
 257             fjy3             = _mm_setzero_pd();
 258             fjz3             = _mm_setzero_pd();
 259
 260             /**************************
 261              * CALCULATE INTERACTIONS *
 262              **************************/
 263
 264             r11              = _mm_mul_pd(rsq11,rinv11);
 265
 266             /* Calculate table index by multiplying r with table scale and truncate to integer */
 267             rt               = _mm_mul_pd(r11,vftabscale);
 268             vfitab           = _mm_cvttpd_epi32(rt);
 269             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 270             vfitab           = _mm_slli_epi32(vfitab,2);
 271
 272             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 273             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 274             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 275             GMX_MM_TRANSPOSE2_PD(Y,F);
 276             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 277             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 278             GMX_MM_TRANSPOSE2_PD(G,H);
 279             Heps             = _mm_mul_pd(vfeps,H);
 280             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 281             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 282             velec            = _mm_mul_pd(qq11,VV);
 283             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 284             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 285
 286             /* Update potential sum for this i atom from the interaction with this j atom. */
 287             velecsum         = _mm_add_pd(velecsum,velec);
 288
 289             fscal            = felec;
 290
 291             /* Calculate temporary vectorial force */
 292             tx               = _mm_mul_pd(fscal,dx11);
 293             ty               = _mm_mul_pd(fscal,dy11);
 294             tz               = _mm_mul_pd(fscal,dz11);
 295
 296             /* Update vectorial force */
 297             fix1             = _mm_add_pd(fix1,tx);
 298             fiy1             = _mm_add_pd(fiy1,ty);
 299             fiz1             = _mm_add_pd(fiz1,tz);
 300
 301             fjx1             = _mm_add_pd(fjx1,tx);
 302             fjy1             = _mm_add_pd(fjy1,ty);
 303             fjz1             = _mm_add_pd(fjz1,tz);
 304
 305             /**************************
 306              * CALCULATE INTERACTIONS *
 307              **************************/
 308
 309             r12              = _mm_mul_pd(rsq12,rinv12);
 310
 311             /* Calculate table index by multiplying r with table scale and truncate to integer */
 312             rt               = _mm_mul_pd(r12,vftabscale);
 313             vfitab           = _mm_cvttpd_epi32(rt);
 314             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 315             vfitab           = _mm_slli_epi32(vfitab,2);
 316
 317             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 318             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 319             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 320             GMX_MM_TRANSPOSE2_PD(Y,F);
 321             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 322             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 323             GMX_MM_TRANSPOSE2_PD(G,H);
 324             Heps             = _mm_mul_pd(vfeps,H);
 325             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 326             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 327             velec            = _mm_mul_pd(qq12,VV);
 328             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 329             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 330
 331             /* Update potential sum for this i atom from the interaction with this j atom. */
 332             velecsum         = _mm_add_pd(velecsum,velec);
 333
 334             fscal            = felec;
 335
 336             /* Calculate temporary vectorial force */
 337             tx               = _mm_mul_pd(fscal,dx12);
 338             ty               = _mm_mul_pd(fscal,dy12);
 339             tz               = _mm_mul_pd(fscal,dz12);
 340
 341             /* Update vectorial force */
 342             fix1             = _mm_add_pd(fix1,tx);
 343             fiy1             = _mm_add_pd(fiy1,ty);
 344             fiz1             = _mm_add_pd(fiz1,tz);
 345
 346             fjx2             = _mm_add_pd(fjx2,tx);
 347             fjy2             = _mm_add_pd(fjy2,ty);
 348             fjz2             = _mm_add_pd(fjz2,tz);
 349
 350             /**************************
 351              * CALCULATE INTERACTIONS *
 352              **************************/
 353
 354             r13              = _mm_mul_pd(rsq13,rinv13);
 355
 356             /* Calculate table index by multiplying r with table scale and truncate to integer */
 357             rt               = _mm_mul_pd(r13,vftabscale);
 358             vfitab           = _mm_cvttpd_epi32(rt);
 359             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 360             vfitab           = _mm_slli_epi32(vfitab,2);
 361
 362             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 363             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 364             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 365             GMX_MM_TRANSPOSE2_PD(Y,F);
 366             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 367             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 368             GMX_MM_TRANSPOSE2_PD(G,H);
 369             Heps             = _mm_mul_pd(vfeps,H);
 370             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 371             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 372             velec            = _mm_mul_pd(qq13,VV);
 373             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 374             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
 375
 376             /* Update potential sum for this i atom from the interaction with this j atom. */
 377             velecsum         = _mm_add_pd(velecsum,velec);
 378
 379             fscal            = felec;
 380
 381             /* Calculate temporary vectorial force */
 382             tx               = _mm_mul_pd(fscal,dx13);
 383             ty               = _mm_mul_pd(fscal,dy13);
 384             tz               = _mm_mul_pd(fscal,dz13);
 385
 386             /* Update vectorial force */
 387             fix1             = _mm_add_pd(fix1,tx);
 388             fiy1             = _mm_add_pd(fiy1,ty);
 389             fiz1             = _mm_add_pd(fiz1,tz);
 390
 391             fjx3             = _mm_add_pd(fjx3,tx);
 392             fjy3             = _mm_add_pd(fjy3,ty);
 393             fjz3             = _mm_add_pd(fjz3,tz);
 394
 395             /**************************
 396              * CALCULATE INTERACTIONS *
 397              **************************/
 398
 399             r21              = _mm_mul_pd(rsq21,rinv21);
 400
 401             /* Calculate table index by multiplying r with table scale and truncate to integer */
 402             rt               = _mm_mul_pd(r21,vftabscale);
 403             vfitab           = _mm_cvttpd_epi32(rt);
 404             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 405             vfitab           = _mm_slli_epi32(vfitab,2);
 406
 407             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 408             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 409             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 410             GMX_MM_TRANSPOSE2_PD(Y,F);
 411             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 412             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 413             GMX_MM_TRANSPOSE2_PD(G,H);
 414             Heps             = _mm_mul_pd(vfeps,H);
 415             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 416             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 417             velec            = _mm_mul_pd(qq21,VV);
 418             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 419             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 420
 421             /* Update potential sum for this i atom from the interaction with this j atom. */
 422             velecsum         = _mm_add_pd(velecsum,velec);
 423
 424             fscal            = felec;
 425
 426             /* Calculate temporary vectorial force */
 427             tx               = _mm_mul_pd(fscal,dx21);
 428             ty               = _mm_mul_pd(fscal,dy21);
 429             tz               = _mm_mul_pd(fscal,dz21);
 430
 431             /* Update vectorial force */
 432             fix2             = _mm_add_pd(fix2,tx);
 433             fiy2             = _mm_add_pd(fiy2,ty);
 434             fiz2             = _mm_add_pd(fiz2,tz);
 435
 436             fjx1             = _mm_add_pd(fjx1,tx);
 437             fjy1             = _mm_add_pd(fjy1,ty);
 438             fjz1             = _mm_add_pd(fjz1,tz);
 439
 440             /**************************
 441              * CALCULATE INTERACTIONS *
 442              **************************/
 443
 444             r22              = _mm_mul_pd(rsq22,rinv22);
 445
 446             /* Calculate table index by multiplying r with table scale and truncate to integer */
 447             rt               = _mm_mul_pd(r22,vftabscale);
 448             vfitab           = _mm_cvttpd_epi32(rt);
 449             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 450             vfitab           = _mm_slli_epi32(vfitab,2);
 451
 452             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 453             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 454             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 455             GMX_MM_TRANSPOSE2_PD(Y,F);
 456             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 457             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 458             GMX_MM_TRANSPOSE2_PD(G,H);
 459             Heps             = _mm_mul_pd(vfeps,H);
 460             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 461             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 462             velec            = _mm_mul_pd(qq22,VV);
 463             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 464             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 465
 466             /* Update potential sum for this i atom from the interaction with this j atom. */
 467             velecsum         = _mm_add_pd(velecsum,velec);
 468
 469             fscal            = felec;
 470
 471             /* Calculate temporary vectorial force */
 472             tx               = _mm_mul_pd(fscal,dx22);
 473             ty               = _mm_mul_pd(fscal,dy22);
 474             tz               = _mm_mul_pd(fscal,dz22);
 475
 476             /* Update vectorial force */
 477             fix2             = _mm_add_pd(fix2,tx);
 478             fiy2             = _mm_add_pd(fiy2,ty);
 479             fiz2             = _mm_add_pd(fiz2,tz);
 480
 481             fjx2             = _mm_add_pd(fjx2,tx);
 482             fjy2             = _mm_add_pd(fjy2,ty);
 483             fjz2             = _mm_add_pd(fjz2,tz);
 484
 485             /**************************
 486              * CALCULATE INTERACTIONS *
 487              **************************/
 488
 489             r23              = _mm_mul_pd(rsq23,rinv23);
 490
 491             /* Calculate table index by multiplying r with table scale and truncate to integer */
 492             rt               = _mm_mul_pd(r23,vftabscale);
 493             vfitab           = _mm_cvttpd_epi32(rt);
 494             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 495             vfitab           = _mm_slli_epi32(vfitab,2);
 496
 497             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 498             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 499             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 500             GMX_MM_TRANSPOSE2_PD(Y,F);
 501             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 502             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 503             GMX_MM_TRANSPOSE2_PD(G,H);
 504             Heps             = _mm_mul_pd(vfeps,H);
 505             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 506             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 507             velec            = _mm_mul_pd(qq23,VV);
 508             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 509             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
 510
 511             /* Update potential sum for this i atom from the interaction with this j atom. */
 512             velecsum         = _mm_add_pd(velecsum,velec);
 513
 514             fscal            = felec;
 515
 516             /* Calculate temporary vectorial force */
 517             tx               = _mm_mul_pd(fscal,dx23);
 518             ty               = _mm_mul_pd(fscal,dy23);
 519             tz               = _mm_mul_pd(fscal,dz23);
 520
 521             /* Update vectorial force */
 522             fix2             = _mm_add_pd(fix2,tx);
 523             fiy2             = _mm_add_pd(fiy2,ty);
 524             fiz2             = _mm_add_pd(fiz2,tz);
 525
 526             fjx3             = _mm_add_pd(fjx3,tx);
 527             fjy3             = _mm_add_pd(fjy3,ty);
 528             fjz3             = _mm_add_pd(fjz3,tz);
 529
 530             /**************************
 531              * CALCULATE INTERACTIONS *
 532              **************************/
 533
 534             r31              = _mm_mul_pd(rsq31,rinv31);
 535
 536             /* Calculate table index by multiplying r with table scale and truncate to integer */
 537             rt               = _mm_mul_pd(r31,vftabscale);
 538             vfitab           = _mm_cvttpd_epi32(rt);
 539             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 540             vfitab           = _mm_slli_epi32(vfitab,2);
 541
 542             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 543             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 544             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 545             GMX_MM_TRANSPOSE2_PD(Y,F);
 546             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 547             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 548             GMX_MM_TRANSPOSE2_PD(G,H);
 549             Heps             = _mm_mul_pd(vfeps,H);
 550             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 551             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 552             velec            = _mm_mul_pd(qq31,VV);
 553             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 554             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
 555
 556             /* Update potential sum for this i atom from the interaction with this j atom. */
 557             velecsum         = _mm_add_pd(velecsum,velec);
 558
 559             fscal            = felec;
 560
 561             /* Calculate temporary vectorial force */
 562             tx               = _mm_mul_pd(fscal,dx31);
 563             ty               = _mm_mul_pd(fscal,dy31);
 564             tz               = _mm_mul_pd(fscal,dz31);
 565
 566             /* Update vectorial force */
 567             fix3             = _mm_add_pd(fix3,tx);
 568             fiy3             = _mm_add_pd(fiy3,ty);
 569             fiz3             = _mm_add_pd(fiz3,tz);
 570
 571             fjx1             = _mm_add_pd(fjx1,tx);
 572             fjy1             = _mm_add_pd(fjy1,ty);
 573             fjz1             = _mm_add_pd(fjz1,tz);
 574
 575             /**************************
 576              * CALCULATE INTERACTIONS *
 577              **************************/
 578
 579             r32              = _mm_mul_pd(rsq32,rinv32);
 580
 581             /* Calculate table index by multiplying r with table scale and truncate to integer */
 582             rt               = _mm_mul_pd(r32,vftabscale);
 583             vfitab           = _mm_cvttpd_epi32(rt);
 584             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 585             vfitab           = _mm_slli_epi32(vfitab,2);
 586
 587             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 588             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 589             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 590             GMX_MM_TRANSPOSE2_PD(Y,F);
 591             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 592             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 593             GMX_MM_TRANSPOSE2_PD(G,H);
 594             Heps             = _mm_mul_pd(vfeps,H);
 595             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 596             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 597             velec            = _mm_mul_pd(qq32,VV);
 598             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 599             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
 600
 601             /* Update potential sum for this i atom from the interaction with this j atom. */
 602             velecsum         = _mm_add_pd(velecsum,velec);
 603
 604             fscal            = felec;
 605
 606             /* Calculate temporary vectorial force */
 607             tx               = _mm_mul_pd(fscal,dx32);
 608             ty               = _mm_mul_pd(fscal,dy32);
 609             tz               = _mm_mul_pd(fscal,dz32);
 610
 611             /* Update vectorial force */
 612             fix3             = _mm_add_pd(fix3,tx);
 613             fiy3             = _mm_add_pd(fiy3,ty);
 614             fiz3             = _mm_add_pd(fiz3,tz);
 615
 616             fjx2             = _mm_add_pd(fjx2,tx);
 617             fjy2             = _mm_add_pd(fjy2,ty);
 618             fjz2             = _mm_add_pd(fjz2,tz);
 619
 620             /**************************
 621              * CALCULATE INTERACTIONS *
 622              **************************/
 623
 624             r33              = _mm_mul_pd(rsq33,rinv33);
 625
 626             /* Calculate table index by multiplying r with table scale and truncate to integer */
 627             rt               = _mm_mul_pd(r33,vftabscale);
 628             vfitab           = _mm_cvttpd_epi32(rt);
 629             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 630             vfitab           = _mm_slli_epi32(vfitab,2);
 631
 632             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 633             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 634             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 635             GMX_MM_TRANSPOSE2_PD(Y,F);
 636             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 637             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 638             GMX_MM_TRANSPOSE2_PD(G,H);
 639             Heps             = _mm_mul_pd(vfeps,H);
 640             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 641             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 642             velec            = _mm_mul_pd(qq33,VV);
 643             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 644             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
 645
 646             /* Update potential sum for this i atom from the interaction with this j atom. */
 647             velecsum         = _mm_add_pd(velecsum,velec);
 648
 649             fscal            = felec;
 650
 651             /* Calculate temporary vectorial force */
 652             tx               = _mm_mul_pd(fscal,dx33);
 653             ty               = _mm_mul_pd(fscal,dy33);
 654             tz               = _mm_mul_pd(fscal,dz33);
 655
 656             /* Update vectorial force */
 657             fix3             = _mm_add_pd(fix3,tx);
 658             fiy3             = _mm_add_pd(fiy3,ty);
 659             fiz3             = _mm_add_pd(fiz3,tz);
 660
 661             fjx3             = _mm_add_pd(fjx3,tx);
 662             fjy3             = _mm_add_pd(fjy3,ty);
 663             fjz3             = _mm_add_pd(fjz3,tz);
 664
 665             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 666
 667             /* Inner loop uses 387 flops */
 668         }
 669
 670         if(jidx<j_index_end)
 671         {
 672
 673             jnrA             = jjnr[jidx];
 674             j_coord_offsetA  = DIM*jnrA;
 675
 676             /* load j atom coordinates */
 677             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
 678                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
 679
 680             /* Calculate displacement vector */
 681             dx11             = _mm_sub_pd(ix1,jx1);
 682             dy11             = _mm_sub_pd(iy1,jy1);
 683             dz11             = _mm_sub_pd(iz1,jz1);
 684             dx12             = _mm_sub_pd(ix1,jx2);
 685             dy12             = _mm_sub_pd(iy1,jy2);
 686             dz12             = _mm_sub_pd(iz1,jz2);
 687             dx13             = _mm_sub_pd(ix1,jx3);
 688             dy13             = _mm_sub_pd(iy1,jy3);
 689             dz13             = _mm_sub_pd(iz1,jz3);
 690             dx21             = _mm_sub_pd(ix2,jx1);
 691             dy21             = _mm_sub_pd(iy2,jy1);
 692             dz21             = _mm_sub_pd(iz2,jz1);
 693             dx22             = _mm_sub_pd(ix2,jx2);
 694             dy22             = _mm_sub_pd(iy2,jy2);
 695             dz22             = _mm_sub_pd(iz2,jz2);
 696             dx23             = _mm_sub_pd(ix2,jx3);
 697             dy23             = _mm_sub_pd(iy2,jy3);
 698             dz23             = _mm_sub_pd(iz2,jz3);
 699             dx31             = _mm_sub_pd(ix3,jx1);
 700             dy31             = _mm_sub_pd(iy3,jy1);
 701             dz31             = _mm_sub_pd(iz3,jz1);
 702             dx32             = _mm_sub_pd(ix3,jx2);
 703             dy32             = _mm_sub_pd(iy3,jy2);
 704             dz32             = _mm_sub_pd(iz3,jz2);
 705             dx33             = _mm_sub_pd(ix3,jx3);
 706             dy33             = _mm_sub_pd(iy3,jy3);
 707             dz33             = _mm_sub_pd(iz3,jz3);
 708
 709             /* Calculate squared distance and things based on it */
 710             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 711             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 712             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 713             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 714             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 715             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 716             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 717             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 718             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 719
 720             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 721             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 722             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 723             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 724             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 725             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 726             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 727             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 728             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 729
 730             fjx1             = _mm_setzero_pd();
 731             fjy1             = _mm_setzero_pd();
 732             fjz1             = _mm_setzero_pd();
 733             fjx2             = _mm_setzero_pd();
 734             fjy2             = _mm_setzero_pd();
 735             fjz2             = _mm_setzero_pd();
 736             fjx3             = _mm_setzero_pd();
 737             fjy3             = _mm_setzero_pd();
 738             fjz3             = _mm_setzero_pd();
 739
 740             /**************************
 741              * CALCULATE INTERACTIONS *
 742              **************************/
 743
 744             r11              = _mm_mul_pd(rsq11,rinv11);
 745
 746             /* Calculate table index by multiplying r with table scale and truncate to integer */
 747             rt               = _mm_mul_pd(r11,vftabscale);
 748             vfitab           = _mm_cvttpd_epi32(rt);
 749             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 750             vfitab           = _mm_slli_epi32(vfitab,2);
 751
 752             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 753             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 754             F                = _mm_setzero_pd();
 755             GMX_MM_TRANSPOSE2_PD(Y,F);
 756             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 757             H                = _mm_setzero_pd();
 758             GMX_MM_TRANSPOSE2_PD(G,H);
 759             Heps             = _mm_mul_pd(vfeps,H);
 760             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 761             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 762             velec            = _mm_mul_pd(qq11,VV);
 763             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 764             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 765
 766             /* Update potential sum for this i atom from the interaction with this j atom. */
 767             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 768             velecsum         = _mm_add_pd(velecsum,velec);
 769
 770             fscal            = felec;
 771
 772             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 773
 774             /* Calculate temporary vectorial force */
 775             tx               = _mm_mul_pd(fscal,dx11);
 776             ty               = _mm_mul_pd(fscal,dy11);
 777             tz               = _mm_mul_pd(fscal,dz11);
 778
 779             /* Update vectorial force */
 780             fix1             = _mm_add_pd(fix1,tx);
 781             fiy1             = _mm_add_pd(fiy1,ty);
 782             fiz1             = _mm_add_pd(fiz1,tz);
 783
 784             fjx1             = _mm_add_pd(fjx1,tx);
 785             fjy1             = _mm_add_pd(fjy1,ty);
 786             fjz1             = _mm_add_pd(fjz1,tz);
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             r12              = _mm_mul_pd(rsq12,rinv12);
 793
 794             /* Calculate table index by multiplying r with table scale and truncate to integer */
 795             rt               = _mm_mul_pd(r12,vftabscale);
 796             vfitab           = _mm_cvttpd_epi32(rt);
 797             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 798             vfitab           = _mm_slli_epi32(vfitab,2);
 799
 800             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 801             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 802             F                = _mm_setzero_pd();
 803             GMX_MM_TRANSPOSE2_PD(Y,F);
 804             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 805             H                = _mm_setzero_pd();
 806             GMX_MM_TRANSPOSE2_PD(G,H);
 807             Heps             = _mm_mul_pd(vfeps,H);
 808             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 809             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 810             velec            = _mm_mul_pd(qq12,VV);
 811             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 812             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 813
 814             /* Update potential sum for this i atom from the interaction with this j atom. */
 815             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 816             velecsum         = _mm_add_pd(velecsum,velec);
 817
 818             fscal            = felec;
 819
 820             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 821
 822             /* Calculate temporary vectorial force */
 823             tx               = _mm_mul_pd(fscal,dx12);
 824             ty               = _mm_mul_pd(fscal,dy12);
 825             tz               = _mm_mul_pd(fscal,dz12);
 826
 827             /* Update vectorial force */
 828             fix1             = _mm_add_pd(fix1,tx);
 829             fiy1             = _mm_add_pd(fiy1,ty);
 830             fiz1             = _mm_add_pd(fiz1,tz);
 831
 832             fjx2             = _mm_add_pd(fjx2,tx);
 833             fjy2             = _mm_add_pd(fjy2,ty);
 834             fjz2             = _mm_add_pd(fjz2,tz);
 835
 836             /**************************
 837              * CALCULATE INTERACTIONS *
 838              **************************/
 839
 840             r13              = _mm_mul_pd(rsq13,rinv13);
 841
 842             /* Calculate table index by multiplying r with table scale and truncate to integer */
 843             rt               = _mm_mul_pd(r13,vftabscale);
 844             vfitab           = _mm_cvttpd_epi32(rt);
 845             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 846             vfitab           = _mm_slli_epi32(vfitab,2);
 847
 848             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 849             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 850             F                = _mm_setzero_pd();
 851             GMX_MM_TRANSPOSE2_PD(Y,F);
 852             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 853             H                = _mm_setzero_pd();
 854             GMX_MM_TRANSPOSE2_PD(G,H);
 855             Heps             = _mm_mul_pd(vfeps,H);
 856             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 857             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 858             velec            = _mm_mul_pd(qq13,VV);
 859             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 860             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
 861
 862             /* Update potential sum for this i atom from the interaction with this j atom. */
 863             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 864             velecsum         = _mm_add_pd(velecsum,velec);
 865
 866             fscal            = felec;
 867
 868             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 869
 870             /* Calculate temporary vectorial force */
 871             tx               = _mm_mul_pd(fscal,dx13);
 872             ty               = _mm_mul_pd(fscal,dy13);
 873             tz               = _mm_mul_pd(fscal,dz13);
 874
 875             /* Update vectorial force */
 876             fix1             = _mm_add_pd(fix1,tx);
 877             fiy1             = _mm_add_pd(fiy1,ty);
 878             fiz1             = _mm_add_pd(fiz1,tz);
 879
 880             fjx3             = _mm_add_pd(fjx3,tx);
 881             fjy3             = _mm_add_pd(fjy3,ty);
 882             fjz3             = _mm_add_pd(fjz3,tz);
 883
 884             /**************************
 885              * CALCULATE INTERACTIONS *
 886              **************************/
 887
 888             r21              = _mm_mul_pd(rsq21,rinv21);
 889
 890             /* Calculate table index by multiplying r with table scale and truncate to integer */
 891             rt               = _mm_mul_pd(r21,vftabscale);
 892             vfitab           = _mm_cvttpd_epi32(rt);
 893             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 894             vfitab           = _mm_slli_epi32(vfitab,2);
 895
 896             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 897             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 898             F                = _mm_setzero_pd();
 899             GMX_MM_TRANSPOSE2_PD(Y,F);
 900             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 901             H                = _mm_setzero_pd();
 902             GMX_MM_TRANSPOSE2_PD(G,H);
 903             Heps             = _mm_mul_pd(vfeps,H);
 904             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 905             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 906             velec            = _mm_mul_pd(qq21,VV);
 907             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 908             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 909
 910             /* Update potential sum for this i atom from the interaction with this j atom. */
 911             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 912             velecsum         = _mm_add_pd(velecsum,velec);
 913
 914             fscal            = felec;
 915
 916             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 917
 918             /* Calculate temporary vectorial force */
 919             tx               = _mm_mul_pd(fscal,dx21);
 920             ty               = _mm_mul_pd(fscal,dy21);
 921             tz               = _mm_mul_pd(fscal,dz21);
 922
 923             /* Update vectorial force */
 924             fix2             = _mm_add_pd(fix2,tx);
 925             fiy2             = _mm_add_pd(fiy2,ty);
 926             fiz2             = _mm_add_pd(fiz2,tz);
 927
 928             fjx1             = _mm_add_pd(fjx1,tx);
 929             fjy1             = _mm_add_pd(fjy1,ty);
 930             fjz1             = _mm_add_pd(fjz1,tz);
 931
 932             /**************************
 933              * CALCULATE INTERACTIONS *
 934              **************************/
 935
 936             r22              = _mm_mul_pd(rsq22,rinv22);
 937
 938             /* Calculate table index by multiplying r with table scale and truncate to integer */
 939             rt               = _mm_mul_pd(r22,vftabscale);
 940             vfitab           = _mm_cvttpd_epi32(rt);
 941             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 942             vfitab           = _mm_slli_epi32(vfitab,2);
 943
 944             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 945             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 946             F                = _mm_setzero_pd();
 947             GMX_MM_TRANSPOSE2_PD(Y,F);
 948             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 949             H                = _mm_setzero_pd();
 950             GMX_MM_TRANSPOSE2_PD(G,H);
 951             Heps             = _mm_mul_pd(vfeps,H);
 952             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 953             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 954             velec            = _mm_mul_pd(qq22,VV);
 955             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 956             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 957
 958             /* Update potential sum for this i atom from the interaction with this j atom. */
 959             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 960             velecsum         = _mm_add_pd(velecsum,velec);
 961
 962             fscal            = felec;
 963
 964             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 965
 966             /* Calculate temporary vectorial force */
 967             tx               = _mm_mul_pd(fscal,dx22);
 968             ty               = _mm_mul_pd(fscal,dy22);
 969             tz               = _mm_mul_pd(fscal,dz22);
 970
 971             /* Update vectorial force */
 972             fix2             = _mm_add_pd(fix2,tx);
 973             fiy2             = _mm_add_pd(fiy2,ty);
 974             fiz2             = _mm_add_pd(fiz2,tz);
 975
 976             fjx2             = _mm_add_pd(fjx2,tx);
 977             fjy2             = _mm_add_pd(fjy2,ty);
 978             fjz2             = _mm_add_pd(fjz2,tz);
 979
 980             /**************************
 981              * CALCULATE INTERACTIONS *
 982              **************************/
 983
 984             r23              = _mm_mul_pd(rsq23,rinv23);
 985
 986             /* Calculate table index by multiplying r with table scale and truncate to integer */
 987             rt               = _mm_mul_pd(r23,vftabscale);
 988             vfitab           = _mm_cvttpd_epi32(rt);
 989             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 990             vfitab           = _mm_slli_epi32(vfitab,2);
 991
 992             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 993             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 994             F                = _mm_setzero_pd();
 995             GMX_MM_TRANSPOSE2_PD(Y,F);
 996             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 997             H                = _mm_setzero_pd();
 998             GMX_MM_TRANSPOSE2_PD(G,H);
 999             Heps             = _mm_mul_pd(vfeps,H);
1000             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1001             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1002             velec            = _mm_mul_pd(qq23,VV);
1003             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1004             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1005
1006             /* Update potential sum for this i atom from the interaction with this j atom. */
1007             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1008             velecsum         = _mm_add_pd(velecsum,velec);
1009
1010             fscal            = felec;
1011
1012             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1013
1014             /* Calculate temporary vectorial force */
1015             tx               = _mm_mul_pd(fscal,dx23);
1016             ty               = _mm_mul_pd(fscal,dy23);
1017             tz               = _mm_mul_pd(fscal,dz23);
1018
1019             /* Update vectorial force */
1020             fix2             = _mm_add_pd(fix2,tx);
1021             fiy2             = _mm_add_pd(fiy2,ty);
1022             fiz2             = _mm_add_pd(fiz2,tz);
1023
1024             fjx3             = _mm_add_pd(fjx3,tx);
1025             fjy3             = _mm_add_pd(fjy3,ty);
1026             fjz3             = _mm_add_pd(fjz3,tz);
1027
1028             /**************************
1029              * CALCULATE INTERACTIONS *
1030              **************************/
1031
1032             r31              = _mm_mul_pd(rsq31,rinv31);
1033
1034             /* Calculate table index by multiplying r with table scale and truncate to integer */
1035             rt               = _mm_mul_pd(r31,vftabscale);
1036             vfitab           = _mm_cvttpd_epi32(rt);
1037             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1038             vfitab           = _mm_slli_epi32(vfitab,2);
1039
1040             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1041             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1042             F                = _mm_setzero_pd();
1043             GMX_MM_TRANSPOSE2_PD(Y,F);
1044             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1045             H                = _mm_setzero_pd();
1046             GMX_MM_TRANSPOSE2_PD(G,H);
1047             Heps             = _mm_mul_pd(vfeps,H);
1048             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1049             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1050             velec            = _mm_mul_pd(qq31,VV);
1051             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1052             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1053
1054             /* Update potential sum for this i atom from the interaction with this j atom. */
1055             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1056             velecsum         = _mm_add_pd(velecsum,velec);
1057
1058             fscal            = felec;
1059
1060             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1061
1062             /* Calculate temporary vectorial force */
1063             tx               = _mm_mul_pd(fscal,dx31);
1064             ty               = _mm_mul_pd(fscal,dy31);
1065             tz               = _mm_mul_pd(fscal,dz31);
1066
1067             /* Update vectorial force */
1068             fix3             = _mm_add_pd(fix3,tx);
1069             fiy3             = _mm_add_pd(fiy3,ty);
1070             fiz3             = _mm_add_pd(fiz3,tz);
1071
1072             fjx1             = _mm_add_pd(fjx1,tx);
1073             fjy1             = _mm_add_pd(fjy1,ty);
1074             fjz1             = _mm_add_pd(fjz1,tz);
1075
1076             /**************************
1077              * CALCULATE INTERACTIONS *
1078              **************************/
1079
1080             r32              = _mm_mul_pd(rsq32,rinv32);
1081
1082             /* Calculate table index by multiplying r with table scale and truncate to integer */
1083             rt               = _mm_mul_pd(r32,vftabscale);
1084             vfitab           = _mm_cvttpd_epi32(rt);
1085             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1086             vfitab           = _mm_slli_epi32(vfitab,2);
1087
1088             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1089             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1090             F                = _mm_setzero_pd();
1091             GMX_MM_TRANSPOSE2_PD(Y,F);
1092             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1093             H                = _mm_setzero_pd();
1094             GMX_MM_TRANSPOSE2_PD(G,H);
1095             Heps             = _mm_mul_pd(vfeps,H);
1096             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1097             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1098             velec            = _mm_mul_pd(qq32,VV);
1099             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1100             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1101
1102             /* Update potential sum for this i atom from the interaction with this j atom. */
1103             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1104             velecsum         = _mm_add_pd(velecsum,velec);
1105
1106             fscal            = felec;
1107
1108             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1109
1110             /* Calculate temporary vectorial force */
1111             tx               = _mm_mul_pd(fscal,dx32);
1112             ty               = _mm_mul_pd(fscal,dy32);
1113             tz               = _mm_mul_pd(fscal,dz32);
1114
1115             /* Update vectorial force */
1116             fix3             = _mm_add_pd(fix3,tx);
1117             fiy3             = _mm_add_pd(fiy3,ty);
1118             fiz3             = _mm_add_pd(fiz3,tz);
1119
1120             fjx2             = _mm_add_pd(fjx2,tx);
1121             fjy2             = _mm_add_pd(fjy2,ty);
1122             fjz2             = _mm_add_pd(fjz2,tz);
1123
1124             /**************************
1125              * CALCULATE INTERACTIONS *
1126              **************************/
1127
1128             r33              = _mm_mul_pd(rsq33,rinv33);
1129
1130             /* Calculate table index by multiplying r with table scale and truncate to integer */
1131             rt               = _mm_mul_pd(r33,vftabscale);
1132             vfitab           = _mm_cvttpd_epi32(rt);
1133             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1134             vfitab           = _mm_slli_epi32(vfitab,2);
1135
1136             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1137             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1138             F                = _mm_setzero_pd();
1139             GMX_MM_TRANSPOSE2_PD(Y,F);
1140             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1141             H                = _mm_setzero_pd();
1142             GMX_MM_TRANSPOSE2_PD(G,H);
1143             Heps             = _mm_mul_pd(vfeps,H);
1144             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1145             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
1146             velec            = _mm_mul_pd(qq33,VV);
1147             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1148             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1149
1150             /* Update potential sum for this i atom from the interaction with this j atom. */
1151             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1152             velecsum         = _mm_add_pd(velecsum,velec);
1153
1154             fscal            = felec;
1155
1156             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1157
1158             /* Calculate temporary vectorial force */
1159             tx               = _mm_mul_pd(fscal,dx33);
1160             ty               = _mm_mul_pd(fscal,dy33);
1161             tz               = _mm_mul_pd(fscal,dz33);
1162
1163             /* Update vectorial force */
1164             fix3             = _mm_add_pd(fix3,tx);
1165             fiy3             = _mm_add_pd(fiy3,ty);
1166             fiz3             = _mm_add_pd(fiz3,tz);
1167
1168             fjx3             = _mm_add_pd(fjx3,tx);
1169             fjy3             = _mm_add_pd(fjy3,ty);
1170             fjz3             = _mm_add_pd(fjz3,tz);
1171
1172             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1173
1174             /* Inner loop uses 387 flops */
1175         }
1176
1177         /* End of innermost loop */
1178
1179         gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1180                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
1181
1182         ggid                        = gid[iidx];
1183         /* Update potential energies */
1184         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1185
1186         /* Increment number of inner iterations */
1187         inneriter                  += j_index_end - j_index_start;
1188
1189         /* Outer loop uses 19 flops */
1190     }
1191
1192     /* Increment number of outer iterations */
1193     outeriter        += nri;
1194
1195     /* Update outer/inner flops */
1196
1197     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_VF,outeriter*19 + inneriter*387);
1198 }
1199 /*
1200  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwNone_GeomW4W4_F_sse4_1_double
1201  * Electrostatics interaction: CubicSplineTable
1202  * VdW interaction:            None
1203  * Geometry:                   Water4-Water4
1204  * Calculate force/pot:        Force
1205  */
1206 void
1207 nb_kernel_ElecCSTab_VdwNone_GeomW4W4_F_sse4_1_double
1208                     (t_nblist                    * gmx_restrict       nlist,
1209                      rvec                        * gmx_restrict          xx,
1210                      rvec                        * gmx_restrict          ff,
1211                      t_forcerec                  * gmx_restrict          fr,
1212                      t_mdatoms                   * gmx_restrict     mdatoms,
1213                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1214                      t_nrnb                      * gmx_restrict        nrnb)
1215 {
1216     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1217      * just 0 for non-waters.
1218      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1219      * jnr indices corresponding to data put in the four positions in the SIMD register.
1220      */
1221     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1222     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1223     int              jnrA,jnrB;
1224     int              j_coord_offsetA,j_coord_offsetB;
1225     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1226     real             rcutoff_scalar;
1227     real             *shiftvec,*fshift,*x,*f;
1228     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1229     int              vdwioffset1;
1230     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1231     int              vdwioffset2;
1232     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1233     int              vdwioffset3;
1234     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1235     int              vdwjidx1A,vdwjidx1B;
1236     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1237     int              vdwjidx2A,vdwjidx2B;
1238     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1239     int              vdwjidx3A,vdwjidx3B;
1240     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1241     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1242     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1243     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1244     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1245     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1246     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1247     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1248     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1249     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1250     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1251     real             *charge;
1252     __m128i          vfitab;
1253     __m128i          ifour       = _mm_set1_epi32(4);
1254     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1255     real             *vftab;
1256     __m128d          dummy_mask,cutoff_mask;
1257     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1258     __m128d          one     = _mm_set1_pd(1.0);
1259     __m128d          two     = _mm_set1_pd(2.0);
1260     x                = xx[0];
1261     f                = ff[0];
1262
1263     nri              = nlist->nri;
1264     iinr             = nlist->iinr;
1265     jindex           = nlist->jindex;
1266     jjnr             = nlist->jjnr;
1267     shiftidx         = nlist->shift;
1268     gid              = nlist->gid;
1269     shiftvec         = fr->shift_vec[0];
1270     fshift           = fr->fshift[0];
1271     facel            = _mm_set1_pd(fr->epsfac);
1272     charge           = mdatoms->chargeA;
1273
1274     vftab            = kernel_data->table_elec->data;
1275     vftabscale       = _mm_set1_pd(kernel_data->table_elec->scale);
1276
1277     /* Setup water-specific parameters */
1278     inr              = nlist->iinr[0];
1279     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1280     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1281     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1282
1283     jq1              = _mm_set1_pd(charge[inr+1]);
1284     jq2              = _mm_set1_pd(charge[inr+2]);
1285     jq3              = _mm_set1_pd(charge[inr+3]);
1286     qq11             = _mm_mul_pd(iq1,jq1);
1287     qq12             = _mm_mul_pd(iq1,jq2);
1288     qq13             = _mm_mul_pd(iq1,jq3);
1289     qq21             = _mm_mul_pd(iq2,jq1);
1290     qq22             = _mm_mul_pd(iq2,jq2);
1291     qq23             = _mm_mul_pd(iq2,jq3);
1292     qq31             = _mm_mul_pd(iq3,jq1);
1293     qq32             = _mm_mul_pd(iq3,jq2);
1294     qq33             = _mm_mul_pd(iq3,jq3);
1295
1296     /* Avoid stupid compiler warnings */
1297     jnrA = jnrB = 0;
1298     j_coord_offsetA = 0;
1299     j_coord_offsetB = 0;
1300
1301     outeriter        = 0;
1302     inneriter        = 0;
1303
1304     /* Start outer loop over neighborlists */
1305     for(iidx=0; iidx<nri; iidx++)
1306     {
1307         /* Load shift vector for this list */
1308         i_shift_offset   = DIM*shiftidx[iidx];
1309
1310         /* Load limits for loop over neighbors */
1311         j_index_start    = jindex[iidx];
1312         j_index_end      = jindex[iidx+1];
1313
1314         /* Get outer coordinate index */
1315         inr              = iinr[iidx];
1316         i_coord_offset   = DIM*inr;
1317
1318         /* Load i particle coords and add shift vector */
1319         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
1320                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1321
1322         fix1             = _mm_setzero_pd();
1323         fiy1             = _mm_setzero_pd();
1324         fiz1             = _mm_setzero_pd();
1325         fix2             = _mm_setzero_pd();
1326         fiy2             = _mm_setzero_pd();
1327         fiz2             = _mm_setzero_pd();
1328         fix3             = _mm_setzero_pd();
1329         fiy3             = _mm_setzero_pd();
1330         fiz3             = _mm_setzero_pd();
1331
1332         /* Start inner kernel loop */
1333         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1334         {
1335
1336             /* Get j neighbor index, and coordinate index */
1337             jnrA             = jjnr[jidx];
1338             jnrB             = jjnr[jidx+1];
1339             j_coord_offsetA  = DIM*jnrA;
1340             j_coord_offsetB  = DIM*jnrB;
1341
1342             /* load j atom coordinates */
1343             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA+DIM,x+j_coord_offsetB+DIM,
1344                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1345
1346             /* Calculate displacement vector */
1347             dx11             = _mm_sub_pd(ix1,jx1);
1348             dy11             = _mm_sub_pd(iy1,jy1);
1349             dz11             = _mm_sub_pd(iz1,jz1);
1350             dx12             = _mm_sub_pd(ix1,jx2);
1351             dy12             = _mm_sub_pd(iy1,jy2);
1352             dz12             = _mm_sub_pd(iz1,jz2);
1353             dx13             = _mm_sub_pd(ix1,jx3);
1354             dy13             = _mm_sub_pd(iy1,jy3);
1355             dz13             = _mm_sub_pd(iz1,jz3);
1356             dx21             = _mm_sub_pd(ix2,jx1);
1357             dy21             = _mm_sub_pd(iy2,jy1);
1358             dz21             = _mm_sub_pd(iz2,jz1);
1359             dx22             = _mm_sub_pd(ix2,jx2);
1360             dy22             = _mm_sub_pd(iy2,jy2);
1361             dz22             = _mm_sub_pd(iz2,jz2);
1362             dx23             = _mm_sub_pd(ix2,jx3);
1363             dy23             = _mm_sub_pd(iy2,jy3);
1364             dz23             = _mm_sub_pd(iz2,jz3);
1365             dx31             = _mm_sub_pd(ix3,jx1);
1366             dy31             = _mm_sub_pd(iy3,jy1);
1367             dz31             = _mm_sub_pd(iz3,jz1);
1368             dx32             = _mm_sub_pd(ix3,jx2);
1369             dy32             = _mm_sub_pd(iy3,jy2);
1370             dz32             = _mm_sub_pd(iz3,jz2);
1371             dx33             = _mm_sub_pd(ix3,jx3);
1372             dy33             = _mm_sub_pd(iy3,jy3);
1373             dz33             = _mm_sub_pd(iz3,jz3);
1374
1375             /* Calculate squared distance and things based on it */
1376             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1377             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1378             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1379             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1380             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1381             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1382             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1383             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1384             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1385
1386             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1387             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1388             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1389             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1390             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1391             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1392             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1393             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1394             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1395
1396             fjx1             = _mm_setzero_pd();
1397             fjy1             = _mm_setzero_pd();
1398             fjz1             = _mm_setzero_pd();
1399             fjx2             = _mm_setzero_pd();
1400             fjy2             = _mm_setzero_pd();
1401             fjz2             = _mm_setzero_pd();
1402             fjx3             = _mm_setzero_pd();
1403             fjy3             = _mm_setzero_pd();
1404             fjz3             = _mm_setzero_pd();
1405
1406             /**************************
1407              * CALCULATE INTERACTIONS *
1408              **************************/
1409
1410             r11              = _mm_mul_pd(rsq11,rinv11);
1411
1412             /* Calculate table index by multiplying r with table scale and truncate to integer */
1413             rt               = _mm_mul_pd(r11,vftabscale);
1414             vfitab           = _mm_cvttpd_epi32(rt);
1415             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1416             vfitab           = _mm_slli_epi32(vfitab,2);
1417
1418             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1419             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1420             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1421             GMX_MM_TRANSPOSE2_PD(Y,F);
1422             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1423             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1424             GMX_MM_TRANSPOSE2_PD(G,H);
1425             Heps             = _mm_mul_pd(vfeps,H);
1426             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1427             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1428             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1429
1430             fscal            = felec;
1431
1432             /* Calculate temporary vectorial force */
1433             tx               = _mm_mul_pd(fscal,dx11);
1434             ty               = _mm_mul_pd(fscal,dy11);
1435             tz               = _mm_mul_pd(fscal,dz11);
1436
1437             /* Update vectorial force */
1438             fix1             = _mm_add_pd(fix1,tx);
1439             fiy1             = _mm_add_pd(fiy1,ty);
1440             fiz1             = _mm_add_pd(fiz1,tz);
1441
1442             fjx1             = _mm_add_pd(fjx1,tx);
1443             fjy1             = _mm_add_pd(fjy1,ty);
1444             fjz1             = _mm_add_pd(fjz1,tz);
1445
1446             /**************************
1447              * CALCULATE INTERACTIONS *
1448              **************************/
1449
1450             r12              = _mm_mul_pd(rsq12,rinv12);
1451
1452             /* Calculate table index by multiplying r with table scale and truncate to integer */
1453             rt               = _mm_mul_pd(r12,vftabscale);
1454             vfitab           = _mm_cvttpd_epi32(rt);
1455             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1456             vfitab           = _mm_slli_epi32(vfitab,2);
1457
1458             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1459             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1460             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1461             GMX_MM_TRANSPOSE2_PD(Y,F);
1462             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1463             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1464             GMX_MM_TRANSPOSE2_PD(G,H);
1465             Heps             = _mm_mul_pd(vfeps,H);
1466             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1467             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1468             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1469
1470             fscal            = felec;
1471
1472             /* Calculate temporary vectorial force */
1473             tx               = _mm_mul_pd(fscal,dx12);
1474             ty               = _mm_mul_pd(fscal,dy12);
1475             tz               = _mm_mul_pd(fscal,dz12);
1476
1477             /* Update vectorial force */
1478             fix1             = _mm_add_pd(fix1,tx);
1479             fiy1             = _mm_add_pd(fiy1,ty);
1480             fiz1             = _mm_add_pd(fiz1,tz);
1481
1482             fjx2             = _mm_add_pd(fjx2,tx);
1483             fjy2             = _mm_add_pd(fjy2,ty);
1484             fjz2             = _mm_add_pd(fjz2,tz);
1485
1486             /**************************
1487              * CALCULATE INTERACTIONS *
1488              **************************/
1489
1490             r13              = _mm_mul_pd(rsq13,rinv13);
1491
1492             /* Calculate table index by multiplying r with table scale and truncate to integer */
1493             rt               = _mm_mul_pd(r13,vftabscale);
1494             vfitab           = _mm_cvttpd_epi32(rt);
1495             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1496             vfitab           = _mm_slli_epi32(vfitab,2);
1497
1498             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1499             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1500             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1501             GMX_MM_TRANSPOSE2_PD(Y,F);
1502             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1503             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1504             GMX_MM_TRANSPOSE2_PD(G,H);
1505             Heps             = _mm_mul_pd(vfeps,H);
1506             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1507             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1508             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1509
1510             fscal            = felec;
1511
1512             /* Calculate temporary vectorial force */
1513             tx               = _mm_mul_pd(fscal,dx13);
1514             ty               = _mm_mul_pd(fscal,dy13);
1515             tz               = _mm_mul_pd(fscal,dz13);
1516
1517             /* Update vectorial force */
1518             fix1             = _mm_add_pd(fix1,tx);
1519             fiy1             = _mm_add_pd(fiy1,ty);
1520             fiz1             = _mm_add_pd(fiz1,tz);
1521
1522             fjx3             = _mm_add_pd(fjx3,tx);
1523             fjy3             = _mm_add_pd(fjy3,ty);
1524             fjz3             = _mm_add_pd(fjz3,tz);
1525
1526             /**************************
1527              * CALCULATE INTERACTIONS *
1528              **************************/
1529
1530             r21              = _mm_mul_pd(rsq21,rinv21);
1531
1532             /* Calculate table index by multiplying r with table scale and truncate to integer */
1533             rt               = _mm_mul_pd(r21,vftabscale);
1534             vfitab           = _mm_cvttpd_epi32(rt);
1535             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1536             vfitab           = _mm_slli_epi32(vfitab,2);
1537
1538             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1539             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1540             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1541             GMX_MM_TRANSPOSE2_PD(Y,F);
1542             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1543             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1544             GMX_MM_TRANSPOSE2_PD(G,H);
1545             Heps             = _mm_mul_pd(vfeps,H);
1546             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1547             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1548             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1549
1550             fscal            = felec;
1551
1552             /* Calculate temporary vectorial force */
1553             tx               = _mm_mul_pd(fscal,dx21);
1554             ty               = _mm_mul_pd(fscal,dy21);
1555             tz               = _mm_mul_pd(fscal,dz21);
1556
1557             /* Update vectorial force */
1558             fix2             = _mm_add_pd(fix2,tx);
1559             fiy2             = _mm_add_pd(fiy2,ty);
1560             fiz2             = _mm_add_pd(fiz2,tz);
1561
1562             fjx1             = _mm_add_pd(fjx1,tx);
1563             fjy1             = _mm_add_pd(fjy1,ty);
1564             fjz1             = _mm_add_pd(fjz1,tz);
1565
1566             /**************************
1567              * CALCULATE INTERACTIONS *
1568              **************************/
1569
1570             r22              = _mm_mul_pd(rsq22,rinv22);
1571
1572             /* Calculate table index by multiplying r with table scale and truncate to integer */
1573             rt               = _mm_mul_pd(r22,vftabscale);
1574             vfitab           = _mm_cvttpd_epi32(rt);
1575             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1576             vfitab           = _mm_slli_epi32(vfitab,2);
1577
1578             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1579             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1580             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1581             GMX_MM_TRANSPOSE2_PD(Y,F);
1582             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1583             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1584             GMX_MM_TRANSPOSE2_PD(G,H);
1585             Heps             = _mm_mul_pd(vfeps,H);
1586             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1587             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1588             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1589
1590             fscal            = felec;
1591
1592             /* Calculate temporary vectorial force */
1593             tx               = _mm_mul_pd(fscal,dx22);
1594             ty               = _mm_mul_pd(fscal,dy22);
1595             tz               = _mm_mul_pd(fscal,dz22);
1596
1597             /* Update vectorial force */
1598             fix2             = _mm_add_pd(fix2,tx);
1599             fiy2             = _mm_add_pd(fiy2,ty);
1600             fiz2             = _mm_add_pd(fiz2,tz);
1601
1602             fjx2             = _mm_add_pd(fjx2,tx);
1603             fjy2             = _mm_add_pd(fjy2,ty);
1604             fjz2             = _mm_add_pd(fjz2,tz);
1605
1606             /**************************
1607              * CALCULATE INTERACTIONS *
1608              **************************/
1609
1610             r23              = _mm_mul_pd(rsq23,rinv23);
1611
1612             /* Calculate table index by multiplying r with table scale and truncate to integer */
1613             rt               = _mm_mul_pd(r23,vftabscale);
1614             vfitab           = _mm_cvttpd_epi32(rt);
1615             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1616             vfitab           = _mm_slli_epi32(vfitab,2);
1617
1618             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1619             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1620             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1621             GMX_MM_TRANSPOSE2_PD(Y,F);
1622             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1623             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1624             GMX_MM_TRANSPOSE2_PD(G,H);
1625             Heps             = _mm_mul_pd(vfeps,H);
1626             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1627             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1628             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1629
1630             fscal            = felec;
1631
1632             /* Calculate temporary vectorial force */
1633             tx               = _mm_mul_pd(fscal,dx23);
1634             ty               = _mm_mul_pd(fscal,dy23);
1635             tz               = _mm_mul_pd(fscal,dz23);
1636
1637             /* Update vectorial force */
1638             fix2             = _mm_add_pd(fix2,tx);
1639             fiy2             = _mm_add_pd(fiy2,ty);
1640             fiz2             = _mm_add_pd(fiz2,tz);
1641
1642             fjx3             = _mm_add_pd(fjx3,tx);
1643             fjy3             = _mm_add_pd(fjy3,ty);
1644             fjz3             = _mm_add_pd(fjz3,tz);
1645
1646             /**************************
1647              * CALCULATE INTERACTIONS *
1648              **************************/
1649
1650             r31              = _mm_mul_pd(rsq31,rinv31);
1651
1652             /* Calculate table index by multiplying r with table scale and truncate to integer */
1653             rt               = _mm_mul_pd(r31,vftabscale);
1654             vfitab           = _mm_cvttpd_epi32(rt);
1655             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1656             vfitab           = _mm_slli_epi32(vfitab,2);
1657
1658             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1659             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1660             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1661             GMX_MM_TRANSPOSE2_PD(Y,F);
1662             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1663             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1664             GMX_MM_TRANSPOSE2_PD(G,H);
1665             Heps             = _mm_mul_pd(vfeps,H);
1666             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1667             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1668             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1669
1670             fscal            = felec;
1671
1672             /* Calculate temporary vectorial force */
1673             tx               = _mm_mul_pd(fscal,dx31);
1674             ty               = _mm_mul_pd(fscal,dy31);
1675             tz               = _mm_mul_pd(fscal,dz31);
1676
1677             /* Update vectorial force */
1678             fix3             = _mm_add_pd(fix3,tx);
1679             fiy3             = _mm_add_pd(fiy3,ty);
1680             fiz3             = _mm_add_pd(fiz3,tz);
1681
1682             fjx1             = _mm_add_pd(fjx1,tx);
1683             fjy1             = _mm_add_pd(fjy1,ty);
1684             fjz1             = _mm_add_pd(fjz1,tz);
1685
1686             /**************************
1687              * CALCULATE INTERACTIONS *
1688              **************************/
1689
1690             r32              = _mm_mul_pd(rsq32,rinv32);
1691
1692             /* Calculate table index by multiplying r with table scale and truncate to integer */
1693             rt               = _mm_mul_pd(r32,vftabscale);
1694             vfitab           = _mm_cvttpd_epi32(rt);
1695             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1696             vfitab           = _mm_slli_epi32(vfitab,2);
1697
1698             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1699             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1700             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1701             GMX_MM_TRANSPOSE2_PD(Y,F);
1702             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1703             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1704             GMX_MM_TRANSPOSE2_PD(G,H);
1705             Heps             = _mm_mul_pd(vfeps,H);
1706             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1707             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1708             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1709
1710             fscal            = felec;
1711
1712             /* Calculate temporary vectorial force */
1713             tx               = _mm_mul_pd(fscal,dx32);
1714             ty               = _mm_mul_pd(fscal,dy32);
1715             tz               = _mm_mul_pd(fscal,dz32);
1716
1717             /* Update vectorial force */
1718             fix3             = _mm_add_pd(fix3,tx);
1719             fiy3             = _mm_add_pd(fiy3,ty);
1720             fiz3             = _mm_add_pd(fiz3,tz);
1721
1722             fjx2             = _mm_add_pd(fjx2,tx);
1723             fjy2             = _mm_add_pd(fjy2,ty);
1724             fjz2             = _mm_add_pd(fjz2,tz);
1725
1726             /**************************
1727              * CALCULATE INTERACTIONS *
1728              **************************/
1729
1730             r33              = _mm_mul_pd(rsq33,rinv33);
1731
1732             /* Calculate table index by multiplying r with table scale and truncate to integer */
1733             rt               = _mm_mul_pd(r33,vftabscale);
1734             vfitab           = _mm_cvttpd_epi32(rt);
1735             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1736             vfitab           = _mm_slli_epi32(vfitab,2);
1737
1738             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1739             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1740             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1741             GMX_MM_TRANSPOSE2_PD(Y,F);
1742             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1743             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1744             GMX_MM_TRANSPOSE2_PD(G,H);
1745             Heps             = _mm_mul_pd(vfeps,H);
1746             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1747             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1748             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1749
1750             fscal            = felec;
1751
1752             /* Calculate temporary vectorial force */
1753             tx               = _mm_mul_pd(fscal,dx33);
1754             ty               = _mm_mul_pd(fscal,dy33);
1755             tz               = _mm_mul_pd(fscal,dz33);
1756
1757             /* Update vectorial force */
1758             fix3             = _mm_add_pd(fix3,tx);
1759             fiy3             = _mm_add_pd(fiy3,ty);
1760             fiz3             = _mm_add_pd(fiz3,tz);
1761
1762             fjx3             = _mm_add_pd(fjx3,tx);
1763             fjy3             = _mm_add_pd(fjy3,ty);
1764             fjz3             = _mm_add_pd(fjz3,tz);
1765
1766             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA+DIM,f+j_coord_offsetB+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1767
1768             /* Inner loop uses 351 flops */
1769         }
1770
1771         if(jidx<j_index_end)
1772         {
1773
1774             jnrA             = jjnr[jidx];
1775             j_coord_offsetA  = DIM*jnrA;
1776
1777             /* load j atom coordinates */
1778             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA+DIM,
1779                                               &jx1,&jy1,&jz1,&jx2,&jy2,&jz2,&jx3,&jy3,&jz3);
1780
1781             /* Calculate displacement vector */
1782             dx11             = _mm_sub_pd(ix1,jx1);
1783             dy11             = _mm_sub_pd(iy1,jy1);
1784             dz11             = _mm_sub_pd(iz1,jz1);
1785             dx12             = _mm_sub_pd(ix1,jx2);
1786             dy12             = _mm_sub_pd(iy1,jy2);
1787             dz12             = _mm_sub_pd(iz1,jz2);
1788             dx13             = _mm_sub_pd(ix1,jx3);
1789             dy13             = _mm_sub_pd(iy1,jy3);
1790             dz13             = _mm_sub_pd(iz1,jz3);
1791             dx21             = _mm_sub_pd(ix2,jx1);
1792             dy21             = _mm_sub_pd(iy2,jy1);
1793             dz21             = _mm_sub_pd(iz2,jz1);
1794             dx22             = _mm_sub_pd(ix2,jx2);
1795             dy22             = _mm_sub_pd(iy2,jy2);
1796             dz22             = _mm_sub_pd(iz2,jz2);
1797             dx23             = _mm_sub_pd(ix2,jx3);
1798             dy23             = _mm_sub_pd(iy2,jy3);
1799             dz23             = _mm_sub_pd(iz2,jz3);
1800             dx31             = _mm_sub_pd(ix3,jx1);
1801             dy31             = _mm_sub_pd(iy3,jy1);
1802             dz31             = _mm_sub_pd(iz3,jz1);
1803             dx32             = _mm_sub_pd(ix3,jx2);
1804             dy32             = _mm_sub_pd(iy3,jy2);
1805             dz32             = _mm_sub_pd(iz3,jz2);
1806             dx33             = _mm_sub_pd(ix3,jx3);
1807             dy33             = _mm_sub_pd(iy3,jy3);
1808             dz33             = _mm_sub_pd(iz3,jz3);
1809
1810             /* Calculate squared distance and things based on it */
1811             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1812             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1813             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1814             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1815             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1816             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1817             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1818             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1819             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1820
1821             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1822             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1823             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1824             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1825             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1826             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1827             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1828             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1829             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1830
1831             fjx1             = _mm_setzero_pd();
1832             fjy1             = _mm_setzero_pd();
1833             fjz1             = _mm_setzero_pd();
1834             fjx2             = _mm_setzero_pd();
1835             fjy2             = _mm_setzero_pd();
1836             fjz2             = _mm_setzero_pd();
1837             fjx3             = _mm_setzero_pd();
1838             fjy3             = _mm_setzero_pd();
1839             fjz3             = _mm_setzero_pd();
1840
1841             /**************************
1842              * CALCULATE INTERACTIONS *
1843              **************************/
1844
1845             r11              = _mm_mul_pd(rsq11,rinv11);
1846
1847             /* Calculate table index by multiplying r with table scale and truncate to integer */
1848             rt               = _mm_mul_pd(r11,vftabscale);
1849             vfitab           = _mm_cvttpd_epi32(rt);
1850             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1851             vfitab           = _mm_slli_epi32(vfitab,2);
1852
1853             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1854             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1855             F                = _mm_setzero_pd();
1856             GMX_MM_TRANSPOSE2_PD(Y,F);
1857             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1858             H                = _mm_setzero_pd();
1859             GMX_MM_TRANSPOSE2_PD(G,H);
1860             Heps             = _mm_mul_pd(vfeps,H);
1861             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1862             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1863             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1864
1865             fscal            = felec;
1866
1867             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1868
1869             /* Calculate temporary vectorial force */
1870             tx               = _mm_mul_pd(fscal,dx11);
1871             ty               = _mm_mul_pd(fscal,dy11);
1872             tz               = _mm_mul_pd(fscal,dz11);
1873
1874             /* Update vectorial force */
1875             fix1             = _mm_add_pd(fix1,tx);
1876             fiy1             = _mm_add_pd(fiy1,ty);
1877             fiz1             = _mm_add_pd(fiz1,tz);
1878
1879             fjx1             = _mm_add_pd(fjx1,tx);
1880             fjy1             = _mm_add_pd(fjy1,ty);
1881             fjz1             = _mm_add_pd(fjz1,tz);
1882
1883             /**************************
1884              * CALCULATE INTERACTIONS *
1885              **************************/
1886
1887             r12              = _mm_mul_pd(rsq12,rinv12);
1888
1889             /* Calculate table index by multiplying r with table scale and truncate to integer */
1890             rt               = _mm_mul_pd(r12,vftabscale);
1891             vfitab           = _mm_cvttpd_epi32(rt);
1892             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1893             vfitab           = _mm_slli_epi32(vfitab,2);
1894
1895             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1896             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1897             F                = _mm_setzero_pd();
1898             GMX_MM_TRANSPOSE2_PD(Y,F);
1899             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1900             H                = _mm_setzero_pd();
1901             GMX_MM_TRANSPOSE2_PD(G,H);
1902             Heps             = _mm_mul_pd(vfeps,H);
1903             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1904             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1905             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1906
1907             fscal            = felec;
1908
1909             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1910
1911             /* Calculate temporary vectorial force */
1912             tx               = _mm_mul_pd(fscal,dx12);
1913             ty               = _mm_mul_pd(fscal,dy12);
1914             tz               = _mm_mul_pd(fscal,dz12);
1915
1916             /* Update vectorial force */
1917             fix1             = _mm_add_pd(fix1,tx);
1918             fiy1             = _mm_add_pd(fiy1,ty);
1919             fiz1             = _mm_add_pd(fiz1,tz);
1920
1921             fjx2             = _mm_add_pd(fjx2,tx);
1922             fjy2             = _mm_add_pd(fjy2,ty);
1923             fjz2             = _mm_add_pd(fjz2,tz);
1924
1925             /**************************
1926              * CALCULATE INTERACTIONS *
1927              **************************/
1928
1929             r13              = _mm_mul_pd(rsq13,rinv13);
1930
1931             /* Calculate table index by multiplying r with table scale and truncate to integer */
1932             rt               = _mm_mul_pd(r13,vftabscale);
1933             vfitab           = _mm_cvttpd_epi32(rt);
1934             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1935             vfitab           = _mm_slli_epi32(vfitab,2);
1936
1937             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1938             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1939             F                = _mm_setzero_pd();
1940             GMX_MM_TRANSPOSE2_PD(Y,F);
1941             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1942             H                = _mm_setzero_pd();
1943             GMX_MM_TRANSPOSE2_PD(G,H);
1944             Heps             = _mm_mul_pd(vfeps,H);
1945             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1946             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1947             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1948
1949             fscal            = felec;
1950
1951             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1952
1953             /* Calculate temporary vectorial force */
1954             tx               = _mm_mul_pd(fscal,dx13);
1955             ty               = _mm_mul_pd(fscal,dy13);
1956             tz               = _mm_mul_pd(fscal,dz13);
1957
1958             /* Update vectorial force */
1959             fix1             = _mm_add_pd(fix1,tx);
1960             fiy1             = _mm_add_pd(fiy1,ty);
1961             fiz1             = _mm_add_pd(fiz1,tz);
1962
1963             fjx3             = _mm_add_pd(fjx3,tx);
1964             fjy3             = _mm_add_pd(fjy3,ty);
1965             fjz3             = _mm_add_pd(fjz3,tz);
1966
1967             /**************************
1968              * CALCULATE INTERACTIONS *
1969              **************************/
1970
1971             r21              = _mm_mul_pd(rsq21,rinv21);
1972
1973             /* Calculate table index by multiplying r with table scale and truncate to integer */
1974             rt               = _mm_mul_pd(r21,vftabscale);
1975             vfitab           = _mm_cvttpd_epi32(rt);
1976             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1977             vfitab           = _mm_slli_epi32(vfitab,2);
1978
1979             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1980             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1981             F                = _mm_setzero_pd();
1982             GMX_MM_TRANSPOSE2_PD(Y,F);
1983             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1984             H                = _mm_setzero_pd();
1985             GMX_MM_TRANSPOSE2_PD(G,H);
1986             Heps             = _mm_mul_pd(vfeps,H);
1987             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1988             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1989             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1990
1991             fscal            = felec;
1992
1993             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1994
1995             /* Calculate temporary vectorial force */
1996             tx               = _mm_mul_pd(fscal,dx21);
1997             ty               = _mm_mul_pd(fscal,dy21);
1998             tz               = _mm_mul_pd(fscal,dz21);
1999
2000             /* Update vectorial force */
2001             fix2             = _mm_add_pd(fix2,tx);
2002             fiy2             = _mm_add_pd(fiy2,ty);
2003             fiz2             = _mm_add_pd(fiz2,tz);
2004
2005             fjx1             = _mm_add_pd(fjx1,tx);
2006             fjy1             = _mm_add_pd(fjy1,ty);
2007             fjz1             = _mm_add_pd(fjz1,tz);
2008
2009             /**************************
2010              * CALCULATE INTERACTIONS *
2011              **************************/
2012
2013             r22              = _mm_mul_pd(rsq22,rinv22);
2014
2015             /* Calculate table index by multiplying r with table scale and truncate to integer */
2016             rt               = _mm_mul_pd(r22,vftabscale);
2017             vfitab           = _mm_cvttpd_epi32(rt);
2018             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2019             vfitab           = _mm_slli_epi32(vfitab,2);
2020
2021             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2022             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2023             F                = _mm_setzero_pd();
2024             GMX_MM_TRANSPOSE2_PD(Y,F);
2025             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2026             H                = _mm_setzero_pd();
2027             GMX_MM_TRANSPOSE2_PD(G,H);
2028             Heps             = _mm_mul_pd(vfeps,H);
2029             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2030             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2031             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2032
2033             fscal            = felec;
2034
2035             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2036
2037             /* Calculate temporary vectorial force */
2038             tx               = _mm_mul_pd(fscal,dx22);
2039             ty               = _mm_mul_pd(fscal,dy22);
2040             tz               = _mm_mul_pd(fscal,dz22);
2041
2042             /* Update vectorial force */
2043             fix2             = _mm_add_pd(fix2,tx);
2044             fiy2             = _mm_add_pd(fiy2,ty);
2045             fiz2             = _mm_add_pd(fiz2,tz);
2046
2047             fjx2             = _mm_add_pd(fjx2,tx);
2048             fjy2             = _mm_add_pd(fjy2,ty);
2049             fjz2             = _mm_add_pd(fjz2,tz);
2050
2051             /**************************
2052              * CALCULATE INTERACTIONS *
2053              **************************/
2054
2055             r23              = _mm_mul_pd(rsq23,rinv23);
2056
2057             /* Calculate table index by multiplying r with table scale and truncate to integer */
2058             rt               = _mm_mul_pd(r23,vftabscale);
2059             vfitab           = _mm_cvttpd_epi32(rt);
2060             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2061             vfitab           = _mm_slli_epi32(vfitab,2);
2062
2063             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2064             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2065             F                = _mm_setzero_pd();
2066             GMX_MM_TRANSPOSE2_PD(Y,F);
2067             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2068             H                = _mm_setzero_pd();
2069             GMX_MM_TRANSPOSE2_PD(G,H);
2070             Heps             = _mm_mul_pd(vfeps,H);
2071             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2072             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2073             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
2074
2075             fscal            = felec;
2076
2077             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2078
2079             /* Calculate temporary vectorial force */
2080             tx               = _mm_mul_pd(fscal,dx23);
2081             ty               = _mm_mul_pd(fscal,dy23);
2082             tz               = _mm_mul_pd(fscal,dz23);
2083
2084             /* Update vectorial force */
2085             fix2             = _mm_add_pd(fix2,tx);
2086             fiy2             = _mm_add_pd(fiy2,ty);
2087             fiz2             = _mm_add_pd(fiz2,tz);
2088
2089             fjx3             = _mm_add_pd(fjx3,tx);
2090             fjy3             = _mm_add_pd(fjy3,ty);
2091             fjz3             = _mm_add_pd(fjz3,tz);
2092
2093             /**************************
2094              * CALCULATE INTERACTIONS *
2095              **************************/
2096
2097             r31              = _mm_mul_pd(rsq31,rinv31);
2098
2099             /* Calculate table index by multiplying r with table scale and truncate to integer */
2100             rt               = _mm_mul_pd(r31,vftabscale);
2101             vfitab           = _mm_cvttpd_epi32(rt);
2102             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2103             vfitab           = _mm_slli_epi32(vfitab,2);
2104
2105             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2106             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2107             F                = _mm_setzero_pd();
2108             GMX_MM_TRANSPOSE2_PD(Y,F);
2109             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2110             H                = _mm_setzero_pd();
2111             GMX_MM_TRANSPOSE2_PD(G,H);
2112             Heps             = _mm_mul_pd(vfeps,H);
2113             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2114             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2115             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
2116
2117             fscal            = felec;
2118
2119             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2120
2121             /* Calculate temporary vectorial force */
2122             tx               = _mm_mul_pd(fscal,dx31);
2123             ty               = _mm_mul_pd(fscal,dy31);
2124             tz               = _mm_mul_pd(fscal,dz31);
2125
2126             /* Update vectorial force */
2127             fix3             = _mm_add_pd(fix3,tx);
2128             fiy3             = _mm_add_pd(fiy3,ty);
2129             fiz3             = _mm_add_pd(fiz3,tz);
2130
2131             fjx1             = _mm_add_pd(fjx1,tx);
2132             fjy1             = _mm_add_pd(fjy1,ty);
2133             fjz1             = _mm_add_pd(fjz1,tz);
2134
2135             /**************************
2136              * CALCULATE INTERACTIONS *
2137              **************************/
2138
2139             r32              = _mm_mul_pd(rsq32,rinv32);
2140
2141             /* Calculate table index by multiplying r with table scale and truncate to integer */
2142             rt               = _mm_mul_pd(r32,vftabscale);
2143             vfitab           = _mm_cvttpd_epi32(rt);
2144             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2145             vfitab           = _mm_slli_epi32(vfitab,2);
2146
2147             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2148             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2149             F                = _mm_setzero_pd();
2150             GMX_MM_TRANSPOSE2_PD(Y,F);
2151             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2152             H                = _mm_setzero_pd();
2153             GMX_MM_TRANSPOSE2_PD(G,H);
2154             Heps             = _mm_mul_pd(vfeps,H);
2155             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2156             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2157             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
2158
2159             fscal            = felec;
2160
2161             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2162
2163             /* Calculate temporary vectorial force */
2164             tx               = _mm_mul_pd(fscal,dx32);
2165             ty               = _mm_mul_pd(fscal,dy32);
2166             tz               = _mm_mul_pd(fscal,dz32);
2167
2168             /* Update vectorial force */
2169             fix3             = _mm_add_pd(fix3,tx);
2170             fiy3             = _mm_add_pd(fiy3,ty);
2171             fiz3             = _mm_add_pd(fiz3,tz);
2172
2173             fjx2             = _mm_add_pd(fjx2,tx);
2174             fjy2             = _mm_add_pd(fjy2,ty);
2175             fjz2             = _mm_add_pd(fjz2,tz);
2176
2177             /**************************
2178              * CALCULATE INTERACTIONS *
2179              **************************/
2180
2181             r33              = _mm_mul_pd(rsq33,rinv33);
2182
2183             /* Calculate table index by multiplying r with table scale and truncate to integer */
2184             rt               = _mm_mul_pd(r33,vftabscale);
2185             vfitab           = _mm_cvttpd_epi32(rt);
2186             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2187             vfitab           = _mm_slli_epi32(vfitab,2);
2188
2189             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2190             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
2191             F                = _mm_setzero_pd();
2192             GMX_MM_TRANSPOSE2_PD(Y,F);
2193             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
2194             H                = _mm_setzero_pd();
2195             GMX_MM_TRANSPOSE2_PD(G,H);
2196             Heps             = _mm_mul_pd(vfeps,H);
2197             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
2198             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
2199             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2200
2201             fscal            = felec;
2202
2203             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2204
2205             /* Calculate temporary vectorial force */
2206             tx               = _mm_mul_pd(fscal,dx33);
2207             ty               = _mm_mul_pd(fscal,dy33);
2208             tz               = _mm_mul_pd(fscal,dz33);
2209
2210             /* Update vectorial force */
2211             fix3             = _mm_add_pd(fix3,tx);
2212             fiy3             = _mm_add_pd(fiy3,ty);
2213             fiz3             = _mm_add_pd(fiz3,tz);
2214
2215             fjx3             = _mm_add_pd(fjx3,tx);
2216             fjy3             = _mm_add_pd(fjy3,ty);
2217             fjz3             = _mm_add_pd(fjz3,tz);
2218
2219             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA+DIM,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2220
2221             /* Inner loop uses 351 flops */
2222         }
2223
2224         /* End of innermost loop */
2225
2226         gmx_mm_update_iforce_3atom_swizzle_pd(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2227                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
2228
2229         /* Increment number of inner iterations */
2230         inneriter                  += j_index_end - j_index_start;
2231
2232         /* Outer loop uses 18 flops */
2233     }
2234
2235     /* Increment number of outer iterations */
2236     outeriter        += nri;
2237
2238     /* Update outer/inner flops */
2239
2240     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4W4_F,outeriter*18 + inneriter*351);
2241 }