src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecEw_VdwCSTab_GeomW3W3_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_ElecEw_VdwCSTab_GeomW3W3_VF_sse4_1_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water3-Water3
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_sse4_1_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwjidx0A,vdwjidx0B;
  87     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  88     int              vdwjidx1A,vdwjidx1B;
  89     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  90     int              vdwjidx2A,vdwjidx2B;
  91     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  92     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  93     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  94     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  95     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  96     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  97     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  98     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  99     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 100     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 101     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 102     real             *charge;
 103     int              nvdwtype;
 104     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 105     int              *vdwtype;
 106     real             *vdwparam;
 107     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 108     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 109     __m128i          vfitab;
 110     __m128i          ifour       = _mm_set1_epi32(4);
 111     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 112     real             *vftab;
 113     __m128i          ewitab;
 114     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 115     real             *ewtab;
 116     __m128d          dummy_mask,cutoff_mask;
 117     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 118     __m128d          one     = _mm_set1_pd(1.0);
 119     __m128d          two     = _mm_set1_pd(2.0);
 120     x                = xx[0];
 121     f                = ff[0];
 122
 123     nri              = nlist->nri;
 124     iinr             = nlist->iinr;
 125     jindex           = nlist->jindex;
 126     jjnr             = nlist->jjnr;
 127     shiftidx         = nlist->shift;
 128     gid              = nlist->gid;
 129     shiftvec         = fr->shift_vec[0];
 130     fshift           = fr->fshift[0];
 131     facel            = _mm_set1_pd(fr->epsfac);
 132     charge           = mdatoms->chargeA;
 133     nvdwtype         = fr->ntype;
 134     vdwparam         = fr->nbfp;
 135     vdwtype          = mdatoms->typeA;
 136
 137     vftab            = kernel_data->table_vdw->data;
 138     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 139
 140     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 141     ewtab            = fr->ic->tabq_coul_FDV0;
 142     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 143     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 144
 145     /* Setup water-specific parameters */
 146     inr              = nlist->iinr[0];
 147     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 148     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 149     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 150     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 151
 152     jq0              = _mm_set1_pd(charge[inr+0]);
 153     jq1              = _mm_set1_pd(charge[inr+1]);
 154     jq2              = _mm_set1_pd(charge[inr+2]);
 155     vdwjidx0A        = 2*vdwtype[inr+0];
 156     qq00             = _mm_mul_pd(iq0,jq0);
 157     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 158     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 159     qq01             = _mm_mul_pd(iq0,jq1);
 160     qq02             = _mm_mul_pd(iq0,jq2);
 161     qq10             = _mm_mul_pd(iq1,jq0);
 162     qq11             = _mm_mul_pd(iq1,jq1);
 163     qq12             = _mm_mul_pd(iq1,jq2);
 164     qq20             = _mm_mul_pd(iq2,jq0);
 165     qq21             = _mm_mul_pd(iq2,jq1);
 166     qq22             = _mm_mul_pd(iq2,jq2);
 167
 168     /* Avoid stupid compiler warnings */
 169     jnrA = jnrB = 0;
 170     j_coord_offsetA = 0;
 171     j_coord_offsetB = 0;
 172
 173     outeriter        = 0;
 174     inneriter        = 0;
 175
 176     /* Start outer loop over neighborlists */
 177     for(iidx=0; iidx<nri; iidx++)
 178     {
 179         /* Load shift vector for this list */
 180         i_shift_offset   = DIM*shiftidx[iidx];
 181
 182         /* Load limits for loop over neighbors */
 183         j_index_start    = jindex[iidx];
 184         j_index_end      = jindex[iidx+1];
 185
 186         /* Get outer coordinate index */
 187         inr              = iinr[iidx];
 188         i_coord_offset   = DIM*inr;
 189
 190         /* Load i particle coords and add shift vector */
 191         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 192                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 193
 194         fix0             = _mm_setzero_pd();
 195         fiy0             = _mm_setzero_pd();
 196         fiz0             = _mm_setzero_pd();
 197         fix1             = _mm_setzero_pd();
 198         fiy1             = _mm_setzero_pd();
 199         fiz1             = _mm_setzero_pd();
 200         fix2             = _mm_setzero_pd();
 201         fiy2             = _mm_setzero_pd();
 202         fiz2             = _mm_setzero_pd();
 203
 204         /* Reset potential sums */
 205         velecsum         = _mm_setzero_pd();
 206         vvdwsum          = _mm_setzero_pd();
 207
 208         /* Start inner kernel loop */
 209         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 210         {
 211
 212             /* Get j neighbor index, and coordinate index */
 213             jnrA             = jjnr[jidx];
 214             jnrB             = jjnr[jidx+1];
 215             j_coord_offsetA  = DIM*jnrA;
 216             j_coord_offsetB  = DIM*jnrB;
 217
 218             /* load j atom coordinates */
 219             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 220                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 221
 222             /* Calculate displacement vector */
 223             dx00             = _mm_sub_pd(ix0,jx0);
 224             dy00             = _mm_sub_pd(iy0,jy0);
 225             dz00             = _mm_sub_pd(iz0,jz0);
 226             dx01             = _mm_sub_pd(ix0,jx1);
 227             dy01             = _mm_sub_pd(iy0,jy1);
 228             dz01             = _mm_sub_pd(iz0,jz1);
 229             dx02             = _mm_sub_pd(ix0,jx2);
 230             dy02             = _mm_sub_pd(iy0,jy2);
 231             dz02             = _mm_sub_pd(iz0,jz2);
 232             dx10             = _mm_sub_pd(ix1,jx0);
 233             dy10             = _mm_sub_pd(iy1,jy0);
 234             dz10             = _mm_sub_pd(iz1,jz0);
 235             dx11             = _mm_sub_pd(ix1,jx1);
 236             dy11             = _mm_sub_pd(iy1,jy1);
 237             dz11             = _mm_sub_pd(iz1,jz1);
 238             dx12             = _mm_sub_pd(ix1,jx2);
 239             dy12             = _mm_sub_pd(iy1,jy2);
 240             dz12             = _mm_sub_pd(iz1,jz2);
 241             dx20             = _mm_sub_pd(ix2,jx0);
 242             dy20             = _mm_sub_pd(iy2,jy0);
 243             dz20             = _mm_sub_pd(iz2,jz0);
 244             dx21             = _mm_sub_pd(ix2,jx1);
 245             dy21             = _mm_sub_pd(iy2,jy1);
 246             dz21             = _mm_sub_pd(iz2,jz1);
 247             dx22             = _mm_sub_pd(ix2,jx2);
 248             dy22             = _mm_sub_pd(iy2,jy2);
 249             dz22             = _mm_sub_pd(iz2,jz2);
 250
 251             /* Calculate squared distance and things based on it */
 252             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 253             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 254             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 255             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 256             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 257             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 258             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 259             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 260             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 261
 262             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 263             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 264             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 265             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 266             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 267             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 268             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 269             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 270             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 271
 272             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 273             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 274             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 275             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 276             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 277             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 278             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 279             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 280             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 281
 282             fjx0             = _mm_setzero_pd();
 283             fjy0             = _mm_setzero_pd();
 284             fjz0             = _mm_setzero_pd();
 285             fjx1             = _mm_setzero_pd();
 286             fjy1             = _mm_setzero_pd();
 287             fjz1             = _mm_setzero_pd();
 288             fjx2             = _mm_setzero_pd();
 289             fjy2             = _mm_setzero_pd();
 290             fjz2             = _mm_setzero_pd();
 291
 292             /**************************
 293              * CALCULATE INTERACTIONS *
 294              **************************/
 295
 296             r00              = _mm_mul_pd(rsq00,rinv00);
 297
 298             /* Calculate table index by multiplying r with table scale and truncate to integer */
 299             rt               = _mm_mul_pd(r00,vftabscale);
 300             vfitab           = _mm_cvttpd_epi32(rt);
 301             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 302             vfitab           = _mm_slli_epi32(vfitab,3);
 303
 304             /* EWALD ELECTROSTATICS */
 305
 306             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 307             ewrt             = _mm_mul_pd(r00,ewtabscale);
 308             ewitab           = _mm_cvttpd_epi32(ewrt);
 309             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 310             ewitab           = _mm_slli_epi32(ewitab,2);
 311             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 312             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 313             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 314             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 315             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 316             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 317             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 318             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 319             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 320             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 321
 322             /* CUBIC SPLINE TABLE DISPERSION */
 323             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 324             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 325             GMX_MM_TRANSPOSE2_PD(Y,F);
 326             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 327             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 328             GMX_MM_TRANSPOSE2_PD(G,H);
 329             Heps             = _mm_mul_pd(vfeps,H);
 330             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 331             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 332             vvdw6            = _mm_mul_pd(c6_00,VV);
 333             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 334             fvdw6            = _mm_mul_pd(c6_00,FF);
 335
 336             /* CUBIC SPLINE TABLE REPULSION */
 337             vfitab           = _mm_add_epi32(vfitab,ifour);
 338             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 339             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 340             GMX_MM_TRANSPOSE2_PD(Y,F);
 341             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 342             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 343             GMX_MM_TRANSPOSE2_PD(G,H);
 344             Heps             = _mm_mul_pd(vfeps,H);
 345             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 346             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 347             vvdw12           = _mm_mul_pd(c12_00,VV);
 348             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 349             fvdw12           = _mm_mul_pd(c12_00,FF);
 350             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 351             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 352
 353             /* Update potential sum for this i atom from the interaction with this j atom. */
 354             velecsum         = _mm_add_pd(velecsum,velec);
 355             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 356
 357             fscal            = _mm_add_pd(felec,fvdw);
 358
 359             /* Calculate temporary vectorial force */
 360             tx               = _mm_mul_pd(fscal,dx00);
 361             ty               = _mm_mul_pd(fscal,dy00);
 362             tz               = _mm_mul_pd(fscal,dz00);
 363
 364             /* Update vectorial force */
 365             fix0             = _mm_add_pd(fix0,tx);
 366             fiy0             = _mm_add_pd(fiy0,ty);
 367             fiz0             = _mm_add_pd(fiz0,tz);
 368
 369             fjx0             = _mm_add_pd(fjx0,tx);
 370             fjy0             = _mm_add_pd(fjy0,ty);
 371             fjz0             = _mm_add_pd(fjz0,tz);
 372
 373             /**************************
 374              * CALCULATE INTERACTIONS *
 375              **************************/
 376
 377             r01              = _mm_mul_pd(rsq01,rinv01);
 378
 379             /* EWALD ELECTROSTATICS */
 380
 381             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 382             ewrt             = _mm_mul_pd(r01,ewtabscale);
 383             ewitab           = _mm_cvttpd_epi32(ewrt);
 384             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 385             ewitab           = _mm_slli_epi32(ewitab,2);
 386             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 387             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 388             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 389             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 390             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 391             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 392             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 393             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 394             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 395             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 396
 397             /* Update potential sum for this i atom from the interaction with this j atom. */
 398             velecsum         = _mm_add_pd(velecsum,velec);
 399
 400             fscal            = felec;
 401
 402             /* Calculate temporary vectorial force */
 403             tx               = _mm_mul_pd(fscal,dx01);
 404             ty               = _mm_mul_pd(fscal,dy01);
 405             tz               = _mm_mul_pd(fscal,dz01);
 406
 407             /* Update vectorial force */
 408             fix0             = _mm_add_pd(fix0,tx);
 409             fiy0             = _mm_add_pd(fiy0,ty);
 410             fiz0             = _mm_add_pd(fiz0,tz);
 411
 412             fjx1             = _mm_add_pd(fjx1,tx);
 413             fjy1             = _mm_add_pd(fjy1,ty);
 414             fjz1             = _mm_add_pd(fjz1,tz);
 415
 416             /**************************
 417              * CALCULATE INTERACTIONS *
 418              **************************/
 419
 420             r02              = _mm_mul_pd(rsq02,rinv02);
 421
 422             /* EWALD ELECTROSTATICS */
 423
 424             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 425             ewrt             = _mm_mul_pd(r02,ewtabscale);
 426             ewitab           = _mm_cvttpd_epi32(ewrt);
 427             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 428             ewitab           = _mm_slli_epi32(ewitab,2);
 429             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 430             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 431             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 432             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 433             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 434             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 435             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 436             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 437             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 438             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 439
 440             /* Update potential sum for this i atom from the interaction with this j atom. */
 441             velecsum         = _mm_add_pd(velecsum,velec);
 442
 443             fscal            = felec;
 444
 445             /* Calculate temporary vectorial force */
 446             tx               = _mm_mul_pd(fscal,dx02);
 447             ty               = _mm_mul_pd(fscal,dy02);
 448             tz               = _mm_mul_pd(fscal,dz02);
 449
 450             /* Update vectorial force */
 451             fix0             = _mm_add_pd(fix0,tx);
 452             fiy0             = _mm_add_pd(fiy0,ty);
 453             fiz0             = _mm_add_pd(fiz0,tz);
 454
 455             fjx2             = _mm_add_pd(fjx2,tx);
 456             fjy2             = _mm_add_pd(fjy2,ty);
 457             fjz2             = _mm_add_pd(fjz2,tz);
 458
 459             /**************************
 460              * CALCULATE INTERACTIONS *
 461              **************************/
 462
 463             r10              = _mm_mul_pd(rsq10,rinv10);
 464
 465             /* EWALD ELECTROSTATICS */
 466
 467             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 468             ewrt             = _mm_mul_pd(r10,ewtabscale);
 469             ewitab           = _mm_cvttpd_epi32(ewrt);
 470             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 471             ewitab           = _mm_slli_epi32(ewitab,2);
 472             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 473             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 474             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 475             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 476             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 477             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 478             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 479             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 480             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 481             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 482
 483             /* Update potential sum for this i atom from the interaction with this j atom. */
 484             velecsum         = _mm_add_pd(velecsum,velec);
 485
 486             fscal            = felec;
 487
 488             /* Calculate temporary vectorial force */
 489             tx               = _mm_mul_pd(fscal,dx10);
 490             ty               = _mm_mul_pd(fscal,dy10);
 491             tz               = _mm_mul_pd(fscal,dz10);
 492
 493             /* Update vectorial force */
 494             fix1             = _mm_add_pd(fix1,tx);
 495             fiy1             = _mm_add_pd(fiy1,ty);
 496             fiz1             = _mm_add_pd(fiz1,tz);
 497
 498             fjx0             = _mm_add_pd(fjx0,tx);
 499             fjy0             = _mm_add_pd(fjy0,ty);
 500             fjz0             = _mm_add_pd(fjz0,tz);
 501
 502             /**************************
 503              * CALCULATE INTERACTIONS *
 504              **************************/
 505
 506             r11              = _mm_mul_pd(rsq11,rinv11);
 507
 508             /* EWALD ELECTROSTATICS */
 509
 510             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 511             ewrt             = _mm_mul_pd(r11,ewtabscale);
 512             ewitab           = _mm_cvttpd_epi32(ewrt);
 513             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 514             ewitab           = _mm_slli_epi32(ewitab,2);
 515             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 516             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 517             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 518             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 519             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 520             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 521             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 522             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 523             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 524             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 525
 526             /* Update potential sum for this i atom from the interaction with this j atom. */
 527             velecsum         = _mm_add_pd(velecsum,velec);
 528
 529             fscal            = felec;
 530
 531             /* Calculate temporary vectorial force */
 532             tx               = _mm_mul_pd(fscal,dx11);
 533             ty               = _mm_mul_pd(fscal,dy11);
 534             tz               = _mm_mul_pd(fscal,dz11);
 535
 536             /* Update vectorial force */
 537             fix1             = _mm_add_pd(fix1,tx);
 538             fiy1             = _mm_add_pd(fiy1,ty);
 539             fiz1             = _mm_add_pd(fiz1,tz);
 540
 541             fjx1             = _mm_add_pd(fjx1,tx);
 542             fjy1             = _mm_add_pd(fjy1,ty);
 543             fjz1             = _mm_add_pd(fjz1,tz);
 544
 545             /**************************
 546              * CALCULATE INTERACTIONS *
 547              **************************/
 548
 549             r12              = _mm_mul_pd(rsq12,rinv12);
 550
 551             /* EWALD ELECTROSTATICS */
 552
 553             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 554             ewrt             = _mm_mul_pd(r12,ewtabscale);
 555             ewitab           = _mm_cvttpd_epi32(ewrt);
 556             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 557             ewitab           = _mm_slli_epi32(ewitab,2);
 558             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 559             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 560             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 561             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 562             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 563             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 564             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 565             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 566             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 567             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 568
 569             /* Update potential sum for this i atom from the interaction with this j atom. */
 570             velecsum         = _mm_add_pd(velecsum,velec);
 571
 572             fscal            = felec;
 573
 574             /* Calculate temporary vectorial force */
 575             tx               = _mm_mul_pd(fscal,dx12);
 576             ty               = _mm_mul_pd(fscal,dy12);
 577             tz               = _mm_mul_pd(fscal,dz12);
 578
 579             /* Update vectorial force */
 580             fix1             = _mm_add_pd(fix1,tx);
 581             fiy1             = _mm_add_pd(fiy1,ty);
 582             fiz1             = _mm_add_pd(fiz1,tz);
 583
 584             fjx2             = _mm_add_pd(fjx2,tx);
 585             fjy2             = _mm_add_pd(fjy2,ty);
 586             fjz2             = _mm_add_pd(fjz2,tz);
 587
 588             /**************************
 589              * CALCULATE INTERACTIONS *
 590              **************************/
 591
 592             r20              = _mm_mul_pd(rsq20,rinv20);
 593
 594             /* EWALD ELECTROSTATICS */
 595
 596             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 597             ewrt             = _mm_mul_pd(r20,ewtabscale);
 598             ewitab           = _mm_cvttpd_epi32(ewrt);
 599             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 600             ewitab           = _mm_slli_epi32(ewitab,2);
 601             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 602             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 603             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 604             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 605             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 606             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 607             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 608             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 609             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 610             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 611
 612             /* Update potential sum for this i atom from the interaction with this j atom. */
 613             velecsum         = _mm_add_pd(velecsum,velec);
 614
 615             fscal            = felec;
 616
 617             /* Calculate temporary vectorial force */
 618             tx               = _mm_mul_pd(fscal,dx20);
 619             ty               = _mm_mul_pd(fscal,dy20);
 620             tz               = _mm_mul_pd(fscal,dz20);
 621
 622             /* Update vectorial force */
 623             fix2             = _mm_add_pd(fix2,tx);
 624             fiy2             = _mm_add_pd(fiy2,ty);
 625             fiz2             = _mm_add_pd(fiz2,tz);
 626
 627             fjx0             = _mm_add_pd(fjx0,tx);
 628             fjy0             = _mm_add_pd(fjy0,ty);
 629             fjz0             = _mm_add_pd(fjz0,tz);
 630
 631             /**************************
 632              * CALCULATE INTERACTIONS *
 633              **************************/
 634
 635             r21              = _mm_mul_pd(rsq21,rinv21);
 636
 637             /* EWALD ELECTROSTATICS */
 638
 639             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 640             ewrt             = _mm_mul_pd(r21,ewtabscale);
 641             ewitab           = _mm_cvttpd_epi32(ewrt);
 642             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 643             ewitab           = _mm_slli_epi32(ewitab,2);
 644             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 645             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 646             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 647             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 648             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 649             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 650             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 651             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 652             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 653             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 654
 655             /* Update potential sum for this i atom from the interaction with this j atom. */
 656             velecsum         = _mm_add_pd(velecsum,velec);
 657
 658             fscal            = felec;
 659
 660             /* Calculate temporary vectorial force */
 661             tx               = _mm_mul_pd(fscal,dx21);
 662             ty               = _mm_mul_pd(fscal,dy21);
 663             tz               = _mm_mul_pd(fscal,dz21);
 664
 665             /* Update vectorial force */
 666             fix2             = _mm_add_pd(fix2,tx);
 667             fiy2             = _mm_add_pd(fiy2,ty);
 668             fiz2             = _mm_add_pd(fiz2,tz);
 669
 670             fjx1             = _mm_add_pd(fjx1,tx);
 671             fjy1             = _mm_add_pd(fjy1,ty);
 672             fjz1             = _mm_add_pd(fjz1,tz);
 673
 674             /**************************
 675              * CALCULATE INTERACTIONS *
 676              **************************/
 677
 678             r22              = _mm_mul_pd(rsq22,rinv22);
 679
 680             /* EWALD ELECTROSTATICS */
 681
 682             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 683             ewrt             = _mm_mul_pd(r22,ewtabscale);
 684             ewitab           = _mm_cvttpd_epi32(ewrt);
 685             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 686             ewitab           = _mm_slli_epi32(ewitab,2);
 687             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 688             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 689             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 690             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 691             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 692             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 693             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 694             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 695             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 696             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 697
 698             /* Update potential sum for this i atom from the interaction with this j atom. */
 699             velecsum         = _mm_add_pd(velecsum,velec);
 700
 701             fscal            = felec;
 702
 703             /* Calculate temporary vectorial force */
 704             tx               = _mm_mul_pd(fscal,dx22);
 705             ty               = _mm_mul_pd(fscal,dy22);
 706             tz               = _mm_mul_pd(fscal,dz22);
 707
 708             /* Update vectorial force */
 709             fix2             = _mm_add_pd(fix2,tx);
 710             fiy2             = _mm_add_pd(fiy2,ty);
 711             fiz2             = _mm_add_pd(fiz2,tz);
 712
 713             fjx2             = _mm_add_pd(fjx2,tx);
 714             fjy2             = _mm_add_pd(fjy2,ty);
 715             fjz2             = _mm_add_pd(fjz2,tz);
 716
 717             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 718
 719             /* Inner loop uses 403 flops */
 720         }
 721
 722         if(jidx<j_index_end)
 723         {
 724
 725             jnrA             = jjnr[jidx];
 726             j_coord_offsetA  = DIM*jnrA;
 727
 728             /* load j atom coordinates */
 729             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 730                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 731
 732             /* Calculate displacement vector */
 733             dx00             = _mm_sub_pd(ix0,jx0);
 734             dy00             = _mm_sub_pd(iy0,jy0);
 735             dz00             = _mm_sub_pd(iz0,jz0);
 736             dx01             = _mm_sub_pd(ix0,jx1);
 737             dy01             = _mm_sub_pd(iy0,jy1);
 738             dz01             = _mm_sub_pd(iz0,jz1);
 739             dx02             = _mm_sub_pd(ix0,jx2);
 740             dy02             = _mm_sub_pd(iy0,jy2);
 741             dz02             = _mm_sub_pd(iz0,jz2);
 742             dx10             = _mm_sub_pd(ix1,jx0);
 743             dy10             = _mm_sub_pd(iy1,jy0);
 744             dz10             = _mm_sub_pd(iz1,jz0);
 745             dx11             = _mm_sub_pd(ix1,jx1);
 746             dy11             = _mm_sub_pd(iy1,jy1);
 747             dz11             = _mm_sub_pd(iz1,jz1);
 748             dx12             = _mm_sub_pd(ix1,jx2);
 749             dy12             = _mm_sub_pd(iy1,jy2);
 750             dz12             = _mm_sub_pd(iz1,jz2);
 751             dx20             = _mm_sub_pd(ix2,jx0);
 752             dy20             = _mm_sub_pd(iy2,jy0);
 753             dz20             = _mm_sub_pd(iz2,jz0);
 754             dx21             = _mm_sub_pd(ix2,jx1);
 755             dy21             = _mm_sub_pd(iy2,jy1);
 756             dz21             = _mm_sub_pd(iz2,jz1);
 757             dx22             = _mm_sub_pd(ix2,jx2);
 758             dy22             = _mm_sub_pd(iy2,jy2);
 759             dz22             = _mm_sub_pd(iz2,jz2);
 760
 761             /* Calculate squared distance and things based on it */
 762             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 763             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 764             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 765             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 766             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 767             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 768             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 769             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 770             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 771
 772             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 773             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 774             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 775             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 776             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 777             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 778             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 779             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 780             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 781
 782             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 783             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 784             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 785             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 786             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 787             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 788             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 789             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 790             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 791
 792             fjx0             = _mm_setzero_pd();
 793             fjy0             = _mm_setzero_pd();
 794             fjz0             = _mm_setzero_pd();
 795             fjx1             = _mm_setzero_pd();
 796             fjy1             = _mm_setzero_pd();
 797             fjz1             = _mm_setzero_pd();
 798             fjx2             = _mm_setzero_pd();
 799             fjy2             = _mm_setzero_pd();
 800             fjz2             = _mm_setzero_pd();
 801
 802             /**************************
 803              * CALCULATE INTERACTIONS *
 804              **************************/
 805
 806             r00              = _mm_mul_pd(rsq00,rinv00);
 807
 808             /* Calculate table index by multiplying r with table scale and truncate to integer */
 809             rt               = _mm_mul_pd(r00,vftabscale);
 810             vfitab           = _mm_cvttpd_epi32(rt);
 811             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 812             vfitab           = _mm_slli_epi32(vfitab,3);
 813
 814             /* EWALD ELECTROSTATICS */
 815
 816             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 817             ewrt             = _mm_mul_pd(r00,ewtabscale);
 818             ewitab           = _mm_cvttpd_epi32(ewrt);
 819             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 820             ewitab           = _mm_slli_epi32(ewitab,2);
 821             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 822             ewtabD           = _mm_setzero_pd();
 823             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 824             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 825             ewtabFn          = _mm_setzero_pd();
 826             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 827             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 828             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 829             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 830             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 831
 832             /* CUBIC SPLINE TABLE DISPERSION */
 833             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 834             F                = _mm_setzero_pd();
 835             GMX_MM_TRANSPOSE2_PD(Y,F);
 836             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 837             H                = _mm_setzero_pd();
 838             GMX_MM_TRANSPOSE2_PD(G,H);
 839             Heps             = _mm_mul_pd(vfeps,H);
 840             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 841             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 842             vvdw6            = _mm_mul_pd(c6_00,VV);
 843             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 844             fvdw6            = _mm_mul_pd(c6_00,FF);
 845
 846             /* CUBIC SPLINE TABLE REPULSION */
 847             vfitab           = _mm_add_epi32(vfitab,ifour);
 848             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 849             F                = _mm_setzero_pd();
 850             GMX_MM_TRANSPOSE2_PD(Y,F);
 851             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 852             H                = _mm_setzero_pd();
 853             GMX_MM_TRANSPOSE2_PD(G,H);
 854             Heps             = _mm_mul_pd(vfeps,H);
 855             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 856             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 857             vvdw12           = _mm_mul_pd(c12_00,VV);
 858             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 859             fvdw12           = _mm_mul_pd(c12_00,FF);
 860             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 861             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 862
 863             /* Update potential sum for this i atom from the interaction with this j atom. */
 864             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 865             velecsum         = _mm_add_pd(velecsum,velec);
 866             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 867             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 868
 869             fscal            = _mm_add_pd(felec,fvdw);
 870
 871             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 872
 873             /* Calculate temporary vectorial force */
 874             tx               = _mm_mul_pd(fscal,dx00);
 875             ty               = _mm_mul_pd(fscal,dy00);
 876             tz               = _mm_mul_pd(fscal,dz00);
 877
 878             /* Update vectorial force */
 879             fix0             = _mm_add_pd(fix0,tx);
 880             fiy0             = _mm_add_pd(fiy0,ty);
 881             fiz0             = _mm_add_pd(fiz0,tz);
 882
 883             fjx0             = _mm_add_pd(fjx0,tx);
 884             fjy0             = _mm_add_pd(fjy0,ty);
 885             fjz0             = _mm_add_pd(fjz0,tz);
 886
 887             /**************************
 888              * CALCULATE INTERACTIONS *
 889              **************************/
 890
 891             r01              = _mm_mul_pd(rsq01,rinv01);
 892
 893             /* EWALD ELECTROSTATICS */
 894
 895             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 896             ewrt             = _mm_mul_pd(r01,ewtabscale);
 897             ewitab           = _mm_cvttpd_epi32(ewrt);
 898             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 899             ewitab           = _mm_slli_epi32(ewitab,2);
 900             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 901             ewtabD           = _mm_setzero_pd();
 902             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 903             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 904             ewtabFn          = _mm_setzero_pd();
 905             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 906             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 907             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 908             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 909             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 910
 911             /* Update potential sum for this i atom from the interaction with this j atom. */
 912             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 913             velecsum         = _mm_add_pd(velecsum,velec);
 914
 915             fscal            = felec;
 916
 917             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 918
 919             /* Calculate temporary vectorial force */
 920             tx               = _mm_mul_pd(fscal,dx01);
 921             ty               = _mm_mul_pd(fscal,dy01);
 922             tz               = _mm_mul_pd(fscal,dz01);
 923
 924             /* Update vectorial force */
 925             fix0             = _mm_add_pd(fix0,tx);
 926             fiy0             = _mm_add_pd(fiy0,ty);
 927             fiz0             = _mm_add_pd(fiz0,tz);
 928
 929             fjx1             = _mm_add_pd(fjx1,tx);
 930             fjy1             = _mm_add_pd(fjy1,ty);
 931             fjz1             = _mm_add_pd(fjz1,tz);
 932
 933             /**************************
 934              * CALCULATE INTERACTIONS *
 935              **************************/
 936
 937             r02              = _mm_mul_pd(rsq02,rinv02);
 938
 939             /* EWALD ELECTROSTATICS */
 940
 941             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 942             ewrt             = _mm_mul_pd(r02,ewtabscale);
 943             ewitab           = _mm_cvttpd_epi32(ewrt);
 944             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 945             ewitab           = _mm_slli_epi32(ewitab,2);
 946             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 947             ewtabD           = _mm_setzero_pd();
 948             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 949             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 950             ewtabFn          = _mm_setzero_pd();
 951             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 952             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 953             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 954             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 955             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 956
 957             /* Update potential sum for this i atom from the interaction with this j atom. */
 958             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 959             velecsum         = _mm_add_pd(velecsum,velec);
 960
 961             fscal            = felec;
 962
 963             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 964
 965             /* Calculate temporary vectorial force */
 966             tx               = _mm_mul_pd(fscal,dx02);
 967             ty               = _mm_mul_pd(fscal,dy02);
 968             tz               = _mm_mul_pd(fscal,dz02);
 969
 970             /* Update vectorial force */
 971             fix0             = _mm_add_pd(fix0,tx);
 972             fiy0             = _mm_add_pd(fiy0,ty);
 973             fiz0             = _mm_add_pd(fiz0,tz);
 974
 975             fjx2             = _mm_add_pd(fjx2,tx);
 976             fjy2             = _mm_add_pd(fjy2,ty);
 977             fjz2             = _mm_add_pd(fjz2,tz);
 978
 979             /**************************
 980              * CALCULATE INTERACTIONS *
 981              **************************/
 982
 983             r10              = _mm_mul_pd(rsq10,rinv10);
 984
 985             /* EWALD ELECTROSTATICS */
 986
 987             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 988             ewrt             = _mm_mul_pd(r10,ewtabscale);
 989             ewitab           = _mm_cvttpd_epi32(ewrt);
 990             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 991             ewitab           = _mm_slli_epi32(ewitab,2);
 992             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 993             ewtabD           = _mm_setzero_pd();
 994             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 995             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 996             ewtabFn          = _mm_setzero_pd();
 997             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 998             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 999             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1000             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
1001             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1002
1003             /* Update potential sum for this i atom from the interaction with this j atom. */
1004             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1005             velecsum         = _mm_add_pd(velecsum,velec);
1006
1007             fscal            = felec;
1008
1009             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1010
1011             /* Calculate temporary vectorial force */
1012             tx               = _mm_mul_pd(fscal,dx10);
1013             ty               = _mm_mul_pd(fscal,dy10);
1014             tz               = _mm_mul_pd(fscal,dz10);
1015
1016             /* Update vectorial force */
1017             fix1             = _mm_add_pd(fix1,tx);
1018             fiy1             = _mm_add_pd(fiy1,ty);
1019             fiz1             = _mm_add_pd(fiz1,tz);
1020
1021             fjx0             = _mm_add_pd(fjx0,tx);
1022             fjy0             = _mm_add_pd(fjy0,ty);
1023             fjz0             = _mm_add_pd(fjz0,tz);
1024
1025             /**************************
1026              * CALCULATE INTERACTIONS *
1027              **************************/
1028
1029             r11              = _mm_mul_pd(rsq11,rinv11);
1030
1031             /* EWALD ELECTROSTATICS */
1032
1033             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1034             ewrt             = _mm_mul_pd(r11,ewtabscale);
1035             ewitab           = _mm_cvttpd_epi32(ewrt);
1036             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1037             ewitab           = _mm_slli_epi32(ewitab,2);
1038             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1039             ewtabD           = _mm_setzero_pd();
1040             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1041             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1042             ewtabFn          = _mm_setzero_pd();
1043             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1044             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1045             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1046             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
1047             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1048
1049             /* Update potential sum for this i atom from the interaction with this j atom. */
1050             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1051             velecsum         = _mm_add_pd(velecsum,velec);
1052
1053             fscal            = felec;
1054
1055             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1056
1057             /* Calculate temporary vectorial force */
1058             tx               = _mm_mul_pd(fscal,dx11);
1059             ty               = _mm_mul_pd(fscal,dy11);
1060             tz               = _mm_mul_pd(fscal,dz11);
1061
1062             /* Update vectorial force */
1063             fix1             = _mm_add_pd(fix1,tx);
1064             fiy1             = _mm_add_pd(fiy1,ty);
1065             fiz1             = _mm_add_pd(fiz1,tz);
1066
1067             fjx1             = _mm_add_pd(fjx1,tx);
1068             fjy1             = _mm_add_pd(fjy1,ty);
1069             fjz1             = _mm_add_pd(fjz1,tz);
1070
1071             /**************************
1072              * CALCULATE INTERACTIONS *
1073              **************************/
1074
1075             r12              = _mm_mul_pd(rsq12,rinv12);
1076
1077             /* EWALD ELECTROSTATICS */
1078
1079             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1080             ewrt             = _mm_mul_pd(r12,ewtabscale);
1081             ewitab           = _mm_cvttpd_epi32(ewrt);
1082             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1083             ewitab           = _mm_slli_epi32(ewitab,2);
1084             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1085             ewtabD           = _mm_setzero_pd();
1086             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1087             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1088             ewtabFn          = _mm_setzero_pd();
1089             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1090             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1091             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1092             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
1093             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1094
1095             /* Update potential sum for this i atom from the interaction with this j atom. */
1096             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1097             velecsum         = _mm_add_pd(velecsum,velec);
1098
1099             fscal            = felec;
1100
1101             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1102
1103             /* Calculate temporary vectorial force */
1104             tx               = _mm_mul_pd(fscal,dx12);
1105             ty               = _mm_mul_pd(fscal,dy12);
1106             tz               = _mm_mul_pd(fscal,dz12);
1107
1108             /* Update vectorial force */
1109             fix1             = _mm_add_pd(fix1,tx);
1110             fiy1             = _mm_add_pd(fiy1,ty);
1111             fiz1             = _mm_add_pd(fiz1,tz);
1112
1113             fjx2             = _mm_add_pd(fjx2,tx);
1114             fjy2             = _mm_add_pd(fjy2,ty);
1115             fjz2             = _mm_add_pd(fjz2,tz);
1116
1117             /**************************
1118              * CALCULATE INTERACTIONS *
1119              **************************/
1120
1121             r20              = _mm_mul_pd(rsq20,rinv20);
1122
1123             /* EWALD ELECTROSTATICS */
1124
1125             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1126             ewrt             = _mm_mul_pd(r20,ewtabscale);
1127             ewitab           = _mm_cvttpd_epi32(ewrt);
1128             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1129             ewitab           = _mm_slli_epi32(ewitab,2);
1130             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1131             ewtabD           = _mm_setzero_pd();
1132             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1133             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1134             ewtabFn          = _mm_setzero_pd();
1135             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1136             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1137             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1138             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1139             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1140
1141             /* Update potential sum for this i atom from the interaction with this j atom. */
1142             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1143             velecsum         = _mm_add_pd(velecsum,velec);
1144
1145             fscal            = felec;
1146
1147             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1148
1149             /* Calculate temporary vectorial force */
1150             tx               = _mm_mul_pd(fscal,dx20);
1151             ty               = _mm_mul_pd(fscal,dy20);
1152             tz               = _mm_mul_pd(fscal,dz20);
1153
1154             /* Update vectorial force */
1155             fix2             = _mm_add_pd(fix2,tx);
1156             fiy2             = _mm_add_pd(fiy2,ty);
1157             fiz2             = _mm_add_pd(fiz2,tz);
1158
1159             fjx0             = _mm_add_pd(fjx0,tx);
1160             fjy0             = _mm_add_pd(fjy0,ty);
1161             fjz0             = _mm_add_pd(fjz0,tz);
1162
1163             /**************************
1164              * CALCULATE INTERACTIONS *
1165              **************************/
1166
1167             r21              = _mm_mul_pd(rsq21,rinv21);
1168
1169             /* EWALD ELECTROSTATICS */
1170
1171             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1172             ewrt             = _mm_mul_pd(r21,ewtabscale);
1173             ewitab           = _mm_cvttpd_epi32(ewrt);
1174             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1175             ewitab           = _mm_slli_epi32(ewitab,2);
1176             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1177             ewtabD           = _mm_setzero_pd();
1178             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1179             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1180             ewtabFn          = _mm_setzero_pd();
1181             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1182             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1183             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1184             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1185             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1186
1187             /* Update potential sum for this i atom from the interaction with this j atom. */
1188             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1189             velecsum         = _mm_add_pd(velecsum,velec);
1190
1191             fscal            = felec;
1192
1193             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1194
1195             /* Calculate temporary vectorial force */
1196             tx               = _mm_mul_pd(fscal,dx21);
1197             ty               = _mm_mul_pd(fscal,dy21);
1198             tz               = _mm_mul_pd(fscal,dz21);
1199
1200             /* Update vectorial force */
1201             fix2             = _mm_add_pd(fix2,tx);
1202             fiy2             = _mm_add_pd(fiy2,ty);
1203             fiz2             = _mm_add_pd(fiz2,tz);
1204
1205             fjx1             = _mm_add_pd(fjx1,tx);
1206             fjy1             = _mm_add_pd(fjy1,ty);
1207             fjz1             = _mm_add_pd(fjz1,tz);
1208
1209             /**************************
1210              * CALCULATE INTERACTIONS *
1211              **************************/
1212
1213             r22              = _mm_mul_pd(rsq22,rinv22);
1214
1215             /* EWALD ELECTROSTATICS */
1216
1217             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1218             ewrt             = _mm_mul_pd(r22,ewtabscale);
1219             ewitab           = _mm_cvttpd_epi32(ewrt);
1220             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1221             ewitab           = _mm_slli_epi32(ewitab,2);
1222             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1223             ewtabD           = _mm_setzero_pd();
1224             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1225             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1226             ewtabFn          = _mm_setzero_pd();
1227             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1228             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1229             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1230             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1231             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1232
1233             /* Update potential sum for this i atom from the interaction with this j atom. */
1234             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1235             velecsum         = _mm_add_pd(velecsum,velec);
1236
1237             fscal            = felec;
1238
1239             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1240
1241             /* Calculate temporary vectorial force */
1242             tx               = _mm_mul_pd(fscal,dx22);
1243             ty               = _mm_mul_pd(fscal,dy22);
1244             tz               = _mm_mul_pd(fscal,dz22);
1245
1246             /* Update vectorial force */
1247             fix2             = _mm_add_pd(fix2,tx);
1248             fiy2             = _mm_add_pd(fiy2,ty);
1249             fiz2             = _mm_add_pd(fiz2,tz);
1250
1251             fjx2             = _mm_add_pd(fjx2,tx);
1252             fjy2             = _mm_add_pd(fjy2,ty);
1253             fjz2             = _mm_add_pd(fjz2,tz);
1254
1255             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1256
1257             /* Inner loop uses 403 flops */
1258         }
1259
1260         /* End of innermost loop */
1261
1262         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1263                                               f+i_coord_offset,fshift+i_shift_offset);
1264
1265         ggid                        = gid[iidx];
1266         /* Update potential energies */
1267         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1268         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1269
1270         /* Increment number of inner iterations */
1271         inneriter                  += j_index_end - j_index_start;
1272
1273         /* Outer loop uses 20 flops */
1274     }
1275
1276     /* Increment number of outer iterations */
1277     outeriter        += nri;
1278
1279     /* Update outer/inner flops */
1280
1281     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*403);
1282 }
1283 /*
1284  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_sse4_1_double
1285  * Electrostatics interaction: Ewald
1286  * VdW interaction:            CubicSplineTable
1287  * Geometry:                   Water3-Water3
1288  * Calculate force/pot:        Force
1289  */
1290 void
1291 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_sse4_1_double
1292                     (t_nblist                    * gmx_restrict       nlist,
1293                      rvec                        * gmx_restrict          xx,
1294                      rvec                        * gmx_restrict          ff,
1295                      t_forcerec                  * gmx_restrict          fr,
1296                      t_mdatoms                   * gmx_restrict     mdatoms,
1297                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1298                      t_nrnb                      * gmx_restrict        nrnb)
1299 {
1300     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1301      * just 0 for non-waters.
1302      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1303      * jnr indices corresponding to data put in the four positions in the SIMD register.
1304      */
1305     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1306     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1307     int              jnrA,jnrB;
1308     int              j_coord_offsetA,j_coord_offsetB;
1309     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1310     real             rcutoff_scalar;
1311     real             *shiftvec,*fshift,*x,*f;
1312     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1313     int              vdwioffset0;
1314     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1315     int              vdwioffset1;
1316     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1317     int              vdwioffset2;
1318     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1319     int              vdwjidx0A,vdwjidx0B;
1320     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1321     int              vdwjidx1A,vdwjidx1B;
1322     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1323     int              vdwjidx2A,vdwjidx2B;
1324     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1325     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1326     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1327     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1328     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1329     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1330     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1331     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1332     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1333     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1334     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1335     real             *charge;
1336     int              nvdwtype;
1337     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1338     int              *vdwtype;
1339     real             *vdwparam;
1340     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1341     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1342     __m128i          vfitab;
1343     __m128i          ifour       = _mm_set1_epi32(4);
1344     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1345     real             *vftab;
1346     __m128i          ewitab;
1347     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1348     real             *ewtab;
1349     __m128d          dummy_mask,cutoff_mask;
1350     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1351     __m128d          one     = _mm_set1_pd(1.0);
1352     __m128d          two     = _mm_set1_pd(2.0);
1353     x                = xx[0];
1354     f                = ff[0];
1355
1356     nri              = nlist->nri;
1357     iinr             = nlist->iinr;
1358     jindex           = nlist->jindex;
1359     jjnr             = nlist->jjnr;
1360     shiftidx         = nlist->shift;
1361     gid              = nlist->gid;
1362     shiftvec         = fr->shift_vec[0];
1363     fshift           = fr->fshift[0];
1364     facel            = _mm_set1_pd(fr->epsfac);
1365     charge           = mdatoms->chargeA;
1366     nvdwtype         = fr->ntype;
1367     vdwparam         = fr->nbfp;
1368     vdwtype          = mdatoms->typeA;
1369
1370     vftab            = kernel_data->table_vdw->data;
1371     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
1372
1373     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1374     ewtab            = fr->ic->tabq_coul_F;
1375     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1376     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1377
1378     /* Setup water-specific parameters */
1379     inr              = nlist->iinr[0];
1380     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1381     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1382     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1383     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1384
1385     jq0              = _mm_set1_pd(charge[inr+0]);
1386     jq1              = _mm_set1_pd(charge[inr+1]);
1387     jq2              = _mm_set1_pd(charge[inr+2]);
1388     vdwjidx0A        = 2*vdwtype[inr+0];
1389     qq00             = _mm_mul_pd(iq0,jq0);
1390     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1391     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1392     qq01             = _mm_mul_pd(iq0,jq1);
1393     qq02             = _mm_mul_pd(iq0,jq2);
1394     qq10             = _mm_mul_pd(iq1,jq0);
1395     qq11             = _mm_mul_pd(iq1,jq1);
1396     qq12             = _mm_mul_pd(iq1,jq2);
1397     qq20             = _mm_mul_pd(iq2,jq0);
1398     qq21             = _mm_mul_pd(iq2,jq1);
1399     qq22             = _mm_mul_pd(iq2,jq2);
1400
1401     /* Avoid stupid compiler warnings */
1402     jnrA = jnrB = 0;
1403     j_coord_offsetA = 0;
1404     j_coord_offsetB = 0;
1405
1406     outeriter        = 0;
1407     inneriter        = 0;
1408
1409     /* Start outer loop over neighborlists */
1410     for(iidx=0; iidx<nri; iidx++)
1411     {
1412         /* Load shift vector for this list */
1413         i_shift_offset   = DIM*shiftidx[iidx];
1414
1415         /* Load limits for loop over neighbors */
1416         j_index_start    = jindex[iidx];
1417         j_index_end      = jindex[iidx+1];
1418
1419         /* Get outer coordinate index */
1420         inr              = iinr[iidx];
1421         i_coord_offset   = DIM*inr;
1422
1423         /* Load i particle coords and add shift vector */
1424         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1425                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1426
1427         fix0             = _mm_setzero_pd();
1428         fiy0             = _mm_setzero_pd();
1429         fiz0             = _mm_setzero_pd();
1430         fix1             = _mm_setzero_pd();
1431         fiy1             = _mm_setzero_pd();
1432         fiz1             = _mm_setzero_pd();
1433         fix2             = _mm_setzero_pd();
1434         fiy2             = _mm_setzero_pd();
1435         fiz2             = _mm_setzero_pd();
1436
1437         /* Start inner kernel loop */
1438         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1439         {
1440
1441             /* Get j neighbor index, and coordinate index */
1442             jnrA             = jjnr[jidx];
1443             jnrB             = jjnr[jidx+1];
1444             j_coord_offsetA  = DIM*jnrA;
1445             j_coord_offsetB  = DIM*jnrB;
1446
1447             /* load j atom coordinates */
1448             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1449                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1450
1451             /* Calculate displacement vector */
1452             dx00             = _mm_sub_pd(ix0,jx0);
1453             dy00             = _mm_sub_pd(iy0,jy0);
1454             dz00             = _mm_sub_pd(iz0,jz0);
1455             dx01             = _mm_sub_pd(ix0,jx1);
1456             dy01             = _mm_sub_pd(iy0,jy1);
1457             dz01             = _mm_sub_pd(iz0,jz1);
1458             dx02             = _mm_sub_pd(ix0,jx2);
1459             dy02             = _mm_sub_pd(iy0,jy2);
1460             dz02             = _mm_sub_pd(iz0,jz2);
1461             dx10             = _mm_sub_pd(ix1,jx0);
1462             dy10             = _mm_sub_pd(iy1,jy0);
1463             dz10             = _mm_sub_pd(iz1,jz0);
1464             dx11             = _mm_sub_pd(ix1,jx1);
1465             dy11             = _mm_sub_pd(iy1,jy1);
1466             dz11             = _mm_sub_pd(iz1,jz1);
1467             dx12             = _mm_sub_pd(ix1,jx2);
1468             dy12             = _mm_sub_pd(iy1,jy2);
1469             dz12             = _mm_sub_pd(iz1,jz2);
1470             dx20             = _mm_sub_pd(ix2,jx0);
1471             dy20             = _mm_sub_pd(iy2,jy0);
1472             dz20             = _mm_sub_pd(iz2,jz0);
1473             dx21             = _mm_sub_pd(ix2,jx1);
1474             dy21             = _mm_sub_pd(iy2,jy1);
1475             dz21             = _mm_sub_pd(iz2,jz1);
1476             dx22             = _mm_sub_pd(ix2,jx2);
1477             dy22             = _mm_sub_pd(iy2,jy2);
1478             dz22             = _mm_sub_pd(iz2,jz2);
1479
1480             /* Calculate squared distance and things based on it */
1481             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1482             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1483             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1484             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1485             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1486             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1487             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1488             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1489             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1490
1491             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1492             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1493             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1494             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1495             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1496             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1497             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1498             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1499             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1500
1501             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1502             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1503             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1504             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1505             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1506             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1507             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1508             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1509             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1510
1511             fjx0             = _mm_setzero_pd();
1512             fjy0             = _mm_setzero_pd();
1513             fjz0             = _mm_setzero_pd();
1514             fjx1             = _mm_setzero_pd();
1515             fjy1             = _mm_setzero_pd();
1516             fjz1             = _mm_setzero_pd();
1517             fjx2             = _mm_setzero_pd();
1518             fjy2             = _mm_setzero_pd();
1519             fjz2             = _mm_setzero_pd();
1520
1521             /**************************
1522              * CALCULATE INTERACTIONS *
1523              **************************/
1524
1525             r00              = _mm_mul_pd(rsq00,rinv00);
1526
1527             /* Calculate table index by multiplying r with table scale and truncate to integer */
1528             rt               = _mm_mul_pd(r00,vftabscale);
1529             vfitab           = _mm_cvttpd_epi32(rt);
1530             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1531             vfitab           = _mm_slli_epi32(vfitab,3);
1532
1533             /* EWALD ELECTROSTATICS */
1534
1535             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1536             ewrt             = _mm_mul_pd(r00,ewtabscale);
1537             ewitab           = _mm_cvttpd_epi32(ewrt);
1538             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1539             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1540                                          &ewtabF,&ewtabFn);
1541             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1542             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1543
1544             /* CUBIC SPLINE TABLE DISPERSION */
1545             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1546             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1547             GMX_MM_TRANSPOSE2_PD(Y,F);
1548             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1549             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1550             GMX_MM_TRANSPOSE2_PD(G,H);
1551             Heps             = _mm_mul_pd(vfeps,H);
1552             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1553             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1554             fvdw6            = _mm_mul_pd(c6_00,FF);
1555
1556             /* CUBIC SPLINE TABLE REPULSION */
1557             vfitab           = _mm_add_epi32(vfitab,ifour);
1558             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1559             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1560             GMX_MM_TRANSPOSE2_PD(Y,F);
1561             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1562             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1563             GMX_MM_TRANSPOSE2_PD(G,H);
1564             Heps             = _mm_mul_pd(vfeps,H);
1565             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1566             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1567             fvdw12           = _mm_mul_pd(c12_00,FF);
1568             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1569
1570             fscal            = _mm_add_pd(felec,fvdw);
1571
1572             /* Calculate temporary vectorial force */
1573             tx               = _mm_mul_pd(fscal,dx00);
1574             ty               = _mm_mul_pd(fscal,dy00);
1575             tz               = _mm_mul_pd(fscal,dz00);
1576
1577             /* Update vectorial force */
1578             fix0             = _mm_add_pd(fix0,tx);
1579             fiy0             = _mm_add_pd(fiy0,ty);
1580             fiz0             = _mm_add_pd(fiz0,tz);
1581
1582             fjx0             = _mm_add_pd(fjx0,tx);
1583             fjy0             = _mm_add_pd(fjy0,ty);
1584             fjz0             = _mm_add_pd(fjz0,tz);
1585
1586             /**************************
1587              * CALCULATE INTERACTIONS *
1588              **************************/
1589
1590             r01              = _mm_mul_pd(rsq01,rinv01);
1591
1592             /* EWALD ELECTROSTATICS */
1593
1594             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1595             ewrt             = _mm_mul_pd(r01,ewtabscale);
1596             ewitab           = _mm_cvttpd_epi32(ewrt);
1597             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1598             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1599                                          &ewtabF,&ewtabFn);
1600             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1601             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1602
1603             fscal            = felec;
1604
1605             /* Calculate temporary vectorial force */
1606             tx               = _mm_mul_pd(fscal,dx01);
1607             ty               = _mm_mul_pd(fscal,dy01);
1608             tz               = _mm_mul_pd(fscal,dz01);
1609
1610             /* Update vectorial force */
1611             fix0             = _mm_add_pd(fix0,tx);
1612             fiy0             = _mm_add_pd(fiy0,ty);
1613             fiz0             = _mm_add_pd(fiz0,tz);
1614
1615             fjx1             = _mm_add_pd(fjx1,tx);
1616             fjy1             = _mm_add_pd(fjy1,ty);
1617             fjz1             = _mm_add_pd(fjz1,tz);
1618
1619             /**************************
1620              * CALCULATE INTERACTIONS *
1621              **************************/
1622
1623             r02              = _mm_mul_pd(rsq02,rinv02);
1624
1625             /* EWALD ELECTROSTATICS */
1626
1627             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1628             ewrt             = _mm_mul_pd(r02,ewtabscale);
1629             ewitab           = _mm_cvttpd_epi32(ewrt);
1630             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1631             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1632                                          &ewtabF,&ewtabFn);
1633             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1634             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1635
1636             fscal            = felec;
1637
1638             /* Calculate temporary vectorial force */
1639             tx               = _mm_mul_pd(fscal,dx02);
1640             ty               = _mm_mul_pd(fscal,dy02);
1641             tz               = _mm_mul_pd(fscal,dz02);
1642
1643             /* Update vectorial force */
1644             fix0             = _mm_add_pd(fix0,tx);
1645             fiy0             = _mm_add_pd(fiy0,ty);
1646             fiz0             = _mm_add_pd(fiz0,tz);
1647
1648             fjx2             = _mm_add_pd(fjx2,tx);
1649             fjy2             = _mm_add_pd(fjy2,ty);
1650             fjz2             = _mm_add_pd(fjz2,tz);
1651
1652             /**************************
1653              * CALCULATE INTERACTIONS *
1654              **************************/
1655
1656             r10              = _mm_mul_pd(rsq10,rinv10);
1657
1658             /* EWALD ELECTROSTATICS */
1659
1660             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1661             ewrt             = _mm_mul_pd(r10,ewtabscale);
1662             ewitab           = _mm_cvttpd_epi32(ewrt);
1663             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1664             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1665                                          &ewtabF,&ewtabFn);
1666             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1667             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1668
1669             fscal            = felec;
1670
1671             /* Calculate temporary vectorial force */
1672             tx               = _mm_mul_pd(fscal,dx10);
1673             ty               = _mm_mul_pd(fscal,dy10);
1674             tz               = _mm_mul_pd(fscal,dz10);
1675
1676             /* Update vectorial force */
1677             fix1             = _mm_add_pd(fix1,tx);
1678             fiy1             = _mm_add_pd(fiy1,ty);
1679             fiz1             = _mm_add_pd(fiz1,tz);
1680
1681             fjx0             = _mm_add_pd(fjx0,tx);
1682             fjy0             = _mm_add_pd(fjy0,ty);
1683             fjz0             = _mm_add_pd(fjz0,tz);
1684
1685             /**************************
1686              * CALCULATE INTERACTIONS *
1687              **************************/
1688
1689             r11              = _mm_mul_pd(rsq11,rinv11);
1690
1691             /* EWALD ELECTROSTATICS */
1692
1693             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1694             ewrt             = _mm_mul_pd(r11,ewtabscale);
1695             ewitab           = _mm_cvttpd_epi32(ewrt);
1696             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1697             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1698                                          &ewtabF,&ewtabFn);
1699             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1700             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1701
1702             fscal            = felec;
1703
1704             /* Calculate temporary vectorial force */
1705             tx               = _mm_mul_pd(fscal,dx11);
1706             ty               = _mm_mul_pd(fscal,dy11);
1707             tz               = _mm_mul_pd(fscal,dz11);
1708
1709             /* Update vectorial force */
1710             fix1             = _mm_add_pd(fix1,tx);
1711             fiy1             = _mm_add_pd(fiy1,ty);
1712             fiz1             = _mm_add_pd(fiz1,tz);
1713
1714             fjx1             = _mm_add_pd(fjx1,tx);
1715             fjy1             = _mm_add_pd(fjy1,ty);
1716             fjz1             = _mm_add_pd(fjz1,tz);
1717
1718             /**************************
1719              * CALCULATE INTERACTIONS *
1720              **************************/
1721
1722             r12              = _mm_mul_pd(rsq12,rinv12);
1723
1724             /* EWALD ELECTROSTATICS */
1725
1726             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1727             ewrt             = _mm_mul_pd(r12,ewtabscale);
1728             ewitab           = _mm_cvttpd_epi32(ewrt);
1729             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1730             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1731                                          &ewtabF,&ewtabFn);
1732             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1733             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1734
1735             fscal            = felec;
1736
1737             /* Calculate temporary vectorial force */
1738             tx               = _mm_mul_pd(fscal,dx12);
1739             ty               = _mm_mul_pd(fscal,dy12);
1740             tz               = _mm_mul_pd(fscal,dz12);
1741
1742             /* Update vectorial force */
1743             fix1             = _mm_add_pd(fix1,tx);
1744             fiy1             = _mm_add_pd(fiy1,ty);
1745             fiz1             = _mm_add_pd(fiz1,tz);
1746
1747             fjx2             = _mm_add_pd(fjx2,tx);
1748             fjy2             = _mm_add_pd(fjy2,ty);
1749             fjz2             = _mm_add_pd(fjz2,tz);
1750
1751             /**************************
1752              * CALCULATE INTERACTIONS *
1753              **************************/
1754
1755             r20              = _mm_mul_pd(rsq20,rinv20);
1756
1757             /* EWALD ELECTROSTATICS */
1758
1759             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1760             ewrt             = _mm_mul_pd(r20,ewtabscale);
1761             ewitab           = _mm_cvttpd_epi32(ewrt);
1762             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1763             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1764                                          &ewtabF,&ewtabFn);
1765             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1766             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1767
1768             fscal            = felec;
1769
1770             /* Calculate temporary vectorial force */
1771             tx               = _mm_mul_pd(fscal,dx20);
1772             ty               = _mm_mul_pd(fscal,dy20);
1773             tz               = _mm_mul_pd(fscal,dz20);
1774
1775             /* Update vectorial force */
1776             fix2             = _mm_add_pd(fix2,tx);
1777             fiy2             = _mm_add_pd(fiy2,ty);
1778             fiz2             = _mm_add_pd(fiz2,tz);
1779
1780             fjx0             = _mm_add_pd(fjx0,tx);
1781             fjy0             = _mm_add_pd(fjy0,ty);
1782             fjz0             = _mm_add_pd(fjz0,tz);
1783
1784             /**************************
1785              * CALCULATE INTERACTIONS *
1786              **************************/
1787
1788             r21              = _mm_mul_pd(rsq21,rinv21);
1789
1790             /* EWALD ELECTROSTATICS */
1791
1792             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1793             ewrt             = _mm_mul_pd(r21,ewtabscale);
1794             ewitab           = _mm_cvttpd_epi32(ewrt);
1795             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1796             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1797                                          &ewtabF,&ewtabFn);
1798             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1799             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1800
1801             fscal            = felec;
1802
1803             /* Calculate temporary vectorial force */
1804             tx               = _mm_mul_pd(fscal,dx21);
1805             ty               = _mm_mul_pd(fscal,dy21);
1806             tz               = _mm_mul_pd(fscal,dz21);
1807
1808             /* Update vectorial force */
1809             fix2             = _mm_add_pd(fix2,tx);
1810             fiy2             = _mm_add_pd(fiy2,ty);
1811             fiz2             = _mm_add_pd(fiz2,tz);
1812
1813             fjx1             = _mm_add_pd(fjx1,tx);
1814             fjy1             = _mm_add_pd(fjy1,ty);
1815             fjz1             = _mm_add_pd(fjz1,tz);
1816
1817             /**************************
1818              * CALCULATE INTERACTIONS *
1819              **************************/
1820
1821             r22              = _mm_mul_pd(rsq22,rinv22);
1822
1823             /* EWALD ELECTROSTATICS */
1824
1825             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1826             ewrt             = _mm_mul_pd(r22,ewtabscale);
1827             ewitab           = _mm_cvttpd_epi32(ewrt);
1828             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1829             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1830                                          &ewtabF,&ewtabFn);
1831             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1832             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1833
1834             fscal            = felec;
1835
1836             /* Calculate temporary vectorial force */
1837             tx               = _mm_mul_pd(fscal,dx22);
1838             ty               = _mm_mul_pd(fscal,dy22);
1839             tz               = _mm_mul_pd(fscal,dz22);
1840
1841             /* Update vectorial force */
1842             fix2             = _mm_add_pd(fix2,tx);
1843             fiy2             = _mm_add_pd(fiy2,ty);
1844             fiz2             = _mm_add_pd(fiz2,tz);
1845
1846             fjx2             = _mm_add_pd(fjx2,tx);
1847             fjy2             = _mm_add_pd(fjy2,ty);
1848             fjz2             = _mm_add_pd(fjz2,tz);
1849
1850             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1851
1852             /* Inner loop uses 350 flops */
1853         }
1854
1855         if(jidx<j_index_end)
1856         {
1857
1858             jnrA             = jjnr[jidx];
1859             j_coord_offsetA  = DIM*jnrA;
1860
1861             /* load j atom coordinates */
1862             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1863                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1864
1865             /* Calculate displacement vector */
1866             dx00             = _mm_sub_pd(ix0,jx0);
1867             dy00             = _mm_sub_pd(iy0,jy0);
1868             dz00             = _mm_sub_pd(iz0,jz0);
1869             dx01             = _mm_sub_pd(ix0,jx1);
1870             dy01             = _mm_sub_pd(iy0,jy1);
1871             dz01             = _mm_sub_pd(iz0,jz1);
1872             dx02             = _mm_sub_pd(ix0,jx2);
1873             dy02             = _mm_sub_pd(iy0,jy2);
1874             dz02             = _mm_sub_pd(iz0,jz2);
1875             dx10             = _mm_sub_pd(ix1,jx0);
1876             dy10             = _mm_sub_pd(iy1,jy0);
1877             dz10             = _mm_sub_pd(iz1,jz0);
1878             dx11             = _mm_sub_pd(ix1,jx1);
1879             dy11             = _mm_sub_pd(iy1,jy1);
1880             dz11             = _mm_sub_pd(iz1,jz1);
1881             dx12             = _mm_sub_pd(ix1,jx2);
1882             dy12             = _mm_sub_pd(iy1,jy2);
1883             dz12             = _mm_sub_pd(iz1,jz2);
1884             dx20             = _mm_sub_pd(ix2,jx0);
1885             dy20             = _mm_sub_pd(iy2,jy0);
1886             dz20             = _mm_sub_pd(iz2,jz0);
1887             dx21             = _mm_sub_pd(ix2,jx1);
1888             dy21             = _mm_sub_pd(iy2,jy1);
1889             dz21             = _mm_sub_pd(iz2,jz1);
1890             dx22             = _mm_sub_pd(ix2,jx2);
1891             dy22             = _mm_sub_pd(iy2,jy2);
1892             dz22             = _mm_sub_pd(iz2,jz2);
1893
1894             /* Calculate squared distance and things based on it */
1895             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1896             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1897             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1898             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1899             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1900             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1901             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1902             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1903             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1904
1905             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1906             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1907             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1908             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1909             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1910             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1911             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1912             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1913             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1914
1915             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1916             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1917             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1918             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1919             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1920             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1921             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1922             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1923             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1924
1925             fjx0             = _mm_setzero_pd();
1926             fjy0             = _mm_setzero_pd();
1927             fjz0             = _mm_setzero_pd();
1928             fjx1             = _mm_setzero_pd();
1929             fjy1             = _mm_setzero_pd();
1930             fjz1             = _mm_setzero_pd();
1931             fjx2             = _mm_setzero_pd();
1932             fjy2             = _mm_setzero_pd();
1933             fjz2             = _mm_setzero_pd();
1934
1935             /**************************
1936              * CALCULATE INTERACTIONS *
1937              **************************/
1938
1939             r00              = _mm_mul_pd(rsq00,rinv00);
1940
1941             /* Calculate table index by multiplying r with table scale and truncate to integer */
1942             rt               = _mm_mul_pd(r00,vftabscale);
1943             vfitab           = _mm_cvttpd_epi32(rt);
1944             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1945             vfitab           = _mm_slli_epi32(vfitab,3);
1946
1947             /* EWALD ELECTROSTATICS */
1948
1949             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1950             ewrt             = _mm_mul_pd(r00,ewtabscale);
1951             ewitab           = _mm_cvttpd_epi32(ewrt);
1952             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1953             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1954             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1955             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1956
1957             /* CUBIC SPLINE TABLE DISPERSION */
1958             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1959             F                = _mm_setzero_pd();
1960             GMX_MM_TRANSPOSE2_PD(Y,F);
1961             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1962             H                = _mm_setzero_pd();
1963             GMX_MM_TRANSPOSE2_PD(G,H);
1964             Heps             = _mm_mul_pd(vfeps,H);
1965             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1966             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1967             fvdw6            = _mm_mul_pd(c6_00,FF);
1968
1969             /* CUBIC SPLINE TABLE REPULSION */
1970             vfitab           = _mm_add_epi32(vfitab,ifour);
1971             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1972             F                = _mm_setzero_pd();
1973             GMX_MM_TRANSPOSE2_PD(Y,F);
1974             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1975             H                = _mm_setzero_pd();
1976             GMX_MM_TRANSPOSE2_PD(G,H);
1977             Heps             = _mm_mul_pd(vfeps,H);
1978             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1979             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1980             fvdw12           = _mm_mul_pd(c12_00,FF);
1981             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1982
1983             fscal            = _mm_add_pd(felec,fvdw);
1984
1985             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1986
1987             /* Calculate temporary vectorial force */
1988             tx               = _mm_mul_pd(fscal,dx00);
1989             ty               = _mm_mul_pd(fscal,dy00);
1990             tz               = _mm_mul_pd(fscal,dz00);
1991
1992             /* Update vectorial force */
1993             fix0             = _mm_add_pd(fix0,tx);
1994             fiy0             = _mm_add_pd(fiy0,ty);
1995             fiz0             = _mm_add_pd(fiz0,tz);
1996
1997             fjx0             = _mm_add_pd(fjx0,tx);
1998             fjy0             = _mm_add_pd(fjy0,ty);
1999             fjz0             = _mm_add_pd(fjz0,tz);
2000
2001             /**************************
2002              * CALCULATE INTERACTIONS *
2003              **************************/
2004
2005             r01              = _mm_mul_pd(rsq01,rinv01);
2006
2007             /* EWALD ELECTROSTATICS */
2008
2009             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2010             ewrt             = _mm_mul_pd(r01,ewtabscale);
2011             ewitab           = _mm_cvttpd_epi32(ewrt);
2012             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2013             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2014             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2015             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2016
2017             fscal            = felec;
2018
2019             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2020
2021             /* Calculate temporary vectorial force */
2022             tx               = _mm_mul_pd(fscal,dx01);
2023             ty               = _mm_mul_pd(fscal,dy01);
2024             tz               = _mm_mul_pd(fscal,dz01);
2025
2026             /* Update vectorial force */
2027             fix0             = _mm_add_pd(fix0,tx);
2028             fiy0             = _mm_add_pd(fiy0,ty);
2029             fiz0             = _mm_add_pd(fiz0,tz);
2030
2031             fjx1             = _mm_add_pd(fjx1,tx);
2032             fjy1             = _mm_add_pd(fjy1,ty);
2033             fjz1             = _mm_add_pd(fjz1,tz);
2034
2035             /**************************
2036              * CALCULATE INTERACTIONS *
2037              **************************/
2038
2039             r02              = _mm_mul_pd(rsq02,rinv02);
2040
2041             /* EWALD ELECTROSTATICS */
2042
2043             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2044             ewrt             = _mm_mul_pd(r02,ewtabscale);
2045             ewitab           = _mm_cvttpd_epi32(ewrt);
2046             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2047             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2048             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2049             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2050
2051             fscal            = felec;
2052
2053             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2054
2055             /* Calculate temporary vectorial force */
2056             tx               = _mm_mul_pd(fscal,dx02);
2057             ty               = _mm_mul_pd(fscal,dy02);
2058             tz               = _mm_mul_pd(fscal,dz02);
2059
2060             /* Update vectorial force */
2061             fix0             = _mm_add_pd(fix0,tx);
2062             fiy0             = _mm_add_pd(fiy0,ty);
2063             fiz0             = _mm_add_pd(fiz0,tz);
2064
2065             fjx2             = _mm_add_pd(fjx2,tx);
2066             fjy2             = _mm_add_pd(fjy2,ty);
2067             fjz2             = _mm_add_pd(fjz2,tz);
2068
2069             /**************************
2070              * CALCULATE INTERACTIONS *
2071              **************************/
2072
2073             r10              = _mm_mul_pd(rsq10,rinv10);
2074
2075             /* EWALD ELECTROSTATICS */
2076
2077             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2078             ewrt             = _mm_mul_pd(r10,ewtabscale);
2079             ewitab           = _mm_cvttpd_epi32(ewrt);
2080             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2081             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2082             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2083             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2084
2085             fscal            = felec;
2086
2087             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2088
2089             /* Calculate temporary vectorial force */
2090             tx               = _mm_mul_pd(fscal,dx10);
2091             ty               = _mm_mul_pd(fscal,dy10);
2092             tz               = _mm_mul_pd(fscal,dz10);
2093
2094             /* Update vectorial force */
2095             fix1             = _mm_add_pd(fix1,tx);
2096             fiy1             = _mm_add_pd(fiy1,ty);
2097             fiz1             = _mm_add_pd(fiz1,tz);
2098
2099             fjx0             = _mm_add_pd(fjx0,tx);
2100             fjy0             = _mm_add_pd(fjy0,ty);
2101             fjz0             = _mm_add_pd(fjz0,tz);
2102
2103             /**************************
2104              * CALCULATE INTERACTIONS *
2105              **************************/
2106
2107             r11              = _mm_mul_pd(rsq11,rinv11);
2108
2109             /* EWALD ELECTROSTATICS */
2110
2111             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2112             ewrt             = _mm_mul_pd(r11,ewtabscale);
2113             ewitab           = _mm_cvttpd_epi32(ewrt);
2114             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2115             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2116             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2117             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2118
2119             fscal            = felec;
2120
2121             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2122
2123             /* Calculate temporary vectorial force */
2124             tx               = _mm_mul_pd(fscal,dx11);
2125             ty               = _mm_mul_pd(fscal,dy11);
2126             tz               = _mm_mul_pd(fscal,dz11);
2127
2128             /* Update vectorial force */
2129             fix1             = _mm_add_pd(fix1,tx);
2130             fiy1             = _mm_add_pd(fiy1,ty);
2131             fiz1             = _mm_add_pd(fiz1,tz);
2132
2133             fjx1             = _mm_add_pd(fjx1,tx);
2134             fjy1             = _mm_add_pd(fjy1,ty);
2135             fjz1             = _mm_add_pd(fjz1,tz);
2136
2137             /**************************
2138              * CALCULATE INTERACTIONS *
2139              **************************/
2140
2141             r12              = _mm_mul_pd(rsq12,rinv12);
2142
2143             /* EWALD ELECTROSTATICS */
2144
2145             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2146             ewrt             = _mm_mul_pd(r12,ewtabscale);
2147             ewitab           = _mm_cvttpd_epi32(ewrt);
2148             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2149             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2150             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2151             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2152
2153             fscal            = felec;
2154
2155             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2156
2157             /* Calculate temporary vectorial force */
2158             tx               = _mm_mul_pd(fscal,dx12);
2159             ty               = _mm_mul_pd(fscal,dy12);
2160             tz               = _mm_mul_pd(fscal,dz12);
2161
2162             /* Update vectorial force */
2163             fix1             = _mm_add_pd(fix1,tx);
2164             fiy1             = _mm_add_pd(fiy1,ty);
2165             fiz1             = _mm_add_pd(fiz1,tz);
2166
2167             fjx2             = _mm_add_pd(fjx2,tx);
2168             fjy2             = _mm_add_pd(fjy2,ty);
2169             fjz2             = _mm_add_pd(fjz2,tz);
2170
2171             /**************************
2172              * CALCULATE INTERACTIONS *
2173              **************************/
2174
2175             r20              = _mm_mul_pd(rsq20,rinv20);
2176
2177             /* EWALD ELECTROSTATICS */
2178
2179             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2180             ewrt             = _mm_mul_pd(r20,ewtabscale);
2181             ewitab           = _mm_cvttpd_epi32(ewrt);
2182             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2183             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2184             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2185             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2186
2187             fscal            = felec;
2188
2189             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2190
2191             /* Calculate temporary vectorial force */
2192             tx               = _mm_mul_pd(fscal,dx20);
2193             ty               = _mm_mul_pd(fscal,dy20);
2194             tz               = _mm_mul_pd(fscal,dz20);
2195
2196             /* Update vectorial force */
2197             fix2             = _mm_add_pd(fix2,tx);
2198             fiy2             = _mm_add_pd(fiy2,ty);
2199             fiz2             = _mm_add_pd(fiz2,tz);
2200
2201             fjx0             = _mm_add_pd(fjx0,tx);
2202             fjy0             = _mm_add_pd(fjy0,ty);
2203             fjz0             = _mm_add_pd(fjz0,tz);
2204
2205             /**************************
2206              * CALCULATE INTERACTIONS *
2207              **************************/
2208
2209             r21              = _mm_mul_pd(rsq21,rinv21);
2210
2211             /* EWALD ELECTROSTATICS */
2212
2213             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2214             ewrt             = _mm_mul_pd(r21,ewtabscale);
2215             ewitab           = _mm_cvttpd_epi32(ewrt);
2216             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2217             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2218             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2219             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2220
2221             fscal            = felec;
2222
2223             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2224
2225             /* Calculate temporary vectorial force */
2226             tx               = _mm_mul_pd(fscal,dx21);
2227             ty               = _mm_mul_pd(fscal,dy21);
2228             tz               = _mm_mul_pd(fscal,dz21);
2229
2230             /* Update vectorial force */
2231             fix2             = _mm_add_pd(fix2,tx);
2232             fiy2             = _mm_add_pd(fiy2,ty);
2233             fiz2             = _mm_add_pd(fiz2,tz);
2234
2235             fjx1             = _mm_add_pd(fjx1,tx);
2236             fjy1             = _mm_add_pd(fjy1,ty);
2237             fjz1             = _mm_add_pd(fjz1,tz);
2238
2239             /**************************
2240              * CALCULATE INTERACTIONS *
2241              **************************/
2242
2243             r22              = _mm_mul_pd(rsq22,rinv22);
2244
2245             /* EWALD ELECTROSTATICS */
2246
2247             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2248             ewrt             = _mm_mul_pd(r22,ewtabscale);
2249             ewitab           = _mm_cvttpd_epi32(ewrt);
2250             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2251             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2252             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2253             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2254
2255             fscal            = felec;
2256
2257             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2258
2259             /* Calculate temporary vectorial force */
2260             tx               = _mm_mul_pd(fscal,dx22);
2261             ty               = _mm_mul_pd(fscal,dy22);
2262             tz               = _mm_mul_pd(fscal,dz22);
2263
2264             /* Update vectorial force */
2265             fix2             = _mm_add_pd(fix2,tx);
2266             fiy2             = _mm_add_pd(fiy2,ty);
2267             fiz2             = _mm_add_pd(fiz2,tz);
2268
2269             fjx2             = _mm_add_pd(fjx2,tx);
2270             fjy2             = _mm_add_pd(fjy2,ty);
2271             fjz2             = _mm_add_pd(fjz2,tz);
2272
2273             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2274
2275             /* Inner loop uses 350 flops */
2276         }
2277
2278         /* End of innermost loop */
2279
2280         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2281                                               f+i_coord_offset,fshift+i_shift_offset);
2282
2283         /* Increment number of inner iterations */
2284         inneriter                  += j_index_end - j_index_start;
2285
2286         /* Outer loop uses 18 flops */
2287     }
2288
2289     /* Increment number of outer iterations */
2290     outeriter        += nri;
2291
2292     /* Update outer/inner flops */
2293
2294     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*350);
2295 }