src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecEw_VdwLJEw_GeomW4W4_sse2_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse2_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_sse2_double.h"
  48 #include "kernelutil_x86_sse2_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJEw_GeomW4W4_VF_sse2_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            LJEwald
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEw_VdwLJEw_GeomW4W4_VF_sse2_double
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwioffset1;
  83     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  84     int              vdwioffset2;
  85     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  86     int              vdwioffset3;
  87     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  88     int              vdwjidx0A,vdwjidx0B;
  89     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  90     int              vdwjidx1A,vdwjidx1B;
  91     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  92     int              vdwjidx2A,vdwjidx2B;
  93     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  94     int              vdwjidx3A,vdwjidx3B;
  95     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  96     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  97     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  98     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  99     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 100     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 101     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 102     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 103     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 104     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 105     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 106     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 107     real             *charge;
 108     int              nvdwtype;
 109     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 110     int              *vdwtype;
 111     real             *vdwparam;
 112     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 113     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 114     __m128d           c6grid_00;
 115     __m128d           c6grid_11;
 116     __m128d           c6grid_12;
 117     __m128d           c6grid_13;
 118     __m128d           c6grid_21;
 119     __m128d           c6grid_22;
 120     __m128d           c6grid_23;
 121     __m128d           c6grid_31;
 122     __m128d           c6grid_32;
 123     __m128d           c6grid_33;
 124     __m128d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 125     real             *vdwgridparam;
 126     __m128d           one_half = _mm_set1_pd(0.5);
 127     __m128d           minus_one = _mm_set1_pd(-1.0);
 128     __m128i          ewitab;
 129     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 130     real             *ewtab;
 131     __m128d          dummy_mask,cutoff_mask;
 132     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 133     __m128d          one     = _mm_set1_pd(1.0);
 134     __m128d          two     = _mm_set1_pd(2.0);
 135     x                = xx[0];
 136     f                = ff[0];
 137
 138     nri              = nlist->nri;
 139     iinr             = nlist->iinr;
 140     jindex           = nlist->jindex;
 141     jjnr             = nlist->jjnr;
 142     shiftidx         = nlist->shift;
 143     gid              = nlist->gid;
 144     shiftvec         = fr->shift_vec[0];
 145     fshift           = fr->fshift[0];
 146     facel            = _mm_set1_pd(fr->epsfac);
 147     charge           = mdatoms->chargeA;
 148     nvdwtype         = fr->ntype;
 149     vdwparam         = fr->nbfp;
 150     vdwtype          = mdatoms->typeA;
 151     vdwgridparam     = fr->ljpme_c6grid;
 152     sh_lj_ewald      = _mm_set1_pd(fr->ic->sh_lj_ewald);
 153     ewclj            = _mm_set1_pd(fr->ewaldcoeff_lj);
 154     ewclj2           = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
 155
 156     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 157     ewtab            = fr->ic->tabq_coul_FDV0;
 158     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 159     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 160
 161     /* Setup water-specific parameters */
 162     inr              = nlist->iinr[0];
 163     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 164     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 165     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 166     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 167
 168     jq1              = _mm_set1_pd(charge[inr+1]);
 169     jq2              = _mm_set1_pd(charge[inr+2]);
 170     jq3              = _mm_set1_pd(charge[inr+3]);
 171     vdwjidx0A        = 2*vdwtype[inr+0];
 172     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 173     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 174     c6grid_00        = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
 175     qq11             = _mm_mul_pd(iq1,jq1);
 176     qq12             = _mm_mul_pd(iq1,jq2);
 177     qq13             = _mm_mul_pd(iq1,jq3);
 178     qq21             = _mm_mul_pd(iq2,jq1);
 179     qq22             = _mm_mul_pd(iq2,jq2);
 180     qq23             = _mm_mul_pd(iq2,jq3);
 181     qq31             = _mm_mul_pd(iq3,jq1);
 182     qq32             = _mm_mul_pd(iq3,jq2);
 183     qq33             = _mm_mul_pd(iq3,jq3);
 184
 185     /* Avoid stupid compiler warnings */
 186     jnrA = jnrB = 0;
 187     j_coord_offsetA = 0;
 188     j_coord_offsetB = 0;
 189
 190     outeriter        = 0;
 191     inneriter        = 0;
 192
 193     /* Start outer loop over neighborlists */
 194     for(iidx=0; iidx<nri; iidx++)
 195     {
 196         /* Load shift vector for this list */
 197         i_shift_offset   = DIM*shiftidx[iidx];
 198
 199         /* Load limits for loop over neighbors */
 200         j_index_start    = jindex[iidx];
 201         j_index_end      = jindex[iidx+1];
 202
 203         /* Get outer coordinate index */
 204         inr              = iinr[iidx];
 205         i_coord_offset   = DIM*inr;
 206
 207         /* Load i particle coords and add shift vector */
 208         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 209                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 210
 211         fix0             = _mm_setzero_pd();
 212         fiy0             = _mm_setzero_pd();
 213         fiz0             = _mm_setzero_pd();
 214         fix1             = _mm_setzero_pd();
 215         fiy1             = _mm_setzero_pd();
 216         fiz1             = _mm_setzero_pd();
 217         fix2             = _mm_setzero_pd();
 218         fiy2             = _mm_setzero_pd();
 219         fiz2             = _mm_setzero_pd();
 220         fix3             = _mm_setzero_pd();
 221         fiy3             = _mm_setzero_pd();
 222         fiz3             = _mm_setzero_pd();
 223
 224         /* Reset potential sums */
 225         velecsum         = _mm_setzero_pd();
 226         vvdwsum          = _mm_setzero_pd();
 227
 228         /* Start inner kernel loop */
 229         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 230         {
 231
 232             /* Get j neighbor index, and coordinate index */
 233             jnrA             = jjnr[jidx];
 234             jnrB             = jjnr[jidx+1];
 235             j_coord_offsetA  = DIM*jnrA;
 236             j_coord_offsetB  = DIM*jnrB;
 237
 238             /* load j atom coordinates */
 239             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 240                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 241                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 242
 243             /* Calculate displacement vector */
 244             dx00             = _mm_sub_pd(ix0,jx0);
 245             dy00             = _mm_sub_pd(iy0,jy0);
 246             dz00             = _mm_sub_pd(iz0,jz0);
 247             dx11             = _mm_sub_pd(ix1,jx1);
 248             dy11             = _mm_sub_pd(iy1,jy1);
 249             dz11             = _mm_sub_pd(iz1,jz1);
 250             dx12             = _mm_sub_pd(ix1,jx2);
 251             dy12             = _mm_sub_pd(iy1,jy2);
 252             dz12             = _mm_sub_pd(iz1,jz2);
 253             dx13             = _mm_sub_pd(ix1,jx3);
 254             dy13             = _mm_sub_pd(iy1,jy3);
 255             dz13             = _mm_sub_pd(iz1,jz3);
 256             dx21             = _mm_sub_pd(ix2,jx1);
 257             dy21             = _mm_sub_pd(iy2,jy1);
 258             dz21             = _mm_sub_pd(iz2,jz1);
 259             dx22             = _mm_sub_pd(ix2,jx2);
 260             dy22             = _mm_sub_pd(iy2,jy2);
 261             dz22             = _mm_sub_pd(iz2,jz2);
 262             dx23             = _mm_sub_pd(ix2,jx3);
 263             dy23             = _mm_sub_pd(iy2,jy3);
 264             dz23             = _mm_sub_pd(iz2,jz3);
 265             dx31             = _mm_sub_pd(ix3,jx1);
 266             dy31             = _mm_sub_pd(iy3,jy1);
 267             dz31             = _mm_sub_pd(iz3,jz1);
 268             dx32             = _mm_sub_pd(ix3,jx2);
 269             dy32             = _mm_sub_pd(iy3,jy2);
 270             dz32             = _mm_sub_pd(iz3,jz2);
 271             dx33             = _mm_sub_pd(ix3,jx3);
 272             dy33             = _mm_sub_pd(iy3,jy3);
 273             dz33             = _mm_sub_pd(iz3,jz3);
 274
 275             /* Calculate squared distance and things based on it */
 276             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 277             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 278             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 279             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 280             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 281             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 282             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 283             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 284             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 285             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 286
 287             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 288             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 289             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 290             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 291             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 292             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 293             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 294             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 295             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 296             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 297
 298             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 299             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 300             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 301             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 302             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 303             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 304             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 305             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 306             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 307             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 308
 309             fjx0             = _mm_setzero_pd();
 310             fjy0             = _mm_setzero_pd();
 311             fjz0             = _mm_setzero_pd();
 312             fjx1             = _mm_setzero_pd();
 313             fjy1             = _mm_setzero_pd();
 314             fjz1             = _mm_setzero_pd();
 315             fjx2             = _mm_setzero_pd();
 316             fjy2             = _mm_setzero_pd();
 317             fjz2             = _mm_setzero_pd();
 318             fjx3             = _mm_setzero_pd();
 319             fjy3             = _mm_setzero_pd();
 320             fjz3             = _mm_setzero_pd();
 321
 322             /**************************
 323              * CALCULATE INTERACTIONS *
 324              **************************/
 325
 326             r00              = _mm_mul_pd(rsq00,rinv00);
 327
 328             /* Analytical LJ-PME */
 329             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 330             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 331             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 332             exponent         = gmx_simd_exp_d(ewcljrsq);
 333             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 334             poly             = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 335             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 336             vvdw6            = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
 337             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 338             vvdw             = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
 339             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 340             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
 341
 342             /* Update potential sum for this i atom from the interaction with this j atom. */
 343             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 344
 345             fscal            = fvdw;
 346
 347             /* Calculate temporary vectorial force */
 348             tx               = _mm_mul_pd(fscal,dx00);
 349             ty               = _mm_mul_pd(fscal,dy00);
 350             tz               = _mm_mul_pd(fscal,dz00);
 351
 352             /* Update vectorial force */
 353             fix0             = _mm_add_pd(fix0,tx);
 354             fiy0             = _mm_add_pd(fiy0,ty);
 355             fiz0             = _mm_add_pd(fiz0,tz);
 356
 357             fjx0             = _mm_add_pd(fjx0,tx);
 358             fjy0             = _mm_add_pd(fjy0,ty);
 359             fjz0             = _mm_add_pd(fjz0,tz);
 360
 361             /**************************
 362              * CALCULATE INTERACTIONS *
 363              **************************/
 364
 365             r11              = _mm_mul_pd(rsq11,rinv11);
 366
 367             /* EWALD ELECTROSTATICS */
 368
 369             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 370             ewrt             = _mm_mul_pd(r11,ewtabscale);
 371             ewitab           = _mm_cvttpd_epi32(ewrt);
 372             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 373             ewitab           = _mm_slli_epi32(ewitab,2);
 374             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 375             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 376             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 377             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 378             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 379             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 380             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 381             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 382             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 383             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 384
 385             /* Update potential sum for this i atom from the interaction with this j atom. */
 386             velecsum         = _mm_add_pd(velecsum,velec);
 387
 388             fscal            = felec;
 389
 390             /* Calculate temporary vectorial force */
 391             tx               = _mm_mul_pd(fscal,dx11);
 392             ty               = _mm_mul_pd(fscal,dy11);
 393             tz               = _mm_mul_pd(fscal,dz11);
 394
 395             /* Update vectorial force */
 396             fix1             = _mm_add_pd(fix1,tx);
 397             fiy1             = _mm_add_pd(fiy1,ty);
 398             fiz1             = _mm_add_pd(fiz1,tz);
 399
 400             fjx1             = _mm_add_pd(fjx1,tx);
 401             fjy1             = _mm_add_pd(fjy1,ty);
 402             fjz1             = _mm_add_pd(fjz1,tz);
 403
 404             /**************************
 405              * CALCULATE INTERACTIONS *
 406              **************************/
 407
 408             r12              = _mm_mul_pd(rsq12,rinv12);
 409
 410             /* EWALD ELECTROSTATICS */
 411
 412             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 413             ewrt             = _mm_mul_pd(r12,ewtabscale);
 414             ewitab           = _mm_cvttpd_epi32(ewrt);
 415             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 416             ewitab           = _mm_slli_epi32(ewitab,2);
 417             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 418             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 419             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 420             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 421             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 422             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 423             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 424             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 425             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 426             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 427
 428             /* Update potential sum for this i atom from the interaction with this j atom. */
 429             velecsum         = _mm_add_pd(velecsum,velec);
 430
 431             fscal            = felec;
 432
 433             /* Calculate temporary vectorial force */
 434             tx               = _mm_mul_pd(fscal,dx12);
 435             ty               = _mm_mul_pd(fscal,dy12);
 436             tz               = _mm_mul_pd(fscal,dz12);
 437
 438             /* Update vectorial force */
 439             fix1             = _mm_add_pd(fix1,tx);
 440             fiy1             = _mm_add_pd(fiy1,ty);
 441             fiz1             = _mm_add_pd(fiz1,tz);
 442
 443             fjx2             = _mm_add_pd(fjx2,tx);
 444             fjy2             = _mm_add_pd(fjy2,ty);
 445             fjz2             = _mm_add_pd(fjz2,tz);
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             r13              = _mm_mul_pd(rsq13,rinv13);
 452
 453             /* EWALD ELECTROSTATICS */
 454
 455             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 456             ewrt             = _mm_mul_pd(r13,ewtabscale);
 457             ewitab           = _mm_cvttpd_epi32(ewrt);
 458             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 459             ewitab           = _mm_slli_epi32(ewitab,2);
 460             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 461             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 462             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 463             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 464             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 465             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 466             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 467             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 468             velec            = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
 469             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
 470
 471             /* Update potential sum for this i atom from the interaction with this j atom. */
 472             velecsum         = _mm_add_pd(velecsum,velec);
 473
 474             fscal            = felec;
 475
 476             /* Calculate temporary vectorial force */
 477             tx               = _mm_mul_pd(fscal,dx13);
 478             ty               = _mm_mul_pd(fscal,dy13);
 479             tz               = _mm_mul_pd(fscal,dz13);
 480
 481             /* Update vectorial force */
 482             fix1             = _mm_add_pd(fix1,tx);
 483             fiy1             = _mm_add_pd(fiy1,ty);
 484             fiz1             = _mm_add_pd(fiz1,tz);
 485
 486             fjx3             = _mm_add_pd(fjx3,tx);
 487             fjy3             = _mm_add_pd(fjy3,ty);
 488             fjz3             = _mm_add_pd(fjz3,tz);
 489
 490             /**************************
 491              * CALCULATE INTERACTIONS *
 492              **************************/
 493
 494             r21              = _mm_mul_pd(rsq21,rinv21);
 495
 496             /* EWALD ELECTROSTATICS */
 497
 498             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 499             ewrt             = _mm_mul_pd(r21,ewtabscale);
 500             ewitab           = _mm_cvttpd_epi32(ewrt);
 501             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 502             ewitab           = _mm_slli_epi32(ewitab,2);
 503             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 504             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 505             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 506             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 507             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 508             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 509             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 510             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 511             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 512             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 513
 514             /* Update potential sum for this i atom from the interaction with this j atom. */
 515             velecsum         = _mm_add_pd(velecsum,velec);
 516
 517             fscal            = felec;
 518
 519             /* Calculate temporary vectorial force */
 520             tx               = _mm_mul_pd(fscal,dx21);
 521             ty               = _mm_mul_pd(fscal,dy21);
 522             tz               = _mm_mul_pd(fscal,dz21);
 523
 524             /* Update vectorial force */
 525             fix2             = _mm_add_pd(fix2,tx);
 526             fiy2             = _mm_add_pd(fiy2,ty);
 527             fiz2             = _mm_add_pd(fiz2,tz);
 528
 529             fjx1             = _mm_add_pd(fjx1,tx);
 530             fjy1             = _mm_add_pd(fjy1,ty);
 531             fjz1             = _mm_add_pd(fjz1,tz);
 532
 533             /**************************
 534              * CALCULATE INTERACTIONS *
 535              **************************/
 536
 537             r22              = _mm_mul_pd(rsq22,rinv22);
 538
 539             /* EWALD ELECTROSTATICS */
 540
 541             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 542             ewrt             = _mm_mul_pd(r22,ewtabscale);
 543             ewitab           = _mm_cvttpd_epi32(ewrt);
 544             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 545             ewitab           = _mm_slli_epi32(ewitab,2);
 546             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 547             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 548             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 549             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 550             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 551             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 552             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 553             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 554             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 555             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 556
 557             /* Update potential sum for this i atom from the interaction with this j atom. */
 558             velecsum         = _mm_add_pd(velecsum,velec);
 559
 560             fscal            = felec;
 561
 562             /* Calculate temporary vectorial force */
 563             tx               = _mm_mul_pd(fscal,dx22);
 564             ty               = _mm_mul_pd(fscal,dy22);
 565             tz               = _mm_mul_pd(fscal,dz22);
 566
 567             /* Update vectorial force */
 568             fix2             = _mm_add_pd(fix2,tx);
 569             fiy2             = _mm_add_pd(fiy2,ty);
 570             fiz2             = _mm_add_pd(fiz2,tz);
 571
 572             fjx2             = _mm_add_pd(fjx2,tx);
 573             fjy2             = _mm_add_pd(fjy2,ty);
 574             fjz2             = _mm_add_pd(fjz2,tz);
 575
 576             /**************************
 577              * CALCULATE INTERACTIONS *
 578              **************************/
 579
 580             r23              = _mm_mul_pd(rsq23,rinv23);
 581
 582             /* EWALD ELECTROSTATICS */
 583
 584             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 585             ewrt             = _mm_mul_pd(r23,ewtabscale);
 586             ewitab           = _mm_cvttpd_epi32(ewrt);
 587             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 588             ewitab           = _mm_slli_epi32(ewitab,2);
 589             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 590             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 591             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 592             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 593             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 594             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 595             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 596             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 597             velec            = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
 598             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
 599
 600             /* Update potential sum for this i atom from the interaction with this j atom. */
 601             velecsum         = _mm_add_pd(velecsum,velec);
 602
 603             fscal            = felec;
 604
 605             /* Calculate temporary vectorial force */
 606             tx               = _mm_mul_pd(fscal,dx23);
 607             ty               = _mm_mul_pd(fscal,dy23);
 608             tz               = _mm_mul_pd(fscal,dz23);
 609
 610             /* Update vectorial force */
 611             fix2             = _mm_add_pd(fix2,tx);
 612             fiy2             = _mm_add_pd(fiy2,ty);
 613             fiz2             = _mm_add_pd(fiz2,tz);
 614
 615             fjx3             = _mm_add_pd(fjx3,tx);
 616             fjy3             = _mm_add_pd(fjy3,ty);
 617             fjz3             = _mm_add_pd(fjz3,tz);
 618
 619             /**************************
 620              * CALCULATE INTERACTIONS *
 621              **************************/
 622
 623             r31              = _mm_mul_pd(rsq31,rinv31);
 624
 625             /* EWALD ELECTROSTATICS */
 626
 627             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 628             ewrt             = _mm_mul_pd(r31,ewtabscale);
 629             ewitab           = _mm_cvttpd_epi32(ewrt);
 630             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 631             ewitab           = _mm_slli_epi32(ewitab,2);
 632             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 633             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 634             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 635             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 636             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 637             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 638             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 639             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 640             velec            = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
 641             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
 642
 643             /* Update potential sum for this i atom from the interaction with this j atom. */
 644             velecsum         = _mm_add_pd(velecsum,velec);
 645
 646             fscal            = felec;
 647
 648             /* Calculate temporary vectorial force */
 649             tx               = _mm_mul_pd(fscal,dx31);
 650             ty               = _mm_mul_pd(fscal,dy31);
 651             tz               = _mm_mul_pd(fscal,dz31);
 652
 653             /* Update vectorial force */
 654             fix3             = _mm_add_pd(fix3,tx);
 655             fiy3             = _mm_add_pd(fiy3,ty);
 656             fiz3             = _mm_add_pd(fiz3,tz);
 657
 658             fjx1             = _mm_add_pd(fjx1,tx);
 659             fjy1             = _mm_add_pd(fjy1,ty);
 660             fjz1             = _mm_add_pd(fjz1,tz);
 661
 662             /**************************
 663              * CALCULATE INTERACTIONS *
 664              **************************/
 665
 666             r32              = _mm_mul_pd(rsq32,rinv32);
 667
 668             /* EWALD ELECTROSTATICS */
 669
 670             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 671             ewrt             = _mm_mul_pd(r32,ewtabscale);
 672             ewitab           = _mm_cvttpd_epi32(ewrt);
 673             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 674             ewitab           = _mm_slli_epi32(ewitab,2);
 675             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 676             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 677             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 678             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 679             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 680             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 681             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 682             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 683             velec            = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
 684             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
 685
 686             /* Update potential sum for this i atom from the interaction with this j atom. */
 687             velecsum         = _mm_add_pd(velecsum,velec);
 688
 689             fscal            = felec;
 690
 691             /* Calculate temporary vectorial force */
 692             tx               = _mm_mul_pd(fscal,dx32);
 693             ty               = _mm_mul_pd(fscal,dy32);
 694             tz               = _mm_mul_pd(fscal,dz32);
 695
 696             /* Update vectorial force */
 697             fix3             = _mm_add_pd(fix3,tx);
 698             fiy3             = _mm_add_pd(fiy3,ty);
 699             fiz3             = _mm_add_pd(fiz3,tz);
 700
 701             fjx2             = _mm_add_pd(fjx2,tx);
 702             fjy2             = _mm_add_pd(fjy2,ty);
 703             fjz2             = _mm_add_pd(fjz2,tz);
 704
 705             /**************************
 706              * CALCULATE INTERACTIONS *
 707              **************************/
 708
 709             r33              = _mm_mul_pd(rsq33,rinv33);
 710
 711             /* EWALD ELECTROSTATICS */
 712
 713             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 714             ewrt             = _mm_mul_pd(r33,ewtabscale);
 715             ewitab           = _mm_cvttpd_epi32(ewrt);
 716             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 717             ewitab           = _mm_slli_epi32(ewitab,2);
 718             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 719             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 720             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 721             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 722             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 723             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 724             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 725             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 726             velec            = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
 727             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
 728
 729             /* Update potential sum for this i atom from the interaction with this j atom. */
 730             velecsum         = _mm_add_pd(velecsum,velec);
 731
 732             fscal            = felec;
 733
 734             /* Calculate temporary vectorial force */
 735             tx               = _mm_mul_pd(fscal,dx33);
 736             ty               = _mm_mul_pd(fscal,dy33);
 737             tz               = _mm_mul_pd(fscal,dz33);
 738
 739             /* Update vectorial force */
 740             fix3             = _mm_add_pd(fix3,tx);
 741             fiy3             = _mm_add_pd(fiy3,ty);
 742             fiz3             = _mm_add_pd(fiz3,tz);
 743
 744             fjx3             = _mm_add_pd(fjx3,tx);
 745             fjy3             = _mm_add_pd(fjy3,ty);
 746             fjz3             = _mm_add_pd(fjz3,tz);
 747
 748             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 749
 750             /* Inner loop uses 423 flops */
 751         }
 752
 753         if(jidx<j_index_end)
 754         {
 755
 756             jnrA             = jjnr[jidx];
 757             j_coord_offsetA  = DIM*jnrA;
 758
 759             /* load j atom coordinates */
 760             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 761                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 762                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 763
 764             /* Calculate displacement vector */
 765             dx00             = _mm_sub_pd(ix0,jx0);
 766             dy00             = _mm_sub_pd(iy0,jy0);
 767             dz00             = _mm_sub_pd(iz0,jz0);
 768             dx11             = _mm_sub_pd(ix1,jx1);
 769             dy11             = _mm_sub_pd(iy1,jy1);
 770             dz11             = _mm_sub_pd(iz1,jz1);
 771             dx12             = _mm_sub_pd(ix1,jx2);
 772             dy12             = _mm_sub_pd(iy1,jy2);
 773             dz12             = _mm_sub_pd(iz1,jz2);
 774             dx13             = _mm_sub_pd(ix1,jx3);
 775             dy13             = _mm_sub_pd(iy1,jy3);
 776             dz13             = _mm_sub_pd(iz1,jz3);
 777             dx21             = _mm_sub_pd(ix2,jx1);
 778             dy21             = _mm_sub_pd(iy2,jy1);
 779             dz21             = _mm_sub_pd(iz2,jz1);
 780             dx22             = _mm_sub_pd(ix2,jx2);
 781             dy22             = _mm_sub_pd(iy2,jy2);
 782             dz22             = _mm_sub_pd(iz2,jz2);
 783             dx23             = _mm_sub_pd(ix2,jx3);
 784             dy23             = _mm_sub_pd(iy2,jy3);
 785             dz23             = _mm_sub_pd(iz2,jz3);
 786             dx31             = _mm_sub_pd(ix3,jx1);
 787             dy31             = _mm_sub_pd(iy3,jy1);
 788             dz31             = _mm_sub_pd(iz3,jz1);
 789             dx32             = _mm_sub_pd(ix3,jx2);
 790             dy32             = _mm_sub_pd(iy3,jy2);
 791             dz32             = _mm_sub_pd(iz3,jz2);
 792             dx33             = _mm_sub_pd(ix3,jx3);
 793             dy33             = _mm_sub_pd(iy3,jy3);
 794             dz33             = _mm_sub_pd(iz3,jz3);
 795
 796             /* Calculate squared distance and things based on it */
 797             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 798             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 799             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 800             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 801             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 802             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 803             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 804             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 805             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 806             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 807
 808             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 809             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 810             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 811             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 812             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 813             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 814             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 815             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 816             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 817             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 818
 819             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 820             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 821             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 822             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 823             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 824             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 825             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 826             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 827             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 828             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 829
 830             fjx0             = _mm_setzero_pd();
 831             fjy0             = _mm_setzero_pd();
 832             fjz0             = _mm_setzero_pd();
 833             fjx1             = _mm_setzero_pd();
 834             fjy1             = _mm_setzero_pd();
 835             fjz1             = _mm_setzero_pd();
 836             fjx2             = _mm_setzero_pd();
 837             fjy2             = _mm_setzero_pd();
 838             fjz2             = _mm_setzero_pd();
 839             fjx3             = _mm_setzero_pd();
 840             fjy3             = _mm_setzero_pd();
 841             fjz3             = _mm_setzero_pd();
 842
 843             /**************************
 844              * CALCULATE INTERACTIONS *
 845              **************************/
 846
 847             r00              = _mm_mul_pd(rsq00,rinv00);
 848
 849             /* Analytical LJ-PME */
 850             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 851             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 852             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 853             exponent         = gmx_simd_exp_d(ewcljrsq);
 854             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 855             poly             = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
 856             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 857             vvdw6            = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
 858             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 859             vvdw             = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
 860             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 861             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
 862
 863             /* Update potential sum for this i atom from the interaction with this j atom. */
 864             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 865             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 866
 867             fscal            = fvdw;
 868
 869             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 870
 871             /* Calculate temporary vectorial force */
 872             tx               = _mm_mul_pd(fscal,dx00);
 873             ty               = _mm_mul_pd(fscal,dy00);
 874             tz               = _mm_mul_pd(fscal,dz00);
 875
 876             /* Update vectorial force */
 877             fix0             = _mm_add_pd(fix0,tx);
 878             fiy0             = _mm_add_pd(fiy0,ty);
 879             fiz0             = _mm_add_pd(fiz0,tz);
 880
 881             fjx0             = _mm_add_pd(fjx0,tx);
 882             fjy0             = _mm_add_pd(fjy0,ty);
 883             fjz0             = _mm_add_pd(fjz0,tz);
 884
 885             /**************************
 886              * CALCULATE INTERACTIONS *
 887              **************************/
 888
 889             r11              = _mm_mul_pd(rsq11,rinv11);
 890
 891             /* EWALD ELECTROSTATICS */
 892
 893             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 894             ewrt             = _mm_mul_pd(r11,ewtabscale);
 895             ewitab           = _mm_cvttpd_epi32(ewrt);
 896             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 897             ewitab           = _mm_slli_epi32(ewitab,2);
 898             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 899             ewtabD           = _mm_setzero_pd();
 900             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 901             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 902             ewtabFn          = _mm_setzero_pd();
 903             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 904             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 905             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 906             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 907             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 908
 909             /* Update potential sum for this i atom from the interaction with this j atom. */
 910             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 911             velecsum         = _mm_add_pd(velecsum,velec);
 912
 913             fscal            = felec;
 914
 915             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 916
 917             /* Calculate temporary vectorial force */
 918             tx               = _mm_mul_pd(fscal,dx11);
 919             ty               = _mm_mul_pd(fscal,dy11);
 920             tz               = _mm_mul_pd(fscal,dz11);
 921
 922             /* Update vectorial force */
 923             fix1             = _mm_add_pd(fix1,tx);
 924             fiy1             = _mm_add_pd(fiy1,ty);
 925             fiz1             = _mm_add_pd(fiz1,tz);
 926
 927             fjx1             = _mm_add_pd(fjx1,tx);
 928             fjy1             = _mm_add_pd(fjy1,ty);
 929             fjz1             = _mm_add_pd(fjz1,tz);
 930
 931             /**************************
 932              * CALCULATE INTERACTIONS *
 933              **************************/
 934
 935             r12              = _mm_mul_pd(rsq12,rinv12);
 936
 937             /* EWALD ELECTROSTATICS */
 938
 939             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 940             ewrt             = _mm_mul_pd(r12,ewtabscale);
 941             ewitab           = _mm_cvttpd_epi32(ewrt);
 942             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 943             ewitab           = _mm_slli_epi32(ewitab,2);
 944             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 945             ewtabD           = _mm_setzero_pd();
 946             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 947             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 948             ewtabFn          = _mm_setzero_pd();
 949             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 950             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 951             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 952             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 953             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 954
 955             /* Update potential sum for this i atom from the interaction with this j atom. */
 956             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 957             velecsum         = _mm_add_pd(velecsum,velec);
 958
 959             fscal            = felec;
 960
 961             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 962
 963             /* Calculate temporary vectorial force */
 964             tx               = _mm_mul_pd(fscal,dx12);
 965             ty               = _mm_mul_pd(fscal,dy12);
 966             tz               = _mm_mul_pd(fscal,dz12);
 967
 968             /* Update vectorial force */
 969             fix1             = _mm_add_pd(fix1,tx);
 970             fiy1             = _mm_add_pd(fiy1,ty);
 971             fiz1             = _mm_add_pd(fiz1,tz);
 972
 973             fjx2             = _mm_add_pd(fjx2,tx);
 974             fjy2             = _mm_add_pd(fjy2,ty);
 975             fjz2             = _mm_add_pd(fjz2,tz);
 976
 977             /**************************
 978              * CALCULATE INTERACTIONS *
 979              **************************/
 980
 981             r13              = _mm_mul_pd(rsq13,rinv13);
 982
 983             /* EWALD ELECTROSTATICS */
 984
 985             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 986             ewrt             = _mm_mul_pd(r13,ewtabscale);
 987             ewitab           = _mm_cvttpd_epi32(ewrt);
 988             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 989             ewitab           = _mm_slli_epi32(ewitab,2);
 990             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 991             ewtabD           = _mm_setzero_pd();
 992             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 993             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 994             ewtabFn          = _mm_setzero_pd();
 995             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 996             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 997             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 998             velec            = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
 999             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1000
1001             /* Update potential sum for this i atom from the interaction with this j atom. */
1002             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1003             velecsum         = _mm_add_pd(velecsum,velec);
1004
1005             fscal            = felec;
1006
1007             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1008
1009             /* Calculate temporary vectorial force */
1010             tx               = _mm_mul_pd(fscal,dx13);
1011             ty               = _mm_mul_pd(fscal,dy13);
1012             tz               = _mm_mul_pd(fscal,dz13);
1013
1014             /* Update vectorial force */
1015             fix1             = _mm_add_pd(fix1,tx);
1016             fiy1             = _mm_add_pd(fiy1,ty);
1017             fiz1             = _mm_add_pd(fiz1,tz);
1018
1019             fjx3             = _mm_add_pd(fjx3,tx);
1020             fjy3             = _mm_add_pd(fjy3,ty);
1021             fjz3             = _mm_add_pd(fjz3,tz);
1022
1023             /**************************
1024              * CALCULATE INTERACTIONS *
1025              **************************/
1026
1027             r21              = _mm_mul_pd(rsq21,rinv21);
1028
1029             /* EWALD ELECTROSTATICS */
1030
1031             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1032             ewrt             = _mm_mul_pd(r21,ewtabscale);
1033             ewitab           = _mm_cvttpd_epi32(ewrt);
1034             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1035             ewitab           = _mm_slli_epi32(ewitab,2);
1036             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1037             ewtabD           = _mm_setzero_pd();
1038             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1039             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1040             ewtabFn          = _mm_setzero_pd();
1041             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1042             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1043             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1044             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1045             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1046
1047             /* Update potential sum for this i atom from the interaction with this j atom. */
1048             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1049             velecsum         = _mm_add_pd(velecsum,velec);
1050
1051             fscal            = felec;
1052
1053             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1054
1055             /* Calculate temporary vectorial force */
1056             tx               = _mm_mul_pd(fscal,dx21);
1057             ty               = _mm_mul_pd(fscal,dy21);
1058             tz               = _mm_mul_pd(fscal,dz21);
1059
1060             /* Update vectorial force */
1061             fix2             = _mm_add_pd(fix2,tx);
1062             fiy2             = _mm_add_pd(fiy2,ty);
1063             fiz2             = _mm_add_pd(fiz2,tz);
1064
1065             fjx1             = _mm_add_pd(fjx1,tx);
1066             fjy1             = _mm_add_pd(fjy1,ty);
1067             fjz1             = _mm_add_pd(fjz1,tz);
1068
1069             /**************************
1070              * CALCULATE INTERACTIONS *
1071              **************************/
1072
1073             r22              = _mm_mul_pd(rsq22,rinv22);
1074
1075             /* EWALD ELECTROSTATICS */
1076
1077             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1078             ewrt             = _mm_mul_pd(r22,ewtabscale);
1079             ewitab           = _mm_cvttpd_epi32(ewrt);
1080             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1081             ewitab           = _mm_slli_epi32(ewitab,2);
1082             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1083             ewtabD           = _mm_setzero_pd();
1084             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1085             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1086             ewtabFn          = _mm_setzero_pd();
1087             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1088             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1089             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1090             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1091             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1092
1093             /* Update potential sum for this i atom from the interaction with this j atom. */
1094             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1095             velecsum         = _mm_add_pd(velecsum,velec);
1096
1097             fscal            = felec;
1098
1099             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1100
1101             /* Calculate temporary vectorial force */
1102             tx               = _mm_mul_pd(fscal,dx22);
1103             ty               = _mm_mul_pd(fscal,dy22);
1104             tz               = _mm_mul_pd(fscal,dz22);
1105
1106             /* Update vectorial force */
1107             fix2             = _mm_add_pd(fix2,tx);
1108             fiy2             = _mm_add_pd(fiy2,ty);
1109             fiz2             = _mm_add_pd(fiz2,tz);
1110
1111             fjx2             = _mm_add_pd(fjx2,tx);
1112             fjy2             = _mm_add_pd(fjy2,ty);
1113             fjz2             = _mm_add_pd(fjz2,tz);
1114
1115             /**************************
1116              * CALCULATE INTERACTIONS *
1117              **************************/
1118
1119             r23              = _mm_mul_pd(rsq23,rinv23);
1120
1121             /* EWALD ELECTROSTATICS */
1122
1123             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1124             ewrt             = _mm_mul_pd(r23,ewtabscale);
1125             ewitab           = _mm_cvttpd_epi32(ewrt);
1126             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1127             ewitab           = _mm_slli_epi32(ewitab,2);
1128             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1129             ewtabD           = _mm_setzero_pd();
1130             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1131             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1132             ewtabFn          = _mm_setzero_pd();
1133             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1134             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1135             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1136             velec            = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
1137             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1138
1139             /* Update potential sum for this i atom from the interaction with this j atom. */
1140             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1141             velecsum         = _mm_add_pd(velecsum,velec);
1142
1143             fscal            = felec;
1144
1145             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1146
1147             /* Calculate temporary vectorial force */
1148             tx               = _mm_mul_pd(fscal,dx23);
1149             ty               = _mm_mul_pd(fscal,dy23);
1150             tz               = _mm_mul_pd(fscal,dz23);
1151
1152             /* Update vectorial force */
1153             fix2             = _mm_add_pd(fix2,tx);
1154             fiy2             = _mm_add_pd(fiy2,ty);
1155             fiz2             = _mm_add_pd(fiz2,tz);
1156
1157             fjx3             = _mm_add_pd(fjx3,tx);
1158             fjy3             = _mm_add_pd(fjy3,ty);
1159             fjz3             = _mm_add_pd(fjz3,tz);
1160
1161             /**************************
1162              * CALCULATE INTERACTIONS *
1163              **************************/
1164
1165             r31              = _mm_mul_pd(rsq31,rinv31);
1166
1167             /* EWALD ELECTROSTATICS */
1168
1169             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1170             ewrt             = _mm_mul_pd(r31,ewtabscale);
1171             ewitab           = _mm_cvttpd_epi32(ewrt);
1172             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1173             ewitab           = _mm_slli_epi32(ewitab,2);
1174             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1175             ewtabD           = _mm_setzero_pd();
1176             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1177             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1178             ewtabFn          = _mm_setzero_pd();
1179             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1180             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1181             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1182             velec            = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
1183             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1184
1185             /* Update potential sum for this i atom from the interaction with this j atom. */
1186             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1187             velecsum         = _mm_add_pd(velecsum,velec);
1188
1189             fscal            = felec;
1190
1191             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1192
1193             /* Calculate temporary vectorial force */
1194             tx               = _mm_mul_pd(fscal,dx31);
1195             ty               = _mm_mul_pd(fscal,dy31);
1196             tz               = _mm_mul_pd(fscal,dz31);
1197
1198             /* Update vectorial force */
1199             fix3             = _mm_add_pd(fix3,tx);
1200             fiy3             = _mm_add_pd(fiy3,ty);
1201             fiz3             = _mm_add_pd(fiz3,tz);
1202
1203             fjx1             = _mm_add_pd(fjx1,tx);
1204             fjy1             = _mm_add_pd(fjy1,ty);
1205             fjz1             = _mm_add_pd(fjz1,tz);
1206
1207             /**************************
1208              * CALCULATE INTERACTIONS *
1209              **************************/
1210
1211             r32              = _mm_mul_pd(rsq32,rinv32);
1212
1213             /* EWALD ELECTROSTATICS */
1214
1215             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1216             ewrt             = _mm_mul_pd(r32,ewtabscale);
1217             ewitab           = _mm_cvttpd_epi32(ewrt);
1218             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1219             ewitab           = _mm_slli_epi32(ewitab,2);
1220             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1221             ewtabD           = _mm_setzero_pd();
1222             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1223             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1224             ewtabFn          = _mm_setzero_pd();
1225             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1226             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1227             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1228             velec            = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
1229             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1230
1231             /* Update potential sum for this i atom from the interaction with this j atom. */
1232             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1233             velecsum         = _mm_add_pd(velecsum,velec);
1234
1235             fscal            = felec;
1236
1237             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1238
1239             /* Calculate temporary vectorial force */
1240             tx               = _mm_mul_pd(fscal,dx32);
1241             ty               = _mm_mul_pd(fscal,dy32);
1242             tz               = _mm_mul_pd(fscal,dz32);
1243
1244             /* Update vectorial force */
1245             fix3             = _mm_add_pd(fix3,tx);
1246             fiy3             = _mm_add_pd(fiy3,ty);
1247             fiz3             = _mm_add_pd(fiz3,tz);
1248
1249             fjx2             = _mm_add_pd(fjx2,tx);
1250             fjy2             = _mm_add_pd(fjy2,ty);
1251             fjz2             = _mm_add_pd(fjz2,tz);
1252
1253             /**************************
1254              * CALCULATE INTERACTIONS *
1255              **************************/
1256
1257             r33              = _mm_mul_pd(rsq33,rinv33);
1258
1259             /* EWALD ELECTROSTATICS */
1260
1261             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1262             ewrt             = _mm_mul_pd(r33,ewtabscale);
1263             ewitab           = _mm_cvttpd_epi32(ewrt);
1264             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1265             ewitab           = _mm_slli_epi32(ewitab,2);
1266             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1267             ewtabD           = _mm_setzero_pd();
1268             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1269             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1270             ewtabFn          = _mm_setzero_pd();
1271             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1272             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1273             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1274             velec            = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
1275             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1276
1277             /* Update potential sum for this i atom from the interaction with this j atom. */
1278             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1279             velecsum         = _mm_add_pd(velecsum,velec);
1280
1281             fscal            = felec;
1282
1283             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1284
1285             /* Calculate temporary vectorial force */
1286             tx               = _mm_mul_pd(fscal,dx33);
1287             ty               = _mm_mul_pd(fscal,dy33);
1288             tz               = _mm_mul_pd(fscal,dz33);
1289
1290             /* Update vectorial force */
1291             fix3             = _mm_add_pd(fix3,tx);
1292             fiy3             = _mm_add_pd(fiy3,ty);
1293             fiz3             = _mm_add_pd(fiz3,tz);
1294
1295             fjx3             = _mm_add_pd(fjx3,tx);
1296             fjy3             = _mm_add_pd(fjy3,ty);
1297             fjz3             = _mm_add_pd(fjz3,tz);
1298
1299             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1300
1301             /* Inner loop uses 423 flops */
1302         }
1303
1304         /* End of innermost loop */
1305
1306         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1307                                               f+i_coord_offset,fshift+i_shift_offset);
1308
1309         ggid                        = gid[iidx];
1310         /* Update potential energies */
1311         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1312         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1313
1314         /* Increment number of inner iterations */
1315         inneriter                  += j_index_end - j_index_start;
1316
1317         /* Outer loop uses 26 flops */
1318     }
1319
1320     /* Increment number of outer iterations */
1321     outeriter        += nri;
1322
1323     /* Update outer/inner flops */
1324
1325     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*423);
1326 }
1327 /*
1328  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJEw_GeomW4W4_F_sse2_double
1329  * Electrostatics interaction: Ewald
1330  * VdW interaction:            LJEwald
1331  * Geometry:                   Water4-Water4
1332  * Calculate force/pot:        Force
1333  */
1334 void
1335 nb_kernel_ElecEw_VdwLJEw_GeomW4W4_F_sse2_double
1336                     (t_nblist                    * gmx_restrict       nlist,
1337                      rvec                        * gmx_restrict          xx,
1338                      rvec                        * gmx_restrict          ff,
1339                      t_forcerec                  * gmx_restrict          fr,
1340                      t_mdatoms                   * gmx_restrict     mdatoms,
1341                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1342                      t_nrnb                      * gmx_restrict        nrnb)
1343 {
1344     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1345      * just 0 for non-waters.
1346      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1347      * jnr indices corresponding to data put in the four positions in the SIMD register.
1348      */
1349     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1350     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1351     int              jnrA,jnrB;
1352     int              j_coord_offsetA,j_coord_offsetB;
1353     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1354     real             rcutoff_scalar;
1355     real             *shiftvec,*fshift,*x,*f;
1356     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1357     int              vdwioffset0;
1358     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1359     int              vdwioffset1;
1360     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1361     int              vdwioffset2;
1362     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1363     int              vdwioffset3;
1364     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1365     int              vdwjidx0A,vdwjidx0B;
1366     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1367     int              vdwjidx1A,vdwjidx1B;
1368     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1369     int              vdwjidx2A,vdwjidx2B;
1370     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1371     int              vdwjidx3A,vdwjidx3B;
1372     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1373     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1374     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1375     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1376     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1377     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1378     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1379     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1380     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1381     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1382     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1383     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1384     real             *charge;
1385     int              nvdwtype;
1386     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1387     int              *vdwtype;
1388     real             *vdwparam;
1389     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1390     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1391     __m128d           c6grid_00;
1392     __m128d           c6grid_11;
1393     __m128d           c6grid_12;
1394     __m128d           c6grid_13;
1395     __m128d           c6grid_21;
1396     __m128d           c6grid_22;
1397     __m128d           c6grid_23;
1398     __m128d           c6grid_31;
1399     __m128d           c6grid_32;
1400     __m128d           c6grid_33;
1401     __m128d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1402     real             *vdwgridparam;
1403     __m128d           one_half = _mm_set1_pd(0.5);
1404     __m128d           minus_one = _mm_set1_pd(-1.0);
1405     __m128i          ewitab;
1406     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1407     real             *ewtab;
1408     __m128d          dummy_mask,cutoff_mask;
1409     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1410     __m128d          one     = _mm_set1_pd(1.0);
1411     __m128d          two     = _mm_set1_pd(2.0);
1412     x                = xx[0];
1413     f                = ff[0];
1414
1415     nri              = nlist->nri;
1416     iinr             = nlist->iinr;
1417     jindex           = nlist->jindex;
1418     jjnr             = nlist->jjnr;
1419     shiftidx         = nlist->shift;
1420     gid              = nlist->gid;
1421     shiftvec         = fr->shift_vec[0];
1422     fshift           = fr->fshift[0];
1423     facel            = _mm_set1_pd(fr->epsfac);
1424     charge           = mdatoms->chargeA;
1425     nvdwtype         = fr->ntype;
1426     vdwparam         = fr->nbfp;
1427     vdwtype          = mdatoms->typeA;
1428     vdwgridparam     = fr->ljpme_c6grid;
1429     sh_lj_ewald      = _mm_set1_pd(fr->ic->sh_lj_ewald);
1430     ewclj            = _mm_set1_pd(fr->ewaldcoeff_lj);
1431     ewclj2           = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
1432
1433     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1434     ewtab            = fr->ic->tabq_coul_F;
1435     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1436     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1437
1438     /* Setup water-specific parameters */
1439     inr              = nlist->iinr[0];
1440     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1441     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1442     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1443     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1444
1445     jq1              = _mm_set1_pd(charge[inr+1]);
1446     jq2              = _mm_set1_pd(charge[inr+2]);
1447     jq3              = _mm_set1_pd(charge[inr+3]);
1448     vdwjidx0A        = 2*vdwtype[inr+0];
1449     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1450     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1451     c6grid_00        = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
1452     qq11             = _mm_mul_pd(iq1,jq1);
1453     qq12             = _mm_mul_pd(iq1,jq2);
1454     qq13             = _mm_mul_pd(iq1,jq3);
1455     qq21             = _mm_mul_pd(iq2,jq1);
1456     qq22             = _mm_mul_pd(iq2,jq2);
1457     qq23             = _mm_mul_pd(iq2,jq3);
1458     qq31             = _mm_mul_pd(iq3,jq1);
1459     qq32             = _mm_mul_pd(iq3,jq2);
1460     qq33             = _mm_mul_pd(iq3,jq3);
1461
1462     /* Avoid stupid compiler warnings */
1463     jnrA = jnrB = 0;
1464     j_coord_offsetA = 0;
1465     j_coord_offsetB = 0;
1466
1467     outeriter        = 0;
1468     inneriter        = 0;
1469
1470     /* Start outer loop over neighborlists */
1471     for(iidx=0; iidx<nri; iidx++)
1472     {
1473         /* Load shift vector for this list */
1474         i_shift_offset   = DIM*shiftidx[iidx];
1475
1476         /* Load limits for loop over neighbors */
1477         j_index_start    = jindex[iidx];
1478         j_index_end      = jindex[iidx+1];
1479
1480         /* Get outer coordinate index */
1481         inr              = iinr[iidx];
1482         i_coord_offset   = DIM*inr;
1483
1484         /* Load i particle coords and add shift vector */
1485         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1486                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1487
1488         fix0             = _mm_setzero_pd();
1489         fiy0             = _mm_setzero_pd();
1490         fiz0             = _mm_setzero_pd();
1491         fix1             = _mm_setzero_pd();
1492         fiy1             = _mm_setzero_pd();
1493         fiz1             = _mm_setzero_pd();
1494         fix2             = _mm_setzero_pd();
1495         fiy2             = _mm_setzero_pd();
1496         fiz2             = _mm_setzero_pd();
1497         fix3             = _mm_setzero_pd();
1498         fiy3             = _mm_setzero_pd();
1499         fiz3             = _mm_setzero_pd();
1500
1501         /* Start inner kernel loop */
1502         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1503         {
1504
1505             /* Get j neighbor index, and coordinate index */
1506             jnrA             = jjnr[jidx];
1507             jnrB             = jjnr[jidx+1];
1508             j_coord_offsetA  = DIM*jnrA;
1509             j_coord_offsetB  = DIM*jnrB;
1510
1511             /* load j atom coordinates */
1512             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1513                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1514                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1515
1516             /* Calculate displacement vector */
1517             dx00             = _mm_sub_pd(ix0,jx0);
1518             dy00             = _mm_sub_pd(iy0,jy0);
1519             dz00             = _mm_sub_pd(iz0,jz0);
1520             dx11             = _mm_sub_pd(ix1,jx1);
1521             dy11             = _mm_sub_pd(iy1,jy1);
1522             dz11             = _mm_sub_pd(iz1,jz1);
1523             dx12             = _mm_sub_pd(ix1,jx2);
1524             dy12             = _mm_sub_pd(iy1,jy2);
1525             dz12             = _mm_sub_pd(iz1,jz2);
1526             dx13             = _mm_sub_pd(ix1,jx3);
1527             dy13             = _mm_sub_pd(iy1,jy3);
1528             dz13             = _mm_sub_pd(iz1,jz3);
1529             dx21             = _mm_sub_pd(ix2,jx1);
1530             dy21             = _mm_sub_pd(iy2,jy1);
1531             dz21             = _mm_sub_pd(iz2,jz1);
1532             dx22             = _mm_sub_pd(ix2,jx2);
1533             dy22             = _mm_sub_pd(iy2,jy2);
1534             dz22             = _mm_sub_pd(iz2,jz2);
1535             dx23             = _mm_sub_pd(ix2,jx3);
1536             dy23             = _mm_sub_pd(iy2,jy3);
1537             dz23             = _mm_sub_pd(iz2,jz3);
1538             dx31             = _mm_sub_pd(ix3,jx1);
1539             dy31             = _mm_sub_pd(iy3,jy1);
1540             dz31             = _mm_sub_pd(iz3,jz1);
1541             dx32             = _mm_sub_pd(ix3,jx2);
1542             dy32             = _mm_sub_pd(iy3,jy2);
1543             dz32             = _mm_sub_pd(iz3,jz2);
1544             dx33             = _mm_sub_pd(ix3,jx3);
1545             dy33             = _mm_sub_pd(iy3,jy3);
1546             dz33             = _mm_sub_pd(iz3,jz3);
1547
1548             /* Calculate squared distance and things based on it */
1549             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1550             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1551             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1552             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1553             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1554             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1555             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1556             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1557             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1558             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1559
1560             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1561             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1562             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1563             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1564             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1565             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1566             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1567             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1568             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1569             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1570
1571             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1572             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1573             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1574             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1575             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1576             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1577             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1578             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1579             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1580             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1581
1582             fjx0             = _mm_setzero_pd();
1583             fjy0             = _mm_setzero_pd();
1584             fjz0             = _mm_setzero_pd();
1585             fjx1             = _mm_setzero_pd();
1586             fjy1             = _mm_setzero_pd();
1587             fjz1             = _mm_setzero_pd();
1588             fjx2             = _mm_setzero_pd();
1589             fjy2             = _mm_setzero_pd();
1590             fjz2             = _mm_setzero_pd();
1591             fjx3             = _mm_setzero_pd();
1592             fjy3             = _mm_setzero_pd();
1593             fjz3             = _mm_setzero_pd();
1594
1595             /**************************
1596              * CALCULATE INTERACTIONS *
1597              **************************/
1598
1599             r00              = _mm_mul_pd(rsq00,rinv00);
1600
1601             /* Analytical LJ-PME */
1602             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1603             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
1604             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
1605             exponent         = gmx_simd_exp_d(ewcljrsq);
1606             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1607             poly             = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
1608             /* f6A = 6 * C6grid * (1 - poly) */
1609             f6A              = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
1610             /* f6B = C6grid * exponent * beta^6 */
1611             f6B              = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
1612             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1613             fvdw              = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1614
1615             fscal            = fvdw;
1616
1617             /* Calculate temporary vectorial force */
1618             tx               = _mm_mul_pd(fscal,dx00);
1619             ty               = _mm_mul_pd(fscal,dy00);
1620             tz               = _mm_mul_pd(fscal,dz00);
1621
1622             /* Update vectorial force */
1623             fix0             = _mm_add_pd(fix0,tx);
1624             fiy0             = _mm_add_pd(fiy0,ty);
1625             fiz0             = _mm_add_pd(fiz0,tz);
1626
1627             fjx0             = _mm_add_pd(fjx0,tx);
1628             fjy0             = _mm_add_pd(fjy0,ty);
1629             fjz0             = _mm_add_pd(fjz0,tz);
1630
1631             /**************************
1632              * CALCULATE INTERACTIONS *
1633              **************************/
1634
1635             r11              = _mm_mul_pd(rsq11,rinv11);
1636
1637             /* EWALD ELECTROSTATICS */
1638
1639             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1640             ewrt             = _mm_mul_pd(r11,ewtabscale);
1641             ewitab           = _mm_cvttpd_epi32(ewrt);
1642             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1643             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1644                                          &ewtabF,&ewtabFn);
1645             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1646             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1647
1648             fscal            = felec;
1649
1650             /* Calculate temporary vectorial force */
1651             tx               = _mm_mul_pd(fscal,dx11);
1652             ty               = _mm_mul_pd(fscal,dy11);
1653             tz               = _mm_mul_pd(fscal,dz11);
1654
1655             /* Update vectorial force */
1656             fix1             = _mm_add_pd(fix1,tx);
1657             fiy1             = _mm_add_pd(fiy1,ty);
1658             fiz1             = _mm_add_pd(fiz1,tz);
1659
1660             fjx1             = _mm_add_pd(fjx1,tx);
1661             fjy1             = _mm_add_pd(fjy1,ty);
1662             fjz1             = _mm_add_pd(fjz1,tz);
1663
1664             /**************************
1665              * CALCULATE INTERACTIONS *
1666              **************************/
1667
1668             r12              = _mm_mul_pd(rsq12,rinv12);
1669
1670             /* EWALD ELECTROSTATICS */
1671
1672             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1673             ewrt             = _mm_mul_pd(r12,ewtabscale);
1674             ewitab           = _mm_cvttpd_epi32(ewrt);
1675             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1676             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1677                                          &ewtabF,&ewtabFn);
1678             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1679             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1680
1681             fscal            = felec;
1682
1683             /* Calculate temporary vectorial force */
1684             tx               = _mm_mul_pd(fscal,dx12);
1685             ty               = _mm_mul_pd(fscal,dy12);
1686             tz               = _mm_mul_pd(fscal,dz12);
1687
1688             /* Update vectorial force */
1689             fix1             = _mm_add_pd(fix1,tx);
1690             fiy1             = _mm_add_pd(fiy1,ty);
1691             fiz1             = _mm_add_pd(fiz1,tz);
1692
1693             fjx2             = _mm_add_pd(fjx2,tx);
1694             fjy2             = _mm_add_pd(fjy2,ty);
1695             fjz2             = _mm_add_pd(fjz2,tz);
1696
1697             /**************************
1698              * CALCULATE INTERACTIONS *
1699              **************************/
1700
1701             r13              = _mm_mul_pd(rsq13,rinv13);
1702
1703             /* EWALD ELECTROSTATICS */
1704
1705             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1706             ewrt             = _mm_mul_pd(r13,ewtabscale);
1707             ewitab           = _mm_cvttpd_epi32(ewrt);
1708             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1709             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1710                                          &ewtabF,&ewtabFn);
1711             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1712             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1713
1714             fscal            = felec;
1715
1716             /* Calculate temporary vectorial force */
1717             tx               = _mm_mul_pd(fscal,dx13);
1718             ty               = _mm_mul_pd(fscal,dy13);
1719             tz               = _mm_mul_pd(fscal,dz13);
1720
1721             /* Update vectorial force */
1722             fix1             = _mm_add_pd(fix1,tx);
1723             fiy1             = _mm_add_pd(fiy1,ty);
1724             fiz1             = _mm_add_pd(fiz1,tz);
1725
1726             fjx3             = _mm_add_pd(fjx3,tx);
1727             fjy3             = _mm_add_pd(fjy3,ty);
1728             fjz3             = _mm_add_pd(fjz3,tz);
1729
1730             /**************************
1731              * CALCULATE INTERACTIONS *
1732              **************************/
1733
1734             r21              = _mm_mul_pd(rsq21,rinv21);
1735
1736             /* EWALD ELECTROSTATICS */
1737
1738             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1739             ewrt             = _mm_mul_pd(r21,ewtabscale);
1740             ewitab           = _mm_cvttpd_epi32(ewrt);
1741             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1742             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1743                                          &ewtabF,&ewtabFn);
1744             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1745             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1746
1747             fscal            = felec;
1748
1749             /* Calculate temporary vectorial force */
1750             tx               = _mm_mul_pd(fscal,dx21);
1751             ty               = _mm_mul_pd(fscal,dy21);
1752             tz               = _mm_mul_pd(fscal,dz21);
1753
1754             /* Update vectorial force */
1755             fix2             = _mm_add_pd(fix2,tx);
1756             fiy2             = _mm_add_pd(fiy2,ty);
1757             fiz2             = _mm_add_pd(fiz2,tz);
1758
1759             fjx1             = _mm_add_pd(fjx1,tx);
1760             fjy1             = _mm_add_pd(fjy1,ty);
1761             fjz1             = _mm_add_pd(fjz1,tz);
1762
1763             /**************************
1764              * CALCULATE INTERACTIONS *
1765              **************************/
1766
1767             r22              = _mm_mul_pd(rsq22,rinv22);
1768
1769             /* EWALD ELECTROSTATICS */
1770
1771             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1772             ewrt             = _mm_mul_pd(r22,ewtabscale);
1773             ewitab           = _mm_cvttpd_epi32(ewrt);
1774             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1775             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1776                                          &ewtabF,&ewtabFn);
1777             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1778             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1779
1780             fscal            = felec;
1781
1782             /* Calculate temporary vectorial force */
1783             tx               = _mm_mul_pd(fscal,dx22);
1784             ty               = _mm_mul_pd(fscal,dy22);
1785             tz               = _mm_mul_pd(fscal,dz22);
1786
1787             /* Update vectorial force */
1788             fix2             = _mm_add_pd(fix2,tx);
1789             fiy2             = _mm_add_pd(fiy2,ty);
1790             fiz2             = _mm_add_pd(fiz2,tz);
1791
1792             fjx2             = _mm_add_pd(fjx2,tx);
1793             fjy2             = _mm_add_pd(fjy2,ty);
1794             fjz2             = _mm_add_pd(fjz2,tz);
1795
1796             /**************************
1797              * CALCULATE INTERACTIONS *
1798              **************************/
1799
1800             r23              = _mm_mul_pd(rsq23,rinv23);
1801
1802             /* EWALD ELECTROSTATICS */
1803
1804             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1805             ewrt             = _mm_mul_pd(r23,ewtabscale);
1806             ewitab           = _mm_cvttpd_epi32(ewrt);
1807             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1808             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1809                                          &ewtabF,&ewtabFn);
1810             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1811             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1812
1813             fscal            = felec;
1814
1815             /* Calculate temporary vectorial force */
1816             tx               = _mm_mul_pd(fscal,dx23);
1817             ty               = _mm_mul_pd(fscal,dy23);
1818             tz               = _mm_mul_pd(fscal,dz23);
1819
1820             /* Update vectorial force */
1821             fix2             = _mm_add_pd(fix2,tx);
1822             fiy2             = _mm_add_pd(fiy2,ty);
1823             fiz2             = _mm_add_pd(fiz2,tz);
1824
1825             fjx3             = _mm_add_pd(fjx3,tx);
1826             fjy3             = _mm_add_pd(fjy3,ty);
1827             fjz3             = _mm_add_pd(fjz3,tz);
1828
1829             /**************************
1830              * CALCULATE INTERACTIONS *
1831              **************************/
1832
1833             r31              = _mm_mul_pd(rsq31,rinv31);
1834
1835             /* EWALD ELECTROSTATICS */
1836
1837             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1838             ewrt             = _mm_mul_pd(r31,ewtabscale);
1839             ewitab           = _mm_cvttpd_epi32(ewrt);
1840             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1841             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1842                                          &ewtabF,&ewtabFn);
1843             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1844             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1845
1846             fscal            = felec;
1847
1848             /* Calculate temporary vectorial force */
1849             tx               = _mm_mul_pd(fscal,dx31);
1850             ty               = _mm_mul_pd(fscal,dy31);
1851             tz               = _mm_mul_pd(fscal,dz31);
1852
1853             /* Update vectorial force */
1854             fix3             = _mm_add_pd(fix3,tx);
1855             fiy3             = _mm_add_pd(fiy3,ty);
1856             fiz3             = _mm_add_pd(fiz3,tz);
1857
1858             fjx1             = _mm_add_pd(fjx1,tx);
1859             fjy1             = _mm_add_pd(fjy1,ty);
1860             fjz1             = _mm_add_pd(fjz1,tz);
1861
1862             /**************************
1863              * CALCULATE INTERACTIONS *
1864              **************************/
1865
1866             r32              = _mm_mul_pd(rsq32,rinv32);
1867
1868             /* EWALD ELECTROSTATICS */
1869
1870             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1871             ewrt             = _mm_mul_pd(r32,ewtabscale);
1872             ewitab           = _mm_cvttpd_epi32(ewrt);
1873             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1874             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1875                                          &ewtabF,&ewtabFn);
1876             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1877             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1878
1879             fscal            = felec;
1880
1881             /* Calculate temporary vectorial force */
1882             tx               = _mm_mul_pd(fscal,dx32);
1883             ty               = _mm_mul_pd(fscal,dy32);
1884             tz               = _mm_mul_pd(fscal,dz32);
1885
1886             /* Update vectorial force */
1887             fix3             = _mm_add_pd(fix3,tx);
1888             fiy3             = _mm_add_pd(fiy3,ty);
1889             fiz3             = _mm_add_pd(fiz3,tz);
1890
1891             fjx2             = _mm_add_pd(fjx2,tx);
1892             fjy2             = _mm_add_pd(fjy2,ty);
1893             fjz2             = _mm_add_pd(fjz2,tz);
1894
1895             /**************************
1896              * CALCULATE INTERACTIONS *
1897              **************************/
1898
1899             r33              = _mm_mul_pd(rsq33,rinv33);
1900
1901             /* EWALD ELECTROSTATICS */
1902
1903             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1904             ewrt             = _mm_mul_pd(r33,ewtabscale);
1905             ewitab           = _mm_cvttpd_epi32(ewrt);
1906             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1907             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1908                                          &ewtabF,&ewtabFn);
1909             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1910             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1911
1912             fscal            = felec;
1913
1914             /* Calculate temporary vectorial force */
1915             tx               = _mm_mul_pd(fscal,dx33);
1916             ty               = _mm_mul_pd(fscal,dy33);
1917             tz               = _mm_mul_pd(fscal,dz33);
1918
1919             /* Update vectorial force */
1920             fix3             = _mm_add_pd(fix3,tx);
1921             fiy3             = _mm_add_pd(fiy3,ty);
1922             fiz3             = _mm_add_pd(fiz3,tz);
1923
1924             fjx3             = _mm_add_pd(fjx3,tx);
1925             fjy3             = _mm_add_pd(fjy3,ty);
1926             fjz3             = _mm_add_pd(fjz3,tz);
1927
1928             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1929
1930             /* Inner loop uses 373 flops */
1931         }
1932
1933         if(jidx<j_index_end)
1934         {
1935
1936             jnrA             = jjnr[jidx];
1937             j_coord_offsetA  = DIM*jnrA;
1938
1939             /* load j atom coordinates */
1940             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1941                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1942                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1943
1944             /* Calculate displacement vector */
1945             dx00             = _mm_sub_pd(ix0,jx0);
1946             dy00             = _mm_sub_pd(iy0,jy0);
1947             dz00             = _mm_sub_pd(iz0,jz0);
1948             dx11             = _mm_sub_pd(ix1,jx1);
1949             dy11             = _mm_sub_pd(iy1,jy1);
1950             dz11             = _mm_sub_pd(iz1,jz1);
1951             dx12             = _mm_sub_pd(ix1,jx2);
1952             dy12             = _mm_sub_pd(iy1,jy2);
1953             dz12             = _mm_sub_pd(iz1,jz2);
1954             dx13             = _mm_sub_pd(ix1,jx3);
1955             dy13             = _mm_sub_pd(iy1,jy3);
1956             dz13             = _mm_sub_pd(iz1,jz3);
1957             dx21             = _mm_sub_pd(ix2,jx1);
1958             dy21             = _mm_sub_pd(iy2,jy1);
1959             dz21             = _mm_sub_pd(iz2,jz1);
1960             dx22             = _mm_sub_pd(ix2,jx2);
1961             dy22             = _mm_sub_pd(iy2,jy2);
1962             dz22             = _mm_sub_pd(iz2,jz2);
1963             dx23             = _mm_sub_pd(ix2,jx3);
1964             dy23             = _mm_sub_pd(iy2,jy3);
1965             dz23             = _mm_sub_pd(iz2,jz3);
1966             dx31             = _mm_sub_pd(ix3,jx1);
1967             dy31             = _mm_sub_pd(iy3,jy1);
1968             dz31             = _mm_sub_pd(iz3,jz1);
1969             dx32             = _mm_sub_pd(ix3,jx2);
1970             dy32             = _mm_sub_pd(iy3,jy2);
1971             dz32             = _mm_sub_pd(iz3,jz2);
1972             dx33             = _mm_sub_pd(ix3,jx3);
1973             dy33             = _mm_sub_pd(iy3,jy3);
1974             dz33             = _mm_sub_pd(iz3,jz3);
1975
1976             /* Calculate squared distance and things based on it */
1977             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1978             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1979             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1980             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1981             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1982             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1983             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1984             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1985             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1986             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1987
1988             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1989             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1990             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1991             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1992             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1993             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1994             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1995             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1996             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1997             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1998
1999             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
2000             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
2001             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
2002             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
2003             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
2004             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
2005             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
2006             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
2007             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
2008             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
2009
2010             fjx0             = _mm_setzero_pd();
2011             fjy0             = _mm_setzero_pd();
2012             fjz0             = _mm_setzero_pd();
2013             fjx1             = _mm_setzero_pd();
2014             fjy1             = _mm_setzero_pd();
2015             fjz1             = _mm_setzero_pd();
2016             fjx2             = _mm_setzero_pd();
2017             fjy2             = _mm_setzero_pd();
2018             fjz2             = _mm_setzero_pd();
2019             fjx3             = _mm_setzero_pd();
2020             fjy3             = _mm_setzero_pd();
2021             fjz3             = _mm_setzero_pd();
2022
2023             /**************************
2024              * CALCULATE INTERACTIONS *
2025              **************************/
2026
2027             r00              = _mm_mul_pd(rsq00,rinv00);
2028
2029             /* Analytical LJ-PME */
2030             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2031             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
2032             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
2033             exponent         = gmx_simd_exp_d(ewcljrsq);
2034             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
2035             poly             = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
2036             /* f6A = 6 * C6grid * (1 - poly) */
2037             f6A              = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
2038             /* f6B = C6grid * exponent * beta^6 */
2039             f6B              = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
2040             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2041             fvdw              = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
2042
2043             fscal            = fvdw;
2044
2045             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2046
2047             /* Calculate temporary vectorial force */
2048             tx               = _mm_mul_pd(fscal,dx00);
2049             ty               = _mm_mul_pd(fscal,dy00);
2050             tz               = _mm_mul_pd(fscal,dz00);
2051
2052             /* Update vectorial force */
2053             fix0             = _mm_add_pd(fix0,tx);
2054             fiy0             = _mm_add_pd(fiy0,ty);
2055             fiz0             = _mm_add_pd(fiz0,tz);
2056
2057             fjx0             = _mm_add_pd(fjx0,tx);
2058             fjy0             = _mm_add_pd(fjy0,ty);
2059             fjz0             = _mm_add_pd(fjz0,tz);
2060
2061             /**************************
2062              * CALCULATE INTERACTIONS *
2063              **************************/
2064
2065             r11              = _mm_mul_pd(rsq11,rinv11);
2066
2067             /* EWALD ELECTROSTATICS */
2068
2069             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2070             ewrt             = _mm_mul_pd(r11,ewtabscale);
2071             ewitab           = _mm_cvttpd_epi32(ewrt);
2072             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2073             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2074             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2075             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2076
2077             fscal            = felec;
2078
2079             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2080
2081             /* Calculate temporary vectorial force */
2082             tx               = _mm_mul_pd(fscal,dx11);
2083             ty               = _mm_mul_pd(fscal,dy11);
2084             tz               = _mm_mul_pd(fscal,dz11);
2085
2086             /* Update vectorial force */
2087             fix1             = _mm_add_pd(fix1,tx);
2088             fiy1             = _mm_add_pd(fiy1,ty);
2089             fiz1             = _mm_add_pd(fiz1,tz);
2090
2091             fjx1             = _mm_add_pd(fjx1,tx);
2092             fjy1             = _mm_add_pd(fjy1,ty);
2093             fjz1             = _mm_add_pd(fjz1,tz);
2094
2095             /**************************
2096              * CALCULATE INTERACTIONS *
2097              **************************/
2098
2099             r12              = _mm_mul_pd(rsq12,rinv12);
2100
2101             /* EWALD ELECTROSTATICS */
2102
2103             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2104             ewrt             = _mm_mul_pd(r12,ewtabscale);
2105             ewitab           = _mm_cvttpd_epi32(ewrt);
2106             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2107             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2108             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2109             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2110
2111             fscal            = felec;
2112
2113             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2114
2115             /* Calculate temporary vectorial force */
2116             tx               = _mm_mul_pd(fscal,dx12);
2117             ty               = _mm_mul_pd(fscal,dy12);
2118             tz               = _mm_mul_pd(fscal,dz12);
2119
2120             /* Update vectorial force */
2121             fix1             = _mm_add_pd(fix1,tx);
2122             fiy1             = _mm_add_pd(fiy1,ty);
2123             fiz1             = _mm_add_pd(fiz1,tz);
2124
2125             fjx2             = _mm_add_pd(fjx2,tx);
2126             fjy2             = _mm_add_pd(fjy2,ty);
2127             fjz2             = _mm_add_pd(fjz2,tz);
2128
2129             /**************************
2130              * CALCULATE INTERACTIONS *
2131              **************************/
2132
2133             r13              = _mm_mul_pd(rsq13,rinv13);
2134
2135             /* EWALD ELECTROSTATICS */
2136
2137             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2138             ewrt             = _mm_mul_pd(r13,ewtabscale);
2139             ewitab           = _mm_cvttpd_epi32(ewrt);
2140             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2141             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2142             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2143             felec            = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
2144
2145             fscal            = felec;
2146
2147             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2148
2149             /* Calculate temporary vectorial force */
2150             tx               = _mm_mul_pd(fscal,dx13);
2151             ty               = _mm_mul_pd(fscal,dy13);
2152             tz               = _mm_mul_pd(fscal,dz13);
2153
2154             /* Update vectorial force */
2155             fix1             = _mm_add_pd(fix1,tx);
2156             fiy1             = _mm_add_pd(fiy1,ty);
2157             fiz1             = _mm_add_pd(fiz1,tz);
2158
2159             fjx3             = _mm_add_pd(fjx3,tx);
2160             fjy3             = _mm_add_pd(fjy3,ty);
2161             fjz3             = _mm_add_pd(fjz3,tz);
2162
2163             /**************************
2164              * CALCULATE INTERACTIONS *
2165              **************************/
2166
2167             r21              = _mm_mul_pd(rsq21,rinv21);
2168
2169             /* EWALD ELECTROSTATICS */
2170
2171             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2172             ewrt             = _mm_mul_pd(r21,ewtabscale);
2173             ewitab           = _mm_cvttpd_epi32(ewrt);
2174             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2175             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2176             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2177             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2178
2179             fscal            = felec;
2180
2181             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2182
2183             /* Calculate temporary vectorial force */
2184             tx               = _mm_mul_pd(fscal,dx21);
2185             ty               = _mm_mul_pd(fscal,dy21);
2186             tz               = _mm_mul_pd(fscal,dz21);
2187
2188             /* Update vectorial force */
2189             fix2             = _mm_add_pd(fix2,tx);
2190             fiy2             = _mm_add_pd(fiy2,ty);
2191             fiz2             = _mm_add_pd(fiz2,tz);
2192
2193             fjx1             = _mm_add_pd(fjx1,tx);
2194             fjy1             = _mm_add_pd(fjy1,ty);
2195             fjz1             = _mm_add_pd(fjz1,tz);
2196
2197             /**************************
2198              * CALCULATE INTERACTIONS *
2199              **************************/
2200
2201             r22              = _mm_mul_pd(rsq22,rinv22);
2202
2203             /* EWALD ELECTROSTATICS */
2204
2205             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2206             ewrt             = _mm_mul_pd(r22,ewtabscale);
2207             ewitab           = _mm_cvttpd_epi32(ewrt);
2208             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2209             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2210             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2211             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2212
2213             fscal            = felec;
2214
2215             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2216
2217             /* Calculate temporary vectorial force */
2218             tx               = _mm_mul_pd(fscal,dx22);
2219             ty               = _mm_mul_pd(fscal,dy22);
2220             tz               = _mm_mul_pd(fscal,dz22);
2221
2222             /* Update vectorial force */
2223             fix2             = _mm_add_pd(fix2,tx);
2224             fiy2             = _mm_add_pd(fiy2,ty);
2225             fiz2             = _mm_add_pd(fiz2,tz);
2226
2227             fjx2             = _mm_add_pd(fjx2,tx);
2228             fjy2             = _mm_add_pd(fjy2,ty);
2229             fjz2             = _mm_add_pd(fjz2,tz);
2230
2231             /**************************
2232              * CALCULATE INTERACTIONS *
2233              **************************/
2234
2235             r23              = _mm_mul_pd(rsq23,rinv23);
2236
2237             /* EWALD ELECTROSTATICS */
2238
2239             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2240             ewrt             = _mm_mul_pd(r23,ewtabscale);
2241             ewitab           = _mm_cvttpd_epi32(ewrt);
2242             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2243             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2244             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2245             felec            = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2246
2247             fscal            = felec;
2248
2249             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2250
2251             /* Calculate temporary vectorial force */
2252             tx               = _mm_mul_pd(fscal,dx23);
2253             ty               = _mm_mul_pd(fscal,dy23);
2254             tz               = _mm_mul_pd(fscal,dz23);
2255
2256             /* Update vectorial force */
2257             fix2             = _mm_add_pd(fix2,tx);
2258             fiy2             = _mm_add_pd(fiy2,ty);
2259             fiz2             = _mm_add_pd(fiz2,tz);
2260
2261             fjx3             = _mm_add_pd(fjx3,tx);
2262             fjy3             = _mm_add_pd(fjy3,ty);
2263             fjz3             = _mm_add_pd(fjz3,tz);
2264
2265             /**************************
2266              * CALCULATE INTERACTIONS *
2267              **************************/
2268
2269             r31              = _mm_mul_pd(rsq31,rinv31);
2270
2271             /* EWALD ELECTROSTATICS */
2272
2273             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2274             ewrt             = _mm_mul_pd(r31,ewtabscale);
2275             ewitab           = _mm_cvttpd_epi32(ewrt);
2276             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2277             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2278             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2279             felec            = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2280
2281             fscal            = felec;
2282
2283             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2284
2285             /* Calculate temporary vectorial force */
2286             tx               = _mm_mul_pd(fscal,dx31);
2287             ty               = _mm_mul_pd(fscal,dy31);
2288             tz               = _mm_mul_pd(fscal,dz31);
2289
2290             /* Update vectorial force */
2291             fix3             = _mm_add_pd(fix3,tx);
2292             fiy3             = _mm_add_pd(fiy3,ty);
2293             fiz3             = _mm_add_pd(fiz3,tz);
2294
2295             fjx1             = _mm_add_pd(fjx1,tx);
2296             fjy1             = _mm_add_pd(fjy1,ty);
2297             fjz1             = _mm_add_pd(fjz1,tz);
2298
2299             /**************************
2300              * CALCULATE INTERACTIONS *
2301              **************************/
2302
2303             r32              = _mm_mul_pd(rsq32,rinv32);
2304
2305             /* EWALD ELECTROSTATICS */
2306
2307             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2308             ewrt             = _mm_mul_pd(r32,ewtabscale);
2309             ewitab           = _mm_cvttpd_epi32(ewrt);
2310             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2311             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2312             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2313             felec            = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2314
2315             fscal            = felec;
2316
2317             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2318
2319             /* Calculate temporary vectorial force */
2320             tx               = _mm_mul_pd(fscal,dx32);
2321             ty               = _mm_mul_pd(fscal,dy32);
2322             tz               = _mm_mul_pd(fscal,dz32);
2323
2324             /* Update vectorial force */
2325             fix3             = _mm_add_pd(fix3,tx);
2326             fiy3             = _mm_add_pd(fiy3,ty);
2327             fiz3             = _mm_add_pd(fiz3,tz);
2328
2329             fjx2             = _mm_add_pd(fjx2,tx);
2330             fjy2             = _mm_add_pd(fjy2,ty);
2331             fjz2             = _mm_add_pd(fjz2,tz);
2332
2333             /**************************
2334              * CALCULATE INTERACTIONS *
2335              **************************/
2336
2337             r33              = _mm_mul_pd(rsq33,rinv33);
2338
2339             /* EWALD ELECTROSTATICS */
2340
2341             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2342             ewrt             = _mm_mul_pd(r33,ewtabscale);
2343             ewitab           = _mm_cvttpd_epi32(ewrt);
2344             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2345             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2346             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2347             felec            = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2348
2349             fscal            = felec;
2350
2351             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2352
2353             /* Calculate temporary vectorial force */
2354             tx               = _mm_mul_pd(fscal,dx33);
2355             ty               = _mm_mul_pd(fscal,dy33);
2356             tz               = _mm_mul_pd(fscal,dz33);
2357
2358             /* Update vectorial force */
2359             fix3             = _mm_add_pd(fix3,tx);
2360             fiy3             = _mm_add_pd(fiy3,ty);
2361             fiz3             = _mm_add_pd(fiz3,tz);
2362
2363             fjx3             = _mm_add_pd(fjx3,tx);
2364             fjy3             = _mm_add_pd(fjy3,ty);
2365             fjz3             = _mm_add_pd(fjz3,tz);
2366
2367             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2368
2369             /* Inner loop uses 373 flops */
2370         }
2371
2372         /* End of innermost loop */
2373
2374         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2375                                               f+i_coord_offset,fshift+i_shift_offset);
2376
2377         /* Increment number of inner iterations */
2378         inneriter                  += j_index_end - j_index_start;
2379
2380         /* Outer loop uses 24 flops */
2381     }
2382
2383     /* Increment number of outer iterations */
2384     outeriter        += nri;
2385
2386     /* Update outer/inner flops */
2387
2388     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*373);
2389 }