src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecEw_VdwLJ_GeomW3W3_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sse4_1_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse4_1_double.h"
  50 #include "kernelutil_x86_sse4_1_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW3W3_VF_sse4_1_double
  54  * Electrostatics interaction: Ewald
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water3-Water3
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecEw_VdwLJ_GeomW3W3_VF_sse4_1_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  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     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  95     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  96     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  97     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  98     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  99     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 100     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 101     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 102     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 103     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 104     real             *charge;
 105     int              nvdwtype;
 106     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 107     int              *vdwtype;
 108     real             *vdwparam;
 109     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 110     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 111     __m128i          ewitab;
 112     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 113     real             *ewtab;
 114     __m128d          dummy_mask,cutoff_mask;
 115     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 116     __m128d          one     = _mm_set1_pd(1.0);
 117     __m128d          two     = _mm_set1_pd(2.0);
 118     x                = xx[0];
 119     f                = ff[0];
 120
 121     nri              = nlist->nri;
 122     iinr             = nlist->iinr;
 123     jindex           = nlist->jindex;
 124     jjnr             = nlist->jjnr;
 125     shiftidx         = nlist->shift;
 126     gid              = nlist->gid;
 127     shiftvec         = fr->shift_vec[0];
 128     fshift           = fr->fshift[0];
 129     facel            = _mm_set1_pd(fr->epsfac);
 130     charge           = mdatoms->chargeA;
 131     nvdwtype         = fr->ntype;
 132     vdwparam         = fr->nbfp;
 133     vdwtype          = mdatoms->typeA;
 134
 135     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 136     ewtab            = fr->ic->tabq_coul_FDV0;
 137     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 138     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 143     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 144     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 145     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 146
 147     jq0              = _mm_set1_pd(charge[inr+0]);
 148     jq1              = _mm_set1_pd(charge[inr+1]);
 149     jq2              = _mm_set1_pd(charge[inr+2]);
 150     vdwjidx0A        = 2*vdwtype[inr+0];
 151     qq00             = _mm_mul_pd(iq0,jq0);
 152     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 153     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 154     qq01             = _mm_mul_pd(iq0,jq1);
 155     qq02             = _mm_mul_pd(iq0,jq2);
 156     qq10             = _mm_mul_pd(iq1,jq0);
 157     qq11             = _mm_mul_pd(iq1,jq1);
 158     qq12             = _mm_mul_pd(iq1,jq2);
 159     qq20             = _mm_mul_pd(iq2,jq0);
 160     qq21             = _mm_mul_pd(iq2,jq1);
 161     qq22             = _mm_mul_pd(iq2,jq2);
 162
 163     /* Avoid stupid compiler warnings */
 164     jnrA = jnrB = 0;
 165     j_coord_offsetA = 0;
 166     j_coord_offsetB = 0;
 167
 168     outeriter        = 0;
 169     inneriter        = 0;
 170
 171     /* Start outer loop over neighborlists */
 172     for(iidx=0; iidx<nri; iidx++)
 173     {
 174         /* Load shift vector for this list */
 175         i_shift_offset   = DIM*shiftidx[iidx];
 176
 177         /* Load limits for loop over neighbors */
 178         j_index_start    = jindex[iidx];
 179         j_index_end      = jindex[iidx+1];
 180
 181         /* Get outer coordinate index */
 182         inr              = iinr[iidx];
 183         i_coord_offset   = DIM*inr;
 184
 185         /* Load i particle coords and add shift vector */
 186         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 187                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 188
 189         fix0             = _mm_setzero_pd();
 190         fiy0             = _mm_setzero_pd();
 191         fiz0             = _mm_setzero_pd();
 192         fix1             = _mm_setzero_pd();
 193         fiy1             = _mm_setzero_pd();
 194         fiz1             = _mm_setzero_pd();
 195         fix2             = _mm_setzero_pd();
 196         fiy2             = _mm_setzero_pd();
 197         fiz2             = _mm_setzero_pd();
 198
 199         /* Reset potential sums */
 200         velecsum         = _mm_setzero_pd();
 201         vvdwsum          = _mm_setzero_pd();
 202
 203         /* Start inner kernel loop */
 204         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 205         {
 206
 207             /* Get j neighbor index, and coordinate index */
 208             jnrA             = jjnr[jidx];
 209             jnrB             = jjnr[jidx+1];
 210             j_coord_offsetA  = DIM*jnrA;
 211             j_coord_offsetB  = DIM*jnrB;
 212
 213             /* load j atom coordinates */
 214             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 215                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 216
 217             /* Calculate displacement vector */
 218             dx00             = _mm_sub_pd(ix0,jx0);
 219             dy00             = _mm_sub_pd(iy0,jy0);
 220             dz00             = _mm_sub_pd(iz0,jz0);
 221             dx01             = _mm_sub_pd(ix0,jx1);
 222             dy01             = _mm_sub_pd(iy0,jy1);
 223             dz01             = _mm_sub_pd(iz0,jz1);
 224             dx02             = _mm_sub_pd(ix0,jx2);
 225             dy02             = _mm_sub_pd(iy0,jy2);
 226             dz02             = _mm_sub_pd(iz0,jz2);
 227             dx10             = _mm_sub_pd(ix1,jx0);
 228             dy10             = _mm_sub_pd(iy1,jy0);
 229             dz10             = _mm_sub_pd(iz1,jz0);
 230             dx11             = _mm_sub_pd(ix1,jx1);
 231             dy11             = _mm_sub_pd(iy1,jy1);
 232             dz11             = _mm_sub_pd(iz1,jz1);
 233             dx12             = _mm_sub_pd(ix1,jx2);
 234             dy12             = _mm_sub_pd(iy1,jy2);
 235             dz12             = _mm_sub_pd(iz1,jz2);
 236             dx20             = _mm_sub_pd(ix2,jx0);
 237             dy20             = _mm_sub_pd(iy2,jy0);
 238             dz20             = _mm_sub_pd(iz2,jz0);
 239             dx21             = _mm_sub_pd(ix2,jx1);
 240             dy21             = _mm_sub_pd(iy2,jy1);
 241             dz21             = _mm_sub_pd(iz2,jz1);
 242             dx22             = _mm_sub_pd(ix2,jx2);
 243             dy22             = _mm_sub_pd(iy2,jy2);
 244             dz22             = _mm_sub_pd(iz2,jz2);
 245
 246             /* Calculate squared distance and things based on it */
 247             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 248             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 249             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 250             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 251             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 252             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 253             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 254             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 255             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 256
 257             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 258             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 259             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 260             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 261             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 262             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 263             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 264             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 265             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 266
 267             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 268             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 269             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 270             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 271             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 272             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 273             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 274             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 275             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 276
 277             fjx0             = _mm_setzero_pd();
 278             fjy0             = _mm_setzero_pd();
 279             fjz0             = _mm_setzero_pd();
 280             fjx1             = _mm_setzero_pd();
 281             fjy1             = _mm_setzero_pd();
 282             fjz1             = _mm_setzero_pd();
 283             fjx2             = _mm_setzero_pd();
 284             fjy2             = _mm_setzero_pd();
 285             fjz2             = _mm_setzero_pd();
 286
 287             /**************************
 288              * CALCULATE INTERACTIONS *
 289              **************************/
 290
 291             r00              = _mm_mul_pd(rsq00,rinv00);
 292
 293             /* EWALD ELECTROSTATICS */
 294
 295             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 296             ewrt             = _mm_mul_pd(r00,ewtabscale);
 297             ewitab           = _mm_cvttpd_epi32(ewrt);
 298             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 299             ewitab           = _mm_slli_epi32(ewitab,2);
 300             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 301             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 302             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 303             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 304             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 305             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 306             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 307             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 308             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 309             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 310
 311             /* LENNARD-JONES DISPERSION/REPULSION */
 312
 313             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 314             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 315             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 316             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 317             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 318
 319             /* Update potential sum for this i atom from the interaction with this j atom. */
 320             velecsum         = _mm_add_pd(velecsum,velec);
 321             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 322
 323             fscal            = _mm_add_pd(felec,fvdw);
 324
 325             /* Calculate temporary vectorial force */
 326             tx               = _mm_mul_pd(fscal,dx00);
 327             ty               = _mm_mul_pd(fscal,dy00);
 328             tz               = _mm_mul_pd(fscal,dz00);
 329
 330             /* Update vectorial force */
 331             fix0             = _mm_add_pd(fix0,tx);
 332             fiy0             = _mm_add_pd(fiy0,ty);
 333             fiz0             = _mm_add_pd(fiz0,tz);
 334
 335             fjx0             = _mm_add_pd(fjx0,tx);
 336             fjy0             = _mm_add_pd(fjy0,ty);
 337             fjz0             = _mm_add_pd(fjz0,tz);
 338
 339             /**************************
 340              * CALCULATE INTERACTIONS *
 341              **************************/
 342
 343             r01              = _mm_mul_pd(rsq01,rinv01);
 344
 345             /* EWALD ELECTROSTATICS */
 346
 347             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 348             ewrt             = _mm_mul_pd(r01,ewtabscale);
 349             ewitab           = _mm_cvttpd_epi32(ewrt);
 350             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 351             ewitab           = _mm_slli_epi32(ewitab,2);
 352             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 353             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 354             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 355             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 356             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 357             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 358             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 359             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 360             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 361             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 362
 363             /* Update potential sum for this i atom from the interaction with this j atom. */
 364             velecsum         = _mm_add_pd(velecsum,velec);
 365
 366             fscal            = felec;
 367
 368             /* Calculate temporary vectorial force */
 369             tx               = _mm_mul_pd(fscal,dx01);
 370             ty               = _mm_mul_pd(fscal,dy01);
 371             tz               = _mm_mul_pd(fscal,dz01);
 372
 373             /* Update vectorial force */
 374             fix0             = _mm_add_pd(fix0,tx);
 375             fiy0             = _mm_add_pd(fiy0,ty);
 376             fiz0             = _mm_add_pd(fiz0,tz);
 377
 378             fjx1             = _mm_add_pd(fjx1,tx);
 379             fjy1             = _mm_add_pd(fjy1,ty);
 380             fjz1             = _mm_add_pd(fjz1,tz);
 381
 382             /**************************
 383              * CALCULATE INTERACTIONS *
 384              **************************/
 385
 386             r02              = _mm_mul_pd(rsq02,rinv02);
 387
 388             /* EWALD ELECTROSTATICS */
 389
 390             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 391             ewrt             = _mm_mul_pd(r02,ewtabscale);
 392             ewitab           = _mm_cvttpd_epi32(ewrt);
 393             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 394             ewitab           = _mm_slli_epi32(ewitab,2);
 395             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 396             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 397             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 398             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 399             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 400             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 401             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 402             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 403             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 404             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 405
 406             /* Update potential sum for this i atom from the interaction with this j atom. */
 407             velecsum         = _mm_add_pd(velecsum,velec);
 408
 409             fscal            = felec;
 410
 411             /* Calculate temporary vectorial force */
 412             tx               = _mm_mul_pd(fscal,dx02);
 413             ty               = _mm_mul_pd(fscal,dy02);
 414             tz               = _mm_mul_pd(fscal,dz02);
 415
 416             /* Update vectorial force */
 417             fix0             = _mm_add_pd(fix0,tx);
 418             fiy0             = _mm_add_pd(fiy0,ty);
 419             fiz0             = _mm_add_pd(fiz0,tz);
 420
 421             fjx2             = _mm_add_pd(fjx2,tx);
 422             fjy2             = _mm_add_pd(fjy2,ty);
 423             fjz2             = _mm_add_pd(fjz2,tz);
 424
 425             /**************************
 426              * CALCULATE INTERACTIONS *
 427              **************************/
 428
 429             r10              = _mm_mul_pd(rsq10,rinv10);
 430
 431             /* EWALD ELECTROSTATICS */
 432
 433             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 434             ewrt             = _mm_mul_pd(r10,ewtabscale);
 435             ewitab           = _mm_cvttpd_epi32(ewrt);
 436             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 437             ewitab           = _mm_slli_epi32(ewitab,2);
 438             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 439             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 440             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 441             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 442             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 443             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 444             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 445             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 446             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 447             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 448
 449             /* Update potential sum for this i atom from the interaction with this j atom. */
 450             velecsum         = _mm_add_pd(velecsum,velec);
 451
 452             fscal            = felec;
 453
 454             /* Calculate temporary vectorial force */
 455             tx               = _mm_mul_pd(fscal,dx10);
 456             ty               = _mm_mul_pd(fscal,dy10);
 457             tz               = _mm_mul_pd(fscal,dz10);
 458
 459             /* Update vectorial force */
 460             fix1             = _mm_add_pd(fix1,tx);
 461             fiy1             = _mm_add_pd(fiy1,ty);
 462             fiz1             = _mm_add_pd(fiz1,tz);
 463
 464             fjx0             = _mm_add_pd(fjx0,tx);
 465             fjy0             = _mm_add_pd(fjy0,ty);
 466             fjz0             = _mm_add_pd(fjz0,tz);
 467
 468             /**************************
 469              * CALCULATE INTERACTIONS *
 470              **************************/
 471
 472             r11              = _mm_mul_pd(rsq11,rinv11);
 473
 474             /* EWALD ELECTROSTATICS */
 475
 476             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 477             ewrt             = _mm_mul_pd(r11,ewtabscale);
 478             ewitab           = _mm_cvttpd_epi32(ewrt);
 479             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 480             ewitab           = _mm_slli_epi32(ewitab,2);
 481             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 482             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 483             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 484             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 485             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 486             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 487             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 488             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 489             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 490             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 491
 492             /* Update potential sum for this i atom from the interaction with this j atom. */
 493             velecsum         = _mm_add_pd(velecsum,velec);
 494
 495             fscal            = felec;
 496
 497             /* Calculate temporary vectorial force */
 498             tx               = _mm_mul_pd(fscal,dx11);
 499             ty               = _mm_mul_pd(fscal,dy11);
 500             tz               = _mm_mul_pd(fscal,dz11);
 501
 502             /* Update vectorial force */
 503             fix1             = _mm_add_pd(fix1,tx);
 504             fiy1             = _mm_add_pd(fiy1,ty);
 505             fiz1             = _mm_add_pd(fiz1,tz);
 506
 507             fjx1             = _mm_add_pd(fjx1,tx);
 508             fjy1             = _mm_add_pd(fjy1,ty);
 509             fjz1             = _mm_add_pd(fjz1,tz);
 510
 511             /**************************
 512              * CALCULATE INTERACTIONS *
 513              **************************/
 514
 515             r12              = _mm_mul_pd(rsq12,rinv12);
 516
 517             /* EWALD ELECTROSTATICS */
 518
 519             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 520             ewrt             = _mm_mul_pd(r12,ewtabscale);
 521             ewitab           = _mm_cvttpd_epi32(ewrt);
 522             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 523             ewitab           = _mm_slli_epi32(ewitab,2);
 524             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 525             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 526             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 527             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 528             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 529             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 530             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 531             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 532             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 533             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 534
 535             /* Update potential sum for this i atom from the interaction with this j atom. */
 536             velecsum         = _mm_add_pd(velecsum,velec);
 537
 538             fscal            = felec;
 539
 540             /* Calculate temporary vectorial force */
 541             tx               = _mm_mul_pd(fscal,dx12);
 542             ty               = _mm_mul_pd(fscal,dy12);
 543             tz               = _mm_mul_pd(fscal,dz12);
 544
 545             /* Update vectorial force */
 546             fix1             = _mm_add_pd(fix1,tx);
 547             fiy1             = _mm_add_pd(fiy1,ty);
 548             fiz1             = _mm_add_pd(fiz1,tz);
 549
 550             fjx2             = _mm_add_pd(fjx2,tx);
 551             fjy2             = _mm_add_pd(fjy2,ty);
 552             fjz2             = _mm_add_pd(fjz2,tz);
 553
 554             /**************************
 555              * CALCULATE INTERACTIONS *
 556              **************************/
 557
 558             r20              = _mm_mul_pd(rsq20,rinv20);
 559
 560             /* EWALD ELECTROSTATICS */
 561
 562             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 563             ewrt             = _mm_mul_pd(r20,ewtabscale);
 564             ewitab           = _mm_cvttpd_epi32(ewrt);
 565             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 566             ewitab           = _mm_slli_epi32(ewitab,2);
 567             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 568             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 569             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 570             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 571             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 572             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 573             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 574             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 575             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 576             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 577
 578             /* Update potential sum for this i atom from the interaction with this j atom. */
 579             velecsum         = _mm_add_pd(velecsum,velec);
 580
 581             fscal            = felec;
 582
 583             /* Calculate temporary vectorial force */
 584             tx               = _mm_mul_pd(fscal,dx20);
 585             ty               = _mm_mul_pd(fscal,dy20);
 586             tz               = _mm_mul_pd(fscal,dz20);
 587
 588             /* Update vectorial force */
 589             fix2             = _mm_add_pd(fix2,tx);
 590             fiy2             = _mm_add_pd(fiy2,ty);
 591             fiz2             = _mm_add_pd(fiz2,tz);
 592
 593             fjx0             = _mm_add_pd(fjx0,tx);
 594             fjy0             = _mm_add_pd(fjy0,ty);
 595             fjz0             = _mm_add_pd(fjz0,tz);
 596
 597             /**************************
 598              * CALCULATE INTERACTIONS *
 599              **************************/
 600
 601             r21              = _mm_mul_pd(rsq21,rinv21);
 602
 603             /* EWALD ELECTROSTATICS */
 604
 605             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 606             ewrt             = _mm_mul_pd(r21,ewtabscale);
 607             ewitab           = _mm_cvttpd_epi32(ewrt);
 608             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 609             ewitab           = _mm_slli_epi32(ewitab,2);
 610             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 611             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 612             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 613             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 614             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 615             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 616             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 617             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 618             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 619             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 620
 621             /* Update potential sum for this i atom from the interaction with this j atom. */
 622             velecsum         = _mm_add_pd(velecsum,velec);
 623
 624             fscal            = felec;
 625
 626             /* Calculate temporary vectorial force */
 627             tx               = _mm_mul_pd(fscal,dx21);
 628             ty               = _mm_mul_pd(fscal,dy21);
 629             tz               = _mm_mul_pd(fscal,dz21);
 630
 631             /* Update vectorial force */
 632             fix2             = _mm_add_pd(fix2,tx);
 633             fiy2             = _mm_add_pd(fiy2,ty);
 634             fiz2             = _mm_add_pd(fiz2,tz);
 635
 636             fjx1             = _mm_add_pd(fjx1,tx);
 637             fjy1             = _mm_add_pd(fjy1,ty);
 638             fjz1             = _mm_add_pd(fjz1,tz);
 639
 640             /**************************
 641              * CALCULATE INTERACTIONS *
 642              **************************/
 643
 644             r22              = _mm_mul_pd(rsq22,rinv22);
 645
 646             /* EWALD ELECTROSTATICS */
 647
 648             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 649             ewrt             = _mm_mul_pd(r22,ewtabscale);
 650             ewitab           = _mm_cvttpd_epi32(ewrt);
 651             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 652             ewitab           = _mm_slli_epi32(ewitab,2);
 653             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 654             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 655             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 656             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 657             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 658             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 659             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 660             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 661             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 662             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 663
 664             /* Update potential sum for this i atom from the interaction with this j atom. */
 665             velecsum         = _mm_add_pd(velecsum,velec);
 666
 667             fscal            = felec;
 668
 669             /* Calculate temporary vectorial force */
 670             tx               = _mm_mul_pd(fscal,dx22);
 671             ty               = _mm_mul_pd(fscal,dy22);
 672             tz               = _mm_mul_pd(fscal,dz22);
 673
 674             /* Update vectorial force */
 675             fix2             = _mm_add_pd(fix2,tx);
 676             fiy2             = _mm_add_pd(fiy2,ty);
 677             fiz2             = _mm_add_pd(fiz2,tz);
 678
 679             fjx2             = _mm_add_pd(fjx2,tx);
 680             fjy2             = _mm_add_pd(fjy2,ty);
 681             fjz2             = _mm_add_pd(fjz2,tz);
 682
 683             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 684
 685             /* Inner loop uses 381 flops */
 686         }
 687
 688         if(jidx<j_index_end)
 689         {
 690
 691             jnrA             = jjnr[jidx];
 692             j_coord_offsetA  = DIM*jnrA;
 693
 694             /* load j atom coordinates */
 695             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 696                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 697
 698             /* Calculate displacement vector */
 699             dx00             = _mm_sub_pd(ix0,jx0);
 700             dy00             = _mm_sub_pd(iy0,jy0);
 701             dz00             = _mm_sub_pd(iz0,jz0);
 702             dx01             = _mm_sub_pd(ix0,jx1);
 703             dy01             = _mm_sub_pd(iy0,jy1);
 704             dz01             = _mm_sub_pd(iz0,jz1);
 705             dx02             = _mm_sub_pd(ix0,jx2);
 706             dy02             = _mm_sub_pd(iy0,jy2);
 707             dz02             = _mm_sub_pd(iz0,jz2);
 708             dx10             = _mm_sub_pd(ix1,jx0);
 709             dy10             = _mm_sub_pd(iy1,jy0);
 710             dz10             = _mm_sub_pd(iz1,jz0);
 711             dx11             = _mm_sub_pd(ix1,jx1);
 712             dy11             = _mm_sub_pd(iy1,jy1);
 713             dz11             = _mm_sub_pd(iz1,jz1);
 714             dx12             = _mm_sub_pd(ix1,jx2);
 715             dy12             = _mm_sub_pd(iy1,jy2);
 716             dz12             = _mm_sub_pd(iz1,jz2);
 717             dx20             = _mm_sub_pd(ix2,jx0);
 718             dy20             = _mm_sub_pd(iy2,jy0);
 719             dz20             = _mm_sub_pd(iz2,jz0);
 720             dx21             = _mm_sub_pd(ix2,jx1);
 721             dy21             = _mm_sub_pd(iy2,jy1);
 722             dz21             = _mm_sub_pd(iz2,jz1);
 723             dx22             = _mm_sub_pd(ix2,jx2);
 724             dy22             = _mm_sub_pd(iy2,jy2);
 725             dz22             = _mm_sub_pd(iz2,jz2);
 726
 727             /* Calculate squared distance and things based on it */
 728             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 729             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 730             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 731             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 732             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 733             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 734             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 735             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 736             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 737
 738             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 739             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 740             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 741             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 742             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 743             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 744             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 745             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 746             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 747
 748             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 749             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 750             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 751             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 752             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 753             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 754             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 755             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 756             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 757
 758             fjx0             = _mm_setzero_pd();
 759             fjy0             = _mm_setzero_pd();
 760             fjz0             = _mm_setzero_pd();
 761             fjx1             = _mm_setzero_pd();
 762             fjy1             = _mm_setzero_pd();
 763             fjz1             = _mm_setzero_pd();
 764             fjx2             = _mm_setzero_pd();
 765             fjy2             = _mm_setzero_pd();
 766             fjz2             = _mm_setzero_pd();
 767
 768             /**************************
 769              * CALCULATE INTERACTIONS *
 770              **************************/
 771
 772             r00              = _mm_mul_pd(rsq00,rinv00);
 773
 774             /* EWALD ELECTROSTATICS */
 775
 776             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 777             ewrt             = _mm_mul_pd(r00,ewtabscale);
 778             ewitab           = _mm_cvttpd_epi32(ewrt);
 779             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 780             ewitab           = _mm_slli_epi32(ewitab,2);
 781             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 782             ewtabD           = _mm_setzero_pd();
 783             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 784             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 785             ewtabFn          = _mm_setzero_pd();
 786             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 787             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 788             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 789             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 790             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 791
 792             /* LENNARD-JONES DISPERSION/REPULSION */
 793
 794             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 795             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 796             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 797             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 798             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 799
 800             /* Update potential sum for this i atom from the interaction with this j atom. */
 801             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 802             velecsum         = _mm_add_pd(velecsum,velec);
 803             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 804             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 805
 806             fscal            = _mm_add_pd(felec,fvdw);
 807
 808             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 809
 810             /* Calculate temporary vectorial force */
 811             tx               = _mm_mul_pd(fscal,dx00);
 812             ty               = _mm_mul_pd(fscal,dy00);
 813             tz               = _mm_mul_pd(fscal,dz00);
 814
 815             /* Update vectorial force */
 816             fix0             = _mm_add_pd(fix0,tx);
 817             fiy0             = _mm_add_pd(fiy0,ty);
 818             fiz0             = _mm_add_pd(fiz0,tz);
 819
 820             fjx0             = _mm_add_pd(fjx0,tx);
 821             fjy0             = _mm_add_pd(fjy0,ty);
 822             fjz0             = _mm_add_pd(fjz0,tz);
 823
 824             /**************************
 825              * CALCULATE INTERACTIONS *
 826              **************************/
 827
 828             r01              = _mm_mul_pd(rsq01,rinv01);
 829
 830             /* EWALD ELECTROSTATICS */
 831
 832             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 833             ewrt             = _mm_mul_pd(r01,ewtabscale);
 834             ewitab           = _mm_cvttpd_epi32(ewrt);
 835             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 836             ewitab           = _mm_slli_epi32(ewitab,2);
 837             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 838             ewtabD           = _mm_setzero_pd();
 839             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 840             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 841             ewtabFn          = _mm_setzero_pd();
 842             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 843             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 844             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 845             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 846             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 847
 848             /* Update potential sum for this i atom from the interaction with this j atom. */
 849             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 850             velecsum         = _mm_add_pd(velecsum,velec);
 851
 852             fscal            = felec;
 853
 854             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 855
 856             /* Calculate temporary vectorial force */
 857             tx               = _mm_mul_pd(fscal,dx01);
 858             ty               = _mm_mul_pd(fscal,dy01);
 859             tz               = _mm_mul_pd(fscal,dz01);
 860
 861             /* Update vectorial force */
 862             fix0             = _mm_add_pd(fix0,tx);
 863             fiy0             = _mm_add_pd(fiy0,ty);
 864             fiz0             = _mm_add_pd(fiz0,tz);
 865
 866             fjx1             = _mm_add_pd(fjx1,tx);
 867             fjy1             = _mm_add_pd(fjy1,ty);
 868             fjz1             = _mm_add_pd(fjz1,tz);
 869
 870             /**************************
 871              * CALCULATE INTERACTIONS *
 872              **************************/
 873
 874             r02              = _mm_mul_pd(rsq02,rinv02);
 875
 876             /* EWALD ELECTROSTATICS */
 877
 878             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 879             ewrt             = _mm_mul_pd(r02,ewtabscale);
 880             ewitab           = _mm_cvttpd_epi32(ewrt);
 881             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 882             ewitab           = _mm_slli_epi32(ewitab,2);
 883             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 884             ewtabD           = _mm_setzero_pd();
 885             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 886             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 887             ewtabFn          = _mm_setzero_pd();
 888             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 889             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 890             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 891             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 892             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 893
 894             /* Update potential sum for this i atom from the interaction with this j atom. */
 895             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 896             velecsum         = _mm_add_pd(velecsum,velec);
 897
 898             fscal            = felec;
 899
 900             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 901
 902             /* Calculate temporary vectorial force */
 903             tx               = _mm_mul_pd(fscal,dx02);
 904             ty               = _mm_mul_pd(fscal,dy02);
 905             tz               = _mm_mul_pd(fscal,dz02);
 906
 907             /* Update vectorial force */
 908             fix0             = _mm_add_pd(fix0,tx);
 909             fiy0             = _mm_add_pd(fiy0,ty);
 910             fiz0             = _mm_add_pd(fiz0,tz);
 911
 912             fjx2             = _mm_add_pd(fjx2,tx);
 913             fjy2             = _mm_add_pd(fjy2,ty);
 914             fjz2             = _mm_add_pd(fjz2,tz);
 915
 916             /**************************
 917              * CALCULATE INTERACTIONS *
 918              **************************/
 919
 920             r10              = _mm_mul_pd(rsq10,rinv10);
 921
 922             /* EWALD ELECTROSTATICS */
 923
 924             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 925             ewrt             = _mm_mul_pd(r10,ewtabscale);
 926             ewitab           = _mm_cvttpd_epi32(ewrt);
 927             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 928             ewitab           = _mm_slli_epi32(ewitab,2);
 929             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 930             ewtabD           = _mm_setzero_pd();
 931             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 932             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 933             ewtabFn          = _mm_setzero_pd();
 934             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 935             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 936             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 937             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 938             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 939
 940             /* Update potential sum for this i atom from the interaction with this j atom. */
 941             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 942             velecsum         = _mm_add_pd(velecsum,velec);
 943
 944             fscal            = felec;
 945
 946             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 947
 948             /* Calculate temporary vectorial force */
 949             tx               = _mm_mul_pd(fscal,dx10);
 950             ty               = _mm_mul_pd(fscal,dy10);
 951             tz               = _mm_mul_pd(fscal,dz10);
 952
 953             /* Update vectorial force */
 954             fix1             = _mm_add_pd(fix1,tx);
 955             fiy1             = _mm_add_pd(fiy1,ty);
 956             fiz1             = _mm_add_pd(fiz1,tz);
 957
 958             fjx0             = _mm_add_pd(fjx0,tx);
 959             fjy0             = _mm_add_pd(fjy0,ty);
 960             fjz0             = _mm_add_pd(fjz0,tz);
 961
 962             /**************************
 963              * CALCULATE INTERACTIONS *
 964              **************************/
 965
 966             r11              = _mm_mul_pd(rsq11,rinv11);
 967
 968             /* EWALD ELECTROSTATICS */
 969
 970             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 971             ewrt             = _mm_mul_pd(r11,ewtabscale);
 972             ewitab           = _mm_cvttpd_epi32(ewrt);
 973             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 974             ewitab           = _mm_slli_epi32(ewitab,2);
 975             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 976             ewtabD           = _mm_setzero_pd();
 977             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 978             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 979             ewtabFn          = _mm_setzero_pd();
 980             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 981             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 982             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 983             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 984             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 985
 986             /* Update potential sum for this i atom from the interaction with this j atom. */
 987             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 988             velecsum         = _mm_add_pd(velecsum,velec);
 989
 990             fscal            = felec;
 991
 992             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 993
 994             /* Calculate temporary vectorial force */
 995             tx               = _mm_mul_pd(fscal,dx11);
 996             ty               = _mm_mul_pd(fscal,dy11);
 997             tz               = _mm_mul_pd(fscal,dz11);
 998
 999             /* Update vectorial force */
1000             fix1             = _mm_add_pd(fix1,tx);
1001             fiy1             = _mm_add_pd(fiy1,ty);
1002             fiz1             = _mm_add_pd(fiz1,tz);
1003
1004             fjx1             = _mm_add_pd(fjx1,tx);
1005             fjy1             = _mm_add_pd(fjy1,ty);
1006             fjz1             = _mm_add_pd(fjz1,tz);
1007
1008             /**************************
1009              * CALCULATE INTERACTIONS *
1010              **************************/
1011
1012             r12              = _mm_mul_pd(rsq12,rinv12);
1013
1014             /* EWALD ELECTROSTATICS */
1015
1016             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1017             ewrt             = _mm_mul_pd(r12,ewtabscale);
1018             ewitab           = _mm_cvttpd_epi32(ewrt);
1019             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1020             ewitab           = _mm_slli_epi32(ewitab,2);
1021             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1022             ewtabD           = _mm_setzero_pd();
1023             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1024             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1025             ewtabFn          = _mm_setzero_pd();
1026             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1027             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1028             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1029             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
1030             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1031
1032             /* Update potential sum for this i atom from the interaction with this j atom. */
1033             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1034             velecsum         = _mm_add_pd(velecsum,velec);
1035
1036             fscal            = felec;
1037
1038             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1039
1040             /* Calculate temporary vectorial force */
1041             tx               = _mm_mul_pd(fscal,dx12);
1042             ty               = _mm_mul_pd(fscal,dy12);
1043             tz               = _mm_mul_pd(fscal,dz12);
1044
1045             /* Update vectorial force */
1046             fix1             = _mm_add_pd(fix1,tx);
1047             fiy1             = _mm_add_pd(fiy1,ty);
1048             fiz1             = _mm_add_pd(fiz1,tz);
1049
1050             fjx2             = _mm_add_pd(fjx2,tx);
1051             fjy2             = _mm_add_pd(fjy2,ty);
1052             fjz2             = _mm_add_pd(fjz2,tz);
1053
1054             /**************************
1055              * CALCULATE INTERACTIONS *
1056              **************************/
1057
1058             r20              = _mm_mul_pd(rsq20,rinv20);
1059
1060             /* EWALD ELECTROSTATICS */
1061
1062             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1063             ewrt             = _mm_mul_pd(r20,ewtabscale);
1064             ewitab           = _mm_cvttpd_epi32(ewrt);
1065             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1066             ewitab           = _mm_slli_epi32(ewitab,2);
1067             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1068             ewtabD           = _mm_setzero_pd();
1069             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1070             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1071             ewtabFn          = _mm_setzero_pd();
1072             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1073             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1074             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1075             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1076             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1077
1078             /* Update potential sum for this i atom from the interaction with this j atom. */
1079             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1080             velecsum         = _mm_add_pd(velecsum,velec);
1081
1082             fscal            = felec;
1083
1084             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1085
1086             /* Calculate temporary vectorial force */
1087             tx               = _mm_mul_pd(fscal,dx20);
1088             ty               = _mm_mul_pd(fscal,dy20);
1089             tz               = _mm_mul_pd(fscal,dz20);
1090
1091             /* Update vectorial force */
1092             fix2             = _mm_add_pd(fix2,tx);
1093             fiy2             = _mm_add_pd(fiy2,ty);
1094             fiz2             = _mm_add_pd(fiz2,tz);
1095
1096             fjx0             = _mm_add_pd(fjx0,tx);
1097             fjy0             = _mm_add_pd(fjy0,ty);
1098             fjz0             = _mm_add_pd(fjz0,tz);
1099
1100             /**************************
1101              * CALCULATE INTERACTIONS *
1102              **************************/
1103
1104             r21              = _mm_mul_pd(rsq21,rinv21);
1105
1106             /* EWALD ELECTROSTATICS */
1107
1108             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1109             ewrt             = _mm_mul_pd(r21,ewtabscale);
1110             ewitab           = _mm_cvttpd_epi32(ewrt);
1111             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1112             ewitab           = _mm_slli_epi32(ewitab,2);
1113             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1114             ewtabD           = _mm_setzero_pd();
1115             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1116             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1117             ewtabFn          = _mm_setzero_pd();
1118             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1119             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1120             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1121             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1122             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1123
1124             /* Update potential sum for this i atom from the interaction with this j atom. */
1125             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1126             velecsum         = _mm_add_pd(velecsum,velec);
1127
1128             fscal            = felec;
1129
1130             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1131
1132             /* Calculate temporary vectorial force */
1133             tx               = _mm_mul_pd(fscal,dx21);
1134             ty               = _mm_mul_pd(fscal,dy21);
1135             tz               = _mm_mul_pd(fscal,dz21);
1136
1137             /* Update vectorial force */
1138             fix2             = _mm_add_pd(fix2,tx);
1139             fiy2             = _mm_add_pd(fiy2,ty);
1140             fiz2             = _mm_add_pd(fiz2,tz);
1141
1142             fjx1             = _mm_add_pd(fjx1,tx);
1143             fjy1             = _mm_add_pd(fjy1,ty);
1144             fjz1             = _mm_add_pd(fjz1,tz);
1145
1146             /**************************
1147              * CALCULATE INTERACTIONS *
1148              **************************/
1149
1150             r22              = _mm_mul_pd(rsq22,rinv22);
1151
1152             /* EWALD ELECTROSTATICS */
1153
1154             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1155             ewrt             = _mm_mul_pd(r22,ewtabscale);
1156             ewitab           = _mm_cvttpd_epi32(ewrt);
1157             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1158             ewitab           = _mm_slli_epi32(ewitab,2);
1159             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1160             ewtabD           = _mm_setzero_pd();
1161             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1162             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1163             ewtabFn          = _mm_setzero_pd();
1164             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1165             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1166             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1167             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1168             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1169
1170             /* Update potential sum for this i atom from the interaction with this j atom. */
1171             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1172             velecsum         = _mm_add_pd(velecsum,velec);
1173
1174             fscal            = felec;
1175
1176             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1177
1178             /* Calculate temporary vectorial force */
1179             tx               = _mm_mul_pd(fscal,dx22);
1180             ty               = _mm_mul_pd(fscal,dy22);
1181             tz               = _mm_mul_pd(fscal,dz22);
1182
1183             /* Update vectorial force */
1184             fix2             = _mm_add_pd(fix2,tx);
1185             fiy2             = _mm_add_pd(fiy2,ty);
1186             fiz2             = _mm_add_pd(fiz2,tz);
1187
1188             fjx2             = _mm_add_pd(fjx2,tx);
1189             fjy2             = _mm_add_pd(fjy2,ty);
1190             fjz2             = _mm_add_pd(fjz2,tz);
1191
1192             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1193
1194             /* Inner loop uses 381 flops */
1195         }
1196
1197         /* End of innermost loop */
1198
1199         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1200                                               f+i_coord_offset,fshift+i_shift_offset);
1201
1202         ggid                        = gid[iidx];
1203         /* Update potential energies */
1204         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1205         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1206
1207         /* Increment number of inner iterations */
1208         inneriter                  += j_index_end - j_index_start;
1209
1210         /* Outer loop uses 20 flops */
1211     }
1212
1213     /* Increment number of outer iterations */
1214     outeriter        += nri;
1215
1216     /* Update outer/inner flops */
1217
1218     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*381);
1219 }
1220 /*
1221  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW3W3_F_sse4_1_double
1222  * Electrostatics interaction: Ewald
1223  * VdW interaction:            LennardJones
1224  * Geometry:                   Water3-Water3
1225  * Calculate force/pot:        Force
1226  */
1227 void
1228 nb_kernel_ElecEw_VdwLJ_GeomW3W3_F_sse4_1_double
1229                     (t_nblist                    * gmx_restrict       nlist,
1230                      rvec                        * gmx_restrict          xx,
1231                      rvec                        * gmx_restrict          ff,
1232                      t_forcerec                  * gmx_restrict          fr,
1233                      t_mdatoms                   * gmx_restrict     mdatoms,
1234                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1235                      t_nrnb                      * gmx_restrict        nrnb)
1236 {
1237     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1238      * just 0 for non-waters.
1239      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1240      * jnr indices corresponding to data put in the four positions in the SIMD register.
1241      */
1242     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1243     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1244     int              jnrA,jnrB;
1245     int              j_coord_offsetA,j_coord_offsetB;
1246     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1247     real             rcutoff_scalar;
1248     real             *shiftvec,*fshift,*x,*f;
1249     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1250     int              vdwioffset0;
1251     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1252     int              vdwioffset1;
1253     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1254     int              vdwioffset2;
1255     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1256     int              vdwjidx0A,vdwjidx0B;
1257     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1258     int              vdwjidx1A,vdwjidx1B;
1259     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1260     int              vdwjidx2A,vdwjidx2B;
1261     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1262     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1263     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1264     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1265     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1266     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1267     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1268     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1269     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1270     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1271     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1272     real             *charge;
1273     int              nvdwtype;
1274     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1275     int              *vdwtype;
1276     real             *vdwparam;
1277     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1278     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1279     __m128i          ewitab;
1280     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1281     real             *ewtab;
1282     __m128d          dummy_mask,cutoff_mask;
1283     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1284     __m128d          one     = _mm_set1_pd(1.0);
1285     __m128d          two     = _mm_set1_pd(2.0);
1286     x                = xx[0];
1287     f                = ff[0];
1288
1289     nri              = nlist->nri;
1290     iinr             = nlist->iinr;
1291     jindex           = nlist->jindex;
1292     jjnr             = nlist->jjnr;
1293     shiftidx         = nlist->shift;
1294     gid              = nlist->gid;
1295     shiftvec         = fr->shift_vec[0];
1296     fshift           = fr->fshift[0];
1297     facel            = _mm_set1_pd(fr->epsfac);
1298     charge           = mdatoms->chargeA;
1299     nvdwtype         = fr->ntype;
1300     vdwparam         = fr->nbfp;
1301     vdwtype          = mdatoms->typeA;
1302
1303     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1304     ewtab            = fr->ic->tabq_coul_F;
1305     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1306     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1307
1308     /* Setup water-specific parameters */
1309     inr              = nlist->iinr[0];
1310     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1311     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1312     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1313     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1314
1315     jq0              = _mm_set1_pd(charge[inr+0]);
1316     jq1              = _mm_set1_pd(charge[inr+1]);
1317     jq2              = _mm_set1_pd(charge[inr+2]);
1318     vdwjidx0A        = 2*vdwtype[inr+0];
1319     qq00             = _mm_mul_pd(iq0,jq0);
1320     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1321     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1322     qq01             = _mm_mul_pd(iq0,jq1);
1323     qq02             = _mm_mul_pd(iq0,jq2);
1324     qq10             = _mm_mul_pd(iq1,jq0);
1325     qq11             = _mm_mul_pd(iq1,jq1);
1326     qq12             = _mm_mul_pd(iq1,jq2);
1327     qq20             = _mm_mul_pd(iq2,jq0);
1328     qq21             = _mm_mul_pd(iq2,jq1);
1329     qq22             = _mm_mul_pd(iq2,jq2);
1330
1331     /* Avoid stupid compiler warnings */
1332     jnrA = jnrB = 0;
1333     j_coord_offsetA = 0;
1334     j_coord_offsetB = 0;
1335
1336     outeriter        = 0;
1337     inneriter        = 0;
1338
1339     /* Start outer loop over neighborlists */
1340     for(iidx=0; iidx<nri; iidx++)
1341     {
1342         /* Load shift vector for this list */
1343         i_shift_offset   = DIM*shiftidx[iidx];
1344
1345         /* Load limits for loop over neighbors */
1346         j_index_start    = jindex[iidx];
1347         j_index_end      = jindex[iidx+1];
1348
1349         /* Get outer coordinate index */
1350         inr              = iinr[iidx];
1351         i_coord_offset   = DIM*inr;
1352
1353         /* Load i particle coords and add shift vector */
1354         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1355                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1356
1357         fix0             = _mm_setzero_pd();
1358         fiy0             = _mm_setzero_pd();
1359         fiz0             = _mm_setzero_pd();
1360         fix1             = _mm_setzero_pd();
1361         fiy1             = _mm_setzero_pd();
1362         fiz1             = _mm_setzero_pd();
1363         fix2             = _mm_setzero_pd();
1364         fiy2             = _mm_setzero_pd();
1365         fiz2             = _mm_setzero_pd();
1366
1367         /* Start inner kernel loop */
1368         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1369         {
1370
1371             /* Get j neighbor index, and coordinate index */
1372             jnrA             = jjnr[jidx];
1373             jnrB             = jjnr[jidx+1];
1374             j_coord_offsetA  = DIM*jnrA;
1375             j_coord_offsetB  = DIM*jnrB;
1376
1377             /* load j atom coordinates */
1378             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1379                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1380
1381             /* Calculate displacement vector */
1382             dx00             = _mm_sub_pd(ix0,jx0);
1383             dy00             = _mm_sub_pd(iy0,jy0);
1384             dz00             = _mm_sub_pd(iz0,jz0);
1385             dx01             = _mm_sub_pd(ix0,jx1);
1386             dy01             = _mm_sub_pd(iy0,jy1);
1387             dz01             = _mm_sub_pd(iz0,jz1);
1388             dx02             = _mm_sub_pd(ix0,jx2);
1389             dy02             = _mm_sub_pd(iy0,jy2);
1390             dz02             = _mm_sub_pd(iz0,jz2);
1391             dx10             = _mm_sub_pd(ix1,jx0);
1392             dy10             = _mm_sub_pd(iy1,jy0);
1393             dz10             = _mm_sub_pd(iz1,jz0);
1394             dx11             = _mm_sub_pd(ix1,jx1);
1395             dy11             = _mm_sub_pd(iy1,jy1);
1396             dz11             = _mm_sub_pd(iz1,jz1);
1397             dx12             = _mm_sub_pd(ix1,jx2);
1398             dy12             = _mm_sub_pd(iy1,jy2);
1399             dz12             = _mm_sub_pd(iz1,jz2);
1400             dx20             = _mm_sub_pd(ix2,jx0);
1401             dy20             = _mm_sub_pd(iy2,jy0);
1402             dz20             = _mm_sub_pd(iz2,jz0);
1403             dx21             = _mm_sub_pd(ix2,jx1);
1404             dy21             = _mm_sub_pd(iy2,jy1);
1405             dz21             = _mm_sub_pd(iz2,jz1);
1406             dx22             = _mm_sub_pd(ix2,jx2);
1407             dy22             = _mm_sub_pd(iy2,jy2);
1408             dz22             = _mm_sub_pd(iz2,jz2);
1409
1410             /* Calculate squared distance and things based on it */
1411             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1412             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1413             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1414             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1415             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1416             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1417             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1418             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1419             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1420
1421             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1422             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1423             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1424             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1425             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1426             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1427             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1428             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1429             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1430
1431             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1432             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1433             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1434             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1435             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1436             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1437             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1438             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1439             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1440
1441             fjx0             = _mm_setzero_pd();
1442             fjy0             = _mm_setzero_pd();
1443             fjz0             = _mm_setzero_pd();
1444             fjx1             = _mm_setzero_pd();
1445             fjy1             = _mm_setzero_pd();
1446             fjz1             = _mm_setzero_pd();
1447             fjx2             = _mm_setzero_pd();
1448             fjy2             = _mm_setzero_pd();
1449             fjz2             = _mm_setzero_pd();
1450
1451             /**************************
1452              * CALCULATE INTERACTIONS *
1453              **************************/
1454
1455             r00              = _mm_mul_pd(rsq00,rinv00);
1456
1457             /* EWALD ELECTROSTATICS */
1458
1459             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1460             ewrt             = _mm_mul_pd(r00,ewtabscale);
1461             ewitab           = _mm_cvttpd_epi32(ewrt);
1462             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1463             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1464                                          &ewtabF,&ewtabFn);
1465             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1466             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1467
1468             /* LENNARD-JONES DISPERSION/REPULSION */
1469
1470             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1471             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1472
1473             fscal            = _mm_add_pd(felec,fvdw);
1474
1475             /* Calculate temporary vectorial force */
1476             tx               = _mm_mul_pd(fscal,dx00);
1477             ty               = _mm_mul_pd(fscal,dy00);
1478             tz               = _mm_mul_pd(fscal,dz00);
1479
1480             /* Update vectorial force */
1481             fix0             = _mm_add_pd(fix0,tx);
1482             fiy0             = _mm_add_pd(fiy0,ty);
1483             fiz0             = _mm_add_pd(fiz0,tz);
1484
1485             fjx0             = _mm_add_pd(fjx0,tx);
1486             fjy0             = _mm_add_pd(fjy0,ty);
1487             fjz0             = _mm_add_pd(fjz0,tz);
1488
1489             /**************************
1490              * CALCULATE INTERACTIONS *
1491              **************************/
1492
1493             r01              = _mm_mul_pd(rsq01,rinv01);
1494
1495             /* EWALD ELECTROSTATICS */
1496
1497             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1498             ewrt             = _mm_mul_pd(r01,ewtabscale);
1499             ewitab           = _mm_cvttpd_epi32(ewrt);
1500             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1501             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1502                                          &ewtabF,&ewtabFn);
1503             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1504             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1505
1506             fscal            = felec;
1507
1508             /* Calculate temporary vectorial force */
1509             tx               = _mm_mul_pd(fscal,dx01);
1510             ty               = _mm_mul_pd(fscal,dy01);
1511             tz               = _mm_mul_pd(fscal,dz01);
1512
1513             /* Update vectorial force */
1514             fix0             = _mm_add_pd(fix0,tx);
1515             fiy0             = _mm_add_pd(fiy0,ty);
1516             fiz0             = _mm_add_pd(fiz0,tz);
1517
1518             fjx1             = _mm_add_pd(fjx1,tx);
1519             fjy1             = _mm_add_pd(fjy1,ty);
1520             fjz1             = _mm_add_pd(fjz1,tz);
1521
1522             /**************************
1523              * CALCULATE INTERACTIONS *
1524              **************************/
1525
1526             r02              = _mm_mul_pd(rsq02,rinv02);
1527
1528             /* EWALD ELECTROSTATICS */
1529
1530             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1531             ewrt             = _mm_mul_pd(r02,ewtabscale);
1532             ewitab           = _mm_cvttpd_epi32(ewrt);
1533             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1534             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1535                                          &ewtabF,&ewtabFn);
1536             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1537             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1538
1539             fscal            = felec;
1540
1541             /* Calculate temporary vectorial force */
1542             tx               = _mm_mul_pd(fscal,dx02);
1543             ty               = _mm_mul_pd(fscal,dy02);
1544             tz               = _mm_mul_pd(fscal,dz02);
1545
1546             /* Update vectorial force */
1547             fix0             = _mm_add_pd(fix0,tx);
1548             fiy0             = _mm_add_pd(fiy0,ty);
1549             fiz0             = _mm_add_pd(fiz0,tz);
1550
1551             fjx2             = _mm_add_pd(fjx2,tx);
1552             fjy2             = _mm_add_pd(fjy2,ty);
1553             fjz2             = _mm_add_pd(fjz2,tz);
1554
1555             /**************************
1556              * CALCULATE INTERACTIONS *
1557              **************************/
1558
1559             r10              = _mm_mul_pd(rsq10,rinv10);
1560
1561             /* EWALD ELECTROSTATICS */
1562
1563             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1564             ewrt             = _mm_mul_pd(r10,ewtabscale);
1565             ewitab           = _mm_cvttpd_epi32(ewrt);
1566             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1567             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1568                                          &ewtabF,&ewtabFn);
1569             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1570             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1571
1572             fscal            = felec;
1573
1574             /* Calculate temporary vectorial force */
1575             tx               = _mm_mul_pd(fscal,dx10);
1576             ty               = _mm_mul_pd(fscal,dy10);
1577             tz               = _mm_mul_pd(fscal,dz10);
1578
1579             /* Update vectorial force */
1580             fix1             = _mm_add_pd(fix1,tx);
1581             fiy1             = _mm_add_pd(fiy1,ty);
1582             fiz1             = _mm_add_pd(fiz1,tz);
1583
1584             fjx0             = _mm_add_pd(fjx0,tx);
1585             fjy0             = _mm_add_pd(fjy0,ty);
1586             fjz0             = _mm_add_pd(fjz0,tz);
1587
1588             /**************************
1589              * CALCULATE INTERACTIONS *
1590              **************************/
1591
1592             r11              = _mm_mul_pd(rsq11,rinv11);
1593
1594             /* EWALD ELECTROSTATICS */
1595
1596             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1597             ewrt             = _mm_mul_pd(r11,ewtabscale);
1598             ewitab           = _mm_cvttpd_epi32(ewrt);
1599             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1600             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1601                                          &ewtabF,&ewtabFn);
1602             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1603             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1604
1605             fscal            = felec;
1606
1607             /* Calculate temporary vectorial force */
1608             tx               = _mm_mul_pd(fscal,dx11);
1609             ty               = _mm_mul_pd(fscal,dy11);
1610             tz               = _mm_mul_pd(fscal,dz11);
1611
1612             /* Update vectorial force */
1613             fix1             = _mm_add_pd(fix1,tx);
1614             fiy1             = _mm_add_pd(fiy1,ty);
1615             fiz1             = _mm_add_pd(fiz1,tz);
1616
1617             fjx1             = _mm_add_pd(fjx1,tx);
1618             fjy1             = _mm_add_pd(fjy1,ty);
1619             fjz1             = _mm_add_pd(fjz1,tz);
1620
1621             /**************************
1622              * CALCULATE INTERACTIONS *
1623              **************************/
1624
1625             r12              = _mm_mul_pd(rsq12,rinv12);
1626
1627             /* EWALD ELECTROSTATICS */
1628
1629             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1630             ewrt             = _mm_mul_pd(r12,ewtabscale);
1631             ewitab           = _mm_cvttpd_epi32(ewrt);
1632             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1633             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1634                                          &ewtabF,&ewtabFn);
1635             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1636             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1637
1638             fscal            = felec;
1639
1640             /* Calculate temporary vectorial force */
1641             tx               = _mm_mul_pd(fscal,dx12);
1642             ty               = _mm_mul_pd(fscal,dy12);
1643             tz               = _mm_mul_pd(fscal,dz12);
1644
1645             /* Update vectorial force */
1646             fix1             = _mm_add_pd(fix1,tx);
1647             fiy1             = _mm_add_pd(fiy1,ty);
1648             fiz1             = _mm_add_pd(fiz1,tz);
1649
1650             fjx2             = _mm_add_pd(fjx2,tx);
1651             fjy2             = _mm_add_pd(fjy2,ty);
1652             fjz2             = _mm_add_pd(fjz2,tz);
1653
1654             /**************************
1655              * CALCULATE INTERACTIONS *
1656              **************************/
1657
1658             r20              = _mm_mul_pd(rsq20,rinv20);
1659
1660             /* EWALD ELECTROSTATICS */
1661
1662             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1663             ewrt             = _mm_mul_pd(r20,ewtabscale);
1664             ewitab           = _mm_cvttpd_epi32(ewrt);
1665             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1666             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1667                                          &ewtabF,&ewtabFn);
1668             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1669             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1670
1671             fscal            = felec;
1672
1673             /* Calculate temporary vectorial force */
1674             tx               = _mm_mul_pd(fscal,dx20);
1675             ty               = _mm_mul_pd(fscal,dy20);
1676             tz               = _mm_mul_pd(fscal,dz20);
1677
1678             /* Update vectorial force */
1679             fix2             = _mm_add_pd(fix2,tx);
1680             fiy2             = _mm_add_pd(fiy2,ty);
1681             fiz2             = _mm_add_pd(fiz2,tz);
1682
1683             fjx0             = _mm_add_pd(fjx0,tx);
1684             fjy0             = _mm_add_pd(fjy0,ty);
1685             fjz0             = _mm_add_pd(fjz0,tz);
1686
1687             /**************************
1688              * CALCULATE INTERACTIONS *
1689              **************************/
1690
1691             r21              = _mm_mul_pd(rsq21,rinv21);
1692
1693             /* EWALD ELECTROSTATICS */
1694
1695             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1696             ewrt             = _mm_mul_pd(r21,ewtabscale);
1697             ewitab           = _mm_cvttpd_epi32(ewrt);
1698             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1699             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1700                                          &ewtabF,&ewtabFn);
1701             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1702             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1703
1704             fscal            = felec;
1705
1706             /* Calculate temporary vectorial force */
1707             tx               = _mm_mul_pd(fscal,dx21);
1708             ty               = _mm_mul_pd(fscal,dy21);
1709             tz               = _mm_mul_pd(fscal,dz21);
1710
1711             /* Update vectorial force */
1712             fix2             = _mm_add_pd(fix2,tx);
1713             fiy2             = _mm_add_pd(fiy2,ty);
1714             fiz2             = _mm_add_pd(fiz2,tz);
1715
1716             fjx1             = _mm_add_pd(fjx1,tx);
1717             fjy1             = _mm_add_pd(fjy1,ty);
1718             fjz1             = _mm_add_pd(fjz1,tz);
1719
1720             /**************************
1721              * CALCULATE INTERACTIONS *
1722              **************************/
1723
1724             r22              = _mm_mul_pd(rsq22,rinv22);
1725
1726             /* EWALD ELECTROSTATICS */
1727
1728             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1729             ewrt             = _mm_mul_pd(r22,ewtabscale);
1730             ewitab           = _mm_cvttpd_epi32(ewrt);
1731             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1732             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1733                                          &ewtabF,&ewtabFn);
1734             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1735             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1736
1737             fscal            = felec;
1738
1739             /* Calculate temporary vectorial force */
1740             tx               = _mm_mul_pd(fscal,dx22);
1741             ty               = _mm_mul_pd(fscal,dy22);
1742             tz               = _mm_mul_pd(fscal,dz22);
1743
1744             /* Update vectorial force */
1745             fix2             = _mm_add_pd(fix2,tx);
1746             fiy2             = _mm_add_pd(fiy2,ty);
1747             fiz2             = _mm_add_pd(fiz2,tz);
1748
1749             fjx2             = _mm_add_pd(fjx2,tx);
1750             fjy2             = _mm_add_pd(fjy2,ty);
1751             fjz2             = _mm_add_pd(fjz2,tz);
1752
1753             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1754
1755             /* Inner loop uses 331 flops */
1756         }
1757
1758         if(jidx<j_index_end)
1759         {
1760
1761             jnrA             = jjnr[jidx];
1762             j_coord_offsetA  = DIM*jnrA;
1763
1764             /* load j atom coordinates */
1765             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1766                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1767
1768             /* Calculate displacement vector */
1769             dx00             = _mm_sub_pd(ix0,jx0);
1770             dy00             = _mm_sub_pd(iy0,jy0);
1771             dz00             = _mm_sub_pd(iz0,jz0);
1772             dx01             = _mm_sub_pd(ix0,jx1);
1773             dy01             = _mm_sub_pd(iy0,jy1);
1774             dz01             = _mm_sub_pd(iz0,jz1);
1775             dx02             = _mm_sub_pd(ix0,jx2);
1776             dy02             = _mm_sub_pd(iy0,jy2);
1777             dz02             = _mm_sub_pd(iz0,jz2);
1778             dx10             = _mm_sub_pd(ix1,jx0);
1779             dy10             = _mm_sub_pd(iy1,jy0);
1780             dz10             = _mm_sub_pd(iz1,jz0);
1781             dx11             = _mm_sub_pd(ix1,jx1);
1782             dy11             = _mm_sub_pd(iy1,jy1);
1783             dz11             = _mm_sub_pd(iz1,jz1);
1784             dx12             = _mm_sub_pd(ix1,jx2);
1785             dy12             = _mm_sub_pd(iy1,jy2);
1786             dz12             = _mm_sub_pd(iz1,jz2);
1787             dx20             = _mm_sub_pd(ix2,jx0);
1788             dy20             = _mm_sub_pd(iy2,jy0);
1789             dz20             = _mm_sub_pd(iz2,jz0);
1790             dx21             = _mm_sub_pd(ix2,jx1);
1791             dy21             = _mm_sub_pd(iy2,jy1);
1792             dz21             = _mm_sub_pd(iz2,jz1);
1793             dx22             = _mm_sub_pd(ix2,jx2);
1794             dy22             = _mm_sub_pd(iy2,jy2);
1795             dz22             = _mm_sub_pd(iz2,jz2);
1796
1797             /* Calculate squared distance and things based on it */
1798             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1799             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1800             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1801             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1802             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1803             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1804             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1805             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1806             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1807
1808             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1809             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1810             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1811             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1812             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1813             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1814             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1815             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1816             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1817
1818             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1819             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1820             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1821             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1822             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1823             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1824             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1825             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1826             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1827
1828             fjx0             = _mm_setzero_pd();
1829             fjy0             = _mm_setzero_pd();
1830             fjz0             = _mm_setzero_pd();
1831             fjx1             = _mm_setzero_pd();
1832             fjy1             = _mm_setzero_pd();
1833             fjz1             = _mm_setzero_pd();
1834             fjx2             = _mm_setzero_pd();
1835             fjy2             = _mm_setzero_pd();
1836             fjz2             = _mm_setzero_pd();
1837
1838             /**************************
1839              * CALCULATE INTERACTIONS *
1840              **************************/
1841
1842             r00              = _mm_mul_pd(rsq00,rinv00);
1843
1844             /* EWALD ELECTROSTATICS */
1845
1846             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1847             ewrt             = _mm_mul_pd(r00,ewtabscale);
1848             ewitab           = _mm_cvttpd_epi32(ewrt);
1849             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1850             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1851             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1852             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1853
1854             /* LENNARD-JONES DISPERSION/REPULSION */
1855
1856             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1857             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1858
1859             fscal            = _mm_add_pd(felec,fvdw);
1860
1861             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1862
1863             /* Calculate temporary vectorial force */
1864             tx               = _mm_mul_pd(fscal,dx00);
1865             ty               = _mm_mul_pd(fscal,dy00);
1866             tz               = _mm_mul_pd(fscal,dz00);
1867
1868             /* Update vectorial force */
1869             fix0             = _mm_add_pd(fix0,tx);
1870             fiy0             = _mm_add_pd(fiy0,ty);
1871             fiz0             = _mm_add_pd(fiz0,tz);
1872
1873             fjx0             = _mm_add_pd(fjx0,tx);
1874             fjy0             = _mm_add_pd(fjy0,ty);
1875             fjz0             = _mm_add_pd(fjz0,tz);
1876
1877             /**************************
1878              * CALCULATE INTERACTIONS *
1879              **************************/
1880
1881             r01              = _mm_mul_pd(rsq01,rinv01);
1882
1883             /* EWALD ELECTROSTATICS */
1884
1885             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1886             ewrt             = _mm_mul_pd(r01,ewtabscale);
1887             ewitab           = _mm_cvttpd_epi32(ewrt);
1888             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1889             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1890             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1891             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1892
1893             fscal            = felec;
1894
1895             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1896
1897             /* Calculate temporary vectorial force */
1898             tx               = _mm_mul_pd(fscal,dx01);
1899             ty               = _mm_mul_pd(fscal,dy01);
1900             tz               = _mm_mul_pd(fscal,dz01);
1901
1902             /* Update vectorial force */
1903             fix0             = _mm_add_pd(fix0,tx);
1904             fiy0             = _mm_add_pd(fiy0,ty);
1905             fiz0             = _mm_add_pd(fiz0,tz);
1906
1907             fjx1             = _mm_add_pd(fjx1,tx);
1908             fjy1             = _mm_add_pd(fjy1,ty);
1909             fjz1             = _mm_add_pd(fjz1,tz);
1910
1911             /**************************
1912              * CALCULATE INTERACTIONS *
1913              **************************/
1914
1915             r02              = _mm_mul_pd(rsq02,rinv02);
1916
1917             /* EWALD ELECTROSTATICS */
1918
1919             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1920             ewrt             = _mm_mul_pd(r02,ewtabscale);
1921             ewitab           = _mm_cvttpd_epi32(ewrt);
1922             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1923             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1924             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1925             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1926
1927             fscal            = felec;
1928
1929             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1930
1931             /* Calculate temporary vectorial force */
1932             tx               = _mm_mul_pd(fscal,dx02);
1933             ty               = _mm_mul_pd(fscal,dy02);
1934             tz               = _mm_mul_pd(fscal,dz02);
1935
1936             /* Update vectorial force */
1937             fix0             = _mm_add_pd(fix0,tx);
1938             fiy0             = _mm_add_pd(fiy0,ty);
1939             fiz0             = _mm_add_pd(fiz0,tz);
1940
1941             fjx2             = _mm_add_pd(fjx2,tx);
1942             fjy2             = _mm_add_pd(fjy2,ty);
1943             fjz2             = _mm_add_pd(fjz2,tz);
1944
1945             /**************************
1946              * CALCULATE INTERACTIONS *
1947              **************************/
1948
1949             r10              = _mm_mul_pd(rsq10,rinv10);
1950
1951             /* EWALD ELECTROSTATICS */
1952
1953             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1954             ewrt             = _mm_mul_pd(r10,ewtabscale);
1955             ewitab           = _mm_cvttpd_epi32(ewrt);
1956             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1957             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1958             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1959             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1960
1961             fscal            = felec;
1962
1963             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1964
1965             /* Calculate temporary vectorial force */
1966             tx               = _mm_mul_pd(fscal,dx10);
1967             ty               = _mm_mul_pd(fscal,dy10);
1968             tz               = _mm_mul_pd(fscal,dz10);
1969
1970             /* Update vectorial force */
1971             fix1             = _mm_add_pd(fix1,tx);
1972             fiy1             = _mm_add_pd(fiy1,ty);
1973             fiz1             = _mm_add_pd(fiz1,tz);
1974
1975             fjx0             = _mm_add_pd(fjx0,tx);
1976             fjy0             = _mm_add_pd(fjy0,ty);
1977             fjz0             = _mm_add_pd(fjz0,tz);
1978
1979             /**************************
1980              * CALCULATE INTERACTIONS *
1981              **************************/
1982
1983             r11              = _mm_mul_pd(rsq11,rinv11);
1984
1985             /* EWALD ELECTROSTATICS */
1986
1987             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1988             ewrt             = _mm_mul_pd(r11,ewtabscale);
1989             ewitab           = _mm_cvttpd_epi32(ewrt);
1990             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1991             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1992             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1993             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1994
1995             fscal            = felec;
1996
1997             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1998
1999             /* Calculate temporary vectorial force */
2000             tx               = _mm_mul_pd(fscal,dx11);
2001             ty               = _mm_mul_pd(fscal,dy11);
2002             tz               = _mm_mul_pd(fscal,dz11);
2003
2004             /* Update vectorial force */
2005             fix1             = _mm_add_pd(fix1,tx);
2006             fiy1             = _mm_add_pd(fiy1,ty);
2007             fiz1             = _mm_add_pd(fiz1,tz);
2008
2009             fjx1             = _mm_add_pd(fjx1,tx);
2010             fjy1             = _mm_add_pd(fjy1,ty);
2011             fjz1             = _mm_add_pd(fjz1,tz);
2012
2013             /**************************
2014              * CALCULATE INTERACTIONS *
2015              **************************/
2016
2017             r12              = _mm_mul_pd(rsq12,rinv12);
2018
2019             /* EWALD ELECTROSTATICS */
2020
2021             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2022             ewrt             = _mm_mul_pd(r12,ewtabscale);
2023             ewitab           = _mm_cvttpd_epi32(ewrt);
2024             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2025             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2026             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2027             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2028
2029             fscal            = felec;
2030
2031             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2032
2033             /* Calculate temporary vectorial force */
2034             tx               = _mm_mul_pd(fscal,dx12);
2035             ty               = _mm_mul_pd(fscal,dy12);
2036             tz               = _mm_mul_pd(fscal,dz12);
2037
2038             /* Update vectorial force */
2039             fix1             = _mm_add_pd(fix1,tx);
2040             fiy1             = _mm_add_pd(fiy1,ty);
2041             fiz1             = _mm_add_pd(fiz1,tz);
2042
2043             fjx2             = _mm_add_pd(fjx2,tx);
2044             fjy2             = _mm_add_pd(fjy2,ty);
2045             fjz2             = _mm_add_pd(fjz2,tz);
2046
2047             /**************************
2048              * CALCULATE INTERACTIONS *
2049              **************************/
2050
2051             r20              = _mm_mul_pd(rsq20,rinv20);
2052
2053             /* EWALD ELECTROSTATICS */
2054
2055             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2056             ewrt             = _mm_mul_pd(r20,ewtabscale);
2057             ewitab           = _mm_cvttpd_epi32(ewrt);
2058             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2059             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2060             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2061             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2062
2063             fscal            = felec;
2064
2065             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2066
2067             /* Calculate temporary vectorial force */
2068             tx               = _mm_mul_pd(fscal,dx20);
2069             ty               = _mm_mul_pd(fscal,dy20);
2070             tz               = _mm_mul_pd(fscal,dz20);
2071
2072             /* Update vectorial force */
2073             fix2             = _mm_add_pd(fix2,tx);
2074             fiy2             = _mm_add_pd(fiy2,ty);
2075             fiz2             = _mm_add_pd(fiz2,tz);
2076
2077             fjx0             = _mm_add_pd(fjx0,tx);
2078             fjy0             = _mm_add_pd(fjy0,ty);
2079             fjz0             = _mm_add_pd(fjz0,tz);
2080
2081             /**************************
2082              * CALCULATE INTERACTIONS *
2083              **************************/
2084
2085             r21              = _mm_mul_pd(rsq21,rinv21);
2086
2087             /* EWALD ELECTROSTATICS */
2088
2089             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2090             ewrt             = _mm_mul_pd(r21,ewtabscale);
2091             ewitab           = _mm_cvttpd_epi32(ewrt);
2092             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2093             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2094             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2095             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2096
2097             fscal            = felec;
2098
2099             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2100
2101             /* Calculate temporary vectorial force */
2102             tx               = _mm_mul_pd(fscal,dx21);
2103             ty               = _mm_mul_pd(fscal,dy21);
2104             tz               = _mm_mul_pd(fscal,dz21);
2105
2106             /* Update vectorial force */
2107             fix2             = _mm_add_pd(fix2,tx);
2108             fiy2             = _mm_add_pd(fiy2,ty);
2109             fiz2             = _mm_add_pd(fiz2,tz);
2110
2111             fjx1             = _mm_add_pd(fjx1,tx);
2112             fjy1             = _mm_add_pd(fjy1,ty);
2113             fjz1             = _mm_add_pd(fjz1,tz);
2114
2115             /**************************
2116              * CALCULATE INTERACTIONS *
2117              **************************/
2118
2119             r22              = _mm_mul_pd(rsq22,rinv22);
2120
2121             /* EWALD ELECTROSTATICS */
2122
2123             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2124             ewrt             = _mm_mul_pd(r22,ewtabscale);
2125             ewitab           = _mm_cvttpd_epi32(ewrt);
2126             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2127             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2128             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2129             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2130
2131             fscal            = felec;
2132
2133             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2134
2135             /* Calculate temporary vectorial force */
2136             tx               = _mm_mul_pd(fscal,dx22);
2137             ty               = _mm_mul_pd(fscal,dy22);
2138             tz               = _mm_mul_pd(fscal,dz22);
2139
2140             /* Update vectorial force */
2141             fix2             = _mm_add_pd(fix2,tx);
2142             fiy2             = _mm_add_pd(fiy2,ty);
2143             fiz2             = _mm_add_pd(fiz2,tz);
2144
2145             fjx2             = _mm_add_pd(fjx2,tx);
2146             fjy2             = _mm_add_pd(fjy2,ty);
2147             fjz2             = _mm_add_pd(fjz2,tz);
2148
2149             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2150
2151             /* Inner loop uses 331 flops */
2152         }
2153
2154         /* End of innermost loop */
2155
2156         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2157                                               f+i_coord_offset,fshift+i_shift_offset);
2158
2159         /* Increment number of inner iterations */
2160         inneriter                  += j_index_end - j_index_start;
2161
2162         /* Outer loop uses 18 flops */
2163     }
2164
2165     /* Increment number of outer iterations */
2166     outeriter        += nri;
2167
2168     /* Update outer/inner flops */
2169
2170     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*331);
2171 }