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

   1 /*
   2  * Note: this file was generated by the Gromacs sse4_1_double kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_sse4_1_double.h"
  34 #include "kernelutil_x86_sse4_1_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW3W3_VF_sse4_1_double
  38  * Electrostatics interaction: Ewald
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water3-Water3
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecEw_VdwLJ_GeomW3W3_VF_sse4_1_double
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB;
  61     int              j_coord_offsetA,j_coord_offsetB;
  62     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  63     real             rcutoff_scalar;
  64     real             *shiftvec,*fshift,*x,*f;
  65     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  66     int              vdwioffset0;
  67     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  68     int              vdwioffset1;
  69     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  70     int              vdwioffset2;
  71     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  72     int              vdwjidx0A,vdwjidx0B;
  73     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  74     int              vdwjidx1A,vdwjidx1B;
  75     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  76     int              vdwjidx2A,vdwjidx2B;
  77     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  78     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  79     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
  80     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
  81     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  82     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  83     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  84     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  85     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  86     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  87     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  88     real             *charge;
  89     int              nvdwtype;
  90     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  91     int              *vdwtype;
  92     real             *vdwparam;
  93     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  94     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  95     __m128i          ewitab;
  96     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  97     real             *ewtab;
  98     __m128d          dummy_mask,cutoff_mask;
  99     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 100     __m128d          one     = _mm_set1_pd(1.0);
 101     __m128d          two     = _mm_set1_pd(2.0);
 102     x                = xx[0];
 103     f                = ff[0];
 104
 105     nri              = nlist->nri;
 106     iinr             = nlist->iinr;
 107     jindex           = nlist->jindex;
 108     jjnr             = nlist->jjnr;
 109     shiftidx         = nlist->shift;
 110     gid              = nlist->gid;
 111     shiftvec         = fr->shift_vec[0];
 112     fshift           = fr->fshift[0];
 113     facel            = _mm_set1_pd(fr->epsfac);
 114     charge           = mdatoms->chargeA;
 115     nvdwtype         = fr->ntype;
 116     vdwparam         = fr->nbfp;
 117     vdwtype          = mdatoms->typeA;
 118
 119     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 120     ewtab            = fr->ic->tabq_coul_FDV0;
 121     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 122     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 123
 124     /* Setup water-specific parameters */
 125     inr              = nlist->iinr[0];
 126     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 127     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 128     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 129     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 130
 131     jq0              = _mm_set1_pd(charge[inr+0]);
 132     jq1              = _mm_set1_pd(charge[inr+1]);
 133     jq2              = _mm_set1_pd(charge[inr+2]);
 134     vdwjidx0A        = 2*vdwtype[inr+0];
 135     qq00             = _mm_mul_pd(iq0,jq0);
 136     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 137     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 138     qq01             = _mm_mul_pd(iq0,jq1);
 139     qq02             = _mm_mul_pd(iq0,jq2);
 140     qq10             = _mm_mul_pd(iq1,jq0);
 141     qq11             = _mm_mul_pd(iq1,jq1);
 142     qq12             = _mm_mul_pd(iq1,jq2);
 143     qq20             = _mm_mul_pd(iq2,jq0);
 144     qq21             = _mm_mul_pd(iq2,jq1);
 145     qq22             = _mm_mul_pd(iq2,jq2);
 146
 147     /* Avoid stupid compiler warnings */
 148     jnrA = jnrB = 0;
 149     j_coord_offsetA = 0;
 150     j_coord_offsetB = 0;
 151
 152     outeriter        = 0;
 153     inneriter        = 0;
 154
 155     /* Start outer loop over neighborlists */
 156     for(iidx=0; iidx<nri; iidx++)
 157     {
 158         /* Load shift vector for this list */
 159         i_shift_offset   = DIM*shiftidx[iidx];
 160
 161         /* Load limits for loop over neighbors */
 162         j_index_start    = jindex[iidx];
 163         j_index_end      = jindex[iidx+1];
 164
 165         /* Get outer coordinate index */
 166         inr              = iinr[iidx];
 167         i_coord_offset   = DIM*inr;
 168
 169         /* Load i particle coords and add shift vector */
 170         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 171                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 172
 173         fix0             = _mm_setzero_pd();
 174         fiy0             = _mm_setzero_pd();
 175         fiz0             = _mm_setzero_pd();
 176         fix1             = _mm_setzero_pd();
 177         fiy1             = _mm_setzero_pd();
 178         fiz1             = _mm_setzero_pd();
 179         fix2             = _mm_setzero_pd();
 180         fiy2             = _mm_setzero_pd();
 181         fiz2             = _mm_setzero_pd();
 182
 183         /* Reset potential sums */
 184         velecsum         = _mm_setzero_pd();
 185         vvdwsum          = _mm_setzero_pd();
 186
 187         /* Start inner kernel loop */
 188         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 189         {
 190
 191             /* Get j neighbor index, and coordinate index */
 192             jnrA             = jjnr[jidx];
 193             jnrB             = jjnr[jidx+1];
 194             j_coord_offsetA  = DIM*jnrA;
 195             j_coord_offsetB  = DIM*jnrB;
 196
 197             /* load j atom coordinates */
 198             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 199                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 200
 201             /* Calculate displacement vector */
 202             dx00             = _mm_sub_pd(ix0,jx0);
 203             dy00             = _mm_sub_pd(iy0,jy0);
 204             dz00             = _mm_sub_pd(iz0,jz0);
 205             dx01             = _mm_sub_pd(ix0,jx1);
 206             dy01             = _mm_sub_pd(iy0,jy1);
 207             dz01             = _mm_sub_pd(iz0,jz1);
 208             dx02             = _mm_sub_pd(ix0,jx2);
 209             dy02             = _mm_sub_pd(iy0,jy2);
 210             dz02             = _mm_sub_pd(iz0,jz2);
 211             dx10             = _mm_sub_pd(ix1,jx0);
 212             dy10             = _mm_sub_pd(iy1,jy0);
 213             dz10             = _mm_sub_pd(iz1,jz0);
 214             dx11             = _mm_sub_pd(ix1,jx1);
 215             dy11             = _mm_sub_pd(iy1,jy1);
 216             dz11             = _mm_sub_pd(iz1,jz1);
 217             dx12             = _mm_sub_pd(ix1,jx2);
 218             dy12             = _mm_sub_pd(iy1,jy2);
 219             dz12             = _mm_sub_pd(iz1,jz2);
 220             dx20             = _mm_sub_pd(ix2,jx0);
 221             dy20             = _mm_sub_pd(iy2,jy0);
 222             dz20             = _mm_sub_pd(iz2,jz0);
 223             dx21             = _mm_sub_pd(ix2,jx1);
 224             dy21             = _mm_sub_pd(iy2,jy1);
 225             dz21             = _mm_sub_pd(iz2,jz1);
 226             dx22             = _mm_sub_pd(ix2,jx2);
 227             dy22             = _mm_sub_pd(iy2,jy2);
 228             dz22             = _mm_sub_pd(iz2,jz2);
 229
 230             /* Calculate squared distance and things based on it */
 231             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 232             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 233             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 234             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 235             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 236             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 237             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 238             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 239             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 240
 241             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 242             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 243             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 244             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 245             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 246             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 247             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 248             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 249             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 250
 251             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 252             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 253             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 254             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 255             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 256             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 257             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 258             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 259             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 260
 261             fjx0             = _mm_setzero_pd();
 262             fjy0             = _mm_setzero_pd();
 263             fjz0             = _mm_setzero_pd();
 264             fjx1             = _mm_setzero_pd();
 265             fjy1             = _mm_setzero_pd();
 266             fjz1             = _mm_setzero_pd();
 267             fjx2             = _mm_setzero_pd();
 268             fjy2             = _mm_setzero_pd();
 269             fjz2             = _mm_setzero_pd();
 270
 271             /**************************
 272              * CALCULATE INTERACTIONS *
 273              **************************/
 274
 275             r00              = _mm_mul_pd(rsq00,rinv00);
 276
 277             /* EWALD ELECTROSTATICS */
 278
 279             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 280             ewrt             = _mm_mul_pd(r00,ewtabscale);
 281             ewitab           = _mm_cvttpd_epi32(ewrt);
 282             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 283             ewitab           = _mm_slli_epi32(ewitab,2);
 284             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 285             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 286             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 287             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 288             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 289             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 290             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 291             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 292             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 293             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 294
 295             /* LENNARD-JONES DISPERSION/REPULSION */
 296
 297             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 298             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 299             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 300             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 301             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 302
 303             /* Update potential sum for this i atom from the interaction with this j atom. */
 304             velecsum         = _mm_add_pd(velecsum,velec);
 305             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 306
 307             fscal            = _mm_add_pd(felec,fvdw);
 308
 309             /* Calculate temporary vectorial force */
 310             tx               = _mm_mul_pd(fscal,dx00);
 311             ty               = _mm_mul_pd(fscal,dy00);
 312             tz               = _mm_mul_pd(fscal,dz00);
 313
 314             /* Update vectorial force */
 315             fix0             = _mm_add_pd(fix0,tx);
 316             fiy0             = _mm_add_pd(fiy0,ty);
 317             fiz0             = _mm_add_pd(fiz0,tz);
 318
 319             fjx0             = _mm_add_pd(fjx0,tx);
 320             fjy0             = _mm_add_pd(fjy0,ty);
 321             fjz0             = _mm_add_pd(fjz0,tz);
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             r01              = _mm_mul_pd(rsq01,rinv01);
 328
 329             /* EWALD ELECTROSTATICS */
 330
 331             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 332             ewrt             = _mm_mul_pd(r01,ewtabscale);
 333             ewitab           = _mm_cvttpd_epi32(ewrt);
 334             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 335             ewitab           = _mm_slli_epi32(ewitab,2);
 336             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 337             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 338             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 339             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 340             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 341             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 342             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 343             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 344             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 345             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 346
 347             /* Update potential sum for this i atom from the interaction with this j atom. */
 348             velecsum         = _mm_add_pd(velecsum,velec);
 349
 350             fscal            = felec;
 351
 352             /* Calculate temporary vectorial force */
 353             tx               = _mm_mul_pd(fscal,dx01);
 354             ty               = _mm_mul_pd(fscal,dy01);
 355             tz               = _mm_mul_pd(fscal,dz01);
 356
 357             /* Update vectorial force */
 358             fix0             = _mm_add_pd(fix0,tx);
 359             fiy0             = _mm_add_pd(fiy0,ty);
 360             fiz0             = _mm_add_pd(fiz0,tz);
 361
 362             fjx1             = _mm_add_pd(fjx1,tx);
 363             fjy1             = _mm_add_pd(fjy1,ty);
 364             fjz1             = _mm_add_pd(fjz1,tz);
 365
 366             /**************************
 367              * CALCULATE INTERACTIONS *
 368              **************************/
 369
 370             r02              = _mm_mul_pd(rsq02,rinv02);
 371
 372             /* EWALD ELECTROSTATICS */
 373
 374             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 375             ewrt             = _mm_mul_pd(r02,ewtabscale);
 376             ewitab           = _mm_cvttpd_epi32(ewrt);
 377             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 378             ewitab           = _mm_slli_epi32(ewitab,2);
 379             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 380             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 381             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 382             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 383             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 384             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 385             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 386             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 387             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 388             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velecsum         = _mm_add_pd(velecsum,velec);
 392
 393             fscal            = felec;
 394
 395             /* Calculate temporary vectorial force */
 396             tx               = _mm_mul_pd(fscal,dx02);
 397             ty               = _mm_mul_pd(fscal,dy02);
 398             tz               = _mm_mul_pd(fscal,dz02);
 399
 400             /* Update vectorial force */
 401             fix0             = _mm_add_pd(fix0,tx);
 402             fiy0             = _mm_add_pd(fiy0,ty);
 403             fiz0             = _mm_add_pd(fiz0,tz);
 404
 405             fjx2             = _mm_add_pd(fjx2,tx);
 406             fjy2             = _mm_add_pd(fjy2,ty);
 407             fjz2             = _mm_add_pd(fjz2,tz);
 408
 409             /**************************
 410              * CALCULATE INTERACTIONS *
 411              **************************/
 412
 413             r10              = _mm_mul_pd(rsq10,rinv10);
 414
 415             /* EWALD ELECTROSTATICS */
 416
 417             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 418             ewrt             = _mm_mul_pd(r10,ewtabscale);
 419             ewitab           = _mm_cvttpd_epi32(ewrt);
 420             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 421             ewitab           = _mm_slli_epi32(ewitab,2);
 422             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 423             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 424             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 425             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 426             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 427             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 428             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 429             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 430             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 431             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 432
 433             /* Update potential sum for this i atom from the interaction with this j atom. */
 434             velecsum         = _mm_add_pd(velecsum,velec);
 435
 436             fscal            = felec;
 437
 438             /* Calculate temporary vectorial force */
 439             tx               = _mm_mul_pd(fscal,dx10);
 440             ty               = _mm_mul_pd(fscal,dy10);
 441             tz               = _mm_mul_pd(fscal,dz10);
 442
 443             /* Update vectorial force */
 444             fix1             = _mm_add_pd(fix1,tx);
 445             fiy1             = _mm_add_pd(fiy1,ty);
 446             fiz1             = _mm_add_pd(fiz1,tz);
 447
 448             fjx0             = _mm_add_pd(fjx0,tx);
 449             fjy0             = _mm_add_pd(fjy0,ty);
 450             fjz0             = _mm_add_pd(fjz0,tz);
 451
 452             /**************************
 453              * CALCULATE INTERACTIONS *
 454              **************************/
 455
 456             r11              = _mm_mul_pd(rsq11,rinv11);
 457
 458             /* EWALD ELECTROSTATICS */
 459
 460             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 461             ewrt             = _mm_mul_pd(r11,ewtabscale);
 462             ewitab           = _mm_cvttpd_epi32(ewrt);
 463             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 464             ewitab           = _mm_slli_epi32(ewitab,2);
 465             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 466             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 467             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 468             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 469             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 470             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 471             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 472             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 473             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 474             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 475
 476             /* Update potential sum for this i atom from the interaction with this j atom. */
 477             velecsum         = _mm_add_pd(velecsum,velec);
 478
 479             fscal            = felec;
 480
 481             /* Calculate temporary vectorial force */
 482             tx               = _mm_mul_pd(fscal,dx11);
 483             ty               = _mm_mul_pd(fscal,dy11);
 484             tz               = _mm_mul_pd(fscal,dz11);
 485
 486             /* Update vectorial force */
 487             fix1             = _mm_add_pd(fix1,tx);
 488             fiy1             = _mm_add_pd(fiy1,ty);
 489             fiz1             = _mm_add_pd(fiz1,tz);
 490
 491             fjx1             = _mm_add_pd(fjx1,tx);
 492             fjy1             = _mm_add_pd(fjy1,ty);
 493             fjz1             = _mm_add_pd(fjz1,tz);
 494
 495             /**************************
 496              * CALCULATE INTERACTIONS *
 497              **************************/
 498
 499             r12              = _mm_mul_pd(rsq12,rinv12);
 500
 501             /* EWALD ELECTROSTATICS */
 502
 503             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 504             ewrt             = _mm_mul_pd(r12,ewtabscale);
 505             ewitab           = _mm_cvttpd_epi32(ewrt);
 506             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 507             ewitab           = _mm_slli_epi32(ewitab,2);
 508             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 509             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 510             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 511             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 512             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 513             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 514             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 515             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 516             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 517             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 518
 519             /* Update potential sum for this i atom from the interaction with this j atom. */
 520             velecsum         = _mm_add_pd(velecsum,velec);
 521
 522             fscal            = felec;
 523
 524             /* Calculate temporary vectorial force */
 525             tx               = _mm_mul_pd(fscal,dx12);
 526             ty               = _mm_mul_pd(fscal,dy12);
 527             tz               = _mm_mul_pd(fscal,dz12);
 528
 529             /* Update vectorial force */
 530             fix1             = _mm_add_pd(fix1,tx);
 531             fiy1             = _mm_add_pd(fiy1,ty);
 532             fiz1             = _mm_add_pd(fiz1,tz);
 533
 534             fjx2             = _mm_add_pd(fjx2,tx);
 535             fjy2             = _mm_add_pd(fjy2,ty);
 536             fjz2             = _mm_add_pd(fjz2,tz);
 537
 538             /**************************
 539              * CALCULATE INTERACTIONS *
 540              **************************/
 541
 542             r20              = _mm_mul_pd(rsq20,rinv20);
 543
 544             /* EWALD ELECTROSTATICS */
 545
 546             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 547             ewrt             = _mm_mul_pd(r20,ewtabscale);
 548             ewitab           = _mm_cvttpd_epi32(ewrt);
 549             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 550             ewitab           = _mm_slli_epi32(ewitab,2);
 551             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 552             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 553             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 554             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 555             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 556             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 557             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 558             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 559             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 560             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 561
 562             /* Update potential sum for this i atom from the interaction with this j atom. */
 563             velecsum         = _mm_add_pd(velecsum,velec);
 564
 565             fscal            = felec;
 566
 567             /* Calculate temporary vectorial force */
 568             tx               = _mm_mul_pd(fscal,dx20);
 569             ty               = _mm_mul_pd(fscal,dy20);
 570             tz               = _mm_mul_pd(fscal,dz20);
 571
 572             /* Update vectorial force */
 573             fix2             = _mm_add_pd(fix2,tx);
 574             fiy2             = _mm_add_pd(fiy2,ty);
 575             fiz2             = _mm_add_pd(fiz2,tz);
 576
 577             fjx0             = _mm_add_pd(fjx0,tx);
 578             fjy0             = _mm_add_pd(fjy0,ty);
 579             fjz0             = _mm_add_pd(fjz0,tz);
 580
 581             /**************************
 582              * CALCULATE INTERACTIONS *
 583              **************************/
 584
 585             r21              = _mm_mul_pd(rsq21,rinv21);
 586
 587             /* EWALD ELECTROSTATICS */
 588
 589             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 590             ewrt             = _mm_mul_pd(r21,ewtabscale);
 591             ewitab           = _mm_cvttpd_epi32(ewrt);
 592             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 593             ewitab           = _mm_slli_epi32(ewitab,2);
 594             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 595             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 596             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 597             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 598             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 599             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 600             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 601             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 602             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 603             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 604
 605             /* Update potential sum for this i atom from the interaction with this j atom. */
 606             velecsum         = _mm_add_pd(velecsum,velec);
 607
 608             fscal            = felec;
 609
 610             /* Calculate temporary vectorial force */
 611             tx               = _mm_mul_pd(fscal,dx21);
 612             ty               = _mm_mul_pd(fscal,dy21);
 613             tz               = _mm_mul_pd(fscal,dz21);
 614
 615             /* Update vectorial force */
 616             fix2             = _mm_add_pd(fix2,tx);
 617             fiy2             = _mm_add_pd(fiy2,ty);
 618             fiz2             = _mm_add_pd(fiz2,tz);
 619
 620             fjx1             = _mm_add_pd(fjx1,tx);
 621             fjy1             = _mm_add_pd(fjy1,ty);
 622             fjz1             = _mm_add_pd(fjz1,tz);
 623
 624             /**************************
 625              * CALCULATE INTERACTIONS *
 626              **************************/
 627
 628             r22              = _mm_mul_pd(rsq22,rinv22);
 629
 630             /* EWALD ELECTROSTATICS */
 631
 632             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 633             ewrt             = _mm_mul_pd(r22,ewtabscale);
 634             ewitab           = _mm_cvttpd_epi32(ewrt);
 635             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 636             ewitab           = _mm_slli_epi32(ewitab,2);
 637             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 638             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 639             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 640             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 641             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 642             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 643             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 644             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 645             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 646             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 647
 648             /* Update potential sum for this i atom from the interaction with this j atom. */
 649             velecsum         = _mm_add_pd(velecsum,velec);
 650
 651             fscal            = felec;
 652
 653             /* Calculate temporary vectorial force */
 654             tx               = _mm_mul_pd(fscal,dx22);
 655             ty               = _mm_mul_pd(fscal,dy22);
 656             tz               = _mm_mul_pd(fscal,dz22);
 657
 658             /* Update vectorial force */
 659             fix2             = _mm_add_pd(fix2,tx);
 660             fiy2             = _mm_add_pd(fiy2,ty);
 661             fiz2             = _mm_add_pd(fiz2,tz);
 662
 663             fjx2             = _mm_add_pd(fjx2,tx);
 664             fjy2             = _mm_add_pd(fjy2,ty);
 665             fjz2             = _mm_add_pd(fjz2,tz);
 666
 667             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 668
 669             /* Inner loop uses 381 flops */
 670         }
 671
 672         if(jidx<j_index_end)
 673         {
 674
 675             jnrA             = jjnr[jidx];
 676             j_coord_offsetA  = DIM*jnrA;
 677
 678             /* load j atom coordinates */
 679             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 680                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 681
 682             /* Calculate displacement vector */
 683             dx00             = _mm_sub_pd(ix0,jx0);
 684             dy00             = _mm_sub_pd(iy0,jy0);
 685             dz00             = _mm_sub_pd(iz0,jz0);
 686             dx01             = _mm_sub_pd(ix0,jx1);
 687             dy01             = _mm_sub_pd(iy0,jy1);
 688             dz01             = _mm_sub_pd(iz0,jz1);
 689             dx02             = _mm_sub_pd(ix0,jx2);
 690             dy02             = _mm_sub_pd(iy0,jy2);
 691             dz02             = _mm_sub_pd(iz0,jz2);
 692             dx10             = _mm_sub_pd(ix1,jx0);
 693             dy10             = _mm_sub_pd(iy1,jy0);
 694             dz10             = _mm_sub_pd(iz1,jz0);
 695             dx11             = _mm_sub_pd(ix1,jx1);
 696             dy11             = _mm_sub_pd(iy1,jy1);
 697             dz11             = _mm_sub_pd(iz1,jz1);
 698             dx12             = _mm_sub_pd(ix1,jx2);
 699             dy12             = _mm_sub_pd(iy1,jy2);
 700             dz12             = _mm_sub_pd(iz1,jz2);
 701             dx20             = _mm_sub_pd(ix2,jx0);
 702             dy20             = _mm_sub_pd(iy2,jy0);
 703             dz20             = _mm_sub_pd(iz2,jz0);
 704             dx21             = _mm_sub_pd(ix2,jx1);
 705             dy21             = _mm_sub_pd(iy2,jy1);
 706             dz21             = _mm_sub_pd(iz2,jz1);
 707             dx22             = _mm_sub_pd(ix2,jx2);
 708             dy22             = _mm_sub_pd(iy2,jy2);
 709             dz22             = _mm_sub_pd(iz2,jz2);
 710
 711             /* Calculate squared distance and things based on it */
 712             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 713             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 714             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 715             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 716             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 717             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 718             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 719             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 720             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 721
 722             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 723             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 724             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 725             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 726             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 727             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 728             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 729             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 730             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 731
 732             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 733             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 734             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 735             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 736             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 737             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 738             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 739             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 740             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 741
 742             fjx0             = _mm_setzero_pd();
 743             fjy0             = _mm_setzero_pd();
 744             fjz0             = _mm_setzero_pd();
 745             fjx1             = _mm_setzero_pd();
 746             fjy1             = _mm_setzero_pd();
 747             fjz1             = _mm_setzero_pd();
 748             fjx2             = _mm_setzero_pd();
 749             fjy2             = _mm_setzero_pd();
 750             fjz2             = _mm_setzero_pd();
 751
 752             /**************************
 753              * CALCULATE INTERACTIONS *
 754              **************************/
 755
 756             r00              = _mm_mul_pd(rsq00,rinv00);
 757
 758             /* EWALD ELECTROSTATICS */
 759
 760             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 761             ewrt             = _mm_mul_pd(r00,ewtabscale);
 762             ewitab           = _mm_cvttpd_epi32(ewrt);
 763             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 764             ewitab           = _mm_slli_epi32(ewitab,2);
 765             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 766             ewtabD           = _mm_setzero_pd();
 767             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 768             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 769             ewtabFn          = _mm_setzero_pd();
 770             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 771             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 772             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 773             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 774             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 775
 776             /* LENNARD-JONES DISPERSION/REPULSION */
 777
 778             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 779             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 780             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 781             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 782             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 783
 784             /* Update potential sum for this i atom from the interaction with this j atom. */
 785             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 786             velecsum         = _mm_add_pd(velecsum,velec);
 787             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 788             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 789
 790             fscal            = _mm_add_pd(felec,fvdw);
 791
 792             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 793
 794             /* Calculate temporary vectorial force */
 795             tx               = _mm_mul_pd(fscal,dx00);
 796             ty               = _mm_mul_pd(fscal,dy00);
 797             tz               = _mm_mul_pd(fscal,dz00);
 798
 799             /* Update vectorial force */
 800             fix0             = _mm_add_pd(fix0,tx);
 801             fiy0             = _mm_add_pd(fiy0,ty);
 802             fiz0             = _mm_add_pd(fiz0,tz);
 803
 804             fjx0             = _mm_add_pd(fjx0,tx);
 805             fjy0             = _mm_add_pd(fjy0,ty);
 806             fjz0             = _mm_add_pd(fjz0,tz);
 807
 808             /**************************
 809              * CALCULATE INTERACTIONS *
 810              **************************/
 811
 812             r01              = _mm_mul_pd(rsq01,rinv01);
 813
 814             /* EWALD ELECTROSTATICS */
 815
 816             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 817             ewrt             = _mm_mul_pd(r01,ewtabscale);
 818             ewitab           = _mm_cvttpd_epi32(ewrt);
 819             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 820             ewitab           = _mm_slli_epi32(ewitab,2);
 821             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 822             ewtabD           = _mm_setzero_pd();
 823             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 824             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 825             ewtabFn          = _mm_setzero_pd();
 826             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 827             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 828             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 829             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 830             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 831
 832             /* Update potential sum for this i atom from the interaction with this j atom. */
 833             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 834             velecsum         = _mm_add_pd(velecsum,velec);
 835
 836             fscal            = felec;
 837
 838             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 839
 840             /* Calculate temporary vectorial force */
 841             tx               = _mm_mul_pd(fscal,dx01);
 842             ty               = _mm_mul_pd(fscal,dy01);
 843             tz               = _mm_mul_pd(fscal,dz01);
 844
 845             /* Update vectorial force */
 846             fix0             = _mm_add_pd(fix0,tx);
 847             fiy0             = _mm_add_pd(fiy0,ty);
 848             fiz0             = _mm_add_pd(fiz0,tz);
 849
 850             fjx1             = _mm_add_pd(fjx1,tx);
 851             fjy1             = _mm_add_pd(fjy1,ty);
 852             fjz1             = _mm_add_pd(fjz1,tz);
 853
 854             /**************************
 855              * CALCULATE INTERACTIONS *
 856              **************************/
 857
 858             r02              = _mm_mul_pd(rsq02,rinv02);
 859
 860             /* EWALD ELECTROSTATICS */
 861
 862             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 863             ewrt             = _mm_mul_pd(r02,ewtabscale);
 864             ewitab           = _mm_cvttpd_epi32(ewrt);
 865             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 866             ewitab           = _mm_slli_epi32(ewitab,2);
 867             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 868             ewtabD           = _mm_setzero_pd();
 869             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 870             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 871             ewtabFn          = _mm_setzero_pd();
 872             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 873             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 874             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 875             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 876             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 877
 878             /* Update potential sum for this i atom from the interaction with this j atom. */
 879             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 880             velecsum         = _mm_add_pd(velecsum,velec);
 881
 882             fscal            = felec;
 883
 884             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 885
 886             /* Calculate temporary vectorial force */
 887             tx               = _mm_mul_pd(fscal,dx02);
 888             ty               = _mm_mul_pd(fscal,dy02);
 889             tz               = _mm_mul_pd(fscal,dz02);
 890
 891             /* Update vectorial force */
 892             fix0             = _mm_add_pd(fix0,tx);
 893             fiy0             = _mm_add_pd(fiy0,ty);
 894             fiz0             = _mm_add_pd(fiz0,tz);
 895
 896             fjx2             = _mm_add_pd(fjx2,tx);
 897             fjy2             = _mm_add_pd(fjy2,ty);
 898             fjz2             = _mm_add_pd(fjz2,tz);
 899
 900             /**************************
 901              * CALCULATE INTERACTIONS *
 902              **************************/
 903
 904             r10              = _mm_mul_pd(rsq10,rinv10);
 905
 906             /* EWALD ELECTROSTATICS */
 907
 908             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 909             ewrt             = _mm_mul_pd(r10,ewtabscale);
 910             ewitab           = _mm_cvttpd_epi32(ewrt);
 911             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 912             ewitab           = _mm_slli_epi32(ewitab,2);
 913             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 914             ewtabD           = _mm_setzero_pd();
 915             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 916             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 917             ewtabFn          = _mm_setzero_pd();
 918             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 919             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 920             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 921             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 922             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 923
 924             /* Update potential sum for this i atom from the interaction with this j atom. */
 925             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 926             velecsum         = _mm_add_pd(velecsum,velec);
 927
 928             fscal            = felec;
 929
 930             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 931
 932             /* Calculate temporary vectorial force */
 933             tx               = _mm_mul_pd(fscal,dx10);
 934             ty               = _mm_mul_pd(fscal,dy10);
 935             tz               = _mm_mul_pd(fscal,dz10);
 936
 937             /* Update vectorial force */
 938             fix1             = _mm_add_pd(fix1,tx);
 939             fiy1             = _mm_add_pd(fiy1,ty);
 940             fiz1             = _mm_add_pd(fiz1,tz);
 941
 942             fjx0             = _mm_add_pd(fjx0,tx);
 943             fjy0             = _mm_add_pd(fjy0,ty);
 944             fjz0             = _mm_add_pd(fjz0,tz);
 945
 946             /**************************
 947              * CALCULATE INTERACTIONS *
 948              **************************/
 949
 950             r11              = _mm_mul_pd(rsq11,rinv11);
 951
 952             /* EWALD ELECTROSTATICS */
 953
 954             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 955             ewrt             = _mm_mul_pd(r11,ewtabscale);
 956             ewitab           = _mm_cvttpd_epi32(ewrt);
 957             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 958             ewitab           = _mm_slli_epi32(ewitab,2);
 959             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 960             ewtabD           = _mm_setzero_pd();
 961             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 962             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 963             ewtabFn          = _mm_setzero_pd();
 964             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 965             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 966             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 967             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 968             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 969
 970             /* Update potential sum for this i atom from the interaction with this j atom. */
 971             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 972             velecsum         = _mm_add_pd(velecsum,velec);
 973
 974             fscal            = felec;
 975
 976             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 977
 978             /* Calculate temporary vectorial force */
 979             tx               = _mm_mul_pd(fscal,dx11);
 980             ty               = _mm_mul_pd(fscal,dy11);
 981             tz               = _mm_mul_pd(fscal,dz11);
 982
 983             /* Update vectorial force */
 984             fix1             = _mm_add_pd(fix1,tx);
 985             fiy1             = _mm_add_pd(fiy1,ty);
 986             fiz1             = _mm_add_pd(fiz1,tz);
 987
 988             fjx1             = _mm_add_pd(fjx1,tx);
 989             fjy1             = _mm_add_pd(fjy1,ty);
 990             fjz1             = _mm_add_pd(fjz1,tz);
 991
 992             /**************************
 993              * CALCULATE INTERACTIONS *
 994              **************************/
 995
 996             r12              = _mm_mul_pd(rsq12,rinv12);
 997
 998             /* EWALD ELECTROSTATICS */
 999
1000             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1001             ewrt             = _mm_mul_pd(r12,ewtabscale);
1002             ewitab           = _mm_cvttpd_epi32(ewrt);
1003             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1004             ewitab           = _mm_slli_epi32(ewitab,2);
1005             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1006             ewtabD           = _mm_setzero_pd();
1007             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1008             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1009             ewtabFn          = _mm_setzero_pd();
1010             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1011             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1012             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1013             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
1014             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1015
1016             /* Update potential sum for this i atom from the interaction with this j atom. */
1017             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1018             velecsum         = _mm_add_pd(velecsum,velec);
1019
1020             fscal            = felec;
1021
1022             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1023
1024             /* Calculate temporary vectorial force */
1025             tx               = _mm_mul_pd(fscal,dx12);
1026             ty               = _mm_mul_pd(fscal,dy12);
1027             tz               = _mm_mul_pd(fscal,dz12);
1028
1029             /* Update vectorial force */
1030             fix1             = _mm_add_pd(fix1,tx);
1031             fiy1             = _mm_add_pd(fiy1,ty);
1032             fiz1             = _mm_add_pd(fiz1,tz);
1033
1034             fjx2             = _mm_add_pd(fjx2,tx);
1035             fjy2             = _mm_add_pd(fjy2,ty);
1036             fjz2             = _mm_add_pd(fjz2,tz);
1037
1038             /**************************
1039              * CALCULATE INTERACTIONS *
1040              **************************/
1041
1042             r20              = _mm_mul_pd(rsq20,rinv20);
1043
1044             /* EWALD ELECTROSTATICS */
1045
1046             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1047             ewrt             = _mm_mul_pd(r20,ewtabscale);
1048             ewitab           = _mm_cvttpd_epi32(ewrt);
1049             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1050             ewitab           = _mm_slli_epi32(ewitab,2);
1051             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1052             ewtabD           = _mm_setzero_pd();
1053             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1054             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1055             ewtabFn          = _mm_setzero_pd();
1056             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1057             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1058             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1059             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1060             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1061
1062             /* Update potential sum for this i atom from the interaction with this j atom. */
1063             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1064             velecsum         = _mm_add_pd(velecsum,velec);
1065
1066             fscal            = felec;
1067
1068             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1069
1070             /* Calculate temporary vectorial force */
1071             tx               = _mm_mul_pd(fscal,dx20);
1072             ty               = _mm_mul_pd(fscal,dy20);
1073             tz               = _mm_mul_pd(fscal,dz20);
1074
1075             /* Update vectorial force */
1076             fix2             = _mm_add_pd(fix2,tx);
1077             fiy2             = _mm_add_pd(fiy2,ty);
1078             fiz2             = _mm_add_pd(fiz2,tz);
1079
1080             fjx0             = _mm_add_pd(fjx0,tx);
1081             fjy0             = _mm_add_pd(fjy0,ty);
1082             fjz0             = _mm_add_pd(fjz0,tz);
1083
1084             /**************************
1085              * CALCULATE INTERACTIONS *
1086              **************************/
1087
1088             r21              = _mm_mul_pd(rsq21,rinv21);
1089
1090             /* EWALD ELECTROSTATICS */
1091
1092             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1093             ewrt             = _mm_mul_pd(r21,ewtabscale);
1094             ewitab           = _mm_cvttpd_epi32(ewrt);
1095             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1096             ewitab           = _mm_slli_epi32(ewitab,2);
1097             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1098             ewtabD           = _mm_setzero_pd();
1099             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1100             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1101             ewtabFn          = _mm_setzero_pd();
1102             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1103             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1104             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1105             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1106             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1107
1108             /* Update potential sum for this i atom from the interaction with this j atom. */
1109             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1110             velecsum         = _mm_add_pd(velecsum,velec);
1111
1112             fscal            = felec;
1113
1114             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1115
1116             /* Calculate temporary vectorial force */
1117             tx               = _mm_mul_pd(fscal,dx21);
1118             ty               = _mm_mul_pd(fscal,dy21);
1119             tz               = _mm_mul_pd(fscal,dz21);
1120
1121             /* Update vectorial force */
1122             fix2             = _mm_add_pd(fix2,tx);
1123             fiy2             = _mm_add_pd(fiy2,ty);
1124             fiz2             = _mm_add_pd(fiz2,tz);
1125
1126             fjx1             = _mm_add_pd(fjx1,tx);
1127             fjy1             = _mm_add_pd(fjy1,ty);
1128             fjz1             = _mm_add_pd(fjz1,tz);
1129
1130             /**************************
1131              * CALCULATE INTERACTIONS *
1132              **************************/
1133
1134             r22              = _mm_mul_pd(rsq22,rinv22);
1135
1136             /* EWALD ELECTROSTATICS */
1137
1138             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1139             ewrt             = _mm_mul_pd(r22,ewtabscale);
1140             ewitab           = _mm_cvttpd_epi32(ewrt);
1141             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1142             ewitab           = _mm_slli_epi32(ewitab,2);
1143             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1144             ewtabD           = _mm_setzero_pd();
1145             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1146             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1147             ewtabFn          = _mm_setzero_pd();
1148             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1149             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1150             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1151             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1152             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1153
1154             /* Update potential sum for this i atom from the interaction with this j atom. */
1155             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1156             velecsum         = _mm_add_pd(velecsum,velec);
1157
1158             fscal            = felec;
1159
1160             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1161
1162             /* Calculate temporary vectorial force */
1163             tx               = _mm_mul_pd(fscal,dx22);
1164             ty               = _mm_mul_pd(fscal,dy22);
1165             tz               = _mm_mul_pd(fscal,dz22);
1166
1167             /* Update vectorial force */
1168             fix2             = _mm_add_pd(fix2,tx);
1169             fiy2             = _mm_add_pd(fiy2,ty);
1170             fiz2             = _mm_add_pd(fiz2,tz);
1171
1172             fjx2             = _mm_add_pd(fjx2,tx);
1173             fjy2             = _mm_add_pd(fjy2,ty);
1174             fjz2             = _mm_add_pd(fjz2,tz);
1175
1176             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1177
1178             /* Inner loop uses 381 flops */
1179         }
1180
1181         /* End of innermost loop */
1182
1183         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1184                                               f+i_coord_offset,fshift+i_shift_offset);
1185
1186         ggid                        = gid[iidx];
1187         /* Update potential energies */
1188         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1189         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1190
1191         /* Increment number of inner iterations */
1192         inneriter                  += j_index_end - j_index_start;
1193
1194         /* Outer loop uses 20 flops */
1195     }
1196
1197     /* Increment number of outer iterations */
1198     outeriter        += nri;
1199
1200     /* Update outer/inner flops */
1201
1202     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*381);
1203 }
1204 /*
1205  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW3W3_F_sse4_1_double
1206  * Electrostatics interaction: Ewald
1207  * VdW interaction:            LennardJones
1208  * Geometry:                   Water3-Water3
1209  * Calculate force/pot:        Force
1210  */
1211 void
1212 nb_kernel_ElecEw_VdwLJ_GeomW3W3_F_sse4_1_double
1213                     (t_nblist * gmx_restrict                nlist,
1214                      rvec * gmx_restrict                    xx,
1215                      rvec * gmx_restrict                    ff,
1216                      t_forcerec * gmx_restrict              fr,
1217                      t_mdatoms * gmx_restrict               mdatoms,
1218                      nb_kernel_data_t * gmx_restrict        kernel_data,
1219                      t_nrnb * gmx_restrict                  nrnb)
1220 {
1221     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1222      * just 0 for non-waters.
1223      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1224      * jnr indices corresponding to data put in the four positions in the SIMD register.
1225      */
1226     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1227     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1228     int              jnrA,jnrB;
1229     int              j_coord_offsetA,j_coord_offsetB;
1230     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1231     real             rcutoff_scalar;
1232     real             *shiftvec,*fshift,*x,*f;
1233     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1234     int              vdwioffset0;
1235     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1236     int              vdwioffset1;
1237     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1238     int              vdwioffset2;
1239     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1240     int              vdwjidx0A,vdwjidx0B;
1241     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1242     int              vdwjidx1A,vdwjidx1B;
1243     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1244     int              vdwjidx2A,vdwjidx2B;
1245     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1246     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1247     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1248     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1249     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1250     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1251     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1252     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1253     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1254     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1255     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1256     real             *charge;
1257     int              nvdwtype;
1258     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1259     int              *vdwtype;
1260     real             *vdwparam;
1261     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1262     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1263     __m128i          ewitab;
1264     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1265     real             *ewtab;
1266     __m128d          dummy_mask,cutoff_mask;
1267     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1268     __m128d          one     = _mm_set1_pd(1.0);
1269     __m128d          two     = _mm_set1_pd(2.0);
1270     x                = xx[0];
1271     f                = ff[0];
1272
1273     nri              = nlist->nri;
1274     iinr             = nlist->iinr;
1275     jindex           = nlist->jindex;
1276     jjnr             = nlist->jjnr;
1277     shiftidx         = nlist->shift;
1278     gid              = nlist->gid;
1279     shiftvec         = fr->shift_vec[0];
1280     fshift           = fr->fshift[0];
1281     facel            = _mm_set1_pd(fr->epsfac);
1282     charge           = mdatoms->chargeA;
1283     nvdwtype         = fr->ntype;
1284     vdwparam         = fr->nbfp;
1285     vdwtype          = mdatoms->typeA;
1286
1287     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1288     ewtab            = fr->ic->tabq_coul_F;
1289     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1290     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1291
1292     /* Setup water-specific parameters */
1293     inr              = nlist->iinr[0];
1294     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1295     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1296     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1297     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1298
1299     jq0              = _mm_set1_pd(charge[inr+0]);
1300     jq1              = _mm_set1_pd(charge[inr+1]);
1301     jq2              = _mm_set1_pd(charge[inr+2]);
1302     vdwjidx0A        = 2*vdwtype[inr+0];
1303     qq00             = _mm_mul_pd(iq0,jq0);
1304     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1305     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1306     qq01             = _mm_mul_pd(iq0,jq1);
1307     qq02             = _mm_mul_pd(iq0,jq2);
1308     qq10             = _mm_mul_pd(iq1,jq0);
1309     qq11             = _mm_mul_pd(iq1,jq1);
1310     qq12             = _mm_mul_pd(iq1,jq2);
1311     qq20             = _mm_mul_pd(iq2,jq0);
1312     qq21             = _mm_mul_pd(iq2,jq1);
1313     qq22             = _mm_mul_pd(iq2,jq2);
1314
1315     /* Avoid stupid compiler warnings */
1316     jnrA = jnrB = 0;
1317     j_coord_offsetA = 0;
1318     j_coord_offsetB = 0;
1319
1320     outeriter        = 0;
1321     inneriter        = 0;
1322
1323     /* Start outer loop over neighborlists */
1324     for(iidx=0; iidx<nri; iidx++)
1325     {
1326         /* Load shift vector for this list */
1327         i_shift_offset   = DIM*shiftidx[iidx];
1328
1329         /* Load limits for loop over neighbors */
1330         j_index_start    = jindex[iidx];
1331         j_index_end      = jindex[iidx+1];
1332
1333         /* Get outer coordinate index */
1334         inr              = iinr[iidx];
1335         i_coord_offset   = DIM*inr;
1336
1337         /* Load i particle coords and add shift vector */
1338         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1339                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1340
1341         fix0             = _mm_setzero_pd();
1342         fiy0             = _mm_setzero_pd();
1343         fiz0             = _mm_setzero_pd();
1344         fix1             = _mm_setzero_pd();
1345         fiy1             = _mm_setzero_pd();
1346         fiz1             = _mm_setzero_pd();
1347         fix2             = _mm_setzero_pd();
1348         fiy2             = _mm_setzero_pd();
1349         fiz2             = _mm_setzero_pd();
1350
1351         /* Start inner kernel loop */
1352         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1353         {
1354
1355             /* Get j neighbor index, and coordinate index */
1356             jnrA             = jjnr[jidx];
1357             jnrB             = jjnr[jidx+1];
1358             j_coord_offsetA  = DIM*jnrA;
1359             j_coord_offsetB  = DIM*jnrB;
1360
1361             /* load j atom coordinates */
1362             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1363                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1364
1365             /* Calculate displacement vector */
1366             dx00             = _mm_sub_pd(ix0,jx0);
1367             dy00             = _mm_sub_pd(iy0,jy0);
1368             dz00             = _mm_sub_pd(iz0,jz0);
1369             dx01             = _mm_sub_pd(ix0,jx1);
1370             dy01             = _mm_sub_pd(iy0,jy1);
1371             dz01             = _mm_sub_pd(iz0,jz1);
1372             dx02             = _mm_sub_pd(ix0,jx2);
1373             dy02             = _mm_sub_pd(iy0,jy2);
1374             dz02             = _mm_sub_pd(iz0,jz2);
1375             dx10             = _mm_sub_pd(ix1,jx0);
1376             dy10             = _mm_sub_pd(iy1,jy0);
1377             dz10             = _mm_sub_pd(iz1,jz0);
1378             dx11             = _mm_sub_pd(ix1,jx1);
1379             dy11             = _mm_sub_pd(iy1,jy1);
1380             dz11             = _mm_sub_pd(iz1,jz1);
1381             dx12             = _mm_sub_pd(ix1,jx2);
1382             dy12             = _mm_sub_pd(iy1,jy2);
1383             dz12             = _mm_sub_pd(iz1,jz2);
1384             dx20             = _mm_sub_pd(ix2,jx0);
1385             dy20             = _mm_sub_pd(iy2,jy0);
1386             dz20             = _mm_sub_pd(iz2,jz0);
1387             dx21             = _mm_sub_pd(ix2,jx1);
1388             dy21             = _mm_sub_pd(iy2,jy1);
1389             dz21             = _mm_sub_pd(iz2,jz1);
1390             dx22             = _mm_sub_pd(ix2,jx2);
1391             dy22             = _mm_sub_pd(iy2,jy2);
1392             dz22             = _mm_sub_pd(iz2,jz2);
1393
1394             /* Calculate squared distance and things based on it */
1395             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1396             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1397             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1398             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1399             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1400             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1401             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1402             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1403             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1404
1405             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1406             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1407             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1408             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1409             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1410             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1411             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1412             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1413             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1414
1415             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1416             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1417             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1418             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1419             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1420             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1421             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1422             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1423             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1424
1425             fjx0             = _mm_setzero_pd();
1426             fjy0             = _mm_setzero_pd();
1427             fjz0             = _mm_setzero_pd();
1428             fjx1             = _mm_setzero_pd();
1429             fjy1             = _mm_setzero_pd();
1430             fjz1             = _mm_setzero_pd();
1431             fjx2             = _mm_setzero_pd();
1432             fjy2             = _mm_setzero_pd();
1433             fjz2             = _mm_setzero_pd();
1434
1435             /**************************
1436              * CALCULATE INTERACTIONS *
1437              **************************/
1438
1439             r00              = _mm_mul_pd(rsq00,rinv00);
1440
1441             /* EWALD ELECTROSTATICS */
1442
1443             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1444             ewrt             = _mm_mul_pd(r00,ewtabscale);
1445             ewitab           = _mm_cvttpd_epi32(ewrt);
1446             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1447             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1448                                          &ewtabF,&ewtabFn);
1449             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1450             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1451
1452             /* LENNARD-JONES DISPERSION/REPULSION */
1453
1454             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1455             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1456
1457             fscal            = _mm_add_pd(felec,fvdw);
1458
1459             /* Calculate temporary vectorial force */
1460             tx               = _mm_mul_pd(fscal,dx00);
1461             ty               = _mm_mul_pd(fscal,dy00);
1462             tz               = _mm_mul_pd(fscal,dz00);
1463
1464             /* Update vectorial force */
1465             fix0             = _mm_add_pd(fix0,tx);
1466             fiy0             = _mm_add_pd(fiy0,ty);
1467             fiz0             = _mm_add_pd(fiz0,tz);
1468
1469             fjx0             = _mm_add_pd(fjx0,tx);
1470             fjy0             = _mm_add_pd(fjy0,ty);
1471             fjz0             = _mm_add_pd(fjz0,tz);
1472
1473             /**************************
1474              * CALCULATE INTERACTIONS *
1475              **************************/
1476
1477             r01              = _mm_mul_pd(rsq01,rinv01);
1478
1479             /* EWALD ELECTROSTATICS */
1480
1481             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1482             ewrt             = _mm_mul_pd(r01,ewtabscale);
1483             ewitab           = _mm_cvttpd_epi32(ewrt);
1484             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1485             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1486                                          &ewtabF,&ewtabFn);
1487             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1488             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1489
1490             fscal            = felec;
1491
1492             /* Calculate temporary vectorial force */
1493             tx               = _mm_mul_pd(fscal,dx01);
1494             ty               = _mm_mul_pd(fscal,dy01);
1495             tz               = _mm_mul_pd(fscal,dz01);
1496
1497             /* Update vectorial force */
1498             fix0             = _mm_add_pd(fix0,tx);
1499             fiy0             = _mm_add_pd(fiy0,ty);
1500             fiz0             = _mm_add_pd(fiz0,tz);
1501
1502             fjx1             = _mm_add_pd(fjx1,tx);
1503             fjy1             = _mm_add_pd(fjy1,ty);
1504             fjz1             = _mm_add_pd(fjz1,tz);
1505
1506             /**************************
1507              * CALCULATE INTERACTIONS *
1508              **************************/
1509
1510             r02              = _mm_mul_pd(rsq02,rinv02);
1511
1512             /* EWALD ELECTROSTATICS */
1513
1514             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1515             ewrt             = _mm_mul_pd(r02,ewtabscale);
1516             ewitab           = _mm_cvttpd_epi32(ewrt);
1517             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1518             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1519                                          &ewtabF,&ewtabFn);
1520             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1521             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1522
1523             fscal            = felec;
1524
1525             /* Calculate temporary vectorial force */
1526             tx               = _mm_mul_pd(fscal,dx02);
1527             ty               = _mm_mul_pd(fscal,dy02);
1528             tz               = _mm_mul_pd(fscal,dz02);
1529
1530             /* Update vectorial force */
1531             fix0             = _mm_add_pd(fix0,tx);
1532             fiy0             = _mm_add_pd(fiy0,ty);
1533             fiz0             = _mm_add_pd(fiz0,tz);
1534
1535             fjx2             = _mm_add_pd(fjx2,tx);
1536             fjy2             = _mm_add_pd(fjy2,ty);
1537             fjz2             = _mm_add_pd(fjz2,tz);
1538
1539             /**************************
1540              * CALCULATE INTERACTIONS *
1541              **************************/
1542
1543             r10              = _mm_mul_pd(rsq10,rinv10);
1544
1545             /* EWALD ELECTROSTATICS */
1546
1547             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1548             ewrt             = _mm_mul_pd(r10,ewtabscale);
1549             ewitab           = _mm_cvttpd_epi32(ewrt);
1550             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1551             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1552                                          &ewtabF,&ewtabFn);
1553             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1554             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1555
1556             fscal            = felec;
1557
1558             /* Calculate temporary vectorial force */
1559             tx               = _mm_mul_pd(fscal,dx10);
1560             ty               = _mm_mul_pd(fscal,dy10);
1561             tz               = _mm_mul_pd(fscal,dz10);
1562
1563             /* Update vectorial force */
1564             fix1             = _mm_add_pd(fix1,tx);
1565             fiy1             = _mm_add_pd(fiy1,ty);
1566             fiz1             = _mm_add_pd(fiz1,tz);
1567
1568             fjx0             = _mm_add_pd(fjx0,tx);
1569             fjy0             = _mm_add_pd(fjy0,ty);
1570             fjz0             = _mm_add_pd(fjz0,tz);
1571
1572             /**************************
1573              * CALCULATE INTERACTIONS *
1574              **************************/
1575
1576             r11              = _mm_mul_pd(rsq11,rinv11);
1577
1578             /* EWALD ELECTROSTATICS */
1579
1580             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1581             ewrt             = _mm_mul_pd(r11,ewtabscale);
1582             ewitab           = _mm_cvttpd_epi32(ewrt);
1583             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1584             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1585                                          &ewtabF,&ewtabFn);
1586             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1587             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1588
1589             fscal            = felec;
1590
1591             /* Calculate temporary vectorial force */
1592             tx               = _mm_mul_pd(fscal,dx11);
1593             ty               = _mm_mul_pd(fscal,dy11);
1594             tz               = _mm_mul_pd(fscal,dz11);
1595
1596             /* Update vectorial force */
1597             fix1             = _mm_add_pd(fix1,tx);
1598             fiy1             = _mm_add_pd(fiy1,ty);
1599             fiz1             = _mm_add_pd(fiz1,tz);
1600
1601             fjx1             = _mm_add_pd(fjx1,tx);
1602             fjy1             = _mm_add_pd(fjy1,ty);
1603             fjz1             = _mm_add_pd(fjz1,tz);
1604
1605             /**************************
1606              * CALCULATE INTERACTIONS *
1607              **************************/
1608
1609             r12              = _mm_mul_pd(rsq12,rinv12);
1610
1611             /* EWALD ELECTROSTATICS */
1612
1613             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1614             ewrt             = _mm_mul_pd(r12,ewtabscale);
1615             ewitab           = _mm_cvttpd_epi32(ewrt);
1616             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1617             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1618                                          &ewtabF,&ewtabFn);
1619             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1620             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1621
1622             fscal            = felec;
1623
1624             /* Calculate temporary vectorial force */
1625             tx               = _mm_mul_pd(fscal,dx12);
1626             ty               = _mm_mul_pd(fscal,dy12);
1627             tz               = _mm_mul_pd(fscal,dz12);
1628
1629             /* Update vectorial force */
1630             fix1             = _mm_add_pd(fix1,tx);
1631             fiy1             = _mm_add_pd(fiy1,ty);
1632             fiz1             = _mm_add_pd(fiz1,tz);
1633
1634             fjx2             = _mm_add_pd(fjx2,tx);
1635             fjy2             = _mm_add_pd(fjy2,ty);
1636             fjz2             = _mm_add_pd(fjz2,tz);
1637
1638             /**************************
1639              * CALCULATE INTERACTIONS *
1640              **************************/
1641
1642             r20              = _mm_mul_pd(rsq20,rinv20);
1643
1644             /* EWALD ELECTROSTATICS */
1645
1646             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1647             ewrt             = _mm_mul_pd(r20,ewtabscale);
1648             ewitab           = _mm_cvttpd_epi32(ewrt);
1649             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1650             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1651                                          &ewtabF,&ewtabFn);
1652             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1653             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1654
1655             fscal            = felec;
1656
1657             /* Calculate temporary vectorial force */
1658             tx               = _mm_mul_pd(fscal,dx20);
1659             ty               = _mm_mul_pd(fscal,dy20);
1660             tz               = _mm_mul_pd(fscal,dz20);
1661
1662             /* Update vectorial force */
1663             fix2             = _mm_add_pd(fix2,tx);
1664             fiy2             = _mm_add_pd(fiy2,ty);
1665             fiz2             = _mm_add_pd(fiz2,tz);
1666
1667             fjx0             = _mm_add_pd(fjx0,tx);
1668             fjy0             = _mm_add_pd(fjy0,ty);
1669             fjz0             = _mm_add_pd(fjz0,tz);
1670
1671             /**************************
1672              * CALCULATE INTERACTIONS *
1673              **************************/
1674
1675             r21              = _mm_mul_pd(rsq21,rinv21);
1676
1677             /* EWALD ELECTROSTATICS */
1678
1679             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1680             ewrt             = _mm_mul_pd(r21,ewtabscale);
1681             ewitab           = _mm_cvttpd_epi32(ewrt);
1682             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1683             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1684                                          &ewtabF,&ewtabFn);
1685             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1686             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1687
1688             fscal            = felec;
1689
1690             /* Calculate temporary vectorial force */
1691             tx               = _mm_mul_pd(fscal,dx21);
1692             ty               = _mm_mul_pd(fscal,dy21);
1693             tz               = _mm_mul_pd(fscal,dz21);
1694
1695             /* Update vectorial force */
1696             fix2             = _mm_add_pd(fix2,tx);
1697             fiy2             = _mm_add_pd(fiy2,ty);
1698             fiz2             = _mm_add_pd(fiz2,tz);
1699
1700             fjx1             = _mm_add_pd(fjx1,tx);
1701             fjy1             = _mm_add_pd(fjy1,ty);
1702             fjz1             = _mm_add_pd(fjz1,tz);
1703
1704             /**************************
1705              * CALCULATE INTERACTIONS *
1706              **************************/
1707
1708             r22              = _mm_mul_pd(rsq22,rinv22);
1709
1710             /* EWALD ELECTROSTATICS */
1711
1712             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1713             ewrt             = _mm_mul_pd(r22,ewtabscale);
1714             ewitab           = _mm_cvttpd_epi32(ewrt);
1715             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1716             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1717                                          &ewtabF,&ewtabFn);
1718             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1719             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1720
1721             fscal            = felec;
1722
1723             /* Calculate temporary vectorial force */
1724             tx               = _mm_mul_pd(fscal,dx22);
1725             ty               = _mm_mul_pd(fscal,dy22);
1726             tz               = _mm_mul_pd(fscal,dz22);
1727
1728             /* Update vectorial force */
1729             fix2             = _mm_add_pd(fix2,tx);
1730             fiy2             = _mm_add_pd(fiy2,ty);
1731             fiz2             = _mm_add_pd(fiz2,tz);
1732
1733             fjx2             = _mm_add_pd(fjx2,tx);
1734             fjy2             = _mm_add_pd(fjy2,ty);
1735             fjz2             = _mm_add_pd(fjz2,tz);
1736
1737             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1738
1739             /* Inner loop uses 331 flops */
1740         }
1741
1742         if(jidx<j_index_end)
1743         {
1744
1745             jnrA             = jjnr[jidx];
1746             j_coord_offsetA  = DIM*jnrA;
1747
1748             /* load j atom coordinates */
1749             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1750                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1751
1752             /* Calculate displacement vector */
1753             dx00             = _mm_sub_pd(ix0,jx0);
1754             dy00             = _mm_sub_pd(iy0,jy0);
1755             dz00             = _mm_sub_pd(iz0,jz0);
1756             dx01             = _mm_sub_pd(ix0,jx1);
1757             dy01             = _mm_sub_pd(iy0,jy1);
1758             dz01             = _mm_sub_pd(iz0,jz1);
1759             dx02             = _mm_sub_pd(ix0,jx2);
1760             dy02             = _mm_sub_pd(iy0,jy2);
1761             dz02             = _mm_sub_pd(iz0,jz2);
1762             dx10             = _mm_sub_pd(ix1,jx0);
1763             dy10             = _mm_sub_pd(iy1,jy0);
1764             dz10             = _mm_sub_pd(iz1,jz0);
1765             dx11             = _mm_sub_pd(ix1,jx1);
1766             dy11             = _mm_sub_pd(iy1,jy1);
1767             dz11             = _mm_sub_pd(iz1,jz1);
1768             dx12             = _mm_sub_pd(ix1,jx2);
1769             dy12             = _mm_sub_pd(iy1,jy2);
1770             dz12             = _mm_sub_pd(iz1,jz2);
1771             dx20             = _mm_sub_pd(ix2,jx0);
1772             dy20             = _mm_sub_pd(iy2,jy0);
1773             dz20             = _mm_sub_pd(iz2,jz0);
1774             dx21             = _mm_sub_pd(ix2,jx1);
1775             dy21             = _mm_sub_pd(iy2,jy1);
1776             dz21             = _mm_sub_pd(iz2,jz1);
1777             dx22             = _mm_sub_pd(ix2,jx2);
1778             dy22             = _mm_sub_pd(iy2,jy2);
1779             dz22             = _mm_sub_pd(iz2,jz2);
1780
1781             /* Calculate squared distance and things based on it */
1782             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1783             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1784             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1785             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1786             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1787             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1788             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1789             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1790             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1791
1792             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1793             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1794             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1795             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1796             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1797             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1798             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1799             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1800             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1801
1802             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1803             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1804             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1805             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1806             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1807             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1808             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1809             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1810             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1811
1812             fjx0             = _mm_setzero_pd();
1813             fjy0             = _mm_setzero_pd();
1814             fjz0             = _mm_setzero_pd();
1815             fjx1             = _mm_setzero_pd();
1816             fjy1             = _mm_setzero_pd();
1817             fjz1             = _mm_setzero_pd();
1818             fjx2             = _mm_setzero_pd();
1819             fjy2             = _mm_setzero_pd();
1820             fjz2             = _mm_setzero_pd();
1821
1822             /**************************
1823              * CALCULATE INTERACTIONS *
1824              **************************/
1825
1826             r00              = _mm_mul_pd(rsq00,rinv00);
1827
1828             /* EWALD ELECTROSTATICS */
1829
1830             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1831             ewrt             = _mm_mul_pd(r00,ewtabscale);
1832             ewitab           = _mm_cvttpd_epi32(ewrt);
1833             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1834             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1835             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1836             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1837
1838             /* LENNARD-JONES DISPERSION/REPULSION */
1839
1840             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1841             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1842
1843             fscal            = _mm_add_pd(felec,fvdw);
1844
1845             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1846
1847             /* Calculate temporary vectorial force */
1848             tx               = _mm_mul_pd(fscal,dx00);
1849             ty               = _mm_mul_pd(fscal,dy00);
1850             tz               = _mm_mul_pd(fscal,dz00);
1851
1852             /* Update vectorial force */
1853             fix0             = _mm_add_pd(fix0,tx);
1854             fiy0             = _mm_add_pd(fiy0,ty);
1855             fiz0             = _mm_add_pd(fiz0,tz);
1856
1857             fjx0             = _mm_add_pd(fjx0,tx);
1858             fjy0             = _mm_add_pd(fjy0,ty);
1859             fjz0             = _mm_add_pd(fjz0,tz);
1860
1861             /**************************
1862              * CALCULATE INTERACTIONS *
1863              **************************/
1864
1865             r01              = _mm_mul_pd(rsq01,rinv01);
1866
1867             /* EWALD ELECTROSTATICS */
1868
1869             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1870             ewrt             = _mm_mul_pd(r01,ewtabscale);
1871             ewitab           = _mm_cvttpd_epi32(ewrt);
1872             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1873             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1874             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1875             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1876
1877             fscal            = felec;
1878
1879             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1880
1881             /* Calculate temporary vectorial force */
1882             tx               = _mm_mul_pd(fscal,dx01);
1883             ty               = _mm_mul_pd(fscal,dy01);
1884             tz               = _mm_mul_pd(fscal,dz01);
1885
1886             /* Update vectorial force */
1887             fix0             = _mm_add_pd(fix0,tx);
1888             fiy0             = _mm_add_pd(fiy0,ty);
1889             fiz0             = _mm_add_pd(fiz0,tz);
1890
1891             fjx1             = _mm_add_pd(fjx1,tx);
1892             fjy1             = _mm_add_pd(fjy1,ty);
1893             fjz1             = _mm_add_pd(fjz1,tz);
1894
1895             /**************************
1896              * CALCULATE INTERACTIONS *
1897              **************************/
1898
1899             r02              = _mm_mul_pd(rsq02,rinv02);
1900
1901             /* EWALD ELECTROSTATICS */
1902
1903             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1904             ewrt             = _mm_mul_pd(r02,ewtabscale);
1905             ewitab           = _mm_cvttpd_epi32(ewrt);
1906             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1907             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1908             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1909             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1910
1911             fscal            = felec;
1912
1913             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1914
1915             /* Calculate temporary vectorial force */
1916             tx               = _mm_mul_pd(fscal,dx02);
1917             ty               = _mm_mul_pd(fscal,dy02);
1918             tz               = _mm_mul_pd(fscal,dz02);
1919
1920             /* Update vectorial force */
1921             fix0             = _mm_add_pd(fix0,tx);
1922             fiy0             = _mm_add_pd(fiy0,ty);
1923             fiz0             = _mm_add_pd(fiz0,tz);
1924
1925             fjx2             = _mm_add_pd(fjx2,tx);
1926             fjy2             = _mm_add_pd(fjy2,ty);
1927             fjz2             = _mm_add_pd(fjz2,tz);
1928
1929             /**************************
1930              * CALCULATE INTERACTIONS *
1931              **************************/
1932
1933             r10              = _mm_mul_pd(rsq10,rinv10);
1934
1935             /* EWALD ELECTROSTATICS */
1936
1937             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1938             ewrt             = _mm_mul_pd(r10,ewtabscale);
1939             ewitab           = _mm_cvttpd_epi32(ewrt);
1940             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1941             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1942             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1943             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1944
1945             fscal            = felec;
1946
1947             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1948
1949             /* Calculate temporary vectorial force */
1950             tx               = _mm_mul_pd(fscal,dx10);
1951             ty               = _mm_mul_pd(fscal,dy10);
1952             tz               = _mm_mul_pd(fscal,dz10);
1953
1954             /* Update vectorial force */
1955             fix1             = _mm_add_pd(fix1,tx);
1956             fiy1             = _mm_add_pd(fiy1,ty);
1957             fiz1             = _mm_add_pd(fiz1,tz);
1958
1959             fjx0             = _mm_add_pd(fjx0,tx);
1960             fjy0             = _mm_add_pd(fjy0,ty);
1961             fjz0             = _mm_add_pd(fjz0,tz);
1962
1963             /**************************
1964              * CALCULATE INTERACTIONS *
1965              **************************/
1966
1967             r11              = _mm_mul_pd(rsq11,rinv11);
1968
1969             /* EWALD ELECTROSTATICS */
1970
1971             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1972             ewrt             = _mm_mul_pd(r11,ewtabscale);
1973             ewitab           = _mm_cvttpd_epi32(ewrt);
1974             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
1975             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1976             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1977             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1978
1979             fscal            = felec;
1980
1981             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1982
1983             /* Calculate temporary vectorial force */
1984             tx               = _mm_mul_pd(fscal,dx11);
1985             ty               = _mm_mul_pd(fscal,dy11);
1986             tz               = _mm_mul_pd(fscal,dz11);
1987
1988             /* Update vectorial force */
1989             fix1             = _mm_add_pd(fix1,tx);
1990             fiy1             = _mm_add_pd(fiy1,ty);
1991             fiz1             = _mm_add_pd(fiz1,tz);
1992
1993             fjx1             = _mm_add_pd(fjx1,tx);
1994             fjy1             = _mm_add_pd(fjy1,ty);
1995             fjz1             = _mm_add_pd(fjz1,tz);
1996
1997             /**************************
1998              * CALCULATE INTERACTIONS *
1999              **************************/
2000
2001             r12              = _mm_mul_pd(rsq12,rinv12);
2002
2003             /* EWALD ELECTROSTATICS */
2004
2005             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2006             ewrt             = _mm_mul_pd(r12,ewtabscale);
2007             ewitab           = _mm_cvttpd_epi32(ewrt);
2008             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2009             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2010             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2011             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2012
2013             fscal            = felec;
2014
2015             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2016
2017             /* Calculate temporary vectorial force */
2018             tx               = _mm_mul_pd(fscal,dx12);
2019             ty               = _mm_mul_pd(fscal,dy12);
2020             tz               = _mm_mul_pd(fscal,dz12);
2021
2022             /* Update vectorial force */
2023             fix1             = _mm_add_pd(fix1,tx);
2024             fiy1             = _mm_add_pd(fiy1,ty);
2025             fiz1             = _mm_add_pd(fiz1,tz);
2026
2027             fjx2             = _mm_add_pd(fjx2,tx);
2028             fjy2             = _mm_add_pd(fjy2,ty);
2029             fjz2             = _mm_add_pd(fjz2,tz);
2030
2031             /**************************
2032              * CALCULATE INTERACTIONS *
2033              **************************/
2034
2035             r20              = _mm_mul_pd(rsq20,rinv20);
2036
2037             /* EWALD ELECTROSTATICS */
2038
2039             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2040             ewrt             = _mm_mul_pd(r20,ewtabscale);
2041             ewitab           = _mm_cvttpd_epi32(ewrt);
2042             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2043             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2044             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2045             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2046
2047             fscal            = felec;
2048
2049             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2050
2051             /* Calculate temporary vectorial force */
2052             tx               = _mm_mul_pd(fscal,dx20);
2053             ty               = _mm_mul_pd(fscal,dy20);
2054             tz               = _mm_mul_pd(fscal,dz20);
2055
2056             /* Update vectorial force */
2057             fix2             = _mm_add_pd(fix2,tx);
2058             fiy2             = _mm_add_pd(fiy2,ty);
2059             fiz2             = _mm_add_pd(fiz2,tz);
2060
2061             fjx0             = _mm_add_pd(fjx0,tx);
2062             fjy0             = _mm_add_pd(fjy0,ty);
2063             fjz0             = _mm_add_pd(fjz0,tz);
2064
2065             /**************************
2066              * CALCULATE INTERACTIONS *
2067              **************************/
2068
2069             r21              = _mm_mul_pd(rsq21,rinv21);
2070
2071             /* EWALD ELECTROSTATICS */
2072
2073             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2074             ewrt             = _mm_mul_pd(r21,ewtabscale);
2075             ewitab           = _mm_cvttpd_epi32(ewrt);
2076             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2077             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2078             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2079             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2080
2081             fscal            = felec;
2082
2083             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2084
2085             /* Calculate temporary vectorial force */
2086             tx               = _mm_mul_pd(fscal,dx21);
2087             ty               = _mm_mul_pd(fscal,dy21);
2088             tz               = _mm_mul_pd(fscal,dz21);
2089
2090             /* Update vectorial force */
2091             fix2             = _mm_add_pd(fix2,tx);
2092             fiy2             = _mm_add_pd(fiy2,ty);
2093             fiz2             = _mm_add_pd(fiz2,tz);
2094
2095             fjx1             = _mm_add_pd(fjx1,tx);
2096             fjy1             = _mm_add_pd(fjy1,ty);
2097             fjz1             = _mm_add_pd(fjz1,tz);
2098
2099             /**************************
2100              * CALCULATE INTERACTIONS *
2101              **************************/
2102
2103             r22              = _mm_mul_pd(rsq22,rinv22);
2104
2105             /* EWALD ELECTROSTATICS */
2106
2107             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2108             ewrt             = _mm_mul_pd(r22,ewtabscale);
2109             ewitab           = _mm_cvttpd_epi32(ewrt);
2110             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
2111             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2112             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2113             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2114
2115             fscal            = felec;
2116
2117             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2118
2119             /* Calculate temporary vectorial force */
2120             tx               = _mm_mul_pd(fscal,dx22);
2121             ty               = _mm_mul_pd(fscal,dy22);
2122             tz               = _mm_mul_pd(fscal,dz22);
2123
2124             /* Update vectorial force */
2125             fix2             = _mm_add_pd(fix2,tx);
2126             fiy2             = _mm_add_pd(fiy2,ty);
2127             fiz2             = _mm_add_pd(fiz2,tz);
2128
2129             fjx2             = _mm_add_pd(fjx2,tx);
2130             fjy2             = _mm_add_pd(fjy2,ty);
2131             fjz2             = _mm_add_pd(fjz2,tz);
2132
2133             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2134
2135             /* Inner loop uses 331 flops */
2136         }
2137
2138         /* End of innermost loop */
2139
2140         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2141                                               f+i_coord_offset,fshift+i_shift_offset);
2142
2143         /* Increment number of inner iterations */
2144         inneriter                  += j_index_end - j_index_start;
2145
2146         /* Outer loop uses 18 flops */
2147     }
2148
2149     /* Increment number of outer iterations */
2150     outeriter        += nri;
2151
2152     /* Update outer/inner flops */
2153
2154     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*331);
2155 }