src/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecEw_VdwCSTab_GeomW3W3_sse2_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs sse2_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_sse2_double.h"
  34 #include "kernelutil_x86_sse2_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_sse2_double
  38  * Electrostatics interaction: Ewald
  39  * VdW interaction:            CubicSplineTable
  40  * Geometry:                   Water3-Water3
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_VF_sse2_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          vfitab;
  96     __m128i          ifour       = _mm_set1_epi32(4);
  97     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
  98     real             *vftab;
  99     __m128i          ewitab;
 100     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 101     real             *ewtab;
 102     __m128d          dummy_mask,cutoff_mask;
 103     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 104     __m128d          one     = _mm_set1_pd(1.0);
 105     __m128d          two     = _mm_set1_pd(2.0);
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = _mm_set1_pd(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     vftab            = kernel_data->table_vdw->data;
 124     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
 125
 126     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 127     ewtab            = fr->ic->tabq_coul_FDV0;
 128     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 129     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
 134     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 135     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 136     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 137
 138     jq0              = _mm_set1_pd(charge[inr+0]);
 139     jq1              = _mm_set1_pd(charge[inr+1]);
 140     jq2              = _mm_set1_pd(charge[inr+2]);
 141     vdwjidx0A        = 2*vdwtype[inr+0];
 142     qq00             = _mm_mul_pd(iq0,jq0);
 143     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 144     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 145     qq01             = _mm_mul_pd(iq0,jq1);
 146     qq02             = _mm_mul_pd(iq0,jq2);
 147     qq10             = _mm_mul_pd(iq1,jq0);
 148     qq11             = _mm_mul_pd(iq1,jq1);
 149     qq12             = _mm_mul_pd(iq1,jq2);
 150     qq20             = _mm_mul_pd(iq2,jq0);
 151     qq21             = _mm_mul_pd(iq2,jq1);
 152     qq22             = _mm_mul_pd(iq2,jq2);
 153
 154     /* Avoid stupid compiler warnings */
 155     jnrA = jnrB = 0;
 156     j_coord_offsetA = 0;
 157     j_coord_offsetB = 0;
 158
 159     outeriter        = 0;
 160     inneriter        = 0;
 161
 162     /* Start outer loop over neighborlists */
 163     for(iidx=0; iidx<nri; iidx++)
 164     {
 165         /* Load shift vector for this list */
 166         i_shift_offset   = DIM*shiftidx[iidx];
 167
 168         /* Load limits for loop over neighbors */
 169         j_index_start    = jindex[iidx];
 170         j_index_end      = jindex[iidx+1];
 171
 172         /* Get outer coordinate index */
 173         inr              = iinr[iidx];
 174         i_coord_offset   = DIM*inr;
 175
 176         /* Load i particle coords and add shift vector */
 177         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 178                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 179
 180         fix0             = _mm_setzero_pd();
 181         fiy0             = _mm_setzero_pd();
 182         fiz0             = _mm_setzero_pd();
 183         fix1             = _mm_setzero_pd();
 184         fiy1             = _mm_setzero_pd();
 185         fiz1             = _mm_setzero_pd();
 186         fix2             = _mm_setzero_pd();
 187         fiy2             = _mm_setzero_pd();
 188         fiz2             = _mm_setzero_pd();
 189
 190         /* Reset potential sums */
 191         velecsum         = _mm_setzero_pd();
 192         vvdwsum          = _mm_setzero_pd();
 193
 194         /* Start inner kernel loop */
 195         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 196         {
 197
 198             /* Get j neighbor index, and coordinate index */
 199             jnrA             = jjnr[jidx];
 200             jnrB             = jjnr[jidx+1];
 201             j_coord_offsetA  = DIM*jnrA;
 202             j_coord_offsetB  = DIM*jnrB;
 203
 204             /* load j atom coordinates */
 205             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 206                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 207
 208             /* Calculate displacement vector */
 209             dx00             = _mm_sub_pd(ix0,jx0);
 210             dy00             = _mm_sub_pd(iy0,jy0);
 211             dz00             = _mm_sub_pd(iz0,jz0);
 212             dx01             = _mm_sub_pd(ix0,jx1);
 213             dy01             = _mm_sub_pd(iy0,jy1);
 214             dz01             = _mm_sub_pd(iz0,jz1);
 215             dx02             = _mm_sub_pd(ix0,jx2);
 216             dy02             = _mm_sub_pd(iy0,jy2);
 217             dz02             = _mm_sub_pd(iz0,jz2);
 218             dx10             = _mm_sub_pd(ix1,jx0);
 219             dy10             = _mm_sub_pd(iy1,jy0);
 220             dz10             = _mm_sub_pd(iz1,jz0);
 221             dx11             = _mm_sub_pd(ix1,jx1);
 222             dy11             = _mm_sub_pd(iy1,jy1);
 223             dz11             = _mm_sub_pd(iz1,jz1);
 224             dx12             = _mm_sub_pd(ix1,jx2);
 225             dy12             = _mm_sub_pd(iy1,jy2);
 226             dz12             = _mm_sub_pd(iz1,jz2);
 227             dx20             = _mm_sub_pd(ix2,jx0);
 228             dy20             = _mm_sub_pd(iy2,jy0);
 229             dz20             = _mm_sub_pd(iz2,jz0);
 230             dx21             = _mm_sub_pd(ix2,jx1);
 231             dy21             = _mm_sub_pd(iy2,jy1);
 232             dz21             = _mm_sub_pd(iz2,jz1);
 233             dx22             = _mm_sub_pd(ix2,jx2);
 234             dy22             = _mm_sub_pd(iy2,jy2);
 235             dz22             = _mm_sub_pd(iz2,jz2);
 236
 237             /* Calculate squared distance and things based on it */
 238             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 239             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 240             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 241             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 242             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 243             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 244             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 245             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 246             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 247
 248             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 249             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 250             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 251             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 252             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 253             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 254             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 255             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 256             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 257
 258             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 259             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 260             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 261             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 262             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 263             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 264             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 265             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 266             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 267
 268             fjx0             = _mm_setzero_pd();
 269             fjy0             = _mm_setzero_pd();
 270             fjz0             = _mm_setzero_pd();
 271             fjx1             = _mm_setzero_pd();
 272             fjy1             = _mm_setzero_pd();
 273             fjz1             = _mm_setzero_pd();
 274             fjx2             = _mm_setzero_pd();
 275             fjy2             = _mm_setzero_pd();
 276             fjz2             = _mm_setzero_pd();
 277
 278             /**************************
 279              * CALCULATE INTERACTIONS *
 280              **************************/
 281
 282             r00              = _mm_mul_pd(rsq00,rinv00);
 283
 284             /* Calculate table index by multiplying r with table scale and truncate to integer */
 285             rt               = _mm_mul_pd(r00,vftabscale);
 286             vfitab           = _mm_cvttpd_epi32(rt);
 287             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 288             vfitab           = _mm_slli_epi32(vfitab,3);
 289
 290             /* EWALD ELECTROSTATICS */
 291
 292             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 293             ewrt             = _mm_mul_pd(r00,ewtabscale);
 294             ewitab           = _mm_cvttpd_epi32(ewrt);
 295             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 296             ewitab           = _mm_slli_epi32(ewitab,2);
 297             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 298             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 299             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 300             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 301             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 302             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 303             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 304             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 305             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 306             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 307
 308             /* CUBIC SPLINE TABLE DISPERSION */
 309             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 310             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 311             GMX_MM_TRANSPOSE2_PD(Y,F);
 312             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 313             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 314             GMX_MM_TRANSPOSE2_PD(G,H);
 315             Heps             = _mm_mul_pd(vfeps,H);
 316             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 317             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 318             vvdw6            = _mm_mul_pd(c6_00,VV);
 319             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 320             fvdw6            = _mm_mul_pd(c6_00,FF);
 321
 322             /* CUBIC SPLINE TABLE REPULSION */
 323             vfitab           = _mm_add_epi32(vfitab,ifour);
 324             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 325             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
 326             GMX_MM_TRANSPOSE2_PD(Y,F);
 327             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 328             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
 329             GMX_MM_TRANSPOSE2_PD(G,H);
 330             Heps             = _mm_mul_pd(vfeps,H);
 331             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 332             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 333             vvdw12           = _mm_mul_pd(c12_00,VV);
 334             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 335             fvdw12           = _mm_mul_pd(c12_00,FF);
 336             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 337             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 338
 339             /* Update potential sum for this i atom from the interaction with this j atom. */
 340             velecsum         = _mm_add_pd(velecsum,velec);
 341             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 342
 343             fscal            = _mm_add_pd(felec,fvdw);
 344
 345             /* Calculate temporary vectorial force */
 346             tx               = _mm_mul_pd(fscal,dx00);
 347             ty               = _mm_mul_pd(fscal,dy00);
 348             tz               = _mm_mul_pd(fscal,dz00);
 349
 350             /* Update vectorial force */
 351             fix0             = _mm_add_pd(fix0,tx);
 352             fiy0             = _mm_add_pd(fiy0,ty);
 353             fiz0             = _mm_add_pd(fiz0,tz);
 354
 355             fjx0             = _mm_add_pd(fjx0,tx);
 356             fjy0             = _mm_add_pd(fjy0,ty);
 357             fjz0             = _mm_add_pd(fjz0,tz);
 358
 359             /**************************
 360              * CALCULATE INTERACTIONS *
 361              **************************/
 362
 363             r01              = _mm_mul_pd(rsq01,rinv01);
 364
 365             /* EWALD ELECTROSTATICS */
 366
 367             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 368             ewrt             = _mm_mul_pd(r01,ewtabscale);
 369             ewitab           = _mm_cvttpd_epi32(ewrt);
 370             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 371             ewitab           = _mm_slli_epi32(ewitab,2);
 372             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 373             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 374             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 375             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 376             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 377             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 378             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 379             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 380             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 381             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 382
 383             /* Update potential sum for this i atom from the interaction with this j atom. */
 384             velecsum         = _mm_add_pd(velecsum,velec);
 385
 386             fscal            = felec;
 387
 388             /* Calculate temporary vectorial force */
 389             tx               = _mm_mul_pd(fscal,dx01);
 390             ty               = _mm_mul_pd(fscal,dy01);
 391             tz               = _mm_mul_pd(fscal,dz01);
 392
 393             /* Update vectorial force */
 394             fix0             = _mm_add_pd(fix0,tx);
 395             fiy0             = _mm_add_pd(fiy0,ty);
 396             fiz0             = _mm_add_pd(fiz0,tz);
 397
 398             fjx1             = _mm_add_pd(fjx1,tx);
 399             fjy1             = _mm_add_pd(fjy1,ty);
 400             fjz1             = _mm_add_pd(fjz1,tz);
 401
 402             /**************************
 403              * CALCULATE INTERACTIONS *
 404              **************************/
 405
 406             r02              = _mm_mul_pd(rsq02,rinv02);
 407
 408             /* EWALD ELECTROSTATICS */
 409
 410             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 411             ewrt             = _mm_mul_pd(r02,ewtabscale);
 412             ewitab           = _mm_cvttpd_epi32(ewrt);
 413             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 414             ewitab           = _mm_slli_epi32(ewitab,2);
 415             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 416             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 417             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 418             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 419             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 420             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 421             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 422             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 423             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 424             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 425
 426             /* Update potential sum for this i atom from the interaction with this j atom. */
 427             velecsum         = _mm_add_pd(velecsum,velec);
 428
 429             fscal            = felec;
 430
 431             /* Calculate temporary vectorial force */
 432             tx               = _mm_mul_pd(fscal,dx02);
 433             ty               = _mm_mul_pd(fscal,dy02);
 434             tz               = _mm_mul_pd(fscal,dz02);
 435
 436             /* Update vectorial force */
 437             fix0             = _mm_add_pd(fix0,tx);
 438             fiy0             = _mm_add_pd(fiy0,ty);
 439             fiz0             = _mm_add_pd(fiz0,tz);
 440
 441             fjx2             = _mm_add_pd(fjx2,tx);
 442             fjy2             = _mm_add_pd(fjy2,ty);
 443             fjz2             = _mm_add_pd(fjz2,tz);
 444
 445             /**************************
 446              * CALCULATE INTERACTIONS *
 447              **************************/
 448
 449             r10              = _mm_mul_pd(rsq10,rinv10);
 450
 451             /* EWALD ELECTROSTATICS */
 452
 453             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 454             ewrt             = _mm_mul_pd(r10,ewtabscale);
 455             ewitab           = _mm_cvttpd_epi32(ewrt);
 456             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 457             ewitab           = _mm_slli_epi32(ewitab,2);
 458             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 459             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 460             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 461             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 462             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 463             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 464             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 465             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 466             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 467             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 468
 469             /* Update potential sum for this i atom from the interaction with this j atom. */
 470             velecsum         = _mm_add_pd(velecsum,velec);
 471
 472             fscal            = felec;
 473
 474             /* Calculate temporary vectorial force */
 475             tx               = _mm_mul_pd(fscal,dx10);
 476             ty               = _mm_mul_pd(fscal,dy10);
 477             tz               = _mm_mul_pd(fscal,dz10);
 478
 479             /* Update vectorial force */
 480             fix1             = _mm_add_pd(fix1,tx);
 481             fiy1             = _mm_add_pd(fiy1,ty);
 482             fiz1             = _mm_add_pd(fiz1,tz);
 483
 484             fjx0             = _mm_add_pd(fjx0,tx);
 485             fjy0             = _mm_add_pd(fjy0,ty);
 486             fjz0             = _mm_add_pd(fjz0,tz);
 487
 488             /**************************
 489              * CALCULATE INTERACTIONS *
 490              **************************/
 491
 492             r11              = _mm_mul_pd(rsq11,rinv11);
 493
 494             /* EWALD ELECTROSTATICS */
 495
 496             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 497             ewrt             = _mm_mul_pd(r11,ewtabscale);
 498             ewitab           = _mm_cvttpd_epi32(ewrt);
 499             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 500             ewitab           = _mm_slli_epi32(ewitab,2);
 501             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 502             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 503             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 504             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 505             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 506             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 507             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 508             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 509             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
 510             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
 511
 512             /* Update potential sum for this i atom from the interaction with this j atom. */
 513             velecsum         = _mm_add_pd(velecsum,velec);
 514
 515             fscal            = felec;
 516
 517             /* Calculate temporary vectorial force */
 518             tx               = _mm_mul_pd(fscal,dx11);
 519             ty               = _mm_mul_pd(fscal,dy11);
 520             tz               = _mm_mul_pd(fscal,dz11);
 521
 522             /* Update vectorial force */
 523             fix1             = _mm_add_pd(fix1,tx);
 524             fiy1             = _mm_add_pd(fiy1,ty);
 525             fiz1             = _mm_add_pd(fiz1,tz);
 526
 527             fjx1             = _mm_add_pd(fjx1,tx);
 528             fjy1             = _mm_add_pd(fjy1,ty);
 529             fjz1             = _mm_add_pd(fjz1,tz);
 530
 531             /**************************
 532              * CALCULATE INTERACTIONS *
 533              **************************/
 534
 535             r12              = _mm_mul_pd(rsq12,rinv12);
 536
 537             /* EWALD ELECTROSTATICS */
 538
 539             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 540             ewrt             = _mm_mul_pd(r12,ewtabscale);
 541             ewitab           = _mm_cvttpd_epi32(ewrt);
 542             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 543             ewitab           = _mm_slli_epi32(ewitab,2);
 544             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 545             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 546             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 547             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 548             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 549             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 550             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 551             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 552             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
 553             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
 554
 555             /* Update potential sum for this i atom from the interaction with this j atom. */
 556             velecsum         = _mm_add_pd(velecsum,velec);
 557
 558             fscal            = felec;
 559
 560             /* Calculate temporary vectorial force */
 561             tx               = _mm_mul_pd(fscal,dx12);
 562             ty               = _mm_mul_pd(fscal,dy12);
 563             tz               = _mm_mul_pd(fscal,dz12);
 564
 565             /* Update vectorial force */
 566             fix1             = _mm_add_pd(fix1,tx);
 567             fiy1             = _mm_add_pd(fiy1,ty);
 568             fiz1             = _mm_add_pd(fiz1,tz);
 569
 570             fjx2             = _mm_add_pd(fjx2,tx);
 571             fjy2             = _mm_add_pd(fjy2,ty);
 572             fjz2             = _mm_add_pd(fjz2,tz);
 573
 574             /**************************
 575              * CALCULATE INTERACTIONS *
 576              **************************/
 577
 578             r20              = _mm_mul_pd(rsq20,rinv20);
 579
 580             /* EWALD ELECTROSTATICS */
 581
 582             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 583             ewrt             = _mm_mul_pd(r20,ewtabscale);
 584             ewitab           = _mm_cvttpd_epi32(ewrt);
 585             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 586             ewitab           = _mm_slli_epi32(ewitab,2);
 587             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 588             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 589             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 590             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 591             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 592             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 593             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 594             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 595             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
 596             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
 597
 598             /* Update potential sum for this i atom from the interaction with this j atom. */
 599             velecsum         = _mm_add_pd(velecsum,velec);
 600
 601             fscal            = felec;
 602
 603             /* Calculate temporary vectorial force */
 604             tx               = _mm_mul_pd(fscal,dx20);
 605             ty               = _mm_mul_pd(fscal,dy20);
 606             tz               = _mm_mul_pd(fscal,dz20);
 607
 608             /* Update vectorial force */
 609             fix2             = _mm_add_pd(fix2,tx);
 610             fiy2             = _mm_add_pd(fiy2,ty);
 611             fiz2             = _mm_add_pd(fiz2,tz);
 612
 613             fjx0             = _mm_add_pd(fjx0,tx);
 614             fjy0             = _mm_add_pd(fjy0,ty);
 615             fjz0             = _mm_add_pd(fjz0,tz);
 616
 617             /**************************
 618              * CALCULATE INTERACTIONS *
 619              **************************/
 620
 621             r21              = _mm_mul_pd(rsq21,rinv21);
 622
 623             /* EWALD ELECTROSTATICS */
 624
 625             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 626             ewrt             = _mm_mul_pd(r21,ewtabscale);
 627             ewitab           = _mm_cvttpd_epi32(ewrt);
 628             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 629             ewitab           = _mm_slli_epi32(ewitab,2);
 630             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 631             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 632             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 633             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 634             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 635             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 636             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 637             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 638             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
 639             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
 640
 641             /* Update potential sum for this i atom from the interaction with this j atom. */
 642             velecsum         = _mm_add_pd(velecsum,velec);
 643
 644             fscal            = felec;
 645
 646             /* Calculate temporary vectorial force */
 647             tx               = _mm_mul_pd(fscal,dx21);
 648             ty               = _mm_mul_pd(fscal,dy21);
 649             tz               = _mm_mul_pd(fscal,dz21);
 650
 651             /* Update vectorial force */
 652             fix2             = _mm_add_pd(fix2,tx);
 653             fiy2             = _mm_add_pd(fiy2,ty);
 654             fiz2             = _mm_add_pd(fiz2,tz);
 655
 656             fjx1             = _mm_add_pd(fjx1,tx);
 657             fjy1             = _mm_add_pd(fjy1,ty);
 658             fjz1             = _mm_add_pd(fjz1,tz);
 659
 660             /**************************
 661              * CALCULATE INTERACTIONS *
 662              **************************/
 663
 664             r22              = _mm_mul_pd(rsq22,rinv22);
 665
 666             /* EWALD ELECTROSTATICS */
 667
 668             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 669             ewrt             = _mm_mul_pd(r22,ewtabscale);
 670             ewitab           = _mm_cvttpd_epi32(ewrt);
 671             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 672             ewitab           = _mm_slli_epi32(ewitab,2);
 673             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 674             ewtabD           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 675             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 676             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 677             ewtabFn          = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
 678             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 679             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 680             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 681             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
 682             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
 683
 684             /* Update potential sum for this i atom from the interaction with this j atom. */
 685             velecsum         = _mm_add_pd(velecsum,velec);
 686
 687             fscal            = felec;
 688
 689             /* Calculate temporary vectorial force */
 690             tx               = _mm_mul_pd(fscal,dx22);
 691             ty               = _mm_mul_pd(fscal,dy22);
 692             tz               = _mm_mul_pd(fscal,dz22);
 693
 694             /* Update vectorial force */
 695             fix2             = _mm_add_pd(fix2,tx);
 696             fiy2             = _mm_add_pd(fiy2,ty);
 697             fiz2             = _mm_add_pd(fiz2,tz);
 698
 699             fjx2             = _mm_add_pd(fjx2,tx);
 700             fjy2             = _mm_add_pd(fjy2,ty);
 701             fjz2             = _mm_add_pd(fjz2,tz);
 702
 703             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
 704
 705             /* Inner loop uses 403 flops */
 706         }
 707
 708         if(jidx<j_index_end)
 709         {
 710
 711             jnrA             = jjnr[jidx];
 712             j_coord_offsetA  = DIM*jnrA;
 713
 714             /* load j atom coordinates */
 715             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 716                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
 717
 718             /* Calculate displacement vector */
 719             dx00             = _mm_sub_pd(ix0,jx0);
 720             dy00             = _mm_sub_pd(iy0,jy0);
 721             dz00             = _mm_sub_pd(iz0,jz0);
 722             dx01             = _mm_sub_pd(ix0,jx1);
 723             dy01             = _mm_sub_pd(iy0,jy1);
 724             dz01             = _mm_sub_pd(iz0,jz1);
 725             dx02             = _mm_sub_pd(ix0,jx2);
 726             dy02             = _mm_sub_pd(iy0,jy2);
 727             dz02             = _mm_sub_pd(iz0,jz2);
 728             dx10             = _mm_sub_pd(ix1,jx0);
 729             dy10             = _mm_sub_pd(iy1,jy0);
 730             dz10             = _mm_sub_pd(iz1,jz0);
 731             dx11             = _mm_sub_pd(ix1,jx1);
 732             dy11             = _mm_sub_pd(iy1,jy1);
 733             dz11             = _mm_sub_pd(iz1,jz1);
 734             dx12             = _mm_sub_pd(ix1,jx2);
 735             dy12             = _mm_sub_pd(iy1,jy2);
 736             dz12             = _mm_sub_pd(iz1,jz2);
 737             dx20             = _mm_sub_pd(ix2,jx0);
 738             dy20             = _mm_sub_pd(iy2,jy0);
 739             dz20             = _mm_sub_pd(iz2,jz0);
 740             dx21             = _mm_sub_pd(ix2,jx1);
 741             dy21             = _mm_sub_pd(iy2,jy1);
 742             dz21             = _mm_sub_pd(iz2,jz1);
 743             dx22             = _mm_sub_pd(ix2,jx2);
 744             dy22             = _mm_sub_pd(iy2,jy2);
 745             dz22             = _mm_sub_pd(iz2,jz2);
 746
 747             /* Calculate squared distance and things based on it */
 748             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 749             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
 750             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
 751             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
 752             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 753             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 754             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
 755             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 756             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 757
 758             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 759             rinv01           = gmx_mm_invsqrt_pd(rsq01);
 760             rinv02           = gmx_mm_invsqrt_pd(rsq02);
 761             rinv10           = gmx_mm_invsqrt_pd(rsq10);
 762             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 763             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 764             rinv20           = gmx_mm_invsqrt_pd(rsq20);
 765             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 766             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 767
 768             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 769             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
 770             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
 771             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
 772             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 773             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 774             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
 775             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 776             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 777
 778             fjx0             = _mm_setzero_pd();
 779             fjy0             = _mm_setzero_pd();
 780             fjz0             = _mm_setzero_pd();
 781             fjx1             = _mm_setzero_pd();
 782             fjy1             = _mm_setzero_pd();
 783             fjz1             = _mm_setzero_pd();
 784             fjx2             = _mm_setzero_pd();
 785             fjy2             = _mm_setzero_pd();
 786             fjz2             = _mm_setzero_pd();
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             r00              = _mm_mul_pd(rsq00,rinv00);
 793
 794             /* Calculate table index by multiplying r with table scale and truncate to integer */
 795             rt               = _mm_mul_pd(r00,vftabscale);
 796             vfitab           = _mm_cvttpd_epi32(rt);
 797             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
 798             vfitab           = _mm_slli_epi32(vfitab,3);
 799
 800             /* EWALD ELECTROSTATICS */
 801
 802             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 803             ewrt             = _mm_mul_pd(r00,ewtabscale);
 804             ewitab           = _mm_cvttpd_epi32(ewrt);
 805             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 806             ewitab           = _mm_slli_epi32(ewitab,2);
 807             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 808             ewtabD           = _mm_setzero_pd();
 809             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 810             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 811             ewtabFn          = _mm_setzero_pd();
 812             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 813             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 814             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 815             velec            = _mm_mul_pd(qq00,_mm_sub_pd(rinv00,velec));
 816             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 817
 818             /* CUBIC SPLINE TABLE DISPERSION */
 819             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 820             F                = _mm_setzero_pd();
 821             GMX_MM_TRANSPOSE2_PD(Y,F);
 822             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 823             H                = _mm_setzero_pd();
 824             GMX_MM_TRANSPOSE2_PD(G,H);
 825             Heps             = _mm_mul_pd(vfeps,H);
 826             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 827             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 828             vvdw6            = _mm_mul_pd(c6_00,VV);
 829             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 830             fvdw6            = _mm_mul_pd(c6_00,FF);
 831
 832             /* CUBIC SPLINE TABLE REPULSION */
 833             vfitab           = _mm_add_epi32(vfitab,ifour);
 834             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
 835             F                = _mm_setzero_pd();
 836             GMX_MM_TRANSPOSE2_PD(Y,F);
 837             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
 838             H                = _mm_setzero_pd();
 839             GMX_MM_TRANSPOSE2_PD(G,H);
 840             Heps             = _mm_mul_pd(vfeps,H);
 841             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
 842             VV               = _mm_add_pd(Y,_mm_mul_pd(vfeps,Fp));
 843             vvdw12           = _mm_mul_pd(c12_00,VV);
 844             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
 845             fvdw12           = _mm_mul_pd(c12_00,FF);
 846             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 847             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 848
 849             /* Update potential sum for this i atom from the interaction with this j atom. */
 850             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 851             velecsum         = _mm_add_pd(velecsum,velec);
 852             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 853             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 854
 855             fscal            = _mm_add_pd(felec,fvdw);
 856
 857             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 858
 859             /* Calculate temporary vectorial force */
 860             tx               = _mm_mul_pd(fscal,dx00);
 861             ty               = _mm_mul_pd(fscal,dy00);
 862             tz               = _mm_mul_pd(fscal,dz00);
 863
 864             /* Update vectorial force */
 865             fix0             = _mm_add_pd(fix0,tx);
 866             fiy0             = _mm_add_pd(fiy0,ty);
 867             fiz0             = _mm_add_pd(fiz0,tz);
 868
 869             fjx0             = _mm_add_pd(fjx0,tx);
 870             fjy0             = _mm_add_pd(fjy0,ty);
 871             fjz0             = _mm_add_pd(fjz0,tz);
 872
 873             /**************************
 874              * CALCULATE INTERACTIONS *
 875              **************************/
 876
 877             r01              = _mm_mul_pd(rsq01,rinv01);
 878
 879             /* EWALD ELECTROSTATICS */
 880
 881             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 882             ewrt             = _mm_mul_pd(r01,ewtabscale);
 883             ewitab           = _mm_cvttpd_epi32(ewrt);
 884             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 885             ewitab           = _mm_slli_epi32(ewitab,2);
 886             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 887             ewtabD           = _mm_setzero_pd();
 888             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 889             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 890             ewtabFn          = _mm_setzero_pd();
 891             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 892             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 893             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 894             velec            = _mm_mul_pd(qq01,_mm_sub_pd(rinv01,velec));
 895             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
 896
 897             /* Update potential sum for this i atom from the interaction with this j atom. */
 898             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 899             velecsum         = _mm_add_pd(velecsum,velec);
 900
 901             fscal            = felec;
 902
 903             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 904
 905             /* Calculate temporary vectorial force */
 906             tx               = _mm_mul_pd(fscal,dx01);
 907             ty               = _mm_mul_pd(fscal,dy01);
 908             tz               = _mm_mul_pd(fscal,dz01);
 909
 910             /* Update vectorial force */
 911             fix0             = _mm_add_pd(fix0,tx);
 912             fiy0             = _mm_add_pd(fiy0,ty);
 913             fiz0             = _mm_add_pd(fiz0,tz);
 914
 915             fjx1             = _mm_add_pd(fjx1,tx);
 916             fjy1             = _mm_add_pd(fjy1,ty);
 917             fjz1             = _mm_add_pd(fjz1,tz);
 918
 919             /**************************
 920              * CALCULATE INTERACTIONS *
 921              **************************/
 922
 923             r02              = _mm_mul_pd(rsq02,rinv02);
 924
 925             /* EWALD ELECTROSTATICS */
 926
 927             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 928             ewrt             = _mm_mul_pd(r02,ewtabscale);
 929             ewitab           = _mm_cvttpd_epi32(ewrt);
 930             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 931             ewitab           = _mm_slli_epi32(ewitab,2);
 932             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 933             ewtabD           = _mm_setzero_pd();
 934             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 935             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 936             ewtabFn          = _mm_setzero_pd();
 937             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 938             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 939             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 940             velec            = _mm_mul_pd(qq02,_mm_sub_pd(rinv02,velec));
 941             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
 942
 943             /* Update potential sum for this i atom from the interaction with this j atom. */
 944             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 945             velecsum         = _mm_add_pd(velecsum,velec);
 946
 947             fscal            = felec;
 948
 949             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 950
 951             /* Calculate temporary vectorial force */
 952             tx               = _mm_mul_pd(fscal,dx02);
 953             ty               = _mm_mul_pd(fscal,dy02);
 954             tz               = _mm_mul_pd(fscal,dz02);
 955
 956             /* Update vectorial force */
 957             fix0             = _mm_add_pd(fix0,tx);
 958             fiy0             = _mm_add_pd(fiy0,ty);
 959             fiz0             = _mm_add_pd(fiz0,tz);
 960
 961             fjx2             = _mm_add_pd(fjx2,tx);
 962             fjy2             = _mm_add_pd(fjy2,ty);
 963             fjz2             = _mm_add_pd(fjz2,tz);
 964
 965             /**************************
 966              * CALCULATE INTERACTIONS *
 967              **************************/
 968
 969             r10              = _mm_mul_pd(rsq10,rinv10);
 970
 971             /* EWALD ELECTROSTATICS */
 972
 973             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 974             ewrt             = _mm_mul_pd(r10,ewtabscale);
 975             ewitab           = _mm_cvttpd_epi32(ewrt);
 976             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
 977             ewitab           = _mm_slli_epi32(ewitab,2);
 978             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 979             ewtabD           = _mm_setzero_pd();
 980             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 981             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
 982             ewtabFn          = _mm_setzero_pd();
 983             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 984             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
 985             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
 986             velec            = _mm_mul_pd(qq10,_mm_sub_pd(rinv10,velec));
 987             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
 988
 989             /* Update potential sum for this i atom from the interaction with this j atom. */
 990             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 991             velecsum         = _mm_add_pd(velecsum,velec);
 992
 993             fscal            = felec;
 994
 995             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 996
 997             /* Calculate temporary vectorial force */
 998             tx               = _mm_mul_pd(fscal,dx10);
 999             ty               = _mm_mul_pd(fscal,dy10);
1000             tz               = _mm_mul_pd(fscal,dz10);
1001
1002             /* Update vectorial force */
1003             fix1             = _mm_add_pd(fix1,tx);
1004             fiy1             = _mm_add_pd(fiy1,ty);
1005             fiz1             = _mm_add_pd(fiz1,tz);
1006
1007             fjx0             = _mm_add_pd(fjx0,tx);
1008             fjy0             = _mm_add_pd(fjy0,ty);
1009             fjz0             = _mm_add_pd(fjz0,tz);
1010
1011             /**************************
1012              * CALCULATE INTERACTIONS *
1013              **************************/
1014
1015             r11              = _mm_mul_pd(rsq11,rinv11);
1016
1017             /* EWALD ELECTROSTATICS */
1018
1019             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1020             ewrt             = _mm_mul_pd(r11,ewtabscale);
1021             ewitab           = _mm_cvttpd_epi32(ewrt);
1022             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1023             ewitab           = _mm_slli_epi32(ewitab,2);
1024             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1025             ewtabD           = _mm_setzero_pd();
1026             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1027             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1028             ewtabFn          = _mm_setzero_pd();
1029             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1030             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1031             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1032             velec            = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
1033             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1034
1035             /* Update potential sum for this i atom from the interaction with this j atom. */
1036             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1037             velecsum         = _mm_add_pd(velecsum,velec);
1038
1039             fscal            = felec;
1040
1041             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1042
1043             /* Calculate temporary vectorial force */
1044             tx               = _mm_mul_pd(fscal,dx11);
1045             ty               = _mm_mul_pd(fscal,dy11);
1046             tz               = _mm_mul_pd(fscal,dz11);
1047
1048             /* Update vectorial force */
1049             fix1             = _mm_add_pd(fix1,tx);
1050             fiy1             = _mm_add_pd(fiy1,ty);
1051             fiz1             = _mm_add_pd(fiz1,tz);
1052
1053             fjx1             = _mm_add_pd(fjx1,tx);
1054             fjy1             = _mm_add_pd(fjy1,ty);
1055             fjz1             = _mm_add_pd(fjz1,tz);
1056
1057             /**************************
1058              * CALCULATE INTERACTIONS *
1059              **************************/
1060
1061             r12              = _mm_mul_pd(rsq12,rinv12);
1062
1063             /* EWALD ELECTROSTATICS */
1064
1065             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1066             ewrt             = _mm_mul_pd(r12,ewtabscale);
1067             ewitab           = _mm_cvttpd_epi32(ewrt);
1068             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1069             ewitab           = _mm_slli_epi32(ewitab,2);
1070             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1071             ewtabD           = _mm_setzero_pd();
1072             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1073             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1074             ewtabFn          = _mm_setzero_pd();
1075             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1076             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1077             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1078             velec            = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
1079             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1080
1081             /* Update potential sum for this i atom from the interaction with this j atom. */
1082             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1083             velecsum         = _mm_add_pd(velecsum,velec);
1084
1085             fscal            = felec;
1086
1087             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1088
1089             /* Calculate temporary vectorial force */
1090             tx               = _mm_mul_pd(fscal,dx12);
1091             ty               = _mm_mul_pd(fscal,dy12);
1092             tz               = _mm_mul_pd(fscal,dz12);
1093
1094             /* Update vectorial force */
1095             fix1             = _mm_add_pd(fix1,tx);
1096             fiy1             = _mm_add_pd(fiy1,ty);
1097             fiz1             = _mm_add_pd(fiz1,tz);
1098
1099             fjx2             = _mm_add_pd(fjx2,tx);
1100             fjy2             = _mm_add_pd(fjy2,ty);
1101             fjz2             = _mm_add_pd(fjz2,tz);
1102
1103             /**************************
1104              * CALCULATE INTERACTIONS *
1105              **************************/
1106
1107             r20              = _mm_mul_pd(rsq20,rinv20);
1108
1109             /* EWALD ELECTROSTATICS */
1110
1111             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1112             ewrt             = _mm_mul_pd(r20,ewtabscale);
1113             ewitab           = _mm_cvttpd_epi32(ewrt);
1114             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1115             ewitab           = _mm_slli_epi32(ewitab,2);
1116             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1117             ewtabD           = _mm_setzero_pd();
1118             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1119             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1120             ewtabFn          = _mm_setzero_pd();
1121             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1122             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1123             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1124             velec            = _mm_mul_pd(qq20,_mm_sub_pd(rinv20,velec));
1125             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1126
1127             /* Update potential sum for this i atom from the interaction with this j atom. */
1128             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1129             velecsum         = _mm_add_pd(velecsum,velec);
1130
1131             fscal            = felec;
1132
1133             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1134
1135             /* Calculate temporary vectorial force */
1136             tx               = _mm_mul_pd(fscal,dx20);
1137             ty               = _mm_mul_pd(fscal,dy20);
1138             tz               = _mm_mul_pd(fscal,dz20);
1139
1140             /* Update vectorial force */
1141             fix2             = _mm_add_pd(fix2,tx);
1142             fiy2             = _mm_add_pd(fiy2,ty);
1143             fiz2             = _mm_add_pd(fiz2,tz);
1144
1145             fjx0             = _mm_add_pd(fjx0,tx);
1146             fjy0             = _mm_add_pd(fjy0,ty);
1147             fjz0             = _mm_add_pd(fjz0,tz);
1148
1149             /**************************
1150              * CALCULATE INTERACTIONS *
1151              **************************/
1152
1153             r21              = _mm_mul_pd(rsq21,rinv21);
1154
1155             /* EWALD ELECTROSTATICS */
1156
1157             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1158             ewrt             = _mm_mul_pd(r21,ewtabscale);
1159             ewitab           = _mm_cvttpd_epi32(ewrt);
1160             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1161             ewitab           = _mm_slli_epi32(ewitab,2);
1162             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1163             ewtabD           = _mm_setzero_pd();
1164             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1165             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1166             ewtabFn          = _mm_setzero_pd();
1167             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1168             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1169             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1170             velec            = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1171             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1172
1173             /* Update potential sum for this i atom from the interaction with this j atom. */
1174             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1175             velecsum         = _mm_add_pd(velecsum,velec);
1176
1177             fscal            = felec;
1178
1179             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1180
1181             /* Calculate temporary vectorial force */
1182             tx               = _mm_mul_pd(fscal,dx21);
1183             ty               = _mm_mul_pd(fscal,dy21);
1184             tz               = _mm_mul_pd(fscal,dz21);
1185
1186             /* Update vectorial force */
1187             fix2             = _mm_add_pd(fix2,tx);
1188             fiy2             = _mm_add_pd(fiy2,ty);
1189             fiz2             = _mm_add_pd(fiz2,tz);
1190
1191             fjx1             = _mm_add_pd(fjx1,tx);
1192             fjy1             = _mm_add_pd(fjy1,ty);
1193             fjz1             = _mm_add_pd(fjz1,tz);
1194
1195             /**************************
1196              * CALCULATE INTERACTIONS *
1197              **************************/
1198
1199             r22              = _mm_mul_pd(rsq22,rinv22);
1200
1201             /* EWALD ELECTROSTATICS */
1202
1203             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1204             ewrt             = _mm_mul_pd(r22,ewtabscale);
1205             ewitab           = _mm_cvttpd_epi32(ewrt);
1206             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1207             ewitab           = _mm_slli_epi32(ewitab,2);
1208             ewtabF           = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1209             ewtabD           = _mm_setzero_pd();
1210             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1211             ewtabV           = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1212             ewtabFn          = _mm_setzero_pd();
1213             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1214             felec            = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1215             velec            = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1216             velec            = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1217             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1218
1219             /* Update potential sum for this i atom from the interaction with this j atom. */
1220             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1221             velecsum         = _mm_add_pd(velecsum,velec);
1222
1223             fscal            = felec;
1224
1225             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1226
1227             /* Calculate temporary vectorial force */
1228             tx               = _mm_mul_pd(fscal,dx22);
1229             ty               = _mm_mul_pd(fscal,dy22);
1230             tz               = _mm_mul_pd(fscal,dz22);
1231
1232             /* Update vectorial force */
1233             fix2             = _mm_add_pd(fix2,tx);
1234             fiy2             = _mm_add_pd(fiy2,ty);
1235             fiz2             = _mm_add_pd(fiz2,tz);
1236
1237             fjx2             = _mm_add_pd(fjx2,tx);
1238             fjy2             = _mm_add_pd(fjy2,ty);
1239             fjz2             = _mm_add_pd(fjz2,tz);
1240
1241             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1242
1243             /* Inner loop uses 403 flops */
1244         }
1245
1246         /* End of innermost loop */
1247
1248         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1249                                               f+i_coord_offset,fshift+i_shift_offset);
1250
1251         ggid                        = gid[iidx];
1252         /* Update potential energies */
1253         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1254         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1255
1256         /* Increment number of inner iterations */
1257         inneriter                  += j_index_end - j_index_start;
1258
1259         /* Outer loop uses 20 flops */
1260     }
1261
1262     /* Increment number of outer iterations */
1263     outeriter        += nri;
1264
1265     /* Update outer/inner flops */
1266
1267     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_VF,outeriter*20 + inneriter*403);
1268 }
1269 /*
1270  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_sse2_double
1271  * Electrostatics interaction: Ewald
1272  * VdW interaction:            CubicSplineTable
1273  * Geometry:                   Water3-Water3
1274  * Calculate force/pot:        Force
1275  */
1276 void
1277 nb_kernel_ElecEw_VdwCSTab_GeomW3W3_F_sse2_double
1278                     (t_nblist * gmx_restrict                nlist,
1279                      rvec * gmx_restrict                    xx,
1280                      rvec * gmx_restrict                    ff,
1281                      t_forcerec * gmx_restrict              fr,
1282                      t_mdatoms * gmx_restrict               mdatoms,
1283                      nb_kernel_data_t * gmx_restrict        kernel_data,
1284                      t_nrnb * gmx_restrict                  nrnb)
1285 {
1286     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1287      * just 0 for non-waters.
1288      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1289      * jnr indices corresponding to data put in the four positions in the SIMD register.
1290      */
1291     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1292     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1293     int              jnrA,jnrB;
1294     int              j_coord_offsetA,j_coord_offsetB;
1295     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1296     real             rcutoff_scalar;
1297     real             *shiftvec,*fshift,*x,*f;
1298     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1299     int              vdwioffset0;
1300     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1301     int              vdwioffset1;
1302     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1303     int              vdwioffset2;
1304     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1305     int              vdwjidx0A,vdwjidx0B;
1306     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1307     int              vdwjidx1A,vdwjidx1B;
1308     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1309     int              vdwjidx2A,vdwjidx2B;
1310     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1311     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1312     __m128d          dx01,dy01,dz01,rsq01,rinv01,rinvsq01,r01,qq01,c6_01,c12_01;
1313     __m128d          dx02,dy02,dz02,rsq02,rinv02,rinvsq02,r02,qq02,c6_02,c12_02;
1314     __m128d          dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
1315     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1316     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1317     __m128d          dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
1318     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1319     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1320     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1321     real             *charge;
1322     int              nvdwtype;
1323     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1324     int              *vdwtype;
1325     real             *vdwparam;
1326     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1327     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1328     __m128i          vfitab;
1329     __m128i          ifour       = _mm_set1_epi32(4);
1330     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1331     real             *vftab;
1332     __m128i          ewitab;
1333     __m128d          ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1334     real             *ewtab;
1335     __m128d          dummy_mask,cutoff_mask;
1336     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1337     __m128d          one     = _mm_set1_pd(1.0);
1338     __m128d          two     = _mm_set1_pd(2.0);
1339     x                = xx[0];
1340     f                = ff[0];
1341
1342     nri              = nlist->nri;
1343     iinr             = nlist->iinr;
1344     jindex           = nlist->jindex;
1345     jjnr             = nlist->jjnr;
1346     shiftidx         = nlist->shift;
1347     gid              = nlist->gid;
1348     shiftvec         = fr->shift_vec[0];
1349     fshift           = fr->fshift[0];
1350     facel            = _mm_set1_pd(fr->epsfac);
1351     charge           = mdatoms->chargeA;
1352     nvdwtype         = fr->ntype;
1353     vdwparam         = fr->nbfp;
1354     vdwtype          = mdatoms->typeA;
1355
1356     vftab            = kernel_data->table_vdw->data;
1357     vftabscale       = _mm_set1_pd(kernel_data->table_vdw->scale);
1358
1359     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
1360     ewtab            = fr->ic->tabq_coul_F;
1361     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
1362     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1363
1364     /* Setup water-specific parameters */
1365     inr              = nlist->iinr[0];
1366     iq0              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
1367     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1368     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1369     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1370
1371     jq0              = _mm_set1_pd(charge[inr+0]);
1372     jq1              = _mm_set1_pd(charge[inr+1]);
1373     jq2              = _mm_set1_pd(charge[inr+2]);
1374     vdwjidx0A        = 2*vdwtype[inr+0];
1375     qq00             = _mm_mul_pd(iq0,jq0);
1376     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1377     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1378     qq01             = _mm_mul_pd(iq0,jq1);
1379     qq02             = _mm_mul_pd(iq0,jq2);
1380     qq10             = _mm_mul_pd(iq1,jq0);
1381     qq11             = _mm_mul_pd(iq1,jq1);
1382     qq12             = _mm_mul_pd(iq1,jq2);
1383     qq20             = _mm_mul_pd(iq2,jq0);
1384     qq21             = _mm_mul_pd(iq2,jq1);
1385     qq22             = _mm_mul_pd(iq2,jq2);
1386
1387     /* Avoid stupid compiler warnings */
1388     jnrA = jnrB = 0;
1389     j_coord_offsetA = 0;
1390     j_coord_offsetB = 0;
1391
1392     outeriter        = 0;
1393     inneriter        = 0;
1394
1395     /* Start outer loop over neighborlists */
1396     for(iidx=0; iidx<nri; iidx++)
1397     {
1398         /* Load shift vector for this list */
1399         i_shift_offset   = DIM*shiftidx[iidx];
1400
1401         /* Load limits for loop over neighbors */
1402         j_index_start    = jindex[iidx];
1403         j_index_end      = jindex[iidx+1];
1404
1405         /* Get outer coordinate index */
1406         inr              = iinr[iidx];
1407         i_coord_offset   = DIM*inr;
1408
1409         /* Load i particle coords and add shift vector */
1410         gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1411                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
1412
1413         fix0             = _mm_setzero_pd();
1414         fiy0             = _mm_setzero_pd();
1415         fiz0             = _mm_setzero_pd();
1416         fix1             = _mm_setzero_pd();
1417         fiy1             = _mm_setzero_pd();
1418         fiz1             = _mm_setzero_pd();
1419         fix2             = _mm_setzero_pd();
1420         fiy2             = _mm_setzero_pd();
1421         fiz2             = _mm_setzero_pd();
1422
1423         /* Start inner kernel loop */
1424         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1425         {
1426
1427             /* Get j neighbor index, and coordinate index */
1428             jnrA             = jjnr[jidx];
1429             jnrB             = jjnr[jidx+1];
1430             j_coord_offsetA  = DIM*jnrA;
1431             j_coord_offsetB  = DIM*jnrB;
1432
1433             /* load j atom coordinates */
1434             gmx_mm_load_3rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1435                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1436
1437             /* Calculate displacement vector */
1438             dx00             = _mm_sub_pd(ix0,jx0);
1439             dy00             = _mm_sub_pd(iy0,jy0);
1440             dz00             = _mm_sub_pd(iz0,jz0);
1441             dx01             = _mm_sub_pd(ix0,jx1);
1442             dy01             = _mm_sub_pd(iy0,jy1);
1443             dz01             = _mm_sub_pd(iz0,jz1);
1444             dx02             = _mm_sub_pd(ix0,jx2);
1445             dy02             = _mm_sub_pd(iy0,jy2);
1446             dz02             = _mm_sub_pd(iz0,jz2);
1447             dx10             = _mm_sub_pd(ix1,jx0);
1448             dy10             = _mm_sub_pd(iy1,jy0);
1449             dz10             = _mm_sub_pd(iz1,jz0);
1450             dx11             = _mm_sub_pd(ix1,jx1);
1451             dy11             = _mm_sub_pd(iy1,jy1);
1452             dz11             = _mm_sub_pd(iz1,jz1);
1453             dx12             = _mm_sub_pd(ix1,jx2);
1454             dy12             = _mm_sub_pd(iy1,jy2);
1455             dz12             = _mm_sub_pd(iz1,jz2);
1456             dx20             = _mm_sub_pd(ix2,jx0);
1457             dy20             = _mm_sub_pd(iy2,jy0);
1458             dz20             = _mm_sub_pd(iz2,jz0);
1459             dx21             = _mm_sub_pd(ix2,jx1);
1460             dy21             = _mm_sub_pd(iy2,jy1);
1461             dz21             = _mm_sub_pd(iz2,jz1);
1462             dx22             = _mm_sub_pd(ix2,jx2);
1463             dy22             = _mm_sub_pd(iy2,jy2);
1464             dz22             = _mm_sub_pd(iz2,jz2);
1465
1466             /* Calculate squared distance and things based on it */
1467             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1468             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1469             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1470             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1471             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1472             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1473             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1474             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1475             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1476
1477             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1478             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1479             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1480             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1481             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1482             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1483             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1484             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1485             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1486
1487             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1488             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1489             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1490             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1491             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1492             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1493             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1494             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1495             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1496
1497             fjx0             = _mm_setzero_pd();
1498             fjy0             = _mm_setzero_pd();
1499             fjz0             = _mm_setzero_pd();
1500             fjx1             = _mm_setzero_pd();
1501             fjy1             = _mm_setzero_pd();
1502             fjz1             = _mm_setzero_pd();
1503             fjx2             = _mm_setzero_pd();
1504             fjy2             = _mm_setzero_pd();
1505             fjz2             = _mm_setzero_pd();
1506
1507             /**************************
1508              * CALCULATE INTERACTIONS *
1509              **************************/
1510
1511             r00              = _mm_mul_pd(rsq00,rinv00);
1512
1513             /* Calculate table index by multiplying r with table scale and truncate to integer */
1514             rt               = _mm_mul_pd(r00,vftabscale);
1515             vfitab           = _mm_cvttpd_epi32(rt);
1516             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1517             vfitab           = _mm_slli_epi32(vfitab,3);
1518
1519             /* EWALD ELECTROSTATICS */
1520
1521             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1522             ewrt             = _mm_mul_pd(r00,ewtabscale);
1523             ewitab           = _mm_cvttpd_epi32(ewrt);
1524             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1525             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1526                                          &ewtabF,&ewtabFn);
1527             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1528             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1529
1530             /* CUBIC SPLINE TABLE DISPERSION */
1531             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1532             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1533             GMX_MM_TRANSPOSE2_PD(Y,F);
1534             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1535             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1536             GMX_MM_TRANSPOSE2_PD(G,H);
1537             Heps             = _mm_mul_pd(vfeps,H);
1538             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1539             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1540             fvdw6            = _mm_mul_pd(c6_00,FF);
1541
1542             /* CUBIC SPLINE TABLE REPULSION */
1543             vfitab           = _mm_add_epi32(vfitab,ifour);
1544             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1545             F                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) );
1546             GMX_MM_TRANSPOSE2_PD(Y,F);
1547             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1548             H                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,1) +2);
1549             GMX_MM_TRANSPOSE2_PD(G,H);
1550             Heps             = _mm_mul_pd(vfeps,H);
1551             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1552             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1553             fvdw12           = _mm_mul_pd(c12_00,FF);
1554             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1555
1556             fscal            = _mm_add_pd(felec,fvdw);
1557
1558             /* Calculate temporary vectorial force */
1559             tx               = _mm_mul_pd(fscal,dx00);
1560             ty               = _mm_mul_pd(fscal,dy00);
1561             tz               = _mm_mul_pd(fscal,dz00);
1562
1563             /* Update vectorial force */
1564             fix0             = _mm_add_pd(fix0,tx);
1565             fiy0             = _mm_add_pd(fiy0,ty);
1566             fiz0             = _mm_add_pd(fiz0,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             r01              = _mm_mul_pd(rsq01,rinv01);
1577
1578             /* EWALD ELECTROSTATICS */
1579
1580             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1581             ewrt             = _mm_mul_pd(r01,ewtabscale);
1582             ewitab           = _mm_cvttpd_epi32(ewrt);
1583             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
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(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
1588
1589             fscal            = felec;
1590
1591             /* Calculate temporary vectorial force */
1592             tx               = _mm_mul_pd(fscal,dx01);
1593             ty               = _mm_mul_pd(fscal,dy01);
1594             tz               = _mm_mul_pd(fscal,dz01);
1595
1596             /* Update vectorial force */
1597             fix0             = _mm_add_pd(fix0,tx);
1598             fiy0             = _mm_add_pd(fiy0,ty);
1599             fiz0             = _mm_add_pd(fiz0,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             r02              = _mm_mul_pd(rsq02,rinv02);
1610
1611             /* EWALD ELECTROSTATICS */
1612
1613             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1614             ewrt             = _mm_mul_pd(r02,ewtabscale);
1615             ewitab           = _mm_cvttpd_epi32(ewrt);
1616             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
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(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
1621
1622             fscal            = felec;
1623
1624             /* Calculate temporary vectorial force */
1625             tx               = _mm_mul_pd(fscal,dx02);
1626             ty               = _mm_mul_pd(fscal,dy02);
1627             tz               = _mm_mul_pd(fscal,dz02);
1628
1629             /* Update vectorial force */
1630             fix0             = _mm_add_pd(fix0,tx);
1631             fiy0             = _mm_add_pd(fiy0,ty);
1632             fiz0             = _mm_add_pd(fiz0,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             r10              = _mm_mul_pd(rsq10,rinv10);
1643
1644             /* EWALD ELECTROSTATICS */
1645
1646             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1647             ewrt             = _mm_mul_pd(r10,ewtabscale);
1648             ewitab           = _mm_cvttpd_epi32(ewrt);
1649             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
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(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
1654
1655             fscal            = felec;
1656
1657             /* Calculate temporary vectorial force */
1658             tx               = _mm_mul_pd(fscal,dx10);
1659             ty               = _mm_mul_pd(fscal,dy10);
1660             tz               = _mm_mul_pd(fscal,dz10);
1661
1662             /* Update vectorial force */
1663             fix1             = _mm_add_pd(fix1,tx);
1664             fiy1             = _mm_add_pd(fiy1,ty);
1665             fiz1             = _mm_add_pd(fiz1,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             r11              = _mm_mul_pd(rsq11,rinv11);
1676
1677             /* EWALD ELECTROSTATICS */
1678
1679             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1680             ewrt             = _mm_mul_pd(r11,ewtabscale);
1681             ewitab           = _mm_cvttpd_epi32(ewrt);
1682             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
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(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1687
1688             fscal            = felec;
1689
1690             /* Calculate temporary vectorial force */
1691             tx               = _mm_mul_pd(fscal,dx11);
1692             ty               = _mm_mul_pd(fscal,dy11);
1693             tz               = _mm_mul_pd(fscal,dz11);
1694
1695             /* Update vectorial force */
1696             fix1             = _mm_add_pd(fix1,tx);
1697             fiy1             = _mm_add_pd(fiy1,ty);
1698             fiz1             = _mm_add_pd(fiz1,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             r12              = _mm_mul_pd(rsq12,rinv12);
1709
1710             /* EWALD ELECTROSTATICS */
1711
1712             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1713             ewrt             = _mm_mul_pd(r12,ewtabscale);
1714             ewitab           = _mm_cvttpd_epi32(ewrt);
1715             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
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(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1720
1721             fscal            = felec;
1722
1723             /* Calculate temporary vectorial force */
1724             tx               = _mm_mul_pd(fscal,dx12);
1725             ty               = _mm_mul_pd(fscal,dy12);
1726             tz               = _mm_mul_pd(fscal,dz12);
1727
1728             /* Update vectorial force */
1729             fix1             = _mm_add_pd(fix1,tx);
1730             fiy1             = _mm_add_pd(fiy1,ty);
1731             fiz1             = _mm_add_pd(fiz1,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             /**************************
1738              * CALCULATE INTERACTIONS *
1739              **************************/
1740
1741             r20              = _mm_mul_pd(rsq20,rinv20);
1742
1743             /* EWALD ELECTROSTATICS */
1744
1745             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1746             ewrt             = _mm_mul_pd(r20,ewtabscale);
1747             ewitab           = _mm_cvttpd_epi32(ewrt);
1748             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1749             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1750                                          &ewtabF,&ewtabFn);
1751             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1752             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
1753
1754             fscal            = felec;
1755
1756             /* Calculate temporary vectorial force */
1757             tx               = _mm_mul_pd(fscal,dx20);
1758             ty               = _mm_mul_pd(fscal,dy20);
1759             tz               = _mm_mul_pd(fscal,dz20);
1760
1761             /* Update vectorial force */
1762             fix2             = _mm_add_pd(fix2,tx);
1763             fiy2             = _mm_add_pd(fiy2,ty);
1764             fiz2             = _mm_add_pd(fiz2,tz);
1765
1766             fjx0             = _mm_add_pd(fjx0,tx);
1767             fjy0             = _mm_add_pd(fjy0,ty);
1768             fjz0             = _mm_add_pd(fjz0,tz);
1769
1770             /**************************
1771              * CALCULATE INTERACTIONS *
1772              **************************/
1773
1774             r21              = _mm_mul_pd(rsq21,rinv21);
1775
1776             /* EWALD ELECTROSTATICS */
1777
1778             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1779             ewrt             = _mm_mul_pd(r21,ewtabscale);
1780             ewitab           = _mm_cvttpd_epi32(ewrt);
1781             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1782             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1783                                          &ewtabF,&ewtabFn);
1784             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1785             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1786
1787             fscal            = felec;
1788
1789             /* Calculate temporary vectorial force */
1790             tx               = _mm_mul_pd(fscal,dx21);
1791             ty               = _mm_mul_pd(fscal,dy21);
1792             tz               = _mm_mul_pd(fscal,dz21);
1793
1794             /* Update vectorial force */
1795             fix2             = _mm_add_pd(fix2,tx);
1796             fiy2             = _mm_add_pd(fiy2,ty);
1797             fiz2             = _mm_add_pd(fiz2,tz);
1798
1799             fjx1             = _mm_add_pd(fjx1,tx);
1800             fjy1             = _mm_add_pd(fjy1,ty);
1801             fjz1             = _mm_add_pd(fjz1,tz);
1802
1803             /**************************
1804              * CALCULATE INTERACTIONS *
1805              **************************/
1806
1807             r22              = _mm_mul_pd(rsq22,rinv22);
1808
1809             /* EWALD ELECTROSTATICS */
1810
1811             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1812             ewrt             = _mm_mul_pd(r22,ewtabscale);
1813             ewitab           = _mm_cvttpd_epi32(ewrt);
1814             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1815             gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1816                                          &ewtabF,&ewtabFn);
1817             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1818             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1819
1820             fscal            = felec;
1821
1822             /* Calculate temporary vectorial force */
1823             tx               = _mm_mul_pd(fscal,dx22);
1824             ty               = _mm_mul_pd(fscal,dy22);
1825             tz               = _mm_mul_pd(fscal,dz22);
1826
1827             /* Update vectorial force */
1828             fix2             = _mm_add_pd(fix2,tx);
1829             fiy2             = _mm_add_pd(fiy2,ty);
1830             fiz2             = _mm_add_pd(fiz2,tz);
1831
1832             fjx2             = _mm_add_pd(fjx2,tx);
1833             fjy2             = _mm_add_pd(fjy2,ty);
1834             fjz2             = _mm_add_pd(fjz2,tz);
1835
1836             gmx_mm_decrement_3rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
1837
1838             /* Inner loop uses 350 flops */
1839         }
1840
1841         if(jidx<j_index_end)
1842         {
1843
1844             jnrA             = jjnr[jidx];
1845             j_coord_offsetA  = DIM*jnrA;
1846
1847             /* load j atom coordinates */
1848             gmx_mm_load_3rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1849                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,&jy2,&jz2);
1850
1851             /* Calculate displacement vector */
1852             dx00             = _mm_sub_pd(ix0,jx0);
1853             dy00             = _mm_sub_pd(iy0,jy0);
1854             dz00             = _mm_sub_pd(iz0,jz0);
1855             dx01             = _mm_sub_pd(ix0,jx1);
1856             dy01             = _mm_sub_pd(iy0,jy1);
1857             dz01             = _mm_sub_pd(iz0,jz1);
1858             dx02             = _mm_sub_pd(ix0,jx2);
1859             dy02             = _mm_sub_pd(iy0,jy2);
1860             dz02             = _mm_sub_pd(iz0,jz2);
1861             dx10             = _mm_sub_pd(ix1,jx0);
1862             dy10             = _mm_sub_pd(iy1,jy0);
1863             dz10             = _mm_sub_pd(iz1,jz0);
1864             dx11             = _mm_sub_pd(ix1,jx1);
1865             dy11             = _mm_sub_pd(iy1,jy1);
1866             dz11             = _mm_sub_pd(iz1,jz1);
1867             dx12             = _mm_sub_pd(ix1,jx2);
1868             dy12             = _mm_sub_pd(iy1,jy2);
1869             dz12             = _mm_sub_pd(iz1,jz2);
1870             dx20             = _mm_sub_pd(ix2,jx0);
1871             dy20             = _mm_sub_pd(iy2,jy0);
1872             dz20             = _mm_sub_pd(iz2,jz0);
1873             dx21             = _mm_sub_pd(ix2,jx1);
1874             dy21             = _mm_sub_pd(iy2,jy1);
1875             dz21             = _mm_sub_pd(iz2,jz1);
1876             dx22             = _mm_sub_pd(ix2,jx2);
1877             dy22             = _mm_sub_pd(iy2,jy2);
1878             dz22             = _mm_sub_pd(iz2,jz2);
1879
1880             /* Calculate squared distance and things based on it */
1881             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1882             rsq01            = gmx_mm_calc_rsq_pd(dx01,dy01,dz01);
1883             rsq02            = gmx_mm_calc_rsq_pd(dx02,dy02,dz02);
1884             rsq10            = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
1885             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1886             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1887             rsq20            = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
1888             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1889             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1890
1891             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1892             rinv01           = gmx_mm_invsqrt_pd(rsq01);
1893             rinv02           = gmx_mm_invsqrt_pd(rsq02);
1894             rinv10           = gmx_mm_invsqrt_pd(rsq10);
1895             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1896             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1897             rinv20           = gmx_mm_invsqrt_pd(rsq20);
1898             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1899             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1900
1901             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
1902             rinvsq01         = _mm_mul_pd(rinv01,rinv01);
1903             rinvsq02         = _mm_mul_pd(rinv02,rinv02);
1904             rinvsq10         = _mm_mul_pd(rinv10,rinv10);
1905             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1906             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1907             rinvsq20         = _mm_mul_pd(rinv20,rinv20);
1908             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1909             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1910
1911             fjx0             = _mm_setzero_pd();
1912             fjy0             = _mm_setzero_pd();
1913             fjz0             = _mm_setzero_pd();
1914             fjx1             = _mm_setzero_pd();
1915             fjy1             = _mm_setzero_pd();
1916             fjz1             = _mm_setzero_pd();
1917             fjx2             = _mm_setzero_pd();
1918             fjy2             = _mm_setzero_pd();
1919             fjz2             = _mm_setzero_pd();
1920
1921             /**************************
1922              * CALCULATE INTERACTIONS *
1923              **************************/
1924
1925             r00              = _mm_mul_pd(rsq00,rinv00);
1926
1927             /* Calculate table index by multiplying r with table scale and truncate to integer */
1928             rt               = _mm_mul_pd(r00,vftabscale);
1929             vfitab           = _mm_cvttpd_epi32(rt);
1930             vfeps            = _mm_sub_pd(rt,_mm_cvtepi32_pd(vfitab));
1931             vfitab           = _mm_slli_epi32(vfitab,3);
1932
1933             /* EWALD ELECTROSTATICS */
1934
1935             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1936             ewrt             = _mm_mul_pd(r00,ewtabscale);
1937             ewitab           = _mm_cvttpd_epi32(ewrt);
1938             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1939             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
1940             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1941             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
1942
1943             /* CUBIC SPLINE TABLE DISPERSION */
1944             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1945             F                = _mm_setzero_pd();
1946             GMX_MM_TRANSPOSE2_PD(Y,F);
1947             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1948             H                = _mm_setzero_pd();
1949             GMX_MM_TRANSPOSE2_PD(G,H);
1950             Heps             = _mm_mul_pd(vfeps,H);
1951             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1952             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1953             fvdw6            = _mm_mul_pd(c6_00,FF);
1954
1955             /* CUBIC SPLINE TABLE REPULSION */
1956             vfitab           = _mm_add_epi32(vfitab,ifour);
1957             Y                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) );
1958             F                = _mm_setzero_pd();
1959             GMX_MM_TRANSPOSE2_PD(Y,F);
1960             G                = _mm_load_pd( vftab + gmx_mm_extract_epi32(vfitab,0) +2);
1961             H                = _mm_setzero_pd();
1962             GMX_MM_TRANSPOSE2_PD(G,H);
1963             Heps             = _mm_mul_pd(vfeps,H);
1964             Fp               = _mm_add_pd(F,_mm_mul_pd(vfeps,_mm_add_pd(G,Heps)));
1965             FF               = _mm_add_pd(Fp,_mm_mul_pd(vfeps,_mm_add_pd(G,_mm_add_pd(Heps,Heps))));
1966             fvdw12           = _mm_mul_pd(c12_00,FF);
1967             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1968
1969             fscal            = _mm_add_pd(felec,fvdw);
1970
1971             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1972
1973             /* Calculate temporary vectorial force */
1974             tx               = _mm_mul_pd(fscal,dx00);
1975             ty               = _mm_mul_pd(fscal,dy00);
1976             tz               = _mm_mul_pd(fscal,dz00);
1977
1978             /* Update vectorial force */
1979             fix0             = _mm_add_pd(fix0,tx);
1980             fiy0             = _mm_add_pd(fiy0,ty);
1981             fiz0             = _mm_add_pd(fiz0,tz);
1982
1983             fjx0             = _mm_add_pd(fjx0,tx);
1984             fjy0             = _mm_add_pd(fjy0,ty);
1985             fjz0             = _mm_add_pd(fjz0,tz);
1986
1987             /**************************
1988              * CALCULATE INTERACTIONS *
1989              **************************/
1990
1991             r01              = _mm_mul_pd(rsq01,rinv01);
1992
1993             /* EWALD ELECTROSTATICS */
1994
1995             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1996             ewrt             = _mm_mul_pd(r01,ewtabscale);
1997             ewitab           = _mm_cvttpd_epi32(ewrt);
1998             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1999             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2000             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2001             felec            = _mm_mul_pd(_mm_mul_pd(qq01,rinv01),_mm_sub_pd(rinvsq01,felec));
2002
2003             fscal            = felec;
2004
2005             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2006
2007             /* Calculate temporary vectorial force */
2008             tx               = _mm_mul_pd(fscal,dx01);
2009             ty               = _mm_mul_pd(fscal,dy01);
2010             tz               = _mm_mul_pd(fscal,dz01);
2011
2012             /* Update vectorial force */
2013             fix0             = _mm_add_pd(fix0,tx);
2014             fiy0             = _mm_add_pd(fiy0,ty);
2015             fiz0             = _mm_add_pd(fiz0,tz);
2016
2017             fjx1             = _mm_add_pd(fjx1,tx);
2018             fjy1             = _mm_add_pd(fjy1,ty);
2019             fjz1             = _mm_add_pd(fjz1,tz);
2020
2021             /**************************
2022              * CALCULATE INTERACTIONS *
2023              **************************/
2024
2025             r02              = _mm_mul_pd(rsq02,rinv02);
2026
2027             /* EWALD ELECTROSTATICS */
2028
2029             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2030             ewrt             = _mm_mul_pd(r02,ewtabscale);
2031             ewitab           = _mm_cvttpd_epi32(ewrt);
2032             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2033             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2034             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2035             felec            = _mm_mul_pd(_mm_mul_pd(qq02,rinv02),_mm_sub_pd(rinvsq02,felec));
2036
2037             fscal            = felec;
2038
2039             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2040
2041             /* Calculate temporary vectorial force */
2042             tx               = _mm_mul_pd(fscal,dx02);
2043             ty               = _mm_mul_pd(fscal,dy02);
2044             tz               = _mm_mul_pd(fscal,dz02);
2045
2046             /* Update vectorial force */
2047             fix0             = _mm_add_pd(fix0,tx);
2048             fiy0             = _mm_add_pd(fiy0,ty);
2049             fiz0             = _mm_add_pd(fiz0,tz);
2050
2051             fjx2             = _mm_add_pd(fjx2,tx);
2052             fjy2             = _mm_add_pd(fjy2,ty);
2053             fjz2             = _mm_add_pd(fjz2,tz);
2054
2055             /**************************
2056              * CALCULATE INTERACTIONS *
2057              **************************/
2058
2059             r10              = _mm_mul_pd(rsq10,rinv10);
2060
2061             /* EWALD ELECTROSTATICS */
2062
2063             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2064             ewrt             = _mm_mul_pd(r10,ewtabscale);
2065             ewitab           = _mm_cvttpd_epi32(ewrt);
2066             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2067             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2068             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2069             felec            = _mm_mul_pd(_mm_mul_pd(qq10,rinv10),_mm_sub_pd(rinvsq10,felec));
2070
2071             fscal            = felec;
2072
2073             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2074
2075             /* Calculate temporary vectorial force */
2076             tx               = _mm_mul_pd(fscal,dx10);
2077             ty               = _mm_mul_pd(fscal,dy10);
2078             tz               = _mm_mul_pd(fscal,dz10);
2079
2080             /* Update vectorial force */
2081             fix1             = _mm_add_pd(fix1,tx);
2082             fiy1             = _mm_add_pd(fiy1,ty);
2083             fiz1             = _mm_add_pd(fiz1,tz);
2084
2085             fjx0             = _mm_add_pd(fjx0,tx);
2086             fjy0             = _mm_add_pd(fjy0,ty);
2087             fjz0             = _mm_add_pd(fjz0,tz);
2088
2089             /**************************
2090              * CALCULATE INTERACTIONS *
2091              **************************/
2092
2093             r11              = _mm_mul_pd(rsq11,rinv11);
2094
2095             /* EWALD ELECTROSTATICS */
2096
2097             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2098             ewrt             = _mm_mul_pd(r11,ewtabscale);
2099             ewitab           = _mm_cvttpd_epi32(ewrt);
2100             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2101             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2102             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2103             felec            = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2104
2105             fscal            = felec;
2106
2107             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2108
2109             /* Calculate temporary vectorial force */
2110             tx               = _mm_mul_pd(fscal,dx11);
2111             ty               = _mm_mul_pd(fscal,dy11);
2112             tz               = _mm_mul_pd(fscal,dz11);
2113
2114             /* Update vectorial force */
2115             fix1             = _mm_add_pd(fix1,tx);
2116             fiy1             = _mm_add_pd(fiy1,ty);
2117             fiz1             = _mm_add_pd(fiz1,tz);
2118
2119             fjx1             = _mm_add_pd(fjx1,tx);
2120             fjy1             = _mm_add_pd(fjy1,ty);
2121             fjz1             = _mm_add_pd(fjz1,tz);
2122
2123             /**************************
2124              * CALCULATE INTERACTIONS *
2125              **************************/
2126
2127             r12              = _mm_mul_pd(rsq12,rinv12);
2128
2129             /* EWALD ELECTROSTATICS */
2130
2131             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2132             ewrt             = _mm_mul_pd(r12,ewtabscale);
2133             ewitab           = _mm_cvttpd_epi32(ewrt);
2134             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2135             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2136             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2137             felec            = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2138
2139             fscal            = felec;
2140
2141             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2142
2143             /* Calculate temporary vectorial force */
2144             tx               = _mm_mul_pd(fscal,dx12);
2145             ty               = _mm_mul_pd(fscal,dy12);
2146             tz               = _mm_mul_pd(fscal,dz12);
2147
2148             /* Update vectorial force */
2149             fix1             = _mm_add_pd(fix1,tx);
2150             fiy1             = _mm_add_pd(fiy1,ty);
2151             fiz1             = _mm_add_pd(fiz1,tz);
2152
2153             fjx2             = _mm_add_pd(fjx2,tx);
2154             fjy2             = _mm_add_pd(fjy2,ty);
2155             fjz2             = _mm_add_pd(fjz2,tz);
2156
2157             /**************************
2158              * CALCULATE INTERACTIONS *
2159              **************************/
2160
2161             r20              = _mm_mul_pd(rsq20,rinv20);
2162
2163             /* EWALD ELECTROSTATICS */
2164
2165             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2166             ewrt             = _mm_mul_pd(r20,ewtabscale);
2167             ewitab           = _mm_cvttpd_epi32(ewrt);
2168             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2169             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2170             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2171             felec            = _mm_mul_pd(_mm_mul_pd(qq20,rinv20),_mm_sub_pd(rinvsq20,felec));
2172
2173             fscal            = felec;
2174
2175             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2176
2177             /* Calculate temporary vectorial force */
2178             tx               = _mm_mul_pd(fscal,dx20);
2179             ty               = _mm_mul_pd(fscal,dy20);
2180             tz               = _mm_mul_pd(fscal,dz20);
2181
2182             /* Update vectorial force */
2183             fix2             = _mm_add_pd(fix2,tx);
2184             fiy2             = _mm_add_pd(fiy2,ty);
2185             fiz2             = _mm_add_pd(fiz2,tz);
2186
2187             fjx0             = _mm_add_pd(fjx0,tx);
2188             fjy0             = _mm_add_pd(fjy0,ty);
2189             fjz0             = _mm_add_pd(fjz0,tz);
2190
2191             /**************************
2192              * CALCULATE INTERACTIONS *
2193              **************************/
2194
2195             r21              = _mm_mul_pd(rsq21,rinv21);
2196
2197             /* EWALD ELECTROSTATICS */
2198
2199             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2200             ewrt             = _mm_mul_pd(r21,ewtabscale);
2201             ewitab           = _mm_cvttpd_epi32(ewrt);
2202             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2203             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2204             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2205             felec            = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2206
2207             fscal            = felec;
2208
2209             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2210
2211             /* Calculate temporary vectorial force */
2212             tx               = _mm_mul_pd(fscal,dx21);
2213             ty               = _mm_mul_pd(fscal,dy21);
2214             tz               = _mm_mul_pd(fscal,dz21);
2215
2216             /* Update vectorial force */
2217             fix2             = _mm_add_pd(fix2,tx);
2218             fiy2             = _mm_add_pd(fiy2,ty);
2219             fiz2             = _mm_add_pd(fiz2,tz);
2220
2221             fjx1             = _mm_add_pd(fjx1,tx);
2222             fjy1             = _mm_add_pd(fjy1,ty);
2223             fjz1             = _mm_add_pd(fjz1,tz);
2224
2225             /**************************
2226              * CALCULATE INTERACTIONS *
2227              **************************/
2228
2229             r22              = _mm_mul_pd(rsq22,rinv22);
2230
2231             /* EWALD ELECTROSTATICS */
2232
2233             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2234             ewrt             = _mm_mul_pd(r22,ewtabscale);
2235             ewitab           = _mm_cvttpd_epi32(ewrt);
2236             eweps            = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2237             gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2238             felec            = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2239             felec            = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2240
2241             fscal            = felec;
2242
2243             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2244
2245             /* Calculate temporary vectorial force */
2246             tx               = _mm_mul_pd(fscal,dx22);
2247             ty               = _mm_mul_pd(fscal,dy22);
2248             tz               = _mm_mul_pd(fscal,dz22);
2249
2250             /* Update vectorial force */
2251             fix2             = _mm_add_pd(fix2,tx);
2252             fiy2             = _mm_add_pd(fiy2,ty);
2253             fiz2             = _mm_add_pd(fiz2,tz);
2254
2255             fjx2             = _mm_add_pd(fjx2,tx);
2256             fjy2             = _mm_add_pd(fjy2,ty);
2257             fjz2             = _mm_add_pd(fjz2,tz);
2258
2259             gmx_mm_decrement_3rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2);
2260
2261             /* Inner loop uses 350 flops */
2262         }
2263
2264         /* End of innermost loop */
2265
2266         gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
2267                                               f+i_coord_offset,fshift+i_shift_offset);
2268
2269         /* Increment number of inner iterations */
2270         inneriter                  += j_index_end - j_index_start;
2271
2272         /* Outer loop uses 18 flops */
2273     }
2274
2275     /* Increment number of outer iterations */
2276     outeriter        += nri;
2277
2278     /* Update outer/inner flops */
2279
2280     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3W3_F,outeriter*18 + inneriter*350);
2281 }