src/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_avx_128_fma_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_128_fma_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_avx_128_fma_double.h"
  34 #include "kernelutil_x86_avx_128_fma_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_avx_128_fma_double
  38  * Electrostatics interaction: CubicSplineTable
  39  * VdW interaction:            CubicSplineTable
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_avx_128_fma_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              vdwioffset3;
  73     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  74     int              vdwjidx0A,vdwjidx0B;
  75     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  76     int              vdwjidx1A,vdwjidx1B;
  77     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  78     int              vdwjidx2A,vdwjidx2B;
  79     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  80     int              vdwjidx3A,vdwjidx3B;
  81     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  82     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  83     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  84     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  85     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  86     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  87     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  88     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  89     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  90     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  91     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  92     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  99     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 100     __m128i          vfitab;
 101     __m128i          ifour       = _mm_set1_epi32(4);
 102     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 103     real             *vftab;
 104     __m128d          dummy_mask,cutoff_mask;
 105     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 106     __m128d          one     = _mm_set1_pd(1.0);
 107     __m128d          two     = _mm_set1_pd(2.0);
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = _mm_set1_pd(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     vftab            = kernel_data->table_elec_vdw->data;
 126     vftabscale       = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 131     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 132     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     jq1              = _mm_set1_pd(charge[inr+1]);
 136     jq2              = _mm_set1_pd(charge[inr+2]);
 137     jq3              = _mm_set1_pd(charge[inr+3]);
 138     vdwjidx0A        = 2*vdwtype[inr+0];
 139     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 140     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 141     qq11             = _mm_mul_pd(iq1,jq1);
 142     qq12             = _mm_mul_pd(iq1,jq2);
 143     qq13             = _mm_mul_pd(iq1,jq3);
 144     qq21             = _mm_mul_pd(iq2,jq1);
 145     qq22             = _mm_mul_pd(iq2,jq2);
 146     qq23             = _mm_mul_pd(iq2,jq3);
 147     qq31             = _mm_mul_pd(iq3,jq1);
 148     qq32             = _mm_mul_pd(iq3,jq2);
 149     qq33             = _mm_mul_pd(iq3,jq3);
 150
 151     /* Avoid stupid compiler warnings */
 152     jnrA = jnrB = 0;
 153     j_coord_offsetA = 0;
 154     j_coord_offsetB = 0;
 155
 156     outeriter        = 0;
 157     inneriter        = 0;
 158
 159     /* Start outer loop over neighborlists */
 160     for(iidx=0; iidx<nri; iidx++)
 161     {
 162         /* Load shift vector for this list */
 163         i_shift_offset   = DIM*shiftidx[iidx];
 164
 165         /* Load limits for loop over neighbors */
 166         j_index_start    = jindex[iidx];
 167         j_index_end      = jindex[iidx+1];
 168
 169         /* Get outer coordinate index */
 170         inr              = iinr[iidx];
 171         i_coord_offset   = DIM*inr;
 172
 173         /* Load i particle coords and add shift vector */
 174         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 175                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 176
 177         fix0             = _mm_setzero_pd();
 178         fiy0             = _mm_setzero_pd();
 179         fiz0             = _mm_setzero_pd();
 180         fix1             = _mm_setzero_pd();
 181         fiy1             = _mm_setzero_pd();
 182         fiz1             = _mm_setzero_pd();
 183         fix2             = _mm_setzero_pd();
 184         fiy2             = _mm_setzero_pd();
 185         fiz2             = _mm_setzero_pd();
 186         fix3             = _mm_setzero_pd();
 187         fiy3             = _mm_setzero_pd();
 188         fiz3             = _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_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 206                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 207                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 208
 209             /* Calculate displacement vector */
 210             dx00             = _mm_sub_pd(ix0,jx0);
 211             dy00             = _mm_sub_pd(iy0,jy0);
 212             dz00             = _mm_sub_pd(iz0,jz0);
 213             dx11             = _mm_sub_pd(ix1,jx1);
 214             dy11             = _mm_sub_pd(iy1,jy1);
 215             dz11             = _mm_sub_pd(iz1,jz1);
 216             dx12             = _mm_sub_pd(ix1,jx2);
 217             dy12             = _mm_sub_pd(iy1,jy2);
 218             dz12             = _mm_sub_pd(iz1,jz2);
 219             dx13             = _mm_sub_pd(ix1,jx3);
 220             dy13             = _mm_sub_pd(iy1,jy3);
 221             dz13             = _mm_sub_pd(iz1,jz3);
 222             dx21             = _mm_sub_pd(ix2,jx1);
 223             dy21             = _mm_sub_pd(iy2,jy1);
 224             dz21             = _mm_sub_pd(iz2,jz1);
 225             dx22             = _mm_sub_pd(ix2,jx2);
 226             dy22             = _mm_sub_pd(iy2,jy2);
 227             dz22             = _mm_sub_pd(iz2,jz2);
 228             dx23             = _mm_sub_pd(ix2,jx3);
 229             dy23             = _mm_sub_pd(iy2,jy3);
 230             dz23             = _mm_sub_pd(iz2,jz3);
 231             dx31             = _mm_sub_pd(ix3,jx1);
 232             dy31             = _mm_sub_pd(iy3,jy1);
 233             dz31             = _mm_sub_pd(iz3,jz1);
 234             dx32             = _mm_sub_pd(ix3,jx2);
 235             dy32             = _mm_sub_pd(iy3,jy2);
 236             dz32             = _mm_sub_pd(iz3,jz2);
 237             dx33             = _mm_sub_pd(ix3,jx3);
 238             dy33             = _mm_sub_pd(iy3,jy3);
 239             dz33             = _mm_sub_pd(iz3,jz3);
 240
 241             /* Calculate squared distance and things based on it */
 242             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 243             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 244             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 245             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 246             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 247             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 248             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 249             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 250             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 251             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 252
 253             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 254             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 255             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 256             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 257             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 258             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 259             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 260             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 261             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 262             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 263
 264             fjx0             = _mm_setzero_pd();
 265             fjy0             = _mm_setzero_pd();
 266             fjz0             = _mm_setzero_pd();
 267             fjx1             = _mm_setzero_pd();
 268             fjy1             = _mm_setzero_pd();
 269             fjz1             = _mm_setzero_pd();
 270             fjx2             = _mm_setzero_pd();
 271             fjy2             = _mm_setzero_pd();
 272             fjz2             = _mm_setzero_pd();
 273             fjx3             = _mm_setzero_pd();
 274             fjy3             = _mm_setzero_pd();
 275             fjz3             = _mm_setzero_pd();
 276
 277             /**************************
 278              * CALCULATE INTERACTIONS *
 279              **************************/
 280
 281             r00              = _mm_mul_pd(rsq00,rinv00);
 282
 283             /* Calculate table index by multiplying r with table scale and truncate to integer */
 284             rt               = _mm_mul_pd(r00,vftabscale);
 285             vfitab           = _mm_cvttpd_epi32(rt);
 286 #ifdef __XOP__
 287             vfeps            = _mm_frcz_pd(rt);
 288 #else
 289             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 290 #endif
 291             twovfeps         = _mm_add_pd(vfeps,vfeps);
 292             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 293
 294             /* CUBIC SPLINE TABLE DISPERSION */
 295             vfitab           = _mm_add_epi32(vfitab,ifour);
 296             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 297             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 298             GMX_MM_TRANSPOSE2_PD(Y,F);
 299             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 300             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 301             GMX_MM_TRANSPOSE2_PD(G,H);
 302             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 303             VV               = _mm_macc_pd(vfeps,Fp,Y);
 304             vvdw6            = _mm_mul_pd(c6_00,VV);
 305             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 306             fvdw6            = _mm_mul_pd(c6_00,FF);
 307
 308             /* CUBIC SPLINE TABLE REPULSION */
 309             vfitab           = _mm_add_epi32(vfitab,ifour);
 310             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 311             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 312             GMX_MM_TRANSPOSE2_PD(Y,F);
 313             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 314             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 315             GMX_MM_TRANSPOSE2_PD(G,H);
 316             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 317             VV               = _mm_macc_pd(vfeps,Fp,Y);
 318             vvdw12           = _mm_mul_pd(c12_00,VV);
 319             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 320             fvdw12           = _mm_mul_pd(c12_00,FF);
 321             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 322             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 323
 324             /* Update potential sum for this i atom from the interaction with this j atom. */
 325             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 326
 327             fscal            = fvdw;
 328
 329             /* Update vectorial force */
 330             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 331             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 332             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 333
 334             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 335             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 336             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 337
 338             /**************************
 339              * CALCULATE INTERACTIONS *
 340              **************************/
 341
 342             r11              = _mm_mul_pd(rsq11,rinv11);
 343
 344             /* Calculate table index by multiplying r with table scale and truncate to integer */
 345             rt               = _mm_mul_pd(r11,vftabscale);
 346             vfitab           = _mm_cvttpd_epi32(rt);
 347 #ifdef __XOP__
 348             vfeps            = _mm_frcz_pd(rt);
 349 #else
 350             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 351 #endif
 352             twovfeps         = _mm_add_pd(vfeps,vfeps);
 353             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 354
 355             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 356             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 357             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 358             GMX_MM_TRANSPOSE2_PD(Y,F);
 359             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 360             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 361             GMX_MM_TRANSPOSE2_PD(G,H);
 362             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 363             VV               = _mm_macc_pd(vfeps,Fp,Y);
 364             velec            = _mm_mul_pd(qq11,VV);
 365             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 366             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 367
 368             /* Update potential sum for this i atom from the interaction with this j atom. */
 369             velecsum         = _mm_add_pd(velecsum,velec);
 370
 371             fscal            = felec;
 372
 373             /* Update vectorial force */
 374             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 375             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 376             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 377
 378             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 379             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 380             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 381
 382             /**************************
 383              * CALCULATE INTERACTIONS *
 384              **************************/
 385
 386             r12              = _mm_mul_pd(rsq12,rinv12);
 387
 388             /* Calculate table index by multiplying r with table scale and truncate to integer */
 389             rt               = _mm_mul_pd(r12,vftabscale);
 390             vfitab           = _mm_cvttpd_epi32(rt);
 391 #ifdef __XOP__
 392             vfeps            = _mm_frcz_pd(rt);
 393 #else
 394             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 395 #endif
 396             twovfeps         = _mm_add_pd(vfeps,vfeps);
 397             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 398
 399             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 400             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 401             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 402             GMX_MM_TRANSPOSE2_PD(Y,F);
 403             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 404             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 405             GMX_MM_TRANSPOSE2_PD(G,H);
 406             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 407             VV               = _mm_macc_pd(vfeps,Fp,Y);
 408             velec            = _mm_mul_pd(qq12,VV);
 409             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 410             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 411
 412             /* Update potential sum for this i atom from the interaction with this j atom. */
 413             velecsum         = _mm_add_pd(velecsum,velec);
 414
 415             fscal            = felec;
 416
 417             /* Update vectorial force */
 418             fix1             = _mm_macc_pd(dx12,fscal,fix1);
 419             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
 420             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
 421
 422             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
 423             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
 424             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
 425
 426             /**************************
 427              * CALCULATE INTERACTIONS *
 428              **************************/
 429
 430             r13              = _mm_mul_pd(rsq13,rinv13);
 431
 432             /* Calculate table index by multiplying r with table scale and truncate to integer */
 433             rt               = _mm_mul_pd(r13,vftabscale);
 434             vfitab           = _mm_cvttpd_epi32(rt);
 435 #ifdef __XOP__
 436             vfeps            = _mm_frcz_pd(rt);
 437 #else
 438             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 439 #endif
 440             twovfeps         = _mm_add_pd(vfeps,vfeps);
 441             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 442
 443             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 444             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 445             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 446             GMX_MM_TRANSPOSE2_PD(Y,F);
 447             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 448             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 449             GMX_MM_TRANSPOSE2_PD(G,H);
 450             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 451             VV               = _mm_macc_pd(vfeps,Fp,Y);
 452             velec            = _mm_mul_pd(qq13,VV);
 453             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 454             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
 455
 456             /* Update potential sum for this i atom from the interaction with this j atom. */
 457             velecsum         = _mm_add_pd(velecsum,velec);
 458
 459             fscal            = felec;
 460
 461             /* Update vectorial force */
 462             fix1             = _mm_macc_pd(dx13,fscal,fix1);
 463             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
 464             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
 465
 466             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
 467             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
 468             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
 469
 470             /**************************
 471              * CALCULATE INTERACTIONS *
 472              **************************/
 473
 474             r21              = _mm_mul_pd(rsq21,rinv21);
 475
 476             /* Calculate table index by multiplying r with table scale and truncate to integer */
 477             rt               = _mm_mul_pd(r21,vftabscale);
 478             vfitab           = _mm_cvttpd_epi32(rt);
 479 #ifdef __XOP__
 480             vfeps            = _mm_frcz_pd(rt);
 481 #else
 482             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 483 #endif
 484             twovfeps         = _mm_add_pd(vfeps,vfeps);
 485             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 486
 487             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 488             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 489             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 490             GMX_MM_TRANSPOSE2_PD(Y,F);
 491             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 492             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 493             GMX_MM_TRANSPOSE2_PD(G,H);
 494             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 495             VV               = _mm_macc_pd(vfeps,Fp,Y);
 496             velec            = _mm_mul_pd(qq21,VV);
 497             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 498             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
 499
 500             /* Update potential sum for this i atom from the interaction with this j atom. */
 501             velecsum         = _mm_add_pd(velecsum,velec);
 502
 503             fscal            = felec;
 504
 505             /* Update vectorial force */
 506             fix2             = _mm_macc_pd(dx21,fscal,fix2);
 507             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
 508             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
 509
 510             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
 511             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
 512             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
 513
 514             /**************************
 515              * CALCULATE INTERACTIONS *
 516              **************************/
 517
 518             r22              = _mm_mul_pd(rsq22,rinv22);
 519
 520             /* Calculate table index by multiplying r with table scale and truncate to integer */
 521             rt               = _mm_mul_pd(r22,vftabscale);
 522             vfitab           = _mm_cvttpd_epi32(rt);
 523 #ifdef __XOP__
 524             vfeps            = _mm_frcz_pd(rt);
 525 #else
 526             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 527 #endif
 528             twovfeps         = _mm_add_pd(vfeps,vfeps);
 529             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 530
 531             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 532             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 533             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 534             GMX_MM_TRANSPOSE2_PD(Y,F);
 535             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 536             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 537             GMX_MM_TRANSPOSE2_PD(G,H);
 538             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 539             VV               = _mm_macc_pd(vfeps,Fp,Y);
 540             velec            = _mm_mul_pd(qq22,VV);
 541             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 542             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
 543
 544             /* Update potential sum for this i atom from the interaction with this j atom. */
 545             velecsum         = _mm_add_pd(velecsum,velec);
 546
 547             fscal            = felec;
 548
 549             /* Update vectorial force */
 550             fix2             = _mm_macc_pd(dx22,fscal,fix2);
 551             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
 552             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
 553
 554             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
 555             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
 556             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
 557
 558             /**************************
 559              * CALCULATE INTERACTIONS *
 560              **************************/
 561
 562             r23              = _mm_mul_pd(rsq23,rinv23);
 563
 564             /* Calculate table index by multiplying r with table scale and truncate to integer */
 565             rt               = _mm_mul_pd(r23,vftabscale);
 566             vfitab           = _mm_cvttpd_epi32(rt);
 567 #ifdef __XOP__
 568             vfeps            = _mm_frcz_pd(rt);
 569 #else
 570             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 571 #endif
 572             twovfeps         = _mm_add_pd(vfeps,vfeps);
 573             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 574
 575             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 576             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 577             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 578             GMX_MM_TRANSPOSE2_PD(Y,F);
 579             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 580             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 581             GMX_MM_TRANSPOSE2_PD(G,H);
 582             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 583             VV               = _mm_macc_pd(vfeps,Fp,Y);
 584             velec            = _mm_mul_pd(qq23,VV);
 585             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 586             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
 587
 588             /* Update potential sum for this i atom from the interaction with this j atom. */
 589             velecsum         = _mm_add_pd(velecsum,velec);
 590
 591             fscal            = felec;
 592
 593             /* Update vectorial force */
 594             fix2             = _mm_macc_pd(dx23,fscal,fix2);
 595             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
 596             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
 597
 598             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
 599             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
 600             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
 601
 602             /**************************
 603              * CALCULATE INTERACTIONS *
 604              **************************/
 605
 606             r31              = _mm_mul_pd(rsq31,rinv31);
 607
 608             /* Calculate table index by multiplying r with table scale and truncate to integer */
 609             rt               = _mm_mul_pd(r31,vftabscale);
 610             vfitab           = _mm_cvttpd_epi32(rt);
 611 #ifdef __XOP__
 612             vfeps            = _mm_frcz_pd(rt);
 613 #else
 614             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 615 #endif
 616             twovfeps         = _mm_add_pd(vfeps,vfeps);
 617             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 618
 619             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 620             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 621             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 622             GMX_MM_TRANSPOSE2_PD(Y,F);
 623             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 624             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 625             GMX_MM_TRANSPOSE2_PD(G,H);
 626             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 627             VV               = _mm_macc_pd(vfeps,Fp,Y);
 628             velec            = _mm_mul_pd(qq31,VV);
 629             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 630             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
 631
 632             /* Update potential sum for this i atom from the interaction with this j atom. */
 633             velecsum         = _mm_add_pd(velecsum,velec);
 634
 635             fscal            = felec;
 636
 637             /* Update vectorial force */
 638             fix3             = _mm_macc_pd(dx31,fscal,fix3);
 639             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
 640             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
 641
 642             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
 643             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
 644             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
 645
 646             /**************************
 647              * CALCULATE INTERACTIONS *
 648              **************************/
 649
 650             r32              = _mm_mul_pd(rsq32,rinv32);
 651
 652             /* Calculate table index by multiplying r with table scale and truncate to integer */
 653             rt               = _mm_mul_pd(r32,vftabscale);
 654             vfitab           = _mm_cvttpd_epi32(rt);
 655 #ifdef __XOP__
 656             vfeps            = _mm_frcz_pd(rt);
 657 #else
 658             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 659 #endif
 660             twovfeps         = _mm_add_pd(vfeps,vfeps);
 661             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 662
 663             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 664             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 665             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 666             GMX_MM_TRANSPOSE2_PD(Y,F);
 667             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 668             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 669             GMX_MM_TRANSPOSE2_PD(G,H);
 670             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 671             VV               = _mm_macc_pd(vfeps,Fp,Y);
 672             velec            = _mm_mul_pd(qq32,VV);
 673             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 674             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
 675
 676             /* Update potential sum for this i atom from the interaction with this j atom. */
 677             velecsum         = _mm_add_pd(velecsum,velec);
 678
 679             fscal            = felec;
 680
 681             /* Update vectorial force */
 682             fix3             = _mm_macc_pd(dx32,fscal,fix3);
 683             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
 684             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
 685
 686             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
 687             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
 688             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
 689
 690             /**************************
 691              * CALCULATE INTERACTIONS *
 692              **************************/
 693
 694             r33              = _mm_mul_pd(rsq33,rinv33);
 695
 696             /* Calculate table index by multiplying r with table scale and truncate to integer */
 697             rt               = _mm_mul_pd(r33,vftabscale);
 698             vfitab           = _mm_cvttpd_epi32(rt);
 699 #ifdef __XOP__
 700             vfeps            = _mm_frcz_pd(rt);
 701 #else
 702             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 703 #endif
 704             twovfeps         = _mm_add_pd(vfeps,vfeps);
 705             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 706
 707             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 708             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 709             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
 710             GMX_MM_TRANSPOSE2_PD(Y,F);
 711             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 712             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
 713             GMX_MM_TRANSPOSE2_PD(G,H);
 714             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 715             VV               = _mm_macc_pd(vfeps,Fp,Y);
 716             velec            = _mm_mul_pd(qq33,VV);
 717             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 718             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
 719
 720             /* Update potential sum for this i atom from the interaction with this j atom. */
 721             velecsum         = _mm_add_pd(velecsum,velec);
 722
 723             fscal            = felec;
 724
 725             /* Update vectorial force */
 726             fix3             = _mm_macc_pd(dx33,fscal,fix3);
 727             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
 728             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
 729
 730             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
 731             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
 732             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
 733
 734             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 735
 736             /* Inner loop uses 476 flops */
 737         }
 738
 739         if(jidx<j_index_end)
 740         {
 741
 742             jnrA             = jjnr[jidx];
 743             j_coord_offsetA  = DIM*jnrA;
 744
 745             /* load j atom coordinates */
 746             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 747                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 748                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 749
 750             /* Calculate displacement vector */
 751             dx00             = _mm_sub_pd(ix0,jx0);
 752             dy00             = _mm_sub_pd(iy0,jy0);
 753             dz00             = _mm_sub_pd(iz0,jz0);
 754             dx11             = _mm_sub_pd(ix1,jx1);
 755             dy11             = _mm_sub_pd(iy1,jy1);
 756             dz11             = _mm_sub_pd(iz1,jz1);
 757             dx12             = _mm_sub_pd(ix1,jx2);
 758             dy12             = _mm_sub_pd(iy1,jy2);
 759             dz12             = _mm_sub_pd(iz1,jz2);
 760             dx13             = _mm_sub_pd(ix1,jx3);
 761             dy13             = _mm_sub_pd(iy1,jy3);
 762             dz13             = _mm_sub_pd(iz1,jz3);
 763             dx21             = _mm_sub_pd(ix2,jx1);
 764             dy21             = _mm_sub_pd(iy2,jy1);
 765             dz21             = _mm_sub_pd(iz2,jz1);
 766             dx22             = _mm_sub_pd(ix2,jx2);
 767             dy22             = _mm_sub_pd(iy2,jy2);
 768             dz22             = _mm_sub_pd(iz2,jz2);
 769             dx23             = _mm_sub_pd(ix2,jx3);
 770             dy23             = _mm_sub_pd(iy2,jy3);
 771             dz23             = _mm_sub_pd(iz2,jz3);
 772             dx31             = _mm_sub_pd(ix3,jx1);
 773             dy31             = _mm_sub_pd(iy3,jy1);
 774             dz31             = _mm_sub_pd(iz3,jz1);
 775             dx32             = _mm_sub_pd(ix3,jx2);
 776             dy32             = _mm_sub_pd(iy3,jy2);
 777             dz32             = _mm_sub_pd(iz3,jz2);
 778             dx33             = _mm_sub_pd(ix3,jx3);
 779             dy33             = _mm_sub_pd(iy3,jy3);
 780             dz33             = _mm_sub_pd(iz3,jz3);
 781
 782             /* Calculate squared distance and things based on it */
 783             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 784             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 785             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 786             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 787             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 788             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 789             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 790             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 791             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 792             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 793
 794             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 795             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 796             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 797             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 798             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 799             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 800             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 801             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 802             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 803             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 804
 805             fjx0             = _mm_setzero_pd();
 806             fjy0             = _mm_setzero_pd();
 807             fjz0             = _mm_setzero_pd();
 808             fjx1             = _mm_setzero_pd();
 809             fjy1             = _mm_setzero_pd();
 810             fjz1             = _mm_setzero_pd();
 811             fjx2             = _mm_setzero_pd();
 812             fjy2             = _mm_setzero_pd();
 813             fjz2             = _mm_setzero_pd();
 814             fjx3             = _mm_setzero_pd();
 815             fjy3             = _mm_setzero_pd();
 816             fjz3             = _mm_setzero_pd();
 817
 818             /**************************
 819              * CALCULATE INTERACTIONS *
 820              **************************/
 821
 822             r00              = _mm_mul_pd(rsq00,rinv00);
 823
 824             /* Calculate table index by multiplying r with table scale and truncate to integer */
 825             rt               = _mm_mul_pd(r00,vftabscale);
 826             vfitab           = _mm_cvttpd_epi32(rt);
 827 #ifdef __XOP__
 828             vfeps            = _mm_frcz_pd(rt);
 829 #else
 830             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 831 #endif
 832             twovfeps         = _mm_add_pd(vfeps,vfeps);
 833             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 834
 835             /* CUBIC SPLINE TABLE DISPERSION */
 836             vfitab           = _mm_add_epi32(vfitab,ifour);
 837             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 838             F                = _mm_setzero_pd();
 839             GMX_MM_TRANSPOSE2_PD(Y,F);
 840             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 841             H                = _mm_setzero_pd();
 842             GMX_MM_TRANSPOSE2_PD(G,H);
 843             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 844             VV               = _mm_macc_pd(vfeps,Fp,Y);
 845             vvdw6            = _mm_mul_pd(c6_00,VV);
 846             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 847             fvdw6            = _mm_mul_pd(c6_00,FF);
 848
 849             /* CUBIC SPLINE TABLE REPULSION */
 850             vfitab           = _mm_add_epi32(vfitab,ifour);
 851             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 852             F                = _mm_setzero_pd();
 853             GMX_MM_TRANSPOSE2_PD(Y,F);
 854             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 855             H                = _mm_setzero_pd();
 856             GMX_MM_TRANSPOSE2_PD(G,H);
 857             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
 858             VV               = _mm_macc_pd(vfeps,Fp,Y);
 859             vvdw12           = _mm_mul_pd(c12_00,VV);
 860             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
 861             fvdw12           = _mm_mul_pd(c12_00,FF);
 862             vvdw             = _mm_add_pd(vvdw12,vvdw6);
 863             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
 864
 865             /* Update potential sum for this i atom from the interaction with this j atom. */
 866             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 867             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 868
 869             fscal            = fvdw;
 870
 871             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 872
 873             /* Update vectorial force */
 874             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 875             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 876             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 877
 878             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
 879             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
 880             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
 881
 882             /**************************
 883              * CALCULATE INTERACTIONS *
 884              **************************/
 885
 886             r11              = _mm_mul_pd(rsq11,rinv11);
 887
 888             /* Calculate table index by multiplying r with table scale and truncate to integer */
 889             rt               = _mm_mul_pd(r11,vftabscale);
 890             vfitab           = _mm_cvttpd_epi32(rt);
 891 #ifdef __XOP__
 892             vfeps            = _mm_frcz_pd(rt);
 893 #else
 894             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 895 #endif
 896             twovfeps         = _mm_add_pd(vfeps,vfeps);
 897             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 898
 899             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 900             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 901             F                = _mm_setzero_pd();
 902             GMX_MM_TRANSPOSE2_PD(Y,F);
 903             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 904             H                = _mm_setzero_pd();
 905             GMX_MM_TRANSPOSE2_PD(G,H);
 906             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 907             VV               = _mm_macc_pd(vfeps,Fp,Y);
 908             velec            = _mm_mul_pd(qq11,VV);
 909             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 910             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
 911
 912             /* Update potential sum for this i atom from the interaction with this j atom. */
 913             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 914             velecsum         = _mm_add_pd(velecsum,velec);
 915
 916             fscal            = felec;
 917
 918             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 919
 920             /* Update vectorial force */
 921             fix1             = _mm_macc_pd(dx11,fscal,fix1);
 922             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
 923             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
 924
 925             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
 926             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
 927             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
 928
 929             /**************************
 930              * CALCULATE INTERACTIONS *
 931              **************************/
 932
 933             r12              = _mm_mul_pd(rsq12,rinv12);
 934
 935             /* Calculate table index by multiplying r with table scale and truncate to integer */
 936             rt               = _mm_mul_pd(r12,vftabscale);
 937             vfitab           = _mm_cvttpd_epi32(rt);
 938 #ifdef __XOP__
 939             vfeps            = _mm_frcz_pd(rt);
 940 #else
 941             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 942 #endif
 943             twovfeps         = _mm_add_pd(vfeps,vfeps);
 944             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 945
 946             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 947             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 948             F                = _mm_setzero_pd();
 949             GMX_MM_TRANSPOSE2_PD(Y,F);
 950             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 951             H                = _mm_setzero_pd();
 952             GMX_MM_TRANSPOSE2_PD(G,H);
 953             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
 954             VV               = _mm_macc_pd(vfeps,Fp,Y);
 955             velec            = _mm_mul_pd(qq12,VV);
 956             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
 957             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
 958
 959             /* Update potential sum for this i atom from the interaction with this j atom. */
 960             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 961             velecsum         = _mm_add_pd(velecsum,velec);
 962
 963             fscal            = felec;
 964
 965             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 966
 967             /* Update vectorial force */
 968             fix1             = _mm_macc_pd(dx12,fscal,fix1);
 969             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
 970             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
 971
 972             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
 973             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
 974             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
 975
 976             /**************************
 977              * CALCULATE INTERACTIONS *
 978              **************************/
 979
 980             r13              = _mm_mul_pd(rsq13,rinv13);
 981
 982             /* Calculate table index by multiplying r with table scale and truncate to integer */
 983             rt               = _mm_mul_pd(r13,vftabscale);
 984             vfitab           = _mm_cvttpd_epi32(rt);
 985 #ifdef __XOP__
 986             vfeps            = _mm_frcz_pd(rt);
 987 #else
 988             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
 989 #endif
 990             twovfeps         = _mm_add_pd(vfeps,vfeps);
 991             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 992
 993             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 994             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
 995             F                = _mm_setzero_pd();
 996             GMX_MM_TRANSPOSE2_PD(Y,F);
 997             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
 998             H                = _mm_setzero_pd();
 999             GMX_MM_TRANSPOSE2_PD(G,H);
1000             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1001             VV               = _mm_macc_pd(vfeps,Fp,Y);
1002             velec            = _mm_mul_pd(qq13,VV);
1003             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1004             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1005
1006             /* Update potential sum for this i atom from the interaction with this j atom. */
1007             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1008             velecsum         = _mm_add_pd(velecsum,velec);
1009
1010             fscal            = felec;
1011
1012             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1013
1014             /* Update vectorial force */
1015             fix1             = _mm_macc_pd(dx13,fscal,fix1);
1016             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
1017             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
1018
1019             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
1020             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
1021             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
1022
1023             /**************************
1024              * CALCULATE INTERACTIONS *
1025              **************************/
1026
1027             r21              = _mm_mul_pd(rsq21,rinv21);
1028
1029             /* Calculate table index by multiplying r with table scale and truncate to integer */
1030             rt               = _mm_mul_pd(r21,vftabscale);
1031             vfitab           = _mm_cvttpd_epi32(rt);
1032 #ifdef __XOP__
1033             vfeps            = _mm_frcz_pd(rt);
1034 #else
1035             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1036 #endif
1037             twovfeps         = _mm_add_pd(vfeps,vfeps);
1038             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1039
1040             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1041             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1042             F                = _mm_setzero_pd();
1043             GMX_MM_TRANSPOSE2_PD(Y,F);
1044             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1045             H                = _mm_setzero_pd();
1046             GMX_MM_TRANSPOSE2_PD(G,H);
1047             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1048             VV               = _mm_macc_pd(vfeps,Fp,Y);
1049             velec            = _mm_mul_pd(qq21,VV);
1050             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1051             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1052
1053             /* Update potential sum for this i atom from the interaction with this j atom. */
1054             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1055             velecsum         = _mm_add_pd(velecsum,velec);
1056
1057             fscal            = felec;
1058
1059             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1060
1061             /* Update vectorial force */
1062             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1063             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1064             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1065
1066             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1067             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1068             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1069
1070             /**************************
1071              * CALCULATE INTERACTIONS *
1072              **************************/
1073
1074             r22              = _mm_mul_pd(rsq22,rinv22);
1075
1076             /* Calculate table index by multiplying r with table scale and truncate to integer */
1077             rt               = _mm_mul_pd(r22,vftabscale);
1078             vfitab           = _mm_cvttpd_epi32(rt);
1079 #ifdef __XOP__
1080             vfeps            = _mm_frcz_pd(rt);
1081 #else
1082             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1083 #endif
1084             twovfeps         = _mm_add_pd(vfeps,vfeps);
1085             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1086
1087             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1088             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1089             F                = _mm_setzero_pd();
1090             GMX_MM_TRANSPOSE2_PD(Y,F);
1091             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1092             H                = _mm_setzero_pd();
1093             GMX_MM_TRANSPOSE2_PD(G,H);
1094             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1095             VV               = _mm_macc_pd(vfeps,Fp,Y);
1096             velec            = _mm_mul_pd(qq22,VV);
1097             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1098             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1099
1100             /* Update potential sum for this i atom from the interaction with this j atom. */
1101             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1102             velecsum         = _mm_add_pd(velecsum,velec);
1103
1104             fscal            = felec;
1105
1106             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1107
1108             /* Update vectorial force */
1109             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1110             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1111             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1112
1113             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1114             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1115             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1116
1117             /**************************
1118              * CALCULATE INTERACTIONS *
1119              **************************/
1120
1121             r23              = _mm_mul_pd(rsq23,rinv23);
1122
1123             /* Calculate table index by multiplying r with table scale and truncate to integer */
1124             rt               = _mm_mul_pd(r23,vftabscale);
1125             vfitab           = _mm_cvttpd_epi32(rt);
1126 #ifdef __XOP__
1127             vfeps            = _mm_frcz_pd(rt);
1128 #else
1129             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1130 #endif
1131             twovfeps         = _mm_add_pd(vfeps,vfeps);
1132             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1133
1134             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1135             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1136             F                = _mm_setzero_pd();
1137             GMX_MM_TRANSPOSE2_PD(Y,F);
1138             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1139             H                = _mm_setzero_pd();
1140             GMX_MM_TRANSPOSE2_PD(G,H);
1141             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1142             VV               = _mm_macc_pd(vfeps,Fp,Y);
1143             velec            = _mm_mul_pd(qq23,VV);
1144             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1145             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1146
1147             /* Update potential sum for this i atom from the interaction with this j atom. */
1148             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1149             velecsum         = _mm_add_pd(velecsum,velec);
1150
1151             fscal            = felec;
1152
1153             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1154
1155             /* Update vectorial force */
1156             fix2             = _mm_macc_pd(dx23,fscal,fix2);
1157             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
1158             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
1159
1160             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
1161             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
1162             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
1163
1164             /**************************
1165              * CALCULATE INTERACTIONS *
1166              **************************/
1167
1168             r31              = _mm_mul_pd(rsq31,rinv31);
1169
1170             /* Calculate table index by multiplying r with table scale and truncate to integer */
1171             rt               = _mm_mul_pd(r31,vftabscale);
1172             vfitab           = _mm_cvttpd_epi32(rt);
1173 #ifdef __XOP__
1174             vfeps            = _mm_frcz_pd(rt);
1175 #else
1176             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1177 #endif
1178             twovfeps         = _mm_add_pd(vfeps,vfeps);
1179             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1180
1181             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1182             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1183             F                = _mm_setzero_pd();
1184             GMX_MM_TRANSPOSE2_PD(Y,F);
1185             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1186             H                = _mm_setzero_pd();
1187             GMX_MM_TRANSPOSE2_PD(G,H);
1188             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1189             VV               = _mm_macc_pd(vfeps,Fp,Y);
1190             velec            = _mm_mul_pd(qq31,VV);
1191             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1192             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1193
1194             /* Update potential sum for this i atom from the interaction with this j atom. */
1195             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1196             velecsum         = _mm_add_pd(velecsum,velec);
1197
1198             fscal            = felec;
1199
1200             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1201
1202             /* Update vectorial force */
1203             fix3             = _mm_macc_pd(dx31,fscal,fix3);
1204             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
1205             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
1206
1207             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
1208             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
1209             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
1210
1211             /**************************
1212              * CALCULATE INTERACTIONS *
1213              **************************/
1214
1215             r32              = _mm_mul_pd(rsq32,rinv32);
1216
1217             /* Calculate table index by multiplying r with table scale and truncate to integer */
1218             rt               = _mm_mul_pd(r32,vftabscale);
1219             vfitab           = _mm_cvttpd_epi32(rt);
1220 #ifdef __XOP__
1221             vfeps            = _mm_frcz_pd(rt);
1222 #else
1223             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1224 #endif
1225             twovfeps         = _mm_add_pd(vfeps,vfeps);
1226             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1227
1228             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1229             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1230             F                = _mm_setzero_pd();
1231             GMX_MM_TRANSPOSE2_PD(Y,F);
1232             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1233             H                = _mm_setzero_pd();
1234             GMX_MM_TRANSPOSE2_PD(G,H);
1235             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1236             VV               = _mm_macc_pd(vfeps,Fp,Y);
1237             velec            = _mm_mul_pd(qq32,VV);
1238             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1239             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1240
1241             /* Update potential sum for this i atom from the interaction with this j atom. */
1242             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1243             velecsum         = _mm_add_pd(velecsum,velec);
1244
1245             fscal            = felec;
1246
1247             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1248
1249             /* Update vectorial force */
1250             fix3             = _mm_macc_pd(dx32,fscal,fix3);
1251             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
1252             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
1253
1254             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
1255             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
1256             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
1257
1258             /**************************
1259              * CALCULATE INTERACTIONS *
1260              **************************/
1261
1262             r33              = _mm_mul_pd(rsq33,rinv33);
1263
1264             /* Calculate table index by multiplying r with table scale and truncate to integer */
1265             rt               = _mm_mul_pd(r33,vftabscale);
1266             vfitab           = _mm_cvttpd_epi32(rt);
1267 #ifdef __XOP__
1268             vfeps            = _mm_frcz_pd(rt);
1269 #else
1270             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1271 #endif
1272             twovfeps         = _mm_add_pd(vfeps,vfeps);
1273             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1274
1275             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1276             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1277             F                = _mm_setzero_pd();
1278             GMX_MM_TRANSPOSE2_PD(Y,F);
1279             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1280             H                = _mm_setzero_pd();
1281             GMX_MM_TRANSPOSE2_PD(G,H);
1282             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1283             VV               = _mm_macc_pd(vfeps,Fp,Y);
1284             velec            = _mm_mul_pd(qq33,VV);
1285             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1286             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1287
1288             /* Update potential sum for this i atom from the interaction with this j atom. */
1289             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1290             velecsum         = _mm_add_pd(velecsum,velec);
1291
1292             fscal            = felec;
1293
1294             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1295
1296             /* Update vectorial force */
1297             fix3             = _mm_macc_pd(dx33,fscal,fix3);
1298             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
1299             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
1300
1301             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
1302             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
1303             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
1304
1305             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1306
1307             /* Inner loop uses 476 flops */
1308         }
1309
1310         /* End of innermost loop */
1311
1312         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1313                                               f+i_coord_offset,fshift+i_shift_offset);
1314
1315         ggid                        = gid[iidx];
1316         /* Update potential energies */
1317         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1318         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1319
1320         /* Increment number of inner iterations */
1321         inneriter                  += j_index_end - j_index_start;
1322
1323         /* Outer loop uses 26 flops */
1324     }
1325
1326     /* Increment number of outer iterations */
1327     outeriter        += nri;
1328
1329     /* Update outer/inner flops */
1330
1331     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*476);
1332 }
1333 /*
1334  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_avx_128_fma_double
1335  * Electrostatics interaction: CubicSplineTable
1336  * VdW interaction:            CubicSplineTable
1337  * Geometry:                   Water4-Water4
1338  * Calculate force/pot:        Force
1339  */
1340 void
1341 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_avx_128_fma_double
1342                     (t_nblist * gmx_restrict                nlist,
1343                      rvec * gmx_restrict                    xx,
1344                      rvec * gmx_restrict                    ff,
1345                      t_forcerec * gmx_restrict              fr,
1346                      t_mdatoms * gmx_restrict               mdatoms,
1347                      nb_kernel_data_t * gmx_restrict        kernel_data,
1348                      t_nrnb * gmx_restrict                  nrnb)
1349 {
1350     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1351      * just 0 for non-waters.
1352      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1353      * jnr indices corresponding to data put in the four positions in the SIMD register.
1354      */
1355     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1356     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1357     int              jnrA,jnrB;
1358     int              j_coord_offsetA,j_coord_offsetB;
1359     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1360     real             rcutoff_scalar;
1361     real             *shiftvec,*fshift,*x,*f;
1362     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1363     int              vdwioffset0;
1364     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1365     int              vdwioffset1;
1366     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1367     int              vdwioffset2;
1368     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1369     int              vdwioffset3;
1370     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1371     int              vdwjidx0A,vdwjidx0B;
1372     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1373     int              vdwjidx1A,vdwjidx1B;
1374     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1375     int              vdwjidx2A,vdwjidx2B;
1376     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1377     int              vdwjidx3A,vdwjidx3B;
1378     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1379     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1380     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1381     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1382     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1383     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1384     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1385     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1386     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1387     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1388     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1389     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1390     real             *charge;
1391     int              nvdwtype;
1392     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1393     int              *vdwtype;
1394     real             *vdwparam;
1395     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1396     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1397     __m128i          vfitab;
1398     __m128i          ifour       = _mm_set1_epi32(4);
1399     __m128d          rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
1400     real             *vftab;
1401     __m128d          dummy_mask,cutoff_mask;
1402     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1403     __m128d          one     = _mm_set1_pd(1.0);
1404     __m128d          two     = _mm_set1_pd(2.0);
1405     x                = xx[0];
1406     f                = ff[0];
1407
1408     nri              = nlist->nri;
1409     iinr             = nlist->iinr;
1410     jindex           = nlist->jindex;
1411     jjnr             = nlist->jjnr;
1412     shiftidx         = nlist->shift;
1413     gid              = nlist->gid;
1414     shiftvec         = fr->shift_vec[0];
1415     fshift           = fr->fshift[0];
1416     facel            = _mm_set1_pd(fr->epsfac);
1417     charge           = mdatoms->chargeA;
1418     nvdwtype         = fr->ntype;
1419     vdwparam         = fr->nbfp;
1420     vdwtype          = mdatoms->typeA;
1421
1422     vftab            = kernel_data->table_elec_vdw->data;
1423     vftabscale       = _mm_set1_pd(kernel_data->table_elec_vdw->scale);
1424
1425     /* Setup water-specific parameters */
1426     inr              = nlist->iinr[0];
1427     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1428     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1429     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1430     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1431
1432     jq1              = _mm_set1_pd(charge[inr+1]);
1433     jq2              = _mm_set1_pd(charge[inr+2]);
1434     jq3              = _mm_set1_pd(charge[inr+3]);
1435     vdwjidx0A        = 2*vdwtype[inr+0];
1436     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1437     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1438     qq11             = _mm_mul_pd(iq1,jq1);
1439     qq12             = _mm_mul_pd(iq1,jq2);
1440     qq13             = _mm_mul_pd(iq1,jq3);
1441     qq21             = _mm_mul_pd(iq2,jq1);
1442     qq22             = _mm_mul_pd(iq2,jq2);
1443     qq23             = _mm_mul_pd(iq2,jq3);
1444     qq31             = _mm_mul_pd(iq3,jq1);
1445     qq32             = _mm_mul_pd(iq3,jq2);
1446     qq33             = _mm_mul_pd(iq3,jq3);
1447
1448     /* Avoid stupid compiler warnings */
1449     jnrA = jnrB = 0;
1450     j_coord_offsetA = 0;
1451     j_coord_offsetB = 0;
1452
1453     outeriter        = 0;
1454     inneriter        = 0;
1455
1456     /* Start outer loop over neighborlists */
1457     for(iidx=0; iidx<nri; iidx++)
1458     {
1459         /* Load shift vector for this list */
1460         i_shift_offset   = DIM*shiftidx[iidx];
1461
1462         /* Load limits for loop over neighbors */
1463         j_index_start    = jindex[iidx];
1464         j_index_end      = jindex[iidx+1];
1465
1466         /* Get outer coordinate index */
1467         inr              = iinr[iidx];
1468         i_coord_offset   = DIM*inr;
1469
1470         /* Load i particle coords and add shift vector */
1471         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1472                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1473
1474         fix0             = _mm_setzero_pd();
1475         fiy0             = _mm_setzero_pd();
1476         fiz0             = _mm_setzero_pd();
1477         fix1             = _mm_setzero_pd();
1478         fiy1             = _mm_setzero_pd();
1479         fiz1             = _mm_setzero_pd();
1480         fix2             = _mm_setzero_pd();
1481         fiy2             = _mm_setzero_pd();
1482         fiz2             = _mm_setzero_pd();
1483         fix3             = _mm_setzero_pd();
1484         fiy3             = _mm_setzero_pd();
1485         fiz3             = _mm_setzero_pd();
1486
1487         /* Start inner kernel loop */
1488         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1489         {
1490
1491             /* Get j neighbor index, and coordinate index */
1492             jnrA             = jjnr[jidx];
1493             jnrB             = jjnr[jidx+1];
1494             j_coord_offsetA  = DIM*jnrA;
1495             j_coord_offsetB  = DIM*jnrB;
1496
1497             /* load j atom coordinates */
1498             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1499                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1500                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1501
1502             /* Calculate displacement vector */
1503             dx00             = _mm_sub_pd(ix0,jx0);
1504             dy00             = _mm_sub_pd(iy0,jy0);
1505             dz00             = _mm_sub_pd(iz0,jz0);
1506             dx11             = _mm_sub_pd(ix1,jx1);
1507             dy11             = _mm_sub_pd(iy1,jy1);
1508             dz11             = _mm_sub_pd(iz1,jz1);
1509             dx12             = _mm_sub_pd(ix1,jx2);
1510             dy12             = _mm_sub_pd(iy1,jy2);
1511             dz12             = _mm_sub_pd(iz1,jz2);
1512             dx13             = _mm_sub_pd(ix1,jx3);
1513             dy13             = _mm_sub_pd(iy1,jy3);
1514             dz13             = _mm_sub_pd(iz1,jz3);
1515             dx21             = _mm_sub_pd(ix2,jx1);
1516             dy21             = _mm_sub_pd(iy2,jy1);
1517             dz21             = _mm_sub_pd(iz2,jz1);
1518             dx22             = _mm_sub_pd(ix2,jx2);
1519             dy22             = _mm_sub_pd(iy2,jy2);
1520             dz22             = _mm_sub_pd(iz2,jz2);
1521             dx23             = _mm_sub_pd(ix2,jx3);
1522             dy23             = _mm_sub_pd(iy2,jy3);
1523             dz23             = _mm_sub_pd(iz2,jz3);
1524             dx31             = _mm_sub_pd(ix3,jx1);
1525             dy31             = _mm_sub_pd(iy3,jy1);
1526             dz31             = _mm_sub_pd(iz3,jz1);
1527             dx32             = _mm_sub_pd(ix3,jx2);
1528             dy32             = _mm_sub_pd(iy3,jy2);
1529             dz32             = _mm_sub_pd(iz3,jz2);
1530             dx33             = _mm_sub_pd(ix3,jx3);
1531             dy33             = _mm_sub_pd(iy3,jy3);
1532             dz33             = _mm_sub_pd(iz3,jz3);
1533
1534             /* Calculate squared distance and things based on it */
1535             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1536             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1537             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1538             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1539             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1540             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1541             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1542             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1543             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1544             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1545
1546             rinv00           = gmx_mm_invsqrt_pd(rsq00);
1547             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1548             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1549             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1550             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1551             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1552             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1553             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1554             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1555             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1556
1557             fjx0             = _mm_setzero_pd();
1558             fjy0             = _mm_setzero_pd();
1559             fjz0             = _mm_setzero_pd();
1560             fjx1             = _mm_setzero_pd();
1561             fjy1             = _mm_setzero_pd();
1562             fjz1             = _mm_setzero_pd();
1563             fjx2             = _mm_setzero_pd();
1564             fjy2             = _mm_setzero_pd();
1565             fjz2             = _mm_setzero_pd();
1566             fjx3             = _mm_setzero_pd();
1567             fjy3             = _mm_setzero_pd();
1568             fjz3             = _mm_setzero_pd();
1569
1570             /**************************
1571              * CALCULATE INTERACTIONS *
1572              **************************/
1573
1574             r00              = _mm_mul_pd(rsq00,rinv00);
1575
1576             /* Calculate table index by multiplying r with table scale and truncate to integer */
1577             rt               = _mm_mul_pd(r00,vftabscale);
1578             vfitab           = _mm_cvttpd_epi32(rt);
1579 #ifdef __XOP__
1580             vfeps            = _mm_frcz_pd(rt);
1581 #else
1582             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1583 #endif
1584             twovfeps         = _mm_add_pd(vfeps,vfeps);
1585             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1586
1587             /* CUBIC SPLINE TABLE DISPERSION */
1588             vfitab           = _mm_add_epi32(vfitab,ifour);
1589             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1590             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1591             GMX_MM_TRANSPOSE2_PD(Y,F);
1592             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1593             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1594             GMX_MM_TRANSPOSE2_PD(G,H);
1595             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1596             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1597             fvdw6            = _mm_mul_pd(c6_00,FF);
1598
1599             /* CUBIC SPLINE TABLE REPULSION */
1600             vfitab           = _mm_add_epi32(vfitab,ifour);
1601             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1602             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1603             GMX_MM_TRANSPOSE2_PD(Y,F);
1604             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1605             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1606             GMX_MM_TRANSPOSE2_PD(G,H);
1607             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
1608             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
1609             fvdw12           = _mm_mul_pd(c12_00,FF);
1610             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
1611
1612             fscal            = fvdw;
1613
1614             /* Update vectorial force */
1615             fix0             = _mm_macc_pd(dx00,fscal,fix0);
1616             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
1617             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
1618
1619             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
1620             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
1621             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
1622
1623             /**************************
1624              * CALCULATE INTERACTIONS *
1625              **************************/
1626
1627             r11              = _mm_mul_pd(rsq11,rinv11);
1628
1629             /* Calculate table index by multiplying r with table scale and truncate to integer */
1630             rt               = _mm_mul_pd(r11,vftabscale);
1631             vfitab           = _mm_cvttpd_epi32(rt);
1632 #ifdef __XOP__
1633             vfeps            = _mm_frcz_pd(rt);
1634 #else
1635             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1636 #endif
1637             twovfeps         = _mm_add_pd(vfeps,vfeps);
1638             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1639
1640             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1641             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1642             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1643             GMX_MM_TRANSPOSE2_PD(Y,F);
1644             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1645             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1646             GMX_MM_TRANSPOSE2_PD(G,H);
1647             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1648             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1649             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
1650
1651             fscal            = felec;
1652
1653             /* Update vectorial force */
1654             fix1             = _mm_macc_pd(dx11,fscal,fix1);
1655             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
1656             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
1657
1658             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
1659             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
1660             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
1661
1662             /**************************
1663              * CALCULATE INTERACTIONS *
1664              **************************/
1665
1666             r12              = _mm_mul_pd(rsq12,rinv12);
1667
1668             /* Calculate table index by multiplying r with table scale and truncate to integer */
1669             rt               = _mm_mul_pd(r12,vftabscale);
1670             vfitab           = _mm_cvttpd_epi32(rt);
1671 #ifdef __XOP__
1672             vfeps            = _mm_frcz_pd(rt);
1673 #else
1674             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1675 #endif
1676             twovfeps         = _mm_add_pd(vfeps,vfeps);
1677             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1678
1679             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1680             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1681             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1682             GMX_MM_TRANSPOSE2_PD(Y,F);
1683             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1684             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1685             GMX_MM_TRANSPOSE2_PD(G,H);
1686             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1687             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1688             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
1689
1690             fscal            = felec;
1691
1692             /* Update vectorial force */
1693             fix1             = _mm_macc_pd(dx12,fscal,fix1);
1694             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
1695             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
1696
1697             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
1698             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
1699             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
1700
1701             /**************************
1702              * CALCULATE INTERACTIONS *
1703              **************************/
1704
1705             r13              = _mm_mul_pd(rsq13,rinv13);
1706
1707             /* Calculate table index by multiplying r with table scale and truncate to integer */
1708             rt               = _mm_mul_pd(r13,vftabscale);
1709             vfitab           = _mm_cvttpd_epi32(rt);
1710 #ifdef __XOP__
1711             vfeps            = _mm_frcz_pd(rt);
1712 #else
1713             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1714 #endif
1715             twovfeps         = _mm_add_pd(vfeps,vfeps);
1716             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1717
1718             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1719             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1720             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1721             GMX_MM_TRANSPOSE2_PD(Y,F);
1722             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1723             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1724             GMX_MM_TRANSPOSE2_PD(G,H);
1725             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1726             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1727             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
1728
1729             fscal            = felec;
1730
1731             /* Update vectorial force */
1732             fix1             = _mm_macc_pd(dx13,fscal,fix1);
1733             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
1734             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
1735
1736             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
1737             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
1738             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
1739
1740             /**************************
1741              * CALCULATE INTERACTIONS *
1742              **************************/
1743
1744             r21              = _mm_mul_pd(rsq21,rinv21);
1745
1746             /* Calculate table index by multiplying r with table scale and truncate to integer */
1747             rt               = _mm_mul_pd(r21,vftabscale);
1748             vfitab           = _mm_cvttpd_epi32(rt);
1749 #ifdef __XOP__
1750             vfeps            = _mm_frcz_pd(rt);
1751 #else
1752             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1753 #endif
1754             twovfeps         = _mm_add_pd(vfeps,vfeps);
1755             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1756
1757             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1758             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1759             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1760             GMX_MM_TRANSPOSE2_PD(Y,F);
1761             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1762             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1763             GMX_MM_TRANSPOSE2_PD(G,H);
1764             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1765             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1766             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
1767
1768             fscal            = felec;
1769
1770             /* Update vectorial force */
1771             fix2             = _mm_macc_pd(dx21,fscal,fix2);
1772             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
1773             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
1774
1775             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
1776             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
1777             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
1778
1779             /**************************
1780              * CALCULATE INTERACTIONS *
1781              **************************/
1782
1783             r22              = _mm_mul_pd(rsq22,rinv22);
1784
1785             /* Calculate table index by multiplying r with table scale and truncate to integer */
1786             rt               = _mm_mul_pd(r22,vftabscale);
1787             vfitab           = _mm_cvttpd_epi32(rt);
1788 #ifdef __XOP__
1789             vfeps            = _mm_frcz_pd(rt);
1790 #else
1791             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1792 #endif
1793             twovfeps         = _mm_add_pd(vfeps,vfeps);
1794             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1795
1796             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1797             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1798             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1799             GMX_MM_TRANSPOSE2_PD(Y,F);
1800             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1801             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1802             GMX_MM_TRANSPOSE2_PD(G,H);
1803             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1804             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1805             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
1806
1807             fscal            = felec;
1808
1809             /* Update vectorial force */
1810             fix2             = _mm_macc_pd(dx22,fscal,fix2);
1811             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
1812             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
1813
1814             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
1815             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
1816             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
1817
1818             /**************************
1819              * CALCULATE INTERACTIONS *
1820              **************************/
1821
1822             r23              = _mm_mul_pd(rsq23,rinv23);
1823
1824             /* Calculate table index by multiplying r with table scale and truncate to integer */
1825             rt               = _mm_mul_pd(r23,vftabscale);
1826             vfitab           = _mm_cvttpd_epi32(rt);
1827 #ifdef __XOP__
1828             vfeps            = _mm_frcz_pd(rt);
1829 #else
1830             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1831 #endif
1832             twovfeps         = _mm_add_pd(vfeps,vfeps);
1833             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1834
1835             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1836             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1837             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1838             GMX_MM_TRANSPOSE2_PD(Y,F);
1839             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1840             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1841             GMX_MM_TRANSPOSE2_PD(G,H);
1842             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1843             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1844             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
1845
1846             fscal            = felec;
1847
1848             /* Update vectorial force */
1849             fix2             = _mm_macc_pd(dx23,fscal,fix2);
1850             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
1851             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
1852
1853             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
1854             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
1855             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
1856
1857             /**************************
1858              * CALCULATE INTERACTIONS *
1859              **************************/
1860
1861             r31              = _mm_mul_pd(rsq31,rinv31);
1862
1863             /* Calculate table index by multiplying r with table scale and truncate to integer */
1864             rt               = _mm_mul_pd(r31,vftabscale);
1865             vfitab           = _mm_cvttpd_epi32(rt);
1866 #ifdef __XOP__
1867             vfeps            = _mm_frcz_pd(rt);
1868 #else
1869             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1870 #endif
1871             twovfeps         = _mm_add_pd(vfeps,vfeps);
1872             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1873
1874             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1875             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1876             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1877             GMX_MM_TRANSPOSE2_PD(Y,F);
1878             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1879             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1880             GMX_MM_TRANSPOSE2_PD(G,H);
1881             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1882             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1883             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
1884
1885             fscal            = felec;
1886
1887             /* Update vectorial force */
1888             fix3             = _mm_macc_pd(dx31,fscal,fix3);
1889             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
1890             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
1891
1892             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
1893             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
1894             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
1895
1896             /**************************
1897              * CALCULATE INTERACTIONS *
1898              **************************/
1899
1900             r32              = _mm_mul_pd(rsq32,rinv32);
1901
1902             /* Calculate table index by multiplying r with table scale and truncate to integer */
1903             rt               = _mm_mul_pd(r32,vftabscale);
1904             vfitab           = _mm_cvttpd_epi32(rt);
1905 #ifdef __XOP__
1906             vfeps            = _mm_frcz_pd(rt);
1907 #else
1908             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1909 #endif
1910             twovfeps         = _mm_add_pd(vfeps,vfeps);
1911             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1912
1913             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1914             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1915             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1916             GMX_MM_TRANSPOSE2_PD(Y,F);
1917             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1918             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1919             GMX_MM_TRANSPOSE2_PD(G,H);
1920             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1921             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1922             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
1923
1924             fscal            = felec;
1925
1926             /* Update vectorial force */
1927             fix3             = _mm_macc_pd(dx32,fscal,fix3);
1928             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
1929             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
1930
1931             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
1932             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
1933             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
1934
1935             /**************************
1936              * CALCULATE INTERACTIONS *
1937              **************************/
1938
1939             r33              = _mm_mul_pd(rsq33,rinv33);
1940
1941             /* Calculate table index by multiplying r with table scale and truncate to integer */
1942             rt               = _mm_mul_pd(r33,vftabscale);
1943             vfitab           = _mm_cvttpd_epi32(rt);
1944 #ifdef __XOP__
1945             vfeps            = _mm_frcz_pd(rt);
1946 #else
1947             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
1948 #endif
1949             twovfeps         = _mm_add_pd(vfeps,vfeps);
1950             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1951
1952             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1953             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
1954             F                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) );
1955             GMX_MM_TRANSPOSE2_PD(Y,F);
1956             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
1957             H                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,1) +2);
1958             GMX_MM_TRANSPOSE2_PD(G,H);
1959             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
1960             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
1961             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
1962
1963             fscal            = felec;
1964
1965             /* Update vectorial force */
1966             fix3             = _mm_macc_pd(dx33,fscal,fix3);
1967             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
1968             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
1969
1970             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
1971             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
1972             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
1973
1974             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1975
1976             /* Inner loop uses 432 flops */
1977         }
1978
1979         if(jidx<j_index_end)
1980         {
1981
1982             jnrA             = jjnr[jidx];
1983             j_coord_offsetA  = DIM*jnrA;
1984
1985             /* load j atom coordinates */
1986             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1987                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1988                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1989
1990             /* Calculate displacement vector */
1991             dx00             = _mm_sub_pd(ix0,jx0);
1992             dy00             = _mm_sub_pd(iy0,jy0);
1993             dz00             = _mm_sub_pd(iz0,jz0);
1994             dx11             = _mm_sub_pd(ix1,jx1);
1995             dy11             = _mm_sub_pd(iy1,jy1);
1996             dz11             = _mm_sub_pd(iz1,jz1);
1997             dx12             = _mm_sub_pd(ix1,jx2);
1998             dy12             = _mm_sub_pd(iy1,jy2);
1999             dz12             = _mm_sub_pd(iz1,jz2);
2000             dx13             = _mm_sub_pd(ix1,jx3);
2001             dy13             = _mm_sub_pd(iy1,jy3);
2002             dz13             = _mm_sub_pd(iz1,jz3);
2003             dx21             = _mm_sub_pd(ix2,jx1);
2004             dy21             = _mm_sub_pd(iy2,jy1);
2005             dz21             = _mm_sub_pd(iz2,jz1);
2006             dx22             = _mm_sub_pd(ix2,jx2);
2007             dy22             = _mm_sub_pd(iy2,jy2);
2008             dz22             = _mm_sub_pd(iz2,jz2);
2009             dx23             = _mm_sub_pd(ix2,jx3);
2010             dy23             = _mm_sub_pd(iy2,jy3);
2011             dz23             = _mm_sub_pd(iz2,jz3);
2012             dx31             = _mm_sub_pd(ix3,jx1);
2013             dy31             = _mm_sub_pd(iy3,jy1);
2014             dz31             = _mm_sub_pd(iz3,jz1);
2015             dx32             = _mm_sub_pd(ix3,jx2);
2016             dy32             = _mm_sub_pd(iy3,jy2);
2017             dz32             = _mm_sub_pd(iz3,jz2);
2018             dx33             = _mm_sub_pd(ix3,jx3);
2019             dy33             = _mm_sub_pd(iy3,jy3);
2020             dz33             = _mm_sub_pd(iz3,jz3);
2021
2022             /* Calculate squared distance and things based on it */
2023             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
2024             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
2025             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
2026             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
2027             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
2028             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
2029             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
2030             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
2031             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
2032             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
2033
2034             rinv00           = gmx_mm_invsqrt_pd(rsq00);
2035             rinv11           = gmx_mm_invsqrt_pd(rsq11);
2036             rinv12           = gmx_mm_invsqrt_pd(rsq12);
2037             rinv13           = gmx_mm_invsqrt_pd(rsq13);
2038             rinv21           = gmx_mm_invsqrt_pd(rsq21);
2039             rinv22           = gmx_mm_invsqrt_pd(rsq22);
2040             rinv23           = gmx_mm_invsqrt_pd(rsq23);
2041             rinv31           = gmx_mm_invsqrt_pd(rsq31);
2042             rinv32           = gmx_mm_invsqrt_pd(rsq32);
2043             rinv33           = gmx_mm_invsqrt_pd(rsq33);
2044
2045             fjx0             = _mm_setzero_pd();
2046             fjy0             = _mm_setzero_pd();
2047             fjz0             = _mm_setzero_pd();
2048             fjx1             = _mm_setzero_pd();
2049             fjy1             = _mm_setzero_pd();
2050             fjz1             = _mm_setzero_pd();
2051             fjx2             = _mm_setzero_pd();
2052             fjy2             = _mm_setzero_pd();
2053             fjz2             = _mm_setzero_pd();
2054             fjx3             = _mm_setzero_pd();
2055             fjy3             = _mm_setzero_pd();
2056             fjz3             = _mm_setzero_pd();
2057
2058             /**************************
2059              * CALCULATE INTERACTIONS *
2060              **************************/
2061
2062             r00              = _mm_mul_pd(rsq00,rinv00);
2063
2064             /* Calculate table index by multiplying r with table scale and truncate to integer */
2065             rt               = _mm_mul_pd(r00,vftabscale);
2066             vfitab           = _mm_cvttpd_epi32(rt);
2067 #ifdef __XOP__
2068             vfeps            = _mm_frcz_pd(rt);
2069 #else
2070             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2071 #endif
2072             twovfeps         = _mm_add_pd(vfeps,vfeps);
2073             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2074
2075             /* CUBIC SPLINE TABLE DISPERSION */
2076             vfitab           = _mm_add_epi32(vfitab,ifour);
2077             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2078             F                = _mm_setzero_pd();
2079             GMX_MM_TRANSPOSE2_PD(Y,F);
2080             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2081             H                = _mm_setzero_pd();
2082             GMX_MM_TRANSPOSE2_PD(G,H);
2083             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
2084             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
2085             fvdw6            = _mm_mul_pd(c6_00,FF);
2086
2087             /* CUBIC SPLINE TABLE REPULSION */
2088             vfitab           = _mm_add_epi32(vfitab,ifour);
2089             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2090             F                = _mm_setzero_pd();
2091             GMX_MM_TRANSPOSE2_PD(Y,F);
2092             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2093             H                = _mm_setzero_pd();
2094             GMX_MM_TRANSPOSE2_PD(G,H);
2095             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(H,vfeps,G),F);
2096             FF               = _mm_macc_pd(vfeps,_mm_macc_pd(twovfeps,H,G),Fp);
2097             fvdw12           = _mm_mul_pd(c12_00,FF);
2098             fvdw             = _mm_xor_pd(signbit,_mm_mul_pd(_mm_add_pd(fvdw6,fvdw12),_mm_mul_pd(vftabscale,rinv00)));
2099
2100             fscal            = fvdw;
2101
2102             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2103
2104             /* Update vectorial force */
2105             fix0             = _mm_macc_pd(dx00,fscal,fix0);
2106             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
2107             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
2108
2109             fjx0             = _mm_macc_pd(dx00,fscal,fjx0);
2110             fjy0             = _mm_macc_pd(dy00,fscal,fjy0);
2111             fjz0             = _mm_macc_pd(dz00,fscal,fjz0);
2112
2113             /**************************
2114              * CALCULATE INTERACTIONS *
2115              **************************/
2116
2117             r11              = _mm_mul_pd(rsq11,rinv11);
2118
2119             /* Calculate table index by multiplying r with table scale and truncate to integer */
2120             rt               = _mm_mul_pd(r11,vftabscale);
2121             vfitab           = _mm_cvttpd_epi32(rt);
2122 #ifdef __XOP__
2123             vfeps            = _mm_frcz_pd(rt);
2124 #else
2125             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2126 #endif
2127             twovfeps         = _mm_add_pd(vfeps,vfeps);
2128             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2129
2130             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2131             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2132             F                = _mm_setzero_pd();
2133             GMX_MM_TRANSPOSE2_PD(Y,F);
2134             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2135             H                = _mm_setzero_pd();
2136             GMX_MM_TRANSPOSE2_PD(G,H);
2137             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2138             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2139             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq11,FF),_mm_mul_pd(vftabscale,rinv11)));
2140
2141             fscal            = felec;
2142
2143             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2144
2145             /* Update vectorial force */
2146             fix1             = _mm_macc_pd(dx11,fscal,fix1);
2147             fiy1             = _mm_macc_pd(dy11,fscal,fiy1);
2148             fiz1             = _mm_macc_pd(dz11,fscal,fiz1);
2149
2150             fjx1             = _mm_macc_pd(dx11,fscal,fjx1);
2151             fjy1             = _mm_macc_pd(dy11,fscal,fjy1);
2152             fjz1             = _mm_macc_pd(dz11,fscal,fjz1);
2153
2154             /**************************
2155              * CALCULATE INTERACTIONS *
2156              **************************/
2157
2158             r12              = _mm_mul_pd(rsq12,rinv12);
2159
2160             /* Calculate table index by multiplying r with table scale and truncate to integer */
2161             rt               = _mm_mul_pd(r12,vftabscale);
2162             vfitab           = _mm_cvttpd_epi32(rt);
2163 #ifdef __XOP__
2164             vfeps            = _mm_frcz_pd(rt);
2165 #else
2166             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2167 #endif
2168             twovfeps         = _mm_add_pd(vfeps,vfeps);
2169             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2170
2171             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2172             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2173             F                = _mm_setzero_pd();
2174             GMX_MM_TRANSPOSE2_PD(Y,F);
2175             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2176             H                = _mm_setzero_pd();
2177             GMX_MM_TRANSPOSE2_PD(G,H);
2178             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2179             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2180             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq12,FF),_mm_mul_pd(vftabscale,rinv12)));
2181
2182             fscal            = felec;
2183
2184             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2185
2186             /* Update vectorial force */
2187             fix1             = _mm_macc_pd(dx12,fscal,fix1);
2188             fiy1             = _mm_macc_pd(dy12,fscal,fiy1);
2189             fiz1             = _mm_macc_pd(dz12,fscal,fiz1);
2190
2191             fjx2             = _mm_macc_pd(dx12,fscal,fjx2);
2192             fjy2             = _mm_macc_pd(dy12,fscal,fjy2);
2193             fjz2             = _mm_macc_pd(dz12,fscal,fjz2);
2194
2195             /**************************
2196              * CALCULATE INTERACTIONS *
2197              **************************/
2198
2199             r13              = _mm_mul_pd(rsq13,rinv13);
2200
2201             /* Calculate table index by multiplying r with table scale and truncate to integer */
2202             rt               = _mm_mul_pd(r13,vftabscale);
2203             vfitab           = _mm_cvttpd_epi32(rt);
2204 #ifdef __XOP__
2205             vfeps            = _mm_frcz_pd(rt);
2206 #else
2207             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2208 #endif
2209             twovfeps         = _mm_add_pd(vfeps,vfeps);
2210             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2211
2212             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2213             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2214             F                = _mm_setzero_pd();
2215             GMX_MM_TRANSPOSE2_PD(Y,F);
2216             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2217             H                = _mm_setzero_pd();
2218             GMX_MM_TRANSPOSE2_PD(G,H);
2219             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2220             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2221             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq13,FF),_mm_mul_pd(vftabscale,rinv13)));
2222
2223             fscal            = felec;
2224
2225             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2226
2227             /* Update vectorial force */
2228             fix1             = _mm_macc_pd(dx13,fscal,fix1);
2229             fiy1             = _mm_macc_pd(dy13,fscal,fiy1);
2230             fiz1             = _mm_macc_pd(dz13,fscal,fiz1);
2231
2232             fjx3             = _mm_macc_pd(dx13,fscal,fjx3);
2233             fjy3             = _mm_macc_pd(dy13,fscal,fjy3);
2234             fjz3             = _mm_macc_pd(dz13,fscal,fjz3);
2235
2236             /**************************
2237              * CALCULATE INTERACTIONS *
2238              **************************/
2239
2240             r21              = _mm_mul_pd(rsq21,rinv21);
2241
2242             /* Calculate table index by multiplying r with table scale and truncate to integer */
2243             rt               = _mm_mul_pd(r21,vftabscale);
2244             vfitab           = _mm_cvttpd_epi32(rt);
2245 #ifdef __XOP__
2246             vfeps            = _mm_frcz_pd(rt);
2247 #else
2248             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2249 #endif
2250             twovfeps         = _mm_add_pd(vfeps,vfeps);
2251             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2252
2253             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2254             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2255             F                = _mm_setzero_pd();
2256             GMX_MM_TRANSPOSE2_PD(Y,F);
2257             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2258             H                = _mm_setzero_pd();
2259             GMX_MM_TRANSPOSE2_PD(G,H);
2260             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2261             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2262             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq21,FF),_mm_mul_pd(vftabscale,rinv21)));
2263
2264             fscal            = felec;
2265
2266             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2267
2268             /* Update vectorial force */
2269             fix2             = _mm_macc_pd(dx21,fscal,fix2);
2270             fiy2             = _mm_macc_pd(dy21,fscal,fiy2);
2271             fiz2             = _mm_macc_pd(dz21,fscal,fiz2);
2272
2273             fjx1             = _mm_macc_pd(dx21,fscal,fjx1);
2274             fjy1             = _mm_macc_pd(dy21,fscal,fjy1);
2275             fjz1             = _mm_macc_pd(dz21,fscal,fjz1);
2276
2277             /**************************
2278              * CALCULATE INTERACTIONS *
2279              **************************/
2280
2281             r22              = _mm_mul_pd(rsq22,rinv22);
2282
2283             /* Calculate table index by multiplying r with table scale and truncate to integer */
2284             rt               = _mm_mul_pd(r22,vftabscale);
2285             vfitab           = _mm_cvttpd_epi32(rt);
2286 #ifdef __XOP__
2287             vfeps            = _mm_frcz_pd(rt);
2288 #else
2289             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2290 #endif
2291             twovfeps         = _mm_add_pd(vfeps,vfeps);
2292             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2293
2294             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2295             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2296             F                = _mm_setzero_pd();
2297             GMX_MM_TRANSPOSE2_PD(Y,F);
2298             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2299             H                = _mm_setzero_pd();
2300             GMX_MM_TRANSPOSE2_PD(G,H);
2301             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2302             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2303             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq22,FF),_mm_mul_pd(vftabscale,rinv22)));
2304
2305             fscal            = felec;
2306
2307             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2308
2309             /* Update vectorial force */
2310             fix2             = _mm_macc_pd(dx22,fscal,fix2);
2311             fiy2             = _mm_macc_pd(dy22,fscal,fiy2);
2312             fiz2             = _mm_macc_pd(dz22,fscal,fiz2);
2313
2314             fjx2             = _mm_macc_pd(dx22,fscal,fjx2);
2315             fjy2             = _mm_macc_pd(dy22,fscal,fjy2);
2316             fjz2             = _mm_macc_pd(dz22,fscal,fjz2);
2317
2318             /**************************
2319              * CALCULATE INTERACTIONS *
2320              **************************/
2321
2322             r23              = _mm_mul_pd(rsq23,rinv23);
2323
2324             /* Calculate table index by multiplying r with table scale and truncate to integer */
2325             rt               = _mm_mul_pd(r23,vftabscale);
2326             vfitab           = _mm_cvttpd_epi32(rt);
2327 #ifdef __XOP__
2328             vfeps            = _mm_frcz_pd(rt);
2329 #else
2330             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2331 #endif
2332             twovfeps         = _mm_add_pd(vfeps,vfeps);
2333             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2334
2335             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2336             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2337             F                = _mm_setzero_pd();
2338             GMX_MM_TRANSPOSE2_PD(Y,F);
2339             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2340             H                = _mm_setzero_pd();
2341             GMX_MM_TRANSPOSE2_PD(G,H);
2342             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2343             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2344             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq23,FF),_mm_mul_pd(vftabscale,rinv23)));
2345
2346             fscal            = felec;
2347
2348             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2349
2350             /* Update vectorial force */
2351             fix2             = _mm_macc_pd(dx23,fscal,fix2);
2352             fiy2             = _mm_macc_pd(dy23,fscal,fiy2);
2353             fiz2             = _mm_macc_pd(dz23,fscal,fiz2);
2354
2355             fjx3             = _mm_macc_pd(dx23,fscal,fjx3);
2356             fjy3             = _mm_macc_pd(dy23,fscal,fjy3);
2357             fjz3             = _mm_macc_pd(dz23,fscal,fjz3);
2358
2359             /**************************
2360              * CALCULATE INTERACTIONS *
2361              **************************/
2362
2363             r31              = _mm_mul_pd(rsq31,rinv31);
2364
2365             /* Calculate table index by multiplying r with table scale and truncate to integer */
2366             rt               = _mm_mul_pd(r31,vftabscale);
2367             vfitab           = _mm_cvttpd_epi32(rt);
2368 #ifdef __XOP__
2369             vfeps            = _mm_frcz_pd(rt);
2370 #else
2371             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2372 #endif
2373             twovfeps         = _mm_add_pd(vfeps,vfeps);
2374             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2375
2376             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2377             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2378             F                = _mm_setzero_pd();
2379             GMX_MM_TRANSPOSE2_PD(Y,F);
2380             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2381             H                = _mm_setzero_pd();
2382             GMX_MM_TRANSPOSE2_PD(G,H);
2383             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2384             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2385             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq31,FF),_mm_mul_pd(vftabscale,rinv31)));
2386
2387             fscal            = felec;
2388
2389             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2390
2391             /* Update vectorial force */
2392             fix3             = _mm_macc_pd(dx31,fscal,fix3);
2393             fiy3             = _mm_macc_pd(dy31,fscal,fiy3);
2394             fiz3             = _mm_macc_pd(dz31,fscal,fiz3);
2395
2396             fjx1             = _mm_macc_pd(dx31,fscal,fjx1);
2397             fjy1             = _mm_macc_pd(dy31,fscal,fjy1);
2398             fjz1             = _mm_macc_pd(dz31,fscal,fjz1);
2399
2400             /**************************
2401              * CALCULATE INTERACTIONS *
2402              **************************/
2403
2404             r32              = _mm_mul_pd(rsq32,rinv32);
2405
2406             /* Calculate table index by multiplying r with table scale and truncate to integer */
2407             rt               = _mm_mul_pd(r32,vftabscale);
2408             vfitab           = _mm_cvttpd_epi32(rt);
2409 #ifdef __XOP__
2410             vfeps            = _mm_frcz_pd(rt);
2411 #else
2412             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2413 #endif
2414             twovfeps         = _mm_add_pd(vfeps,vfeps);
2415             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2416
2417             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2418             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2419             F                = _mm_setzero_pd();
2420             GMX_MM_TRANSPOSE2_PD(Y,F);
2421             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2422             H                = _mm_setzero_pd();
2423             GMX_MM_TRANSPOSE2_PD(G,H);
2424             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2425             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2426             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq32,FF),_mm_mul_pd(vftabscale,rinv32)));
2427
2428             fscal            = felec;
2429
2430             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2431
2432             /* Update vectorial force */
2433             fix3             = _mm_macc_pd(dx32,fscal,fix3);
2434             fiy3             = _mm_macc_pd(dy32,fscal,fiy3);
2435             fiz3             = _mm_macc_pd(dz32,fscal,fiz3);
2436
2437             fjx2             = _mm_macc_pd(dx32,fscal,fjx2);
2438             fjy2             = _mm_macc_pd(dy32,fscal,fjy2);
2439             fjz2             = _mm_macc_pd(dz32,fscal,fjz2);
2440
2441             /**************************
2442              * CALCULATE INTERACTIONS *
2443              **************************/
2444
2445             r33              = _mm_mul_pd(rsq33,rinv33);
2446
2447             /* Calculate table index by multiplying r with table scale and truncate to integer */
2448             rt               = _mm_mul_pd(r33,vftabscale);
2449             vfitab           = _mm_cvttpd_epi32(rt);
2450 #ifdef __XOP__
2451             vfeps            = _mm_frcz_pd(rt);
2452 #else
2453             vfeps            = _mm_sub_pd(rt,_mm_round_pd(rt, _MM_FROUND_FLOOR));
2454 #endif
2455             twovfeps         = _mm_add_pd(vfeps,vfeps);
2456             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2457
2458             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2459             Y                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) );
2460             F                = _mm_setzero_pd();
2461             GMX_MM_TRANSPOSE2_PD(Y,F);
2462             G                = _mm_load_pd( vftab + _mm_extract_epi32(vfitab,0) +2);
2463             H                = _mm_setzero_pd();
2464             GMX_MM_TRANSPOSE2_PD(G,H);
2465             Fp               = _mm_macc_pd(vfeps,_mm_macc_pd(vfeps,H,G),F);
2466             FF               = _mm_macc_pd(_mm_macc_pd(twovfeps,H,G),vfeps,Fp);
2467             felec            = _mm_xor_pd(signbit,_mm_mul_pd(_mm_mul_pd(qq33,FF),_mm_mul_pd(vftabscale,rinv33)));
2468
2469             fscal            = felec;
2470
2471             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2472
2473             /* Update vectorial force */
2474             fix3             = _mm_macc_pd(dx33,fscal,fix3);
2475             fiy3             = _mm_macc_pd(dy33,fscal,fiy3);
2476             fiz3             = _mm_macc_pd(dz33,fscal,fiz3);
2477
2478             fjx3             = _mm_macc_pd(dx33,fscal,fjx3);
2479             fjy3             = _mm_macc_pd(dy33,fscal,fjy3);
2480             fjz3             = _mm_macc_pd(dz33,fscal,fjz3);
2481
2482             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2483
2484             /* Inner loop uses 432 flops */
2485         }
2486
2487         /* End of innermost loop */
2488
2489         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2490                                               f+i_coord_offset,fshift+i_shift_offset);
2491
2492         /* Increment number of inner iterations */
2493         inneriter                  += j_index_end - j_index_start;
2494
2495         /* Outer loop uses 24 flops */
2496     }
2497
2498     /* Increment number of outer iterations */
2499     outeriter        += nri;
2500
2501     /* Update outer/inner flops */
2502
2503     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*432);
2504 }