src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_single/nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_sse2_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs sse2_single kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_sse2_single.h"
  34 #include "kernelutil_x86_sse2_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_VF_sse2_single
  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_sse2_single
  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,C,D refer to j loop unrolling done with SSE, e.g. for the four 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,jnrC,jnrD;
  61     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  62     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  63     real             shX,shY,shZ,rcutoff_scalar;
  64     real             *shiftvec,*fshift,*x,*f;
  65     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  66     int              vdwioffset0;
  67     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  68     int              vdwioffset1;
  69     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  70     int              vdwioffset2;
  71     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  72     int              vdwioffset3;
  73     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  74     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  75     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  76     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  77     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  78     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  79     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  80     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  81     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  82     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  83     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  84     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  85     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  86     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  87     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  88     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  89     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  90     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  91     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  92     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
  99     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 100     __m128i          vfitab;
 101     __m128i          ifour       = _mm_set1_epi32(4);
 102     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 103     real             *vftab;
 104     __m128           dummy_mask,cutoff_mask;
 105     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 106     __m128           one     = _mm_set1_ps(1.0);
 107     __m128           two     = _mm_set1_ps(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_ps(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_ps(kernel_data->table_elec_vdw->scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 131     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 132     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     jq1              = _mm_set1_ps(charge[inr+1]);
 136     jq2              = _mm_set1_ps(charge[inr+2]);
 137     jq3              = _mm_set1_ps(charge[inr+3]);
 138     vdwjidx0A        = 2*vdwtype[inr+0];
 139     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 140     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 141     qq11             = _mm_mul_ps(iq1,jq1);
 142     qq12             = _mm_mul_ps(iq1,jq2);
 143     qq13             = _mm_mul_ps(iq1,jq3);
 144     qq21             = _mm_mul_ps(iq2,jq1);
 145     qq22             = _mm_mul_ps(iq2,jq2);
 146     qq23             = _mm_mul_ps(iq2,jq3);
 147     qq31             = _mm_mul_ps(iq3,jq1);
 148     qq32             = _mm_mul_ps(iq3,jq2);
 149     qq33             = _mm_mul_ps(iq3,jq3);
 150
 151     /* Avoid stupid compiler warnings */
 152     jnrA = jnrB = jnrC = jnrD = 0;
 153     j_coord_offsetA = 0;
 154     j_coord_offsetB = 0;
 155     j_coord_offsetC = 0;
 156     j_coord_offsetD = 0;
 157
 158     outeriter        = 0;
 159     inneriter        = 0;
 160
 161     /* Start outer loop over neighborlists */
 162     for(iidx=0; iidx<nri; iidx++)
 163     {
 164         /* Load shift vector for this list */
 165         i_shift_offset   = DIM*shiftidx[iidx];
 166         shX              = shiftvec[i_shift_offset+XX];
 167         shY              = shiftvec[i_shift_offset+YY];
 168         shZ              = shiftvec[i_shift_offset+ZZ];
 169
 170         /* Load limits for loop over neighbors */
 171         j_index_start    = jindex[iidx];
 172         j_index_end      = jindex[iidx+1];
 173
 174         /* Get outer coordinate index */
 175         inr              = iinr[iidx];
 176         i_coord_offset   = DIM*inr;
 177
 178         /* Load i particle coords and add shift vector */
 179         ix0              = _mm_set1_ps(shX + x[i_coord_offset+DIM*0+XX]);
 180         iy0              = _mm_set1_ps(shY + x[i_coord_offset+DIM*0+YY]);
 181         iz0              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*0+ZZ]);
 182         ix1              = _mm_set1_ps(shX + x[i_coord_offset+DIM*1+XX]);
 183         iy1              = _mm_set1_ps(shY + x[i_coord_offset+DIM*1+YY]);
 184         iz1              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*1+ZZ]);
 185         ix2              = _mm_set1_ps(shX + x[i_coord_offset+DIM*2+XX]);
 186         iy2              = _mm_set1_ps(shY + x[i_coord_offset+DIM*2+YY]);
 187         iz2              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*2+ZZ]);
 188         ix3              = _mm_set1_ps(shX + x[i_coord_offset+DIM*3+XX]);
 189         iy3              = _mm_set1_ps(shY + x[i_coord_offset+DIM*3+YY]);
 190         iz3              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*3+ZZ]);
 191
 192         fix0             = _mm_setzero_ps();
 193         fiy0             = _mm_setzero_ps();
 194         fiz0             = _mm_setzero_ps();
 195         fix1             = _mm_setzero_ps();
 196         fiy1             = _mm_setzero_ps();
 197         fiz1             = _mm_setzero_ps();
 198         fix2             = _mm_setzero_ps();
 199         fiy2             = _mm_setzero_ps();
 200         fiz2             = _mm_setzero_ps();
 201         fix3             = _mm_setzero_ps();
 202         fiy3             = _mm_setzero_ps();
 203         fiz3             = _mm_setzero_ps();
 204
 205         /* Reset potential sums */
 206         velecsum         = _mm_setzero_ps();
 207         vvdwsum          = _mm_setzero_ps();
 208
 209         /* Start inner kernel loop */
 210         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 211         {
 212
 213             /* Get j neighbor index, and coordinate index */
 214             jnrA             = jjnr[jidx];
 215             jnrB             = jjnr[jidx+1];
 216             jnrC             = jjnr[jidx+2];
 217             jnrD             = jjnr[jidx+3];
 218
 219             j_coord_offsetA  = DIM*jnrA;
 220             j_coord_offsetB  = DIM*jnrB;
 221             j_coord_offsetC  = DIM*jnrC;
 222             j_coord_offsetD  = DIM*jnrD;
 223
 224             /* load j atom coordinates */
 225             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 226                                               x+j_coord_offsetC,x+j_coord_offsetD,
 227                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 228                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 229
 230             /* Calculate displacement vector */
 231             dx00             = _mm_sub_ps(ix0,jx0);
 232             dy00             = _mm_sub_ps(iy0,jy0);
 233             dz00             = _mm_sub_ps(iz0,jz0);
 234             dx11             = _mm_sub_ps(ix1,jx1);
 235             dy11             = _mm_sub_ps(iy1,jy1);
 236             dz11             = _mm_sub_ps(iz1,jz1);
 237             dx12             = _mm_sub_ps(ix1,jx2);
 238             dy12             = _mm_sub_ps(iy1,jy2);
 239             dz12             = _mm_sub_ps(iz1,jz2);
 240             dx13             = _mm_sub_ps(ix1,jx3);
 241             dy13             = _mm_sub_ps(iy1,jy3);
 242             dz13             = _mm_sub_ps(iz1,jz3);
 243             dx21             = _mm_sub_ps(ix2,jx1);
 244             dy21             = _mm_sub_ps(iy2,jy1);
 245             dz21             = _mm_sub_ps(iz2,jz1);
 246             dx22             = _mm_sub_ps(ix2,jx2);
 247             dy22             = _mm_sub_ps(iy2,jy2);
 248             dz22             = _mm_sub_ps(iz2,jz2);
 249             dx23             = _mm_sub_ps(ix2,jx3);
 250             dy23             = _mm_sub_ps(iy2,jy3);
 251             dz23             = _mm_sub_ps(iz2,jz3);
 252             dx31             = _mm_sub_ps(ix3,jx1);
 253             dy31             = _mm_sub_ps(iy3,jy1);
 254             dz31             = _mm_sub_ps(iz3,jz1);
 255             dx32             = _mm_sub_ps(ix3,jx2);
 256             dy32             = _mm_sub_ps(iy3,jy2);
 257             dz32             = _mm_sub_ps(iz3,jz2);
 258             dx33             = _mm_sub_ps(ix3,jx3);
 259             dy33             = _mm_sub_ps(iy3,jy3);
 260             dz33             = _mm_sub_ps(iz3,jz3);
 261
 262             /* Calculate squared distance and things based on it */
 263             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 264             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 265             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 266             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 267             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 268             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 269             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 270             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 271             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 272             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 273
 274             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 275             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 276             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 277             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 278             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 279             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 280             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 281             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 282             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 283             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 284
 285             fjx0             = _mm_setzero_ps();
 286             fjy0             = _mm_setzero_ps();
 287             fjz0             = _mm_setzero_ps();
 288             fjx1             = _mm_setzero_ps();
 289             fjy1             = _mm_setzero_ps();
 290             fjz1             = _mm_setzero_ps();
 291             fjx2             = _mm_setzero_ps();
 292             fjy2             = _mm_setzero_ps();
 293             fjz2             = _mm_setzero_ps();
 294             fjx3             = _mm_setzero_ps();
 295             fjy3             = _mm_setzero_ps();
 296             fjz3             = _mm_setzero_ps();
 297
 298             /**************************
 299              * CALCULATE INTERACTIONS *
 300              **************************/
 301
 302             r00              = _mm_mul_ps(rsq00,rinv00);
 303
 304             /* Calculate table index by multiplying r with table scale and truncate to integer */
 305             rt               = _mm_mul_ps(r00,vftabscale);
 306             vfitab           = _mm_cvttps_epi32(rt);
 307             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 308             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 309
 310             /* CUBIC SPLINE TABLE DISPERSION */
 311             vfitab           = _mm_add_epi32(vfitab,ifour);
 312             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 313             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 314             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 315             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 316             _MM_TRANSPOSE4_PS(Y,F,G,H);
 317             Heps             = _mm_mul_ps(vfeps,H);
 318             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 319             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 320             vvdw6            = _mm_mul_ps(c6_00,VV);
 321             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 322             fvdw6            = _mm_mul_ps(c6_00,FF);
 323
 324             /* CUBIC SPLINE TABLE REPULSION */
 325             vfitab           = _mm_add_epi32(vfitab,ifour);
 326             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 327             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 328             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 329             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 330             _MM_TRANSPOSE4_PS(Y,F,G,H);
 331             Heps             = _mm_mul_ps(vfeps,H);
 332             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 333             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 334             vvdw12           = _mm_mul_ps(c12_00,VV);
 335             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 336             fvdw12           = _mm_mul_ps(c12_00,FF);
 337             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 338             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 339
 340             /* Update potential sum for this i atom from the interaction with this j atom. */
 341             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 342
 343             fscal            = fvdw;
 344
 345             /* Calculate temporary vectorial force */
 346             tx               = _mm_mul_ps(fscal,dx00);
 347             ty               = _mm_mul_ps(fscal,dy00);
 348             tz               = _mm_mul_ps(fscal,dz00);
 349
 350             /* Update vectorial force */
 351             fix0             = _mm_add_ps(fix0,tx);
 352             fiy0             = _mm_add_ps(fiy0,ty);
 353             fiz0             = _mm_add_ps(fiz0,tz);
 354
 355             fjx0             = _mm_add_ps(fjx0,tx);
 356             fjy0             = _mm_add_ps(fjy0,ty);
 357             fjz0             = _mm_add_ps(fjz0,tz);
 358
 359             /**************************
 360              * CALCULATE INTERACTIONS *
 361              **************************/
 362
 363             r11              = _mm_mul_ps(rsq11,rinv11);
 364
 365             /* Calculate table index by multiplying r with table scale and truncate to integer */
 366             rt               = _mm_mul_ps(r11,vftabscale);
 367             vfitab           = _mm_cvttps_epi32(rt);
 368             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 369             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 370
 371             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 372             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 373             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 374             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 375             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 376             _MM_TRANSPOSE4_PS(Y,F,G,H);
 377             Heps             = _mm_mul_ps(vfeps,H);
 378             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 379             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 380             velec            = _mm_mul_ps(qq11,VV);
 381             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 382             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 383
 384             /* Update potential sum for this i atom from the interaction with this j atom. */
 385             velecsum         = _mm_add_ps(velecsum,velec);
 386
 387             fscal            = felec;
 388
 389             /* Calculate temporary vectorial force */
 390             tx               = _mm_mul_ps(fscal,dx11);
 391             ty               = _mm_mul_ps(fscal,dy11);
 392             tz               = _mm_mul_ps(fscal,dz11);
 393
 394             /* Update vectorial force */
 395             fix1             = _mm_add_ps(fix1,tx);
 396             fiy1             = _mm_add_ps(fiy1,ty);
 397             fiz1             = _mm_add_ps(fiz1,tz);
 398
 399             fjx1             = _mm_add_ps(fjx1,tx);
 400             fjy1             = _mm_add_ps(fjy1,ty);
 401             fjz1             = _mm_add_ps(fjz1,tz);
 402
 403             /**************************
 404              * CALCULATE INTERACTIONS *
 405              **************************/
 406
 407             r12              = _mm_mul_ps(rsq12,rinv12);
 408
 409             /* Calculate table index by multiplying r with table scale and truncate to integer */
 410             rt               = _mm_mul_ps(r12,vftabscale);
 411             vfitab           = _mm_cvttps_epi32(rt);
 412             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 413             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 414
 415             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 416             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 417             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 418             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 419             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 420             _MM_TRANSPOSE4_PS(Y,F,G,H);
 421             Heps             = _mm_mul_ps(vfeps,H);
 422             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 423             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 424             velec            = _mm_mul_ps(qq12,VV);
 425             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 426             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 427
 428             /* Update potential sum for this i atom from the interaction with this j atom. */
 429             velecsum         = _mm_add_ps(velecsum,velec);
 430
 431             fscal            = felec;
 432
 433             /* Calculate temporary vectorial force */
 434             tx               = _mm_mul_ps(fscal,dx12);
 435             ty               = _mm_mul_ps(fscal,dy12);
 436             tz               = _mm_mul_ps(fscal,dz12);
 437
 438             /* Update vectorial force */
 439             fix1             = _mm_add_ps(fix1,tx);
 440             fiy1             = _mm_add_ps(fiy1,ty);
 441             fiz1             = _mm_add_ps(fiz1,tz);
 442
 443             fjx2             = _mm_add_ps(fjx2,tx);
 444             fjy2             = _mm_add_ps(fjy2,ty);
 445             fjz2             = _mm_add_ps(fjz2,tz);
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             r13              = _mm_mul_ps(rsq13,rinv13);
 452
 453             /* Calculate table index by multiplying r with table scale and truncate to integer */
 454             rt               = _mm_mul_ps(r13,vftabscale);
 455             vfitab           = _mm_cvttps_epi32(rt);
 456             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 457             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 458
 459             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 460             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 461             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 462             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 463             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 464             _MM_TRANSPOSE4_PS(Y,F,G,H);
 465             Heps             = _mm_mul_ps(vfeps,H);
 466             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 467             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 468             velec            = _mm_mul_ps(qq13,VV);
 469             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 470             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
 471
 472             /* Update potential sum for this i atom from the interaction with this j atom. */
 473             velecsum         = _mm_add_ps(velecsum,velec);
 474
 475             fscal            = felec;
 476
 477             /* Calculate temporary vectorial force */
 478             tx               = _mm_mul_ps(fscal,dx13);
 479             ty               = _mm_mul_ps(fscal,dy13);
 480             tz               = _mm_mul_ps(fscal,dz13);
 481
 482             /* Update vectorial force */
 483             fix1             = _mm_add_ps(fix1,tx);
 484             fiy1             = _mm_add_ps(fiy1,ty);
 485             fiz1             = _mm_add_ps(fiz1,tz);
 486
 487             fjx3             = _mm_add_ps(fjx3,tx);
 488             fjy3             = _mm_add_ps(fjy3,ty);
 489             fjz3             = _mm_add_ps(fjz3,tz);
 490
 491             /**************************
 492              * CALCULATE INTERACTIONS *
 493              **************************/
 494
 495             r21              = _mm_mul_ps(rsq21,rinv21);
 496
 497             /* Calculate table index by multiplying r with table scale and truncate to integer */
 498             rt               = _mm_mul_ps(r21,vftabscale);
 499             vfitab           = _mm_cvttps_epi32(rt);
 500             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 501             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 502
 503             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 504             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 505             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 506             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 507             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 508             _MM_TRANSPOSE4_PS(Y,F,G,H);
 509             Heps             = _mm_mul_ps(vfeps,H);
 510             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 511             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 512             velec            = _mm_mul_ps(qq21,VV);
 513             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 514             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
 515
 516             /* Update potential sum for this i atom from the interaction with this j atom. */
 517             velecsum         = _mm_add_ps(velecsum,velec);
 518
 519             fscal            = felec;
 520
 521             /* Calculate temporary vectorial force */
 522             tx               = _mm_mul_ps(fscal,dx21);
 523             ty               = _mm_mul_ps(fscal,dy21);
 524             tz               = _mm_mul_ps(fscal,dz21);
 525
 526             /* Update vectorial force */
 527             fix2             = _mm_add_ps(fix2,tx);
 528             fiy2             = _mm_add_ps(fiy2,ty);
 529             fiz2             = _mm_add_ps(fiz2,tz);
 530
 531             fjx1             = _mm_add_ps(fjx1,tx);
 532             fjy1             = _mm_add_ps(fjy1,ty);
 533             fjz1             = _mm_add_ps(fjz1,tz);
 534
 535             /**************************
 536              * CALCULATE INTERACTIONS *
 537              **************************/
 538
 539             r22              = _mm_mul_ps(rsq22,rinv22);
 540
 541             /* Calculate table index by multiplying r with table scale and truncate to integer */
 542             rt               = _mm_mul_ps(r22,vftabscale);
 543             vfitab           = _mm_cvttps_epi32(rt);
 544             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 545             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 546
 547             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 548             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 549             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 550             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 551             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 552             _MM_TRANSPOSE4_PS(Y,F,G,H);
 553             Heps             = _mm_mul_ps(vfeps,H);
 554             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 555             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 556             velec            = _mm_mul_ps(qq22,VV);
 557             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 558             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
 559
 560             /* Update potential sum for this i atom from the interaction with this j atom. */
 561             velecsum         = _mm_add_ps(velecsum,velec);
 562
 563             fscal            = felec;
 564
 565             /* Calculate temporary vectorial force */
 566             tx               = _mm_mul_ps(fscal,dx22);
 567             ty               = _mm_mul_ps(fscal,dy22);
 568             tz               = _mm_mul_ps(fscal,dz22);
 569
 570             /* Update vectorial force */
 571             fix2             = _mm_add_ps(fix2,tx);
 572             fiy2             = _mm_add_ps(fiy2,ty);
 573             fiz2             = _mm_add_ps(fiz2,tz);
 574
 575             fjx2             = _mm_add_ps(fjx2,tx);
 576             fjy2             = _mm_add_ps(fjy2,ty);
 577             fjz2             = _mm_add_ps(fjz2,tz);
 578
 579             /**************************
 580              * CALCULATE INTERACTIONS *
 581              **************************/
 582
 583             r23              = _mm_mul_ps(rsq23,rinv23);
 584
 585             /* Calculate table index by multiplying r with table scale and truncate to integer */
 586             rt               = _mm_mul_ps(r23,vftabscale);
 587             vfitab           = _mm_cvttps_epi32(rt);
 588             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 589             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 590
 591             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 592             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 593             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 594             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 595             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 596             _MM_TRANSPOSE4_PS(Y,F,G,H);
 597             Heps             = _mm_mul_ps(vfeps,H);
 598             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 599             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 600             velec            = _mm_mul_ps(qq23,VV);
 601             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 602             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
 603
 604             /* Update potential sum for this i atom from the interaction with this j atom. */
 605             velecsum         = _mm_add_ps(velecsum,velec);
 606
 607             fscal            = felec;
 608
 609             /* Calculate temporary vectorial force */
 610             tx               = _mm_mul_ps(fscal,dx23);
 611             ty               = _mm_mul_ps(fscal,dy23);
 612             tz               = _mm_mul_ps(fscal,dz23);
 613
 614             /* Update vectorial force */
 615             fix2             = _mm_add_ps(fix2,tx);
 616             fiy2             = _mm_add_ps(fiy2,ty);
 617             fiz2             = _mm_add_ps(fiz2,tz);
 618
 619             fjx3             = _mm_add_ps(fjx3,tx);
 620             fjy3             = _mm_add_ps(fjy3,ty);
 621             fjz3             = _mm_add_ps(fjz3,tz);
 622
 623             /**************************
 624              * CALCULATE INTERACTIONS *
 625              **************************/
 626
 627             r31              = _mm_mul_ps(rsq31,rinv31);
 628
 629             /* Calculate table index by multiplying r with table scale and truncate to integer */
 630             rt               = _mm_mul_ps(r31,vftabscale);
 631             vfitab           = _mm_cvttps_epi32(rt);
 632             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 633             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 634
 635             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 636             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 637             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 638             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 639             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 640             _MM_TRANSPOSE4_PS(Y,F,G,H);
 641             Heps             = _mm_mul_ps(vfeps,H);
 642             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 643             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 644             velec            = _mm_mul_ps(qq31,VV);
 645             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 646             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
 647
 648             /* Update potential sum for this i atom from the interaction with this j atom. */
 649             velecsum         = _mm_add_ps(velecsum,velec);
 650
 651             fscal            = felec;
 652
 653             /* Calculate temporary vectorial force */
 654             tx               = _mm_mul_ps(fscal,dx31);
 655             ty               = _mm_mul_ps(fscal,dy31);
 656             tz               = _mm_mul_ps(fscal,dz31);
 657
 658             /* Update vectorial force */
 659             fix3             = _mm_add_ps(fix3,tx);
 660             fiy3             = _mm_add_ps(fiy3,ty);
 661             fiz3             = _mm_add_ps(fiz3,tz);
 662
 663             fjx1             = _mm_add_ps(fjx1,tx);
 664             fjy1             = _mm_add_ps(fjy1,ty);
 665             fjz1             = _mm_add_ps(fjz1,tz);
 666
 667             /**************************
 668              * CALCULATE INTERACTIONS *
 669              **************************/
 670
 671             r32              = _mm_mul_ps(rsq32,rinv32);
 672
 673             /* Calculate table index by multiplying r with table scale and truncate to integer */
 674             rt               = _mm_mul_ps(r32,vftabscale);
 675             vfitab           = _mm_cvttps_epi32(rt);
 676             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 677             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 678
 679             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 680             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 681             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 682             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 683             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 684             _MM_TRANSPOSE4_PS(Y,F,G,H);
 685             Heps             = _mm_mul_ps(vfeps,H);
 686             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 687             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 688             velec            = _mm_mul_ps(qq32,VV);
 689             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 690             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
 691
 692             /* Update potential sum for this i atom from the interaction with this j atom. */
 693             velecsum         = _mm_add_ps(velecsum,velec);
 694
 695             fscal            = felec;
 696
 697             /* Calculate temporary vectorial force */
 698             tx               = _mm_mul_ps(fscal,dx32);
 699             ty               = _mm_mul_ps(fscal,dy32);
 700             tz               = _mm_mul_ps(fscal,dz32);
 701
 702             /* Update vectorial force */
 703             fix3             = _mm_add_ps(fix3,tx);
 704             fiy3             = _mm_add_ps(fiy3,ty);
 705             fiz3             = _mm_add_ps(fiz3,tz);
 706
 707             fjx2             = _mm_add_ps(fjx2,tx);
 708             fjy2             = _mm_add_ps(fjy2,ty);
 709             fjz2             = _mm_add_ps(fjz2,tz);
 710
 711             /**************************
 712              * CALCULATE INTERACTIONS *
 713              **************************/
 714
 715             r33              = _mm_mul_ps(rsq33,rinv33);
 716
 717             /* Calculate table index by multiplying r with table scale and truncate to integer */
 718             rt               = _mm_mul_ps(r33,vftabscale);
 719             vfitab           = _mm_cvttps_epi32(rt);
 720             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 721             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 722
 723             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 724             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 725             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 726             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 727             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 728             _MM_TRANSPOSE4_PS(Y,F,G,H);
 729             Heps             = _mm_mul_ps(vfeps,H);
 730             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 731             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 732             velec            = _mm_mul_ps(qq33,VV);
 733             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 734             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
 735
 736             /* Update potential sum for this i atom from the interaction with this j atom. */
 737             velecsum         = _mm_add_ps(velecsum,velec);
 738
 739             fscal            = felec;
 740
 741             /* Calculate temporary vectorial force */
 742             tx               = _mm_mul_ps(fscal,dx33);
 743             ty               = _mm_mul_ps(fscal,dy33);
 744             tz               = _mm_mul_ps(fscal,dz33);
 745
 746             /* Update vectorial force */
 747             fix3             = _mm_add_ps(fix3,tx);
 748             fiy3             = _mm_add_ps(fiy3,ty);
 749             fiz3             = _mm_add_ps(fiz3,tz);
 750
 751             fjx3             = _mm_add_ps(fjx3,tx);
 752             fjy3             = _mm_add_ps(fjy3,ty);
 753             fjz3             = _mm_add_ps(fjz3,tz);
 754
 755             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
 756                                                    f+j_coord_offsetC,f+j_coord_offsetD,
 757                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 758                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 759
 760             /* Inner loop uses 446 flops */
 761         }
 762
 763         if(jidx<j_index_end)
 764         {
 765
 766             /* Get j neighbor index, and coordinate index */
 767             jnrA             = jjnr[jidx];
 768             jnrB             = jjnr[jidx+1];
 769             jnrC             = jjnr[jidx+2];
 770             jnrD             = jjnr[jidx+3];
 771
 772             /* Sign of each element will be negative for non-real atoms.
 773              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 774              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 775              */
 776             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 777             jnrA       = (jnrA>=0) ? jnrA : 0;
 778             jnrB       = (jnrB>=0) ? jnrB : 0;
 779             jnrC       = (jnrC>=0) ? jnrC : 0;
 780             jnrD       = (jnrD>=0) ? jnrD : 0;
 781
 782             j_coord_offsetA  = DIM*jnrA;
 783             j_coord_offsetB  = DIM*jnrB;
 784             j_coord_offsetC  = DIM*jnrC;
 785             j_coord_offsetD  = DIM*jnrD;
 786
 787             /* load j atom coordinates */
 788             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 789                                               x+j_coord_offsetC,x+j_coord_offsetD,
 790                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 791                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 792
 793             /* Calculate displacement vector */
 794             dx00             = _mm_sub_ps(ix0,jx0);
 795             dy00             = _mm_sub_ps(iy0,jy0);
 796             dz00             = _mm_sub_ps(iz0,jz0);
 797             dx11             = _mm_sub_ps(ix1,jx1);
 798             dy11             = _mm_sub_ps(iy1,jy1);
 799             dz11             = _mm_sub_ps(iz1,jz1);
 800             dx12             = _mm_sub_ps(ix1,jx2);
 801             dy12             = _mm_sub_ps(iy1,jy2);
 802             dz12             = _mm_sub_ps(iz1,jz2);
 803             dx13             = _mm_sub_ps(ix1,jx3);
 804             dy13             = _mm_sub_ps(iy1,jy3);
 805             dz13             = _mm_sub_ps(iz1,jz3);
 806             dx21             = _mm_sub_ps(ix2,jx1);
 807             dy21             = _mm_sub_ps(iy2,jy1);
 808             dz21             = _mm_sub_ps(iz2,jz1);
 809             dx22             = _mm_sub_ps(ix2,jx2);
 810             dy22             = _mm_sub_ps(iy2,jy2);
 811             dz22             = _mm_sub_ps(iz2,jz2);
 812             dx23             = _mm_sub_ps(ix2,jx3);
 813             dy23             = _mm_sub_ps(iy2,jy3);
 814             dz23             = _mm_sub_ps(iz2,jz3);
 815             dx31             = _mm_sub_ps(ix3,jx1);
 816             dy31             = _mm_sub_ps(iy3,jy1);
 817             dz31             = _mm_sub_ps(iz3,jz1);
 818             dx32             = _mm_sub_ps(ix3,jx2);
 819             dy32             = _mm_sub_ps(iy3,jy2);
 820             dz32             = _mm_sub_ps(iz3,jz2);
 821             dx33             = _mm_sub_ps(ix3,jx3);
 822             dy33             = _mm_sub_ps(iy3,jy3);
 823             dz33             = _mm_sub_ps(iz3,jz3);
 824
 825             /* Calculate squared distance and things based on it */
 826             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 827             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 828             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 829             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 830             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 831             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 832             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 833             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 834             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 835             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 836
 837             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 838             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 839             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 840             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 841             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 842             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 843             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 844             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 845             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 846             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 847
 848             fjx0             = _mm_setzero_ps();
 849             fjy0             = _mm_setzero_ps();
 850             fjz0             = _mm_setzero_ps();
 851             fjx1             = _mm_setzero_ps();
 852             fjy1             = _mm_setzero_ps();
 853             fjz1             = _mm_setzero_ps();
 854             fjx2             = _mm_setzero_ps();
 855             fjy2             = _mm_setzero_ps();
 856             fjz2             = _mm_setzero_ps();
 857             fjx3             = _mm_setzero_ps();
 858             fjy3             = _mm_setzero_ps();
 859             fjz3             = _mm_setzero_ps();
 860
 861             /**************************
 862              * CALCULATE INTERACTIONS *
 863              **************************/
 864
 865             r00              = _mm_mul_ps(rsq00,rinv00);
 866             r00              = _mm_andnot_ps(dummy_mask,r00);
 867
 868             /* Calculate table index by multiplying r with table scale and truncate to integer */
 869             rt               = _mm_mul_ps(r00,vftabscale);
 870             vfitab           = _mm_cvttps_epi32(rt);
 871             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 872             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 873
 874             /* CUBIC SPLINE TABLE DISPERSION */
 875             vfitab           = _mm_add_epi32(vfitab,ifour);
 876             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 877             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 878             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 879             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 880             _MM_TRANSPOSE4_PS(Y,F,G,H);
 881             Heps             = _mm_mul_ps(vfeps,H);
 882             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 883             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 884             vvdw6            = _mm_mul_ps(c6_00,VV);
 885             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 886             fvdw6            = _mm_mul_ps(c6_00,FF);
 887
 888             /* CUBIC SPLINE TABLE REPULSION */
 889             vfitab           = _mm_add_epi32(vfitab,ifour);
 890             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 891             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 892             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 893             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 894             _MM_TRANSPOSE4_PS(Y,F,G,H);
 895             Heps             = _mm_mul_ps(vfeps,H);
 896             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 897             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 898             vvdw12           = _mm_mul_ps(c12_00,VV);
 899             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 900             fvdw12           = _mm_mul_ps(c12_00,FF);
 901             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 902             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 903
 904             /* Update potential sum for this i atom from the interaction with this j atom. */
 905             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 906             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 907
 908             fscal            = fvdw;
 909
 910             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 911
 912             /* Calculate temporary vectorial force */
 913             tx               = _mm_mul_ps(fscal,dx00);
 914             ty               = _mm_mul_ps(fscal,dy00);
 915             tz               = _mm_mul_ps(fscal,dz00);
 916
 917             /* Update vectorial force */
 918             fix0             = _mm_add_ps(fix0,tx);
 919             fiy0             = _mm_add_ps(fiy0,ty);
 920             fiz0             = _mm_add_ps(fiz0,tz);
 921
 922             fjx0             = _mm_add_ps(fjx0,tx);
 923             fjy0             = _mm_add_ps(fjy0,ty);
 924             fjz0             = _mm_add_ps(fjz0,tz);
 925
 926             /**************************
 927              * CALCULATE INTERACTIONS *
 928              **************************/
 929
 930             r11              = _mm_mul_ps(rsq11,rinv11);
 931             r11              = _mm_andnot_ps(dummy_mask,r11);
 932
 933             /* Calculate table index by multiplying r with table scale and truncate to integer */
 934             rt               = _mm_mul_ps(r11,vftabscale);
 935             vfitab           = _mm_cvttps_epi32(rt);
 936             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 937             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 938
 939             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 940             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 941             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 942             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 943             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 944             _MM_TRANSPOSE4_PS(Y,F,G,H);
 945             Heps             = _mm_mul_ps(vfeps,H);
 946             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 947             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 948             velec            = _mm_mul_ps(qq11,VV);
 949             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 950             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 951
 952             /* Update potential sum for this i atom from the interaction with this j atom. */
 953             velec            = _mm_andnot_ps(dummy_mask,velec);
 954             velecsum         = _mm_add_ps(velecsum,velec);
 955
 956             fscal            = felec;
 957
 958             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 959
 960             /* Calculate temporary vectorial force */
 961             tx               = _mm_mul_ps(fscal,dx11);
 962             ty               = _mm_mul_ps(fscal,dy11);
 963             tz               = _mm_mul_ps(fscal,dz11);
 964
 965             /* Update vectorial force */
 966             fix1             = _mm_add_ps(fix1,tx);
 967             fiy1             = _mm_add_ps(fiy1,ty);
 968             fiz1             = _mm_add_ps(fiz1,tz);
 969
 970             fjx1             = _mm_add_ps(fjx1,tx);
 971             fjy1             = _mm_add_ps(fjy1,ty);
 972             fjz1             = _mm_add_ps(fjz1,tz);
 973
 974             /**************************
 975              * CALCULATE INTERACTIONS *
 976              **************************/
 977
 978             r12              = _mm_mul_ps(rsq12,rinv12);
 979             r12              = _mm_andnot_ps(dummy_mask,r12);
 980
 981             /* Calculate table index by multiplying r with table scale and truncate to integer */
 982             rt               = _mm_mul_ps(r12,vftabscale);
 983             vfitab           = _mm_cvttps_epi32(rt);
 984             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 985             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 986
 987             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 988             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 989             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 990             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 991             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 992             _MM_TRANSPOSE4_PS(Y,F,G,H);
 993             Heps             = _mm_mul_ps(vfeps,H);
 994             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 995             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 996             velec            = _mm_mul_ps(qq12,VV);
 997             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 998             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 999
1000             /* Update potential sum for this i atom from the interaction with this j atom. */
1001             velec            = _mm_andnot_ps(dummy_mask,velec);
1002             velecsum         = _mm_add_ps(velecsum,velec);
1003
1004             fscal            = felec;
1005
1006             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1007
1008             /* Calculate temporary vectorial force */
1009             tx               = _mm_mul_ps(fscal,dx12);
1010             ty               = _mm_mul_ps(fscal,dy12);
1011             tz               = _mm_mul_ps(fscal,dz12);
1012
1013             /* Update vectorial force */
1014             fix1             = _mm_add_ps(fix1,tx);
1015             fiy1             = _mm_add_ps(fiy1,ty);
1016             fiz1             = _mm_add_ps(fiz1,tz);
1017
1018             fjx2             = _mm_add_ps(fjx2,tx);
1019             fjy2             = _mm_add_ps(fjy2,ty);
1020             fjz2             = _mm_add_ps(fjz2,tz);
1021
1022             /**************************
1023              * CALCULATE INTERACTIONS *
1024              **************************/
1025
1026             r13              = _mm_mul_ps(rsq13,rinv13);
1027             r13              = _mm_andnot_ps(dummy_mask,r13);
1028
1029             /* Calculate table index by multiplying r with table scale and truncate to integer */
1030             rt               = _mm_mul_ps(r13,vftabscale);
1031             vfitab           = _mm_cvttps_epi32(rt);
1032             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1033             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1034
1035             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1036             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1037             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1038             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1039             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1040             _MM_TRANSPOSE4_PS(Y,F,G,H);
1041             Heps             = _mm_mul_ps(vfeps,H);
1042             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1043             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1044             velec            = _mm_mul_ps(qq13,VV);
1045             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1046             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1047
1048             /* Update potential sum for this i atom from the interaction with this j atom. */
1049             velec            = _mm_andnot_ps(dummy_mask,velec);
1050             velecsum         = _mm_add_ps(velecsum,velec);
1051
1052             fscal            = felec;
1053
1054             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1055
1056             /* Calculate temporary vectorial force */
1057             tx               = _mm_mul_ps(fscal,dx13);
1058             ty               = _mm_mul_ps(fscal,dy13);
1059             tz               = _mm_mul_ps(fscal,dz13);
1060
1061             /* Update vectorial force */
1062             fix1             = _mm_add_ps(fix1,tx);
1063             fiy1             = _mm_add_ps(fiy1,ty);
1064             fiz1             = _mm_add_ps(fiz1,tz);
1065
1066             fjx3             = _mm_add_ps(fjx3,tx);
1067             fjy3             = _mm_add_ps(fjy3,ty);
1068             fjz3             = _mm_add_ps(fjz3,tz);
1069
1070             /**************************
1071              * CALCULATE INTERACTIONS *
1072              **************************/
1073
1074             r21              = _mm_mul_ps(rsq21,rinv21);
1075             r21              = _mm_andnot_ps(dummy_mask,r21);
1076
1077             /* Calculate table index by multiplying r with table scale and truncate to integer */
1078             rt               = _mm_mul_ps(r21,vftabscale);
1079             vfitab           = _mm_cvttps_epi32(rt);
1080             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1081             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1082
1083             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1084             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1085             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1086             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1087             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1088             _MM_TRANSPOSE4_PS(Y,F,G,H);
1089             Heps             = _mm_mul_ps(vfeps,H);
1090             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1091             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1092             velec            = _mm_mul_ps(qq21,VV);
1093             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1094             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1095
1096             /* Update potential sum for this i atom from the interaction with this j atom. */
1097             velec            = _mm_andnot_ps(dummy_mask,velec);
1098             velecsum         = _mm_add_ps(velecsum,velec);
1099
1100             fscal            = felec;
1101
1102             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1103
1104             /* Calculate temporary vectorial force */
1105             tx               = _mm_mul_ps(fscal,dx21);
1106             ty               = _mm_mul_ps(fscal,dy21);
1107             tz               = _mm_mul_ps(fscal,dz21);
1108
1109             /* Update vectorial force */
1110             fix2             = _mm_add_ps(fix2,tx);
1111             fiy2             = _mm_add_ps(fiy2,ty);
1112             fiz2             = _mm_add_ps(fiz2,tz);
1113
1114             fjx1             = _mm_add_ps(fjx1,tx);
1115             fjy1             = _mm_add_ps(fjy1,ty);
1116             fjz1             = _mm_add_ps(fjz1,tz);
1117
1118             /**************************
1119              * CALCULATE INTERACTIONS *
1120              **************************/
1121
1122             r22              = _mm_mul_ps(rsq22,rinv22);
1123             r22              = _mm_andnot_ps(dummy_mask,r22);
1124
1125             /* Calculate table index by multiplying r with table scale and truncate to integer */
1126             rt               = _mm_mul_ps(r22,vftabscale);
1127             vfitab           = _mm_cvttps_epi32(rt);
1128             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1129             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1130
1131             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1132             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1133             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1134             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1135             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1136             _MM_TRANSPOSE4_PS(Y,F,G,H);
1137             Heps             = _mm_mul_ps(vfeps,H);
1138             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1139             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1140             velec            = _mm_mul_ps(qq22,VV);
1141             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1142             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1143
1144             /* Update potential sum for this i atom from the interaction with this j atom. */
1145             velec            = _mm_andnot_ps(dummy_mask,velec);
1146             velecsum         = _mm_add_ps(velecsum,velec);
1147
1148             fscal            = felec;
1149
1150             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1151
1152             /* Calculate temporary vectorial force */
1153             tx               = _mm_mul_ps(fscal,dx22);
1154             ty               = _mm_mul_ps(fscal,dy22);
1155             tz               = _mm_mul_ps(fscal,dz22);
1156
1157             /* Update vectorial force */
1158             fix2             = _mm_add_ps(fix2,tx);
1159             fiy2             = _mm_add_ps(fiy2,ty);
1160             fiz2             = _mm_add_ps(fiz2,tz);
1161
1162             fjx2             = _mm_add_ps(fjx2,tx);
1163             fjy2             = _mm_add_ps(fjy2,ty);
1164             fjz2             = _mm_add_ps(fjz2,tz);
1165
1166             /**************************
1167              * CALCULATE INTERACTIONS *
1168              **************************/
1169
1170             r23              = _mm_mul_ps(rsq23,rinv23);
1171             r23              = _mm_andnot_ps(dummy_mask,r23);
1172
1173             /* Calculate table index by multiplying r with table scale and truncate to integer */
1174             rt               = _mm_mul_ps(r23,vftabscale);
1175             vfitab           = _mm_cvttps_epi32(rt);
1176             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1177             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1178
1179             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1180             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1181             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1182             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1183             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1184             _MM_TRANSPOSE4_PS(Y,F,G,H);
1185             Heps             = _mm_mul_ps(vfeps,H);
1186             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1187             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1188             velec            = _mm_mul_ps(qq23,VV);
1189             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1190             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1191
1192             /* Update potential sum for this i atom from the interaction with this j atom. */
1193             velec            = _mm_andnot_ps(dummy_mask,velec);
1194             velecsum         = _mm_add_ps(velecsum,velec);
1195
1196             fscal            = felec;
1197
1198             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1199
1200             /* Calculate temporary vectorial force */
1201             tx               = _mm_mul_ps(fscal,dx23);
1202             ty               = _mm_mul_ps(fscal,dy23);
1203             tz               = _mm_mul_ps(fscal,dz23);
1204
1205             /* Update vectorial force */
1206             fix2             = _mm_add_ps(fix2,tx);
1207             fiy2             = _mm_add_ps(fiy2,ty);
1208             fiz2             = _mm_add_ps(fiz2,tz);
1209
1210             fjx3             = _mm_add_ps(fjx3,tx);
1211             fjy3             = _mm_add_ps(fjy3,ty);
1212             fjz3             = _mm_add_ps(fjz3,tz);
1213
1214             /**************************
1215              * CALCULATE INTERACTIONS *
1216              **************************/
1217
1218             r31              = _mm_mul_ps(rsq31,rinv31);
1219             r31              = _mm_andnot_ps(dummy_mask,r31);
1220
1221             /* Calculate table index by multiplying r with table scale and truncate to integer */
1222             rt               = _mm_mul_ps(r31,vftabscale);
1223             vfitab           = _mm_cvttps_epi32(rt);
1224             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1225             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1226
1227             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1228             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1229             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1230             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1231             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1232             _MM_TRANSPOSE4_PS(Y,F,G,H);
1233             Heps             = _mm_mul_ps(vfeps,H);
1234             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1235             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1236             velec            = _mm_mul_ps(qq31,VV);
1237             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1238             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1239
1240             /* Update potential sum for this i atom from the interaction with this j atom. */
1241             velec            = _mm_andnot_ps(dummy_mask,velec);
1242             velecsum         = _mm_add_ps(velecsum,velec);
1243
1244             fscal            = felec;
1245
1246             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1247
1248             /* Calculate temporary vectorial force */
1249             tx               = _mm_mul_ps(fscal,dx31);
1250             ty               = _mm_mul_ps(fscal,dy31);
1251             tz               = _mm_mul_ps(fscal,dz31);
1252
1253             /* Update vectorial force */
1254             fix3             = _mm_add_ps(fix3,tx);
1255             fiy3             = _mm_add_ps(fiy3,ty);
1256             fiz3             = _mm_add_ps(fiz3,tz);
1257
1258             fjx1             = _mm_add_ps(fjx1,tx);
1259             fjy1             = _mm_add_ps(fjy1,ty);
1260             fjz1             = _mm_add_ps(fjz1,tz);
1261
1262             /**************************
1263              * CALCULATE INTERACTIONS *
1264              **************************/
1265
1266             r32              = _mm_mul_ps(rsq32,rinv32);
1267             r32              = _mm_andnot_ps(dummy_mask,r32);
1268
1269             /* Calculate table index by multiplying r with table scale and truncate to integer */
1270             rt               = _mm_mul_ps(r32,vftabscale);
1271             vfitab           = _mm_cvttps_epi32(rt);
1272             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
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_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1277             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1278             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1279             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1280             _MM_TRANSPOSE4_PS(Y,F,G,H);
1281             Heps             = _mm_mul_ps(vfeps,H);
1282             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1283             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1284             velec            = _mm_mul_ps(qq32,VV);
1285             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1286             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1287
1288             /* Update potential sum for this i atom from the interaction with this j atom. */
1289             velec            = _mm_andnot_ps(dummy_mask,velec);
1290             velecsum         = _mm_add_ps(velecsum,velec);
1291
1292             fscal            = felec;
1293
1294             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1295
1296             /* Calculate temporary vectorial force */
1297             tx               = _mm_mul_ps(fscal,dx32);
1298             ty               = _mm_mul_ps(fscal,dy32);
1299             tz               = _mm_mul_ps(fscal,dz32);
1300
1301             /* Update vectorial force */
1302             fix3             = _mm_add_ps(fix3,tx);
1303             fiy3             = _mm_add_ps(fiy3,ty);
1304             fiz3             = _mm_add_ps(fiz3,tz);
1305
1306             fjx2             = _mm_add_ps(fjx2,tx);
1307             fjy2             = _mm_add_ps(fjy2,ty);
1308             fjz2             = _mm_add_ps(fjz2,tz);
1309
1310             /**************************
1311              * CALCULATE INTERACTIONS *
1312              **************************/
1313
1314             r33              = _mm_mul_ps(rsq33,rinv33);
1315             r33              = _mm_andnot_ps(dummy_mask,r33);
1316
1317             /* Calculate table index by multiplying r with table scale and truncate to integer */
1318             rt               = _mm_mul_ps(r33,vftabscale);
1319             vfitab           = _mm_cvttps_epi32(rt);
1320             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1321             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1322
1323             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1324             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1325             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1326             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1327             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1328             _MM_TRANSPOSE4_PS(Y,F,G,H);
1329             Heps             = _mm_mul_ps(vfeps,H);
1330             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1331             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1332             velec            = _mm_mul_ps(qq33,VV);
1333             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1334             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
1335
1336             /* Update potential sum for this i atom from the interaction with this j atom. */
1337             velec            = _mm_andnot_ps(dummy_mask,velec);
1338             velecsum         = _mm_add_ps(velecsum,velec);
1339
1340             fscal            = felec;
1341
1342             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1343
1344             /* Calculate temporary vectorial force */
1345             tx               = _mm_mul_ps(fscal,dx33);
1346             ty               = _mm_mul_ps(fscal,dy33);
1347             tz               = _mm_mul_ps(fscal,dz33);
1348
1349             /* Update vectorial force */
1350             fix3             = _mm_add_ps(fix3,tx);
1351             fiy3             = _mm_add_ps(fiy3,ty);
1352             fiz3             = _mm_add_ps(fiz3,tz);
1353
1354             fjx3             = _mm_add_ps(fjx3,tx);
1355             fjy3             = _mm_add_ps(fjy3,ty);
1356             fjz3             = _mm_add_ps(fjz3,tz);
1357
1358             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
1359                                                    f+j_coord_offsetC,f+j_coord_offsetD,
1360                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1361                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1362
1363             /* Inner loop uses 456 flops */
1364         }
1365
1366         /* End of innermost loop */
1367
1368         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1369                                               f+i_coord_offset,fshift+i_shift_offset);
1370
1371         ggid                        = gid[iidx];
1372         /* Update potential energies */
1373         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1374         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1375
1376         /* Increment number of inner iterations */
1377         inneriter                  += j_index_end - j_index_start;
1378
1379         /* Outer loop uses 38 flops */
1380     }
1381
1382     /* Increment number of outer iterations */
1383     outeriter        += nri;
1384
1385     /* Update outer/inner flops */
1386
1387     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*38 + inneriter*456);
1388 }
1389 /*
1390  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_sse2_single
1391  * Electrostatics interaction: CubicSplineTable
1392  * VdW interaction:            CubicSplineTable
1393  * Geometry:                   Water4-Water4
1394  * Calculate force/pot:        Force
1395  */
1396 void
1397 nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4_F_sse2_single
1398                     (t_nblist * gmx_restrict                nlist,
1399                      rvec * gmx_restrict                    xx,
1400                      rvec * gmx_restrict                    ff,
1401                      t_forcerec * gmx_restrict              fr,
1402                      t_mdatoms * gmx_restrict               mdatoms,
1403                      nb_kernel_data_t * gmx_restrict        kernel_data,
1404                      t_nrnb * gmx_restrict                  nrnb)
1405 {
1406     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1407      * just 0 for non-waters.
1408      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1409      * jnr indices corresponding to data put in the four positions in the SIMD register.
1410      */
1411     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1412     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1413     int              jnrA,jnrB,jnrC,jnrD;
1414     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1415     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1416     real             shX,shY,shZ,rcutoff_scalar;
1417     real             *shiftvec,*fshift,*x,*f;
1418     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1419     int              vdwioffset0;
1420     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1421     int              vdwioffset1;
1422     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1423     int              vdwioffset2;
1424     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1425     int              vdwioffset3;
1426     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1427     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1428     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1429     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1430     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1431     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1432     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1433     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1434     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1435     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1436     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1437     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1438     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1439     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1440     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1441     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1442     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1443     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1444     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1445     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1446     real             *charge;
1447     int              nvdwtype;
1448     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1449     int              *vdwtype;
1450     real             *vdwparam;
1451     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1452     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1453     __m128i          vfitab;
1454     __m128i          ifour       = _mm_set1_epi32(4);
1455     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1456     real             *vftab;
1457     __m128           dummy_mask,cutoff_mask;
1458     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1459     __m128           one     = _mm_set1_ps(1.0);
1460     __m128           two     = _mm_set1_ps(2.0);
1461     x                = xx[0];
1462     f                = ff[0];
1463
1464     nri              = nlist->nri;
1465     iinr             = nlist->iinr;
1466     jindex           = nlist->jindex;
1467     jjnr             = nlist->jjnr;
1468     shiftidx         = nlist->shift;
1469     gid              = nlist->gid;
1470     shiftvec         = fr->shift_vec[0];
1471     fshift           = fr->fshift[0];
1472     facel            = _mm_set1_ps(fr->epsfac);
1473     charge           = mdatoms->chargeA;
1474     nvdwtype         = fr->ntype;
1475     vdwparam         = fr->nbfp;
1476     vdwtype          = mdatoms->typeA;
1477
1478     vftab            = kernel_data->table_elec_vdw->data;
1479     vftabscale       = _mm_set1_ps(kernel_data->table_elec_vdw->scale);
1480
1481     /* Setup water-specific parameters */
1482     inr              = nlist->iinr[0];
1483     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1484     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1485     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1486     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1487
1488     jq1              = _mm_set1_ps(charge[inr+1]);
1489     jq2              = _mm_set1_ps(charge[inr+2]);
1490     jq3              = _mm_set1_ps(charge[inr+3]);
1491     vdwjidx0A        = 2*vdwtype[inr+0];
1492     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1493     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1494     qq11             = _mm_mul_ps(iq1,jq1);
1495     qq12             = _mm_mul_ps(iq1,jq2);
1496     qq13             = _mm_mul_ps(iq1,jq3);
1497     qq21             = _mm_mul_ps(iq2,jq1);
1498     qq22             = _mm_mul_ps(iq2,jq2);
1499     qq23             = _mm_mul_ps(iq2,jq3);
1500     qq31             = _mm_mul_ps(iq3,jq1);
1501     qq32             = _mm_mul_ps(iq3,jq2);
1502     qq33             = _mm_mul_ps(iq3,jq3);
1503
1504     /* Avoid stupid compiler warnings */
1505     jnrA = jnrB = jnrC = jnrD = 0;
1506     j_coord_offsetA = 0;
1507     j_coord_offsetB = 0;
1508     j_coord_offsetC = 0;
1509     j_coord_offsetD = 0;
1510
1511     outeriter        = 0;
1512     inneriter        = 0;
1513
1514     /* Start outer loop over neighborlists */
1515     for(iidx=0; iidx<nri; iidx++)
1516     {
1517         /* Load shift vector for this list */
1518         i_shift_offset   = DIM*shiftidx[iidx];
1519         shX              = shiftvec[i_shift_offset+XX];
1520         shY              = shiftvec[i_shift_offset+YY];
1521         shZ              = shiftvec[i_shift_offset+ZZ];
1522
1523         /* Load limits for loop over neighbors */
1524         j_index_start    = jindex[iidx];
1525         j_index_end      = jindex[iidx+1];
1526
1527         /* Get outer coordinate index */
1528         inr              = iinr[iidx];
1529         i_coord_offset   = DIM*inr;
1530
1531         /* Load i particle coords and add shift vector */
1532         ix0              = _mm_set1_ps(shX + x[i_coord_offset+DIM*0+XX]);
1533         iy0              = _mm_set1_ps(shY + x[i_coord_offset+DIM*0+YY]);
1534         iz0              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*0+ZZ]);
1535         ix1              = _mm_set1_ps(shX + x[i_coord_offset+DIM*1+XX]);
1536         iy1              = _mm_set1_ps(shY + x[i_coord_offset+DIM*1+YY]);
1537         iz1              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*1+ZZ]);
1538         ix2              = _mm_set1_ps(shX + x[i_coord_offset+DIM*2+XX]);
1539         iy2              = _mm_set1_ps(shY + x[i_coord_offset+DIM*2+YY]);
1540         iz2              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*2+ZZ]);
1541         ix3              = _mm_set1_ps(shX + x[i_coord_offset+DIM*3+XX]);
1542         iy3              = _mm_set1_ps(shY + x[i_coord_offset+DIM*3+YY]);
1543         iz3              = _mm_set1_ps(shZ + x[i_coord_offset+DIM*3+ZZ]);
1544
1545         fix0             = _mm_setzero_ps();
1546         fiy0             = _mm_setzero_ps();
1547         fiz0             = _mm_setzero_ps();
1548         fix1             = _mm_setzero_ps();
1549         fiy1             = _mm_setzero_ps();
1550         fiz1             = _mm_setzero_ps();
1551         fix2             = _mm_setzero_ps();
1552         fiy2             = _mm_setzero_ps();
1553         fiz2             = _mm_setzero_ps();
1554         fix3             = _mm_setzero_ps();
1555         fiy3             = _mm_setzero_ps();
1556         fiz3             = _mm_setzero_ps();
1557
1558         /* Start inner kernel loop */
1559         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1560         {
1561
1562             /* Get j neighbor index, and coordinate index */
1563             jnrA             = jjnr[jidx];
1564             jnrB             = jjnr[jidx+1];
1565             jnrC             = jjnr[jidx+2];
1566             jnrD             = jjnr[jidx+3];
1567
1568             j_coord_offsetA  = DIM*jnrA;
1569             j_coord_offsetB  = DIM*jnrB;
1570             j_coord_offsetC  = DIM*jnrC;
1571             j_coord_offsetD  = DIM*jnrD;
1572
1573             /* load j atom coordinates */
1574             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1575                                               x+j_coord_offsetC,x+j_coord_offsetD,
1576                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1577                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1578
1579             /* Calculate displacement vector */
1580             dx00             = _mm_sub_ps(ix0,jx0);
1581             dy00             = _mm_sub_ps(iy0,jy0);
1582             dz00             = _mm_sub_ps(iz0,jz0);
1583             dx11             = _mm_sub_ps(ix1,jx1);
1584             dy11             = _mm_sub_ps(iy1,jy1);
1585             dz11             = _mm_sub_ps(iz1,jz1);
1586             dx12             = _mm_sub_ps(ix1,jx2);
1587             dy12             = _mm_sub_ps(iy1,jy2);
1588             dz12             = _mm_sub_ps(iz1,jz2);
1589             dx13             = _mm_sub_ps(ix1,jx3);
1590             dy13             = _mm_sub_ps(iy1,jy3);
1591             dz13             = _mm_sub_ps(iz1,jz3);
1592             dx21             = _mm_sub_ps(ix2,jx1);
1593             dy21             = _mm_sub_ps(iy2,jy1);
1594             dz21             = _mm_sub_ps(iz2,jz1);
1595             dx22             = _mm_sub_ps(ix2,jx2);
1596             dy22             = _mm_sub_ps(iy2,jy2);
1597             dz22             = _mm_sub_ps(iz2,jz2);
1598             dx23             = _mm_sub_ps(ix2,jx3);
1599             dy23             = _mm_sub_ps(iy2,jy3);
1600             dz23             = _mm_sub_ps(iz2,jz3);
1601             dx31             = _mm_sub_ps(ix3,jx1);
1602             dy31             = _mm_sub_ps(iy3,jy1);
1603             dz31             = _mm_sub_ps(iz3,jz1);
1604             dx32             = _mm_sub_ps(ix3,jx2);
1605             dy32             = _mm_sub_ps(iy3,jy2);
1606             dz32             = _mm_sub_ps(iz3,jz2);
1607             dx33             = _mm_sub_ps(ix3,jx3);
1608             dy33             = _mm_sub_ps(iy3,jy3);
1609             dz33             = _mm_sub_ps(iz3,jz3);
1610
1611             /* Calculate squared distance and things based on it */
1612             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1613             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1614             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1615             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1616             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1617             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1618             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1619             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1620             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1621             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1622
1623             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1624             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1625             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1626             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1627             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1628             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1629             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1630             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1631             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1632             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1633
1634             fjx0             = _mm_setzero_ps();
1635             fjy0             = _mm_setzero_ps();
1636             fjz0             = _mm_setzero_ps();
1637             fjx1             = _mm_setzero_ps();
1638             fjy1             = _mm_setzero_ps();
1639             fjz1             = _mm_setzero_ps();
1640             fjx2             = _mm_setzero_ps();
1641             fjy2             = _mm_setzero_ps();
1642             fjz2             = _mm_setzero_ps();
1643             fjx3             = _mm_setzero_ps();
1644             fjy3             = _mm_setzero_ps();
1645             fjz3             = _mm_setzero_ps();
1646
1647             /**************************
1648              * CALCULATE INTERACTIONS *
1649              **************************/
1650
1651             r00              = _mm_mul_ps(rsq00,rinv00);
1652
1653             /* Calculate table index by multiplying r with table scale and truncate to integer */
1654             rt               = _mm_mul_ps(r00,vftabscale);
1655             vfitab           = _mm_cvttps_epi32(rt);
1656             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1657             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1658
1659             /* CUBIC SPLINE TABLE DISPERSION */
1660             vfitab           = _mm_add_epi32(vfitab,ifour);
1661             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1662             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1663             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1664             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1665             _MM_TRANSPOSE4_PS(Y,F,G,H);
1666             Heps             = _mm_mul_ps(vfeps,H);
1667             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1668             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1669             fvdw6            = _mm_mul_ps(c6_00,FF);
1670
1671             /* CUBIC SPLINE TABLE REPULSION */
1672             vfitab           = _mm_add_epi32(vfitab,ifour);
1673             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1674             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1675             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1676             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1677             _MM_TRANSPOSE4_PS(Y,F,G,H);
1678             Heps             = _mm_mul_ps(vfeps,H);
1679             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1680             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1681             fvdw12           = _mm_mul_ps(c12_00,FF);
1682             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1683
1684             fscal            = fvdw;
1685
1686             /* Calculate temporary vectorial force */
1687             tx               = _mm_mul_ps(fscal,dx00);
1688             ty               = _mm_mul_ps(fscal,dy00);
1689             tz               = _mm_mul_ps(fscal,dz00);
1690
1691             /* Update vectorial force */
1692             fix0             = _mm_add_ps(fix0,tx);
1693             fiy0             = _mm_add_ps(fiy0,ty);
1694             fiz0             = _mm_add_ps(fiz0,tz);
1695
1696             fjx0             = _mm_add_ps(fjx0,tx);
1697             fjy0             = _mm_add_ps(fjy0,ty);
1698             fjz0             = _mm_add_ps(fjz0,tz);
1699
1700             /**************************
1701              * CALCULATE INTERACTIONS *
1702              **************************/
1703
1704             r11              = _mm_mul_ps(rsq11,rinv11);
1705
1706             /* Calculate table index by multiplying r with table scale and truncate to integer */
1707             rt               = _mm_mul_ps(r11,vftabscale);
1708             vfitab           = _mm_cvttps_epi32(rt);
1709             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1710             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1711
1712             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1713             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1714             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1715             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1716             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1717             _MM_TRANSPOSE4_PS(Y,F,G,H);
1718             Heps             = _mm_mul_ps(vfeps,H);
1719             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1720             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1721             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
1722
1723             fscal            = felec;
1724
1725             /* Calculate temporary vectorial force */
1726             tx               = _mm_mul_ps(fscal,dx11);
1727             ty               = _mm_mul_ps(fscal,dy11);
1728             tz               = _mm_mul_ps(fscal,dz11);
1729
1730             /* Update vectorial force */
1731             fix1             = _mm_add_ps(fix1,tx);
1732             fiy1             = _mm_add_ps(fiy1,ty);
1733             fiz1             = _mm_add_ps(fiz1,tz);
1734
1735             fjx1             = _mm_add_ps(fjx1,tx);
1736             fjy1             = _mm_add_ps(fjy1,ty);
1737             fjz1             = _mm_add_ps(fjz1,tz);
1738
1739             /**************************
1740              * CALCULATE INTERACTIONS *
1741              **************************/
1742
1743             r12              = _mm_mul_ps(rsq12,rinv12);
1744
1745             /* Calculate table index by multiplying r with table scale and truncate to integer */
1746             rt               = _mm_mul_ps(r12,vftabscale);
1747             vfitab           = _mm_cvttps_epi32(rt);
1748             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1749             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1750
1751             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1752             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1753             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1754             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1755             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1756             _MM_TRANSPOSE4_PS(Y,F,G,H);
1757             Heps             = _mm_mul_ps(vfeps,H);
1758             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1759             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1760             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
1761
1762             fscal            = felec;
1763
1764             /* Calculate temporary vectorial force */
1765             tx               = _mm_mul_ps(fscal,dx12);
1766             ty               = _mm_mul_ps(fscal,dy12);
1767             tz               = _mm_mul_ps(fscal,dz12);
1768
1769             /* Update vectorial force */
1770             fix1             = _mm_add_ps(fix1,tx);
1771             fiy1             = _mm_add_ps(fiy1,ty);
1772             fiz1             = _mm_add_ps(fiz1,tz);
1773
1774             fjx2             = _mm_add_ps(fjx2,tx);
1775             fjy2             = _mm_add_ps(fjy2,ty);
1776             fjz2             = _mm_add_ps(fjz2,tz);
1777
1778             /**************************
1779              * CALCULATE INTERACTIONS *
1780              **************************/
1781
1782             r13              = _mm_mul_ps(rsq13,rinv13);
1783
1784             /* Calculate table index by multiplying r with table scale and truncate to integer */
1785             rt               = _mm_mul_ps(r13,vftabscale);
1786             vfitab           = _mm_cvttps_epi32(rt);
1787             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1788             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1789
1790             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1791             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1792             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1793             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1794             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1795             _MM_TRANSPOSE4_PS(Y,F,G,H);
1796             Heps             = _mm_mul_ps(vfeps,H);
1797             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1798             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1799             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1800
1801             fscal            = felec;
1802
1803             /* Calculate temporary vectorial force */
1804             tx               = _mm_mul_ps(fscal,dx13);
1805             ty               = _mm_mul_ps(fscal,dy13);
1806             tz               = _mm_mul_ps(fscal,dz13);
1807
1808             /* Update vectorial force */
1809             fix1             = _mm_add_ps(fix1,tx);
1810             fiy1             = _mm_add_ps(fiy1,ty);
1811             fiz1             = _mm_add_ps(fiz1,tz);
1812
1813             fjx3             = _mm_add_ps(fjx3,tx);
1814             fjy3             = _mm_add_ps(fjy3,ty);
1815             fjz3             = _mm_add_ps(fjz3,tz);
1816
1817             /**************************
1818              * CALCULATE INTERACTIONS *
1819              **************************/
1820
1821             r21              = _mm_mul_ps(rsq21,rinv21);
1822
1823             /* Calculate table index by multiplying r with table scale and truncate to integer */
1824             rt               = _mm_mul_ps(r21,vftabscale);
1825             vfitab           = _mm_cvttps_epi32(rt);
1826             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1827             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1828
1829             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1830             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1831             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1832             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1833             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1834             _MM_TRANSPOSE4_PS(Y,F,G,H);
1835             Heps             = _mm_mul_ps(vfeps,H);
1836             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1837             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1838             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1839
1840             fscal            = felec;
1841
1842             /* Calculate temporary vectorial force */
1843             tx               = _mm_mul_ps(fscal,dx21);
1844             ty               = _mm_mul_ps(fscal,dy21);
1845             tz               = _mm_mul_ps(fscal,dz21);
1846
1847             /* Update vectorial force */
1848             fix2             = _mm_add_ps(fix2,tx);
1849             fiy2             = _mm_add_ps(fiy2,ty);
1850             fiz2             = _mm_add_ps(fiz2,tz);
1851
1852             fjx1             = _mm_add_ps(fjx1,tx);
1853             fjy1             = _mm_add_ps(fjy1,ty);
1854             fjz1             = _mm_add_ps(fjz1,tz);
1855
1856             /**************************
1857              * CALCULATE INTERACTIONS *
1858              **************************/
1859
1860             r22              = _mm_mul_ps(rsq22,rinv22);
1861
1862             /* Calculate table index by multiplying r with table scale and truncate to integer */
1863             rt               = _mm_mul_ps(r22,vftabscale);
1864             vfitab           = _mm_cvttps_epi32(rt);
1865             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1866             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1867
1868             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1869             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1870             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1871             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1872             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1873             _MM_TRANSPOSE4_PS(Y,F,G,H);
1874             Heps             = _mm_mul_ps(vfeps,H);
1875             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1876             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1877             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1878
1879             fscal            = felec;
1880
1881             /* Calculate temporary vectorial force */
1882             tx               = _mm_mul_ps(fscal,dx22);
1883             ty               = _mm_mul_ps(fscal,dy22);
1884             tz               = _mm_mul_ps(fscal,dz22);
1885
1886             /* Update vectorial force */
1887             fix2             = _mm_add_ps(fix2,tx);
1888             fiy2             = _mm_add_ps(fiy2,ty);
1889             fiz2             = _mm_add_ps(fiz2,tz);
1890
1891             fjx2             = _mm_add_ps(fjx2,tx);
1892             fjy2             = _mm_add_ps(fjy2,ty);
1893             fjz2             = _mm_add_ps(fjz2,tz);
1894
1895             /**************************
1896              * CALCULATE INTERACTIONS *
1897              **************************/
1898
1899             r23              = _mm_mul_ps(rsq23,rinv23);
1900
1901             /* Calculate table index by multiplying r with table scale and truncate to integer */
1902             rt               = _mm_mul_ps(r23,vftabscale);
1903             vfitab           = _mm_cvttps_epi32(rt);
1904             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1905             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1906
1907             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1908             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1909             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1910             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1911             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1912             _MM_TRANSPOSE4_PS(Y,F,G,H);
1913             Heps             = _mm_mul_ps(vfeps,H);
1914             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1915             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1916             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1917
1918             fscal            = felec;
1919
1920             /* Calculate temporary vectorial force */
1921             tx               = _mm_mul_ps(fscal,dx23);
1922             ty               = _mm_mul_ps(fscal,dy23);
1923             tz               = _mm_mul_ps(fscal,dz23);
1924
1925             /* Update vectorial force */
1926             fix2             = _mm_add_ps(fix2,tx);
1927             fiy2             = _mm_add_ps(fiy2,ty);
1928             fiz2             = _mm_add_ps(fiz2,tz);
1929
1930             fjx3             = _mm_add_ps(fjx3,tx);
1931             fjy3             = _mm_add_ps(fjy3,ty);
1932             fjz3             = _mm_add_ps(fjz3,tz);
1933
1934             /**************************
1935              * CALCULATE INTERACTIONS *
1936              **************************/
1937
1938             r31              = _mm_mul_ps(rsq31,rinv31);
1939
1940             /* Calculate table index by multiplying r with table scale and truncate to integer */
1941             rt               = _mm_mul_ps(r31,vftabscale);
1942             vfitab           = _mm_cvttps_epi32(rt);
1943             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1944             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1945
1946             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1947             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1948             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1949             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1950             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1951             _MM_TRANSPOSE4_PS(Y,F,G,H);
1952             Heps             = _mm_mul_ps(vfeps,H);
1953             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1954             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1955             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1956
1957             fscal            = felec;
1958
1959             /* Calculate temporary vectorial force */
1960             tx               = _mm_mul_ps(fscal,dx31);
1961             ty               = _mm_mul_ps(fscal,dy31);
1962             tz               = _mm_mul_ps(fscal,dz31);
1963
1964             /* Update vectorial force */
1965             fix3             = _mm_add_ps(fix3,tx);
1966             fiy3             = _mm_add_ps(fiy3,ty);
1967             fiz3             = _mm_add_ps(fiz3,tz);
1968
1969             fjx1             = _mm_add_ps(fjx1,tx);
1970             fjy1             = _mm_add_ps(fjy1,ty);
1971             fjz1             = _mm_add_ps(fjz1,tz);
1972
1973             /**************************
1974              * CALCULATE INTERACTIONS *
1975              **************************/
1976
1977             r32              = _mm_mul_ps(rsq32,rinv32);
1978
1979             /* Calculate table index by multiplying r with table scale and truncate to integer */
1980             rt               = _mm_mul_ps(r32,vftabscale);
1981             vfitab           = _mm_cvttps_epi32(rt);
1982             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1983             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1984
1985             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1986             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1987             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1988             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1989             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1990             _MM_TRANSPOSE4_PS(Y,F,G,H);
1991             Heps             = _mm_mul_ps(vfeps,H);
1992             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1993             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1994             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1995
1996             fscal            = felec;
1997
1998             /* Calculate temporary vectorial force */
1999             tx               = _mm_mul_ps(fscal,dx32);
2000             ty               = _mm_mul_ps(fscal,dy32);
2001             tz               = _mm_mul_ps(fscal,dz32);
2002
2003             /* Update vectorial force */
2004             fix3             = _mm_add_ps(fix3,tx);
2005             fiy3             = _mm_add_ps(fiy3,ty);
2006             fiz3             = _mm_add_ps(fiz3,tz);
2007
2008             fjx2             = _mm_add_ps(fjx2,tx);
2009             fjy2             = _mm_add_ps(fjy2,ty);
2010             fjz2             = _mm_add_ps(fjz2,tz);
2011
2012             /**************************
2013              * CALCULATE INTERACTIONS *
2014              **************************/
2015
2016             r33              = _mm_mul_ps(rsq33,rinv33);
2017
2018             /* Calculate table index by multiplying r with table scale and truncate to integer */
2019             rt               = _mm_mul_ps(r33,vftabscale);
2020             vfitab           = _mm_cvttps_epi32(rt);
2021             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2022             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2023
2024             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2025             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2026             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2027             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2028             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2029             _MM_TRANSPOSE4_PS(Y,F,G,H);
2030             Heps             = _mm_mul_ps(vfeps,H);
2031             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2032             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2033             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2034
2035             fscal            = felec;
2036
2037             /* Calculate temporary vectorial force */
2038             tx               = _mm_mul_ps(fscal,dx33);
2039             ty               = _mm_mul_ps(fscal,dy33);
2040             tz               = _mm_mul_ps(fscal,dz33);
2041
2042             /* Update vectorial force */
2043             fix3             = _mm_add_ps(fix3,tx);
2044             fiy3             = _mm_add_ps(fiy3,ty);
2045             fiz3             = _mm_add_ps(fiz3,tz);
2046
2047             fjx3             = _mm_add_ps(fjx3,tx);
2048             fjy3             = _mm_add_ps(fjy3,ty);
2049             fjz3             = _mm_add_ps(fjz3,tz);
2050
2051             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
2052                                                    f+j_coord_offsetC,f+j_coord_offsetD,
2053                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2054                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2055
2056             /* Inner loop uses 402 flops */
2057         }
2058
2059         if(jidx<j_index_end)
2060         {
2061
2062             /* Get j neighbor index, and coordinate index */
2063             jnrA             = jjnr[jidx];
2064             jnrB             = jjnr[jidx+1];
2065             jnrC             = jjnr[jidx+2];
2066             jnrD             = jjnr[jidx+3];
2067
2068             /* Sign of each element will be negative for non-real atoms.
2069              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2070              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2071              */
2072             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2073             jnrA       = (jnrA>=0) ? jnrA : 0;
2074             jnrB       = (jnrB>=0) ? jnrB : 0;
2075             jnrC       = (jnrC>=0) ? jnrC : 0;
2076             jnrD       = (jnrD>=0) ? jnrD : 0;
2077
2078             j_coord_offsetA  = DIM*jnrA;
2079             j_coord_offsetB  = DIM*jnrB;
2080             j_coord_offsetC  = DIM*jnrC;
2081             j_coord_offsetD  = DIM*jnrD;
2082
2083             /* load j atom coordinates */
2084             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2085                                               x+j_coord_offsetC,x+j_coord_offsetD,
2086                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2087                                               &jy2,&jz2,&jx3,&jy3,&jz3);
2088
2089             /* Calculate displacement vector */
2090             dx00             = _mm_sub_ps(ix0,jx0);
2091             dy00             = _mm_sub_ps(iy0,jy0);
2092             dz00             = _mm_sub_ps(iz0,jz0);
2093             dx11             = _mm_sub_ps(ix1,jx1);
2094             dy11             = _mm_sub_ps(iy1,jy1);
2095             dz11             = _mm_sub_ps(iz1,jz1);
2096             dx12             = _mm_sub_ps(ix1,jx2);
2097             dy12             = _mm_sub_ps(iy1,jy2);
2098             dz12             = _mm_sub_ps(iz1,jz2);
2099             dx13             = _mm_sub_ps(ix1,jx3);
2100             dy13             = _mm_sub_ps(iy1,jy3);
2101             dz13             = _mm_sub_ps(iz1,jz3);
2102             dx21             = _mm_sub_ps(ix2,jx1);
2103             dy21             = _mm_sub_ps(iy2,jy1);
2104             dz21             = _mm_sub_ps(iz2,jz1);
2105             dx22             = _mm_sub_ps(ix2,jx2);
2106             dy22             = _mm_sub_ps(iy2,jy2);
2107             dz22             = _mm_sub_ps(iz2,jz2);
2108             dx23             = _mm_sub_ps(ix2,jx3);
2109             dy23             = _mm_sub_ps(iy2,jy3);
2110             dz23             = _mm_sub_ps(iz2,jz3);
2111             dx31             = _mm_sub_ps(ix3,jx1);
2112             dy31             = _mm_sub_ps(iy3,jy1);
2113             dz31             = _mm_sub_ps(iz3,jz1);
2114             dx32             = _mm_sub_ps(ix3,jx2);
2115             dy32             = _mm_sub_ps(iy3,jy2);
2116             dz32             = _mm_sub_ps(iz3,jz2);
2117             dx33             = _mm_sub_ps(ix3,jx3);
2118             dy33             = _mm_sub_ps(iy3,jy3);
2119             dz33             = _mm_sub_ps(iz3,jz3);
2120
2121             /* Calculate squared distance and things based on it */
2122             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2123             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2124             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2125             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2126             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2127             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2128             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2129             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2130             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2131             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2132
2133             rinv00           = gmx_mm_invsqrt_ps(rsq00);
2134             rinv11           = gmx_mm_invsqrt_ps(rsq11);
2135             rinv12           = gmx_mm_invsqrt_ps(rsq12);
2136             rinv13           = gmx_mm_invsqrt_ps(rsq13);
2137             rinv21           = gmx_mm_invsqrt_ps(rsq21);
2138             rinv22           = gmx_mm_invsqrt_ps(rsq22);
2139             rinv23           = gmx_mm_invsqrt_ps(rsq23);
2140             rinv31           = gmx_mm_invsqrt_ps(rsq31);
2141             rinv32           = gmx_mm_invsqrt_ps(rsq32);
2142             rinv33           = gmx_mm_invsqrt_ps(rsq33);
2143
2144             fjx0             = _mm_setzero_ps();
2145             fjy0             = _mm_setzero_ps();
2146             fjz0             = _mm_setzero_ps();
2147             fjx1             = _mm_setzero_ps();
2148             fjy1             = _mm_setzero_ps();
2149             fjz1             = _mm_setzero_ps();
2150             fjx2             = _mm_setzero_ps();
2151             fjy2             = _mm_setzero_ps();
2152             fjz2             = _mm_setzero_ps();
2153             fjx3             = _mm_setzero_ps();
2154             fjy3             = _mm_setzero_ps();
2155             fjz3             = _mm_setzero_ps();
2156
2157             /**************************
2158              * CALCULATE INTERACTIONS *
2159              **************************/
2160
2161             r00              = _mm_mul_ps(rsq00,rinv00);
2162             r00              = _mm_andnot_ps(dummy_mask,r00);
2163
2164             /* Calculate table index by multiplying r with table scale and truncate to integer */
2165             rt               = _mm_mul_ps(r00,vftabscale);
2166             vfitab           = _mm_cvttps_epi32(rt);
2167             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2168             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2169
2170             /* CUBIC SPLINE TABLE DISPERSION */
2171             vfitab           = _mm_add_epi32(vfitab,ifour);
2172             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2173             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2174             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2175             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2176             _MM_TRANSPOSE4_PS(Y,F,G,H);
2177             Heps             = _mm_mul_ps(vfeps,H);
2178             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2179             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2180             fvdw6            = _mm_mul_ps(c6_00,FF);
2181
2182             /* CUBIC SPLINE TABLE REPULSION */
2183             vfitab           = _mm_add_epi32(vfitab,ifour);
2184             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2185             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2186             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2187             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2188             _MM_TRANSPOSE4_PS(Y,F,G,H);
2189             Heps             = _mm_mul_ps(vfeps,H);
2190             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2191             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2192             fvdw12           = _mm_mul_ps(c12_00,FF);
2193             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
2194
2195             fscal            = fvdw;
2196
2197             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2198
2199             /* Calculate temporary vectorial force */
2200             tx               = _mm_mul_ps(fscal,dx00);
2201             ty               = _mm_mul_ps(fscal,dy00);
2202             tz               = _mm_mul_ps(fscal,dz00);
2203
2204             /* Update vectorial force */
2205             fix0             = _mm_add_ps(fix0,tx);
2206             fiy0             = _mm_add_ps(fiy0,ty);
2207             fiz0             = _mm_add_ps(fiz0,tz);
2208
2209             fjx0             = _mm_add_ps(fjx0,tx);
2210             fjy0             = _mm_add_ps(fjy0,ty);
2211             fjz0             = _mm_add_ps(fjz0,tz);
2212
2213             /**************************
2214              * CALCULATE INTERACTIONS *
2215              **************************/
2216
2217             r11              = _mm_mul_ps(rsq11,rinv11);
2218             r11              = _mm_andnot_ps(dummy_mask,r11);
2219
2220             /* Calculate table index by multiplying r with table scale and truncate to integer */
2221             rt               = _mm_mul_ps(r11,vftabscale);
2222             vfitab           = _mm_cvttps_epi32(rt);
2223             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2224             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2225
2226             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2227             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2228             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2229             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2230             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2231             _MM_TRANSPOSE4_PS(Y,F,G,H);
2232             Heps             = _mm_mul_ps(vfeps,H);
2233             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2234             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2235             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
2236
2237             fscal            = felec;
2238
2239             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2240
2241             /* Calculate temporary vectorial force */
2242             tx               = _mm_mul_ps(fscal,dx11);
2243             ty               = _mm_mul_ps(fscal,dy11);
2244             tz               = _mm_mul_ps(fscal,dz11);
2245
2246             /* Update vectorial force */
2247             fix1             = _mm_add_ps(fix1,tx);
2248             fiy1             = _mm_add_ps(fiy1,ty);
2249             fiz1             = _mm_add_ps(fiz1,tz);
2250
2251             fjx1             = _mm_add_ps(fjx1,tx);
2252             fjy1             = _mm_add_ps(fjy1,ty);
2253             fjz1             = _mm_add_ps(fjz1,tz);
2254
2255             /**************************
2256              * CALCULATE INTERACTIONS *
2257              **************************/
2258
2259             r12              = _mm_mul_ps(rsq12,rinv12);
2260             r12              = _mm_andnot_ps(dummy_mask,r12);
2261
2262             /* Calculate table index by multiplying r with table scale and truncate to integer */
2263             rt               = _mm_mul_ps(r12,vftabscale);
2264             vfitab           = _mm_cvttps_epi32(rt);
2265             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2266             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2267
2268             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2269             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2270             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2271             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2272             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2273             _MM_TRANSPOSE4_PS(Y,F,G,H);
2274             Heps             = _mm_mul_ps(vfeps,H);
2275             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2276             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2277             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
2278
2279             fscal            = felec;
2280
2281             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2282
2283             /* Calculate temporary vectorial force */
2284             tx               = _mm_mul_ps(fscal,dx12);
2285             ty               = _mm_mul_ps(fscal,dy12);
2286             tz               = _mm_mul_ps(fscal,dz12);
2287
2288             /* Update vectorial force */
2289             fix1             = _mm_add_ps(fix1,tx);
2290             fiy1             = _mm_add_ps(fiy1,ty);
2291             fiz1             = _mm_add_ps(fiz1,tz);
2292
2293             fjx2             = _mm_add_ps(fjx2,tx);
2294             fjy2             = _mm_add_ps(fjy2,ty);
2295             fjz2             = _mm_add_ps(fjz2,tz);
2296
2297             /**************************
2298              * CALCULATE INTERACTIONS *
2299              **************************/
2300
2301             r13              = _mm_mul_ps(rsq13,rinv13);
2302             r13              = _mm_andnot_ps(dummy_mask,r13);
2303
2304             /* Calculate table index by multiplying r with table scale and truncate to integer */
2305             rt               = _mm_mul_ps(r13,vftabscale);
2306             vfitab           = _mm_cvttps_epi32(rt);
2307             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2308             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2309
2310             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2311             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2312             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2313             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2314             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2315             _MM_TRANSPOSE4_PS(Y,F,G,H);
2316             Heps             = _mm_mul_ps(vfeps,H);
2317             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2318             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2319             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
2320
2321             fscal            = felec;
2322
2323             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2324
2325             /* Calculate temporary vectorial force */
2326             tx               = _mm_mul_ps(fscal,dx13);
2327             ty               = _mm_mul_ps(fscal,dy13);
2328             tz               = _mm_mul_ps(fscal,dz13);
2329
2330             /* Update vectorial force */
2331             fix1             = _mm_add_ps(fix1,tx);
2332             fiy1             = _mm_add_ps(fiy1,ty);
2333             fiz1             = _mm_add_ps(fiz1,tz);
2334
2335             fjx3             = _mm_add_ps(fjx3,tx);
2336             fjy3             = _mm_add_ps(fjy3,ty);
2337             fjz3             = _mm_add_ps(fjz3,tz);
2338
2339             /**************************
2340              * CALCULATE INTERACTIONS *
2341              **************************/
2342
2343             r21              = _mm_mul_ps(rsq21,rinv21);
2344             r21              = _mm_andnot_ps(dummy_mask,r21);
2345
2346             /* Calculate table index by multiplying r with table scale and truncate to integer */
2347             rt               = _mm_mul_ps(r21,vftabscale);
2348             vfitab           = _mm_cvttps_epi32(rt);
2349             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2350             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2351
2352             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2353             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2354             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2355             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2356             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2357             _MM_TRANSPOSE4_PS(Y,F,G,H);
2358             Heps             = _mm_mul_ps(vfeps,H);
2359             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2360             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2361             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
2362
2363             fscal            = felec;
2364
2365             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2366
2367             /* Calculate temporary vectorial force */
2368             tx               = _mm_mul_ps(fscal,dx21);
2369             ty               = _mm_mul_ps(fscal,dy21);
2370             tz               = _mm_mul_ps(fscal,dz21);
2371
2372             /* Update vectorial force */
2373             fix2             = _mm_add_ps(fix2,tx);
2374             fiy2             = _mm_add_ps(fiy2,ty);
2375             fiz2             = _mm_add_ps(fiz2,tz);
2376
2377             fjx1             = _mm_add_ps(fjx1,tx);
2378             fjy1             = _mm_add_ps(fjy1,ty);
2379             fjz1             = _mm_add_ps(fjz1,tz);
2380
2381             /**************************
2382              * CALCULATE INTERACTIONS *
2383              **************************/
2384
2385             r22              = _mm_mul_ps(rsq22,rinv22);
2386             r22              = _mm_andnot_ps(dummy_mask,r22);
2387
2388             /* Calculate table index by multiplying r with table scale and truncate to integer */
2389             rt               = _mm_mul_ps(r22,vftabscale);
2390             vfitab           = _mm_cvttps_epi32(rt);
2391             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2392             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2393
2394             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2395             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2396             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2397             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2398             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2399             _MM_TRANSPOSE4_PS(Y,F,G,H);
2400             Heps             = _mm_mul_ps(vfeps,H);
2401             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2402             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2403             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
2404
2405             fscal            = felec;
2406
2407             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2408
2409             /* Calculate temporary vectorial force */
2410             tx               = _mm_mul_ps(fscal,dx22);
2411             ty               = _mm_mul_ps(fscal,dy22);
2412             tz               = _mm_mul_ps(fscal,dz22);
2413
2414             /* Update vectorial force */
2415             fix2             = _mm_add_ps(fix2,tx);
2416             fiy2             = _mm_add_ps(fiy2,ty);
2417             fiz2             = _mm_add_ps(fiz2,tz);
2418
2419             fjx2             = _mm_add_ps(fjx2,tx);
2420             fjy2             = _mm_add_ps(fjy2,ty);
2421             fjz2             = _mm_add_ps(fjz2,tz);
2422
2423             /**************************
2424              * CALCULATE INTERACTIONS *
2425              **************************/
2426
2427             r23              = _mm_mul_ps(rsq23,rinv23);
2428             r23              = _mm_andnot_ps(dummy_mask,r23);
2429
2430             /* Calculate table index by multiplying r with table scale and truncate to integer */
2431             rt               = _mm_mul_ps(r23,vftabscale);
2432             vfitab           = _mm_cvttps_epi32(rt);
2433             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2434             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2435
2436             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2437             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2438             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2439             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2440             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2441             _MM_TRANSPOSE4_PS(Y,F,G,H);
2442             Heps             = _mm_mul_ps(vfeps,H);
2443             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2444             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2445             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
2446
2447             fscal            = felec;
2448
2449             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2450
2451             /* Calculate temporary vectorial force */
2452             tx               = _mm_mul_ps(fscal,dx23);
2453             ty               = _mm_mul_ps(fscal,dy23);
2454             tz               = _mm_mul_ps(fscal,dz23);
2455
2456             /* Update vectorial force */
2457             fix2             = _mm_add_ps(fix2,tx);
2458             fiy2             = _mm_add_ps(fiy2,ty);
2459             fiz2             = _mm_add_ps(fiz2,tz);
2460
2461             fjx3             = _mm_add_ps(fjx3,tx);
2462             fjy3             = _mm_add_ps(fjy3,ty);
2463             fjz3             = _mm_add_ps(fjz3,tz);
2464
2465             /**************************
2466              * CALCULATE INTERACTIONS *
2467              **************************/
2468
2469             r31              = _mm_mul_ps(rsq31,rinv31);
2470             r31              = _mm_andnot_ps(dummy_mask,r31);
2471
2472             /* Calculate table index by multiplying r with table scale and truncate to integer */
2473             rt               = _mm_mul_ps(r31,vftabscale);
2474             vfitab           = _mm_cvttps_epi32(rt);
2475             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2476             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2477
2478             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2479             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2480             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2481             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2482             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2483             _MM_TRANSPOSE4_PS(Y,F,G,H);
2484             Heps             = _mm_mul_ps(vfeps,H);
2485             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2486             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2487             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
2488
2489             fscal            = felec;
2490
2491             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2492
2493             /* Calculate temporary vectorial force */
2494             tx               = _mm_mul_ps(fscal,dx31);
2495             ty               = _mm_mul_ps(fscal,dy31);
2496             tz               = _mm_mul_ps(fscal,dz31);
2497
2498             /* Update vectorial force */
2499             fix3             = _mm_add_ps(fix3,tx);
2500             fiy3             = _mm_add_ps(fiy3,ty);
2501             fiz3             = _mm_add_ps(fiz3,tz);
2502
2503             fjx1             = _mm_add_ps(fjx1,tx);
2504             fjy1             = _mm_add_ps(fjy1,ty);
2505             fjz1             = _mm_add_ps(fjz1,tz);
2506
2507             /**************************
2508              * CALCULATE INTERACTIONS *
2509              **************************/
2510
2511             r32              = _mm_mul_ps(rsq32,rinv32);
2512             r32              = _mm_andnot_ps(dummy_mask,r32);
2513
2514             /* Calculate table index by multiplying r with table scale and truncate to integer */
2515             rt               = _mm_mul_ps(r32,vftabscale);
2516             vfitab           = _mm_cvttps_epi32(rt);
2517             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2518             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2519
2520             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2521             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2522             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2523             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2524             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2525             _MM_TRANSPOSE4_PS(Y,F,G,H);
2526             Heps             = _mm_mul_ps(vfeps,H);
2527             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2528             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2529             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
2530
2531             fscal            = felec;
2532
2533             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2534
2535             /* Calculate temporary vectorial force */
2536             tx               = _mm_mul_ps(fscal,dx32);
2537             ty               = _mm_mul_ps(fscal,dy32);
2538             tz               = _mm_mul_ps(fscal,dz32);
2539
2540             /* Update vectorial force */
2541             fix3             = _mm_add_ps(fix3,tx);
2542             fiy3             = _mm_add_ps(fiy3,ty);
2543             fiz3             = _mm_add_ps(fiz3,tz);
2544
2545             fjx2             = _mm_add_ps(fjx2,tx);
2546             fjy2             = _mm_add_ps(fjy2,ty);
2547             fjz2             = _mm_add_ps(fjz2,tz);
2548
2549             /**************************
2550              * CALCULATE INTERACTIONS *
2551              **************************/
2552
2553             r33              = _mm_mul_ps(rsq33,rinv33);
2554             r33              = _mm_andnot_ps(dummy_mask,r33);
2555
2556             /* Calculate table index by multiplying r with table scale and truncate to integer */
2557             rt               = _mm_mul_ps(r33,vftabscale);
2558             vfitab           = _mm_cvttps_epi32(rt);
2559             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2560             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2561
2562             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2563             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2564             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2565             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2566             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2567             _MM_TRANSPOSE4_PS(Y,F,G,H);
2568             Heps             = _mm_mul_ps(vfeps,H);
2569             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2570             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2571             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2572
2573             fscal            = felec;
2574
2575             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2576
2577             /* Calculate temporary vectorial force */
2578             tx               = _mm_mul_ps(fscal,dx33);
2579             ty               = _mm_mul_ps(fscal,dy33);
2580             tz               = _mm_mul_ps(fscal,dz33);
2581
2582             /* Update vectorial force */
2583             fix3             = _mm_add_ps(fix3,tx);
2584             fiy3             = _mm_add_ps(fiy3,ty);
2585             fiz3             = _mm_add_ps(fiz3,tz);
2586
2587             fjx3             = _mm_add_ps(fjx3,tx);
2588             fjy3             = _mm_add_ps(fjy3,ty);
2589             fjz3             = _mm_add_ps(fjz3,tz);
2590
2591             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(f+j_coord_offsetA,f+j_coord_offsetB,
2592                                                    f+j_coord_offsetC,f+j_coord_offsetD,
2593                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2594                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2595
2596             /* Inner loop uses 412 flops */
2597         }
2598
2599         /* End of innermost loop */
2600
2601         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2602                                               f+i_coord_offset,fshift+i_shift_offset);
2603
2604         /* Increment number of inner iterations */
2605         inneriter                  += j_index_end - j_index_start;
2606
2607         /* Outer loop uses 36 flops */
2608     }
2609
2610     /* Increment number of outer iterations */
2611     outeriter        += nri;
2612
2613     /* Update outer/inner flops */
2614
2615     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*36 + inneriter*412);
2616 }