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              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  62     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  63     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  64     real             rcutoff_scalar;
  65     real             *shiftvec,*fshift,*x,*f;
  66     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  67     real             scratch[4*DIM];
  68     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  69     int              vdwioffset0;
  70     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  71     int              vdwioffset1;
  72     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  73     int              vdwioffset2;
  74     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  75     int              vdwioffset3;
  76     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  77     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  78     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  79     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  80     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  81     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  82     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  83     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  84     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  85     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  86     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  87     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  88     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  89     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  90     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  91     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  92     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  93     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  94     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  95     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 102     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 103     __m128i          vfitab;
 104     __m128i          ifour       = _mm_set1_epi32(4);
 105     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
 106     real             *vftab;
 107     __m128           dummy_mask,cutoff_mask;
 108     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 109     __m128           one     = _mm_set1_ps(1.0);
 110     __m128           two     = _mm_set1_ps(2.0);
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = _mm_set1_ps(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124     nvdwtype         = fr->ntype;
 125     vdwparam         = fr->nbfp;
 126     vdwtype          = mdatoms->typeA;
 127
 128     vftab            = kernel_data->table_elec_vdw->data;
 129     vftabscale       = _mm_set1_ps(kernel_data->table_elec_vdw->scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 134     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 135     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 136     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 137
 138     jq1              = _mm_set1_ps(charge[inr+1]);
 139     jq2              = _mm_set1_ps(charge[inr+2]);
 140     jq3              = _mm_set1_ps(charge[inr+3]);
 141     vdwjidx0A        = 2*vdwtype[inr+0];
 142     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 143     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 144     qq11             = _mm_mul_ps(iq1,jq1);
 145     qq12             = _mm_mul_ps(iq1,jq2);
 146     qq13             = _mm_mul_ps(iq1,jq3);
 147     qq21             = _mm_mul_ps(iq2,jq1);
 148     qq22             = _mm_mul_ps(iq2,jq2);
 149     qq23             = _mm_mul_ps(iq2,jq3);
 150     qq31             = _mm_mul_ps(iq3,jq1);
 151     qq32             = _mm_mul_ps(iq3,jq2);
 152     qq33             = _mm_mul_ps(iq3,jq3);
 153
 154     /* Avoid stupid compiler warnings */
 155     jnrA = jnrB = jnrC = jnrD = 0;
 156     j_coord_offsetA = 0;
 157     j_coord_offsetB = 0;
 158     j_coord_offsetC = 0;
 159     j_coord_offsetD = 0;
 160
 161     outeriter        = 0;
 162     inneriter        = 0;
 163
 164     for(iidx=0;iidx<4*DIM;iidx++)
 165     {
 166         scratch[iidx] = 0.0;
 167     }
 168
 169     /* Start outer loop over neighborlists */
 170     for(iidx=0; iidx<nri; iidx++)
 171     {
 172         /* Load shift vector for this list */
 173         i_shift_offset   = DIM*shiftidx[iidx];
 174
 175         /* Load limits for loop over neighbors */
 176         j_index_start    = jindex[iidx];
 177         j_index_end      = jindex[iidx+1];
 178
 179         /* Get outer coordinate index */
 180         inr              = iinr[iidx];
 181         i_coord_offset   = DIM*inr;
 182
 183         /* Load i particle coords and add shift vector */
 184         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 185                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 186
 187         fix0             = _mm_setzero_ps();
 188         fiy0             = _mm_setzero_ps();
 189         fiz0             = _mm_setzero_ps();
 190         fix1             = _mm_setzero_ps();
 191         fiy1             = _mm_setzero_ps();
 192         fiz1             = _mm_setzero_ps();
 193         fix2             = _mm_setzero_ps();
 194         fiy2             = _mm_setzero_ps();
 195         fiz2             = _mm_setzero_ps();
 196         fix3             = _mm_setzero_ps();
 197         fiy3             = _mm_setzero_ps();
 198         fiz3             = _mm_setzero_ps();
 199
 200         /* Reset potential sums */
 201         velecsum         = _mm_setzero_ps();
 202         vvdwsum          = _mm_setzero_ps();
 203
 204         /* Start inner kernel loop */
 205         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 206         {
 207
 208             /* Get j neighbor index, and coordinate index */
 209             jnrA             = jjnr[jidx];
 210             jnrB             = jjnr[jidx+1];
 211             jnrC             = jjnr[jidx+2];
 212             jnrD             = jjnr[jidx+3];
 213             j_coord_offsetA  = DIM*jnrA;
 214             j_coord_offsetB  = DIM*jnrB;
 215             j_coord_offsetC  = DIM*jnrC;
 216             j_coord_offsetD  = DIM*jnrD;
 217
 218             /* load j atom coordinates */
 219             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 220                                               x+j_coord_offsetC,x+j_coord_offsetD,
 221                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 222                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 223
 224             /* Calculate displacement vector */
 225             dx00             = _mm_sub_ps(ix0,jx0);
 226             dy00             = _mm_sub_ps(iy0,jy0);
 227             dz00             = _mm_sub_ps(iz0,jz0);
 228             dx11             = _mm_sub_ps(ix1,jx1);
 229             dy11             = _mm_sub_ps(iy1,jy1);
 230             dz11             = _mm_sub_ps(iz1,jz1);
 231             dx12             = _mm_sub_ps(ix1,jx2);
 232             dy12             = _mm_sub_ps(iy1,jy2);
 233             dz12             = _mm_sub_ps(iz1,jz2);
 234             dx13             = _mm_sub_ps(ix1,jx3);
 235             dy13             = _mm_sub_ps(iy1,jy3);
 236             dz13             = _mm_sub_ps(iz1,jz3);
 237             dx21             = _mm_sub_ps(ix2,jx1);
 238             dy21             = _mm_sub_ps(iy2,jy1);
 239             dz21             = _mm_sub_ps(iz2,jz1);
 240             dx22             = _mm_sub_ps(ix2,jx2);
 241             dy22             = _mm_sub_ps(iy2,jy2);
 242             dz22             = _mm_sub_ps(iz2,jz2);
 243             dx23             = _mm_sub_ps(ix2,jx3);
 244             dy23             = _mm_sub_ps(iy2,jy3);
 245             dz23             = _mm_sub_ps(iz2,jz3);
 246             dx31             = _mm_sub_ps(ix3,jx1);
 247             dy31             = _mm_sub_ps(iy3,jy1);
 248             dz31             = _mm_sub_ps(iz3,jz1);
 249             dx32             = _mm_sub_ps(ix3,jx2);
 250             dy32             = _mm_sub_ps(iy3,jy2);
 251             dz32             = _mm_sub_ps(iz3,jz2);
 252             dx33             = _mm_sub_ps(ix3,jx3);
 253             dy33             = _mm_sub_ps(iy3,jy3);
 254             dz33             = _mm_sub_ps(iz3,jz3);
 255
 256             /* Calculate squared distance and things based on it */
 257             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 258             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 259             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 260             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 261             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 262             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 263             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 264             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 265             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 266             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 267
 268             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 269             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 270             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 271             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 272             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 273             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 274             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 275             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 276             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 277             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 278
 279             fjx0             = _mm_setzero_ps();
 280             fjy0             = _mm_setzero_ps();
 281             fjz0             = _mm_setzero_ps();
 282             fjx1             = _mm_setzero_ps();
 283             fjy1             = _mm_setzero_ps();
 284             fjz1             = _mm_setzero_ps();
 285             fjx2             = _mm_setzero_ps();
 286             fjy2             = _mm_setzero_ps();
 287             fjz2             = _mm_setzero_ps();
 288             fjx3             = _mm_setzero_ps();
 289             fjy3             = _mm_setzero_ps();
 290             fjz3             = _mm_setzero_ps();
 291
 292             /**************************
 293              * CALCULATE INTERACTIONS *
 294              **************************/
 295
 296             r00              = _mm_mul_ps(rsq00,rinv00);
 297
 298             /* Calculate table index by multiplying r with table scale and truncate to integer */
 299             rt               = _mm_mul_ps(r00,vftabscale);
 300             vfitab           = _mm_cvttps_epi32(rt);
 301             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 302             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 303
 304             /* CUBIC SPLINE TABLE DISPERSION */
 305             vfitab           = _mm_add_epi32(vfitab,ifour);
 306             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 307             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 308             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 309             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 310             _MM_TRANSPOSE4_PS(Y,F,G,H);
 311             Heps             = _mm_mul_ps(vfeps,H);
 312             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 313             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 314             vvdw6            = _mm_mul_ps(c6_00,VV);
 315             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 316             fvdw6            = _mm_mul_ps(c6_00,FF);
 317
 318             /* CUBIC SPLINE TABLE REPULSION */
 319             vfitab           = _mm_add_epi32(vfitab,ifour);
 320             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 321             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 322             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 323             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 324             _MM_TRANSPOSE4_PS(Y,F,G,H);
 325             Heps             = _mm_mul_ps(vfeps,H);
 326             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 327             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 328             vvdw12           = _mm_mul_ps(c12_00,VV);
 329             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 330             fvdw12           = _mm_mul_ps(c12_00,FF);
 331             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 332             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 333
 334             /* Update potential sum for this i atom from the interaction with this j atom. */
 335             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 336
 337             fscal            = fvdw;
 338
 339             /* Calculate temporary vectorial force */
 340             tx               = _mm_mul_ps(fscal,dx00);
 341             ty               = _mm_mul_ps(fscal,dy00);
 342             tz               = _mm_mul_ps(fscal,dz00);
 343
 344             /* Update vectorial force */
 345             fix0             = _mm_add_ps(fix0,tx);
 346             fiy0             = _mm_add_ps(fiy0,ty);
 347             fiz0             = _mm_add_ps(fiz0,tz);
 348
 349             fjx0             = _mm_add_ps(fjx0,tx);
 350             fjy0             = _mm_add_ps(fjy0,ty);
 351             fjz0             = _mm_add_ps(fjz0,tz);
 352
 353             /**************************
 354              * CALCULATE INTERACTIONS *
 355              **************************/
 356
 357             r11              = _mm_mul_ps(rsq11,rinv11);
 358
 359             /* Calculate table index by multiplying r with table scale and truncate to integer */
 360             rt               = _mm_mul_ps(r11,vftabscale);
 361             vfitab           = _mm_cvttps_epi32(rt);
 362             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 363             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 364
 365             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 366             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 367             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 368             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 369             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 370             _MM_TRANSPOSE4_PS(Y,F,G,H);
 371             Heps             = _mm_mul_ps(vfeps,H);
 372             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 373             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 374             velec            = _mm_mul_ps(qq11,VV);
 375             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 376             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 377
 378             /* Update potential sum for this i atom from the interaction with this j atom. */
 379             velecsum         = _mm_add_ps(velecsum,velec);
 380
 381             fscal            = felec;
 382
 383             /* Calculate temporary vectorial force */
 384             tx               = _mm_mul_ps(fscal,dx11);
 385             ty               = _mm_mul_ps(fscal,dy11);
 386             tz               = _mm_mul_ps(fscal,dz11);
 387
 388             /* Update vectorial force */
 389             fix1             = _mm_add_ps(fix1,tx);
 390             fiy1             = _mm_add_ps(fiy1,ty);
 391             fiz1             = _mm_add_ps(fiz1,tz);
 392
 393             fjx1             = _mm_add_ps(fjx1,tx);
 394             fjy1             = _mm_add_ps(fjy1,ty);
 395             fjz1             = _mm_add_ps(fjz1,tz);
 396
 397             /**************************
 398              * CALCULATE INTERACTIONS *
 399              **************************/
 400
 401             r12              = _mm_mul_ps(rsq12,rinv12);
 402
 403             /* Calculate table index by multiplying r with table scale and truncate to integer */
 404             rt               = _mm_mul_ps(r12,vftabscale);
 405             vfitab           = _mm_cvttps_epi32(rt);
 406             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 407             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 408
 409             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 410             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 411             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 412             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 413             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 414             _MM_TRANSPOSE4_PS(Y,F,G,H);
 415             Heps             = _mm_mul_ps(vfeps,H);
 416             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 417             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 418             velec            = _mm_mul_ps(qq12,VV);
 419             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 420             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 421
 422             /* Update potential sum for this i atom from the interaction with this j atom. */
 423             velecsum         = _mm_add_ps(velecsum,velec);
 424
 425             fscal            = felec;
 426
 427             /* Calculate temporary vectorial force */
 428             tx               = _mm_mul_ps(fscal,dx12);
 429             ty               = _mm_mul_ps(fscal,dy12);
 430             tz               = _mm_mul_ps(fscal,dz12);
 431
 432             /* Update vectorial force */
 433             fix1             = _mm_add_ps(fix1,tx);
 434             fiy1             = _mm_add_ps(fiy1,ty);
 435             fiz1             = _mm_add_ps(fiz1,tz);
 436
 437             fjx2             = _mm_add_ps(fjx2,tx);
 438             fjy2             = _mm_add_ps(fjy2,ty);
 439             fjz2             = _mm_add_ps(fjz2,tz);
 440
 441             /**************************
 442              * CALCULATE INTERACTIONS *
 443              **************************/
 444
 445             r13              = _mm_mul_ps(rsq13,rinv13);
 446
 447             /* Calculate table index by multiplying r with table scale and truncate to integer */
 448             rt               = _mm_mul_ps(r13,vftabscale);
 449             vfitab           = _mm_cvttps_epi32(rt);
 450             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 451             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 452
 453             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 454             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 455             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 456             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 457             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 458             _MM_TRANSPOSE4_PS(Y,F,G,H);
 459             Heps             = _mm_mul_ps(vfeps,H);
 460             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 461             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 462             velec            = _mm_mul_ps(qq13,VV);
 463             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 464             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
 465
 466             /* Update potential sum for this i atom from the interaction with this j atom. */
 467             velecsum         = _mm_add_ps(velecsum,velec);
 468
 469             fscal            = felec;
 470
 471             /* Calculate temporary vectorial force */
 472             tx               = _mm_mul_ps(fscal,dx13);
 473             ty               = _mm_mul_ps(fscal,dy13);
 474             tz               = _mm_mul_ps(fscal,dz13);
 475
 476             /* Update vectorial force */
 477             fix1             = _mm_add_ps(fix1,tx);
 478             fiy1             = _mm_add_ps(fiy1,ty);
 479             fiz1             = _mm_add_ps(fiz1,tz);
 480
 481             fjx3             = _mm_add_ps(fjx3,tx);
 482             fjy3             = _mm_add_ps(fjy3,ty);
 483             fjz3             = _mm_add_ps(fjz3,tz);
 484
 485             /**************************
 486              * CALCULATE INTERACTIONS *
 487              **************************/
 488
 489             r21              = _mm_mul_ps(rsq21,rinv21);
 490
 491             /* Calculate table index by multiplying r with table scale and truncate to integer */
 492             rt               = _mm_mul_ps(r21,vftabscale);
 493             vfitab           = _mm_cvttps_epi32(rt);
 494             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 495             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 496
 497             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 498             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 499             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 500             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 501             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 502             _MM_TRANSPOSE4_PS(Y,F,G,H);
 503             Heps             = _mm_mul_ps(vfeps,H);
 504             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 505             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 506             velec            = _mm_mul_ps(qq21,VV);
 507             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 508             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
 509
 510             /* Update potential sum for this i atom from the interaction with this j atom. */
 511             velecsum         = _mm_add_ps(velecsum,velec);
 512
 513             fscal            = felec;
 514
 515             /* Calculate temporary vectorial force */
 516             tx               = _mm_mul_ps(fscal,dx21);
 517             ty               = _mm_mul_ps(fscal,dy21);
 518             tz               = _mm_mul_ps(fscal,dz21);
 519
 520             /* Update vectorial force */
 521             fix2             = _mm_add_ps(fix2,tx);
 522             fiy2             = _mm_add_ps(fiy2,ty);
 523             fiz2             = _mm_add_ps(fiz2,tz);
 524
 525             fjx1             = _mm_add_ps(fjx1,tx);
 526             fjy1             = _mm_add_ps(fjy1,ty);
 527             fjz1             = _mm_add_ps(fjz1,tz);
 528
 529             /**************************
 530              * CALCULATE INTERACTIONS *
 531              **************************/
 532
 533             r22              = _mm_mul_ps(rsq22,rinv22);
 534
 535             /* Calculate table index by multiplying r with table scale and truncate to integer */
 536             rt               = _mm_mul_ps(r22,vftabscale);
 537             vfitab           = _mm_cvttps_epi32(rt);
 538             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 539             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 540
 541             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 542             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 543             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 544             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 545             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 546             _MM_TRANSPOSE4_PS(Y,F,G,H);
 547             Heps             = _mm_mul_ps(vfeps,H);
 548             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 549             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 550             velec            = _mm_mul_ps(qq22,VV);
 551             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 552             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
 553
 554             /* Update potential sum for this i atom from the interaction with this j atom. */
 555             velecsum         = _mm_add_ps(velecsum,velec);
 556
 557             fscal            = felec;
 558
 559             /* Calculate temporary vectorial force */
 560             tx               = _mm_mul_ps(fscal,dx22);
 561             ty               = _mm_mul_ps(fscal,dy22);
 562             tz               = _mm_mul_ps(fscal,dz22);
 563
 564             /* Update vectorial force */
 565             fix2             = _mm_add_ps(fix2,tx);
 566             fiy2             = _mm_add_ps(fiy2,ty);
 567             fiz2             = _mm_add_ps(fiz2,tz);
 568
 569             fjx2             = _mm_add_ps(fjx2,tx);
 570             fjy2             = _mm_add_ps(fjy2,ty);
 571             fjz2             = _mm_add_ps(fjz2,tz);
 572
 573             /**************************
 574              * CALCULATE INTERACTIONS *
 575              **************************/
 576
 577             r23              = _mm_mul_ps(rsq23,rinv23);
 578
 579             /* Calculate table index by multiplying r with table scale and truncate to integer */
 580             rt               = _mm_mul_ps(r23,vftabscale);
 581             vfitab           = _mm_cvttps_epi32(rt);
 582             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 583             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 584
 585             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 586             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 587             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 588             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 589             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 590             _MM_TRANSPOSE4_PS(Y,F,G,H);
 591             Heps             = _mm_mul_ps(vfeps,H);
 592             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 593             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 594             velec            = _mm_mul_ps(qq23,VV);
 595             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 596             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
 597
 598             /* Update potential sum for this i atom from the interaction with this j atom. */
 599             velecsum         = _mm_add_ps(velecsum,velec);
 600
 601             fscal            = felec;
 602
 603             /* Calculate temporary vectorial force */
 604             tx               = _mm_mul_ps(fscal,dx23);
 605             ty               = _mm_mul_ps(fscal,dy23);
 606             tz               = _mm_mul_ps(fscal,dz23);
 607
 608             /* Update vectorial force */
 609             fix2             = _mm_add_ps(fix2,tx);
 610             fiy2             = _mm_add_ps(fiy2,ty);
 611             fiz2             = _mm_add_ps(fiz2,tz);
 612
 613             fjx3             = _mm_add_ps(fjx3,tx);
 614             fjy3             = _mm_add_ps(fjy3,ty);
 615             fjz3             = _mm_add_ps(fjz3,tz);
 616
 617             /**************************
 618              * CALCULATE INTERACTIONS *
 619              **************************/
 620
 621             r31              = _mm_mul_ps(rsq31,rinv31);
 622
 623             /* Calculate table index by multiplying r with table scale and truncate to integer */
 624             rt               = _mm_mul_ps(r31,vftabscale);
 625             vfitab           = _mm_cvttps_epi32(rt);
 626             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 627             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 628
 629             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 630             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 631             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 632             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 633             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 634             _MM_TRANSPOSE4_PS(Y,F,G,H);
 635             Heps             = _mm_mul_ps(vfeps,H);
 636             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 637             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 638             velec            = _mm_mul_ps(qq31,VV);
 639             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 640             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
 641
 642             /* Update potential sum for this i atom from the interaction with this j atom. */
 643             velecsum         = _mm_add_ps(velecsum,velec);
 644
 645             fscal            = felec;
 646
 647             /* Calculate temporary vectorial force */
 648             tx               = _mm_mul_ps(fscal,dx31);
 649             ty               = _mm_mul_ps(fscal,dy31);
 650             tz               = _mm_mul_ps(fscal,dz31);
 651
 652             /* Update vectorial force */
 653             fix3             = _mm_add_ps(fix3,tx);
 654             fiy3             = _mm_add_ps(fiy3,ty);
 655             fiz3             = _mm_add_ps(fiz3,tz);
 656
 657             fjx1             = _mm_add_ps(fjx1,tx);
 658             fjy1             = _mm_add_ps(fjy1,ty);
 659             fjz1             = _mm_add_ps(fjz1,tz);
 660
 661             /**************************
 662              * CALCULATE INTERACTIONS *
 663              **************************/
 664
 665             r32              = _mm_mul_ps(rsq32,rinv32);
 666
 667             /* Calculate table index by multiplying r with table scale and truncate to integer */
 668             rt               = _mm_mul_ps(r32,vftabscale);
 669             vfitab           = _mm_cvttps_epi32(rt);
 670             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 671             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 672
 673             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 674             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 675             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 676             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 677             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 678             _MM_TRANSPOSE4_PS(Y,F,G,H);
 679             Heps             = _mm_mul_ps(vfeps,H);
 680             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 681             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 682             velec            = _mm_mul_ps(qq32,VV);
 683             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 684             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
 685
 686             /* Update potential sum for this i atom from the interaction with this j atom. */
 687             velecsum         = _mm_add_ps(velecsum,velec);
 688
 689             fscal            = felec;
 690
 691             /* Calculate temporary vectorial force */
 692             tx               = _mm_mul_ps(fscal,dx32);
 693             ty               = _mm_mul_ps(fscal,dy32);
 694             tz               = _mm_mul_ps(fscal,dz32);
 695
 696             /* Update vectorial force */
 697             fix3             = _mm_add_ps(fix3,tx);
 698             fiy3             = _mm_add_ps(fiy3,ty);
 699             fiz3             = _mm_add_ps(fiz3,tz);
 700
 701             fjx2             = _mm_add_ps(fjx2,tx);
 702             fjy2             = _mm_add_ps(fjy2,ty);
 703             fjz2             = _mm_add_ps(fjz2,tz);
 704
 705             /**************************
 706              * CALCULATE INTERACTIONS *
 707              **************************/
 708
 709             r33              = _mm_mul_ps(rsq33,rinv33);
 710
 711             /* Calculate table index by multiplying r with table scale and truncate to integer */
 712             rt               = _mm_mul_ps(r33,vftabscale);
 713             vfitab           = _mm_cvttps_epi32(rt);
 714             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 715             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 716
 717             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 718             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 719             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 720             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 721             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 722             _MM_TRANSPOSE4_PS(Y,F,G,H);
 723             Heps             = _mm_mul_ps(vfeps,H);
 724             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 725             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 726             velec            = _mm_mul_ps(qq33,VV);
 727             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 728             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
 729
 730             /* Update potential sum for this i atom from the interaction with this j atom. */
 731             velecsum         = _mm_add_ps(velecsum,velec);
 732
 733             fscal            = felec;
 734
 735             /* Calculate temporary vectorial force */
 736             tx               = _mm_mul_ps(fscal,dx33);
 737             ty               = _mm_mul_ps(fscal,dy33);
 738             tz               = _mm_mul_ps(fscal,dz33);
 739
 740             /* Update vectorial force */
 741             fix3             = _mm_add_ps(fix3,tx);
 742             fiy3             = _mm_add_ps(fiy3,ty);
 743             fiz3             = _mm_add_ps(fiz3,tz);
 744
 745             fjx3             = _mm_add_ps(fjx3,tx);
 746             fjy3             = _mm_add_ps(fjy3,ty);
 747             fjz3             = _mm_add_ps(fjz3,tz);
 748
 749             fjptrA             = f+j_coord_offsetA;
 750             fjptrB             = f+j_coord_offsetB;
 751             fjptrC             = f+j_coord_offsetC;
 752             fjptrD             = f+j_coord_offsetD;
 753
 754             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 755                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 756                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 757
 758             /* Inner loop uses 446 flops */
 759         }
 760
 761         if(jidx<j_index_end)
 762         {
 763
 764             /* Get j neighbor index, and coordinate index */
 765             jnrlistA         = jjnr[jidx];
 766             jnrlistB         = jjnr[jidx+1];
 767             jnrlistC         = jjnr[jidx+2];
 768             jnrlistD         = jjnr[jidx+3];
 769             /* Sign of each element will be negative for non-real atoms.
 770              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 771              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 772              */
 773             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 774             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 775             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 776             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 777             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 778             j_coord_offsetA  = DIM*jnrA;
 779             j_coord_offsetB  = DIM*jnrB;
 780             j_coord_offsetC  = DIM*jnrC;
 781             j_coord_offsetD  = DIM*jnrD;
 782
 783             /* load j atom coordinates */
 784             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 785                                               x+j_coord_offsetC,x+j_coord_offsetD,
 786                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 787                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 788
 789             /* Calculate displacement vector */
 790             dx00             = _mm_sub_ps(ix0,jx0);
 791             dy00             = _mm_sub_ps(iy0,jy0);
 792             dz00             = _mm_sub_ps(iz0,jz0);
 793             dx11             = _mm_sub_ps(ix1,jx1);
 794             dy11             = _mm_sub_ps(iy1,jy1);
 795             dz11             = _mm_sub_ps(iz1,jz1);
 796             dx12             = _mm_sub_ps(ix1,jx2);
 797             dy12             = _mm_sub_ps(iy1,jy2);
 798             dz12             = _mm_sub_ps(iz1,jz2);
 799             dx13             = _mm_sub_ps(ix1,jx3);
 800             dy13             = _mm_sub_ps(iy1,jy3);
 801             dz13             = _mm_sub_ps(iz1,jz3);
 802             dx21             = _mm_sub_ps(ix2,jx1);
 803             dy21             = _mm_sub_ps(iy2,jy1);
 804             dz21             = _mm_sub_ps(iz2,jz1);
 805             dx22             = _mm_sub_ps(ix2,jx2);
 806             dy22             = _mm_sub_ps(iy2,jy2);
 807             dz22             = _mm_sub_ps(iz2,jz2);
 808             dx23             = _mm_sub_ps(ix2,jx3);
 809             dy23             = _mm_sub_ps(iy2,jy3);
 810             dz23             = _mm_sub_ps(iz2,jz3);
 811             dx31             = _mm_sub_ps(ix3,jx1);
 812             dy31             = _mm_sub_ps(iy3,jy1);
 813             dz31             = _mm_sub_ps(iz3,jz1);
 814             dx32             = _mm_sub_ps(ix3,jx2);
 815             dy32             = _mm_sub_ps(iy3,jy2);
 816             dz32             = _mm_sub_ps(iz3,jz2);
 817             dx33             = _mm_sub_ps(ix3,jx3);
 818             dy33             = _mm_sub_ps(iy3,jy3);
 819             dz33             = _mm_sub_ps(iz3,jz3);
 820
 821             /* Calculate squared distance and things based on it */
 822             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 823             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 824             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 825             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 826             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 827             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 828             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 829             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 830             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 831             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 832
 833             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 834             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 835             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 836             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 837             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 838             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 839             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 840             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 841             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 842             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 843
 844             fjx0             = _mm_setzero_ps();
 845             fjy0             = _mm_setzero_ps();
 846             fjz0             = _mm_setzero_ps();
 847             fjx1             = _mm_setzero_ps();
 848             fjy1             = _mm_setzero_ps();
 849             fjz1             = _mm_setzero_ps();
 850             fjx2             = _mm_setzero_ps();
 851             fjy2             = _mm_setzero_ps();
 852             fjz2             = _mm_setzero_ps();
 853             fjx3             = _mm_setzero_ps();
 854             fjy3             = _mm_setzero_ps();
 855             fjz3             = _mm_setzero_ps();
 856
 857             /**************************
 858              * CALCULATE INTERACTIONS *
 859              **************************/
 860
 861             r00              = _mm_mul_ps(rsq00,rinv00);
 862             r00              = _mm_andnot_ps(dummy_mask,r00);
 863
 864             /* Calculate table index by multiplying r with table scale and truncate to integer */
 865             rt               = _mm_mul_ps(r00,vftabscale);
 866             vfitab           = _mm_cvttps_epi32(rt);
 867             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 868             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 869
 870             /* CUBIC SPLINE TABLE DISPERSION */
 871             vfitab           = _mm_add_epi32(vfitab,ifour);
 872             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 873             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 874             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 875             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 876             _MM_TRANSPOSE4_PS(Y,F,G,H);
 877             Heps             = _mm_mul_ps(vfeps,H);
 878             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 879             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 880             vvdw6            = _mm_mul_ps(c6_00,VV);
 881             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 882             fvdw6            = _mm_mul_ps(c6_00,FF);
 883
 884             /* CUBIC SPLINE TABLE REPULSION */
 885             vfitab           = _mm_add_epi32(vfitab,ifour);
 886             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 887             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 888             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 889             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 890             _MM_TRANSPOSE4_PS(Y,F,G,H);
 891             Heps             = _mm_mul_ps(vfeps,H);
 892             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 893             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 894             vvdw12           = _mm_mul_ps(c12_00,VV);
 895             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 896             fvdw12           = _mm_mul_ps(c12_00,FF);
 897             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 898             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 899
 900             /* Update potential sum for this i atom from the interaction with this j atom. */
 901             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 902             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 903
 904             fscal            = fvdw;
 905
 906             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 907
 908             /* Calculate temporary vectorial force */
 909             tx               = _mm_mul_ps(fscal,dx00);
 910             ty               = _mm_mul_ps(fscal,dy00);
 911             tz               = _mm_mul_ps(fscal,dz00);
 912
 913             /* Update vectorial force */
 914             fix0             = _mm_add_ps(fix0,tx);
 915             fiy0             = _mm_add_ps(fiy0,ty);
 916             fiz0             = _mm_add_ps(fiz0,tz);
 917
 918             fjx0             = _mm_add_ps(fjx0,tx);
 919             fjy0             = _mm_add_ps(fjy0,ty);
 920             fjz0             = _mm_add_ps(fjz0,tz);
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             r11              = _mm_mul_ps(rsq11,rinv11);
 927             r11              = _mm_andnot_ps(dummy_mask,r11);
 928
 929             /* Calculate table index by multiplying r with table scale and truncate to integer */
 930             rt               = _mm_mul_ps(r11,vftabscale);
 931             vfitab           = _mm_cvttps_epi32(rt);
 932             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 933             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 934
 935             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 936             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 937             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 938             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 939             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 940             _MM_TRANSPOSE4_PS(Y,F,G,H);
 941             Heps             = _mm_mul_ps(vfeps,H);
 942             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 943             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 944             velec            = _mm_mul_ps(qq11,VV);
 945             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 946             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 947
 948             /* Update potential sum for this i atom from the interaction with this j atom. */
 949             velec            = _mm_andnot_ps(dummy_mask,velec);
 950             velecsum         = _mm_add_ps(velecsum,velec);
 951
 952             fscal            = felec;
 953
 954             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 955
 956             /* Calculate temporary vectorial force */
 957             tx               = _mm_mul_ps(fscal,dx11);
 958             ty               = _mm_mul_ps(fscal,dy11);
 959             tz               = _mm_mul_ps(fscal,dz11);
 960
 961             /* Update vectorial force */
 962             fix1             = _mm_add_ps(fix1,tx);
 963             fiy1             = _mm_add_ps(fiy1,ty);
 964             fiz1             = _mm_add_ps(fiz1,tz);
 965
 966             fjx1             = _mm_add_ps(fjx1,tx);
 967             fjy1             = _mm_add_ps(fjy1,ty);
 968             fjz1             = _mm_add_ps(fjz1,tz);
 969
 970             /**************************
 971              * CALCULATE INTERACTIONS *
 972              **************************/
 973
 974             r12              = _mm_mul_ps(rsq12,rinv12);
 975             r12              = _mm_andnot_ps(dummy_mask,r12);
 976
 977             /* Calculate table index by multiplying r with table scale and truncate to integer */
 978             rt               = _mm_mul_ps(r12,vftabscale);
 979             vfitab           = _mm_cvttps_epi32(rt);
 980             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 981             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
 982
 983             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 984             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 985             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 986             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 987             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 988             _MM_TRANSPOSE4_PS(Y,F,G,H);
 989             Heps             = _mm_mul_ps(vfeps,H);
 990             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 991             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 992             velec            = _mm_mul_ps(qq12,VV);
 993             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 994             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 995
 996             /* Update potential sum for this i atom from the interaction with this j atom. */
 997             velec            = _mm_andnot_ps(dummy_mask,velec);
 998             velecsum         = _mm_add_ps(velecsum,velec);
 999
1000             fscal            = felec;
1001
1002             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1003
1004             /* Calculate temporary vectorial force */
1005             tx               = _mm_mul_ps(fscal,dx12);
1006             ty               = _mm_mul_ps(fscal,dy12);
1007             tz               = _mm_mul_ps(fscal,dz12);
1008
1009             /* Update vectorial force */
1010             fix1             = _mm_add_ps(fix1,tx);
1011             fiy1             = _mm_add_ps(fiy1,ty);
1012             fiz1             = _mm_add_ps(fiz1,tz);
1013
1014             fjx2             = _mm_add_ps(fjx2,tx);
1015             fjy2             = _mm_add_ps(fjy2,ty);
1016             fjz2             = _mm_add_ps(fjz2,tz);
1017
1018             /**************************
1019              * CALCULATE INTERACTIONS *
1020              **************************/
1021
1022             r13              = _mm_mul_ps(rsq13,rinv13);
1023             r13              = _mm_andnot_ps(dummy_mask,r13);
1024
1025             /* Calculate table index by multiplying r with table scale and truncate to integer */
1026             rt               = _mm_mul_ps(r13,vftabscale);
1027             vfitab           = _mm_cvttps_epi32(rt);
1028             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1029             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1030
1031             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1032             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1033             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1034             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1035             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1036             _MM_TRANSPOSE4_PS(Y,F,G,H);
1037             Heps             = _mm_mul_ps(vfeps,H);
1038             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1039             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1040             velec            = _mm_mul_ps(qq13,VV);
1041             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1042             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1043
1044             /* Update potential sum for this i atom from the interaction with this j atom. */
1045             velec            = _mm_andnot_ps(dummy_mask,velec);
1046             velecsum         = _mm_add_ps(velecsum,velec);
1047
1048             fscal            = felec;
1049
1050             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1051
1052             /* Calculate temporary vectorial force */
1053             tx               = _mm_mul_ps(fscal,dx13);
1054             ty               = _mm_mul_ps(fscal,dy13);
1055             tz               = _mm_mul_ps(fscal,dz13);
1056
1057             /* Update vectorial force */
1058             fix1             = _mm_add_ps(fix1,tx);
1059             fiy1             = _mm_add_ps(fiy1,ty);
1060             fiz1             = _mm_add_ps(fiz1,tz);
1061
1062             fjx3             = _mm_add_ps(fjx3,tx);
1063             fjy3             = _mm_add_ps(fjy3,ty);
1064             fjz3             = _mm_add_ps(fjz3,tz);
1065
1066             /**************************
1067              * CALCULATE INTERACTIONS *
1068              **************************/
1069
1070             r21              = _mm_mul_ps(rsq21,rinv21);
1071             r21              = _mm_andnot_ps(dummy_mask,r21);
1072
1073             /* Calculate table index by multiplying r with table scale and truncate to integer */
1074             rt               = _mm_mul_ps(r21,vftabscale);
1075             vfitab           = _mm_cvttps_epi32(rt);
1076             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1077             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1078
1079             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1080             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1081             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1082             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1083             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1084             _MM_TRANSPOSE4_PS(Y,F,G,H);
1085             Heps             = _mm_mul_ps(vfeps,H);
1086             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1087             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1088             velec            = _mm_mul_ps(qq21,VV);
1089             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1090             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1091
1092             /* Update potential sum for this i atom from the interaction with this j atom. */
1093             velec            = _mm_andnot_ps(dummy_mask,velec);
1094             velecsum         = _mm_add_ps(velecsum,velec);
1095
1096             fscal            = felec;
1097
1098             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1099
1100             /* Calculate temporary vectorial force */
1101             tx               = _mm_mul_ps(fscal,dx21);
1102             ty               = _mm_mul_ps(fscal,dy21);
1103             tz               = _mm_mul_ps(fscal,dz21);
1104
1105             /* Update vectorial force */
1106             fix2             = _mm_add_ps(fix2,tx);
1107             fiy2             = _mm_add_ps(fiy2,ty);
1108             fiz2             = _mm_add_ps(fiz2,tz);
1109
1110             fjx1             = _mm_add_ps(fjx1,tx);
1111             fjy1             = _mm_add_ps(fjy1,ty);
1112             fjz1             = _mm_add_ps(fjz1,tz);
1113
1114             /**************************
1115              * CALCULATE INTERACTIONS *
1116              **************************/
1117
1118             r22              = _mm_mul_ps(rsq22,rinv22);
1119             r22              = _mm_andnot_ps(dummy_mask,r22);
1120
1121             /* Calculate table index by multiplying r with table scale and truncate to integer */
1122             rt               = _mm_mul_ps(r22,vftabscale);
1123             vfitab           = _mm_cvttps_epi32(rt);
1124             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1125             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1126
1127             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1128             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1129             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1130             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1131             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1132             _MM_TRANSPOSE4_PS(Y,F,G,H);
1133             Heps             = _mm_mul_ps(vfeps,H);
1134             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1135             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1136             velec            = _mm_mul_ps(qq22,VV);
1137             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1138             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1139
1140             /* Update potential sum for this i atom from the interaction with this j atom. */
1141             velec            = _mm_andnot_ps(dummy_mask,velec);
1142             velecsum         = _mm_add_ps(velecsum,velec);
1143
1144             fscal            = felec;
1145
1146             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1147
1148             /* Calculate temporary vectorial force */
1149             tx               = _mm_mul_ps(fscal,dx22);
1150             ty               = _mm_mul_ps(fscal,dy22);
1151             tz               = _mm_mul_ps(fscal,dz22);
1152
1153             /* Update vectorial force */
1154             fix2             = _mm_add_ps(fix2,tx);
1155             fiy2             = _mm_add_ps(fiy2,ty);
1156             fiz2             = _mm_add_ps(fiz2,tz);
1157
1158             fjx2             = _mm_add_ps(fjx2,tx);
1159             fjy2             = _mm_add_ps(fjy2,ty);
1160             fjz2             = _mm_add_ps(fjz2,tz);
1161
1162             /**************************
1163              * CALCULATE INTERACTIONS *
1164              **************************/
1165
1166             r23              = _mm_mul_ps(rsq23,rinv23);
1167             r23              = _mm_andnot_ps(dummy_mask,r23);
1168
1169             /* Calculate table index by multiplying r with table scale and truncate to integer */
1170             rt               = _mm_mul_ps(r23,vftabscale);
1171             vfitab           = _mm_cvttps_epi32(rt);
1172             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1173             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1174
1175             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1176             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1177             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1178             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1179             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1180             _MM_TRANSPOSE4_PS(Y,F,G,H);
1181             Heps             = _mm_mul_ps(vfeps,H);
1182             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1183             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1184             velec            = _mm_mul_ps(qq23,VV);
1185             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1186             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1187
1188             /* Update potential sum for this i atom from the interaction with this j atom. */
1189             velec            = _mm_andnot_ps(dummy_mask,velec);
1190             velecsum         = _mm_add_ps(velecsum,velec);
1191
1192             fscal            = felec;
1193
1194             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1195
1196             /* Calculate temporary vectorial force */
1197             tx               = _mm_mul_ps(fscal,dx23);
1198             ty               = _mm_mul_ps(fscal,dy23);
1199             tz               = _mm_mul_ps(fscal,dz23);
1200
1201             /* Update vectorial force */
1202             fix2             = _mm_add_ps(fix2,tx);
1203             fiy2             = _mm_add_ps(fiy2,ty);
1204             fiz2             = _mm_add_ps(fiz2,tz);
1205
1206             fjx3             = _mm_add_ps(fjx3,tx);
1207             fjy3             = _mm_add_ps(fjy3,ty);
1208             fjz3             = _mm_add_ps(fjz3,tz);
1209
1210             /**************************
1211              * CALCULATE INTERACTIONS *
1212              **************************/
1213
1214             r31              = _mm_mul_ps(rsq31,rinv31);
1215             r31              = _mm_andnot_ps(dummy_mask,r31);
1216
1217             /* Calculate table index by multiplying r with table scale and truncate to integer */
1218             rt               = _mm_mul_ps(r31,vftabscale);
1219             vfitab           = _mm_cvttps_epi32(rt);
1220             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1221             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1222
1223             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1224             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1225             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1226             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1227             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1228             _MM_TRANSPOSE4_PS(Y,F,G,H);
1229             Heps             = _mm_mul_ps(vfeps,H);
1230             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1231             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1232             velec            = _mm_mul_ps(qq31,VV);
1233             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1234             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1235
1236             /* Update potential sum for this i atom from the interaction with this j atom. */
1237             velec            = _mm_andnot_ps(dummy_mask,velec);
1238             velecsum         = _mm_add_ps(velecsum,velec);
1239
1240             fscal            = felec;
1241
1242             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1243
1244             /* Calculate temporary vectorial force */
1245             tx               = _mm_mul_ps(fscal,dx31);
1246             ty               = _mm_mul_ps(fscal,dy31);
1247             tz               = _mm_mul_ps(fscal,dz31);
1248
1249             /* Update vectorial force */
1250             fix3             = _mm_add_ps(fix3,tx);
1251             fiy3             = _mm_add_ps(fiy3,ty);
1252             fiz3             = _mm_add_ps(fiz3,tz);
1253
1254             fjx1             = _mm_add_ps(fjx1,tx);
1255             fjy1             = _mm_add_ps(fjy1,ty);
1256             fjz1             = _mm_add_ps(fjz1,tz);
1257
1258             /**************************
1259              * CALCULATE INTERACTIONS *
1260              **************************/
1261
1262             r32              = _mm_mul_ps(rsq32,rinv32);
1263             r32              = _mm_andnot_ps(dummy_mask,r32);
1264
1265             /* Calculate table index by multiplying r with table scale and truncate to integer */
1266             rt               = _mm_mul_ps(r32,vftabscale);
1267             vfitab           = _mm_cvttps_epi32(rt);
1268             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1269             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1270
1271             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1272             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1273             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1274             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1275             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1276             _MM_TRANSPOSE4_PS(Y,F,G,H);
1277             Heps             = _mm_mul_ps(vfeps,H);
1278             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1279             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1280             velec            = _mm_mul_ps(qq32,VV);
1281             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1282             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1283
1284             /* Update potential sum for this i atom from the interaction with this j atom. */
1285             velec            = _mm_andnot_ps(dummy_mask,velec);
1286             velecsum         = _mm_add_ps(velecsum,velec);
1287
1288             fscal            = felec;
1289
1290             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1291
1292             /* Calculate temporary vectorial force */
1293             tx               = _mm_mul_ps(fscal,dx32);
1294             ty               = _mm_mul_ps(fscal,dy32);
1295             tz               = _mm_mul_ps(fscal,dz32);
1296
1297             /* Update vectorial force */
1298             fix3             = _mm_add_ps(fix3,tx);
1299             fiy3             = _mm_add_ps(fiy3,ty);
1300             fiz3             = _mm_add_ps(fiz3,tz);
1301
1302             fjx2             = _mm_add_ps(fjx2,tx);
1303             fjy2             = _mm_add_ps(fjy2,ty);
1304             fjz2             = _mm_add_ps(fjz2,tz);
1305
1306             /**************************
1307              * CALCULATE INTERACTIONS *
1308              **************************/
1309
1310             r33              = _mm_mul_ps(rsq33,rinv33);
1311             r33              = _mm_andnot_ps(dummy_mask,r33);
1312
1313             /* Calculate table index by multiplying r with table scale and truncate to integer */
1314             rt               = _mm_mul_ps(r33,vftabscale);
1315             vfitab           = _mm_cvttps_epi32(rt);
1316             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1317             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1318
1319             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1320             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1321             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1322             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1323             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1324             _MM_TRANSPOSE4_PS(Y,F,G,H);
1325             Heps             = _mm_mul_ps(vfeps,H);
1326             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1327             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1328             velec            = _mm_mul_ps(qq33,VV);
1329             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1330             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
1331
1332             /* Update potential sum for this i atom from the interaction with this j atom. */
1333             velec            = _mm_andnot_ps(dummy_mask,velec);
1334             velecsum         = _mm_add_ps(velecsum,velec);
1335
1336             fscal            = felec;
1337
1338             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1339
1340             /* Calculate temporary vectorial force */
1341             tx               = _mm_mul_ps(fscal,dx33);
1342             ty               = _mm_mul_ps(fscal,dy33);
1343             tz               = _mm_mul_ps(fscal,dz33);
1344
1345             /* Update vectorial force */
1346             fix3             = _mm_add_ps(fix3,tx);
1347             fiy3             = _mm_add_ps(fiy3,ty);
1348             fiz3             = _mm_add_ps(fiz3,tz);
1349
1350             fjx3             = _mm_add_ps(fjx3,tx);
1351             fjy3             = _mm_add_ps(fjy3,ty);
1352             fjz3             = _mm_add_ps(fjz3,tz);
1353
1354             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1355             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1356             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1357             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1358
1359             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
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 26 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*26 + 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              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1415     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1416     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1417     real             rcutoff_scalar;
1418     real             *shiftvec,*fshift,*x,*f;
1419     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1420     real             scratch[4*DIM];
1421     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1422     int              vdwioffset0;
1423     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1424     int              vdwioffset1;
1425     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1426     int              vdwioffset2;
1427     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1428     int              vdwioffset3;
1429     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1430     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1431     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1432     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1433     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1434     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1435     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1436     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1437     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1438     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1439     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1440     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1441     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1442     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1443     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1444     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1445     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1446     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1447     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1448     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1449     real             *charge;
1450     int              nvdwtype;
1451     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1452     int              *vdwtype;
1453     real             *vdwparam;
1454     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1455     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1456     __m128i          vfitab;
1457     __m128i          ifour       = _mm_set1_epi32(4);
1458     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1459     real             *vftab;
1460     __m128           dummy_mask,cutoff_mask;
1461     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1462     __m128           one     = _mm_set1_ps(1.0);
1463     __m128           two     = _mm_set1_ps(2.0);
1464     x                = xx[0];
1465     f                = ff[0];
1466
1467     nri              = nlist->nri;
1468     iinr             = nlist->iinr;
1469     jindex           = nlist->jindex;
1470     jjnr             = nlist->jjnr;
1471     shiftidx         = nlist->shift;
1472     gid              = nlist->gid;
1473     shiftvec         = fr->shift_vec[0];
1474     fshift           = fr->fshift[0];
1475     facel            = _mm_set1_ps(fr->epsfac);
1476     charge           = mdatoms->chargeA;
1477     nvdwtype         = fr->ntype;
1478     vdwparam         = fr->nbfp;
1479     vdwtype          = mdatoms->typeA;
1480
1481     vftab            = kernel_data->table_elec_vdw->data;
1482     vftabscale       = _mm_set1_ps(kernel_data->table_elec_vdw->scale);
1483
1484     /* Setup water-specific parameters */
1485     inr              = nlist->iinr[0];
1486     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1487     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1488     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1489     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1490
1491     jq1              = _mm_set1_ps(charge[inr+1]);
1492     jq2              = _mm_set1_ps(charge[inr+2]);
1493     jq3              = _mm_set1_ps(charge[inr+3]);
1494     vdwjidx0A        = 2*vdwtype[inr+0];
1495     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1496     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1497     qq11             = _mm_mul_ps(iq1,jq1);
1498     qq12             = _mm_mul_ps(iq1,jq2);
1499     qq13             = _mm_mul_ps(iq1,jq3);
1500     qq21             = _mm_mul_ps(iq2,jq1);
1501     qq22             = _mm_mul_ps(iq2,jq2);
1502     qq23             = _mm_mul_ps(iq2,jq3);
1503     qq31             = _mm_mul_ps(iq3,jq1);
1504     qq32             = _mm_mul_ps(iq3,jq2);
1505     qq33             = _mm_mul_ps(iq3,jq3);
1506
1507     /* Avoid stupid compiler warnings */
1508     jnrA = jnrB = jnrC = jnrD = 0;
1509     j_coord_offsetA = 0;
1510     j_coord_offsetB = 0;
1511     j_coord_offsetC = 0;
1512     j_coord_offsetD = 0;
1513
1514     outeriter        = 0;
1515     inneriter        = 0;
1516
1517     for(iidx=0;iidx<4*DIM;iidx++)
1518     {
1519         scratch[iidx] = 0.0;
1520     }
1521
1522     /* Start outer loop over neighborlists */
1523     for(iidx=0; iidx<nri; iidx++)
1524     {
1525         /* Load shift vector for this list */
1526         i_shift_offset   = DIM*shiftidx[iidx];
1527
1528         /* Load limits for loop over neighbors */
1529         j_index_start    = jindex[iidx];
1530         j_index_end      = jindex[iidx+1];
1531
1532         /* Get outer coordinate index */
1533         inr              = iinr[iidx];
1534         i_coord_offset   = DIM*inr;
1535
1536         /* Load i particle coords and add shift vector */
1537         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1538                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1539
1540         fix0             = _mm_setzero_ps();
1541         fiy0             = _mm_setzero_ps();
1542         fiz0             = _mm_setzero_ps();
1543         fix1             = _mm_setzero_ps();
1544         fiy1             = _mm_setzero_ps();
1545         fiz1             = _mm_setzero_ps();
1546         fix2             = _mm_setzero_ps();
1547         fiy2             = _mm_setzero_ps();
1548         fiz2             = _mm_setzero_ps();
1549         fix3             = _mm_setzero_ps();
1550         fiy3             = _mm_setzero_ps();
1551         fiz3             = _mm_setzero_ps();
1552
1553         /* Start inner kernel loop */
1554         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1555         {
1556
1557             /* Get j neighbor index, and coordinate index */
1558             jnrA             = jjnr[jidx];
1559             jnrB             = jjnr[jidx+1];
1560             jnrC             = jjnr[jidx+2];
1561             jnrD             = jjnr[jidx+3];
1562             j_coord_offsetA  = DIM*jnrA;
1563             j_coord_offsetB  = DIM*jnrB;
1564             j_coord_offsetC  = DIM*jnrC;
1565             j_coord_offsetD  = DIM*jnrD;
1566
1567             /* load j atom coordinates */
1568             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1569                                               x+j_coord_offsetC,x+j_coord_offsetD,
1570                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1571                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1572
1573             /* Calculate displacement vector */
1574             dx00             = _mm_sub_ps(ix0,jx0);
1575             dy00             = _mm_sub_ps(iy0,jy0);
1576             dz00             = _mm_sub_ps(iz0,jz0);
1577             dx11             = _mm_sub_ps(ix1,jx1);
1578             dy11             = _mm_sub_ps(iy1,jy1);
1579             dz11             = _mm_sub_ps(iz1,jz1);
1580             dx12             = _mm_sub_ps(ix1,jx2);
1581             dy12             = _mm_sub_ps(iy1,jy2);
1582             dz12             = _mm_sub_ps(iz1,jz2);
1583             dx13             = _mm_sub_ps(ix1,jx3);
1584             dy13             = _mm_sub_ps(iy1,jy3);
1585             dz13             = _mm_sub_ps(iz1,jz3);
1586             dx21             = _mm_sub_ps(ix2,jx1);
1587             dy21             = _mm_sub_ps(iy2,jy1);
1588             dz21             = _mm_sub_ps(iz2,jz1);
1589             dx22             = _mm_sub_ps(ix2,jx2);
1590             dy22             = _mm_sub_ps(iy2,jy2);
1591             dz22             = _mm_sub_ps(iz2,jz2);
1592             dx23             = _mm_sub_ps(ix2,jx3);
1593             dy23             = _mm_sub_ps(iy2,jy3);
1594             dz23             = _mm_sub_ps(iz2,jz3);
1595             dx31             = _mm_sub_ps(ix3,jx1);
1596             dy31             = _mm_sub_ps(iy3,jy1);
1597             dz31             = _mm_sub_ps(iz3,jz1);
1598             dx32             = _mm_sub_ps(ix3,jx2);
1599             dy32             = _mm_sub_ps(iy3,jy2);
1600             dz32             = _mm_sub_ps(iz3,jz2);
1601             dx33             = _mm_sub_ps(ix3,jx3);
1602             dy33             = _mm_sub_ps(iy3,jy3);
1603             dz33             = _mm_sub_ps(iz3,jz3);
1604
1605             /* Calculate squared distance and things based on it */
1606             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1607             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1608             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1609             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1610             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1611             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1612             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1613             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1614             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1615             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1616
1617             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1618             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1619             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1620             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1621             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1622             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1623             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1624             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1625             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1626             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1627
1628             fjx0             = _mm_setzero_ps();
1629             fjy0             = _mm_setzero_ps();
1630             fjz0             = _mm_setzero_ps();
1631             fjx1             = _mm_setzero_ps();
1632             fjy1             = _mm_setzero_ps();
1633             fjz1             = _mm_setzero_ps();
1634             fjx2             = _mm_setzero_ps();
1635             fjy2             = _mm_setzero_ps();
1636             fjz2             = _mm_setzero_ps();
1637             fjx3             = _mm_setzero_ps();
1638             fjy3             = _mm_setzero_ps();
1639             fjz3             = _mm_setzero_ps();
1640
1641             /**************************
1642              * CALCULATE INTERACTIONS *
1643              **************************/
1644
1645             r00              = _mm_mul_ps(rsq00,rinv00);
1646
1647             /* Calculate table index by multiplying r with table scale and truncate to integer */
1648             rt               = _mm_mul_ps(r00,vftabscale);
1649             vfitab           = _mm_cvttps_epi32(rt);
1650             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1651             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1652
1653             /* CUBIC SPLINE TABLE DISPERSION */
1654             vfitab           = _mm_add_epi32(vfitab,ifour);
1655             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1656             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1657             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1658             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1659             _MM_TRANSPOSE4_PS(Y,F,G,H);
1660             Heps             = _mm_mul_ps(vfeps,H);
1661             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1662             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1663             fvdw6            = _mm_mul_ps(c6_00,FF);
1664
1665             /* CUBIC SPLINE TABLE REPULSION */
1666             vfitab           = _mm_add_epi32(vfitab,ifour);
1667             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1668             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1669             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1670             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1671             _MM_TRANSPOSE4_PS(Y,F,G,H);
1672             Heps             = _mm_mul_ps(vfeps,H);
1673             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1674             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1675             fvdw12           = _mm_mul_ps(c12_00,FF);
1676             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1677
1678             fscal            = fvdw;
1679
1680             /* Calculate temporary vectorial force */
1681             tx               = _mm_mul_ps(fscal,dx00);
1682             ty               = _mm_mul_ps(fscal,dy00);
1683             tz               = _mm_mul_ps(fscal,dz00);
1684
1685             /* Update vectorial force */
1686             fix0             = _mm_add_ps(fix0,tx);
1687             fiy0             = _mm_add_ps(fiy0,ty);
1688             fiz0             = _mm_add_ps(fiz0,tz);
1689
1690             fjx0             = _mm_add_ps(fjx0,tx);
1691             fjy0             = _mm_add_ps(fjy0,ty);
1692             fjz0             = _mm_add_ps(fjz0,tz);
1693
1694             /**************************
1695              * CALCULATE INTERACTIONS *
1696              **************************/
1697
1698             r11              = _mm_mul_ps(rsq11,rinv11);
1699
1700             /* Calculate table index by multiplying r with table scale and truncate to integer */
1701             rt               = _mm_mul_ps(r11,vftabscale);
1702             vfitab           = _mm_cvttps_epi32(rt);
1703             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1704             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1705
1706             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1707             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1708             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1709             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1710             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1711             _MM_TRANSPOSE4_PS(Y,F,G,H);
1712             Heps             = _mm_mul_ps(vfeps,H);
1713             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1714             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1715             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
1716
1717             fscal            = felec;
1718
1719             /* Calculate temporary vectorial force */
1720             tx               = _mm_mul_ps(fscal,dx11);
1721             ty               = _mm_mul_ps(fscal,dy11);
1722             tz               = _mm_mul_ps(fscal,dz11);
1723
1724             /* Update vectorial force */
1725             fix1             = _mm_add_ps(fix1,tx);
1726             fiy1             = _mm_add_ps(fiy1,ty);
1727             fiz1             = _mm_add_ps(fiz1,tz);
1728
1729             fjx1             = _mm_add_ps(fjx1,tx);
1730             fjy1             = _mm_add_ps(fjy1,ty);
1731             fjz1             = _mm_add_ps(fjz1,tz);
1732
1733             /**************************
1734              * CALCULATE INTERACTIONS *
1735              **************************/
1736
1737             r12              = _mm_mul_ps(rsq12,rinv12);
1738
1739             /* Calculate table index by multiplying r with table scale and truncate to integer */
1740             rt               = _mm_mul_ps(r12,vftabscale);
1741             vfitab           = _mm_cvttps_epi32(rt);
1742             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1743             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1744
1745             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1746             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1747             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1748             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1749             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1750             _MM_TRANSPOSE4_PS(Y,F,G,H);
1751             Heps             = _mm_mul_ps(vfeps,H);
1752             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1753             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1754             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
1755
1756             fscal            = felec;
1757
1758             /* Calculate temporary vectorial force */
1759             tx               = _mm_mul_ps(fscal,dx12);
1760             ty               = _mm_mul_ps(fscal,dy12);
1761             tz               = _mm_mul_ps(fscal,dz12);
1762
1763             /* Update vectorial force */
1764             fix1             = _mm_add_ps(fix1,tx);
1765             fiy1             = _mm_add_ps(fiy1,ty);
1766             fiz1             = _mm_add_ps(fiz1,tz);
1767
1768             fjx2             = _mm_add_ps(fjx2,tx);
1769             fjy2             = _mm_add_ps(fjy2,ty);
1770             fjz2             = _mm_add_ps(fjz2,tz);
1771
1772             /**************************
1773              * CALCULATE INTERACTIONS *
1774              **************************/
1775
1776             r13              = _mm_mul_ps(rsq13,rinv13);
1777
1778             /* Calculate table index by multiplying r with table scale and truncate to integer */
1779             rt               = _mm_mul_ps(r13,vftabscale);
1780             vfitab           = _mm_cvttps_epi32(rt);
1781             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1782             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1783
1784             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1785             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1786             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1787             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1788             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1789             _MM_TRANSPOSE4_PS(Y,F,G,H);
1790             Heps             = _mm_mul_ps(vfeps,H);
1791             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1792             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1793             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1794
1795             fscal            = felec;
1796
1797             /* Calculate temporary vectorial force */
1798             tx               = _mm_mul_ps(fscal,dx13);
1799             ty               = _mm_mul_ps(fscal,dy13);
1800             tz               = _mm_mul_ps(fscal,dz13);
1801
1802             /* Update vectorial force */
1803             fix1             = _mm_add_ps(fix1,tx);
1804             fiy1             = _mm_add_ps(fiy1,ty);
1805             fiz1             = _mm_add_ps(fiz1,tz);
1806
1807             fjx3             = _mm_add_ps(fjx3,tx);
1808             fjy3             = _mm_add_ps(fjy3,ty);
1809             fjz3             = _mm_add_ps(fjz3,tz);
1810
1811             /**************************
1812              * CALCULATE INTERACTIONS *
1813              **************************/
1814
1815             r21              = _mm_mul_ps(rsq21,rinv21);
1816
1817             /* Calculate table index by multiplying r with table scale and truncate to integer */
1818             rt               = _mm_mul_ps(r21,vftabscale);
1819             vfitab           = _mm_cvttps_epi32(rt);
1820             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1821             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1822
1823             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1824             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1825             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1826             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1827             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1828             _MM_TRANSPOSE4_PS(Y,F,G,H);
1829             Heps             = _mm_mul_ps(vfeps,H);
1830             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1831             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1832             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1833
1834             fscal            = felec;
1835
1836             /* Calculate temporary vectorial force */
1837             tx               = _mm_mul_ps(fscal,dx21);
1838             ty               = _mm_mul_ps(fscal,dy21);
1839             tz               = _mm_mul_ps(fscal,dz21);
1840
1841             /* Update vectorial force */
1842             fix2             = _mm_add_ps(fix2,tx);
1843             fiy2             = _mm_add_ps(fiy2,ty);
1844             fiz2             = _mm_add_ps(fiz2,tz);
1845
1846             fjx1             = _mm_add_ps(fjx1,tx);
1847             fjy1             = _mm_add_ps(fjy1,ty);
1848             fjz1             = _mm_add_ps(fjz1,tz);
1849
1850             /**************************
1851              * CALCULATE INTERACTIONS *
1852              **************************/
1853
1854             r22              = _mm_mul_ps(rsq22,rinv22);
1855
1856             /* Calculate table index by multiplying r with table scale and truncate to integer */
1857             rt               = _mm_mul_ps(r22,vftabscale);
1858             vfitab           = _mm_cvttps_epi32(rt);
1859             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1860             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1861
1862             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1863             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1864             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1865             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1866             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1867             _MM_TRANSPOSE4_PS(Y,F,G,H);
1868             Heps             = _mm_mul_ps(vfeps,H);
1869             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1870             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1871             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1872
1873             fscal            = felec;
1874
1875             /* Calculate temporary vectorial force */
1876             tx               = _mm_mul_ps(fscal,dx22);
1877             ty               = _mm_mul_ps(fscal,dy22);
1878             tz               = _mm_mul_ps(fscal,dz22);
1879
1880             /* Update vectorial force */
1881             fix2             = _mm_add_ps(fix2,tx);
1882             fiy2             = _mm_add_ps(fiy2,ty);
1883             fiz2             = _mm_add_ps(fiz2,tz);
1884
1885             fjx2             = _mm_add_ps(fjx2,tx);
1886             fjy2             = _mm_add_ps(fjy2,ty);
1887             fjz2             = _mm_add_ps(fjz2,tz);
1888
1889             /**************************
1890              * CALCULATE INTERACTIONS *
1891              **************************/
1892
1893             r23              = _mm_mul_ps(rsq23,rinv23);
1894
1895             /* Calculate table index by multiplying r with table scale and truncate to integer */
1896             rt               = _mm_mul_ps(r23,vftabscale);
1897             vfitab           = _mm_cvttps_epi32(rt);
1898             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1899             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1900
1901             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1902             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1903             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1904             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1905             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1906             _MM_TRANSPOSE4_PS(Y,F,G,H);
1907             Heps             = _mm_mul_ps(vfeps,H);
1908             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1909             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1910             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1911
1912             fscal            = felec;
1913
1914             /* Calculate temporary vectorial force */
1915             tx               = _mm_mul_ps(fscal,dx23);
1916             ty               = _mm_mul_ps(fscal,dy23);
1917             tz               = _mm_mul_ps(fscal,dz23);
1918
1919             /* Update vectorial force */
1920             fix2             = _mm_add_ps(fix2,tx);
1921             fiy2             = _mm_add_ps(fiy2,ty);
1922             fiz2             = _mm_add_ps(fiz2,tz);
1923
1924             fjx3             = _mm_add_ps(fjx3,tx);
1925             fjy3             = _mm_add_ps(fjy3,ty);
1926             fjz3             = _mm_add_ps(fjz3,tz);
1927
1928             /**************************
1929              * CALCULATE INTERACTIONS *
1930              **************************/
1931
1932             r31              = _mm_mul_ps(rsq31,rinv31);
1933
1934             /* Calculate table index by multiplying r with table scale and truncate to integer */
1935             rt               = _mm_mul_ps(r31,vftabscale);
1936             vfitab           = _mm_cvttps_epi32(rt);
1937             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1938             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1939
1940             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1941             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1942             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1943             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1944             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1945             _MM_TRANSPOSE4_PS(Y,F,G,H);
1946             Heps             = _mm_mul_ps(vfeps,H);
1947             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1948             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1949             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1950
1951             fscal            = felec;
1952
1953             /* Calculate temporary vectorial force */
1954             tx               = _mm_mul_ps(fscal,dx31);
1955             ty               = _mm_mul_ps(fscal,dy31);
1956             tz               = _mm_mul_ps(fscal,dz31);
1957
1958             /* Update vectorial force */
1959             fix3             = _mm_add_ps(fix3,tx);
1960             fiy3             = _mm_add_ps(fiy3,ty);
1961             fiz3             = _mm_add_ps(fiz3,tz);
1962
1963             fjx1             = _mm_add_ps(fjx1,tx);
1964             fjy1             = _mm_add_ps(fjy1,ty);
1965             fjz1             = _mm_add_ps(fjz1,tz);
1966
1967             /**************************
1968              * CALCULATE INTERACTIONS *
1969              **************************/
1970
1971             r32              = _mm_mul_ps(rsq32,rinv32);
1972
1973             /* Calculate table index by multiplying r with table scale and truncate to integer */
1974             rt               = _mm_mul_ps(r32,vftabscale);
1975             vfitab           = _mm_cvttps_epi32(rt);
1976             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1977             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
1978
1979             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1980             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1981             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1982             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1983             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1984             _MM_TRANSPOSE4_PS(Y,F,G,H);
1985             Heps             = _mm_mul_ps(vfeps,H);
1986             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1987             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1988             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1989
1990             fscal            = felec;
1991
1992             /* Calculate temporary vectorial force */
1993             tx               = _mm_mul_ps(fscal,dx32);
1994             ty               = _mm_mul_ps(fscal,dy32);
1995             tz               = _mm_mul_ps(fscal,dz32);
1996
1997             /* Update vectorial force */
1998             fix3             = _mm_add_ps(fix3,tx);
1999             fiy3             = _mm_add_ps(fiy3,ty);
2000             fiz3             = _mm_add_ps(fiz3,tz);
2001
2002             fjx2             = _mm_add_ps(fjx2,tx);
2003             fjy2             = _mm_add_ps(fjy2,ty);
2004             fjz2             = _mm_add_ps(fjz2,tz);
2005
2006             /**************************
2007              * CALCULATE INTERACTIONS *
2008              **************************/
2009
2010             r33              = _mm_mul_ps(rsq33,rinv33);
2011
2012             /* Calculate table index by multiplying r with table scale and truncate to integer */
2013             rt               = _mm_mul_ps(r33,vftabscale);
2014             vfitab           = _mm_cvttps_epi32(rt);
2015             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2016             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2017
2018             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2019             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2020             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2021             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2022             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2023             _MM_TRANSPOSE4_PS(Y,F,G,H);
2024             Heps             = _mm_mul_ps(vfeps,H);
2025             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2026             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2027             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2028
2029             fscal            = felec;
2030
2031             /* Calculate temporary vectorial force */
2032             tx               = _mm_mul_ps(fscal,dx33);
2033             ty               = _mm_mul_ps(fscal,dy33);
2034             tz               = _mm_mul_ps(fscal,dz33);
2035
2036             /* Update vectorial force */
2037             fix3             = _mm_add_ps(fix3,tx);
2038             fiy3             = _mm_add_ps(fiy3,ty);
2039             fiz3             = _mm_add_ps(fiz3,tz);
2040
2041             fjx3             = _mm_add_ps(fjx3,tx);
2042             fjy3             = _mm_add_ps(fjy3,ty);
2043             fjz3             = _mm_add_ps(fjz3,tz);
2044
2045             fjptrA             = f+j_coord_offsetA;
2046             fjptrB             = f+j_coord_offsetB;
2047             fjptrC             = f+j_coord_offsetC;
2048             fjptrD             = f+j_coord_offsetD;
2049
2050             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2051                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2052                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2053
2054             /* Inner loop uses 402 flops */
2055         }
2056
2057         if(jidx<j_index_end)
2058         {
2059
2060             /* Get j neighbor index, and coordinate index */
2061             jnrlistA         = jjnr[jidx];
2062             jnrlistB         = jjnr[jidx+1];
2063             jnrlistC         = jjnr[jidx+2];
2064             jnrlistD         = jjnr[jidx+3];
2065             /* Sign of each element will be negative for non-real atoms.
2066              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
2067              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
2068              */
2069             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
2070             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
2071             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
2072             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
2073             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
2074             j_coord_offsetA  = DIM*jnrA;
2075             j_coord_offsetB  = DIM*jnrB;
2076             j_coord_offsetC  = DIM*jnrC;
2077             j_coord_offsetD  = DIM*jnrD;
2078
2079             /* load j atom coordinates */
2080             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
2081                                               x+j_coord_offsetC,x+j_coord_offsetD,
2082                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
2083                                               &jy2,&jz2,&jx3,&jy3,&jz3);
2084
2085             /* Calculate displacement vector */
2086             dx00             = _mm_sub_ps(ix0,jx0);
2087             dy00             = _mm_sub_ps(iy0,jy0);
2088             dz00             = _mm_sub_ps(iz0,jz0);
2089             dx11             = _mm_sub_ps(ix1,jx1);
2090             dy11             = _mm_sub_ps(iy1,jy1);
2091             dz11             = _mm_sub_ps(iz1,jz1);
2092             dx12             = _mm_sub_ps(ix1,jx2);
2093             dy12             = _mm_sub_ps(iy1,jy2);
2094             dz12             = _mm_sub_ps(iz1,jz2);
2095             dx13             = _mm_sub_ps(ix1,jx3);
2096             dy13             = _mm_sub_ps(iy1,jy3);
2097             dz13             = _mm_sub_ps(iz1,jz3);
2098             dx21             = _mm_sub_ps(ix2,jx1);
2099             dy21             = _mm_sub_ps(iy2,jy1);
2100             dz21             = _mm_sub_ps(iz2,jz1);
2101             dx22             = _mm_sub_ps(ix2,jx2);
2102             dy22             = _mm_sub_ps(iy2,jy2);
2103             dz22             = _mm_sub_ps(iz2,jz2);
2104             dx23             = _mm_sub_ps(ix2,jx3);
2105             dy23             = _mm_sub_ps(iy2,jy3);
2106             dz23             = _mm_sub_ps(iz2,jz3);
2107             dx31             = _mm_sub_ps(ix3,jx1);
2108             dy31             = _mm_sub_ps(iy3,jy1);
2109             dz31             = _mm_sub_ps(iz3,jz1);
2110             dx32             = _mm_sub_ps(ix3,jx2);
2111             dy32             = _mm_sub_ps(iy3,jy2);
2112             dz32             = _mm_sub_ps(iz3,jz2);
2113             dx33             = _mm_sub_ps(ix3,jx3);
2114             dy33             = _mm_sub_ps(iy3,jy3);
2115             dz33             = _mm_sub_ps(iz3,jz3);
2116
2117             /* Calculate squared distance and things based on it */
2118             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2119             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2120             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2121             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2122             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2123             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2124             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2125             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2126             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2127             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2128
2129             rinv00           = gmx_mm_invsqrt_ps(rsq00);
2130             rinv11           = gmx_mm_invsqrt_ps(rsq11);
2131             rinv12           = gmx_mm_invsqrt_ps(rsq12);
2132             rinv13           = gmx_mm_invsqrt_ps(rsq13);
2133             rinv21           = gmx_mm_invsqrt_ps(rsq21);
2134             rinv22           = gmx_mm_invsqrt_ps(rsq22);
2135             rinv23           = gmx_mm_invsqrt_ps(rsq23);
2136             rinv31           = gmx_mm_invsqrt_ps(rsq31);
2137             rinv32           = gmx_mm_invsqrt_ps(rsq32);
2138             rinv33           = gmx_mm_invsqrt_ps(rsq33);
2139
2140             fjx0             = _mm_setzero_ps();
2141             fjy0             = _mm_setzero_ps();
2142             fjz0             = _mm_setzero_ps();
2143             fjx1             = _mm_setzero_ps();
2144             fjy1             = _mm_setzero_ps();
2145             fjz1             = _mm_setzero_ps();
2146             fjx2             = _mm_setzero_ps();
2147             fjy2             = _mm_setzero_ps();
2148             fjz2             = _mm_setzero_ps();
2149             fjx3             = _mm_setzero_ps();
2150             fjy3             = _mm_setzero_ps();
2151             fjz3             = _mm_setzero_ps();
2152
2153             /**************************
2154              * CALCULATE INTERACTIONS *
2155              **************************/
2156
2157             r00              = _mm_mul_ps(rsq00,rinv00);
2158             r00              = _mm_andnot_ps(dummy_mask,r00);
2159
2160             /* Calculate table index by multiplying r with table scale and truncate to integer */
2161             rt               = _mm_mul_ps(r00,vftabscale);
2162             vfitab           = _mm_cvttps_epi32(rt);
2163             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2164             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2165
2166             /* CUBIC SPLINE TABLE DISPERSION */
2167             vfitab           = _mm_add_epi32(vfitab,ifour);
2168             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2169             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2170             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2171             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2172             _MM_TRANSPOSE4_PS(Y,F,G,H);
2173             Heps             = _mm_mul_ps(vfeps,H);
2174             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2175             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2176             fvdw6            = _mm_mul_ps(c6_00,FF);
2177
2178             /* CUBIC SPLINE TABLE REPULSION */
2179             vfitab           = _mm_add_epi32(vfitab,ifour);
2180             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2181             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2182             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2183             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2184             _MM_TRANSPOSE4_PS(Y,F,G,H);
2185             Heps             = _mm_mul_ps(vfeps,H);
2186             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2187             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2188             fvdw12           = _mm_mul_ps(c12_00,FF);
2189             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
2190
2191             fscal            = fvdw;
2192
2193             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2194
2195             /* Calculate temporary vectorial force */
2196             tx               = _mm_mul_ps(fscal,dx00);
2197             ty               = _mm_mul_ps(fscal,dy00);
2198             tz               = _mm_mul_ps(fscal,dz00);
2199
2200             /* Update vectorial force */
2201             fix0             = _mm_add_ps(fix0,tx);
2202             fiy0             = _mm_add_ps(fiy0,ty);
2203             fiz0             = _mm_add_ps(fiz0,tz);
2204
2205             fjx0             = _mm_add_ps(fjx0,tx);
2206             fjy0             = _mm_add_ps(fjy0,ty);
2207             fjz0             = _mm_add_ps(fjz0,tz);
2208
2209             /**************************
2210              * CALCULATE INTERACTIONS *
2211              **************************/
2212
2213             r11              = _mm_mul_ps(rsq11,rinv11);
2214             r11              = _mm_andnot_ps(dummy_mask,r11);
2215
2216             /* Calculate table index by multiplying r with table scale and truncate to integer */
2217             rt               = _mm_mul_ps(r11,vftabscale);
2218             vfitab           = _mm_cvttps_epi32(rt);
2219             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2220             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2221
2222             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2223             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2224             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2225             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2226             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2227             _MM_TRANSPOSE4_PS(Y,F,G,H);
2228             Heps             = _mm_mul_ps(vfeps,H);
2229             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2230             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2231             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
2232
2233             fscal            = felec;
2234
2235             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2236
2237             /* Calculate temporary vectorial force */
2238             tx               = _mm_mul_ps(fscal,dx11);
2239             ty               = _mm_mul_ps(fscal,dy11);
2240             tz               = _mm_mul_ps(fscal,dz11);
2241
2242             /* Update vectorial force */
2243             fix1             = _mm_add_ps(fix1,tx);
2244             fiy1             = _mm_add_ps(fiy1,ty);
2245             fiz1             = _mm_add_ps(fiz1,tz);
2246
2247             fjx1             = _mm_add_ps(fjx1,tx);
2248             fjy1             = _mm_add_ps(fjy1,ty);
2249             fjz1             = _mm_add_ps(fjz1,tz);
2250
2251             /**************************
2252              * CALCULATE INTERACTIONS *
2253              **************************/
2254
2255             r12              = _mm_mul_ps(rsq12,rinv12);
2256             r12              = _mm_andnot_ps(dummy_mask,r12);
2257
2258             /* Calculate table index by multiplying r with table scale and truncate to integer */
2259             rt               = _mm_mul_ps(r12,vftabscale);
2260             vfitab           = _mm_cvttps_epi32(rt);
2261             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2262             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2263
2264             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2265             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2266             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2267             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2268             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2269             _MM_TRANSPOSE4_PS(Y,F,G,H);
2270             Heps             = _mm_mul_ps(vfeps,H);
2271             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2272             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2273             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
2274
2275             fscal            = felec;
2276
2277             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2278
2279             /* Calculate temporary vectorial force */
2280             tx               = _mm_mul_ps(fscal,dx12);
2281             ty               = _mm_mul_ps(fscal,dy12);
2282             tz               = _mm_mul_ps(fscal,dz12);
2283
2284             /* Update vectorial force */
2285             fix1             = _mm_add_ps(fix1,tx);
2286             fiy1             = _mm_add_ps(fiy1,ty);
2287             fiz1             = _mm_add_ps(fiz1,tz);
2288
2289             fjx2             = _mm_add_ps(fjx2,tx);
2290             fjy2             = _mm_add_ps(fjy2,ty);
2291             fjz2             = _mm_add_ps(fjz2,tz);
2292
2293             /**************************
2294              * CALCULATE INTERACTIONS *
2295              **************************/
2296
2297             r13              = _mm_mul_ps(rsq13,rinv13);
2298             r13              = _mm_andnot_ps(dummy_mask,r13);
2299
2300             /* Calculate table index by multiplying r with table scale and truncate to integer */
2301             rt               = _mm_mul_ps(r13,vftabscale);
2302             vfitab           = _mm_cvttps_epi32(rt);
2303             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2304             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2305
2306             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2307             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2308             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2309             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2310             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2311             _MM_TRANSPOSE4_PS(Y,F,G,H);
2312             Heps             = _mm_mul_ps(vfeps,H);
2313             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2314             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2315             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
2316
2317             fscal            = felec;
2318
2319             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2320
2321             /* Calculate temporary vectorial force */
2322             tx               = _mm_mul_ps(fscal,dx13);
2323             ty               = _mm_mul_ps(fscal,dy13);
2324             tz               = _mm_mul_ps(fscal,dz13);
2325
2326             /* Update vectorial force */
2327             fix1             = _mm_add_ps(fix1,tx);
2328             fiy1             = _mm_add_ps(fiy1,ty);
2329             fiz1             = _mm_add_ps(fiz1,tz);
2330
2331             fjx3             = _mm_add_ps(fjx3,tx);
2332             fjy3             = _mm_add_ps(fjy3,ty);
2333             fjz3             = _mm_add_ps(fjz3,tz);
2334
2335             /**************************
2336              * CALCULATE INTERACTIONS *
2337              **************************/
2338
2339             r21              = _mm_mul_ps(rsq21,rinv21);
2340             r21              = _mm_andnot_ps(dummy_mask,r21);
2341
2342             /* Calculate table index by multiplying r with table scale and truncate to integer */
2343             rt               = _mm_mul_ps(r21,vftabscale);
2344             vfitab           = _mm_cvttps_epi32(rt);
2345             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2346             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2347
2348             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2349             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2350             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2351             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2352             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2353             _MM_TRANSPOSE4_PS(Y,F,G,H);
2354             Heps             = _mm_mul_ps(vfeps,H);
2355             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2356             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2357             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
2358
2359             fscal            = felec;
2360
2361             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2362
2363             /* Calculate temporary vectorial force */
2364             tx               = _mm_mul_ps(fscal,dx21);
2365             ty               = _mm_mul_ps(fscal,dy21);
2366             tz               = _mm_mul_ps(fscal,dz21);
2367
2368             /* Update vectorial force */
2369             fix2             = _mm_add_ps(fix2,tx);
2370             fiy2             = _mm_add_ps(fiy2,ty);
2371             fiz2             = _mm_add_ps(fiz2,tz);
2372
2373             fjx1             = _mm_add_ps(fjx1,tx);
2374             fjy1             = _mm_add_ps(fjy1,ty);
2375             fjz1             = _mm_add_ps(fjz1,tz);
2376
2377             /**************************
2378              * CALCULATE INTERACTIONS *
2379              **************************/
2380
2381             r22              = _mm_mul_ps(rsq22,rinv22);
2382             r22              = _mm_andnot_ps(dummy_mask,r22);
2383
2384             /* Calculate table index by multiplying r with table scale and truncate to integer */
2385             rt               = _mm_mul_ps(r22,vftabscale);
2386             vfitab           = _mm_cvttps_epi32(rt);
2387             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2388             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2389
2390             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2391             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2392             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2393             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2394             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2395             _MM_TRANSPOSE4_PS(Y,F,G,H);
2396             Heps             = _mm_mul_ps(vfeps,H);
2397             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2398             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2399             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
2400
2401             fscal            = felec;
2402
2403             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2404
2405             /* Calculate temporary vectorial force */
2406             tx               = _mm_mul_ps(fscal,dx22);
2407             ty               = _mm_mul_ps(fscal,dy22);
2408             tz               = _mm_mul_ps(fscal,dz22);
2409
2410             /* Update vectorial force */
2411             fix2             = _mm_add_ps(fix2,tx);
2412             fiy2             = _mm_add_ps(fiy2,ty);
2413             fiz2             = _mm_add_ps(fiz2,tz);
2414
2415             fjx2             = _mm_add_ps(fjx2,tx);
2416             fjy2             = _mm_add_ps(fjy2,ty);
2417             fjz2             = _mm_add_ps(fjz2,tz);
2418
2419             /**************************
2420              * CALCULATE INTERACTIONS *
2421              **************************/
2422
2423             r23              = _mm_mul_ps(rsq23,rinv23);
2424             r23              = _mm_andnot_ps(dummy_mask,r23);
2425
2426             /* Calculate table index by multiplying r with table scale and truncate to integer */
2427             rt               = _mm_mul_ps(r23,vftabscale);
2428             vfitab           = _mm_cvttps_epi32(rt);
2429             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2430             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2431
2432             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2433             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2434             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2435             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2436             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2437             _MM_TRANSPOSE4_PS(Y,F,G,H);
2438             Heps             = _mm_mul_ps(vfeps,H);
2439             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2440             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2441             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
2442
2443             fscal            = felec;
2444
2445             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2446
2447             /* Calculate temporary vectorial force */
2448             tx               = _mm_mul_ps(fscal,dx23);
2449             ty               = _mm_mul_ps(fscal,dy23);
2450             tz               = _mm_mul_ps(fscal,dz23);
2451
2452             /* Update vectorial force */
2453             fix2             = _mm_add_ps(fix2,tx);
2454             fiy2             = _mm_add_ps(fiy2,ty);
2455             fiz2             = _mm_add_ps(fiz2,tz);
2456
2457             fjx3             = _mm_add_ps(fjx3,tx);
2458             fjy3             = _mm_add_ps(fjy3,ty);
2459             fjz3             = _mm_add_ps(fjz3,tz);
2460
2461             /**************************
2462              * CALCULATE INTERACTIONS *
2463              **************************/
2464
2465             r31              = _mm_mul_ps(rsq31,rinv31);
2466             r31              = _mm_andnot_ps(dummy_mask,r31);
2467
2468             /* Calculate table index by multiplying r with table scale and truncate to integer */
2469             rt               = _mm_mul_ps(r31,vftabscale);
2470             vfitab           = _mm_cvttps_epi32(rt);
2471             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2472             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2473
2474             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2475             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2476             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2477             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2478             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2479             _MM_TRANSPOSE4_PS(Y,F,G,H);
2480             Heps             = _mm_mul_ps(vfeps,H);
2481             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2482             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2483             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
2484
2485             fscal            = felec;
2486
2487             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2488
2489             /* Calculate temporary vectorial force */
2490             tx               = _mm_mul_ps(fscal,dx31);
2491             ty               = _mm_mul_ps(fscal,dy31);
2492             tz               = _mm_mul_ps(fscal,dz31);
2493
2494             /* Update vectorial force */
2495             fix3             = _mm_add_ps(fix3,tx);
2496             fiy3             = _mm_add_ps(fiy3,ty);
2497             fiz3             = _mm_add_ps(fiz3,tz);
2498
2499             fjx1             = _mm_add_ps(fjx1,tx);
2500             fjy1             = _mm_add_ps(fjy1,ty);
2501             fjz1             = _mm_add_ps(fjz1,tz);
2502
2503             /**************************
2504              * CALCULATE INTERACTIONS *
2505              **************************/
2506
2507             r32              = _mm_mul_ps(rsq32,rinv32);
2508             r32              = _mm_andnot_ps(dummy_mask,r32);
2509
2510             /* Calculate table index by multiplying r with table scale and truncate to integer */
2511             rt               = _mm_mul_ps(r32,vftabscale);
2512             vfitab           = _mm_cvttps_epi32(rt);
2513             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2514             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2515
2516             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2517             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2518             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2519             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2520             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2521             _MM_TRANSPOSE4_PS(Y,F,G,H);
2522             Heps             = _mm_mul_ps(vfeps,H);
2523             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2524             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2525             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
2526
2527             fscal            = felec;
2528
2529             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2530
2531             /* Calculate temporary vectorial force */
2532             tx               = _mm_mul_ps(fscal,dx32);
2533             ty               = _mm_mul_ps(fscal,dy32);
2534             tz               = _mm_mul_ps(fscal,dz32);
2535
2536             /* Update vectorial force */
2537             fix3             = _mm_add_ps(fix3,tx);
2538             fiy3             = _mm_add_ps(fiy3,ty);
2539             fiz3             = _mm_add_ps(fiz3,tz);
2540
2541             fjx2             = _mm_add_ps(fjx2,tx);
2542             fjy2             = _mm_add_ps(fjy2,ty);
2543             fjz2             = _mm_add_ps(fjz2,tz);
2544
2545             /**************************
2546              * CALCULATE INTERACTIONS *
2547              **************************/
2548
2549             r33              = _mm_mul_ps(rsq33,rinv33);
2550             r33              = _mm_andnot_ps(dummy_mask,r33);
2551
2552             /* Calculate table index by multiplying r with table scale and truncate to integer */
2553             rt               = _mm_mul_ps(r33,vftabscale);
2554             vfitab           = _mm_cvttps_epi32(rt);
2555             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2556             vfitab           = _mm_slli_epi32(_mm_add_epi32(vfitab,_mm_slli_epi32(vfitab,1)),2);
2557
2558             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2559             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2560             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2561             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2562             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2563             _MM_TRANSPOSE4_PS(Y,F,G,H);
2564             Heps             = _mm_mul_ps(vfeps,H);
2565             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2566             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2567             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2568
2569             fscal            = felec;
2570
2571             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2572
2573             /* Calculate temporary vectorial force */
2574             tx               = _mm_mul_ps(fscal,dx33);
2575             ty               = _mm_mul_ps(fscal,dy33);
2576             tz               = _mm_mul_ps(fscal,dz33);
2577
2578             /* Update vectorial force */
2579             fix3             = _mm_add_ps(fix3,tx);
2580             fiy3             = _mm_add_ps(fiy3,ty);
2581             fiz3             = _mm_add_ps(fiz3,tz);
2582
2583             fjx3             = _mm_add_ps(fjx3,tx);
2584             fjy3             = _mm_add_ps(fjy3,ty);
2585             fjz3             = _mm_add_ps(fjz3,tz);
2586
2587             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2588             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2589             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2590             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2591
2592             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
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 24 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*24 + inneriter*412);
2616 }