src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_single/nb_kernel_ElecCSTab_VdwLJ_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_VdwLJ_GeomW4W4_VF_sse2_single
  38  * Electrostatics interaction: CubicSplineTable
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCSTab_VdwLJ_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->data;
 129     vftabscale       = _mm_set1_ps(kernel_data->table_elec->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             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 269             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 270             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 271             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 272             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 273             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 274             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 275             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 276             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 277
 278             rinvsq00         = gmx_mm_inv_ps(rsq00);
 279
 280             fjx0             = _mm_setzero_ps();
 281             fjy0             = _mm_setzero_ps();
 282             fjz0             = _mm_setzero_ps();
 283             fjx1             = _mm_setzero_ps();
 284             fjy1             = _mm_setzero_ps();
 285             fjz1             = _mm_setzero_ps();
 286             fjx2             = _mm_setzero_ps();
 287             fjy2             = _mm_setzero_ps();
 288             fjz2             = _mm_setzero_ps();
 289             fjx3             = _mm_setzero_ps();
 290             fjy3             = _mm_setzero_ps();
 291             fjz3             = _mm_setzero_ps();
 292
 293             /**************************
 294              * CALCULATE INTERACTIONS *
 295              **************************/
 296
 297             /* LENNARD-JONES DISPERSION/REPULSION */
 298
 299             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 300             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 301             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 302             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 303             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 304
 305             /* Update potential sum for this i atom from the interaction with this j atom. */
 306             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 307
 308             fscal            = fvdw;
 309
 310             /* Calculate temporary vectorial force */
 311             tx               = _mm_mul_ps(fscal,dx00);
 312             ty               = _mm_mul_ps(fscal,dy00);
 313             tz               = _mm_mul_ps(fscal,dz00);
 314
 315             /* Update vectorial force */
 316             fix0             = _mm_add_ps(fix0,tx);
 317             fiy0             = _mm_add_ps(fiy0,ty);
 318             fiz0             = _mm_add_ps(fiz0,tz);
 319
 320             fjx0             = _mm_add_ps(fjx0,tx);
 321             fjy0             = _mm_add_ps(fjy0,ty);
 322             fjz0             = _mm_add_ps(fjz0,tz);
 323
 324             /**************************
 325              * CALCULATE INTERACTIONS *
 326              **************************/
 327
 328             r11              = _mm_mul_ps(rsq11,rinv11);
 329
 330             /* Calculate table index by multiplying r with table scale and truncate to integer */
 331             rt               = _mm_mul_ps(r11,vftabscale);
 332             vfitab           = _mm_cvttps_epi32(rt);
 333             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 334             vfitab           = _mm_slli_epi32(vfitab,2);
 335
 336             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 337             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 338             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 339             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 340             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 341             _MM_TRANSPOSE4_PS(Y,F,G,H);
 342             Heps             = _mm_mul_ps(vfeps,H);
 343             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 344             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 345             velec            = _mm_mul_ps(qq11,VV);
 346             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 347             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 348
 349             /* Update potential sum for this i atom from the interaction with this j atom. */
 350             velecsum         = _mm_add_ps(velecsum,velec);
 351
 352             fscal            = felec;
 353
 354             /* Calculate temporary vectorial force */
 355             tx               = _mm_mul_ps(fscal,dx11);
 356             ty               = _mm_mul_ps(fscal,dy11);
 357             tz               = _mm_mul_ps(fscal,dz11);
 358
 359             /* Update vectorial force */
 360             fix1             = _mm_add_ps(fix1,tx);
 361             fiy1             = _mm_add_ps(fiy1,ty);
 362             fiz1             = _mm_add_ps(fiz1,tz);
 363
 364             fjx1             = _mm_add_ps(fjx1,tx);
 365             fjy1             = _mm_add_ps(fjy1,ty);
 366             fjz1             = _mm_add_ps(fjz1,tz);
 367
 368             /**************************
 369              * CALCULATE INTERACTIONS *
 370              **************************/
 371
 372             r12              = _mm_mul_ps(rsq12,rinv12);
 373
 374             /* Calculate table index by multiplying r with table scale and truncate to integer */
 375             rt               = _mm_mul_ps(r12,vftabscale);
 376             vfitab           = _mm_cvttps_epi32(rt);
 377             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 378             vfitab           = _mm_slli_epi32(vfitab,2);
 379
 380             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 381             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 382             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 383             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 384             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 385             _MM_TRANSPOSE4_PS(Y,F,G,H);
 386             Heps             = _mm_mul_ps(vfeps,H);
 387             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 388             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 389             velec            = _mm_mul_ps(qq12,VV);
 390             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 391             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 392
 393             /* Update potential sum for this i atom from the interaction with this j atom. */
 394             velecsum         = _mm_add_ps(velecsum,velec);
 395
 396             fscal            = felec;
 397
 398             /* Calculate temporary vectorial force */
 399             tx               = _mm_mul_ps(fscal,dx12);
 400             ty               = _mm_mul_ps(fscal,dy12);
 401             tz               = _mm_mul_ps(fscal,dz12);
 402
 403             /* Update vectorial force */
 404             fix1             = _mm_add_ps(fix1,tx);
 405             fiy1             = _mm_add_ps(fiy1,ty);
 406             fiz1             = _mm_add_ps(fiz1,tz);
 407
 408             fjx2             = _mm_add_ps(fjx2,tx);
 409             fjy2             = _mm_add_ps(fjy2,ty);
 410             fjz2             = _mm_add_ps(fjz2,tz);
 411
 412             /**************************
 413              * CALCULATE INTERACTIONS *
 414              **************************/
 415
 416             r13              = _mm_mul_ps(rsq13,rinv13);
 417
 418             /* Calculate table index by multiplying r with table scale and truncate to integer */
 419             rt               = _mm_mul_ps(r13,vftabscale);
 420             vfitab           = _mm_cvttps_epi32(rt);
 421             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 422             vfitab           = _mm_slli_epi32(vfitab,2);
 423
 424             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 425             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 426             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 427             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 428             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 429             _MM_TRANSPOSE4_PS(Y,F,G,H);
 430             Heps             = _mm_mul_ps(vfeps,H);
 431             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 432             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 433             velec            = _mm_mul_ps(qq13,VV);
 434             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 435             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
 436
 437             /* Update potential sum for this i atom from the interaction with this j atom. */
 438             velecsum         = _mm_add_ps(velecsum,velec);
 439
 440             fscal            = felec;
 441
 442             /* Calculate temporary vectorial force */
 443             tx               = _mm_mul_ps(fscal,dx13);
 444             ty               = _mm_mul_ps(fscal,dy13);
 445             tz               = _mm_mul_ps(fscal,dz13);
 446
 447             /* Update vectorial force */
 448             fix1             = _mm_add_ps(fix1,tx);
 449             fiy1             = _mm_add_ps(fiy1,ty);
 450             fiz1             = _mm_add_ps(fiz1,tz);
 451
 452             fjx3             = _mm_add_ps(fjx3,tx);
 453             fjy3             = _mm_add_ps(fjy3,ty);
 454             fjz3             = _mm_add_ps(fjz3,tz);
 455
 456             /**************************
 457              * CALCULATE INTERACTIONS *
 458              **************************/
 459
 460             r21              = _mm_mul_ps(rsq21,rinv21);
 461
 462             /* Calculate table index by multiplying r with table scale and truncate to integer */
 463             rt               = _mm_mul_ps(r21,vftabscale);
 464             vfitab           = _mm_cvttps_epi32(rt);
 465             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 466             vfitab           = _mm_slli_epi32(vfitab,2);
 467
 468             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 469             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 470             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 471             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 472             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 473             _MM_TRANSPOSE4_PS(Y,F,G,H);
 474             Heps             = _mm_mul_ps(vfeps,H);
 475             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 476             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 477             velec            = _mm_mul_ps(qq21,VV);
 478             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 479             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
 480
 481             /* Update potential sum for this i atom from the interaction with this j atom. */
 482             velecsum         = _mm_add_ps(velecsum,velec);
 483
 484             fscal            = felec;
 485
 486             /* Calculate temporary vectorial force */
 487             tx               = _mm_mul_ps(fscal,dx21);
 488             ty               = _mm_mul_ps(fscal,dy21);
 489             tz               = _mm_mul_ps(fscal,dz21);
 490
 491             /* Update vectorial force */
 492             fix2             = _mm_add_ps(fix2,tx);
 493             fiy2             = _mm_add_ps(fiy2,ty);
 494             fiz2             = _mm_add_ps(fiz2,tz);
 495
 496             fjx1             = _mm_add_ps(fjx1,tx);
 497             fjy1             = _mm_add_ps(fjy1,ty);
 498             fjz1             = _mm_add_ps(fjz1,tz);
 499
 500             /**************************
 501              * CALCULATE INTERACTIONS *
 502              **************************/
 503
 504             r22              = _mm_mul_ps(rsq22,rinv22);
 505
 506             /* Calculate table index by multiplying r with table scale and truncate to integer */
 507             rt               = _mm_mul_ps(r22,vftabscale);
 508             vfitab           = _mm_cvttps_epi32(rt);
 509             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 510             vfitab           = _mm_slli_epi32(vfitab,2);
 511
 512             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 513             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 514             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 515             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 516             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 517             _MM_TRANSPOSE4_PS(Y,F,G,H);
 518             Heps             = _mm_mul_ps(vfeps,H);
 519             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 520             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 521             velec            = _mm_mul_ps(qq22,VV);
 522             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 523             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
 524
 525             /* Update potential sum for this i atom from the interaction with this j atom. */
 526             velecsum         = _mm_add_ps(velecsum,velec);
 527
 528             fscal            = felec;
 529
 530             /* Calculate temporary vectorial force */
 531             tx               = _mm_mul_ps(fscal,dx22);
 532             ty               = _mm_mul_ps(fscal,dy22);
 533             tz               = _mm_mul_ps(fscal,dz22);
 534
 535             /* Update vectorial force */
 536             fix2             = _mm_add_ps(fix2,tx);
 537             fiy2             = _mm_add_ps(fiy2,ty);
 538             fiz2             = _mm_add_ps(fiz2,tz);
 539
 540             fjx2             = _mm_add_ps(fjx2,tx);
 541             fjy2             = _mm_add_ps(fjy2,ty);
 542             fjz2             = _mm_add_ps(fjz2,tz);
 543
 544             /**************************
 545              * CALCULATE INTERACTIONS *
 546              **************************/
 547
 548             r23              = _mm_mul_ps(rsq23,rinv23);
 549
 550             /* Calculate table index by multiplying r with table scale and truncate to integer */
 551             rt               = _mm_mul_ps(r23,vftabscale);
 552             vfitab           = _mm_cvttps_epi32(rt);
 553             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 554             vfitab           = _mm_slli_epi32(vfitab,2);
 555
 556             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 557             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 558             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 559             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 560             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 561             _MM_TRANSPOSE4_PS(Y,F,G,H);
 562             Heps             = _mm_mul_ps(vfeps,H);
 563             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 564             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 565             velec            = _mm_mul_ps(qq23,VV);
 566             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 567             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
 568
 569             /* Update potential sum for this i atom from the interaction with this j atom. */
 570             velecsum         = _mm_add_ps(velecsum,velec);
 571
 572             fscal            = felec;
 573
 574             /* Calculate temporary vectorial force */
 575             tx               = _mm_mul_ps(fscal,dx23);
 576             ty               = _mm_mul_ps(fscal,dy23);
 577             tz               = _mm_mul_ps(fscal,dz23);
 578
 579             /* Update vectorial force */
 580             fix2             = _mm_add_ps(fix2,tx);
 581             fiy2             = _mm_add_ps(fiy2,ty);
 582             fiz2             = _mm_add_ps(fiz2,tz);
 583
 584             fjx3             = _mm_add_ps(fjx3,tx);
 585             fjy3             = _mm_add_ps(fjy3,ty);
 586             fjz3             = _mm_add_ps(fjz3,tz);
 587
 588             /**************************
 589              * CALCULATE INTERACTIONS *
 590              **************************/
 591
 592             r31              = _mm_mul_ps(rsq31,rinv31);
 593
 594             /* Calculate table index by multiplying r with table scale and truncate to integer */
 595             rt               = _mm_mul_ps(r31,vftabscale);
 596             vfitab           = _mm_cvttps_epi32(rt);
 597             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 598             vfitab           = _mm_slli_epi32(vfitab,2);
 599
 600             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 601             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 602             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 603             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 604             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 605             _MM_TRANSPOSE4_PS(Y,F,G,H);
 606             Heps             = _mm_mul_ps(vfeps,H);
 607             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 608             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 609             velec            = _mm_mul_ps(qq31,VV);
 610             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 611             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
 612
 613             /* Update potential sum for this i atom from the interaction with this j atom. */
 614             velecsum         = _mm_add_ps(velecsum,velec);
 615
 616             fscal            = felec;
 617
 618             /* Calculate temporary vectorial force */
 619             tx               = _mm_mul_ps(fscal,dx31);
 620             ty               = _mm_mul_ps(fscal,dy31);
 621             tz               = _mm_mul_ps(fscal,dz31);
 622
 623             /* Update vectorial force */
 624             fix3             = _mm_add_ps(fix3,tx);
 625             fiy3             = _mm_add_ps(fiy3,ty);
 626             fiz3             = _mm_add_ps(fiz3,tz);
 627
 628             fjx1             = _mm_add_ps(fjx1,tx);
 629             fjy1             = _mm_add_ps(fjy1,ty);
 630             fjz1             = _mm_add_ps(fjz1,tz);
 631
 632             /**************************
 633              * CALCULATE INTERACTIONS *
 634              **************************/
 635
 636             r32              = _mm_mul_ps(rsq32,rinv32);
 637
 638             /* Calculate table index by multiplying r with table scale and truncate to integer */
 639             rt               = _mm_mul_ps(r32,vftabscale);
 640             vfitab           = _mm_cvttps_epi32(rt);
 641             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 642             vfitab           = _mm_slli_epi32(vfitab,2);
 643
 644             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 645             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 646             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 647             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 648             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 649             _MM_TRANSPOSE4_PS(Y,F,G,H);
 650             Heps             = _mm_mul_ps(vfeps,H);
 651             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 652             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 653             velec            = _mm_mul_ps(qq32,VV);
 654             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 655             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
 656
 657             /* Update potential sum for this i atom from the interaction with this j atom. */
 658             velecsum         = _mm_add_ps(velecsum,velec);
 659
 660             fscal            = felec;
 661
 662             /* Calculate temporary vectorial force */
 663             tx               = _mm_mul_ps(fscal,dx32);
 664             ty               = _mm_mul_ps(fscal,dy32);
 665             tz               = _mm_mul_ps(fscal,dz32);
 666
 667             /* Update vectorial force */
 668             fix3             = _mm_add_ps(fix3,tx);
 669             fiy3             = _mm_add_ps(fiy3,ty);
 670             fiz3             = _mm_add_ps(fiz3,tz);
 671
 672             fjx2             = _mm_add_ps(fjx2,tx);
 673             fjy2             = _mm_add_ps(fjy2,ty);
 674             fjz2             = _mm_add_ps(fjz2,tz);
 675
 676             /**************************
 677              * CALCULATE INTERACTIONS *
 678              **************************/
 679
 680             r33              = _mm_mul_ps(rsq33,rinv33);
 681
 682             /* Calculate table index by multiplying r with table scale and truncate to integer */
 683             rt               = _mm_mul_ps(r33,vftabscale);
 684             vfitab           = _mm_cvttps_epi32(rt);
 685             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 686             vfitab           = _mm_slli_epi32(vfitab,2);
 687
 688             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 689             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 690             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 691             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 692             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 693             _MM_TRANSPOSE4_PS(Y,F,G,H);
 694             Heps             = _mm_mul_ps(vfeps,H);
 695             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 696             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 697             velec            = _mm_mul_ps(qq33,VV);
 698             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 699             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
 700
 701             /* Update potential sum for this i atom from the interaction with this j atom. */
 702             velecsum         = _mm_add_ps(velecsum,velec);
 703
 704             fscal            = felec;
 705
 706             /* Calculate temporary vectorial force */
 707             tx               = _mm_mul_ps(fscal,dx33);
 708             ty               = _mm_mul_ps(fscal,dy33);
 709             tz               = _mm_mul_ps(fscal,dz33);
 710
 711             /* Update vectorial force */
 712             fix3             = _mm_add_ps(fix3,tx);
 713             fiy3             = _mm_add_ps(fiy3,ty);
 714             fiz3             = _mm_add_ps(fiz3,tz);
 715
 716             fjx3             = _mm_add_ps(fjx3,tx);
 717             fjy3             = _mm_add_ps(fjy3,ty);
 718             fjz3             = _mm_add_ps(fjz3,tz);
 719
 720             fjptrA             = f+j_coord_offsetA;
 721             fjptrB             = f+j_coord_offsetB;
 722             fjptrC             = f+j_coord_offsetC;
 723             fjptrD             = f+j_coord_offsetD;
 724
 725             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 726                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 727                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 728
 729             /* Inner loop uses 422 flops */
 730         }
 731
 732         if(jidx<j_index_end)
 733         {
 734
 735             /* Get j neighbor index, and coordinate index */
 736             jnrlistA         = jjnr[jidx];
 737             jnrlistB         = jjnr[jidx+1];
 738             jnrlistC         = jjnr[jidx+2];
 739             jnrlistD         = jjnr[jidx+3];
 740             /* Sign of each element will be negative for non-real atoms.
 741              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 742              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 743              */
 744             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 745             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 746             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 747             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 748             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 749             j_coord_offsetA  = DIM*jnrA;
 750             j_coord_offsetB  = DIM*jnrB;
 751             j_coord_offsetC  = DIM*jnrC;
 752             j_coord_offsetD  = DIM*jnrD;
 753
 754             /* load j atom coordinates */
 755             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 756                                               x+j_coord_offsetC,x+j_coord_offsetD,
 757                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 758                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 759
 760             /* Calculate displacement vector */
 761             dx00             = _mm_sub_ps(ix0,jx0);
 762             dy00             = _mm_sub_ps(iy0,jy0);
 763             dz00             = _mm_sub_ps(iz0,jz0);
 764             dx11             = _mm_sub_ps(ix1,jx1);
 765             dy11             = _mm_sub_ps(iy1,jy1);
 766             dz11             = _mm_sub_ps(iz1,jz1);
 767             dx12             = _mm_sub_ps(ix1,jx2);
 768             dy12             = _mm_sub_ps(iy1,jy2);
 769             dz12             = _mm_sub_ps(iz1,jz2);
 770             dx13             = _mm_sub_ps(ix1,jx3);
 771             dy13             = _mm_sub_ps(iy1,jy3);
 772             dz13             = _mm_sub_ps(iz1,jz3);
 773             dx21             = _mm_sub_ps(ix2,jx1);
 774             dy21             = _mm_sub_ps(iy2,jy1);
 775             dz21             = _mm_sub_ps(iz2,jz1);
 776             dx22             = _mm_sub_ps(ix2,jx2);
 777             dy22             = _mm_sub_ps(iy2,jy2);
 778             dz22             = _mm_sub_ps(iz2,jz2);
 779             dx23             = _mm_sub_ps(ix2,jx3);
 780             dy23             = _mm_sub_ps(iy2,jy3);
 781             dz23             = _mm_sub_ps(iz2,jz3);
 782             dx31             = _mm_sub_ps(ix3,jx1);
 783             dy31             = _mm_sub_ps(iy3,jy1);
 784             dz31             = _mm_sub_ps(iz3,jz1);
 785             dx32             = _mm_sub_ps(ix3,jx2);
 786             dy32             = _mm_sub_ps(iy3,jy2);
 787             dz32             = _mm_sub_ps(iz3,jz2);
 788             dx33             = _mm_sub_ps(ix3,jx3);
 789             dy33             = _mm_sub_ps(iy3,jy3);
 790             dz33             = _mm_sub_ps(iz3,jz3);
 791
 792             /* Calculate squared distance and things based on it */
 793             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 794             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 795             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 796             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 797             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 798             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 799             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 800             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 801             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 802             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 803
 804             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 805             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 806             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 807             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 808             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 809             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 810             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 811             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 812             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 813
 814             rinvsq00         = gmx_mm_inv_ps(rsq00);
 815
 816             fjx0             = _mm_setzero_ps();
 817             fjy0             = _mm_setzero_ps();
 818             fjz0             = _mm_setzero_ps();
 819             fjx1             = _mm_setzero_ps();
 820             fjy1             = _mm_setzero_ps();
 821             fjz1             = _mm_setzero_ps();
 822             fjx2             = _mm_setzero_ps();
 823             fjy2             = _mm_setzero_ps();
 824             fjz2             = _mm_setzero_ps();
 825             fjx3             = _mm_setzero_ps();
 826             fjy3             = _mm_setzero_ps();
 827             fjz3             = _mm_setzero_ps();
 828
 829             /**************************
 830              * CALCULATE INTERACTIONS *
 831              **************************/
 832
 833             /* LENNARD-JONES DISPERSION/REPULSION */
 834
 835             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 836             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 837             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 838             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 839             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 840
 841             /* Update potential sum for this i atom from the interaction with this j atom. */
 842             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 843             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 844
 845             fscal            = fvdw;
 846
 847             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 848
 849             /* Calculate temporary vectorial force */
 850             tx               = _mm_mul_ps(fscal,dx00);
 851             ty               = _mm_mul_ps(fscal,dy00);
 852             tz               = _mm_mul_ps(fscal,dz00);
 853
 854             /* Update vectorial force */
 855             fix0             = _mm_add_ps(fix0,tx);
 856             fiy0             = _mm_add_ps(fiy0,ty);
 857             fiz0             = _mm_add_ps(fiz0,tz);
 858
 859             fjx0             = _mm_add_ps(fjx0,tx);
 860             fjy0             = _mm_add_ps(fjy0,ty);
 861             fjz0             = _mm_add_ps(fjz0,tz);
 862
 863             /**************************
 864              * CALCULATE INTERACTIONS *
 865              **************************/
 866
 867             r11              = _mm_mul_ps(rsq11,rinv11);
 868             r11              = _mm_andnot_ps(dummy_mask,r11);
 869
 870             /* Calculate table index by multiplying r with table scale and truncate to integer */
 871             rt               = _mm_mul_ps(r11,vftabscale);
 872             vfitab           = _mm_cvttps_epi32(rt);
 873             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 874             vfitab           = _mm_slli_epi32(vfitab,2);
 875
 876             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 877             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 878             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 879             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 880             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 881             _MM_TRANSPOSE4_PS(Y,F,G,H);
 882             Heps             = _mm_mul_ps(vfeps,H);
 883             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 884             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 885             velec            = _mm_mul_ps(qq11,VV);
 886             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 887             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
 888
 889             /* Update potential sum for this i atom from the interaction with this j atom. */
 890             velec            = _mm_andnot_ps(dummy_mask,velec);
 891             velecsum         = _mm_add_ps(velecsum,velec);
 892
 893             fscal            = felec;
 894
 895             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 896
 897             /* Calculate temporary vectorial force */
 898             tx               = _mm_mul_ps(fscal,dx11);
 899             ty               = _mm_mul_ps(fscal,dy11);
 900             tz               = _mm_mul_ps(fscal,dz11);
 901
 902             /* Update vectorial force */
 903             fix1             = _mm_add_ps(fix1,tx);
 904             fiy1             = _mm_add_ps(fiy1,ty);
 905             fiz1             = _mm_add_ps(fiz1,tz);
 906
 907             fjx1             = _mm_add_ps(fjx1,tx);
 908             fjy1             = _mm_add_ps(fjy1,ty);
 909             fjz1             = _mm_add_ps(fjz1,tz);
 910
 911             /**************************
 912              * CALCULATE INTERACTIONS *
 913              **************************/
 914
 915             r12              = _mm_mul_ps(rsq12,rinv12);
 916             r12              = _mm_andnot_ps(dummy_mask,r12);
 917
 918             /* Calculate table index by multiplying r with table scale and truncate to integer */
 919             rt               = _mm_mul_ps(r12,vftabscale);
 920             vfitab           = _mm_cvttps_epi32(rt);
 921             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 922             vfitab           = _mm_slli_epi32(vfitab,2);
 923
 924             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 925             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 926             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 927             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 928             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 929             _MM_TRANSPOSE4_PS(Y,F,G,H);
 930             Heps             = _mm_mul_ps(vfeps,H);
 931             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 932             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 933             velec            = _mm_mul_ps(qq12,VV);
 934             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 935             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
 936
 937             /* Update potential sum for this i atom from the interaction with this j atom. */
 938             velec            = _mm_andnot_ps(dummy_mask,velec);
 939             velecsum         = _mm_add_ps(velecsum,velec);
 940
 941             fscal            = felec;
 942
 943             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 944
 945             /* Calculate temporary vectorial force */
 946             tx               = _mm_mul_ps(fscal,dx12);
 947             ty               = _mm_mul_ps(fscal,dy12);
 948             tz               = _mm_mul_ps(fscal,dz12);
 949
 950             /* Update vectorial force */
 951             fix1             = _mm_add_ps(fix1,tx);
 952             fiy1             = _mm_add_ps(fiy1,ty);
 953             fiz1             = _mm_add_ps(fiz1,tz);
 954
 955             fjx2             = _mm_add_ps(fjx2,tx);
 956             fjy2             = _mm_add_ps(fjy2,ty);
 957             fjz2             = _mm_add_ps(fjz2,tz);
 958
 959             /**************************
 960              * CALCULATE INTERACTIONS *
 961              **************************/
 962
 963             r13              = _mm_mul_ps(rsq13,rinv13);
 964             r13              = _mm_andnot_ps(dummy_mask,r13);
 965
 966             /* Calculate table index by multiplying r with table scale and truncate to integer */
 967             rt               = _mm_mul_ps(r13,vftabscale);
 968             vfitab           = _mm_cvttps_epi32(rt);
 969             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
 970             vfitab           = _mm_slli_epi32(vfitab,2);
 971
 972             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 973             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
 974             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
 975             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
 976             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
 977             _MM_TRANSPOSE4_PS(Y,F,G,H);
 978             Heps             = _mm_mul_ps(vfeps,H);
 979             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
 980             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
 981             velec            = _mm_mul_ps(qq13,VV);
 982             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
 983             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
 984
 985             /* Update potential sum for this i atom from the interaction with this j atom. */
 986             velec            = _mm_andnot_ps(dummy_mask,velec);
 987             velecsum         = _mm_add_ps(velecsum,velec);
 988
 989             fscal            = felec;
 990
 991             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 992
 993             /* Calculate temporary vectorial force */
 994             tx               = _mm_mul_ps(fscal,dx13);
 995             ty               = _mm_mul_ps(fscal,dy13);
 996             tz               = _mm_mul_ps(fscal,dz13);
 997
 998             /* Update vectorial force */
 999             fix1             = _mm_add_ps(fix1,tx);
1000             fiy1             = _mm_add_ps(fiy1,ty);
1001             fiz1             = _mm_add_ps(fiz1,tz);
1002
1003             fjx3             = _mm_add_ps(fjx3,tx);
1004             fjy3             = _mm_add_ps(fjy3,ty);
1005             fjz3             = _mm_add_ps(fjz3,tz);
1006
1007             /**************************
1008              * CALCULATE INTERACTIONS *
1009              **************************/
1010
1011             r21              = _mm_mul_ps(rsq21,rinv21);
1012             r21              = _mm_andnot_ps(dummy_mask,r21);
1013
1014             /* Calculate table index by multiplying r with table scale and truncate to integer */
1015             rt               = _mm_mul_ps(r21,vftabscale);
1016             vfitab           = _mm_cvttps_epi32(rt);
1017             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1018             vfitab           = _mm_slli_epi32(vfitab,2);
1019
1020             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1021             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1022             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1023             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1024             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1025             _MM_TRANSPOSE4_PS(Y,F,G,H);
1026             Heps             = _mm_mul_ps(vfeps,H);
1027             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1028             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1029             velec            = _mm_mul_ps(qq21,VV);
1030             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1031             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1032
1033             /* Update potential sum for this i atom from the interaction with this j atom. */
1034             velec            = _mm_andnot_ps(dummy_mask,velec);
1035             velecsum         = _mm_add_ps(velecsum,velec);
1036
1037             fscal            = felec;
1038
1039             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1040
1041             /* Calculate temporary vectorial force */
1042             tx               = _mm_mul_ps(fscal,dx21);
1043             ty               = _mm_mul_ps(fscal,dy21);
1044             tz               = _mm_mul_ps(fscal,dz21);
1045
1046             /* Update vectorial force */
1047             fix2             = _mm_add_ps(fix2,tx);
1048             fiy2             = _mm_add_ps(fiy2,ty);
1049             fiz2             = _mm_add_ps(fiz2,tz);
1050
1051             fjx1             = _mm_add_ps(fjx1,tx);
1052             fjy1             = _mm_add_ps(fjy1,ty);
1053             fjz1             = _mm_add_ps(fjz1,tz);
1054
1055             /**************************
1056              * CALCULATE INTERACTIONS *
1057              **************************/
1058
1059             r22              = _mm_mul_ps(rsq22,rinv22);
1060             r22              = _mm_andnot_ps(dummy_mask,r22);
1061
1062             /* Calculate table index by multiplying r with table scale and truncate to integer */
1063             rt               = _mm_mul_ps(r22,vftabscale);
1064             vfitab           = _mm_cvttps_epi32(rt);
1065             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1066             vfitab           = _mm_slli_epi32(vfitab,2);
1067
1068             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1069             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1070             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1071             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1072             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1073             _MM_TRANSPOSE4_PS(Y,F,G,H);
1074             Heps             = _mm_mul_ps(vfeps,H);
1075             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1076             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1077             velec            = _mm_mul_ps(qq22,VV);
1078             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1079             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1080
1081             /* Update potential sum for this i atom from the interaction with this j atom. */
1082             velec            = _mm_andnot_ps(dummy_mask,velec);
1083             velecsum         = _mm_add_ps(velecsum,velec);
1084
1085             fscal            = felec;
1086
1087             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1088
1089             /* Calculate temporary vectorial force */
1090             tx               = _mm_mul_ps(fscal,dx22);
1091             ty               = _mm_mul_ps(fscal,dy22);
1092             tz               = _mm_mul_ps(fscal,dz22);
1093
1094             /* Update vectorial force */
1095             fix2             = _mm_add_ps(fix2,tx);
1096             fiy2             = _mm_add_ps(fiy2,ty);
1097             fiz2             = _mm_add_ps(fiz2,tz);
1098
1099             fjx2             = _mm_add_ps(fjx2,tx);
1100             fjy2             = _mm_add_ps(fjy2,ty);
1101             fjz2             = _mm_add_ps(fjz2,tz);
1102
1103             /**************************
1104              * CALCULATE INTERACTIONS *
1105              **************************/
1106
1107             r23              = _mm_mul_ps(rsq23,rinv23);
1108             r23              = _mm_andnot_ps(dummy_mask,r23);
1109
1110             /* Calculate table index by multiplying r with table scale and truncate to integer */
1111             rt               = _mm_mul_ps(r23,vftabscale);
1112             vfitab           = _mm_cvttps_epi32(rt);
1113             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1114             vfitab           = _mm_slli_epi32(vfitab,2);
1115
1116             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1117             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1118             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1119             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1120             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1121             _MM_TRANSPOSE4_PS(Y,F,G,H);
1122             Heps             = _mm_mul_ps(vfeps,H);
1123             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1124             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1125             velec            = _mm_mul_ps(qq23,VV);
1126             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1127             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1128
1129             /* Update potential sum for this i atom from the interaction with this j atom. */
1130             velec            = _mm_andnot_ps(dummy_mask,velec);
1131             velecsum         = _mm_add_ps(velecsum,velec);
1132
1133             fscal            = felec;
1134
1135             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1136
1137             /* Calculate temporary vectorial force */
1138             tx               = _mm_mul_ps(fscal,dx23);
1139             ty               = _mm_mul_ps(fscal,dy23);
1140             tz               = _mm_mul_ps(fscal,dz23);
1141
1142             /* Update vectorial force */
1143             fix2             = _mm_add_ps(fix2,tx);
1144             fiy2             = _mm_add_ps(fiy2,ty);
1145             fiz2             = _mm_add_ps(fiz2,tz);
1146
1147             fjx3             = _mm_add_ps(fjx3,tx);
1148             fjy3             = _mm_add_ps(fjy3,ty);
1149             fjz3             = _mm_add_ps(fjz3,tz);
1150
1151             /**************************
1152              * CALCULATE INTERACTIONS *
1153              **************************/
1154
1155             r31              = _mm_mul_ps(rsq31,rinv31);
1156             r31              = _mm_andnot_ps(dummy_mask,r31);
1157
1158             /* Calculate table index by multiplying r with table scale and truncate to integer */
1159             rt               = _mm_mul_ps(r31,vftabscale);
1160             vfitab           = _mm_cvttps_epi32(rt);
1161             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1162             vfitab           = _mm_slli_epi32(vfitab,2);
1163
1164             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1165             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1166             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1167             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1168             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1169             _MM_TRANSPOSE4_PS(Y,F,G,H);
1170             Heps             = _mm_mul_ps(vfeps,H);
1171             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1172             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1173             velec            = _mm_mul_ps(qq31,VV);
1174             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1175             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1176
1177             /* Update potential sum for this i atom from the interaction with this j atom. */
1178             velec            = _mm_andnot_ps(dummy_mask,velec);
1179             velecsum         = _mm_add_ps(velecsum,velec);
1180
1181             fscal            = felec;
1182
1183             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1184
1185             /* Calculate temporary vectorial force */
1186             tx               = _mm_mul_ps(fscal,dx31);
1187             ty               = _mm_mul_ps(fscal,dy31);
1188             tz               = _mm_mul_ps(fscal,dz31);
1189
1190             /* Update vectorial force */
1191             fix3             = _mm_add_ps(fix3,tx);
1192             fiy3             = _mm_add_ps(fiy3,ty);
1193             fiz3             = _mm_add_ps(fiz3,tz);
1194
1195             fjx1             = _mm_add_ps(fjx1,tx);
1196             fjy1             = _mm_add_ps(fjy1,ty);
1197             fjz1             = _mm_add_ps(fjz1,tz);
1198
1199             /**************************
1200              * CALCULATE INTERACTIONS *
1201              **************************/
1202
1203             r32              = _mm_mul_ps(rsq32,rinv32);
1204             r32              = _mm_andnot_ps(dummy_mask,r32);
1205
1206             /* Calculate table index by multiplying r with table scale and truncate to integer */
1207             rt               = _mm_mul_ps(r32,vftabscale);
1208             vfitab           = _mm_cvttps_epi32(rt);
1209             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1210             vfitab           = _mm_slli_epi32(vfitab,2);
1211
1212             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1213             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1214             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1215             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1216             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1217             _MM_TRANSPOSE4_PS(Y,F,G,H);
1218             Heps             = _mm_mul_ps(vfeps,H);
1219             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1220             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1221             velec            = _mm_mul_ps(qq32,VV);
1222             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1223             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1224
1225             /* Update potential sum for this i atom from the interaction with this j atom. */
1226             velec            = _mm_andnot_ps(dummy_mask,velec);
1227             velecsum         = _mm_add_ps(velecsum,velec);
1228
1229             fscal            = felec;
1230
1231             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1232
1233             /* Calculate temporary vectorial force */
1234             tx               = _mm_mul_ps(fscal,dx32);
1235             ty               = _mm_mul_ps(fscal,dy32);
1236             tz               = _mm_mul_ps(fscal,dz32);
1237
1238             /* Update vectorial force */
1239             fix3             = _mm_add_ps(fix3,tx);
1240             fiy3             = _mm_add_ps(fiy3,ty);
1241             fiz3             = _mm_add_ps(fiz3,tz);
1242
1243             fjx2             = _mm_add_ps(fjx2,tx);
1244             fjy2             = _mm_add_ps(fjy2,ty);
1245             fjz2             = _mm_add_ps(fjz2,tz);
1246
1247             /**************************
1248              * CALCULATE INTERACTIONS *
1249              **************************/
1250
1251             r33              = _mm_mul_ps(rsq33,rinv33);
1252             r33              = _mm_andnot_ps(dummy_mask,r33);
1253
1254             /* Calculate table index by multiplying r with table scale and truncate to integer */
1255             rt               = _mm_mul_ps(r33,vftabscale);
1256             vfitab           = _mm_cvttps_epi32(rt);
1257             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1258             vfitab           = _mm_slli_epi32(vfitab,2);
1259
1260             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1261             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1262             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1263             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1264             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1265             _MM_TRANSPOSE4_PS(Y,F,G,H);
1266             Heps             = _mm_mul_ps(vfeps,H);
1267             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1268             VV               = _mm_add_ps(Y,_mm_mul_ps(vfeps,Fp));
1269             velec            = _mm_mul_ps(qq33,VV);
1270             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1271             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
1272
1273             /* Update potential sum for this i atom from the interaction with this j atom. */
1274             velec            = _mm_andnot_ps(dummy_mask,velec);
1275             velecsum         = _mm_add_ps(velecsum,velec);
1276
1277             fscal            = felec;
1278
1279             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1280
1281             /* Calculate temporary vectorial force */
1282             tx               = _mm_mul_ps(fscal,dx33);
1283             ty               = _mm_mul_ps(fscal,dy33);
1284             tz               = _mm_mul_ps(fscal,dz33);
1285
1286             /* Update vectorial force */
1287             fix3             = _mm_add_ps(fix3,tx);
1288             fiy3             = _mm_add_ps(fiy3,ty);
1289             fiz3             = _mm_add_ps(fiz3,tz);
1290
1291             fjx3             = _mm_add_ps(fjx3,tx);
1292             fjy3             = _mm_add_ps(fjy3,ty);
1293             fjz3             = _mm_add_ps(fjz3,tz);
1294
1295             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1296             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1297             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1298             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1299
1300             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1301                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1302                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1303
1304             /* Inner loop uses 431 flops */
1305         }
1306
1307         /* End of innermost loop */
1308
1309         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1310                                               f+i_coord_offset,fshift+i_shift_offset);
1311
1312         ggid                        = gid[iidx];
1313         /* Update potential energies */
1314         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1315         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1316
1317         /* Increment number of inner iterations */
1318         inneriter                  += j_index_end - j_index_start;
1319
1320         /* Outer loop uses 26 flops */
1321     }
1322
1323     /* Increment number of outer iterations */
1324     outeriter        += nri;
1325
1326     /* Update outer/inner flops */
1327
1328     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*431);
1329 }
1330 /*
1331  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse2_single
1332  * Electrostatics interaction: CubicSplineTable
1333  * VdW interaction:            LennardJones
1334  * Geometry:                   Water4-Water4
1335  * Calculate force/pot:        Force
1336  */
1337 void
1338 nb_kernel_ElecCSTab_VdwLJ_GeomW4W4_F_sse2_single
1339                     (t_nblist * gmx_restrict                nlist,
1340                      rvec * gmx_restrict                    xx,
1341                      rvec * gmx_restrict                    ff,
1342                      t_forcerec * gmx_restrict              fr,
1343                      t_mdatoms * gmx_restrict               mdatoms,
1344                      nb_kernel_data_t * gmx_restrict        kernel_data,
1345                      t_nrnb * gmx_restrict                  nrnb)
1346 {
1347     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1348      * just 0 for non-waters.
1349      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1350      * jnr indices corresponding to data put in the four positions in the SIMD register.
1351      */
1352     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1353     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1354     int              jnrA,jnrB,jnrC,jnrD;
1355     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1356     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1357     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1358     real             rcutoff_scalar;
1359     real             *shiftvec,*fshift,*x,*f;
1360     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1361     real             scratch[4*DIM];
1362     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1363     int              vdwioffset0;
1364     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1365     int              vdwioffset1;
1366     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1367     int              vdwioffset2;
1368     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1369     int              vdwioffset3;
1370     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1371     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1372     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1373     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1374     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1375     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1376     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1377     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1378     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1379     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1380     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1381     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1382     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1383     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1384     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1385     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1386     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1387     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1388     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1389     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1390     real             *charge;
1391     int              nvdwtype;
1392     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1393     int              *vdwtype;
1394     real             *vdwparam;
1395     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1396     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1397     __m128i          vfitab;
1398     __m128i          ifour       = _mm_set1_epi32(4);
1399     __m128           rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF;
1400     real             *vftab;
1401     __m128           dummy_mask,cutoff_mask;
1402     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1403     __m128           one     = _mm_set1_ps(1.0);
1404     __m128           two     = _mm_set1_ps(2.0);
1405     x                = xx[0];
1406     f                = ff[0];
1407
1408     nri              = nlist->nri;
1409     iinr             = nlist->iinr;
1410     jindex           = nlist->jindex;
1411     jjnr             = nlist->jjnr;
1412     shiftidx         = nlist->shift;
1413     gid              = nlist->gid;
1414     shiftvec         = fr->shift_vec[0];
1415     fshift           = fr->fshift[0];
1416     facel            = _mm_set1_ps(fr->epsfac);
1417     charge           = mdatoms->chargeA;
1418     nvdwtype         = fr->ntype;
1419     vdwparam         = fr->nbfp;
1420     vdwtype          = mdatoms->typeA;
1421
1422     vftab            = kernel_data->table_elec->data;
1423     vftabscale       = _mm_set1_ps(kernel_data->table_elec->scale);
1424
1425     /* Setup water-specific parameters */
1426     inr              = nlist->iinr[0];
1427     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1428     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1429     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1430     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1431
1432     jq1              = _mm_set1_ps(charge[inr+1]);
1433     jq2              = _mm_set1_ps(charge[inr+2]);
1434     jq3              = _mm_set1_ps(charge[inr+3]);
1435     vdwjidx0A        = 2*vdwtype[inr+0];
1436     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1437     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1438     qq11             = _mm_mul_ps(iq1,jq1);
1439     qq12             = _mm_mul_ps(iq1,jq2);
1440     qq13             = _mm_mul_ps(iq1,jq3);
1441     qq21             = _mm_mul_ps(iq2,jq1);
1442     qq22             = _mm_mul_ps(iq2,jq2);
1443     qq23             = _mm_mul_ps(iq2,jq3);
1444     qq31             = _mm_mul_ps(iq3,jq1);
1445     qq32             = _mm_mul_ps(iq3,jq2);
1446     qq33             = _mm_mul_ps(iq3,jq3);
1447
1448     /* Avoid stupid compiler warnings */
1449     jnrA = jnrB = jnrC = jnrD = 0;
1450     j_coord_offsetA = 0;
1451     j_coord_offsetB = 0;
1452     j_coord_offsetC = 0;
1453     j_coord_offsetD = 0;
1454
1455     outeriter        = 0;
1456     inneriter        = 0;
1457
1458     for(iidx=0;iidx<4*DIM;iidx++)
1459     {
1460         scratch[iidx] = 0.0;
1461     }
1462
1463     /* Start outer loop over neighborlists */
1464     for(iidx=0; iidx<nri; iidx++)
1465     {
1466         /* Load shift vector for this list */
1467         i_shift_offset   = DIM*shiftidx[iidx];
1468
1469         /* Load limits for loop over neighbors */
1470         j_index_start    = jindex[iidx];
1471         j_index_end      = jindex[iidx+1];
1472
1473         /* Get outer coordinate index */
1474         inr              = iinr[iidx];
1475         i_coord_offset   = DIM*inr;
1476
1477         /* Load i particle coords and add shift vector */
1478         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1479                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1480
1481         fix0             = _mm_setzero_ps();
1482         fiy0             = _mm_setzero_ps();
1483         fiz0             = _mm_setzero_ps();
1484         fix1             = _mm_setzero_ps();
1485         fiy1             = _mm_setzero_ps();
1486         fiz1             = _mm_setzero_ps();
1487         fix2             = _mm_setzero_ps();
1488         fiy2             = _mm_setzero_ps();
1489         fiz2             = _mm_setzero_ps();
1490         fix3             = _mm_setzero_ps();
1491         fiy3             = _mm_setzero_ps();
1492         fiz3             = _mm_setzero_ps();
1493
1494         /* Start inner kernel loop */
1495         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1496         {
1497
1498             /* Get j neighbor index, and coordinate index */
1499             jnrA             = jjnr[jidx];
1500             jnrB             = jjnr[jidx+1];
1501             jnrC             = jjnr[jidx+2];
1502             jnrD             = jjnr[jidx+3];
1503             j_coord_offsetA  = DIM*jnrA;
1504             j_coord_offsetB  = DIM*jnrB;
1505             j_coord_offsetC  = DIM*jnrC;
1506             j_coord_offsetD  = DIM*jnrD;
1507
1508             /* load j atom coordinates */
1509             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1510                                               x+j_coord_offsetC,x+j_coord_offsetD,
1511                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1512                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1513
1514             /* Calculate displacement vector */
1515             dx00             = _mm_sub_ps(ix0,jx0);
1516             dy00             = _mm_sub_ps(iy0,jy0);
1517             dz00             = _mm_sub_ps(iz0,jz0);
1518             dx11             = _mm_sub_ps(ix1,jx1);
1519             dy11             = _mm_sub_ps(iy1,jy1);
1520             dz11             = _mm_sub_ps(iz1,jz1);
1521             dx12             = _mm_sub_ps(ix1,jx2);
1522             dy12             = _mm_sub_ps(iy1,jy2);
1523             dz12             = _mm_sub_ps(iz1,jz2);
1524             dx13             = _mm_sub_ps(ix1,jx3);
1525             dy13             = _mm_sub_ps(iy1,jy3);
1526             dz13             = _mm_sub_ps(iz1,jz3);
1527             dx21             = _mm_sub_ps(ix2,jx1);
1528             dy21             = _mm_sub_ps(iy2,jy1);
1529             dz21             = _mm_sub_ps(iz2,jz1);
1530             dx22             = _mm_sub_ps(ix2,jx2);
1531             dy22             = _mm_sub_ps(iy2,jy2);
1532             dz22             = _mm_sub_ps(iz2,jz2);
1533             dx23             = _mm_sub_ps(ix2,jx3);
1534             dy23             = _mm_sub_ps(iy2,jy3);
1535             dz23             = _mm_sub_ps(iz2,jz3);
1536             dx31             = _mm_sub_ps(ix3,jx1);
1537             dy31             = _mm_sub_ps(iy3,jy1);
1538             dz31             = _mm_sub_ps(iz3,jz1);
1539             dx32             = _mm_sub_ps(ix3,jx2);
1540             dy32             = _mm_sub_ps(iy3,jy2);
1541             dz32             = _mm_sub_ps(iz3,jz2);
1542             dx33             = _mm_sub_ps(ix3,jx3);
1543             dy33             = _mm_sub_ps(iy3,jy3);
1544             dz33             = _mm_sub_ps(iz3,jz3);
1545
1546             /* Calculate squared distance and things based on it */
1547             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1548             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1549             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1550             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1551             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1552             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1553             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1554             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1555             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1556             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1557
1558             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1559             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1560             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1561             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1562             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1563             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1564             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1565             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1566             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1567
1568             rinvsq00         = gmx_mm_inv_ps(rsq00);
1569
1570             fjx0             = _mm_setzero_ps();
1571             fjy0             = _mm_setzero_ps();
1572             fjz0             = _mm_setzero_ps();
1573             fjx1             = _mm_setzero_ps();
1574             fjy1             = _mm_setzero_ps();
1575             fjz1             = _mm_setzero_ps();
1576             fjx2             = _mm_setzero_ps();
1577             fjy2             = _mm_setzero_ps();
1578             fjz2             = _mm_setzero_ps();
1579             fjx3             = _mm_setzero_ps();
1580             fjy3             = _mm_setzero_ps();
1581             fjz3             = _mm_setzero_ps();
1582
1583             /**************************
1584              * CALCULATE INTERACTIONS *
1585              **************************/
1586
1587             /* LENNARD-JONES DISPERSION/REPULSION */
1588
1589             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1590             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
1591
1592             fscal            = fvdw;
1593
1594             /* Calculate temporary vectorial force */
1595             tx               = _mm_mul_ps(fscal,dx00);
1596             ty               = _mm_mul_ps(fscal,dy00);
1597             tz               = _mm_mul_ps(fscal,dz00);
1598
1599             /* Update vectorial force */
1600             fix0             = _mm_add_ps(fix0,tx);
1601             fiy0             = _mm_add_ps(fiy0,ty);
1602             fiz0             = _mm_add_ps(fiz0,tz);
1603
1604             fjx0             = _mm_add_ps(fjx0,tx);
1605             fjy0             = _mm_add_ps(fjy0,ty);
1606             fjz0             = _mm_add_ps(fjz0,tz);
1607
1608             /**************************
1609              * CALCULATE INTERACTIONS *
1610              **************************/
1611
1612             r11              = _mm_mul_ps(rsq11,rinv11);
1613
1614             /* Calculate table index by multiplying r with table scale and truncate to integer */
1615             rt               = _mm_mul_ps(r11,vftabscale);
1616             vfitab           = _mm_cvttps_epi32(rt);
1617             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1618             vfitab           = _mm_slli_epi32(vfitab,2);
1619
1620             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1621             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1622             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1623             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1624             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1625             _MM_TRANSPOSE4_PS(Y,F,G,H);
1626             Heps             = _mm_mul_ps(vfeps,H);
1627             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1628             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1629             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
1630
1631             fscal            = felec;
1632
1633             /* Calculate temporary vectorial force */
1634             tx               = _mm_mul_ps(fscal,dx11);
1635             ty               = _mm_mul_ps(fscal,dy11);
1636             tz               = _mm_mul_ps(fscal,dz11);
1637
1638             /* Update vectorial force */
1639             fix1             = _mm_add_ps(fix1,tx);
1640             fiy1             = _mm_add_ps(fiy1,ty);
1641             fiz1             = _mm_add_ps(fiz1,tz);
1642
1643             fjx1             = _mm_add_ps(fjx1,tx);
1644             fjy1             = _mm_add_ps(fjy1,ty);
1645             fjz1             = _mm_add_ps(fjz1,tz);
1646
1647             /**************************
1648              * CALCULATE INTERACTIONS *
1649              **************************/
1650
1651             r12              = _mm_mul_ps(rsq12,rinv12);
1652
1653             /* Calculate table index by multiplying r with table scale and truncate to integer */
1654             rt               = _mm_mul_ps(r12,vftabscale);
1655             vfitab           = _mm_cvttps_epi32(rt);
1656             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1657             vfitab           = _mm_slli_epi32(vfitab,2);
1658
1659             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1660             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1661             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1662             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1663             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1664             _MM_TRANSPOSE4_PS(Y,F,G,H);
1665             Heps             = _mm_mul_ps(vfeps,H);
1666             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1667             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1668             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
1669
1670             fscal            = felec;
1671
1672             /* Calculate temporary vectorial force */
1673             tx               = _mm_mul_ps(fscal,dx12);
1674             ty               = _mm_mul_ps(fscal,dy12);
1675             tz               = _mm_mul_ps(fscal,dz12);
1676
1677             /* Update vectorial force */
1678             fix1             = _mm_add_ps(fix1,tx);
1679             fiy1             = _mm_add_ps(fiy1,ty);
1680             fiz1             = _mm_add_ps(fiz1,tz);
1681
1682             fjx2             = _mm_add_ps(fjx2,tx);
1683             fjy2             = _mm_add_ps(fjy2,ty);
1684             fjz2             = _mm_add_ps(fjz2,tz);
1685
1686             /**************************
1687              * CALCULATE INTERACTIONS *
1688              **************************/
1689
1690             r13              = _mm_mul_ps(rsq13,rinv13);
1691
1692             /* Calculate table index by multiplying r with table scale and truncate to integer */
1693             rt               = _mm_mul_ps(r13,vftabscale);
1694             vfitab           = _mm_cvttps_epi32(rt);
1695             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1696             vfitab           = _mm_slli_epi32(vfitab,2);
1697
1698             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1699             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1700             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1701             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1702             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1703             _MM_TRANSPOSE4_PS(Y,F,G,H);
1704             Heps             = _mm_mul_ps(vfeps,H);
1705             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1706             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1707             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
1708
1709             fscal            = felec;
1710
1711             /* Calculate temporary vectorial force */
1712             tx               = _mm_mul_ps(fscal,dx13);
1713             ty               = _mm_mul_ps(fscal,dy13);
1714             tz               = _mm_mul_ps(fscal,dz13);
1715
1716             /* Update vectorial force */
1717             fix1             = _mm_add_ps(fix1,tx);
1718             fiy1             = _mm_add_ps(fiy1,ty);
1719             fiz1             = _mm_add_ps(fiz1,tz);
1720
1721             fjx3             = _mm_add_ps(fjx3,tx);
1722             fjy3             = _mm_add_ps(fjy3,ty);
1723             fjz3             = _mm_add_ps(fjz3,tz);
1724
1725             /**************************
1726              * CALCULATE INTERACTIONS *
1727              **************************/
1728
1729             r21              = _mm_mul_ps(rsq21,rinv21);
1730
1731             /* Calculate table index by multiplying r with table scale and truncate to integer */
1732             rt               = _mm_mul_ps(r21,vftabscale);
1733             vfitab           = _mm_cvttps_epi32(rt);
1734             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1735             vfitab           = _mm_slli_epi32(vfitab,2);
1736
1737             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1738             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1739             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1740             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1741             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1742             _MM_TRANSPOSE4_PS(Y,F,G,H);
1743             Heps             = _mm_mul_ps(vfeps,H);
1744             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1745             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1746             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
1747
1748             fscal            = felec;
1749
1750             /* Calculate temporary vectorial force */
1751             tx               = _mm_mul_ps(fscal,dx21);
1752             ty               = _mm_mul_ps(fscal,dy21);
1753             tz               = _mm_mul_ps(fscal,dz21);
1754
1755             /* Update vectorial force */
1756             fix2             = _mm_add_ps(fix2,tx);
1757             fiy2             = _mm_add_ps(fiy2,ty);
1758             fiz2             = _mm_add_ps(fiz2,tz);
1759
1760             fjx1             = _mm_add_ps(fjx1,tx);
1761             fjy1             = _mm_add_ps(fjy1,ty);
1762             fjz1             = _mm_add_ps(fjz1,tz);
1763
1764             /**************************
1765              * CALCULATE INTERACTIONS *
1766              **************************/
1767
1768             r22              = _mm_mul_ps(rsq22,rinv22);
1769
1770             /* Calculate table index by multiplying r with table scale and truncate to integer */
1771             rt               = _mm_mul_ps(r22,vftabscale);
1772             vfitab           = _mm_cvttps_epi32(rt);
1773             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1774             vfitab           = _mm_slli_epi32(vfitab,2);
1775
1776             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1777             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1778             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1779             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1780             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1781             _MM_TRANSPOSE4_PS(Y,F,G,H);
1782             Heps             = _mm_mul_ps(vfeps,H);
1783             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1784             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1785             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
1786
1787             fscal            = felec;
1788
1789             /* Calculate temporary vectorial force */
1790             tx               = _mm_mul_ps(fscal,dx22);
1791             ty               = _mm_mul_ps(fscal,dy22);
1792             tz               = _mm_mul_ps(fscal,dz22);
1793
1794             /* Update vectorial force */
1795             fix2             = _mm_add_ps(fix2,tx);
1796             fiy2             = _mm_add_ps(fiy2,ty);
1797             fiz2             = _mm_add_ps(fiz2,tz);
1798
1799             fjx2             = _mm_add_ps(fjx2,tx);
1800             fjy2             = _mm_add_ps(fjy2,ty);
1801             fjz2             = _mm_add_ps(fjz2,tz);
1802
1803             /**************************
1804              * CALCULATE INTERACTIONS *
1805              **************************/
1806
1807             r23              = _mm_mul_ps(rsq23,rinv23);
1808
1809             /* Calculate table index by multiplying r with table scale and truncate to integer */
1810             rt               = _mm_mul_ps(r23,vftabscale);
1811             vfitab           = _mm_cvttps_epi32(rt);
1812             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1813             vfitab           = _mm_slli_epi32(vfitab,2);
1814
1815             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1816             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1817             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1818             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1819             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1820             _MM_TRANSPOSE4_PS(Y,F,G,H);
1821             Heps             = _mm_mul_ps(vfeps,H);
1822             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1823             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1824             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
1825
1826             fscal            = felec;
1827
1828             /* Calculate temporary vectorial force */
1829             tx               = _mm_mul_ps(fscal,dx23);
1830             ty               = _mm_mul_ps(fscal,dy23);
1831             tz               = _mm_mul_ps(fscal,dz23);
1832
1833             /* Update vectorial force */
1834             fix2             = _mm_add_ps(fix2,tx);
1835             fiy2             = _mm_add_ps(fiy2,ty);
1836             fiz2             = _mm_add_ps(fiz2,tz);
1837
1838             fjx3             = _mm_add_ps(fjx3,tx);
1839             fjy3             = _mm_add_ps(fjy3,ty);
1840             fjz3             = _mm_add_ps(fjz3,tz);
1841
1842             /**************************
1843              * CALCULATE INTERACTIONS *
1844              **************************/
1845
1846             r31              = _mm_mul_ps(rsq31,rinv31);
1847
1848             /* Calculate table index by multiplying r with table scale and truncate to integer */
1849             rt               = _mm_mul_ps(r31,vftabscale);
1850             vfitab           = _mm_cvttps_epi32(rt);
1851             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1852             vfitab           = _mm_slli_epi32(vfitab,2);
1853
1854             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1855             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1856             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1857             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1858             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1859             _MM_TRANSPOSE4_PS(Y,F,G,H);
1860             Heps             = _mm_mul_ps(vfeps,H);
1861             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1862             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1863             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
1864
1865             fscal            = felec;
1866
1867             /* Calculate temporary vectorial force */
1868             tx               = _mm_mul_ps(fscal,dx31);
1869             ty               = _mm_mul_ps(fscal,dy31);
1870             tz               = _mm_mul_ps(fscal,dz31);
1871
1872             /* Update vectorial force */
1873             fix3             = _mm_add_ps(fix3,tx);
1874             fiy3             = _mm_add_ps(fiy3,ty);
1875             fiz3             = _mm_add_ps(fiz3,tz);
1876
1877             fjx1             = _mm_add_ps(fjx1,tx);
1878             fjy1             = _mm_add_ps(fjy1,ty);
1879             fjz1             = _mm_add_ps(fjz1,tz);
1880
1881             /**************************
1882              * CALCULATE INTERACTIONS *
1883              **************************/
1884
1885             r32              = _mm_mul_ps(rsq32,rinv32);
1886
1887             /* Calculate table index by multiplying r with table scale and truncate to integer */
1888             rt               = _mm_mul_ps(r32,vftabscale);
1889             vfitab           = _mm_cvttps_epi32(rt);
1890             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1891             vfitab           = _mm_slli_epi32(vfitab,2);
1892
1893             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1894             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1895             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1896             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1897             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1898             _MM_TRANSPOSE4_PS(Y,F,G,H);
1899             Heps             = _mm_mul_ps(vfeps,H);
1900             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1901             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1902             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
1903
1904             fscal            = felec;
1905
1906             /* Calculate temporary vectorial force */
1907             tx               = _mm_mul_ps(fscal,dx32);
1908             ty               = _mm_mul_ps(fscal,dy32);
1909             tz               = _mm_mul_ps(fscal,dz32);
1910
1911             /* Update vectorial force */
1912             fix3             = _mm_add_ps(fix3,tx);
1913             fiy3             = _mm_add_ps(fiy3,ty);
1914             fiz3             = _mm_add_ps(fiz3,tz);
1915
1916             fjx2             = _mm_add_ps(fjx2,tx);
1917             fjy2             = _mm_add_ps(fjy2,ty);
1918             fjz2             = _mm_add_ps(fjz2,tz);
1919
1920             /**************************
1921              * CALCULATE INTERACTIONS *
1922              **************************/
1923
1924             r33              = _mm_mul_ps(rsq33,rinv33);
1925
1926             /* Calculate table index by multiplying r with table scale and truncate to integer */
1927             rt               = _mm_mul_ps(r33,vftabscale);
1928             vfitab           = _mm_cvttps_epi32(rt);
1929             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
1930             vfitab           = _mm_slli_epi32(vfitab,2);
1931
1932             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1933             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
1934             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
1935             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
1936             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
1937             _MM_TRANSPOSE4_PS(Y,F,G,H);
1938             Heps             = _mm_mul_ps(vfeps,H);
1939             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
1940             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
1941             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
1942
1943             fscal            = felec;
1944
1945             /* Calculate temporary vectorial force */
1946             tx               = _mm_mul_ps(fscal,dx33);
1947             ty               = _mm_mul_ps(fscal,dy33);
1948             tz               = _mm_mul_ps(fscal,dz33);
1949
1950             /* Update vectorial force */
1951             fix3             = _mm_add_ps(fix3,tx);
1952             fiy3             = _mm_add_ps(fiy3,ty);
1953             fiz3             = _mm_add_ps(fiz3,tz);
1954
1955             fjx3             = _mm_add_ps(fjx3,tx);
1956             fjy3             = _mm_add_ps(fjy3,ty);
1957             fjz3             = _mm_add_ps(fjz3,tz);
1958
1959             fjptrA             = f+j_coord_offsetA;
1960             fjptrB             = f+j_coord_offsetB;
1961             fjptrC             = f+j_coord_offsetC;
1962             fjptrD             = f+j_coord_offsetD;
1963
1964             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1965                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1966                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1967
1968             /* Inner loop uses 381 flops */
1969         }
1970
1971         if(jidx<j_index_end)
1972         {
1973
1974             /* Get j neighbor index, and coordinate index */
1975             jnrlistA         = jjnr[jidx];
1976             jnrlistB         = jjnr[jidx+1];
1977             jnrlistC         = jjnr[jidx+2];
1978             jnrlistD         = jjnr[jidx+3];
1979             /* Sign of each element will be negative for non-real atoms.
1980              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1981              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1982              */
1983             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1984             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1985             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1986             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1987             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1988             j_coord_offsetA  = DIM*jnrA;
1989             j_coord_offsetB  = DIM*jnrB;
1990             j_coord_offsetC  = DIM*jnrC;
1991             j_coord_offsetD  = DIM*jnrD;
1992
1993             /* load j atom coordinates */
1994             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1995                                               x+j_coord_offsetC,x+j_coord_offsetD,
1996                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1997                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1998
1999             /* Calculate displacement vector */
2000             dx00             = _mm_sub_ps(ix0,jx0);
2001             dy00             = _mm_sub_ps(iy0,jy0);
2002             dz00             = _mm_sub_ps(iz0,jz0);
2003             dx11             = _mm_sub_ps(ix1,jx1);
2004             dy11             = _mm_sub_ps(iy1,jy1);
2005             dz11             = _mm_sub_ps(iz1,jz1);
2006             dx12             = _mm_sub_ps(ix1,jx2);
2007             dy12             = _mm_sub_ps(iy1,jy2);
2008             dz12             = _mm_sub_ps(iz1,jz2);
2009             dx13             = _mm_sub_ps(ix1,jx3);
2010             dy13             = _mm_sub_ps(iy1,jy3);
2011             dz13             = _mm_sub_ps(iz1,jz3);
2012             dx21             = _mm_sub_ps(ix2,jx1);
2013             dy21             = _mm_sub_ps(iy2,jy1);
2014             dz21             = _mm_sub_ps(iz2,jz1);
2015             dx22             = _mm_sub_ps(ix2,jx2);
2016             dy22             = _mm_sub_ps(iy2,jy2);
2017             dz22             = _mm_sub_ps(iz2,jz2);
2018             dx23             = _mm_sub_ps(ix2,jx3);
2019             dy23             = _mm_sub_ps(iy2,jy3);
2020             dz23             = _mm_sub_ps(iz2,jz3);
2021             dx31             = _mm_sub_ps(ix3,jx1);
2022             dy31             = _mm_sub_ps(iy3,jy1);
2023             dz31             = _mm_sub_ps(iz3,jz1);
2024             dx32             = _mm_sub_ps(ix3,jx2);
2025             dy32             = _mm_sub_ps(iy3,jy2);
2026             dz32             = _mm_sub_ps(iz3,jz2);
2027             dx33             = _mm_sub_ps(ix3,jx3);
2028             dy33             = _mm_sub_ps(iy3,jy3);
2029             dz33             = _mm_sub_ps(iz3,jz3);
2030
2031             /* Calculate squared distance and things based on it */
2032             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
2033             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
2034             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
2035             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
2036             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
2037             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
2038             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2039             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2040             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2041             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2042
2043             rinv11           = gmx_mm_invsqrt_ps(rsq11);
2044             rinv12           = gmx_mm_invsqrt_ps(rsq12);
2045             rinv13           = gmx_mm_invsqrt_ps(rsq13);
2046             rinv21           = gmx_mm_invsqrt_ps(rsq21);
2047             rinv22           = gmx_mm_invsqrt_ps(rsq22);
2048             rinv23           = gmx_mm_invsqrt_ps(rsq23);
2049             rinv31           = gmx_mm_invsqrt_ps(rsq31);
2050             rinv32           = gmx_mm_invsqrt_ps(rsq32);
2051             rinv33           = gmx_mm_invsqrt_ps(rsq33);
2052
2053             rinvsq00         = gmx_mm_inv_ps(rsq00);
2054
2055             fjx0             = _mm_setzero_ps();
2056             fjy0             = _mm_setzero_ps();
2057             fjz0             = _mm_setzero_ps();
2058             fjx1             = _mm_setzero_ps();
2059             fjy1             = _mm_setzero_ps();
2060             fjz1             = _mm_setzero_ps();
2061             fjx2             = _mm_setzero_ps();
2062             fjy2             = _mm_setzero_ps();
2063             fjz2             = _mm_setzero_ps();
2064             fjx3             = _mm_setzero_ps();
2065             fjy3             = _mm_setzero_ps();
2066             fjz3             = _mm_setzero_ps();
2067
2068             /**************************
2069              * CALCULATE INTERACTIONS *
2070              **************************/
2071
2072             /* LENNARD-JONES DISPERSION/REPULSION */
2073
2074             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2075             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
2076
2077             fscal            = fvdw;
2078
2079             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2080
2081             /* Calculate temporary vectorial force */
2082             tx               = _mm_mul_ps(fscal,dx00);
2083             ty               = _mm_mul_ps(fscal,dy00);
2084             tz               = _mm_mul_ps(fscal,dz00);
2085
2086             /* Update vectorial force */
2087             fix0             = _mm_add_ps(fix0,tx);
2088             fiy0             = _mm_add_ps(fiy0,ty);
2089             fiz0             = _mm_add_ps(fiz0,tz);
2090
2091             fjx0             = _mm_add_ps(fjx0,tx);
2092             fjy0             = _mm_add_ps(fjy0,ty);
2093             fjz0             = _mm_add_ps(fjz0,tz);
2094
2095             /**************************
2096              * CALCULATE INTERACTIONS *
2097              **************************/
2098
2099             r11              = _mm_mul_ps(rsq11,rinv11);
2100             r11              = _mm_andnot_ps(dummy_mask,r11);
2101
2102             /* Calculate table index by multiplying r with table scale and truncate to integer */
2103             rt               = _mm_mul_ps(r11,vftabscale);
2104             vfitab           = _mm_cvttps_epi32(rt);
2105             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2106             vfitab           = _mm_slli_epi32(vfitab,2);
2107
2108             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2109             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2110             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2111             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2112             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2113             _MM_TRANSPOSE4_PS(Y,F,G,H);
2114             Heps             = _mm_mul_ps(vfeps,H);
2115             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2116             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2117             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq11,FF),_mm_mul_ps(vftabscale,rinv11)));
2118
2119             fscal            = felec;
2120
2121             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2122
2123             /* Calculate temporary vectorial force */
2124             tx               = _mm_mul_ps(fscal,dx11);
2125             ty               = _mm_mul_ps(fscal,dy11);
2126             tz               = _mm_mul_ps(fscal,dz11);
2127
2128             /* Update vectorial force */
2129             fix1             = _mm_add_ps(fix1,tx);
2130             fiy1             = _mm_add_ps(fiy1,ty);
2131             fiz1             = _mm_add_ps(fiz1,tz);
2132
2133             fjx1             = _mm_add_ps(fjx1,tx);
2134             fjy1             = _mm_add_ps(fjy1,ty);
2135             fjz1             = _mm_add_ps(fjz1,tz);
2136
2137             /**************************
2138              * CALCULATE INTERACTIONS *
2139              **************************/
2140
2141             r12              = _mm_mul_ps(rsq12,rinv12);
2142             r12              = _mm_andnot_ps(dummy_mask,r12);
2143
2144             /* Calculate table index by multiplying r with table scale and truncate to integer */
2145             rt               = _mm_mul_ps(r12,vftabscale);
2146             vfitab           = _mm_cvttps_epi32(rt);
2147             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2148             vfitab           = _mm_slli_epi32(vfitab,2);
2149
2150             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2151             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2152             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2153             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2154             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2155             _MM_TRANSPOSE4_PS(Y,F,G,H);
2156             Heps             = _mm_mul_ps(vfeps,H);
2157             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2158             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2159             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq12,FF),_mm_mul_ps(vftabscale,rinv12)));
2160
2161             fscal            = felec;
2162
2163             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2164
2165             /* Calculate temporary vectorial force */
2166             tx               = _mm_mul_ps(fscal,dx12);
2167             ty               = _mm_mul_ps(fscal,dy12);
2168             tz               = _mm_mul_ps(fscal,dz12);
2169
2170             /* Update vectorial force */
2171             fix1             = _mm_add_ps(fix1,tx);
2172             fiy1             = _mm_add_ps(fiy1,ty);
2173             fiz1             = _mm_add_ps(fiz1,tz);
2174
2175             fjx2             = _mm_add_ps(fjx2,tx);
2176             fjy2             = _mm_add_ps(fjy2,ty);
2177             fjz2             = _mm_add_ps(fjz2,tz);
2178
2179             /**************************
2180              * CALCULATE INTERACTIONS *
2181              **************************/
2182
2183             r13              = _mm_mul_ps(rsq13,rinv13);
2184             r13              = _mm_andnot_ps(dummy_mask,r13);
2185
2186             /* Calculate table index by multiplying r with table scale and truncate to integer */
2187             rt               = _mm_mul_ps(r13,vftabscale);
2188             vfitab           = _mm_cvttps_epi32(rt);
2189             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2190             vfitab           = _mm_slli_epi32(vfitab,2);
2191
2192             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2193             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2194             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2195             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2196             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2197             _MM_TRANSPOSE4_PS(Y,F,G,H);
2198             Heps             = _mm_mul_ps(vfeps,H);
2199             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2200             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2201             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq13,FF),_mm_mul_ps(vftabscale,rinv13)));
2202
2203             fscal            = felec;
2204
2205             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2206
2207             /* Calculate temporary vectorial force */
2208             tx               = _mm_mul_ps(fscal,dx13);
2209             ty               = _mm_mul_ps(fscal,dy13);
2210             tz               = _mm_mul_ps(fscal,dz13);
2211
2212             /* Update vectorial force */
2213             fix1             = _mm_add_ps(fix1,tx);
2214             fiy1             = _mm_add_ps(fiy1,ty);
2215             fiz1             = _mm_add_ps(fiz1,tz);
2216
2217             fjx3             = _mm_add_ps(fjx3,tx);
2218             fjy3             = _mm_add_ps(fjy3,ty);
2219             fjz3             = _mm_add_ps(fjz3,tz);
2220
2221             /**************************
2222              * CALCULATE INTERACTIONS *
2223              **************************/
2224
2225             r21              = _mm_mul_ps(rsq21,rinv21);
2226             r21              = _mm_andnot_ps(dummy_mask,r21);
2227
2228             /* Calculate table index by multiplying r with table scale and truncate to integer */
2229             rt               = _mm_mul_ps(r21,vftabscale);
2230             vfitab           = _mm_cvttps_epi32(rt);
2231             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2232             vfitab           = _mm_slli_epi32(vfitab,2);
2233
2234             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2235             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2236             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2237             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2238             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2239             _MM_TRANSPOSE4_PS(Y,F,G,H);
2240             Heps             = _mm_mul_ps(vfeps,H);
2241             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2242             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2243             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq21,FF),_mm_mul_ps(vftabscale,rinv21)));
2244
2245             fscal            = felec;
2246
2247             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2248
2249             /* Calculate temporary vectorial force */
2250             tx               = _mm_mul_ps(fscal,dx21);
2251             ty               = _mm_mul_ps(fscal,dy21);
2252             tz               = _mm_mul_ps(fscal,dz21);
2253
2254             /* Update vectorial force */
2255             fix2             = _mm_add_ps(fix2,tx);
2256             fiy2             = _mm_add_ps(fiy2,ty);
2257             fiz2             = _mm_add_ps(fiz2,tz);
2258
2259             fjx1             = _mm_add_ps(fjx1,tx);
2260             fjy1             = _mm_add_ps(fjy1,ty);
2261             fjz1             = _mm_add_ps(fjz1,tz);
2262
2263             /**************************
2264              * CALCULATE INTERACTIONS *
2265              **************************/
2266
2267             r22              = _mm_mul_ps(rsq22,rinv22);
2268             r22              = _mm_andnot_ps(dummy_mask,r22);
2269
2270             /* Calculate table index by multiplying r with table scale and truncate to integer */
2271             rt               = _mm_mul_ps(r22,vftabscale);
2272             vfitab           = _mm_cvttps_epi32(rt);
2273             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2274             vfitab           = _mm_slli_epi32(vfitab,2);
2275
2276             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2277             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2278             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2279             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2280             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2281             _MM_TRANSPOSE4_PS(Y,F,G,H);
2282             Heps             = _mm_mul_ps(vfeps,H);
2283             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2284             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2285             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq22,FF),_mm_mul_ps(vftabscale,rinv22)));
2286
2287             fscal            = felec;
2288
2289             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2290
2291             /* Calculate temporary vectorial force */
2292             tx               = _mm_mul_ps(fscal,dx22);
2293             ty               = _mm_mul_ps(fscal,dy22);
2294             tz               = _mm_mul_ps(fscal,dz22);
2295
2296             /* Update vectorial force */
2297             fix2             = _mm_add_ps(fix2,tx);
2298             fiy2             = _mm_add_ps(fiy2,ty);
2299             fiz2             = _mm_add_ps(fiz2,tz);
2300
2301             fjx2             = _mm_add_ps(fjx2,tx);
2302             fjy2             = _mm_add_ps(fjy2,ty);
2303             fjz2             = _mm_add_ps(fjz2,tz);
2304
2305             /**************************
2306              * CALCULATE INTERACTIONS *
2307              **************************/
2308
2309             r23              = _mm_mul_ps(rsq23,rinv23);
2310             r23              = _mm_andnot_ps(dummy_mask,r23);
2311
2312             /* Calculate table index by multiplying r with table scale and truncate to integer */
2313             rt               = _mm_mul_ps(r23,vftabscale);
2314             vfitab           = _mm_cvttps_epi32(rt);
2315             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2316             vfitab           = _mm_slli_epi32(vfitab,2);
2317
2318             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2319             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2320             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2321             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2322             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2323             _MM_TRANSPOSE4_PS(Y,F,G,H);
2324             Heps             = _mm_mul_ps(vfeps,H);
2325             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2326             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2327             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq23,FF),_mm_mul_ps(vftabscale,rinv23)));
2328
2329             fscal            = felec;
2330
2331             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2332
2333             /* Calculate temporary vectorial force */
2334             tx               = _mm_mul_ps(fscal,dx23);
2335             ty               = _mm_mul_ps(fscal,dy23);
2336             tz               = _mm_mul_ps(fscal,dz23);
2337
2338             /* Update vectorial force */
2339             fix2             = _mm_add_ps(fix2,tx);
2340             fiy2             = _mm_add_ps(fiy2,ty);
2341             fiz2             = _mm_add_ps(fiz2,tz);
2342
2343             fjx3             = _mm_add_ps(fjx3,tx);
2344             fjy3             = _mm_add_ps(fjy3,ty);
2345             fjz3             = _mm_add_ps(fjz3,tz);
2346
2347             /**************************
2348              * CALCULATE INTERACTIONS *
2349              **************************/
2350
2351             r31              = _mm_mul_ps(rsq31,rinv31);
2352             r31              = _mm_andnot_ps(dummy_mask,r31);
2353
2354             /* Calculate table index by multiplying r with table scale and truncate to integer */
2355             rt               = _mm_mul_ps(r31,vftabscale);
2356             vfitab           = _mm_cvttps_epi32(rt);
2357             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2358             vfitab           = _mm_slli_epi32(vfitab,2);
2359
2360             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2361             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2362             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2363             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2364             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2365             _MM_TRANSPOSE4_PS(Y,F,G,H);
2366             Heps             = _mm_mul_ps(vfeps,H);
2367             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2368             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2369             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq31,FF),_mm_mul_ps(vftabscale,rinv31)));
2370
2371             fscal            = felec;
2372
2373             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2374
2375             /* Calculate temporary vectorial force */
2376             tx               = _mm_mul_ps(fscal,dx31);
2377             ty               = _mm_mul_ps(fscal,dy31);
2378             tz               = _mm_mul_ps(fscal,dz31);
2379
2380             /* Update vectorial force */
2381             fix3             = _mm_add_ps(fix3,tx);
2382             fiy3             = _mm_add_ps(fiy3,ty);
2383             fiz3             = _mm_add_ps(fiz3,tz);
2384
2385             fjx1             = _mm_add_ps(fjx1,tx);
2386             fjy1             = _mm_add_ps(fjy1,ty);
2387             fjz1             = _mm_add_ps(fjz1,tz);
2388
2389             /**************************
2390              * CALCULATE INTERACTIONS *
2391              **************************/
2392
2393             r32              = _mm_mul_ps(rsq32,rinv32);
2394             r32              = _mm_andnot_ps(dummy_mask,r32);
2395
2396             /* Calculate table index by multiplying r with table scale and truncate to integer */
2397             rt               = _mm_mul_ps(r32,vftabscale);
2398             vfitab           = _mm_cvttps_epi32(rt);
2399             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2400             vfitab           = _mm_slli_epi32(vfitab,2);
2401
2402             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2403             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2404             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2405             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2406             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2407             _MM_TRANSPOSE4_PS(Y,F,G,H);
2408             Heps             = _mm_mul_ps(vfeps,H);
2409             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2410             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2411             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq32,FF),_mm_mul_ps(vftabscale,rinv32)));
2412
2413             fscal            = felec;
2414
2415             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2416
2417             /* Calculate temporary vectorial force */
2418             tx               = _mm_mul_ps(fscal,dx32);
2419             ty               = _mm_mul_ps(fscal,dy32);
2420             tz               = _mm_mul_ps(fscal,dz32);
2421
2422             /* Update vectorial force */
2423             fix3             = _mm_add_ps(fix3,tx);
2424             fiy3             = _mm_add_ps(fiy3,ty);
2425             fiz3             = _mm_add_ps(fiz3,tz);
2426
2427             fjx2             = _mm_add_ps(fjx2,tx);
2428             fjy2             = _mm_add_ps(fjy2,ty);
2429             fjz2             = _mm_add_ps(fjz2,tz);
2430
2431             /**************************
2432              * CALCULATE INTERACTIONS *
2433              **************************/
2434
2435             r33              = _mm_mul_ps(rsq33,rinv33);
2436             r33              = _mm_andnot_ps(dummy_mask,r33);
2437
2438             /* Calculate table index by multiplying r with table scale and truncate to integer */
2439             rt               = _mm_mul_ps(r33,vftabscale);
2440             vfitab           = _mm_cvttps_epi32(rt);
2441             vfeps            = _mm_sub_ps(rt,_mm_cvtepi32_ps(vfitab));
2442             vfitab           = _mm_slli_epi32(vfitab,2);
2443
2444             /* CUBIC SPLINE TABLE ELECTROSTATICS */
2445             Y                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,0) );
2446             F                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,1) );
2447             G                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,2) );
2448             H                = _mm_load_ps( vftab + gmx_mm_extract_epi32(vfitab,3) );
2449             _MM_TRANSPOSE4_PS(Y,F,G,H);
2450             Heps             = _mm_mul_ps(vfeps,H);
2451             Fp               = _mm_add_ps(F,_mm_mul_ps(vfeps,_mm_add_ps(G,Heps)));
2452             FF               = _mm_add_ps(Fp,_mm_mul_ps(vfeps,_mm_add_ps(G,_mm_add_ps(Heps,Heps))));
2453             felec            = _mm_xor_ps(signbit,_mm_mul_ps(_mm_mul_ps(qq33,FF),_mm_mul_ps(vftabscale,rinv33)));
2454
2455             fscal            = felec;
2456
2457             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2458
2459             /* Calculate temporary vectorial force */
2460             tx               = _mm_mul_ps(fscal,dx33);
2461             ty               = _mm_mul_ps(fscal,dy33);
2462             tz               = _mm_mul_ps(fscal,dz33);
2463
2464             /* Update vectorial force */
2465             fix3             = _mm_add_ps(fix3,tx);
2466             fiy3             = _mm_add_ps(fiy3,ty);
2467             fiz3             = _mm_add_ps(fiz3,tz);
2468
2469             fjx3             = _mm_add_ps(fjx3,tx);
2470             fjy3             = _mm_add_ps(fjy3,ty);
2471             fjz3             = _mm_add_ps(fjz3,tz);
2472
2473             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2474             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2475             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2476             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2477
2478             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2479                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2480                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2481
2482             /* Inner loop uses 390 flops */
2483         }
2484
2485         /* End of innermost loop */
2486
2487         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2488                                               f+i_coord_offset,fshift+i_shift_offset);
2489
2490         /* Increment number of inner iterations */
2491         inneriter                  += j_index_end - j_index_start;
2492
2493         /* Outer loop uses 24 flops */
2494     }
2495
2496     /* Increment number of outer iterations */
2497     outeriter        += nri;
2498
2499     /* Update outer/inner flops */
2500
2501     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*390);
2502 }