src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_single/nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_avx_128_fma_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_128_fma_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
  48 #include "kernelutil_x86_avx_128_fma_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_128_fma_single
  52  * Electrostatics interaction: Coulomb
  53  * VdW interaction:            CubicSplineTable
  54  * Geometry:                   Water4-Water4
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_128_fma_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m128           fscal,rcutoff,rcutoff2,jidxall;
  83     int              vdwioffset0;
  84     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  85     int              vdwioffset1;
  86     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  87     int              vdwioffset2;
  88     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  89     int              vdwioffset3;
  90     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  92     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  94     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  95     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  96     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  97     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  98     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  99     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 100     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 101     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 102     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 103     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 104     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 105     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 106     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 107     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 108     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 109     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 110     real             *charge;
 111     int              nvdwtype;
 112     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 113     int              *vdwtype;
 114     real             *vdwparam;
 115     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 116     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 117     __m128i          vfitab;
 118     __m128i          ifour       = _mm_set1_epi32(4);
 119     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
 120     real             *vftab;
 121     __m128           dummy_mask,cutoff_mask;
 122     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 123     __m128           one     = _mm_set1_ps(1.0);
 124     __m128           two     = _mm_set1_ps(2.0);
 125     x                = xx[0];
 126     f                = ff[0];
 127
 128     nri              = nlist->nri;
 129     iinr             = nlist->iinr;
 130     jindex           = nlist->jindex;
 131     jjnr             = nlist->jjnr;
 132     shiftidx         = nlist->shift;
 133     gid              = nlist->gid;
 134     shiftvec         = fr->shift_vec[0];
 135     fshift           = fr->fshift[0];
 136     facel            = _mm_set1_ps(fr->epsfac);
 137     charge           = mdatoms->chargeA;
 138     nvdwtype         = fr->ntype;
 139     vdwparam         = fr->nbfp;
 140     vdwtype          = mdatoms->typeA;
 141
 142     vftab            = kernel_data->table_vdw->data;
 143     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 144
 145     /* Setup water-specific parameters */
 146     inr              = nlist->iinr[0];
 147     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 148     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 149     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 150     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 151
 152     jq1              = _mm_set1_ps(charge[inr+1]);
 153     jq2              = _mm_set1_ps(charge[inr+2]);
 154     jq3              = _mm_set1_ps(charge[inr+3]);
 155     vdwjidx0A        = 2*vdwtype[inr+0];
 156     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 157     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 158     qq11             = _mm_mul_ps(iq1,jq1);
 159     qq12             = _mm_mul_ps(iq1,jq2);
 160     qq13             = _mm_mul_ps(iq1,jq3);
 161     qq21             = _mm_mul_ps(iq2,jq1);
 162     qq22             = _mm_mul_ps(iq2,jq2);
 163     qq23             = _mm_mul_ps(iq2,jq3);
 164     qq31             = _mm_mul_ps(iq3,jq1);
 165     qq32             = _mm_mul_ps(iq3,jq2);
 166     qq33             = _mm_mul_ps(iq3,jq3);
 167
 168     /* Avoid stupid compiler warnings */
 169     jnrA = jnrB = jnrC = jnrD = 0;
 170     j_coord_offsetA = 0;
 171     j_coord_offsetB = 0;
 172     j_coord_offsetC = 0;
 173     j_coord_offsetD = 0;
 174
 175     outeriter        = 0;
 176     inneriter        = 0;
 177
 178     for(iidx=0;iidx<4*DIM;iidx++)
 179     {
 180         scratch[iidx] = 0.0;
 181     }
 182
 183     /* Start outer loop over neighborlists */
 184     for(iidx=0; iidx<nri; iidx++)
 185     {
 186         /* Load shift vector for this list */
 187         i_shift_offset   = DIM*shiftidx[iidx];
 188
 189         /* Load limits for loop over neighbors */
 190         j_index_start    = jindex[iidx];
 191         j_index_end      = jindex[iidx+1];
 192
 193         /* Get outer coordinate index */
 194         inr              = iinr[iidx];
 195         i_coord_offset   = DIM*inr;
 196
 197         /* Load i particle coords and add shift vector */
 198         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 199                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 200
 201         fix0             = _mm_setzero_ps();
 202         fiy0             = _mm_setzero_ps();
 203         fiz0             = _mm_setzero_ps();
 204         fix1             = _mm_setzero_ps();
 205         fiy1             = _mm_setzero_ps();
 206         fiz1             = _mm_setzero_ps();
 207         fix2             = _mm_setzero_ps();
 208         fiy2             = _mm_setzero_ps();
 209         fiz2             = _mm_setzero_ps();
 210         fix3             = _mm_setzero_ps();
 211         fiy3             = _mm_setzero_ps();
 212         fiz3             = _mm_setzero_ps();
 213
 214         /* Reset potential sums */
 215         velecsum         = _mm_setzero_ps();
 216         vvdwsum          = _mm_setzero_ps();
 217
 218         /* Start inner kernel loop */
 219         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 220         {
 221
 222             /* Get j neighbor index, and coordinate index */
 223             jnrA             = jjnr[jidx];
 224             jnrB             = jjnr[jidx+1];
 225             jnrC             = jjnr[jidx+2];
 226             jnrD             = jjnr[jidx+3];
 227             j_coord_offsetA  = DIM*jnrA;
 228             j_coord_offsetB  = DIM*jnrB;
 229             j_coord_offsetC  = DIM*jnrC;
 230             j_coord_offsetD  = DIM*jnrD;
 231
 232             /* load j atom coordinates */
 233             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 234                                               x+j_coord_offsetC,x+j_coord_offsetD,
 235                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 236                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 237
 238             /* Calculate displacement vector */
 239             dx00             = _mm_sub_ps(ix0,jx0);
 240             dy00             = _mm_sub_ps(iy0,jy0);
 241             dz00             = _mm_sub_ps(iz0,jz0);
 242             dx11             = _mm_sub_ps(ix1,jx1);
 243             dy11             = _mm_sub_ps(iy1,jy1);
 244             dz11             = _mm_sub_ps(iz1,jz1);
 245             dx12             = _mm_sub_ps(ix1,jx2);
 246             dy12             = _mm_sub_ps(iy1,jy2);
 247             dz12             = _mm_sub_ps(iz1,jz2);
 248             dx13             = _mm_sub_ps(ix1,jx3);
 249             dy13             = _mm_sub_ps(iy1,jy3);
 250             dz13             = _mm_sub_ps(iz1,jz3);
 251             dx21             = _mm_sub_ps(ix2,jx1);
 252             dy21             = _mm_sub_ps(iy2,jy1);
 253             dz21             = _mm_sub_ps(iz2,jz1);
 254             dx22             = _mm_sub_ps(ix2,jx2);
 255             dy22             = _mm_sub_ps(iy2,jy2);
 256             dz22             = _mm_sub_ps(iz2,jz2);
 257             dx23             = _mm_sub_ps(ix2,jx3);
 258             dy23             = _mm_sub_ps(iy2,jy3);
 259             dz23             = _mm_sub_ps(iz2,jz3);
 260             dx31             = _mm_sub_ps(ix3,jx1);
 261             dy31             = _mm_sub_ps(iy3,jy1);
 262             dz31             = _mm_sub_ps(iz3,jz1);
 263             dx32             = _mm_sub_ps(ix3,jx2);
 264             dy32             = _mm_sub_ps(iy3,jy2);
 265             dz32             = _mm_sub_ps(iz3,jz2);
 266             dx33             = _mm_sub_ps(ix3,jx3);
 267             dy33             = _mm_sub_ps(iy3,jy3);
 268             dz33             = _mm_sub_ps(iz3,jz3);
 269
 270             /* Calculate squared distance and things based on it */
 271             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 272             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 273             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 274             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 275             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 276             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 277             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 278             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 279             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 280             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 281
 282             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 283             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 284             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 285             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 286             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 287             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 288             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 289             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 290             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 291             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 292
 293             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 294             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 295             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 296             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 297             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 298             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 299             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 300             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 301             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 302
 303             fjx0             = _mm_setzero_ps();
 304             fjy0             = _mm_setzero_ps();
 305             fjz0             = _mm_setzero_ps();
 306             fjx1             = _mm_setzero_ps();
 307             fjy1             = _mm_setzero_ps();
 308             fjz1             = _mm_setzero_ps();
 309             fjx2             = _mm_setzero_ps();
 310             fjy2             = _mm_setzero_ps();
 311             fjz2             = _mm_setzero_ps();
 312             fjx3             = _mm_setzero_ps();
 313             fjy3             = _mm_setzero_ps();
 314             fjz3             = _mm_setzero_ps();
 315
 316             /**************************
 317              * CALCULATE INTERACTIONS *
 318              **************************/
 319
 320             r00              = _mm_mul_ps(rsq00,rinv00);
 321
 322             /* Calculate table index by multiplying r with table scale and truncate to integer */
 323             rt               = _mm_mul_ps(r00,vftabscale);
 324             vfitab           = _mm_cvttps_epi32(rt);
 325 #ifdef __XOP__
 326             vfeps            = _mm_frcz_ps(rt);
 327 #else
 328             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 329 #endif
 330             twovfeps         = _mm_add_ps(vfeps,vfeps);
 331             vfitab           = _mm_slli_epi32(vfitab,3);
 332
 333             /* CUBIC SPLINE TABLE DISPERSION */
 334             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 335             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 336             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 337             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 338             _MM_TRANSPOSE4_PS(Y,F,G,H);
 339             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 340             VV               = _mm_macc_ps(vfeps,Fp,Y);
 341             vvdw6            = _mm_mul_ps(c6_00,VV);
 342             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 343             fvdw6            = _mm_mul_ps(c6_00,FF);
 344
 345             /* CUBIC SPLINE TABLE REPULSION */
 346             vfitab           = _mm_add_epi32(vfitab,ifour);
 347             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 348             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 349             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 350             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 351             _MM_TRANSPOSE4_PS(Y,F,G,H);
 352             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 353             VV               = _mm_macc_ps(vfeps,Fp,Y);
 354             vvdw12           = _mm_mul_ps(c12_00,VV);
 355             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 356             fvdw12           = _mm_mul_ps(c12_00,FF);
 357             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 358             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 359
 360             /* Update potential sum for this i atom from the interaction with this j atom. */
 361             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 362
 363             fscal            = fvdw;
 364
 365              /* Update vectorial force */
 366             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 367             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 368             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 369
 370             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 371             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 372             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 373
 374             /**************************
 375              * CALCULATE INTERACTIONS *
 376              **************************/
 377
 378             /* COULOMB ELECTROSTATICS */
 379             velec            = _mm_mul_ps(qq11,rinv11);
 380             felec            = _mm_mul_ps(velec,rinvsq11);
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velecsum         = _mm_add_ps(velecsum,velec);
 384
 385             fscal            = felec;
 386
 387              /* Update vectorial force */
 388             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 389             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 390             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 391
 392             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 393             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 394             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 395
 396             /**************************
 397              * CALCULATE INTERACTIONS *
 398              **************************/
 399
 400             /* COULOMB ELECTROSTATICS */
 401             velec            = _mm_mul_ps(qq12,rinv12);
 402             felec            = _mm_mul_ps(velec,rinvsq12);
 403
 404             /* Update potential sum for this i atom from the interaction with this j atom. */
 405             velecsum         = _mm_add_ps(velecsum,velec);
 406
 407             fscal            = felec;
 408
 409              /* Update vectorial force */
 410             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 411             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
 412             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
 413
 414             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
 415             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
 416             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
 417
 418             /**************************
 419              * CALCULATE INTERACTIONS *
 420              **************************/
 421
 422             /* COULOMB ELECTROSTATICS */
 423             velec            = _mm_mul_ps(qq13,rinv13);
 424             felec            = _mm_mul_ps(velec,rinvsq13);
 425
 426             /* Update potential sum for this i atom from the interaction with this j atom. */
 427             velecsum         = _mm_add_ps(velecsum,velec);
 428
 429             fscal            = felec;
 430
 431              /* Update vectorial force */
 432             fix1             = _mm_macc_ps(dx13,fscal,fix1);
 433             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
 434             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
 435
 436             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
 437             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
 438             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
 439
 440             /**************************
 441              * CALCULATE INTERACTIONS *
 442              **************************/
 443
 444             /* COULOMB ELECTROSTATICS */
 445             velec            = _mm_mul_ps(qq21,rinv21);
 446             felec            = _mm_mul_ps(velec,rinvsq21);
 447
 448             /* Update potential sum for this i atom from the interaction with this j atom. */
 449             velecsum         = _mm_add_ps(velecsum,velec);
 450
 451             fscal            = felec;
 452
 453              /* Update vectorial force */
 454             fix2             = _mm_macc_ps(dx21,fscal,fix2);
 455             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
 456             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
 457
 458             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
 459             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
 460             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
 461
 462             /**************************
 463              * CALCULATE INTERACTIONS *
 464              **************************/
 465
 466             /* COULOMB ELECTROSTATICS */
 467             velec            = _mm_mul_ps(qq22,rinv22);
 468             felec            = _mm_mul_ps(velec,rinvsq22);
 469
 470             /* Update potential sum for this i atom from the interaction with this j atom. */
 471             velecsum         = _mm_add_ps(velecsum,velec);
 472
 473             fscal            = felec;
 474
 475              /* Update vectorial force */
 476             fix2             = _mm_macc_ps(dx22,fscal,fix2);
 477             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
 478             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
 479
 480             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
 481             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
 482             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
 483
 484             /**************************
 485              * CALCULATE INTERACTIONS *
 486              **************************/
 487
 488             /* COULOMB ELECTROSTATICS */
 489             velec            = _mm_mul_ps(qq23,rinv23);
 490             felec            = _mm_mul_ps(velec,rinvsq23);
 491
 492             /* Update potential sum for this i atom from the interaction with this j atom. */
 493             velecsum         = _mm_add_ps(velecsum,velec);
 494
 495             fscal            = felec;
 496
 497              /* Update vectorial force */
 498             fix2             = _mm_macc_ps(dx23,fscal,fix2);
 499             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
 500             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
 501
 502             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
 503             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
 504             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
 505
 506             /**************************
 507              * CALCULATE INTERACTIONS *
 508              **************************/
 509
 510             /* COULOMB ELECTROSTATICS */
 511             velec            = _mm_mul_ps(qq31,rinv31);
 512             felec            = _mm_mul_ps(velec,rinvsq31);
 513
 514             /* Update potential sum for this i atom from the interaction with this j atom. */
 515             velecsum         = _mm_add_ps(velecsum,velec);
 516
 517             fscal            = felec;
 518
 519              /* Update vectorial force */
 520             fix3             = _mm_macc_ps(dx31,fscal,fix3);
 521             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
 522             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
 523
 524             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
 525             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
 526             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
 527
 528             /**************************
 529              * CALCULATE INTERACTIONS *
 530              **************************/
 531
 532             /* COULOMB ELECTROSTATICS */
 533             velec            = _mm_mul_ps(qq32,rinv32);
 534             felec            = _mm_mul_ps(velec,rinvsq32);
 535
 536             /* Update potential sum for this i atom from the interaction with this j atom. */
 537             velecsum         = _mm_add_ps(velecsum,velec);
 538
 539             fscal            = felec;
 540
 541              /* Update vectorial force */
 542             fix3             = _mm_macc_ps(dx32,fscal,fix3);
 543             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
 544             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
 545
 546             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
 547             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
 548             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
 549
 550             /**************************
 551              * CALCULATE INTERACTIONS *
 552              **************************/
 553
 554             /* COULOMB ELECTROSTATICS */
 555             velec            = _mm_mul_ps(qq33,rinv33);
 556             felec            = _mm_mul_ps(velec,rinvsq33);
 557
 558             /* Update potential sum for this i atom from the interaction with this j atom. */
 559             velecsum         = _mm_add_ps(velecsum,velec);
 560
 561             fscal            = felec;
 562
 563              /* Update vectorial force */
 564             fix3             = _mm_macc_ps(dx33,fscal,fix3);
 565             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
 566             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
 567
 568             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
 569             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
 570             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
 571
 572             fjptrA             = f+j_coord_offsetA;
 573             fjptrB             = f+j_coord_offsetB;
 574             fjptrC             = f+j_coord_offsetC;
 575             fjptrD             = f+j_coord_offsetD;
 576
 577             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 578                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 579                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 580
 581             /* Inner loop uses 341 flops */
 582         }
 583
 584         if(jidx<j_index_end)
 585         {
 586
 587             /* Get j neighbor index, and coordinate index */
 588             jnrlistA         = jjnr[jidx];
 589             jnrlistB         = jjnr[jidx+1];
 590             jnrlistC         = jjnr[jidx+2];
 591             jnrlistD         = jjnr[jidx+3];
 592             /* Sign of each element will be negative for non-real atoms.
 593              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 594              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 595              */
 596             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 597             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 598             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 599             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 600             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 601             j_coord_offsetA  = DIM*jnrA;
 602             j_coord_offsetB  = DIM*jnrB;
 603             j_coord_offsetC  = DIM*jnrC;
 604             j_coord_offsetD  = DIM*jnrD;
 605
 606             /* load j atom coordinates */
 607             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 608                                               x+j_coord_offsetC,x+j_coord_offsetD,
 609                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 610                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 611
 612             /* Calculate displacement vector */
 613             dx00             = _mm_sub_ps(ix0,jx0);
 614             dy00             = _mm_sub_ps(iy0,jy0);
 615             dz00             = _mm_sub_ps(iz0,jz0);
 616             dx11             = _mm_sub_ps(ix1,jx1);
 617             dy11             = _mm_sub_ps(iy1,jy1);
 618             dz11             = _mm_sub_ps(iz1,jz1);
 619             dx12             = _mm_sub_ps(ix1,jx2);
 620             dy12             = _mm_sub_ps(iy1,jy2);
 621             dz12             = _mm_sub_ps(iz1,jz2);
 622             dx13             = _mm_sub_ps(ix1,jx3);
 623             dy13             = _mm_sub_ps(iy1,jy3);
 624             dz13             = _mm_sub_ps(iz1,jz3);
 625             dx21             = _mm_sub_ps(ix2,jx1);
 626             dy21             = _mm_sub_ps(iy2,jy1);
 627             dz21             = _mm_sub_ps(iz2,jz1);
 628             dx22             = _mm_sub_ps(ix2,jx2);
 629             dy22             = _mm_sub_ps(iy2,jy2);
 630             dz22             = _mm_sub_ps(iz2,jz2);
 631             dx23             = _mm_sub_ps(ix2,jx3);
 632             dy23             = _mm_sub_ps(iy2,jy3);
 633             dz23             = _mm_sub_ps(iz2,jz3);
 634             dx31             = _mm_sub_ps(ix3,jx1);
 635             dy31             = _mm_sub_ps(iy3,jy1);
 636             dz31             = _mm_sub_ps(iz3,jz1);
 637             dx32             = _mm_sub_ps(ix3,jx2);
 638             dy32             = _mm_sub_ps(iy3,jy2);
 639             dz32             = _mm_sub_ps(iz3,jz2);
 640             dx33             = _mm_sub_ps(ix3,jx3);
 641             dy33             = _mm_sub_ps(iy3,jy3);
 642             dz33             = _mm_sub_ps(iz3,jz3);
 643
 644             /* Calculate squared distance and things based on it */
 645             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 646             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 647             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 648             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 649             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 650             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 651             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 652             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 653             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 654             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 655
 656             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 657             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 658             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 659             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 660             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 661             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 662             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 663             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 664             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 665             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 666
 667             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 668             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 669             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 670             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 671             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 672             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 673             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 674             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 675             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 676
 677             fjx0             = _mm_setzero_ps();
 678             fjy0             = _mm_setzero_ps();
 679             fjz0             = _mm_setzero_ps();
 680             fjx1             = _mm_setzero_ps();
 681             fjy1             = _mm_setzero_ps();
 682             fjz1             = _mm_setzero_ps();
 683             fjx2             = _mm_setzero_ps();
 684             fjy2             = _mm_setzero_ps();
 685             fjz2             = _mm_setzero_ps();
 686             fjx3             = _mm_setzero_ps();
 687             fjy3             = _mm_setzero_ps();
 688             fjz3             = _mm_setzero_ps();
 689
 690             /**************************
 691              * CALCULATE INTERACTIONS *
 692              **************************/
 693
 694             r00              = _mm_mul_ps(rsq00,rinv00);
 695             r00              = _mm_andnot_ps(dummy_mask,r00);
 696
 697             /* Calculate table index by multiplying r with table scale and truncate to integer */
 698             rt               = _mm_mul_ps(r00,vftabscale);
 699             vfitab           = _mm_cvttps_epi32(rt);
 700 #ifdef __XOP__
 701             vfeps            = _mm_frcz_ps(rt);
 702 #else
 703             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 704 #endif
 705             twovfeps         = _mm_add_ps(vfeps,vfeps);
 706             vfitab           = _mm_slli_epi32(vfitab,3);
 707
 708             /* CUBIC SPLINE TABLE DISPERSION */
 709             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 710             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 711             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 712             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 713             _MM_TRANSPOSE4_PS(Y,F,G,H);
 714             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 715             VV               = _mm_macc_ps(vfeps,Fp,Y);
 716             vvdw6            = _mm_mul_ps(c6_00,VV);
 717             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 718             fvdw6            = _mm_mul_ps(c6_00,FF);
 719
 720             /* CUBIC SPLINE TABLE REPULSION */
 721             vfitab           = _mm_add_epi32(vfitab,ifour);
 722             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 723             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 724             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 725             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 726             _MM_TRANSPOSE4_PS(Y,F,G,H);
 727             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 728             VV               = _mm_macc_ps(vfeps,Fp,Y);
 729             vvdw12           = _mm_mul_ps(c12_00,VV);
 730             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 731             fvdw12           = _mm_mul_ps(c12_00,FF);
 732             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 733             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 734
 735             /* Update potential sum for this i atom from the interaction with this j atom. */
 736             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 737             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 738
 739             fscal            = fvdw;
 740
 741             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 742
 743              /* Update vectorial force */
 744             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 745             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 746             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 747
 748             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 749             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 750             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 751
 752             /**************************
 753              * CALCULATE INTERACTIONS *
 754              **************************/
 755
 756             /* COULOMB ELECTROSTATICS */
 757             velec            = _mm_mul_ps(qq11,rinv11);
 758             felec            = _mm_mul_ps(velec,rinvsq11);
 759
 760             /* Update potential sum for this i atom from the interaction with this j atom. */
 761             velec            = _mm_andnot_ps(dummy_mask,velec);
 762             velecsum         = _mm_add_ps(velecsum,velec);
 763
 764             fscal            = felec;
 765
 766             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 767
 768              /* Update vectorial force */
 769             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 770             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 771             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 772
 773             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 774             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 775             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 776
 777             /**************************
 778              * CALCULATE INTERACTIONS *
 779              **************************/
 780
 781             /* COULOMB ELECTROSTATICS */
 782             velec            = _mm_mul_ps(qq12,rinv12);
 783             felec            = _mm_mul_ps(velec,rinvsq12);
 784
 785             /* Update potential sum for this i atom from the interaction with this j atom. */
 786             velec            = _mm_andnot_ps(dummy_mask,velec);
 787             velecsum         = _mm_add_ps(velecsum,velec);
 788
 789             fscal            = felec;
 790
 791             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 792
 793              /* Update vectorial force */
 794             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 795             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
 796             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
 797
 798             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
 799             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
 800             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
 801
 802             /**************************
 803              * CALCULATE INTERACTIONS *
 804              **************************/
 805
 806             /* COULOMB ELECTROSTATICS */
 807             velec            = _mm_mul_ps(qq13,rinv13);
 808             felec            = _mm_mul_ps(velec,rinvsq13);
 809
 810             /* Update potential sum for this i atom from the interaction with this j atom. */
 811             velec            = _mm_andnot_ps(dummy_mask,velec);
 812             velecsum         = _mm_add_ps(velecsum,velec);
 813
 814             fscal            = felec;
 815
 816             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 817
 818              /* Update vectorial force */
 819             fix1             = _mm_macc_ps(dx13,fscal,fix1);
 820             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
 821             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
 822
 823             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
 824             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
 825             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
 826
 827             /**************************
 828              * CALCULATE INTERACTIONS *
 829              **************************/
 830
 831             /* COULOMB ELECTROSTATICS */
 832             velec            = _mm_mul_ps(qq21,rinv21);
 833             felec            = _mm_mul_ps(velec,rinvsq21);
 834
 835             /* Update potential sum for this i atom from the interaction with this j atom. */
 836             velec            = _mm_andnot_ps(dummy_mask,velec);
 837             velecsum         = _mm_add_ps(velecsum,velec);
 838
 839             fscal            = felec;
 840
 841             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 842
 843              /* Update vectorial force */
 844             fix2             = _mm_macc_ps(dx21,fscal,fix2);
 845             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
 846             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
 847
 848             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
 849             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
 850             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
 851
 852             /**************************
 853              * CALCULATE INTERACTIONS *
 854              **************************/
 855
 856             /* COULOMB ELECTROSTATICS */
 857             velec            = _mm_mul_ps(qq22,rinv22);
 858             felec            = _mm_mul_ps(velec,rinvsq22);
 859
 860             /* Update potential sum for this i atom from the interaction with this j atom. */
 861             velec            = _mm_andnot_ps(dummy_mask,velec);
 862             velecsum         = _mm_add_ps(velecsum,velec);
 863
 864             fscal            = felec;
 865
 866             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 867
 868              /* Update vectorial force */
 869             fix2             = _mm_macc_ps(dx22,fscal,fix2);
 870             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
 871             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
 872
 873             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
 874             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
 875             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
 876
 877             /**************************
 878              * CALCULATE INTERACTIONS *
 879              **************************/
 880
 881             /* COULOMB ELECTROSTATICS */
 882             velec            = _mm_mul_ps(qq23,rinv23);
 883             felec            = _mm_mul_ps(velec,rinvsq23);
 884
 885             /* Update potential sum for this i atom from the interaction with this j atom. */
 886             velec            = _mm_andnot_ps(dummy_mask,velec);
 887             velecsum         = _mm_add_ps(velecsum,velec);
 888
 889             fscal            = felec;
 890
 891             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 892
 893              /* Update vectorial force */
 894             fix2             = _mm_macc_ps(dx23,fscal,fix2);
 895             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
 896             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
 897
 898             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
 899             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
 900             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
 901
 902             /**************************
 903              * CALCULATE INTERACTIONS *
 904              **************************/
 905
 906             /* COULOMB ELECTROSTATICS */
 907             velec            = _mm_mul_ps(qq31,rinv31);
 908             felec            = _mm_mul_ps(velec,rinvsq31);
 909
 910             /* Update potential sum for this i atom from the interaction with this j atom. */
 911             velec            = _mm_andnot_ps(dummy_mask,velec);
 912             velecsum         = _mm_add_ps(velecsum,velec);
 913
 914             fscal            = felec;
 915
 916             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 917
 918              /* Update vectorial force */
 919             fix3             = _mm_macc_ps(dx31,fscal,fix3);
 920             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
 921             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
 922
 923             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
 924             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
 925             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
 926
 927             /**************************
 928              * CALCULATE INTERACTIONS *
 929              **************************/
 930
 931             /* COULOMB ELECTROSTATICS */
 932             velec            = _mm_mul_ps(qq32,rinv32);
 933             felec            = _mm_mul_ps(velec,rinvsq32);
 934
 935             /* Update potential sum for this i atom from the interaction with this j atom. */
 936             velec            = _mm_andnot_ps(dummy_mask,velec);
 937             velecsum         = _mm_add_ps(velecsum,velec);
 938
 939             fscal            = felec;
 940
 941             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 942
 943              /* Update vectorial force */
 944             fix3             = _mm_macc_ps(dx32,fscal,fix3);
 945             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
 946             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
 947
 948             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
 949             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
 950             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
 951
 952             /**************************
 953              * CALCULATE INTERACTIONS *
 954              **************************/
 955
 956             /* COULOMB ELECTROSTATICS */
 957             velec            = _mm_mul_ps(qq33,rinv33);
 958             felec            = _mm_mul_ps(velec,rinvsq33);
 959
 960             /* Update potential sum for this i atom from the interaction with this j atom. */
 961             velec            = _mm_andnot_ps(dummy_mask,velec);
 962             velecsum         = _mm_add_ps(velecsum,velec);
 963
 964             fscal            = felec;
 965
 966             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 967
 968              /* Update vectorial force */
 969             fix3             = _mm_macc_ps(dx33,fscal,fix3);
 970             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
 971             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
 972
 973             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
 974             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
 975             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
 976
 977             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 978             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 979             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 980             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 981
 982             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 983                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 984                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 985
 986             /* Inner loop uses 342 flops */
 987         }
 988
 989         /* End of innermost loop */
 990
 991         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 992                                               f+i_coord_offset,fshift+i_shift_offset);
 993
 994         ggid                        = gid[iidx];
 995         /* Update potential energies */
 996         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 997         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 998
 999         /* Increment number of inner iterations */
1000         inneriter                  += j_index_end - j_index_start;
1001
1002         /* Outer loop uses 26 flops */
1003     }
1004
1005     /* Increment number of outer iterations */
1006     outeriter        += nri;
1007
1008     /* Update outer/inner flops */
1009
1010     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*342);
1011 }
1012 /*
1013  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_128_fma_single
1014  * Electrostatics interaction: Coulomb
1015  * VdW interaction:            CubicSplineTable
1016  * Geometry:                   Water4-Water4
1017  * Calculate force/pot:        Force
1018  */
1019 void
1020 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_128_fma_single
1021                     (t_nblist                    * gmx_restrict       nlist,
1022                      rvec                        * gmx_restrict          xx,
1023                      rvec                        * gmx_restrict          ff,
1024                      t_forcerec                  * gmx_restrict          fr,
1025                      t_mdatoms                   * gmx_restrict     mdatoms,
1026                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1027                      t_nrnb                      * gmx_restrict        nrnb)
1028 {
1029     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1030      * just 0 for non-waters.
1031      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
1032      * jnr indices corresponding to data put in the four positions in the SIMD register.
1033      */
1034     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1035     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1036     int              jnrA,jnrB,jnrC,jnrD;
1037     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1038     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1039     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1040     real             rcutoff_scalar;
1041     real             *shiftvec,*fshift,*x,*f;
1042     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1043     real             scratch[4*DIM];
1044     __m128           fscal,rcutoff,rcutoff2,jidxall;
1045     int              vdwioffset0;
1046     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1047     int              vdwioffset1;
1048     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1049     int              vdwioffset2;
1050     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1051     int              vdwioffset3;
1052     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1053     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1054     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1055     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1056     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1057     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1058     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1059     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1060     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1061     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1062     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1063     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1064     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1065     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1066     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1067     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1068     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1069     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1070     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1071     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1072     real             *charge;
1073     int              nvdwtype;
1074     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1075     int              *vdwtype;
1076     real             *vdwparam;
1077     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1078     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1079     __m128i          vfitab;
1080     __m128i          ifour       = _mm_set1_epi32(4);
1081     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
1082     real             *vftab;
1083     __m128           dummy_mask,cutoff_mask;
1084     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1085     __m128           one     = _mm_set1_ps(1.0);
1086     __m128           two     = _mm_set1_ps(2.0);
1087     x                = xx[0];
1088     f                = ff[0];
1089
1090     nri              = nlist->nri;
1091     iinr             = nlist->iinr;
1092     jindex           = nlist->jindex;
1093     jjnr             = nlist->jjnr;
1094     shiftidx         = nlist->shift;
1095     gid              = nlist->gid;
1096     shiftvec         = fr->shift_vec[0];
1097     fshift           = fr->fshift[0];
1098     facel            = _mm_set1_ps(fr->epsfac);
1099     charge           = mdatoms->chargeA;
1100     nvdwtype         = fr->ntype;
1101     vdwparam         = fr->nbfp;
1102     vdwtype          = mdatoms->typeA;
1103
1104     vftab            = kernel_data->table_vdw->data;
1105     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
1106
1107     /* Setup water-specific parameters */
1108     inr              = nlist->iinr[0];
1109     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1110     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1111     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1112     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1113
1114     jq1              = _mm_set1_ps(charge[inr+1]);
1115     jq2              = _mm_set1_ps(charge[inr+2]);
1116     jq3              = _mm_set1_ps(charge[inr+3]);
1117     vdwjidx0A        = 2*vdwtype[inr+0];
1118     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1119     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1120     qq11             = _mm_mul_ps(iq1,jq1);
1121     qq12             = _mm_mul_ps(iq1,jq2);
1122     qq13             = _mm_mul_ps(iq1,jq3);
1123     qq21             = _mm_mul_ps(iq2,jq1);
1124     qq22             = _mm_mul_ps(iq2,jq2);
1125     qq23             = _mm_mul_ps(iq2,jq3);
1126     qq31             = _mm_mul_ps(iq3,jq1);
1127     qq32             = _mm_mul_ps(iq3,jq2);
1128     qq33             = _mm_mul_ps(iq3,jq3);
1129
1130     /* Avoid stupid compiler warnings */
1131     jnrA = jnrB = jnrC = jnrD = 0;
1132     j_coord_offsetA = 0;
1133     j_coord_offsetB = 0;
1134     j_coord_offsetC = 0;
1135     j_coord_offsetD = 0;
1136
1137     outeriter        = 0;
1138     inneriter        = 0;
1139
1140     for(iidx=0;iidx<4*DIM;iidx++)
1141     {
1142         scratch[iidx] = 0.0;
1143     }
1144
1145     /* Start outer loop over neighborlists */
1146     for(iidx=0; iidx<nri; iidx++)
1147     {
1148         /* Load shift vector for this list */
1149         i_shift_offset   = DIM*shiftidx[iidx];
1150
1151         /* Load limits for loop over neighbors */
1152         j_index_start    = jindex[iidx];
1153         j_index_end      = jindex[iidx+1];
1154
1155         /* Get outer coordinate index */
1156         inr              = iinr[iidx];
1157         i_coord_offset   = DIM*inr;
1158
1159         /* Load i particle coords and add shift vector */
1160         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1161                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1162
1163         fix0             = _mm_setzero_ps();
1164         fiy0             = _mm_setzero_ps();
1165         fiz0             = _mm_setzero_ps();
1166         fix1             = _mm_setzero_ps();
1167         fiy1             = _mm_setzero_ps();
1168         fiz1             = _mm_setzero_ps();
1169         fix2             = _mm_setzero_ps();
1170         fiy2             = _mm_setzero_ps();
1171         fiz2             = _mm_setzero_ps();
1172         fix3             = _mm_setzero_ps();
1173         fiy3             = _mm_setzero_ps();
1174         fiz3             = _mm_setzero_ps();
1175
1176         /* Start inner kernel loop */
1177         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1178         {
1179
1180             /* Get j neighbor index, and coordinate index */
1181             jnrA             = jjnr[jidx];
1182             jnrB             = jjnr[jidx+1];
1183             jnrC             = jjnr[jidx+2];
1184             jnrD             = jjnr[jidx+3];
1185             j_coord_offsetA  = DIM*jnrA;
1186             j_coord_offsetB  = DIM*jnrB;
1187             j_coord_offsetC  = DIM*jnrC;
1188             j_coord_offsetD  = DIM*jnrD;
1189
1190             /* load j atom coordinates */
1191             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1192                                               x+j_coord_offsetC,x+j_coord_offsetD,
1193                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1194                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1195
1196             /* Calculate displacement vector */
1197             dx00             = _mm_sub_ps(ix0,jx0);
1198             dy00             = _mm_sub_ps(iy0,jy0);
1199             dz00             = _mm_sub_ps(iz0,jz0);
1200             dx11             = _mm_sub_ps(ix1,jx1);
1201             dy11             = _mm_sub_ps(iy1,jy1);
1202             dz11             = _mm_sub_ps(iz1,jz1);
1203             dx12             = _mm_sub_ps(ix1,jx2);
1204             dy12             = _mm_sub_ps(iy1,jy2);
1205             dz12             = _mm_sub_ps(iz1,jz2);
1206             dx13             = _mm_sub_ps(ix1,jx3);
1207             dy13             = _mm_sub_ps(iy1,jy3);
1208             dz13             = _mm_sub_ps(iz1,jz3);
1209             dx21             = _mm_sub_ps(ix2,jx1);
1210             dy21             = _mm_sub_ps(iy2,jy1);
1211             dz21             = _mm_sub_ps(iz2,jz1);
1212             dx22             = _mm_sub_ps(ix2,jx2);
1213             dy22             = _mm_sub_ps(iy2,jy2);
1214             dz22             = _mm_sub_ps(iz2,jz2);
1215             dx23             = _mm_sub_ps(ix2,jx3);
1216             dy23             = _mm_sub_ps(iy2,jy3);
1217             dz23             = _mm_sub_ps(iz2,jz3);
1218             dx31             = _mm_sub_ps(ix3,jx1);
1219             dy31             = _mm_sub_ps(iy3,jy1);
1220             dz31             = _mm_sub_ps(iz3,jz1);
1221             dx32             = _mm_sub_ps(ix3,jx2);
1222             dy32             = _mm_sub_ps(iy3,jy2);
1223             dz32             = _mm_sub_ps(iz3,jz2);
1224             dx33             = _mm_sub_ps(ix3,jx3);
1225             dy33             = _mm_sub_ps(iy3,jy3);
1226             dz33             = _mm_sub_ps(iz3,jz3);
1227
1228             /* Calculate squared distance and things based on it */
1229             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1230             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1231             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1232             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1233             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1234             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1235             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1236             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1237             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1238             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1239
1240             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1241             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1242             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1243             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1244             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1245             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1246             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1247             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1248             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1249             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1250
1251             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1252             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1253             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1254             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1255             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1256             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1257             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1258             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1259             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1260
1261             fjx0             = _mm_setzero_ps();
1262             fjy0             = _mm_setzero_ps();
1263             fjz0             = _mm_setzero_ps();
1264             fjx1             = _mm_setzero_ps();
1265             fjy1             = _mm_setzero_ps();
1266             fjz1             = _mm_setzero_ps();
1267             fjx2             = _mm_setzero_ps();
1268             fjy2             = _mm_setzero_ps();
1269             fjz2             = _mm_setzero_ps();
1270             fjx3             = _mm_setzero_ps();
1271             fjy3             = _mm_setzero_ps();
1272             fjz3             = _mm_setzero_ps();
1273
1274             /**************************
1275              * CALCULATE INTERACTIONS *
1276              **************************/
1277
1278             r00              = _mm_mul_ps(rsq00,rinv00);
1279
1280             /* Calculate table index by multiplying r with table scale and truncate to integer */
1281             rt               = _mm_mul_ps(r00,vftabscale);
1282             vfitab           = _mm_cvttps_epi32(rt);
1283 #ifdef __XOP__
1284             vfeps            = _mm_frcz_ps(rt);
1285 #else
1286             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1287 #endif
1288             twovfeps         = _mm_add_ps(vfeps,vfeps);
1289             vfitab           = _mm_slli_epi32(vfitab,3);
1290
1291             /* CUBIC SPLINE TABLE DISPERSION */
1292             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1293             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1294             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1295             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1296             _MM_TRANSPOSE4_PS(Y,F,G,H);
1297             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1298             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1299             fvdw6            = _mm_mul_ps(c6_00,FF);
1300
1301             /* CUBIC SPLINE TABLE REPULSION */
1302             vfitab           = _mm_add_epi32(vfitab,ifour);
1303             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1304             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1305             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1306             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1307             _MM_TRANSPOSE4_PS(Y,F,G,H);
1308             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1309             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1310             fvdw12           = _mm_mul_ps(c12_00,FF);
1311             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1312
1313             fscal            = fvdw;
1314
1315              /* Update vectorial force */
1316             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1317             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1318             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1319
1320             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1321             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1322             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1323
1324             /**************************
1325              * CALCULATE INTERACTIONS *
1326              **************************/
1327
1328             /* COULOMB ELECTROSTATICS */
1329             velec            = _mm_mul_ps(qq11,rinv11);
1330             felec            = _mm_mul_ps(velec,rinvsq11);
1331
1332             fscal            = felec;
1333
1334              /* Update vectorial force */
1335             fix1             = _mm_macc_ps(dx11,fscal,fix1);
1336             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
1337             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
1338
1339             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
1340             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
1341             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
1342
1343             /**************************
1344              * CALCULATE INTERACTIONS *
1345              **************************/
1346
1347             /* COULOMB ELECTROSTATICS */
1348             velec            = _mm_mul_ps(qq12,rinv12);
1349             felec            = _mm_mul_ps(velec,rinvsq12);
1350
1351             fscal            = felec;
1352
1353              /* Update vectorial force */
1354             fix1             = _mm_macc_ps(dx12,fscal,fix1);
1355             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1356             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1357
1358             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1359             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1360             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1361
1362             /**************************
1363              * CALCULATE INTERACTIONS *
1364              **************************/
1365
1366             /* COULOMB ELECTROSTATICS */
1367             velec            = _mm_mul_ps(qq13,rinv13);
1368             felec            = _mm_mul_ps(velec,rinvsq13);
1369
1370             fscal            = felec;
1371
1372              /* Update vectorial force */
1373             fix1             = _mm_macc_ps(dx13,fscal,fix1);
1374             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
1375             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
1376
1377             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
1378             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
1379             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
1380
1381             /**************************
1382              * CALCULATE INTERACTIONS *
1383              **************************/
1384
1385             /* COULOMB ELECTROSTATICS */
1386             velec            = _mm_mul_ps(qq21,rinv21);
1387             felec            = _mm_mul_ps(velec,rinvsq21);
1388
1389             fscal            = felec;
1390
1391              /* Update vectorial force */
1392             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1393             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1394             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1395
1396             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1397             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1398             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1399
1400             /**************************
1401              * CALCULATE INTERACTIONS *
1402              **************************/
1403
1404             /* COULOMB ELECTROSTATICS */
1405             velec            = _mm_mul_ps(qq22,rinv22);
1406             felec            = _mm_mul_ps(velec,rinvsq22);
1407
1408             fscal            = felec;
1409
1410              /* Update vectorial force */
1411             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1412             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1413             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1414
1415             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1416             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1417             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1418
1419             /**************************
1420              * CALCULATE INTERACTIONS *
1421              **************************/
1422
1423             /* COULOMB ELECTROSTATICS */
1424             velec            = _mm_mul_ps(qq23,rinv23);
1425             felec            = _mm_mul_ps(velec,rinvsq23);
1426
1427             fscal            = felec;
1428
1429              /* Update vectorial force */
1430             fix2             = _mm_macc_ps(dx23,fscal,fix2);
1431             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
1432             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
1433
1434             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
1435             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
1436             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
1437
1438             /**************************
1439              * CALCULATE INTERACTIONS *
1440              **************************/
1441
1442             /* COULOMB ELECTROSTATICS */
1443             velec            = _mm_mul_ps(qq31,rinv31);
1444             felec            = _mm_mul_ps(velec,rinvsq31);
1445
1446             fscal            = felec;
1447
1448              /* Update vectorial force */
1449             fix3             = _mm_macc_ps(dx31,fscal,fix3);
1450             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
1451             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
1452
1453             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
1454             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
1455             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
1456
1457             /**************************
1458              * CALCULATE INTERACTIONS *
1459              **************************/
1460
1461             /* COULOMB ELECTROSTATICS */
1462             velec            = _mm_mul_ps(qq32,rinv32);
1463             felec            = _mm_mul_ps(velec,rinvsq32);
1464
1465             fscal            = felec;
1466
1467              /* Update vectorial force */
1468             fix3             = _mm_macc_ps(dx32,fscal,fix3);
1469             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
1470             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
1471
1472             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
1473             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
1474             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
1475
1476             /**************************
1477              * CALCULATE INTERACTIONS *
1478              **************************/
1479
1480             /* COULOMB ELECTROSTATICS */
1481             velec            = _mm_mul_ps(qq33,rinv33);
1482             felec            = _mm_mul_ps(velec,rinvsq33);
1483
1484             fscal            = felec;
1485
1486              /* Update vectorial force */
1487             fix3             = _mm_macc_ps(dx33,fscal,fix3);
1488             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
1489             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
1490
1491             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
1492             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
1493             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
1494
1495             fjptrA             = f+j_coord_offsetA;
1496             fjptrB             = f+j_coord_offsetB;
1497             fjptrC             = f+j_coord_offsetC;
1498             fjptrD             = f+j_coord_offsetD;
1499
1500             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1501                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1502                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1503
1504             /* Inner loop uses 324 flops */
1505         }
1506
1507         if(jidx<j_index_end)
1508         {
1509
1510             /* Get j neighbor index, and coordinate index */
1511             jnrlistA         = jjnr[jidx];
1512             jnrlistB         = jjnr[jidx+1];
1513             jnrlistC         = jjnr[jidx+2];
1514             jnrlistD         = jjnr[jidx+3];
1515             /* Sign of each element will be negative for non-real atoms.
1516              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1517              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1518              */
1519             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1520             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1521             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1522             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1523             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1524             j_coord_offsetA  = DIM*jnrA;
1525             j_coord_offsetB  = DIM*jnrB;
1526             j_coord_offsetC  = DIM*jnrC;
1527             j_coord_offsetD  = DIM*jnrD;
1528
1529             /* load j atom coordinates */
1530             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1531                                               x+j_coord_offsetC,x+j_coord_offsetD,
1532                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1533                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1534
1535             /* Calculate displacement vector */
1536             dx00             = _mm_sub_ps(ix0,jx0);
1537             dy00             = _mm_sub_ps(iy0,jy0);
1538             dz00             = _mm_sub_ps(iz0,jz0);
1539             dx11             = _mm_sub_ps(ix1,jx1);
1540             dy11             = _mm_sub_ps(iy1,jy1);
1541             dz11             = _mm_sub_ps(iz1,jz1);
1542             dx12             = _mm_sub_ps(ix1,jx2);
1543             dy12             = _mm_sub_ps(iy1,jy2);
1544             dz12             = _mm_sub_ps(iz1,jz2);
1545             dx13             = _mm_sub_ps(ix1,jx3);
1546             dy13             = _mm_sub_ps(iy1,jy3);
1547             dz13             = _mm_sub_ps(iz1,jz3);
1548             dx21             = _mm_sub_ps(ix2,jx1);
1549             dy21             = _mm_sub_ps(iy2,jy1);
1550             dz21             = _mm_sub_ps(iz2,jz1);
1551             dx22             = _mm_sub_ps(ix2,jx2);
1552             dy22             = _mm_sub_ps(iy2,jy2);
1553             dz22             = _mm_sub_ps(iz2,jz2);
1554             dx23             = _mm_sub_ps(ix2,jx3);
1555             dy23             = _mm_sub_ps(iy2,jy3);
1556             dz23             = _mm_sub_ps(iz2,jz3);
1557             dx31             = _mm_sub_ps(ix3,jx1);
1558             dy31             = _mm_sub_ps(iy3,jy1);
1559             dz31             = _mm_sub_ps(iz3,jz1);
1560             dx32             = _mm_sub_ps(ix3,jx2);
1561             dy32             = _mm_sub_ps(iy3,jy2);
1562             dz32             = _mm_sub_ps(iz3,jz2);
1563             dx33             = _mm_sub_ps(ix3,jx3);
1564             dy33             = _mm_sub_ps(iy3,jy3);
1565             dz33             = _mm_sub_ps(iz3,jz3);
1566
1567             /* Calculate squared distance and things based on it */
1568             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1569             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1570             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1571             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1572             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1573             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1574             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1575             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1576             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1577             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1578
1579             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1580             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1581             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1582             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1583             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1584             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1585             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1586             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1587             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1588             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1589
1590             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1591             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1592             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1593             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1594             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1595             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1596             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1597             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1598             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1599
1600             fjx0             = _mm_setzero_ps();
1601             fjy0             = _mm_setzero_ps();
1602             fjz0             = _mm_setzero_ps();
1603             fjx1             = _mm_setzero_ps();
1604             fjy1             = _mm_setzero_ps();
1605             fjz1             = _mm_setzero_ps();
1606             fjx2             = _mm_setzero_ps();
1607             fjy2             = _mm_setzero_ps();
1608             fjz2             = _mm_setzero_ps();
1609             fjx3             = _mm_setzero_ps();
1610             fjy3             = _mm_setzero_ps();
1611             fjz3             = _mm_setzero_ps();
1612
1613             /**************************
1614              * CALCULATE INTERACTIONS *
1615              **************************/
1616
1617             r00              = _mm_mul_ps(rsq00,rinv00);
1618             r00              = _mm_andnot_ps(dummy_mask,r00);
1619
1620             /* Calculate table index by multiplying r with table scale and truncate to integer */
1621             rt               = _mm_mul_ps(r00,vftabscale);
1622             vfitab           = _mm_cvttps_epi32(rt);
1623 #ifdef __XOP__
1624             vfeps            = _mm_frcz_ps(rt);
1625 #else
1626             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1627 #endif
1628             twovfeps         = _mm_add_ps(vfeps,vfeps);
1629             vfitab           = _mm_slli_epi32(vfitab,3);
1630
1631             /* CUBIC SPLINE TABLE DISPERSION */
1632             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1633             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1634             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1635             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1636             _MM_TRANSPOSE4_PS(Y,F,G,H);
1637             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1638             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1639             fvdw6            = _mm_mul_ps(c6_00,FF);
1640
1641             /* CUBIC SPLINE TABLE REPULSION */
1642             vfitab           = _mm_add_epi32(vfitab,ifour);
1643             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1644             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1645             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1646             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1647             _MM_TRANSPOSE4_PS(Y,F,G,H);
1648             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1649             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1650             fvdw12           = _mm_mul_ps(c12_00,FF);
1651             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1652
1653             fscal            = fvdw;
1654
1655             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1656
1657              /* Update vectorial force */
1658             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1659             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1660             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1661
1662             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1663             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1664             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1665
1666             /**************************
1667              * CALCULATE INTERACTIONS *
1668              **************************/
1669
1670             /* COULOMB ELECTROSTATICS */
1671             velec            = _mm_mul_ps(qq11,rinv11);
1672             felec            = _mm_mul_ps(velec,rinvsq11);
1673
1674             fscal            = felec;
1675
1676             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1677
1678              /* Update vectorial force */
1679             fix1             = _mm_macc_ps(dx11,fscal,fix1);
1680             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
1681             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
1682
1683             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
1684             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
1685             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
1686
1687             /**************************
1688              * CALCULATE INTERACTIONS *
1689              **************************/
1690
1691             /* COULOMB ELECTROSTATICS */
1692             velec            = _mm_mul_ps(qq12,rinv12);
1693             felec            = _mm_mul_ps(velec,rinvsq12);
1694
1695             fscal            = felec;
1696
1697             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1698
1699              /* Update vectorial force */
1700             fix1             = _mm_macc_ps(dx12,fscal,fix1);
1701             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1702             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1703
1704             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1705             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1706             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1707
1708             /**************************
1709              * CALCULATE INTERACTIONS *
1710              **************************/
1711
1712             /* COULOMB ELECTROSTATICS */
1713             velec            = _mm_mul_ps(qq13,rinv13);
1714             felec            = _mm_mul_ps(velec,rinvsq13);
1715
1716             fscal            = felec;
1717
1718             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1719
1720              /* Update vectorial force */
1721             fix1             = _mm_macc_ps(dx13,fscal,fix1);
1722             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
1723             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
1724
1725             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
1726             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
1727             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
1728
1729             /**************************
1730              * CALCULATE INTERACTIONS *
1731              **************************/
1732
1733             /* COULOMB ELECTROSTATICS */
1734             velec            = _mm_mul_ps(qq21,rinv21);
1735             felec            = _mm_mul_ps(velec,rinvsq21);
1736
1737             fscal            = felec;
1738
1739             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1740
1741              /* Update vectorial force */
1742             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1743             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1744             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1745
1746             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1747             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1748             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1749
1750             /**************************
1751              * CALCULATE INTERACTIONS *
1752              **************************/
1753
1754             /* COULOMB ELECTROSTATICS */
1755             velec            = _mm_mul_ps(qq22,rinv22);
1756             felec            = _mm_mul_ps(velec,rinvsq22);
1757
1758             fscal            = felec;
1759
1760             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1761
1762              /* Update vectorial force */
1763             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1764             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1765             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1766
1767             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1768             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1769             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1770
1771             /**************************
1772              * CALCULATE INTERACTIONS *
1773              **************************/
1774
1775             /* COULOMB ELECTROSTATICS */
1776             velec            = _mm_mul_ps(qq23,rinv23);
1777             felec            = _mm_mul_ps(velec,rinvsq23);
1778
1779             fscal            = felec;
1780
1781             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1782
1783              /* Update vectorial force */
1784             fix2             = _mm_macc_ps(dx23,fscal,fix2);
1785             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
1786             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
1787
1788             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
1789             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
1790             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
1791
1792             /**************************
1793              * CALCULATE INTERACTIONS *
1794              **************************/
1795
1796             /* COULOMB ELECTROSTATICS */
1797             velec            = _mm_mul_ps(qq31,rinv31);
1798             felec            = _mm_mul_ps(velec,rinvsq31);
1799
1800             fscal            = felec;
1801
1802             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1803
1804              /* Update vectorial force */
1805             fix3             = _mm_macc_ps(dx31,fscal,fix3);
1806             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
1807             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
1808
1809             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
1810             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
1811             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
1812
1813             /**************************
1814              * CALCULATE INTERACTIONS *
1815              **************************/
1816
1817             /* COULOMB ELECTROSTATICS */
1818             velec            = _mm_mul_ps(qq32,rinv32);
1819             felec            = _mm_mul_ps(velec,rinvsq32);
1820
1821             fscal            = felec;
1822
1823             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1824
1825              /* Update vectorial force */
1826             fix3             = _mm_macc_ps(dx32,fscal,fix3);
1827             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
1828             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
1829
1830             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
1831             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
1832             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
1833
1834             /**************************
1835              * CALCULATE INTERACTIONS *
1836              **************************/
1837
1838             /* COULOMB ELECTROSTATICS */
1839             velec            = _mm_mul_ps(qq33,rinv33);
1840             felec            = _mm_mul_ps(velec,rinvsq33);
1841
1842             fscal            = felec;
1843
1844             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1845
1846              /* Update vectorial force */
1847             fix3             = _mm_macc_ps(dx33,fscal,fix3);
1848             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
1849             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
1850
1851             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
1852             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
1853             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
1854
1855             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1856             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1857             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1858             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1859
1860             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1861                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1862                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1863
1864             /* Inner loop uses 325 flops */
1865         }
1866
1867         /* End of innermost loop */
1868
1869         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1870                                               f+i_coord_offset,fshift+i_shift_offset);
1871
1872         /* Increment number of inner iterations */
1873         inneriter                  += j_index_end - j_index_start;
1874
1875         /* Outer loop uses 24 flops */
1876     }
1877
1878     /* Increment number of outer iterations */
1879     outeriter        += nri;
1880
1881     /* Update outer/inner flops */
1882
1883     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*325);
1884 }