src/gmxlib/nonbonded/nb_kernel_avx_128_fma_single/nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_avx_128_fma_single.c

   1 /*
   2  * Note: this file was generated by the Gromacs avx_128_fma_single kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_avx_128_fma_single.h"
  34 #include "kernelutil_x86_avx_128_fma_single.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_128_fma_single
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            CubicSplineTable
  40  * Geometry:                   Water4-Water4
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_VF_avx_128_fma_single
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB,jnrC,jnrD;
  61     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  62     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  63     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  64     real             rcutoff_scalar;
  65     real             *shiftvec,*fshift,*x,*f;
  66     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  67     real             scratch[4*DIM];
  68     __m128           fscal,rcutoff,rcutoff2,jidxall;
  69     int              vdwioffset0;
  70     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  71     int              vdwioffset1;
  72     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  73     int              vdwioffset2;
  74     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  75     int              vdwioffset3;
  76     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  77     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  78     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  79     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  80     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  81     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  82     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  83     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  84     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  85     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  86     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
  87     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
  88     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
  89     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
  90     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
  91     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
  92     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
  93     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
  94     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
  95     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 102     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 103     __m128i          vfitab;
 104     __m128i          ifour       = _mm_set1_epi32(4);
 105     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
 106     real             *vftab;
 107     __m128           dummy_mask,cutoff_mask;
 108     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 109     __m128           one     = _mm_set1_ps(1.0);
 110     __m128           two     = _mm_set1_ps(2.0);
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = _mm_set1_ps(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124     nvdwtype         = fr->ntype;
 125     vdwparam         = fr->nbfp;
 126     vdwtype          = mdatoms->typeA;
 127
 128     vftab            = kernel_data->table_vdw->data;
 129     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 134     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 135     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 136     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 137
 138     jq1              = _mm_set1_ps(charge[inr+1]);
 139     jq2              = _mm_set1_ps(charge[inr+2]);
 140     jq3              = _mm_set1_ps(charge[inr+3]);
 141     vdwjidx0A        = 2*vdwtype[inr+0];
 142     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
 143     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
 144     qq11             = _mm_mul_ps(iq1,jq1);
 145     qq12             = _mm_mul_ps(iq1,jq2);
 146     qq13             = _mm_mul_ps(iq1,jq3);
 147     qq21             = _mm_mul_ps(iq2,jq1);
 148     qq22             = _mm_mul_ps(iq2,jq2);
 149     qq23             = _mm_mul_ps(iq2,jq3);
 150     qq31             = _mm_mul_ps(iq3,jq1);
 151     qq32             = _mm_mul_ps(iq3,jq2);
 152     qq33             = _mm_mul_ps(iq3,jq3);
 153
 154     /* Avoid stupid compiler warnings */
 155     jnrA = jnrB = jnrC = jnrD = 0;
 156     j_coord_offsetA = 0;
 157     j_coord_offsetB = 0;
 158     j_coord_offsetC = 0;
 159     j_coord_offsetD = 0;
 160
 161     outeriter        = 0;
 162     inneriter        = 0;
 163
 164     for(iidx=0;iidx<4*DIM;iidx++)
 165     {
 166         scratch[iidx] = 0.0;
 167     }
 168
 169     /* Start outer loop over neighborlists */
 170     for(iidx=0; iidx<nri; iidx++)
 171     {
 172         /* Load shift vector for this list */
 173         i_shift_offset   = DIM*shiftidx[iidx];
 174
 175         /* Load limits for loop over neighbors */
 176         j_index_start    = jindex[iidx];
 177         j_index_end      = jindex[iidx+1];
 178
 179         /* Get outer coordinate index */
 180         inr              = iinr[iidx];
 181         i_coord_offset   = DIM*inr;
 182
 183         /* Load i particle coords and add shift vector */
 184         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 185                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 186
 187         fix0             = _mm_setzero_ps();
 188         fiy0             = _mm_setzero_ps();
 189         fiz0             = _mm_setzero_ps();
 190         fix1             = _mm_setzero_ps();
 191         fiy1             = _mm_setzero_ps();
 192         fiz1             = _mm_setzero_ps();
 193         fix2             = _mm_setzero_ps();
 194         fiy2             = _mm_setzero_ps();
 195         fiz2             = _mm_setzero_ps();
 196         fix3             = _mm_setzero_ps();
 197         fiy3             = _mm_setzero_ps();
 198         fiz3             = _mm_setzero_ps();
 199
 200         /* Reset potential sums */
 201         velecsum         = _mm_setzero_ps();
 202         vvdwsum          = _mm_setzero_ps();
 203
 204         /* Start inner kernel loop */
 205         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 206         {
 207
 208             /* Get j neighbor index, and coordinate index */
 209             jnrA             = jjnr[jidx];
 210             jnrB             = jjnr[jidx+1];
 211             jnrC             = jjnr[jidx+2];
 212             jnrD             = jjnr[jidx+3];
 213             j_coord_offsetA  = DIM*jnrA;
 214             j_coord_offsetB  = DIM*jnrB;
 215             j_coord_offsetC  = DIM*jnrC;
 216             j_coord_offsetD  = DIM*jnrD;
 217
 218             /* load j atom coordinates */
 219             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 220                                               x+j_coord_offsetC,x+j_coord_offsetD,
 221                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 222                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 223
 224             /* Calculate displacement vector */
 225             dx00             = _mm_sub_ps(ix0,jx0);
 226             dy00             = _mm_sub_ps(iy0,jy0);
 227             dz00             = _mm_sub_ps(iz0,jz0);
 228             dx11             = _mm_sub_ps(ix1,jx1);
 229             dy11             = _mm_sub_ps(iy1,jy1);
 230             dz11             = _mm_sub_ps(iz1,jz1);
 231             dx12             = _mm_sub_ps(ix1,jx2);
 232             dy12             = _mm_sub_ps(iy1,jy2);
 233             dz12             = _mm_sub_ps(iz1,jz2);
 234             dx13             = _mm_sub_ps(ix1,jx3);
 235             dy13             = _mm_sub_ps(iy1,jy3);
 236             dz13             = _mm_sub_ps(iz1,jz3);
 237             dx21             = _mm_sub_ps(ix2,jx1);
 238             dy21             = _mm_sub_ps(iy2,jy1);
 239             dz21             = _mm_sub_ps(iz2,jz1);
 240             dx22             = _mm_sub_ps(ix2,jx2);
 241             dy22             = _mm_sub_ps(iy2,jy2);
 242             dz22             = _mm_sub_ps(iz2,jz2);
 243             dx23             = _mm_sub_ps(ix2,jx3);
 244             dy23             = _mm_sub_ps(iy2,jy3);
 245             dz23             = _mm_sub_ps(iz2,jz3);
 246             dx31             = _mm_sub_ps(ix3,jx1);
 247             dy31             = _mm_sub_ps(iy3,jy1);
 248             dz31             = _mm_sub_ps(iz3,jz1);
 249             dx32             = _mm_sub_ps(ix3,jx2);
 250             dy32             = _mm_sub_ps(iy3,jy2);
 251             dz32             = _mm_sub_ps(iz3,jz2);
 252             dx33             = _mm_sub_ps(ix3,jx3);
 253             dy33             = _mm_sub_ps(iy3,jy3);
 254             dz33             = _mm_sub_ps(iz3,jz3);
 255
 256             /* Calculate squared distance and things based on it */
 257             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 258             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 259             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 260             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 261             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 262             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 263             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 264             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 265             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 266             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 267
 268             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 269             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 270             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 271             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 272             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 273             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 274             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 275             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 276             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 277             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 278
 279             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 280             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 281             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 282             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 283             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 284             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 285             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 286             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 287             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 288
 289             fjx0             = _mm_setzero_ps();
 290             fjy0             = _mm_setzero_ps();
 291             fjz0             = _mm_setzero_ps();
 292             fjx1             = _mm_setzero_ps();
 293             fjy1             = _mm_setzero_ps();
 294             fjz1             = _mm_setzero_ps();
 295             fjx2             = _mm_setzero_ps();
 296             fjy2             = _mm_setzero_ps();
 297             fjz2             = _mm_setzero_ps();
 298             fjx3             = _mm_setzero_ps();
 299             fjy3             = _mm_setzero_ps();
 300             fjz3             = _mm_setzero_ps();
 301
 302             /**************************
 303              * CALCULATE INTERACTIONS *
 304              **************************/
 305
 306             r00              = _mm_mul_ps(rsq00,rinv00);
 307
 308             /* Calculate table index by multiplying r with table scale and truncate to integer */
 309             rt               = _mm_mul_ps(r00,vftabscale);
 310             vfitab           = _mm_cvttps_epi32(rt);
 311 #ifdef __XOP__
 312             vfeps            = _mm_frcz_ps(rt);
 313 #else
 314             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 315 #endif
 316             twovfeps         = _mm_add_ps(vfeps,vfeps);
 317             vfitab           = _mm_slli_epi32(vfitab,3);
 318
 319             /* CUBIC SPLINE TABLE DISPERSION */
 320             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 321             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 322             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 323             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 324             _MM_TRANSPOSE4_PS(Y,F,G,H);
 325             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 326             VV               = _mm_macc_ps(vfeps,Fp,Y);
 327             vvdw6            = _mm_mul_ps(c6_00,VV);
 328             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 329             fvdw6            = _mm_mul_ps(c6_00,FF);
 330
 331             /* CUBIC SPLINE TABLE REPULSION */
 332             vfitab           = _mm_add_epi32(vfitab,ifour);
 333             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 334             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 335             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 336             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 337             _MM_TRANSPOSE4_PS(Y,F,G,H);
 338             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 339             VV               = _mm_macc_ps(vfeps,Fp,Y);
 340             vvdw12           = _mm_mul_ps(c12_00,VV);
 341             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 342             fvdw12           = _mm_mul_ps(c12_00,FF);
 343             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 344             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 345
 346             /* Update potential sum for this i atom from the interaction with this j atom. */
 347             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 348
 349             fscal            = fvdw;
 350
 351              /* Update vectorial force */
 352             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 353             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 354             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 355
 356             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 357             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 358             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 359
 360             /**************************
 361              * CALCULATE INTERACTIONS *
 362              **************************/
 363
 364             /* COULOMB ELECTROSTATICS */
 365             velec            = _mm_mul_ps(qq11,rinv11);
 366             felec            = _mm_mul_ps(velec,rinvsq11);
 367
 368             /* Update potential sum for this i atom from the interaction with this j atom. */
 369             velecsum         = _mm_add_ps(velecsum,velec);
 370
 371             fscal            = felec;
 372
 373              /* Update vectorial force */
 374             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 375             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 376             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 377
 378             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 379             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 380             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 381
 382             /**************************
 383              * CALCULATE INTERACTIONS *
 384              **************************/
 385
 386             /* COULOMB ELECTROSTATICS */
 387             velec            = _mm_mul_ps(qq12,rinv12);
 388             felec            = _mm_mul_ps(velec,rinvsq12);
 389
 390             /* Update potential sum for this i atom from the interaction with this j atom. */
 391             velecsum         = _mm_add_ps(velecsum,velec);
 392
 393             fscal            = felec;
 394
 395              /* Update vectorial force */
 396             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 397             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
 398             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
 399
 400             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
 401             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
 402             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
 403
 404             /**************************
 405              * CALCULATE INTERACTIONS *
 406              **************************/
 407
 408             /* COULOMB ELECTROSTATICS */
 409             velec            = _mm_mul_ps(qq13,rinv13);
 410             felec            = _mm_mul_ps(velec,rinvsq13);
 411
 412             /* Update potential sum for this i atom from the interaction with this j atom. */
 413             velecsum         = _mm_add_ps(velecsum,velec);
 414
 415             fscal            = felec;
 416
 417              /* Update vectorial force */
 418             fix1             = _mm_macc_ps(dx13,fscal,fix1);
 419             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
 420             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
 421
 422             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
 423             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
 424             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
 425
 426             /**************************
 427              * CALCULATE INTERACTIONS *
 428              **************************/
 429
 430             /* COULOMB ELECTROSTATICS */
 431             velec            = _mm_mul_ps(qq21,rinv21);
 432             felec            = _mm_mul_ps(velec,rinvsq21);
 433
 434             /* Update potential sum for this i atom from the interaction with this j atom. */
 435             velecsum         = _mm_add_ps(velecsum,velec);
 436
 437             fscal            = felec;
 438
 439              /* Update vectorial force */
 440             fix2             = _mm_macc_ps(dx21,fscal,fix2);
 441             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
 442             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
 443
 444             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
 445             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
 446             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
 447
 448             /**************************
 449              * CALCULATE INTERACTIONS *
 450              **************************/
 451
 452             /* COULOMB ELECTROSTATICS */
 453             velec            = _mm_mul_ps(qq22,rinv22);
 454             felec            = _mm_mul_ps(velec,rinvsq22);
 455
 456             /* Update potential sum for this i atom from the interaction with this j atom. */
 457             velecsum         = _mm_add_ps(velecsum,velec);
 458
 459             fscal            = felec;
 460
 461              /* Update vectorial force */
 462             fix2             = _mm_macc_ps(dx22,fscal,fix2);
 463             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
 464             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
 465
 466             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
 467             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
 468             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
 469
 470             /**************************
 471              * CALCULATE INTERACTIONS *
 472              **************************/
 473
 474             /* COULOMB ELECTROSTATICS */
 475             velec            = _mm_mul_ps(qq23,rinv23);
 476             felec            = _mm_mul_ps(velec,rinvsq23);
 477
 478             /* Update potential sum for this i atom from the interaction with this j atom. */
 479             velecsum         = _mm_add_ps(velecsum,velec);
 480
 481             fscal            = felec;
 482
 483              /* Update vectorial force */
 484             fix2             = _mm_macc_ps(dx23,fscal,fix2);
 485             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
 486             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
 487
 488             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
 489             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
 490             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
 491
 492             /**************************
 493              * CALCULATE INTERACTIONS *
 494              **************************/
 495
 496             /* COULOMB ELECTROSTATICS */
 497             velec            = _mm_mul_ps(qq31,rinv31);
 498             felec            = _mm_mul_ps(velec,rinvsq31);
 499
 500             /* Update potential sum for this i atom from the interaction with this j atom. */
 501             velecsum         = _mm_add_ps(velecsum,velec);
 502
 503             fscal            = felec;
 504
 505              /* Update vectorial force */
 506             fix3             = _mm_macc_ps(dx31,fscal,fix3);
 507             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
 508             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
 509
 510             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
 511             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
 512             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
 513
 514             /**************************
 515              * CALCULATE INTERACTIONS *
 516              **************************/
 517
 518             /* COULOMB ELECTROSTATICS */
 519             velec            = _mm_mul_ps(qq32,rinv32);
 520             felec            = _mm_mul_ps(velec,rinvsq32);
 521
 522             /* Update potential sum for this i atom from the interaction with this j atom. */
 523             velecsum         = _mm_add_ps(velecsum,velec);
 524
 525             fscal            = felec;
 526
 527              /* Update vectorial force */
 528             fix3             = _mm_macc_ps(dx32,fscal,fix3);
 529             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
 530             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
 531
 532             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
 533             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
 534             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
 535
 536             /**************************
 537              * CALCULATE INTERACTIONS *
 538              **************************/
 539
 540             /* COULOMB ELECTROSTATICS */
 541             velec            = _mm_mul_ps(qq33,rinv33);
 542             felec            = _mm_mul_ps(velec,rinvsq33);
 543
 544             /* Update potential sum for this i atom from the interaction with this j atom. */
 545             velecsum         = _mm_add_ps(velecsum,velec);
 546
 547             fscal            = felec;
 548
 549              /* Update vectorial force */
 550             fix3             = _mm_macc_ps(dx33,fscal,fix3);
 551             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
 552             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
 553
 554             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
 555             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
 556             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
 557
 558             fjptrA             = f+j_coord_offsetA;
 559             fjptrB             = f+j_coord_offsetB;
 560             fjptrC             = f+j_coord_offsetC;
 561             fjptrD             = f+j_coord_offsetD;
 562
 563             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 564                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 565                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 566
 567             /* Inner loop uses 341 flops */
 568         }
 569
 570         if(jidx<j_index_end)
 571         {
 572
 573             /* Get j neighbor index, and coordinate index */
 574             jnrlistA         = jjnr[jidx];
 575             jnrlistB         = jjnr[jidx+1];
 576             jnrlistC         = jjnr[jidx+2];
 577             jnrlistD         = jjnr[jidx+3];
 578             /* Sign of each element will be negative for non-real atoms.
 579              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 580              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 581              */
 582             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 583             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 584             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 585             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 586             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 587             j_coord_offsetA  = DIM*jnrA;
 588             j_coord_offsetB  = DIM*jnrB;
 589             j_coord_offsetC  = DIM*jnrC;
 590             j_coord_offsetD  = DIM*jnrD;
 591
 592             /* load j atom coordinates */
 593             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 594                                               x+j_coord_offsetC,x+j_coord_offsetD,
 595                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 596                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 597
 598             /* Calculate displacement vector */
 599             dx00             = _mm_sub_ps(ix0,jx0);
 600             dy00             = _mm_sub_ps(iy0,jy0);
 601             dz00             = _mm_sub_ps(iz0,jz0);
 602             dx11             = _mm_sub_ps(ix1,jx1);
 603             dy11             = _mm_sub_ps(iy1,jy1);
 604             dz11             = _mm_sub_ps(iz1,jz1);
 605             dx12             = _mm_sub_ps(ix1,jx2);
 606             dy12             = _mm_sub_ps(iy1,jy2);
 607             dz12             = _mm_sub_ps(iz1,jz2);
 608             dx13             = _mm_sub_ps(ix1,jx3);
 609             dy13             = _mm_sub_ps(iy1,jy3);
 610             dz13             = _mm_sub_ps(iz1,jz3);
 611             dx21             = _mm_sub_ps(ix2,jx1);
 612             dy21             = _mm_sub_ps(iy2,jy1);
 613             dz21             = _mm_sub_ps(iz2,jz1);
 614             dx22             = _mm_sub_ps(ix2,jx2);
 615             dy22             = _mm_sub_ps(iy2,jy2);
 616             dz22             = _mm_sub_ps(iz2,jz2);
 617             dx23             = _mm_sub_ps(ix2,jx3);
 618             dy23             = _mm_sub_ps(iy2,jy3);
 619             dz23             = _mm_sub_ps(iz2,jz3);
 620             dx31             = _mm_sub_ps(ix3,jx1);
 621             dy31             = _mm_sub_ps(iy3,jy1);
 622             dz31             = _mm_sub_ps(iz3,jz1);
 623             dx32             = _mm_sub_ps(ix3,jx2);
 624             dy32             = _mm_sub_ps(iy3,jy2);
 625             dz32             = _mm_sub_ps(iz3,jz2);
 626             dx33             = _mm_sub_ps(ix3,jx3);
 627             dy33             = _mm_sub_ps(iy3,jy3);
 628             dz33             = _mm_sub_ps(iz3,jz3);
 629
 630             /* Calculate squared distance and things based on it */
 631             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 632             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 633             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 634             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 635             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 636             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 637             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 638             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 639             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 640             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 641
 642             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 643             rinv11           = gmx_mm_invsqrt_ps(rsq11);
 644             rinv12           = gmx_mm_invsqrt_ps(rsq12);
 645             rinv13           = gmx_mm_invsqrt_ps(rsq13);
 646             rinv21           = gmx_mm_invsqrt_ps(rsq21);
 647             rinv22           = gmx_mm_invsqrt_ps(rsq22);
 648             rinv23           = gmx_mm_invsqrt_ps(rsq23);
 649             rinv31           = gmx_mm_invsqrt_ps(rsq31);
 650             rinv32           = gmx_mm_invsqrt_ps(rsq32);
 651             rinv33           = gmx_mm_invsqrt_ps(rsq33);
 652
 653             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 654             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 655             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 656             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 657             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 658             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 659             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 660             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 661             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 662
 663             fjx0             = _mm_setzero_ps();
 664             fjy0             = _mm_setzero_ps();
 665             fjz0             = _mm_setzero_ps();
 666             fjx1             = _mm_setzero_ps();
 667             fjy1             = _mm_setzero_ps();
 668             fjz1             = _mm_setzero_ps();
 669             fjx2             = _mm_setzero_ps();
 670             fjy2             = _mm_setzero_ps();
 671             fjz2             = _mm_setzero_ps();
 672             fjx3             = _mm_setzero_ps();
 673             fjy3             = _mm_setzero_ps();
 674             fjz3             = _mm_setzero_ps();
 675
 676             /**************************
 677              * CALCULATE INTERACTIONS *
 678              **************************/
 679
 680             r00              = _mm_mul_ps(rsq00,rinv00);
 681             r00              = _mm_andnot_ps(dummy_mask,r00);
 682
 683             /* Calculate table index by multiplying r with table scale and truncate to integer */
 684             rt               = _mm_mul_ps(r00,vftabscale);
 685             vfitab           = _mm_cvttps_epi32(rt);
 686 #ifdef __XOP__
 687             vfeps            = _mm_frcz_ps(rt);
 688 #else
 689             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
 690 #endif
 691             twovfeps         = _mm_add_ps(vfeps,vfeps);
 692             vfitab           = _mm_slli_epi32(vfitab,3);
 693
 694             /* CUBIC SPLINE TABLE DISPERSION */
 695             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 696             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 697             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 698             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 699             _MM_TRANSPOSE4_PS(Y,F,G,H);
 700             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 701             VV               = _mm_macc_ps(vfeps,Fp,Y);
 702             vvdw6            = _mm_mul_ps(c6_00,VV);
 703             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 704             fvdw6            = _mm_mul_ps(c6_00,FF);
 705
 706             /* CUBIC SPLINE TABLE REPULSION */
 707             vfitab           = _mm_add_epi32(vfitab,ifour);
 708             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
 709             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
 710             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
 711             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
 712             _MM_TRANSPOSE4_PS(Y,F,G,H);
 713             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
 714             VV               = _mm_macc_ps(vfeps,Fp,Y);
 715             vvdw12           = _mm_mul_ps(c12_00,VV);
 716             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
 717             fvdw12           = _mm_mul_ps(c12_00,FF);
 718             vvdw             = _mm_add_ps(vvdw12,vvdw6);
 719             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
 720
 721             /* Update potential sum for this i atom from the interaction with this j atom. */
 722             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 723             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 724
 725             fscal            = fvdw;
 726
 727             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 728
 729              /* Update vectorial force */
 730             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 731             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 732             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 733
 734             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 735             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 736             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 737
 738             /**************************
 739              * CALCULATE INTERACTIONS *
 740              **************************/
 741
 742             /* COULOMB ELECTROSTATICS */
 743             velec            = _mm_mul_ps(qq11,rinv11);
 744             felec            = _mm_mul_ps(velec,rinvsq11);
 745
 746             /* Update potential sum for this i atom from the interaction with this j atom. */
 747             velec            = _mm_andnot_ps(dummy_mask,velec);
 748             velecsum         = _mm_add_ps(velecsum,velec);
 749
 750             fscal            = felec;
 751
 752             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 753
 754              /* Update vectorial force */
 755             fix1             = _mm_macc_ps(dx11,fscal,fix1);
 756             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
 757             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
 758
 759             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
 760             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
 761             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
 762
 763             /**************************
 764              * CALCULATE INTERACTIONS *
 765              **************************/
 766
 767             /* COULOMB ELECTROSTATICS */
 768             velec            = _mm_mul_ps(qq12,rinv12);
 769             felec            = _mm_mul_ps(velec,rinvsq12);
 770
 771             /* Update potential sum for this i atom from the interaction with this j atom. */
 772             velec            = _mm_andnot_ps(dummy_mask,velec);
 773             velecsum         = _mm_add_ps(velecsum,velec);
 774
 775             fscal            = felec;
 776
 777             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 778
 779              /* Update vectorial force */
 780             fix1             = _mm_macc_ps(dx12,fscal,fix1);
 781             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
 782             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
 783
 784             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
 785             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
 786             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
 787
 788             /**************************
 789              * CALCULATE INTERACTIONS *
 790              **************************/
 791
 792             /* COULOMB ELECTROSTATICS */
 793             velec            = _mm_mul_ps(qq13,rinv13);
 794             felec            = _mm_mul_ps(velec,rinvsq13);
 795
 796             /* Update potential sum for this i atom from the interaction with this j atom. */
 797             velec            = _mm_andnot_ps(dummy_mask,velec);
 798             velecsum         = _mm_add_ps(velecsum,velec);
 799
 800             fscal            = felec;
 801
 802             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 803
 804              /* Update vectorial force */
 805             fix1             = _mm_macc_ps(dx13,fscal,fix1);
 806             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
 807             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
 808
 809             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
 810             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
 811             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
 812
 813             /**************************
 814              * CALCULATE INTERACTIONS *
 815              **************************/
 816
 817             /* COULOMB ELECTROSTATICS */
 818             velec            = _mm_mul_ps(qq21,rinv21);
 819             felec            = _mm_mul_ps(velec,rinvsq21);
 820
 821             /* Update potential sum for this i atom from the interaction with this j atom. */
 822             velec            = _mm_andnot_ps(dummy_mask,velec);
 823             velecsum         = _mm_add_ps(velecsum,velec);
 824
 825             fscal            = felec;
 826
 827             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 828
 829              /* Update vectorial force */
 830             fix2             = _mm_macc_ps(dx21,fscal,fix2);
 831             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
 832             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
 833
 834             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
 835             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
 836             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
 837
 838             /**************************
 839              * CALCULATE INTERACTIONS *
 840              **************************/
 841
 842             /* COULOMB ELECTROSTATICS */
 843             velec            = _mm_mul_ps(qq22,rinv22);
 844             felec            = _mm_mul_ps(velec,rinvsq22);
 845
 846             /* Update potential sum for this i atom from the interaction with this j atom. */
 847             velec            = _mm_andnot_ps(dummy_mask,velec);
 848             velecsum         = _mm_add_ps(velecsum,velec);
 849
 850             fscal            = felec;
 851
 852             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 853
 854              /* Update vectorial force */
 855             fix2             = _mm_macc_ps(dx22,fscal,fix2);
 856             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
 857             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
 858
 859             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
 860             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
 861             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
 862
 863             /**************************
 864              * CALCULATE INTERACTIONS *
 865              **************************/
 866
 867             /* COULOMB ELECTROSTATICS */
 868             velec            = _mm_mul_ps(qq23,rinv23);
 869             felec            = _mm_mul_ps(velec,rinvsq23);
 870
 871             /* Update potential sum for this i atom from the interaction with this j atom. */
 872             velec            = _mm_andnot_ps(dummy_mask,velec);
 873             velecsum         = _mm_add_ps(velecsum,velec);
 874
 875             fscal            = felec;
 876
 877             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 878
 879              /* Update vectorial force */
 880             fix2             = _mm_macc_ps(dx23,fscal,fix2);
 881             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
 882             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
 883
 884             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
 885             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
 886             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
 887
 888             /**************************
 889              * CALCULATE INTERACTIONS *
 890              **************************/
 891
 892             /* COULOMB ELECTROSTATICS */
 893             velec            = _mm_mul_ps(qq31,rinv31);
 894             felec            = _mm_mul_ps(velec,rinvsq31);
 895
 896             /* Update potential sum for this i atom from the interaction with this j atom. */
 897             velec            = _mm_andnot_ps(dummy_mask,velec);
 898             velecsum         = _mm_add_ps(velecsum,velec);
 899
 900             fscal            = felec;
 901
 902             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 903
 904              /* Update vectorial force */
 905             fix3             = _mm_macc_ps(dx31,fscal,fix3);
 906             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
 907             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
 908
 909             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
 910             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
 911             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
 912
 913             /**************************
 914              * CALCULATE INTERACTIONS *
 915              **************************/
 916
 917             /* COULOMB ELECTROSTATICS */
 918             velec            = _mm_mul_ps(qq32,rinv32);
 919             felec            = _mm_mul_ps(velec,rinvsq32);
 920
 921             /* Update potential sum for this i atom from the interaction with this j atom. */
 922             velec            = _mm_andnot_ps(dummy_mask,velec);
 923             velecsum         = _mm_add_ps(velecsum,velec);
 924
 925             fscal            = felec;
 926
 927             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 928
 929              /* Update vectorial force */
 930             fix3             = _mm_macc_ps(dx32,fscal,fix3);
 931             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
 932             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
 933
 934             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
 935             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
 936             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
 937
 938             /**************************
 939              * CALCULATE INTERACTIONS *
 940              **************************/
 941
 942             /* COULOMB ELECTROSTATICS */
 943             velec            = _mm_mul_ps(qq33,rinv33);
 944             felec            = _mm_mul_ps(velec,rinvsq33);
 945
 946             /* Update potential sum for this i atom from the interaction with this j atom. */
 947             velec            = _mm_andnot_ps(dummy_mask,velec);
 948             velecsum         = _mm_add_ps(velecsum,velec);
 949
 950             fscal            = felec;
 951
 952             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 953
 954              /* Update vectorial force */
 955             fix3             = _mm_macc_ps(dx33,fscal,fix3);
 956             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
 957             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
 958
 959             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
 960             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
 961             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
 962
 963             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 964             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 965             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 966             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 967
 968             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 969                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 970                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 971
 972             /* Inner loop uses 342 flops */
 973         }
 974
 975         /* End of innermost loop */
 976
 977         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 978                                               f+i_coord_offset,fshift+i_shift_offset);
 979
 980         ggid                        = gid[iidx];
 981         /* Update potential energies */
 982         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 983         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 984
 985         /* Increment number of inner iterations */
 986         inneriter                  += j_index_end - j_index_start;
 987
 988         /* Outer loop uses 26 flops */
 989     }
 990
 991     /* Increment number of outer iterations */
 992     outeriter        += nri;
 993
 994     /* Update outer/inner flops */
 995
 996     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*342);
 997 }
 998 /*
 999  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_128_fma_single
1000  * Electrostatics interaction: Coulomb
1001  * VdW interaction:            CubicSplineTable
1002  * Geometry:                   Water4-Water4
1003  * Calculate force/pot:        Force
1004  */
1005 void
1006 nb_kernel_ElecCoul_VdwCSTab_GeomW4W4_F_avx_128_fma_single
1007                     (t_nblist * gmx_restrict                nlist,
1008                      rvec * gmx_restrict                    xx,
1009                      rvec * gmx_restrict                    ff,
1010                      t_forcerec * gmx_restrict              fr,
1011                      t_mdatoms * gmx_restrict               mdatoms,
1012                      nb_kernel_data_t * gmx_restrict        kernel_data,
1013                      t_nrnb * gmx_restrict                  nrnb)
1014 {
1015     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1016      * just 0 for non-waters.
1017      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
1018      * jnr indices corresponding to data put in the four positions in the SIMD register.
1019      */
1020     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1021     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1022     int              jnrA,jnrB,jnrC,jnrD;
1023     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1024     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1025     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1026     real             rcutoff_scalar;
1027     real             *shiftvec,*fshift,*x,*f;
1028     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1029     real             scratch[4*DIM];
1030     __m128           fscal,rcutoff,rcutoff2,jidxall;
1031     int              vdwioffset0;
1032     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1033     int              vdwioffset1;
1034     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1035     int              vdwioffset2;
1036     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1037     int              vdwioffset3;
1038     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1039     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1040     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1041     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1042     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1043     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1044     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1045     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1046     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1047     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1048     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1049     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1050     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1051     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1052     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1053     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1054     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1055     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1056     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1057     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1058     real             *charge;
1059     int              nvdwtype;
1060     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1061     int              *vdwtype;
1062     real             *vdwparam;
1063     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1064     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1065     __m128i          vfitab;
1066     __m128i          ifour       = _mm_set1_epi32(4);
1067     __m128           rt,vfeps,twovfeps,vftabscale,Y,F,G,H,Fp,VV,FF;
1068     real             *vftab;
1069     __m128           dummy_mask,cutoff_mask;
1070     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1071     __m128           one     = _mm_set1_ps(1.0);
1072     __m128           two     = _mm_set1_ps(2.0);
1073     x                = xx[0];
1074     f                = ff[0];
1075
1076     nri              = nlist->nri;
1077     iinr             = nlist->iinr;
1078     jindex           = nlist->jindex;
1079     jjnr             = nlist->jjnr;
1080     shiftidx         = nlist->shift;
1081     gid              = nlist->gid;
1082     shiftvec         = fr->shift_vec[0];
1083     fshift           = fr->fshift[0];
1084     facel            = _mm_set1_ps(fr->epsfac);
1085     charge           = mdatoms->chargeA;
1086     nvdwtype         = fr->ntype;
1087     vdwparam         = fr->nbfp;
1088     vdwtype          = mdatoms->typeA;
1089
1090     vftab            = kernel_data->table_vdw->data;
1091     vftabscale       = _mm_set1_ps(kernel_data->table_vdw->scale);
1092
1093     /* Setup water-specific parameters */
1094     inr              = nlist->iinr[0];
1095     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1096     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1097     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1098     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1099
1100     jq1              = _mm_set1_ps(charge[inr+1]);
1101     jq2              = _mm_set1_ps(charge[inr+2]);
1102     jq3              = _mm_set1_ps(charge[inr+3]);
1103     vdwjidx0A        = 2*vdwtype[inr+0];
1104     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1105     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1106     qq11             = _mm_mul_ps(iq1,jq1);
1107     qq12             = _mm_mul_ps(iq1,jq2);
1108     qq13             = _mm_mul_ps(iq1,jq3);
1109     qq21             = _mm_mul_ps(iq2,jq1);
1110     qq22             = _mm_mul_ps(iq2,jq2);
1111     qq23             = _mm_mul_ps(iq2,jq3);
1112     qq31             = _mm_mul_ps(iq3,jq1);
1113     qq32             = _mm_mul_ps(iq3,jq2);
1114     qq33             = _mm_mul_ps(iq3,jq3);
1115
1116     /* Avoid stupid compiler warnings */
1117     jnrA = jnrB = jnrC = jnrD = 0;
1118     j_coord_offsetA = 0;
1119     j_coord_offsetB = 0;
1120     j_coord_offsetC = 0;
1121     j_coord_offsetD = 0;
1122
1123     outeriter        = 0;
1124     inneriter        = 0;
1125
1126     for(iidx=0;iidx<4*DIM;iidx++)
1127     {
1128         scratch[iidx] = 0.0;
1129     }
1130
1131     /* Start outer loop over neighborlists */
1132     for(iidx=0; iidx<nri; iidx++)
1133     {
1134         /* Load shift vector for this list */
1135         i_shift_offset   = DIM*shiftidx[iidx];
1136
1137         /* Load limits for loop over neighbors */
1138         j_index_start    = jindex[iidx];
1139         j_index_end      = jindex[iidx+1];
1140
1141         /* Get outer coordinate index */
1142         inr              = iinr[iidx];
1143         i_coord_offset   = DIM*inr;
1144
1145         /* Load i particle coords and add shift vector */
1146         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1147                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1148
1149         fix0             = _mm_setzero_ps();
1150         fiy0             = _mm_setzero_ps();
1151         fiz0             = _mm_setzero_ps();
1152         fix1             = _mm_setzero_ps();
1153         fiy1             = _mm_setzero_ps();
1154         fiz1             = _mm_setzero_ps();
1155         fix2             = _mm_setzero_ps();
1156         fiy2             = _mm_setzero_ps();
1157         fiz2             = _mm_setzero_ps();
1158         fix3             = _mm_setzero_ps();
1159         fiy3             = _mm_setzero_ps();
1160         fiz3             = _mm_setzero_ps();
1161
1162         /* Start inner kernel loop */
1163         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1164         {
1165
1166             /* Get j neighbor index, and coordinate index */
1167             jnrA             = jjnr[jidx];
1168             jnrB             = jjnr[jidx+1];
1169             jnrC             = jjnr[jidx+2];
1170             jnrD             = jjnr[jidx+3];
1171             j_coord_offsetA  = DIM*jnrA;
1172             j_coord_offsetB  = DIM*jnrB;
1173             j_coord_offsetC  = DIM*jnrC;
1174             j_coord_offsetD  = DIM*jnrD;
1175
1176             /* load j atom coordinates */
1177             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1178                                               x+j_coord_offsetC,x+j_coord_offsetD,
1179                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1180                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1181
1182             /* Calculate displacement vector */
1183             dx00             = _mm_sub_ps(ix0,jx0);
1184             dy00             = _mm_sub_ps(iy0,jy0);
1185             dz00             = _mm_sub_ps(iz0,jz0);
1186             dx11             = _mm_sub_ps(ix1,jx1);
1187             dy11             = _mm_sub_ps(iy1,jy1);
1188             dz11             = _mm_sub_ps(iz1,jz1);
1189             dx12             = _mm_sub_ps(ix1,jx2);
1190             dy12             = _mm_sub_ps(iy1,jy2);
1191             dz12             = _mm_sub_ps(iz1,jz2);
1192             dx13             = _mm_sub_ps(ix1,jx3);
1193             dy13             = _mm_sub_ps(iy1,jy3);
1194             dz13             = _mm_sub_ps(iz1,jz3);
1195             dx21             = _mm_sub_ps(ix2,jx1);
1196             dy21             = _mm_sub_ps(iy2,jy1);
1197             dz21             = _mm_sub_ps(iz2,jz1);
1198             dx22             = _mm_sub_ps(ix2,jx2);
1199             dy22             = _mm_sub_ps(iy2,jy2);
1200             dz22             = _mm_sub_ps(iz2,jz2);
1201             dx23             = _mm_sub_ps(ix2,jx3);
1202             dy23             = _mm_sub_ps(iy2,jy3);
1203             dz23             = _mm_sub_ps(iz2,jz3);
1204             dx31             = _mm_sub_ps(ix3,jx1);
1205             dy31             = _mm_sub_ps(iy3,jy1);
1206             dz31             = _mm_sub_ps(iz3,jz1);
1207             dx32             = _mm_sub_ps(ix3,jx2);
1208             dy32             = _mm_sub_ps(iy3,jy2);
1209             dz32             = _mm_sub_ps(iz3,jz2);
1210             dx33             = _mm_sub_ps(ix3,jx3);
1211             dy33             = _mm_sub_ps(iy3,jy3);
1212             dz33             = _mm_sub_ps(iz3,jz3);
1213
1214             /* Calculate squared distance and things based on it */
1215             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1216             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1217             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1218             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1219             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1220             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1221             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1222             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1223             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1224             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1225
1226             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1227             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1228             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1229             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1230             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1231             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1232             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1233             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1234             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1235             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1236
1237             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1238             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1239             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1240             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1241             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1242             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1243             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1244             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1245             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1246
1247             fjx0             = _mm_setzero_ps();
1248             fjy0             = _mm_setzero_ps();
1249             fjz0             = _mm_setzero_ps();
1250             fjx1             = _mm_setzero_ps();
1251             fjy1             = _mm_setzero_ps();
1252             fjz1             = _mm_setzero_ps();
1253             fjx2             = _mm_setzero_ps();
1254             fjy2             = _mm_setzero_ps();
1255             fjz2             = _mm_setzero_ps();
1256             fjx3             = _mm_setzero_ps();
1257             fjy3             = _mm_setzero_ps();
1258             fjz3             = _mm_setzero_ps();
1259
1260             /**************************
1261              * CALCULATE INTERACTIONS *
1262              **************************/
1263
1264             r00              = _mm_mul_ps(rsq00,rinv00);
1265
1266             /* Calculate table index by multiplying r with table scale and truncate to integer */
1267             rt               = _mm_mul_ps(r00,vftabscale);
1268             vfitab           = _mm_cvttps_epi32(rt);
1269 #ifdef __XOP__
1270             vfeps            = _mm_frcz_ps(rt);
1271 #else
1272             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1273 #endif
1274             twovfeps         = _mm_add_ps(vfeps,vfeps);
1275             vfitab           = _mm_slli_epi32(vfitab,3);
1276
1277             /* CUBIC SPLINE TABLE DISPERSION */
1278             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1279             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1280             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1281             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1282             _MM_TRANSPOSE4_PS(Y,F,G,H);
1283             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1284             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1285             fvdw6            = _mm_mul_ps(c6_00,FF);
1286
1287             /* CUBIC SPLINE TABLE REPULSION */
1288             vfitab           = _mm_add_epi32(vfitab,ifour);
1289             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1290             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1291             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1292             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1293             _MM_TRANSPOSE4_PS(Y,F,G,H);
1294             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1295             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1296             fvdw12           = _mm_mul_ps(c12_00,FF);
1297             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1298
1299             fscal            = fvdw;
1300
1301              /* Update vectorial force */
1302             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1303             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1304             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1305
1306             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1307             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1308             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1309
1310             /**************************
1311              * CALCULATE INTERACTIONS *
1312              **************************/
1313
1314             /* COULOMB ELECTROSTATICS */
1315             velec            = _mm_mul_ps(qq11,rinv11);
1316             felec            = _mm_mul_ps(velec,rinvsq11);
1317
1318             fscal            = felec;
1319
1320              /* Update vectorial force */
1321             fix1             = _mm_macc_ps(dx11,fscal,fix1);
1322             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
1323             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
1324
1325             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
1326             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
1327             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
1328
1329             /**************************
1330              * CALCULATE INTERACTIONS *
1331              **************************/
1332
1333             /* COULOMB ELECTROSTATICS */
1334             velec            = _mm_mul_ps(qq12,rinv12);
1335             felec            = _mm_mul_ps(velec,rinvsq12);
1336
1337             fscal            = felec;
1338
1339              /* Update vectorial force */
1340             fix1             = _mm_macc_ps(dx12,fscal,fix1);
1341             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1342             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1343
1344             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1345             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1346             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1347
1348             /**************************
1349              * CALCULATE INTERACTIONS *
1350              **************************/
1351
1352             /* COULOMB ELECTROSTATICS */
1353             velec            = _mm_mul_ps(qq13,rinv13);
1354             felec            = _mm_mul_ps(velec,rinvsq13);
1355
1356             fscal            = felec;
1357
1358              /* Update vectorial force */
1359             fix1             = _mm_macc_ps(dx13,fscal,fix1);
1360             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
1361             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
1362
1363             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
1364             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
1365             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
1366
1367             /**************************
1368              * CALCULATE INTERACTIONS *
1369              **************************/
1370
1371             /* COULOMB ELECTROSTATICS */
1372             velec            = _mm_mul_ps(qq21,rinv21);
1373             felec            = _mm_mul_ps(velec,rinvsq21);
1374
1375             fscal            = felec;
1376
1377              /* Update vectorial force */
1378             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1379             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1380             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1381
1382             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1383             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1384             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1385
1386             /**************************
1387              * CALCULATE INTERACTIONS *
1388              **************************/
1389
1390             /* COULOMB ELECTROSTATICS */
1391             velec            = _mm_mul_ps(qq22,rinv22);
1392             felec            = _mm_mul_ps(velec,rinvsq22);
1393
1394             fscal            = felec;
1395
1396              /* Update vectorial force */
1397             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1398             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1399             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1400
1401             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1402             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1403             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1404
1405             /**************************
1406              * CALCULATE INTERACTIONS *
1407              **************************/
1408
1409             /* COULOMB ELECTROSTATICS */
1410             velec            = _mm_mul_ps(qq23,rinv23);
1411             felec            = _mm_mul_ps(velec,rinvsq23);
1412
1413             fscal            = felec;
1414
1415              /* Update vectorial force */
1416             fix2             = _mm_macc_ps(dx23,fscal,fix2);
1417             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
1418             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
1419
1420             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
1421             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
1422             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
1423
1424             /**************************
1425              * CALCULATE INTERACTIONS *
1426              **************************/
1427
1428             /* COULOMB ELECTROSTATICS */
1429             velec            = _mm_mul_ps(qq31,rinv31);
1430             felec            = _mm_mul_ps(velec,rinvsq31);
1431
1432             fscal            = felec;
1433
1434              /* Update vectorial force */
1435             fix3             = _mm_macc_ps(dx31,fscal,fix3);
1436             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
1437             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
1438
1439             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
1440             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
1441             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
1442
1443             /**************************
1444              * CALCULATE INTERACTIONS *
1445              **************************/
1446
1447             /* COULOMB ELECTROSTATICS */
1448             velec            = _mm_mul_ps(qq32,rinv32);
1449             felec            = _mm_mul_ps(velec,rinvsq32);
1450
1451             fscal            = felec;
1452
1453              /* Update vectorial force */
1454             fix3             = _mm_macc_ps(dx32,fscal,fix3);
1455             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
1456             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
1457
1458             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
1459             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
1460             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
1461
1462             /**************************
1463              * CALCULATE INTERACTIONS *
1464              **************************/
1465
1466             /* COULOMB ELECTROSTATICS */
1467             velec            = _mm_mul_ps(qq33,rinv33);
1468             felec            = _mm_mul_ps(velec,rinvsq33);
1469
1470             fscal            = felec;
1471
1472              /* Update vectorial force */
1473             fix3             = _mm_macc_ps(dx33,fscal,fix3);
1474             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
1475             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
1476
1477             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
1478             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
1479             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
1480
1481             fjptrA             = f+j_coord_offsetA;
1482             fjptrB             = f+j_coord_offsetB;
1483             fjptrC             = f+j_coord_offsetC;
1484             fjptrD             = f+j_coord_offsetD;
1485
1486             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1487                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1488                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1489
1490             /* Inner loop uses 324 flops */
1491         }
1492
1493         if(jidx<j_index_end)
1494         {
1495
1496             /* Get j neighbor index, and coordinate index */
1497             jnrlistA         = jjnr[jidx];
1498             jnrlistB         = jjnr[jidx+1];
1499             jnrlistC         = jjnr[jidx+2];
1500             jnrlistD         = jjnr[jidx+3];
1501             /* Sign of each element will be negative for non-real atoms.
1502              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1503              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1504              */
1505             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1506             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1507             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1508             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1509             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1510             j_coord_offsetA  = DIM*jnrA;
1511             j_coord_offsetB  = DIM*jnrB;
1512             j_coord_offsetC  = DIM*jnrC;
1513             j_coord_offsetD  = DIM*jnrD;
1514
1515             /* load j atom coordinates */
1516             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1517                                               x+j_coord_offsetC,x+j_coord_offsetD,
1518                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1519                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1520
1521             /* Calculate displacement vector */
1522             dx00             = _mm_sub_ps(ix0,jx0);
1523             dy00             = _mm_sub_ps(iy0,jy0);
1524             dz00             = _mm_sub_ps(iz0,jz0);
1525             dx11             = _mm_sub_ps(ix1,jx1);
1526             dy11             = _mm_sub_ps(iy1,jy1);
1527             dz11             = _mm_sub_ps(iz1,jz1);
1528             dx12             = _mm_sub_ps(ix1,jx2);
1529             dy12             = _mm_sub_ps(iy1,jy2);
1530             dz12             = _mm_sub_ps(iz1,jz2);
1531             dx13             = _mm_sub_ps(ix1,jx3);
1532             dy13             = _mm_sub_ps(iy1,jy3);
1533             dz13             = _mm_sub_ps(iz1,jz3);
1534             dx21             = _mm_sub_ps(ix2,jx1);
1535             dy21             = _mm_sub_ps(iy2,jy1);
1536             dz21             = _mm_sub_ps(iz2,jz1);
1537             dx22             = _mm_sub_ps(ix2,jx2);
1538             dy22             = _mm_sub_ps(iy2,jy2);
1539             dz22             = _mm_sub_ps(iz2,jz2);
1540             dx23             = _mm_sub_ps(ix2,jx3);
1541             dy23             = _mm_sub_ps(iy2,jy3);
1542             dz23             = _mm_sub_ps(iz2,jz3);
1543             dx31             = _mm_sub_ps(ix3,jx1);
1544             dy31             = _mm_sub_ps(iy3,jy1);
1545             dz31             = _mm_sub_ps(iz3,jz1);
1546             dx32             = _mm_sub_ps(ix3,jx2);
1547             dy32             = _mm_sub_ps(iy3,jy2);
1548             dz32             = _mm_sub_ps(iz3,jz2);
1549             dx33             = _mm_sub_ps(ix3,jx3);
1550             dy33             = _mm_sub_ps(iy3,jy3);
1551             dz33             = _mm_sub_ps(iz3,jz3);
1552
1553             /* Calculate squared distance and things based on it */
1554             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1555             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1556             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1557             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1558             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1559             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1560             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1561             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1562             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1563             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1564
1565             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1566             rinv11           = gmx_mm_invsqrt_ps(rsq11);
1567             rinv12           = gmx_mm_invsqrt_ps(rsq12);
1568             rinv13           = gmx_mm_invsqrt_ps(rsq13);
1569             rinv21           = gmx_mm_invsqrt_ps(rsq21);
1570             rinv22           = gmx_mm_invsqrt_ps(rsq22);
1571             rinv23           = gmx_mm_invsqrt_ps(rsq23);
1572             rinv31           = gmx_mm_invsqrt_ps(rsq31);
1573             rinv32           = gmx_mm_invsqrt_ps(rsq32);
1574             rinv33           = gmx_mm_invsqrt_ps(rsq33);
1575
1576             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1577             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1578             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1579             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1580             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1581             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1582             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1583             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1584             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1585
1586             fjx0             = _mm_setzero_ps();
1587             fjy0             = _mm_setzero_ps();
1588             fjz0             = _mm_setzero_ps();
1589             fjx1             = _mm_setzero_ps();
1590             fjy1             = _mm_setzero_ps();
1591             fjz1             = _mm_setzero_ps();
1592             fjx2             = _mm_setzero_ps();
1593             fjy2             = _mm_setzero_ps();
1594             fjz2             = _mm_setzero_ps();
1595             fjx3             = _mm_setzero_ps();
1596             fjy3             = _mm_setzero_ps();
1597             fjz3             = _mm_setzero_ps();
1598
1599             /**************************
1600              * CALCULATE INTERACTIONS *
1601              **************************/
1602
1603             r00              = _mm_mul_ps(rsq00,rinv00);
1604             r00              = _mm_andnot_ps(dummy_mask,r00);
1605
1606             /* Calculate table index by multiplying r with table scale and truncate to integer */
1607             rt               = _mm_mul_ps(r00,vftabscale);
1608             vfitab           = _mm_cvttps_epi32(rt);
1609 #ifdef __XOP__
1610             vfeps            = _mm_frcz_ps(rt);
1611 #else
1612             vfeps            = _mm_sub_ps(rt,_mm_round_ps(rt, _MM_FROUND_FLOOR));
1613 #endif
1614             twovfeps         = _mm_add_ps(vfeps,vfeps);
1615             vfitab           = _mm_slli_epi32(vfitab,3);
1616
1617             /* CUBIC SPLINE TABLE DISPERSION */
1618             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1619             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1620             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1621             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1622             _MM_TRANSPOSE4_PS(Y,F,G,H);
1623             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1624             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1625             fvdw6            = _mm_mul_ps(c6_00,FF);
1626
1627             /* CUBIC SPLINE TABLE REPULSION */
1628             vfitab           = _mm_add_epi32(vfitab,ifour);
1629             Y                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,0) );
1630             F                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,1) );
1631             G                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,2) );
1632             H                = _mm_load_ps( vftab + _mm_extract_epi32(vfitab,3) );
1633             _MM_TRANSPOSE4_PS(Y,F,G,H);
1634             Fp               = _mm_macc_ps(vfeps,_mm_macc_ps(H,vfeps,G),F);
1635             FF               = _mm_macc_ps(vfeps,_mm_macc_ps(twovfeps,H,G),Fp);
1636             fvdw12           = _mm_mul_ps(c12_00,FF);
1637             fvdw             = _mm_xor_ps(signbit,_mm_mul_ps(_mm_add_ps(fvdw6,fvdw12),_mm_mul_ps(vftabscale,rinv00)));
1638
1639             fscal            = fvdw;
1640
1641             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1642
1643              /* Update vectorial force */
1644             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1645             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1646             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1647
1648             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1649             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1650             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1651
1652             /**************************
1653              * CALCULATE INTERACTIONS *
1654              **************************/
1655
1656             /* COULOMB ELECTROSTATICS */
1657             velec            = _mm_mul_ps(qq11,rinv11);
1658             felec            = _mm_mul_ps(velec,rinvsq11);
1659
1660             fscal            = felec;
1661
1662             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1663
1664              /* Update vectorial force */
1665             fix1             = _mm_macc_ps(dx11,fscal,fix1);
1666             fiy1             = _mm_macc_ps(dy11,fscal,fiy1);
1667             fiz1             = _mm_macc_ps(dz11,fscal,fiz1);
1668
1669             fjx1             = _mm_macc_ps(dx11,fscal,fjx1);
1670             fjy1             = _mm_macc_ps(dy11,fscal,fjy1);
1671             fjz1             = _mm_macc_ps(dz11,fscal,fjz1);
1672
1673             /**************************
1674              * CALCULATE INTERACTIONS *
1675              **************************/
1676
1677             /* COULOMB ELECTROSTATICS */
1678             velec            = _mm_mul_ps(qq12,rinv12);
1679             felec            = _mm_mul_ps(velec,rinvsq12);
1680
1681             fscal            = felec;
1682
1683             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1684
1685              /* Update vectorial force */
1686             fix1             = _mm_macc_ps(dx12,fscal,fix1);
1687             fiy1             = _mm_macc_ps(dy12,fscal,fiy1);
1688             fiz1             = _mm_macc_ps(dz12,fscal,fiz1);
1689
1690             fjx2             = _mm_macc_ps(dx12,fscal,fjx2);
1691             fjy2             = _mm_macc_ps(dy12,fscal,fjy2);
1692             fjz2             = _mm_macc_ps(dz12,fscal,fjz2);
1693
1694             /**************************
1695              * CALCULATE INTERACTIONS *
1696              **************************/
1697
1698             /* COULOMB ELECTROSTATICS */
1699             velec            = _mm_mul_ps(qq13,rinv13);
1700             felec            = _mm_mul_ps(velec,rinvsq13);
1701
1702             fscal            = felec;
1703
1704             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1705
1706              /* Update vectorial force */
1707             fix1             = _mm_macc_ps(dx13,fscal,fix1);
1708             fiy1             = _mm_macc_ps(dy13,fscal,fiy1);
1709             fiz1             = _mm_macc_ps(dz13,fscal,fiz1);
1710
1711             fjx3             = _mm_macc_ps(dx13,fscal,fjx3);
1712             fjy3             = _mm_macc_ps(dy13,fscal,fjy3);
1713             fjz3             = _mm_macc_ps(dz13,fscal,fjz3);
1714
1715             /**************************
1716              * CALCULATE INTERACTIONS *
1717              **************************/
1718
1719             /* COULOMB ELECTROSTATICS */
1720             velec            = _mm_mul_ps(qq21,rinv21);
1721             felec            = _mm_mul_ps(velec,rinvsq21);
1722
1723             fscal            = felec;
1724
1725             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1726
1727              /* Update vectorial force */
1728             fix2             = _mm_macc_ps(dx21,fscal,fix2);
1729             fiy2             = _mm_macc_ps(dy21,fscal,fiy2);
1730             fiz2             = _mm_macc_ps(dz21,fscal,fiz2);
1731
1732             fjx1             = _mm_macc_ps(dx21,fscal,fjx1);
1733             fjy1             = _mm_macc_ps(dy21,fscal,fjy1);
1734             fjz1             = _mm_macc_ps(dz21,fscal,fjz1);
1735
1736             /**************************
1737              * CALCULATE INTERACTIONS *
1738              **************************/
1739
1740             /* COULOMB ELECTROSTATICS */
1741             velec            = _mm_mul_ps(qq22,rinv22);
1742             felec            = _mm_mul_ps(velec,rinvsq22);
1743
1744             fscal            = felec;
1745
1746             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1747
1748              /* Update vectorial force */
1749             fix2             = _mm_macc_ps(dx22,fscal,fix2);
1750             fiy2             = _mm_macc_ps(dy22,fscal,fiy2);
1751             fiz2             = _mm_macc_ps(dz22,fscal,fiz2);
1752
1753             fjx2             = _mm_macc_ps(dx22,fscal,fjx2);
1754             fjy2             = _mm_macc_ps(dy22,fscal,fjy2);
1755             fjz2             = _mm_macc_ps(dz22,fscal,fjz2);
1756
1757             /**************************
1758              * CALCULATE INTERACTIONS *
1759              **************************/
1760
1761             /* COULOMB ELECTROSTATICS */
1762             velec            = _mm_mul_ps(qq23,rinv23);
1763             felec            = _mm_mul_ps(velec,rinvsq23);
1764
1765             fscal            = felec;
1766
1767             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1768
1769              /* Update vectorial force */
1770             fix2             = _mm_macc_ps(dx23,fscal,fix2);
1771             fiy2             = _mm_macc_ps(dy23,fscal,fiy2);
1772             fiz2             = _mm_macc_ps(dz23,fscal,fiz2);
1773
1774             fjx3             = _mm_macc_ps(dx23,fscal,fjx3);
1775             fjy3             = _mm_macc_ps(dy23,fscal,fjy3);
1776             fjz3             = _mm_macc_ps(dz23,fscal,fjz3);
1777
1778             /**************************
1779              * CALCULATE INTERACTIONS *
1780              **************************/
1781
1782             /* COULOMB ELECTROSTATICS */
1783             velec            = _mm_mul_ps(qq31,rinv31);
1784             felec            = _mm_mul_ps(velec,rinvsq31);
1785
1786             fscal            = felec;
1787
1788             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1789
1790              /* Update vectorial force */
1791             fix3             = _mm_macc_ps(dx31,fscal,fix3);
1792             fiy3             = _mm_macc_ps(dy31,fscal,fiy3);
1793             fiz3             = _mm_macc_ps(dz31,fscal,fiz3);
1794
1795             fjx1             = _mm_macc_ps(dx31,fscal,fjx1);
1796             fjy1             = _mm_macc_ps(dy31,fscal,fjy1);
1797             fjz1             = _mm_macc_ps(dz31,fscal,fjz1);
1798
1799             /**************************
1800              * CALCULATE INTERACTIONS *
1801              **************************/
1802
1803             /* COULOMB ELECTROSTATICS */
1804             velec            = _mm_mul_ps(qq32,rinv32);
1805             felec            = _mm_mul_ps(velec,rinvsq32);
1806
1807             fscal            = felec;
1808
1809             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1810
1811              /* Update vectorial force */
1812             fix3             = _mm_macc_ps(dx32,fscal,fix3);
1813             fiy3             = _mm_macc_ps(dy32,fscal,fiy3);
1814             fiz3             = _mm_macc_ps(dz32,fscal,fiz3);
1815
1816             fjx2             = _mm_macc_ps(dx32,fscal,fjx2);
1817             fjy2             = _mm_macc_ps(dy32,fscal,fjy2);
1818             fjz2             = _mm_macc_ps(dz32,fscal,fjz2);
1819
1820             /**************************
1821              * CALCULATE INTERACTIONS *
1822              **************************/
1823
1824             /* COULOMB ELECTROSTATICS */
1825             velec            = _mm_mul_ps(qq33,rinv33);
1826             felec            = _mm_mul_ps(velec,rinvsq33);
1827
1828             fscal            = felec;
1829
1830             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1831
1832              /* Update vectorial force */
1833             fix3             = _mm_macc_ps(dx33,fscal,fix3);
1834             fiy3             = _mm_macc_ps(dy33,fscal,fiy3);
1835             fiz3             = _mm_macc_ps(dz33,fscal,fiz3);
1836
1837             fjx3             = _mm_macc_ps(dx33,fscal,fjx3);
1838             fjy3             = _mm_macc_ps(dy33,fscal,fjy3);
1839             fjz3             = _mm_macc_ps(dz33,fscal,fjz3);
1840
1841             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1842             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1843             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1844             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1845
1846             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1847                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1848                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1849
1850             /* Inner loop uses 325 flops */
1851         }
1852
1853         /* End of innermost loop */
1854
1855         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1856                                               f+i_coord_offset,fshift+i_shift_offset);
1857
1858         /* Increment number of inner iterations */
1859         inneriter                  += j_index_end - j_index_start;
1860
1861         /* Outer loop uses 24 flops */
1862     }
1863
1864     /* Increment number of outer iterations */
1865     outeriter        += nri;
1866
1867     /* Update outer/inner flops */
1868
1869     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*325);
1870 }