src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_single/nb_kernel_ElecEw_VdwLJ_GeomW4W4_sse2_single.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017, 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 sse2_single kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_x86_sse2_single.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_sse2_single
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water4-Water4
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEw_VdwLJ_GeomW4W4_VF_sse2_single
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB,jnrC,jnrD;
  74     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  75     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  80     real             scratch[4*DIM];
  81     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwioffset3;
  89     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  90     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  91     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
  93     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  94     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
  95     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  96     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
  97     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  98     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 100     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 101     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 102     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 103     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 104     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 105     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 106     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 107     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 108     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 109     real             *charge;
 110     int              nvdwtype;
 111     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 112     int              *vdwtype;
 113     real             *vdwparam;
 114     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 115     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 116     __m128i          ewitab;
 117     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 118     real             *ewtab;
 119     __m128           dummy_mask,cutoff_mask;
 120     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 121     __m128           one     = _mm_set1_ps(1.0);
 122     __m128           two     = _mm_set1_ps(2.0);
 123     x                = xx[0];
 124     f                = ff[0];
 125
 126     nri              = nlist->nri;
 127     iinr             = nlist->iinr;
 128     jindex           = nlist->jindex;
 129     jjnr             = nlist->jjnr;
 130     shiftidx         = nlist->shift;
 131     gid              = nlist->gid;
 132     shiftvec         = fr->shift_vec[0];
 133     fshift           = fr->fshift[0];
 134     facel            = _mm_set1_ps(fr->ic->epsfac);
 135     charge           = mdatoms->chargeA;
 136     nvdwtype         = fr->ntype;
 137     vdwparam         = fr->nbfp;
 138     vdwtype          = mdatoms->typeA;
 139
 140     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 141     ewtab            = fr->ic->tabq_coul_FDV0;
 142     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 143     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_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             rinv11           = sse2_invsqrt_f(rsq11);
 283             rinv12           = sse2_invsqrt_f(rsq12);
 284             rinv13           = sse2_invsqrt_f(rsq13);
 285             rinv21           = sse2_invsqrt_f(rsq21);
 286             rinv22           = sse2_invsqrt_f(rsq22);
 287             rinv23           = sse2_invsqrt_f(rsq23);
 288             rinv31           = sse2_invsqrt_f(rsq31);
 289             rinv32           = sse2_invsqrt_f(rsq32);
 290             rinv33           = sse2_invsqrt_f(rsq33);
 291
 292             rinvsq00         = sse2_inv_f(rsq00);
 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             /* LENNARD-JONES DISPERSION/REPULSION */
 321
 322             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 323             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 324             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 325             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 326             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 327
 328             /* Update potential sum for this i atom from the interaction with this j atom. */
 329             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 330
 331             fscal            = fvdw;
 332
 333             /* Calculate temporary vectorial force */
 334             tx               = _mm_mul_ps(fscal,dx00);
 335             ty               = _mm_mul_ps(fscal,dy00);
 336             tz               = _mm_mul_ps(fscal,dz00);
 337
 338             /* Update vectorial force */
 339             fix0             = _mm_add_ps(fix0,tx);
 340             fiy0             = _mm_add_ps(fiy0,ty);
 341             fiz0             = _mm_add_ps(fiz0,tz);
 342
 343             fjx0             = _mm_add_ps(fjx0,tx);
 344             fjy0             = _mm_add_ps(fjy0,ty);
 345             fjz0             = _mm_add_ps(fjz0,tz);
 346
 347             /**************************
 348              * CALCULATE INTERACTIONS *
 349              **************************/
 350
 351             r11              = _mm_mul_ps(rsq11,rinv11);
 352
 353             /* EWALD ELECTROSTATICS */
 354
 355             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 356             ewrt             = _mm_mul_ps(r11,ewtabscale);
 357             ewitab           = _mm_cvttps_epi32(ewrt);
 358             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 359             ewitab           = _mm_slli_epi32(ewitab,2);
 360             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 361             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 362             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 363             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 364             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 365             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 366             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 367             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 368             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 369
 370             /* Update potential sum for this i atom from the interaction with this j atom. */
 371             velecsum         = _mm_add_ps(velecsum,velec);
 372
 373             fscal            = felec;
 374
 375             /* Calculate temporary vectorial force */
 376             tx               = _mm_mul_ps(fscal,dx11);
 377             ty               = _mm_mul_ps(fscal,dy11);
 378             tz               = _mm_mul_ps(fscal,dz11);
 379
 380             /* Update vectorial force */
 381             fix1             = _mm_add_ps(fix1,tx);
 382             fiy1             = _mm_add_ps(fiy1,ty);
 383             fiz1             = _mm_add_ps(fiz1,tz);
 384
 385             fjx1             = _mm_add_ps(fjx1,tx);
 386             fjy1             = _mm_add_ps(fjy1,ty);
 387             fjz1             = _mm_add_ps(fjz1,tz);
 388
 389             /**************************
 390              * CALCULATE INTERACTIONS *
 391              **************************/
 392
 393             r12              = _mm_mul_ps(rsq12,rinv12);
 394
 395             /* EWALD ELECTROSTATICS */
 396
 397             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 398             ewrt             = _mm_mul_ps(r12,ewtabscale);
 399             ewitab           = _mm_cvttps_epi32(ewrt);
 400             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 401             ewitab           = _mm_slli_epi32(ewitab,2);
 402             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 403             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 404             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 405             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 406             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 407             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 408             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 409             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
 410             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
 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             /* Calculate temporary vectorial force */
 418             tx               = _mm_mul_ps(fscal,dx12);
 419             ty               = _mm_mul_ps(fscal,dy12);
 420             tz               = _mm_mul_ps(fscal,dz12);
 421
 422             /* Update vectorial force */
 423             fix1             = _mm_add_ps(fix1,tx);
 424             fiy1             = _mm_add_ps(fiy1,ty);
 425             fiz1             = _mm_add_ps(fiz1,tz);
 426
 427             fjx2             = _mm_add_ps(fjx2,tx);
 428             fjy2             = _mm_add_ps(fjy2,ty);
 429             fjz2             = _mm_add_ps(fjz2,tz);
 430
 431             /**************************
 432              * CALCULATE INTERACTIONS *
 433              **************************/
 434
 435             r13              = _mm_mul_ps(rsq13,rinv13);
 436
 437             /* EWALD ELECTROSTATICS */
 438
 439             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 440             ewrt             = _mm_mul_ps(r13,ewtabscale);
 441             ewitab           = _mm_cvttps_epi32(ewrt);
 442             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 443             ewitab           = _mm_slli_epi32(ewitab,2);
 444             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 445             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 446             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 447             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 448             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 449             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 450             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 451             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
 452             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
 453
 454             /* Update potential sum for this i atom from the interaction with this j atom. */
 455             velecsum         = _mm_add_ps(velecsum,velec);
 456
 457             fscal            = felec;
 458
 459             /* Calculate temporary vectorial force */
 460             tx               = _mm_mul_ps(fscal,dx13);
 461             ty               = _mm_mul_ps(fscal,dy13);
 462             tz               = _mm_mul_ps(fscal,dz13);
 463
 464             /* Update vectorial force */
 465             fix1             = _mm_add_ps(fix1,tx);
 466             fiy1             = _mm_add_ps(fiy1,ty);
 467             fiz1             = _mm_add_ps(fiz1,tz);
 468
 469             fjx3             = _mm_add_ps(fjx3,tx);
 470             fjy3             = _mm_add_ps(fjy3,ty);
 471             fjz3             = _mm_add_ps(fjz3,tz);
 472
 473             /**************************
 474              * CALCULATE INTERACTIONS *
 475              **************************/
 476
 477             r21              = _mm_mul_ps(rsq21,rinv21);
 478
 479             /* EWALD ELECTROSTATICS */
 480
 481             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 482             ewrt             = _mm_mul_ps(r21,ewtabscale);
 483             ewitab           = _mm_cvttps_epi32(ewrt);
 484             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 485             ewitab           = _mm_slli_epi32(ewitab,2);
 486             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 487             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 488             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 489             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 490             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 491             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 492             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 493             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
 494             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
 495
 496             /* Update potential sum for this i atom from the interaction with this j atom. */
 497             velecsum         = _mm_add_ps(velecsum,velec);
 498
 499             fscal            = felec;
 500
 501             /* Calculate temporary vectorial force */
 502             tx               = _mm_mul_ps(fscal,dx21);
 503             ty               = _mm_mul_ps(fscal,dy21);
 504             tz               = _mm_mul_ps(fscal,dz21);
 505
 506             /* Update vectorial force */
 507             fix2             = _mm_add_ps(fix2,tx);
 508             fiy2             = _mm_add_ps(fiy2,ty);
 509             fiz2             = _mm_add_ps(fiz2,tz);
 510
 511             fjx1             = _mm_add_ps(fjx1,tx);
 512             fjy1             = _mm_add_ps(fjy1,ty);
 513             fjz1             = _mm_add_ps(fjz1,tz);
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             r22              = _mm_mul_ps(rsq22,rinv22);
 520
 521             /* EWALD ELECTROSTATICS */
 522
 523             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 524             ewrt             = _mm_mul_ps(r22,ewtabscale);
 525             ewitab           = _mm_cvttps_epi32(ewrt);
 526             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 527             ewitab           = _mm_slli_epi32(ewitab,2);
 528             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 529             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 530             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 531             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 532             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 533             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 534             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 535             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
 536             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
 537
 538             /* Update potential sum for this i atom from the interaction with this j atom. */
 539             velecsum         = _mm_add_ps(velecsum,velec);
 540
 541             fscal            = felec;
 542
 543             /* Calculate temporary vectorial force */
 544             tx               = _mm_mul_ps(fscal,dx22);
 545             ty               = _mm_mul_ps(fscal,dy22);
 546             tz               = _mm_mul_ps(fscal,dz22);
 547
 548             /* Update vectorial force */
 549             fix2             = _mm_add_ps(fix2,tx);
 550             fiy2             = _mm_add_ps(fiy2,ty);
 551             fiz2             = _mm_add_ps(fiz2,tz);
 552
 553             fjx2             = _mm_add_ps(fjx2,tx);
 554             fjy2             = _mm_add_ps(fjy2,ty);
 555             fjz2             = _mm_add_ps(fjz2,tz);
 556
 557             /**************************
 558              * CALCULATE INTERACTIONS *
 559              **************************/
 560
 561             r23              = _mm_mul_ps(rsq23,rinv23);
 562
 563             /* EWALD ELECTROSTATICS */
 564
 565             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 566             ewrt             = _mm_mul_ps(r23,ewtabscale);
 567             ewitab           = _mm_cvttps_epi32(ewrt);
 568             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 569             ewitab           = _mm_slli_epi32(ewitab,2);
 570             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 571             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 572             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 573             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 574             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 575             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 576             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 577             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
 578             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
 579
 580             /* Update potential sum for this i atom from the interaction with this j atom. */
 581             velecsum         = _mm_add_ps(velecsum,velec);
 582
 583             fscal            = felec;
 584
 585             /* Calculate temporary vectorial force */
 586             tx               = _mm_mul_ps(fscal,dx23);
 587             ty               = _mm_mul_ps(fscal,dy23);
 588             tz               = _mm_mul_ps(fscal,dz23);
 589
 590             /* Update vectorial force */
 591             fix2             = _mm_add_ps(fix2,tx);
 592             fiy2             = _mm_add_ps(fiy2,ty);
 593             fiz2             = _mm_add_ps(fiz2,tz);
 594
 595             fjx3             = _mm_add_ps(fjx3,tx);
 596             fjy3             = _mm_add_ps(fjy3,ty);
 597             fjz3             = _mm_add_ps(fjz3,tz);
 598
 599             /**************************
 600              * CALCULATE INTERACTIONS *
 601              **************************/
 602
 603             r31              = _mm_mul_ps(rsq31,rinv31);
 604
 605             /* EWALD ELECTROSTATICS */
 606
 607             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 608             ewrt             = _mm_mul_ps(r31,ewtabscale);
 609             ewitab           = _mm_cvttps_epi32(ewrt);
 610             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 611             ewitab           = _mm_slli_epi32(ewitab,2);
 612             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 613             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 614             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 615             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 616             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 617             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 618             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 619             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
 620             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
 621
 622             /* Update potential sum for this i atom from the interaction with this j atom. */
 623             velecsum         = _mm_add_ps(velecsum,velec);
 624
 625             fscal            = felec;
 626
 627             /* Calculate temporary vectorial force */
 628             tx               = _mm_mul_ps(fscal,dx31);
 629             ty               = _mm_mul_ps(fscal,dy31);
 630             tz               = _mm_mul_ps(fscal,dz31);
 631
 632             /* Update vectorial force */
 633             fix3             = _mm_add_ps(fix3,tx);
 634             fiy3             = _mm_add_ps(fiy3,ty);
 635             fiz3             = _mm_add_ps(fiz3,tz);
 636
 637             fjx1             = _mm_add_ps(fjx1,tx);
 638             fjy1             = _mm_add_ps(fjy1,ty);
 639             fjz1             = _mm_add_ps(fjz1,tz);
 640
 641             /**************************
 642              * CALCULATE INTERACTIONS *
 643              **************************/
 644
 645             r32              = _mm_mul_ps(rsq32,rinv32);
 646
 647             /* EWALD ELECTROSTATICS */
 648
 649             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 650             ewrt             = _mm_mul_ps(r32,ewtabscale);
 651             ewitab           = _mm_cvttps_epi32(ewrt);
 652             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 653             ewitab           = _mm_slli_epi32(ewitab,2);
 654             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 655             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 656             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 657             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 658             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 659             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 660             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 661             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
 662             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
 663
 664             /* Update potential sum for this i atom from the interaction with this j atom. */
 665             velecsum         = _mm_add_ps(velecsum,velec);
 666
 667             fscal            = felec;
 668
 669             /* Calculate temporary vectorial force */
 670             tx               = _mm_mul_ps(fscal,dx32);
 671             ty               = _mm_mul_ps(fscal,dy32);
 672             tz               = _mm_mul_ps(fscal,dz32);
 673
 674             /* Update vectorial force */
 675             fix3             = _mm_add_ps(fix3,tx);
 676             fiy3             = _mm_add_ps(fiy3,ty);
 677             fiz3             = _mm_add_ps(fiz3,tz);
 678
 679             fjx2             = _mm_add_ps(fjx2,tx);
 680             fjy2             = _mm_add_ps(fjy2,ty);
 681             fjz2             = _mm_add_ps(fjz2,tz);
 682
 683             /**************************
 684              * CALCULATE INTERACTIONS *
 685              **************************/
 686
 687             r33              = _mm_mul_ps(rsq33,rinv33);
 688
 689             /* EWALD ELECTROSTATICS */
 690
 691             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 692             ewrt             = _mm_mul_ps(r33,ewtabscale);
 693             ewitab           = _mm_cvttps_epi32(ewrt);
 694             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 695             ewitab           = _mm_slli_epi32(ewitab,2);
 696             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 697             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 698             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 699             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 700             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 701             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 702             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 703             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
 704             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
 705
 706             /* Update potential sum for this i atom from the interaction with this j atom. */
 707             velecsum         = _mm_add_ps(velecsum,velec);
 708
 709             fscal            = felec;
 710
 711             /* Calculate temporary vectorial force */
 712             tx               = _mm_mul_ps(fscal,dx33);
 713             ty               = _mm_mul_ps(fscal,dy33);
 714             tz               = _mm_mul_ps(fscal,dz33);
 715
 716             /* Update vectorial force */
 717             fix3             = _mm_add_ps(fix3,tx);
 718             fiy3             = _mm_add_ps(fiy3,ty);
 719             fiz3             = _mm_add_ps(fiz3,tz);
 720
 721             fjx3             = _mm_add_ps(fjx3,tx);
 722             fjy3             = _mm_add_ps(fjy3,ty);
 723             fjz3             = _mm_add_ps(fjz3,tz);
 724
 725             fjptrA             = f+j_coord_offsetA;
 726             fjptrB             = f+j_coord_offsetB;
 727             fjptrC             = f+j_coord_offsetC;
 728             fjptrD             = f+j_coord_offsetD;
 729
 730             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 731                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
 732                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 733
 734             /* Inner loop uses 404 flops */
 735         }
 736
 737         if(jidx<j_index_end)
 738         {
 739
 740             /* Get j neighbor index, and coordinate index */
 741             jnrlistA         = jjnr[jidx];
 742             jnrlistB         = jjnr[jidx+1];
 743             jnrlistC         = jjnr[jidx+2];
 744             jnrlistD         = jjnr[jidx+3];
 745             /* Sign of each element will be negative for non-real atoms.
 746              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 747              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 748              */
 749             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 750             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 751             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 752             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 753             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 754             j_coord_offsetA  = DIM*jnrA;
 755             j_coord_offsetB  = DIM*jnrB;
 756             j_coord_offsetC  = DIM*jnrC;
 757             j_coord_offsetD  = DIM*jnrD;
 758
 759             /* load j atom coordinates */
 760             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 761                                               x+j_coord_offsetC,x+j_coord_offsetD,
 762                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 763                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 764
 765             /* Calculate displacement vector */
 766             dx00             = _mm_sub_ps(ix0,jx0);
 767             dy00             = _mm_sub_ps(iy0,jy0);
 768             dz00             = _mm_sub_ps(iz0,jz0);
 769             dx11             = _mm_sub_ps(ix1,jx1);
 770             dy11             = _mm_sub_ps(iy1,jy1);
 771             dz11             = _mm_sub_ps(iz1,jz1);
 772             dx12             = _mm_sub_ps(ix1,jx2);
 773             dy12             = _mm_sub_ps(iy1,jy2);
 774             dz12             = _mm_sub_ps(iz1,jz2);
 775             dx13             = _mm_sub_ps(ix1,jx3);
 776             dy13             = _mm_sub_ps(iy1,jy3);
 777             dz13             = _mm_sub_ps(iz1,jz3);
 778             dx21             = _mm_sub_ps(ix2,jx1);
 779             dy21             = _mm_sub_ps(iy2,jy1);
 780             dz21             = _mm_sub_ps(iz2,jz1);
 781             dx22             = _mm_sub_ps(ix2,jx2);
 782             dy22             = _mm_sub_ps(iy2,jy2);
 783             dz22             = _mm_sub_ps(iz2,jz2);
 784             dx23             = _mm_sub_ps(ix2,jx3);
 785             dy23             = _mm_sub_ps(iy2,jy3);
 786             dz23             = _mm_sub_ps(iz2,jz3);
 787             dx31             = _mm_sub_ps(ix3,jx1);
 788             dy31             = _mm_sub_ps(iy3,jy1);
 789             dz31             = _mm_sub_ps(iz3,jz1);
 790             dx32             = _mm_sub_ps(ix3,jx2);
 791             dy32             = _mm_sub_ps(iy3,jy2);
 792             dz32             = _mm_sub_ps(iz3,jz2);
 793             dx33             = _mm_sub_ps(ix3,jx3);
 794             dy33             = _mm_sub_ps(iy3,jy3);
 795             dz33             = _mm_sub_ps(iz3,jz3);
 796
 797             /* Calculate squared distance and things based on it */
 798             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 799             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
 800             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
 801             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
 802             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
 803             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
 804             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
 805             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
 806             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
 807             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
 808
 809             rinv11           = sse2_invsqrt_f(rsq11);
 810             rinv12           = sse2_invsqrt_f(rsq12);
 811             rinv13           = sse2_invsqrt_f(rsq13);
 812             rinv21           = sse2_invsqrt_f(rsq21);
 813             rinv22           = sse2_invsqrt_f(rsq22);
 814             rinv23           = sse2_invsqrt_f(rsq23);
 815             rinv31           = sse2_invsqrt_f(rsq31);
 816             rinv32           = sse2_invsqrt_f(rsq32);
 817             rinv33           = sse2_invsqrt_f(rsq33);
 818
 819             rinvsq00         = sse2_inv_f(rsq00);
 820             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
 821             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
 822             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
 823             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
 824             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
 825             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
 826             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
 827             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
 828             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
 829
 830             fjx0             = _mm_setzero_ps();
 831             fjy0             = _mm_setzero_ps();
 832             fjz0             = _mm_setzero_ps();
 833             fjx1             = _mm_setzero_ps();
 834             fjy1             = _mm_setzero_ps();
 835             fjz1             = _mm_setzero_ps();
 836             fjx2             = _mm_setzero_ps();
 837             fjy2             = _mm_setzero_ps();
 838             fjz2             = _mm_setzero_ps();
 839             fjx3             = _mm_setzero_ps();
 840             fjy3             = _mm_setzero_ps();
 841             fjz3             = _mm_setzero_ps();
 842
 843             /**************************
 844              * CALCULATE INTERACTIONS *
 845              **************************/
 846
 847             /* LENNARD-JONES DISPERSION/REPULSION */
 848
 849             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 850             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 851             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 852             vvdw             = _mm_sub_ps( _mm_mul_ps(vvdw12,one_twelfth) , _mm_mul_ps(vvdw6,one_sixth) );
 853             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 854
 855             /* Update potential sum for this i atom from the interaction with this j atom. */
 856             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 857             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 858
 859             fscal            = fvdw;
 860
 861             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 862
 863             /* Calculate temporary vectorial force */
 864             tx               = _mm_mul_ps(fscal,dx00);
 865             ty               = _mm_mul_ps(fscal,dy00);
 866             tz               = _mm_mul_ps(fscal,dz00);
 867
 868             /* Update vectorial force */
 869             fix0             = _mm_add_ps(fix0,tx);
 870             fiy0             = _mm_add_ps(fiy0,ty);
 871             fiz0             = _mm_add_ps(fiz0,tz);
 872
 873             fjx0             = _mm_add_ps(fjx0,tx);
 874             fjy0             = _mm_add_ps(fjy0,ty);
 875             fjz0             = _mm_add_ps(fjz0,tz);
 876
 877             /**************************
 878              * CALCULATE INTERACTIONS *
 879              **************************/
 880
 881             r11              = _mm_mul_ps(rsq11,rinv11);
 882             r11              = _mm_andnot_ps(dummy_mask,r11);
 883
 884             /* EWALD ELECTROSTATICS */
 885
 886             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 887             ewrt             = _mm_mul_ps(r11,ewtabscale);
 888             ewitab           = _mm_cvttps_epi32(ewrt);
 889             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 890             ewitab           = _mm_slli_epi32(ewitab,2);
 891             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 892             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 893             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 894             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 895             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 896             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 897             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 898             velec            = _mm_mul_ps(qq11,_mm_sub_ps(rinv11,velec));
 899             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
 900
 901             /* Update potential sum for this i atom from the interaction with this j atom. */
 902             velec            = _mm_andnot_ps(dummy_mask,velec);
 903             velecsum         = _mm_add_ps(velecsum,velec);
 904
 905             fscal            = felec;
 906
 907             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 908
 909             /* Calculate temporary vectorial force */
 910             tx               = _mm_mul_ps(fscal,dx11);
 911             ty               = _mm_mul_ps(fscal,dy11);
 912             tz               = _mm_mul_ps(fscal,dz11);
 913
 914             /* Update vectorial force */
 915             fix1             = _mm_add_ps(fix1,tx);
 916             fiy1             = _mm_add_ps(fiy1,ty);
 917             fiz1             = _mm_add_ps(fiz1,tz);
 918
 919             fjx1             = _mm_add_ps(fjx1,tx);
 920             fjy1             = _mm_add_ps(fjy1,ty);
 921             fjz1             = _mm_add_ps(fjz1,tz);
 922
 923             /**************************
 924              * CALCULATE INTERACTIONS *
 925              **************************/
 926
 927             r12              = _mm_mul_ps(rsq12,rinv12);
 928             r12              = _mm_andnot_ps(dummy_mask,r12);
 929
 930             /* EWALD ELECTROSTATICS */
 931
 932             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 933             ewrt             = _mm_mul_ps(r12,ewtabscale);
 934             ewitab           = _mm_cvttps_epi32(ewrt);
 935             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 936             ewitab           = _mm_slli_epi32(ewitab,2);
 937             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 938             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 939             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 940             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 941             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 942             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 943             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 944             velec            = _mm_mul_ps(qq12,_mm_sub_ps(rinv12,velec));
 945             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
 946
 947             /* Update potential sum for this i atom from the interaction with this j atom. */
 948             velec            = _mm_andnot_ps(dummy_mask,velec);
 949             velecsum         = _mm_add_ps(velecsum,velec);
 950
 951             fscal            = felec;
 952
 953             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 954
 955             /* Calculate temporary vectorial force */
 956             tx               = _mm_mul_ps(fscal,dx12);
 957             ty               = _mm_mul_ps(fscal,dy12);
 958             tz               = _mm_mul_ps(fscal,dz12);
 959
 960             /* Update vectorial force */
 961             fix1             = _mm_add_ps(fix1,tx);
 962             fiy1             = _mm_add_ps(fiy1,ty);
 963             fiz1             = _mm_add_ps(fiz1,tz);
 964
 965             fjx2             = _mm_add_ps(fjx2,tx);
 966             fjy2             = _mm_add_ps(fjy2,ty);
 967             fjz2             = _mm_add_ps(fjz2,tz);
 968
 969             /**************************
 970              * CALCULATE INTERACTIONS *
 971              **************************/
 972
 973             r13              = _mm_mul_ps(rsq13,rinv13);
 974             r13              = _mm_andnot_ps(dummy_mask,r13);
 975
 976             /* EWALD ELECTROSTATICS */
 977
 978             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 979             ewrt             = _mm_mul_ps(r13,ewtabscale);
 980             ewitab           = _mm_cvttps_epi32(ewrt);
 981             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
 982             ewitab           = _mm_slli_epi32(ewitab,2);
 983             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
 984             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
 985             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
 986             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
 987             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
 988             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
 989             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
 990             velec            = _mm_mul_ps(qq13,_mm_sub_ps(rinv13,velec));
 991             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
 992
 993             /* Update potential sum for this i atom from the interaction with this j atom. */
 994             velec            = _mm_andnot_ps(dummy_mask,velec);
 995             velecsum         = _mm_add_ps(velecsum,velec);
 996
 997             fscal            = felec;
 998
 999             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1000
1001             /* Calculate temporary vectorial force */
1002             tx               = _mm_mul_ps(fscal,dx13);
1003             ty               = _mm_mul_ps(fscal,dy13);
1004             tz               = _mm_mul_ps(fscal,dz13);
1005
1006             /* Update vectorial force */
1007             fix1             = _mm_add_ps(fix1,tx);
1008             fiy1             = _mm_add_ps(fiy1,ty);
1009             fiz1             = _mm_add_ps(fiz1,tz);
1010
1011             fjx3             = _mm_add_ps(fjx3,tx);
1012             fjy3             = _mm_add_ps(fjy3,ty);
1013             fjz3             = _mm_add_ps(fjz3,tz);
1014
1015             /**************************
1016              * CALCULATE INTERACTIONS *
1017              **************************/
1018
1019             r21              = _mm_mul_ps(rsq21,rinv21);
1020             r21              = _mm_andnot_ps(dummy_mask,r21);
1021
1022             /* EWALD ELECTROSTATICS */
1023
1024             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1025             ewrt             = _mm_mul_ps(r21,ewtabscale);
1026             ewitab           = _mm_cvttps_epi32(ewrt);
1027             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1028             ewitab           = _mm_slli_epi32(ewitab,2);
1029             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1030             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1031             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1032             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1033             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1034             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1035             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1036             velec            = _mm_mul_ps(qq21,_mm_sub_ps(rinv21,velec));
1037             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1038
1039             /* Update potential sum for this i atom from the interaction with this j atom. */
1040             velec            = _mm_andnot_ps(dummy_mask,velec);
1041             velecsum         = _mm_add_ps(velecsum,velec);
1042
1043             fscal            = felec;
1044
1045             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1046
1047             /* Calculate temporary vectorial force */
1048             tx               = _mm_mul_ps(fscal,dx21);
1049             ty               = _mm_mul_ps(fscal,dy21);
1050             tz               = _mm_mul_ps(fscal,dz21);
1051
1052             /* Update vectorial force */
1053             fix2             = _mm_add_ps(fix2,tx);
1054             fiy2             = _mm_add_ps(fiy2,ty);
1055             fiz2             = _mm_add_ps(fiz2,tz);
1056
1057             fjx1             = _mm_add_ps(fjx1,tx);
1058             fjy1             = _mm_add_ps(fjy1,ty);
1059             fjz1             = _mm_add_ps(fjz1,tz);
1060
1061             /**************************
1062              * CALCULATE INTERACTIONS *
1063              **************************/
1064
1065             r22              = _mm_mul_ps(rsq22,rinv22);
1066             r22              = _mm_andnot_ps(dummy_mask,r22);
1067
1068             /* EWALD ELECTROSTATICS */
1069
1070             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1071             ewrt             = _mm_mul_ps(r22,ewtabscale);
1072             ewitab           = _mm_cvttps_epi32(ewrt);
1073             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1074             ewitab           = _mm_slli_epi32(ewitab,2);
1075             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1076             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1077             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1078             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1079             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1080             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1081             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1082             velec            = _mm_mul_ps(qq22,_mm_sub_ps(rinv22,velec));
1083             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1084
1085             /* Update potential sum for this i atom from the interaction with this j atom. */
1086             velec            = _mm_andnot_ps(dummy_mask,velec);
1087             velecsum         = _mm_add_ps(velecsum,velec);
1088
1089             fscal            = felec;
1090
1091             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1092
1093             /* Calculate temporary vectorial force */
1094             tx               = _mm_mul_ps(fscal,dx22);
1095             ty               = _mm_mul_ps(fscal,dy22);
1096             tz               = _mm_mul_ps(fscal,dz22);
1097
1098             /* Update vectorial force */
1099             fix2             = _mm_add_ps(fix2,tx);
1100             fiy2             = _mm_add_ps(fiy2,ty);
1101             fiz2             = _mm_add_ps(fiz2,tz);
1102
1103             fjx2             = _mm_add_ps(fjx2,tx);
1104             fjy2             = _mm_add_ps(fjy2,ty);
1105             fjz2             = _mm_add_ps(fjz2,tz);
1106
1107             /**************************
1108              * CALCULATE INTERACTIONS *
1109              **************************/
1110
1111             r23              = _mm_mul_ps(rsq23,rinv23);
1112             r23              = _mm_andnot_ps(dummy_mask,r23);
1113
1114             /* EWALD ELECTROSTATICS */
1115
1116             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1117             ewrt             = _mm_mul_ps(r23,ewtabscale);
1118             ewitab           = _mm_cvttps_epi32(ewrt);
1119             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1120             ewitab           = _mm_slli_epi32(ewitab,2);
1121             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1122             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1123             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1124             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1125             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1126             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1127             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1128             velec            = _mm_mul_ps(qq23,_mm_sub_ps(rinv23,velec));
1129             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1130
1131             /* Update potential sum for this i atom from the interaction with this j atom. */
1132             velec            = _mm_andnot_ps(dummy_mask,velec);
1133             velecsum         = _mm_add_ps(velecsum,velec);
1134
1135             fscal            = felec;
1136
1137             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1138
1139             /* Calculate temporary vectorial force */
1140             tx               = _mm_mul_ps(fscal,dx23);
1141             ty               = _mm_mul_ps(fscal,dy23);
1142             tz               = _mm_mul_ps(fscal,dz23);
1143
1144             /* Update vectorial force */
1145             fix2             = _mm_add_ps(fix2,tx);
1146             fiy2             = _mm_add_ps(fiy2,ty);
1147             fiz2             = _mm_add_ps(fiz2,tz);
1148
1149             fjx3             = _mm_add_ps(fjx3,tx);
1150             fjy3             = _mm_add_ps(fjy3,ty);
1151             fjz3             = _mm_add_ps(fjz3,tz);
1152
1153             /**************************
1154              * CALCULATE INTERACTIONS *
1155              **************************/
1156
1157             r31              = _mm_mul_ps(rsq31,rinv31);
1158             r31              = _mm_andnot_ps(dummy_mask,r31);
1159
1160             /* EWALD ELECTROSTATICS */
1161
1162             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1163             ewrt             = _mm_mul_ps(r31,ewtabscale);
1164             ewitab           = _mm_cvttps_epi32(ewrt);
1165             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1166             ewitab           = _mm_slli_epi32(ewitab,2);
1167             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1168             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1169             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1170             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1171             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1172             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1173             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1174             velec            = _mm_mul_ps(qq31,_mm_sub_ps(rinv31,velec));
1175             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1176
1177             /* Update potential sum for this i atom from the interaction with this j atom. */
1178             velec            = _mm_andnot_ps(dummy_mask,velec);
1179             velecsum         = _mm_add_ps(velecsum,velec);
1180
1181             fscal            = felec;
1182
1183             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1184
1185             /* Calculate temporary vectorial force */
1186             tx               = _mm_mul_ps(fscal,dx31);
1187             ty               = _mm_mul_ps(fscal,dy31);
1188             tz               = _mm_mul_ps(fscal,dz31);
1189
1190             /* Update vectorial force */
1191             fix3             = _mm_add_ps(fix3,tx);
1192             fiy3             = _mm_add_ps(fiy3,ty);
1193             fiz3             = _mm_add_ps(fiz3,tz);
1194
1195             fjx1             = _mm_add_ps(fjx1,tx);
1196             fjy1             = _mm_add_ps(fjy1,ty);
1197             fjz1             = _mm_add_ps(fjz1,tz);
1198
1199             /**************************
1200              * CALCULATE INTERACTIONS *
1201              **************************/
1202
1203             r32              = _mm_mul_ps(rsq32,rinv32);
1204             r32              = _mm_andnot_ps(dummy_mask,r32);
1205
1206             /* EWALD ELECTROSTATICS */
1207
1208             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1209             ewrt             = _mm_mul_ps(r32,ewtabscale);
1210             ewitab           = _mm_cvttps_epi32(ewrt);
1211             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1212             ewitab           = _mm_slli_epi32(ewitab,2);
1213             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1214             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1215             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1216             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1217             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1218             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1219             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1220             velec            = _mm_mul_ps(qq32,_mm_sub_ps(rinv32,velec));
1221             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1222
1223             /* Update potential sum for this i atom from the interaction with this j atom. */
1224             velec            = _mm_andnot_ps(dummy_mask,velec);
1225             velecsum         = _mm_add_ps(velecsum,velec);
1226
1227             fscal            = felec;
1228
1229             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1230
1231             /* Calculate temporary vectorial force */
1232             tx               = _mm_mul_ps(fscal,dx32);
1233             ty               = _mm_mul_ps(fscal,dy32);
1234             tz               = _mm_mul_ps(fscal,dz32);
1235
1236             /* Update vectorial force */
1237             fix3             = _mm_add_ps(fix3,tx);
1238             fiy3             = _mm_add_ps(fiy3,ty);
1239             fiz3             = _mm_add_ps(fiz3,tz);
1240
1241             fjx2             = _mm_add_ps(fjx2,tx);
1242             fjy2             = _mm_add_ps(fjy2,ty);
1243             fjz2             = _mm_add_ps(fjz2,tz);
1244
1245             /**************************
1246              * CALCULATE INTERACTIONS *
1247              **************************/
1248
1249             r33              = _mm_mul_ps(rsq33,rinv33);
1250             r33              = _mm_andnot_ps(dummy_mask,r33);
1251
1252             /* EWALD ELECTROSTATICS */
1253
1254             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1255             ewrt             = _mm_mul_ps(r33,ewtabscale);
1256             ewitab           = _mm_cvttps_epi32(ewrt);
1257             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1258             ewitab           = _mm_slli_epi32(ewitab,2);
1259             ewtabF           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1260             ewtabD           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,1) );
1261             ewtabV           = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,2) );
1262             ewtabFn          = _mm_load_ps( ewtab + gmx_mm_extract_epi32(ewitab,3) );
1263             _MM_TRANSPOSE4_PS(ewtabF,ewtabD,ewtabV,ewtabFn);
1264             felec            = _mm_add_ps(ewtabF,_mm_mul_ps(eweps,ewtabD));
1265             velec            = _mm_sub_ps(ewtabV,_mm_mul_ps(_mm_mul_ps(ewtabhalfspace,eweps),_mm_add_ps(ewtabF,felec)));
1266             velec            = _mm_mul_ps(qq33,_mm_sub_ps(rinv33,velec));
1267             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1268
1269             /* Update potential sum for this i atom from the interaction with this j atom. */
1270             velec            = _mm_andnot_ps(dummy_mask,velec);
1271             velecsum         = _mm_add_ps(velecsum,velec);
1272
1273             fscal            = felec;
1274
1275             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1276
1277             /* Calculate temporary vectorial force */
1278             tx               = _mm_mul_ps(fscal,dx33);
1279             ty               = _mm_mul_ps(fscal,dy33);
1280             tz               = _mm_mul_ps(fscal,dz33);
1281
1282             /* Update vectorial force */
1283             fix3             = _mm_add_ps(fix3,tx);
1284             fiy3             = _mm_add_ps(fiy3,ty);
1285             fiz3             = _mm_add_ps(fiz3,tz);
1286
1287             fjx3             = _mm_add_ps(fjx3,tx);
1288             fjy3             = _mm_add_ps(fjy3,ty);
1289             fjz3             = _mm_add_ps(fjz3,tz);
1290
1291             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1292             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1293             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1294             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1295
1296             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1297                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1298                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1299
1300             /* Inner loop uses 413 flops */
1301         }
1302
1303         /* End of innermost loop */
1304
1305         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1306                                               f+i_coord_offset,fshift+i_shift_offset);
1307
1308         ggid                        = gid[iidx];
1309         /* Update potential energies */
1310         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
1311         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
1312
1313         /* Increment number of inner iterations */
1314         inneriter                  += j_index_end - j_index_start;
1315
1316         /* Outer loop uses 26 flops */
1317     }
1318
1319     /* Increment number of outer iterations */
1320     outeriter        += nri;
1321
1322     /* Update outer/inner flops */
1323
1324     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*413);
1325 }
1326 /*
1327  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_sse2_single
1328  * Electrostatics interaction: Ewald
1329  * VdW interaction:            LennardJones
1330  * Geometry:                   Water4-Water4
1331  * Calculate force/pot:        Force
1332  */
1333 void
1334 nb_kernel_ElecEw_VdwLJ_GeomW4W4_F_sse2_single
1335                     (t_nblist                    * gmx_restrict       nlist,
1336                      rvec                        * gmx_restrict          xx,
1337                      rvec                        * gmx_restrict          ff,
1338                      struct t_forcerec           * gmx_restrict          fr,
1339                      t_mdatoms                   * gmx_restrict     mdatoms,
1340                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1341                      t_nrnb                      * gmx_restrict        nrnb)
1342 {
1343     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1344      * just 0 for non-waters.
1345      * Suffixes A,B,C,D refer to j loop unrolling done with SSE, e.g. for the four different
1346      * jnr indices corresponding to data put in the four positions in the SIMD register.
1347      */
1348     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1349     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1350     int              jnrA,jnrB,jnrC,jnrD;
1351     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
1352     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
1353     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1354     real             rcutoff_scalar;
1355     real             *shiftvec,*fshift,*x,*f;
1356     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
1357     real             scratch[4*DIM];
1358     __m128           tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1359     int              vdwioffset0;
1360     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1361     int              vdwioffset1;
1362     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1363     int              vdwioffset2;
1364     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1365     int              vdwioffset3;
1366     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1367     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
1368     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1369     int              vdwjidx1A,vdwjidx1B,vdwjidx1C,vdwjidx1D;
1370     __m128           jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1371     int              vdwjidx2A,vdwjidx2B,vdwjidx2C,vdwjidx2D;
1372     __m128           jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1373     int              vdwjidx3A,vdwjidx3B,vdwjidx3C,vdwjidx3D;
1374     __m128           jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1375     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1376     __m128           dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1377     __m128           dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1378     __m128           dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1379     __m128           dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1380     __m128           dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1381     __m128           dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1382     __m128           dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1383     __m128           dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1384     __m128           dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1385     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
1386     real             *charge;
1387     int              nvdwtype;
1388     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1389     int              *vdwtype;
1390     real             *vdwparam;
1391     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
1392     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
1393     __m128i          ewitab;
1394     __m128           ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1395     real             *ewtab;
1396     __m128           dummy_mask,cutoff_mask;
1397     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
1398     __m128           one     = _mm_set1_ps(1.0);
1399     __m128           two     = _mm_set1_ps(2.0);
1400     x                = xx[0];
1401     f                = ff[0];
1402
1403     nri              = nlist->nri;
1404     iinr             = nlist->iinr;
1405     jindex           = nlist->jindex;
1406     jjnr             = nlist->jjnr;
1407     shiftidx         = nlist->shift;
1408     gid              = nlist->gid;
1409     shiftvec         = fr->shift_vec[0];
1410     fshift           = fr->fshift[0];
1411     facel            = _mm_set1_ps(fr->ic->epsfac);
1412     charge           = mdatoms->chargeA;
1413     nvdwtype         = fr->ntype;
1414     vdwparam         = fr->nbfp;
1415     vdwtype          = mdatoms->typeA;
1416
1417     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
1418     ewtab            = fr->ic->tabq_coul_F;
1419     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
1420     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
1421
1422     /* Setup water-specific parameters */
1423     inr              = nlist->iinr[0];
1424     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
1425     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
1426     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
1427     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1428
1429     jq1              = _mm_set1_ps(charge[inr+1]);
1430     jq2              = _mm_set1_ps(charge[inr+2]);
1431     jq3              = _mm_set1_ps(charge[inr+3]);
1432     vdwjidx0A        = 2*vdwtype[inr+0];
1433     c6_00            = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A]);
1434     c12_00           = _mm_set1_ps(vdwparam[vdwioffset0+vdwjidx0A+1]);
1435     qq11             = _mm_mul_ps(iq1,jq1);
1436     qq12             = _mm_mul_ps(iq1,jq2);
1437     qq13             = _mm_mul_ps(iq1,jq3);
1438     qq21             = _mm_mul_ps(iq2,jq1);
1439     qq22             = _mm_mul_ps(iq2,jq2);
1440     qq23             = _mm_mul_ps(iq2,jq3);
1441     qq31             = _mm_mul_ps(iq3,jq1);
1442     qq32             = _mm_mul_ps(iq3,jq2);
1443     qq33             = _mm_mul_ps(iq3,jq3);
1444
1445     /* Avoid stupid compiler warnings */
1446     jnrA = jnrB = jnrC = jnrD = 0;
1447     j_coord_offsetA = 0;
1448     j_coord_offsetB = 0;
1449     j_coord_offsetC = 0;
1450     j_coord_offsetD = 0;
1451
1452     outeriter        = 0;
1453     inneriter        = 0;
1454
1455     for(iidx=0;iidx<4*DIM;iidx++)
1456     {
1457         scratch[iidx] = 0.0;
1458     }
1459
1460     /* Start outer loop over neighborlists */
1461     for(iidx=0; iidx<nri; iidx++)
1462     {
1463         /* Load shift vector for this list */
1464         i_shift_offset   = DIM*shiftidx[iidx];
1465
1466         /* Load limits for loop over neighbors */
1467         j_index_start    = jindex[iidx];
1468         j_index_end      = jindex[iidx+1];
1469
1470         /* Get outer coordinate index */
1471         inr              = iinr[iidx];
1472         i_coord_offset   = DIM*inr;
1473
1474         /* Load i particle coords and add shift vector */
1475         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
1476                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1477
1478         fix0             = _mm_setzero_ps();
1479         fiy0             = _mm_setzero_ps();
1480         fiz0             = _mm_setzero_ps();
1481         fix1             = _mm_setzero_ps();
1482         fiy1             = _mm_setzero_ps();
1483         fiz1             = _mm_setzero_ps();
1484         fix2             = _mm_setzero_ps();
1485         fiy2             = _mm_setzero_ps();
1486         fiz2             = _mm_setzero_ps();
1487         fix3             = _mm_setzero_ps();
1488         fiy3             = _mm_setzero_ps();
1489         fiz3             = _mm_setzero_ps();
1490
1491         /* Start inner kernel loop */
1492         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1493         {
1494
1495             /* Get j neighbor index, and coordinate index */
1496             jnrA             = jjnr[jidx];
1497             jnrB             = jjnr[jidx+1];
1498             jnrC             = jjnr[jidx+2];
1499             jnrD             = jjnr[jidx+3];
1500             j_coord_offsetA  = DIM*jnrA;
1501             j_coord_offsetB  = DIM*jnrB;
1502             j_coord_offsetC  = DIM*jnrC;
1503             j_coord_offsetD  = DIM*jnrD;
1504
1505             /* load j atom coordinates */
1506             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1507                                               x+j_coord_offsetC,x+j_coord_offsetD,
1508                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1509                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1510
1511             /* Calculate displacement vector */
1512             dx00             = _mm_sub_ps(ix0,jx0);
1513             dy00             = _mm_sub_ps(iy0,jy0);
1514             dz00             = _mm_sub_ps(iz0,jz0);
1515             dx11             = _mm_sub_ps(ix1,jx1);
1516             dy11             = _mm_sub_ps(iy1,jy1);
1517             dz11             = _mm_sub_ps(iz1,jz1);
1518             dx12             = _mm_sub_ps(ix1,jx2);
1519             dy12             = _mm_sub_ps(iy1,jy2);
1520             dz12             = _mm_sub_ps(iz1,jz2);
1521             dx13             = _mm_sub_ps(ix1,jx3);
1522             dy13             = _mm_sub_ps(iy1,jy3);
1523             dz13             = _mm_sub_ps(iz1,jz3);
1524             dx21             = _mm_sub_ps(ix2,jx1);
1525             dy21             = _mm_sub_ps(iy2,jy1);
1526             dz21             = _mm_sub_ps(iz2,jz1);
1527             dx22             = _mm_sub_ps(ix2,jx2);
1528             dy22             = _mm_sub_ps(iy2,jy2);
1529             dz22             = _mm_sub_ps(iz2,jz2);
1530             dx23             = _mm_sub_ps(ix2,jx3);
1531             dy23             = _mm_sub_ps(iy2,jy3);
1532             dz23             = _mm_sub_ps(iz2,jz3);
1533             dx31             = _mm_sub_ps(ix3,jx1);
1534             dy31             = _mm_sub_ps(iy3,jy1);
1535             dz31             = _mm_sub_ps(iz3,jz1);
1536             dx32             = _mm_sub_ps(ix3,jx2);
1537             dy32             = _mm_sub_ps(iy3,jy2);
1538             dz32             = _mm_sub_ps(iz3,jz2);
1539             dx33             = _mm_sub_ps(ix3,jx3);
1540             dy33             = _mm_sub_ps(iy3,jy3);
1541             dz33             = _mm_sub_ps(iz3,jz3);
1542
1543             /* Calculate squared distance and things based on it */
1544             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1545             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1546             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1547             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1548             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1549             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1550             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
1551             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
1552             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
1553             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
1554
1555             rinv11           = sse2_invsqrt_f(rsq11);
1556             rinv12           = sse2_invsqrt_f(rsq12);
1557             rinv13           = sse2_invsqrt_f(rsq13);
1558             rinv21           = sse2_invsqrt_f(rsq21);
1559             rinv22           = sse2_invsqrt_f(rsq22);
1560             rinv23           = sse2_invsqrt_f(rsq23);
1561             rinv31           = sse2_invsqrt_f(rsq31);
1562             rinv32           = sse2_invsqrt_f(rsq32);
1563             rinv33           = sse2_invsqrt_f(rsq33);
1564
1565             rinvsq00         = sse2_inv_f(rsq00);
1566             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
1567             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
1568             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
1569             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
1570             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
1571             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
1572             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
1573             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
1574             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
1575
1576             fjx0             = _mm_setzero_ps();
1577             fjy0             = _mm_setzero_ps();
1578             fjz0             = _mm_setzero_ps();
1579             fjx1             = _mm_setzero_ps();
1580             fjy1             = _mm_setzero_ps();
1581             fjz1             = _mm_setzero_ps();
1582             fjx2             = _mm_setzero_ps();
1583             fjy2             = _mm_setzero_ps();
1584             fjz2             = _mm_setzero_ps();
1585             fjx3             = _mm_setzero_ps();
1586             fjy3             = _mm_setzero_ps();
1587             fjz3             = _mm_setzero_ps();
1588
1589             /**************************
1590              * CALCULATE INTERACTIONS *
1591              **************************/
1592
1593             /* LENNARD-JONES DISPERSION/REPULSION */
1594
1595             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1596             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
1597
1598             fscal            = fvdw;
1599
1600             /* Calculate temporary vectorial force */
1601             tx               = _mm_mul_ps(fscal,dx00);
1602             ty               = _mm_mul_ps(fscal,dy00);
1603             tz               = _mm_mul_ps(fscal,dz00);
1604
1605             /* Update vectorial force */
1606             fix0             = _mm_add_ps(fix0,tx);
1607             fiy0             = _mm_add_ps(fiy0,ty);
1608             fiz0             = _mm_add_ps(fiz0,tz);
1609
1610             fjx0             = _mm_add_ps(fjx0,tx);
1611             fjy0             = _mm_add_ps(fjy0,ty);
1612             fjz0             = _mm_add_ps(fjz0,tz);
1613
1614             /**************************
1615              * CALCULATE INTERACTIONS *
1616              **************************/
1617
1618             r11              = _mm_mul_ps(rsq11,rinv11);
1619
1620             /* EWALD ELECTROSTATICS */
1621
1622             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1623             ewrt             = _mm_mul_ps(r11,ewtabscale);
1624             ewitab           = _mm_cvttps_epi32(ewrt);
1625             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1626             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1627                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1628                                          &ewtabF,&ewtabFn);
1629             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1630             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
1631
1632             fscal            = felec;
1633
1634             /* Calculate temporary vectorial force */
1635             tx               = _mm_mul_ps(fscal,dx11);
1636             ty               = _mm_mul_ps(fscal,dy11);
1637             tz               = _mm_mul_ps(fscal,dz11);
1638
1639             /* Update vectorial force */
1640             fix1             = _mm_add_ps(fix1,tx);
1641             fiy1             = _mm_add_ps(fiy1,ty);
1642             fiz1             = _mm_add_ps(fiz1,tz);
1643
1644             fjx1             = _mm_add_ps(fjx1,tx);
1645             fjy1             = _mm_add_ps(fjy1,ty);
1646             fjz1             = _mm_add_ps(fjz1,tz);
1647
1648             /**************************
1649              * CALCULATE INTERACTIONS *
1650              **************************/
1651
1652             r12              = _mm_mul_ps(rsq12,rinv12);
1653
1654             /* EWALD ELECTROSTATICS */
1655
1656             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1657             ewrt             = _mm_mul_ps(r12,ewtabscale);
1658             ewitab           = _mm_cvttps_epi32(ewrt);
1659             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1660             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1661                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1662                                          &ewtabF,&ewtabFn);
1663             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1664             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
1665
1666             fscal            = felec;
1667
1668             /* Calculate temporary vectorial force */
1669             tx               = _mm_mul_ps(fscal,dx12);
1670             ty               = _mm_mul_ps(fscal,dy12);
1671             tz               = _mm_mul_ps(fscal,dz12);
1672
1673             /* Update vectorial force */
1674             fix1             = _mm_add_ps(fix1,tx);
1675             fiy1             = _mm_add_ps(fiy1,ty);
1676             fiz1             = _mm_add_ps(fiz1,tz);
1677
1678             fjx2             = _mm_add_ps(fjx2,tx);
1679             fjy2             = _mm_add_ps(fjy2,ty);
1680             fjz2             = _mm_add_ps(fjz2,tz);
1681
1682             /**************************
1683              * CALCULATE INTERACTIONS *
1684              **************************/
1685
1686             r13              = _mm_mul_ps(rsq13,rinv13);
1687
1688             /* EWALD ELECTROSTATICS */
1689
1690             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1691             ewrt             = _mm_mul_ps(r13,ewtabscale);
1692             ewitab           = _mm_cvttps_epi32(ewrt);
1693             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1694             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1695                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1696                                          &ewtabF,&ewtabFn);
1697             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1698             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
1699
1700             fscal            = felec;
1701
1702             /* Calculate temporary vectorial force */
1703             tx               = _mm_mul_ps(fscal,dx13);
1704             ty               = _mm_mul_ps(fscal,dy13);
1705             tz               = _mm_mul_ps(fscal,dz13);
1706
1707             /* Update vectorial force */
1708             fix1             = _mm_add_ps(fix1,tx);
1709             fiy1             = _mm_add_ps(fiy1,ty);
1710             fiz1             = _mm_add_ps(fiz1,tz);
1711
1712             fjx3             = _mm_add_ps(fjx3,tx);
1713             fjy3             = _mm_add_ps(fjy3,ty);
1714             fjz3             = _mm_add_ps(fjz3,tz);
1715
1716             /**************************
1717              * CALCULATE INTERACTIONS *
1718              **************************/
1719
1720             r21              = _mm_mul_ps(rsq21,rinv21);
1721
1722             /* EWALD ELECTROSTATICS */
1723
1724             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1725             ewrt             = _mm_mul_ps(r21,ewtabscale);
1726             ewitab           = _mm_cvttps_epi32(ewrt);
1727             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1728             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1729                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1730                                          &ewtabF,&ewtabFn);
1731             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1732             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
1733
1734             fscal            = felec;
1735
1736             /* Calculate temporary vectorial force */
1737             tx               = _mm_mul_ps(fscal,dx21);
1738             ty               = _mm_mul_ps(fscal,dy21);
1739             tz               = _mm_mul_ps(fscal,dz21);
1740
1741             /* Update vectorial force */
1742             fix2             = _mm_add_ps(fix2,tx);
1743             fiy2             = _mm_add_ps(fiy2,ty);
1744             fiz2             = _mm_add_ps(fiz2,tz);
1745
1746             fjx1             = _mm_add_ps(fjx1,tx);
1747             fjy1             = _mm_add_ps(fjy1,ty);
1748             fjz1             = _mm_add_ps(fjz1,tz);
1749
1750             /**************************
1751              * CALCULATE INTERACTIONS *
1752              **************************/
1753
1754             r22              = _mm_mul_ps(rsq22,rinv22);
1755
1756             /* EWALD ELECTROSTATICS */
1757
1758             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1759             ewrt             = _mm_mul_ps(r22,ewtabscale);
1760             ewitab           = _mm_cvttps_epi32(ewrt);
1761             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1762             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1763                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1764                                          &ewtabF,&ewtabFn);
1765             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1766             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
1767
1768             fscal            = felec;
1769
1770             /* Calculate temporary vectorial force */
1771             tx               = _mm_mul_ps(fscal,dx22);
1772             ty               = _mm_mul_ps(fscal,dy22);
1773             tz               = _mm_mul_ps(fscal,dz22);
1774
1775             /* Update vectorial force */
1776             fix2             = _mm_add_ps(fix2,tx);
1777             fiy2             = _mm_add_ps(fiy2,ty);
1778             fiz2             = _mm_add_ps(fiz2,tz);
1779
1780             fjx2             = _mm_add_ps(fjx2,tx);
1781             fjy2             = _mm_add_ps(fjy2,ty);
1782             fjz2             = _mm_add_ps(fjz2,tz);
1783
1784             /**************************
1785              * CALCULATE INTERACTIONS *
1786              **************************/
1787
1788             r23              = _mm_mul_ps(rsq23,rinv23);
1789
1790             /* EWALD ELECTROSTATICS */
1791
1792             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1793             ewrt             = _mm_mul_ps(r23,ewtabscale);
1794             ewitab           = _mm_cvttps_epi32(ewrt);
1795             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1796             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1797                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1798                                          &ewtabF,&ewtabFn);
1799             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1800             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
1801
1802             fscal            = felec;
1803
1804             /* Calculate temporary vectorial force */
1805             tx               = _mm_mul_ps(fscal,dx23);
1806             ty               = _mm_mul_ps(fscal,dy23);
1807             tz               = _mm_mul_ps(fscal,dz23);
1808
1809             /* Update vectorial force */
1810             fix2             = _mm_add_ps(fix2,tx);
1811             fiy2             = _mm_add_ps(fiy2,ty);
1812             fiz2             = _mm_add_ps(fiz2,tz);
1813
1814             fjx3             = _mm_add_ps(fjx3,tx);
1815             fjy3             = _mm_add_ps(fjy3,ty);
1816             fjz3             = _mm_add_ps(fjz3,tz);
1817
1818             /**************************
1819              * CALCULATE INTERACTIONS *
1820              **************************/
1821
1822             r31              = _mm_mul_ps(rsq31,rinv31);
1823
1824             /* EWALD ELECTROSTATICS */
1825
1826             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1827             ewrt             = _mm_mul_ps(r31,ewtabscale);
1828             ewitab           = _mm_cvttps_epi32(ewrt);
1829             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1830             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1831                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1832                                          &ewtabF,&ewtabFn);
1833             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1834             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
1835
1836             fscal            = felec;
1837
1838             /* Calculate temporary vectorial force */
1839             tx               = _mm_mul_ps(fscal,dx31);
1840             ty               = _mm_mul_ps(fscal,dy31);
1841             tz               = _mm_mul_ps(fscal,dz31);
1842
1843             /* Update vectorial force */
1844             fix3             = _mm_add_ps(fix3,tx);
1845             fiy3             = _mm_add_ps(fiy3,ty);
1846             fiz3             = _mm_add_ps(fiz3,tz);
1847
1848             fjx1             = _mm_add_ps(fjx1,tx);
1849             fjy1             = _mm_add_ps(fjy1,ty);
1850             fjz1             = _mm_add_ps(fjz1,tz);
1851
1852             /**************************
1853              * CALCULATE INTERACTIONS *
1854              **************************/
1855
1856             r32              = _mm_mul_ps(rsq32,rinv32);
1857
1858             /* EWALD ELECTROSTATICS */
1859
1860             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1861             ewrt             = _mm_mul_ps(r32,ewtabscale);
1862             ewitab           = _mm_cvttps_epi32(ewrt);
1863             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1864             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1865                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1866                                          &ewtabF,&ewtabFn);
1867             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1868             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
1869
1870             fscal            = felec;
1871
1872             /* Calculate temporary vectorial force */
1873             tx               = _mm_mul_ps(fscal,dx32);
1874             ty               = _mm_mul_ps(fscal,dy32);
1875             tz               = _mm_mul_ps(fscal,dz32);
1876
1877             /* Update vectorial force */
1878             fix3             = _mm_add_ps(fix3,tx);
1879             fiy3             = _mm_add_ps(fiy3,ty);
1880             fiz3             = _mm_add_ps(fiz3,tz);
1881
1882             fjx2             = _mm_add_ps(fjx2,tx);
1883             fjy2             = _mm_add_ps(fjy2,ty);
1884             fjz2             = _mm_add_ps(fjz2,tz);
1885
1886             /**************************
1887              * CALCULATE INTERACTIONS *
1888              **************************/
1889
1890             r33              = _mm_mul_ps(rsq33,rinv33);
1891
1892             /* EWALD ELECTROSTATICS */
1893
1894             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1895             ewrt             = _mm_mul_ps(r33,ewtabscale);
1896             ewitab           = _mm_cvttps_epi32(ewrt);
1897             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
1898             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1899                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
1900                                          &ewtabF,&ewtabFn);
1901             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
1902             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
1903
1904             fscal            = felec;
1905
1906             /* Calculate temporary vectorial force */
1907             tx               = _mm_mul_ps(fscal,dx33);
1908             ty               = _mm_mul_ps(fscal,dy33);
1909             tz               = _mm_mul_ps(fscal,dz33);
1910
1911             /* Update vectorial force */
1912             fix3             = _mm_add_ps(fix3,tx);
1913             fiy3             = _mm_add_ps(fiy3,ty);
1914             fiz3             = _mm_add_ps(fiz3,tz);
1915
1916             fjx3             = _mm_add_ps(fjx3,tx);
1917             fjy3             = _mm_add_ps(fjy3,ty);
1918             fjz3             = _mm_add_ps(fjz3,tz);
1919
1920             fjptrA             = f+j_coord_offsetA;
1921             fjptrB             = f+j_coord_offsetB;
1922             fjptrC             = f+j_coord_offsetC;
1923             fjptrD             = f+j_coord_offsetD;
1924
1925             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
1926                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
1927                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1928
1929             /* Inner loop uses 354 flops */
1930         }
1931
1932         if(jidx<j_index_end)
1933         {
1934
1935             /* Get j neighbor index, and coordinate index */
1936             jnrlistA         = jjnr[jidx];
1937             jnrlistB         = jjnr[jidx+1];
1938             jnrlistC         = jjnr[jidx+2];
1939             jnrlistD         = jjnr[jidx+3];
1940             /* Sign of each element will be negative for non-real atoms.
1941              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1942              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1943              */
1944             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1945             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1946             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1947             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1948             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1949             j_coord_offsetA  = DIM*jnrA;
1950             j_coord_offsetB  = DIM*jnrB;
1951             j_coord_offsetC  = DIM*jnrC;
1952             j_coord_offsetD  = DIM*jnrD;
1953
1954             /* load j atom coordinates */
1955             gmx_mm_load_4rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1956                                               x+j_coord_offsetC,x+j_coord_offsetD,
1957                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1958                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1959
1960             /* Calculate displacement vector */
1961             dx00             = _mm_sub_ps(ix0,jx0);
1962             dy00             = _mm_sub_ps(iy0,jy0);
1963             dz00             = _mm_sub_ps(iz0,jz0);
1964             dx11             = _mm_sub_ps(ix1,jx1);
1965             dy11             = _mm_sub_ps(iy1,jy1);
1966             dz11             = _mm_sub_ps(iz1,jz1);
1967             dx12             = _mm_sub_ps(ix1,jx2);
1968             dy12             = _mm_sub_ps(iy1,jy2);
1969             dz12             = _mm_sub_ps(iz1,jz2);
1970             dx13             = _mm_sub_ps(ix1,jx3);
1971             dy13             = _mm_sub_ps(iy1,jy3);
1972             dz13             = _mm_sub_ps(iz1,jz3);
1973             dx21             = _mm_sub_ps(ix2,jx1);
1974             dy21             = _mm_sub_ps(iy2,jy1);
1975             dz21             = _mm_sub_ps(iz2,jz1);
1976             dx22             = _mm_sub_ps(ix2,jx2);
1977             dy22             = _mm_sub_ps(iy2,jy2);
1978             dz22             = _mm_sub_ps(iz2,jz2);
1979             dx23             = _mm_sub_ps(ix2,jx3);
1980             dy23             = _mm_sub_ps(iy2,jy3);
1981             dz23             = _mm_sub_ps(iz2,jz3);
1982             dx31             = _mm_sub_ps(ix3,jx1);
1983             dy31             = _mm_sub_ps(iy3,jy1);
1984             dz31             = _mm_sub_ps(iz3,jz1);
1985             dx32             = _mm_sub_ps(ix3,jx2);
1986             dy32             = _mm_sub_ps(iy3,jy2);
1987             dz32             = _mm_sub_ps(iz3,jz2);
1988             dx33             = _mm_sub_ps(ix3,jx3);
1989             dy33             = _mm_sub_ps(iy3,jy3);
1990             dz33             = _mm_sub_ps(iz3,jz3);
1991
1992             /* Calculate squared distance and things based on it */
1993             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1994             rsq11            = gmx_mm_calc_rsq_ps(dx11,dy11,dz11);
1995             rsq12            = gmx_mm_calc_rsq_ps(dx12,dy12,dz12);
1996             rsq13            = gmx_mm_calc_rsq_ps(dx13,dy13,dz13);
1997             rsq21            = gmx_mm_calc_rsq_ps(dx21,dy21,dz21);
1998             rsq22            = gmx_mm_calc_rsq_ps(dx22,dy22,dz22);
1999             rsq23            = gmx_mm_calc_rsq_ps(dx23,dy23,dz23);
2000             rsq31            = gmx_mm_calc_rsq_ps(dx31,dy31,dz31);
2001             rsq32            = gmx_mm_calc_rsq_ps(dx32,dy32,dz32);
2002             rsq33            = gmx_mm_calc_rsq_ps(dx33,dy33,dz33);
2003
2004             rinv11           = sse2_invsqrt_f(rsq11);
2005             rinv12           = sse2_invsqrt_f(rsq12);
2006             rinv13           = sse2_invsqrt_f(rsq13);
2007             rinv21           = sse2_invsqrt_f(rsq21);
2008             rinv22           = sse2_invsqrt_f(rsq22);
2009             rinv23           = sse2_invsqrt_f(rsq23);
2010             rinv31           = sse2_invsqrt_f(rsq31);
2011             rinv32           = sse2_invsqrt_f(rsq32);
2012             rinv33           = sse2_invsqrt_f(rsq33);
2013
2014             rinvsq00         = sse2_inv_f(rsq00);
2015             rinvsq11         = _mm_mul_ps(rinv11,rinv11);
2016             rinvsq12         = _mm_mul_ps(rinv12,rinv12);
2017             rinvsq13         = _mm_mul_ps(rinv13,rinv13);
2018             rinvsq21         = _mm_mul_ps(rinv21,rinv21);
2019             rinvsq22         = _mm_mul_ps(rinv22,rinv22);
2020             rinvsq23         = _mm_mul_ps(rinv23,rinv23);
2021             rinvsq31         = _mm_mul_ps(rinv31,rinv31);
2022             rinvsq32         = _mm_mul_ps(rinv32,rinv32);
2023             rinvsq33         = _mm_mul_ps(rinv33,rinv33);
2024
2025             fjx0             = _mm_setzero_ps();
2026             fjy0             = _mm_setzero_ps();
2027             fjz0             = _mm_setzero_ps();
2028             fjx1             = _mm_setzero_ps();
2029             fjy1             = _mm_setzero_ps();
2030             fjz1             = _mm_setzero_ps();
2031             fjx2             = _mm_setzero_ps();
2032             fjy2             = _mm_setzero_ps();
2033             fjz2             = _mm_setzero_ps();
2034             fjx3             = _mm_setzero_ps();
2035             fjy3             = _mm_setzero_ps();
2036             fjz3             = _mm_setzero_ps();
2037
2038             /**************************
2039              * CALCULATE INTERACTIONS *
2040              **************************/
2041
2042             /* LENNARD-JONES DISPERSION/REPULSION */
2043
2044             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
2045             fvdw             = _mm_mul_ps(_mm_sub_ps(_mm_mul_ps(c12_00,rinvsix),c6_00),_mm_mul_ps(rinvsix,rinvsq00));
2046
2047             fscal            = fvdw;
2048
2049             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2050
2051             /* Calculate temporary vectorial force */
2052             tx               = _mm_mul_ps(fscal,dx00);
2053             ty               = _mm_mul_ps(fscal,dy00);
2054             tz               = _mm_mul_ps(fscal,dz00);
2055
2056             /* Update vectorial force */
2057             fix0             = _mm_add_ps(fix0,tx);
2058             fiy0             = _mm_add_ps(fiy0,ty);
2059             fiz0             = _mm_add_ps(fiz0,tz);
2060
2061             fjx0             = _mm_add_ps(fjx0,tx);
2062             fjy0             = _mm_add_ps(fjy0,ty);
2063             fjz0             = _mm_add_ps(fjz0,tz);
2064
2065             /**************************
2066              * CALCULATE INTERACTIONS *
2067              **************************/
2068
2069             r11              = _mm_mul_ps(rsq11,rinv11);
2070             r11              = _mm_andnot_ps(dummy_mask,r11);
2071
2072             /* EWALD ELECTROSTATICS */
2073
2074             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2075             ewrt             = _mm_mul_ps(r11,ewtabscale);
2076             ewitab           = _mm_cvttps_epi32(ewrt);
2077             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2078             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2079                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2080                                          &ewtabF,&ewtabFn);
2081             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2082             felec            = _mm_mul_ps(_mm_mul_ps(qq11,rinv11),_mm_sub_ps(rinvsq11,felec));
2083
2084             fscal            = felec;
2085
2086             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2087
2088             /* Calculate temporary vectorial force */
2089             tx               = _mm_mul_ps(fscal,dx11);
2090             ty               = _mm_mul_ps(fscal,dy11);
2091             tz               = _mm_mul_ps(fscal,dz11);
2092
2093             /* Update vectorial force */
2094             fix1             = _mm_add_ps(fix1,tx);
2095             fiy1             = _mm_add_ps(fiy1,ty);
2096             fiz1             = _mm_add_ps(fiz1,tz);
2097
2098             fjx1             = _mm_add_ps(fjx1,tx);
2099             fjy1             = _mm_add_ps(fjy1,ty);
2100             fjz1             = _mm_add_ps(fjz1,tz);
2101
2102             /**************************
2103              * CALCULATE INTERACTIONS *
2104              **************************/
2105
2106             r12              = _mm_mul_ps(rsq12,rinv12);
2107             r12              = _mm_andnot_ps(dummy_mask,r12);
2108
2109             /* EWALD ELECTROSTATICS */
2110
2111             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2112             ewrt             = _mm_mul_ps(r12,ewtabscale);
2113             ewitab           = _mm_cvttps_epi32(ewrt);
2114             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2115             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2116                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2117                                          &ewtabF,&ewtabFn);
2118             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2119             felec            = _mm_mul_ps(_mm_mul_ps(qq12,rinv12),_mm_sub_ps(rinvsq12,felec));
2120
2121             fscal            = felec;
2122
2123             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2124
2125             /* Calculate temporary vectorial force */
2126             tx               = _mm_mul_ps(fscal,dx12);
2127             ty               = _mm_mul_ps(fscal,dy12);
2128             tz               = _mm_mul_ps(fscal,dz12);
2129
2130             /* Update vectorial force */
2131             fix1             = _mm_add_ps(fix1,tx);
2132             fiy1             = _mm_add_ps(fiy1,ty);
2133             fiz1             = _mm_add_ps(fiz1,tz);
2134
2135             fjx2             = _mm_add_ps(fjx2,tx);
2136             fjy2             = _mm_add_ps(fjy2,ty);
2137             fjz2             = _mm_add_ps(fjz2,tz);
2138
2139             /**************************
2140              * CALCULATE INTERACTIONS *
2141              **************************/
2142
2143             r13              = _mm_mul_ps(rsq13,rinv13);
2144             r13              = _mm_andnot_ps(dummy_mask,r13);
2145
2146             /* EWALD ELECTROSTATICS */
2147
2148             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2149             ewrt             = _mm_mul_ps(r13,ewtabscale);
2150             ewitab           = _mm_cvttps_epi32(ewrt);
2151             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2152             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2153                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2154                                          &ewtabF,&ewtabFn);
2155             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2156             felec            = _mm_mul_ps(_mm_mul_ps(qq13,rinv13),_mm_sub_ps(rinvsq13,felec));
2157
2158             fscal            = felec;
2159
2160             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2161
2162             /* Calculate temporary vectorial force */
2163             tx               = _mm_mul_ps(fscal,dx13);
2164             ty               = _mm_mul_ps(fscal,dy13);
2165             tz               = _mm_mul_ps(fscal,dz13);
2166
2167             /* Update vectorial force */
2168             fix1             = _mm_add_ps(fix1,tx);
2169             fiy1             = _mm_add_ps(fiy1,ty);
2170             fiz1             = _mm_add_ps(fiz1,tz);
2171
2172             fjx3             = _mm_add_ps(fjx3,tx);
2173             fjy3             = _mm_add_ps(fjy3,ty);
2174             fjz3             = _mm_add_ps(fjz3,tz);
2175
2176             /**************************
2177              * CALCULATE INTERACTIONS *
2178              **************************/
2179
2180             r21              = _mm_mul_ps(rsq21,rinv21);
2181             r21              = _mm_andnot_ps(dummy_mask,r21);
2182
2183             /* EWALD ELECTROSTATICS */
2184
2185             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2186             ewrt             = _mm_mul_ps(r21,ewtabscale);
2187             ewitab           = _mm_cvttps_epi32(ewrt);
2188             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2189             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2190                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2191                                          &ewtabF,&ewtabFn);
2192             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2193             felec            = _mm_mul_ps(_mm_mul_ps(qq21,rinv21),_mm_sub_ps(rinvsq21,felec));
2194
2195             fscal            = felec;
2196
2197             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2198
2199             /* Calculate temporary vectorial force */
2200             tx               = _mm_mul_ps(fscal,dx21);
2201             ty               = _mm_mul_ps(fscal,dy21);
2202             tz               = _mm_mul_ps(fscal,dz21);
2203
2204             /* Update vectorial force */
2205             fix2             = _mm_add_ps(fix2,tx);
2206             fiy2             = _mm_add_ps(fiy2,ty);
2207             fiz2             = _mm_add_ps(fiz2,tz);
2208
2209             fjx1             = _mm_add_ps(fjx1,tx);
2210             fjy1             = _mm_add_ps(fjy1,ty);
2211             fjz1             = _mm_add_ps(fjz1,tz);
2212
2213             /**************************
2214              * CALCULATE INTERACTIONS *
2215              **************************/
2216
2217             r22              = _mm_mul_ps(rsq22,rinv22);
2218             r22              = _mm_andnot_ps(dummy_mask,r22);
2219
2220             /* EWALD ELECTROSTATICS */
2221
2222             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2223             ewrt             = _mm_mul_ps(r22,ewtabscale);
2224             ewitab           = _mm_cvttps_epi32(ewrt);
2225             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2226             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2227                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2228                                          &ewtabF,&ewtabFn);
2229             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2230             felec            = _mm_mul_ps(_mm_mul_ps(qq22,rinv22),_mm_sub_ps(rinvsq22,felec));
2231
2232             fscal            = felec;
2233
2234             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2235
2236             /* Calculate temporary vectorial force */
2237             tx               = _mm_mul_ps(fscal,dx22);
2238             ty               = _mm_mul_ps(fscal,dy22);
2239             tz               = _mm_mul_ps(fscal,dz22);
2240
2241             /* Update vectorial force */
2242             fix2             = _mm_add_ps(fix2,tx);
2243             fiy2             = _mm_add_ps(fiy2,ty);
2244             fiz2             = _mm_add_ps(fiz2,tz);
2245
2246             fjx2             = _mm_add_ps(fjx2,tx);
2247             fjy2             = _mm_add_ps(fjy2,ty);
2248             fjz2             = _mm_add_ps(fjz2,tz);
2249
2250             /**************************
2251              * CALCULATE INTERACTIONS *
2252              **************************/
2253
2254             r23              = _mm_mul_ps(rsq23,rinv23);
2255             r23              = _mm_andnot_ps(dummy_mask,r23);
2256
2257             /* EWALD ELECTROSTATICS */
2258
2259             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2260             ewrt             = _mm_mul_ps(r23,ewtabscale);
2261             ewitab           = _mm_cvttps_epi32(ewrt);
2262             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2263             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2264                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2265                                          &ewtabF,&ewtabFn);
2266             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2267             felec            = _mm_mul_ps(_mm_mul_ps(qq23,rinv23),_mm_sub_ps(rinvsq23,felec));
2268
2269             fscal            = felec;
2270
2271             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2272
2273             /* Calculate temporary vectorial force */
2274             tx               = _mm_mul_ps(fscal,dx23);
2275             ty               = _mm_mul_ps(fscal,dy23);
2276             tz               = _mm_mul_ps(fscal,dz23);
2277
2278             /* Update vectorial force */
2279             fix2             = _mm_add_ps(fix2,tx);
2280             fiy2             = _mm_add_ps(fiy2,ty);
2281             fiz2             = _mm_add_ps(fiz2,tz);
2282
2283             fjx3             = _mm_add_ps(fjx3,tx);
2284             fjy3             = _mm_add_ps(fjy3,ty);
2285             fjz3             = _mm_add_ps(fjz3,tz);
2286
2287             /**************************
2288              * CALCULATE INTERACTIONS *
2289              **************************/
2290
2291             r31              = _mm_mul_ps(rsq31,rinv31);
2292             r31              = _mm_andnot_ps(dummy_mask,r31);
2293
2294             /* EWALD ELECTROSTATICS */
2295
2296             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2297             ewrt             = _mm_mul_ps(r31,ewtabscale);
2298             ewitab           = _mm_cvttps_epi32(ewrt);
2299             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2300             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2301                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2302                                          &ewtabF,&ewtabFn);
2303             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2304             felec            = _mm_mul_ps(_mm_mul_ps(qq31,rinv31),_mm_sub_ps(rinvsq31,felec));
2305
2306             fscal            = felec;
2307
2308             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2309
2310             /* Calculate temporary vectorial force */
2311             tx               = _mm_mul_ps(fscal,dx31);
2312             ty               = _mm_mul_ps(fscal,dy31);
2313             tz               = _mm_mul_ps(fscal,dz31);
2314
2315             /* Update vectorial force */
2316             fix3             = _mm_add_ps(fix3,tx);
2317             fiy3             = _mm_add_ps(fiy3,ty);
2318             fiz3             = _mm_add_ps(fiz3,tz);
2319
2320             fjx1             = _mm_add_ps(fjx1,tx);
2321             fjy1             = _mm_add_ps(fjy1,ty);
2322             fjz1             = _mm_add_ps(fjz1,tz);
2323
2324             /**************************
2325              * CALCULATE INTERACTIONS *
2326              **************************/
2327
2328             r32              = _mm_mul_ps(rsq32,rinv32);
2329             r32              = _mm_andnot_ps(dummy_mask,r32);
2330
2331             /* EWALD ELECTROSTATICS */
2332
2333             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2334             ewrt             = _mm_mul_ps(r32,ewtabscale);
2335             ewitab           = _mm_cvttps_epi32(ewrt);
2336             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2337             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2338                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2339                                          &ewtabF,&ewtabFn);
2340             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2341             felec            = _mm_mul_ps(_mm_mul_ps(qq32,rinv32),_mm_sub_ps(rinvsq32,felec));
2342
2343             fscal            = felec;
2344
2345             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2346
2347             /* Calculate temporary vectorial force */
2348             tx               = _mm_mul_ps(fscal,dx32);
2349             ty               = _mm_mul_ps(fscal,dy32);
2350             tz               = _mm_mul_ps(fscal,dz32);
2351
2352             /* Update vectorial force */
2353             fix3             = _mm_add_ps(fix3,tx);
2354             fiy3             = _mm_add_ps(fiy3,ty);
2355             fiz3             = _mm_add_ps(fiz3,tz);
2356
2357             fjx2             = _mm_add_ps(fjx2,tx);
2358             fjy2             = _mm_add_ps(fjy2,ty);
2359             fjz2             = _mm_add_ps(fjz2,tz);
2360
2361             /**************************
2362              * CALCULATE INTERACTIONS *
2363              **************************/
2364
2365             r33              = _mm_mul_ps(rsq33,rinv33);
2366             r33              = _mm_andnot_ps(dummy_mask,r33);
2367
2368             /* EWALD ELECTROSTATICS */
2369
2370             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2371             ewrt             = _mm_mul_ps(r33,ewtabscale);
2372             ewitab           = _mm_cvttps_epi32(ewrt);
2373             eweps            = _mm_sub_ps(ewrt,_mm_cvtepi32_ps(ewitab));
2374             gmx_mm_load_4pair_swizzle_ps(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
2375                                          ewtab+gmx_mm_extract_epi32(ewitab,2),ewtab+gmx_mm_extract_epi32(ewitab,3),
2376                                          &ewtabF,&ewtabFn);
2377             felec            = _mm_add_ps(_mm_mul_ps( _mm_sub_ps(one,eweps),ewtabF),_mm_mul_ps(eweps,ewtabFn));
2378             felec            = _mm_mul_ps(_mm_mul_ps(qq33,rinv33),_mm_sub_ps(rinvsq33,felec));
2379
2380             fscal            = felec;
2381
2382             fscal            = _mm_andnot_ps(dummy_mask,fscal);
2383
2384             /* Calculate temporary vectorial force */
2385             tx               = _mm_mul_ps(fscal,dx33);
2386             ty               = _mm_mul_ps(fscal,dy33);
2387             tz               = _mm_mul_ps(fscal,dz33);
2388
2389             /* Update vectorial force */
2390             fix3             = _mm_add_ps(fix3,tx);
2391             fiy3             = _mm_add_ps(fiy3,ty);
2392             fiz3             = _mm_add_ps(fiz3,tz);
2393
2394             fjx3             = _mm_add_ps(fjx3,tx);
2395             fjy3             = _mm_add_ps(fjy3,ty);
2396             fjz3             = _mm_add_ps(fjz3,tz);
2397
2398             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
2399             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
2400             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
2401             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
2402
2403             gmx_mm_decrement_4rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
2404                                                    fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,
2405                                                    fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2406
2407             /* Inner loop uses 363 flops */
2408         }
2409
2410         /* End of innermost loop */
2411
2412         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2413                                               f+i_coord_offset,fshift+i_shift_offset);
2414
2415         /* Increment number of inner iterations */
2416         inneriter                  += j_index_end - j_index_start;
2417
2418         /* Outer loop uses 24 flops */
2419     }
2420
2421     /* Increment number of outer iterations */
2422     outeriter        += nri;
2423
2424     /* Update outer/inner flops */
2425
2426     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*363);
2427 }