src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecRFCut_VdwLJSh_GeomW4W4_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, 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 sse4_1_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "gromacs/simd/math_x86_sse4_1_double.h"
  50 #include "kernelutil_x86_sse4_1_double.h"
  51
  52 /*
  53  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomW4W4_VF_sse4_1_double
  54  * Electrostatics interaction: ReactionField
  55  * VdW interaction:            LennardJones
  56  * Geometry:                   Water4-Water4
  57  * Calculate force/pot:        PotentialAndForce
  58  */
  59 void
  60 nb_kernel_ElecRFCut_VdwLJSh_GeomW4W4_VF_sse4_1_double
  61                     (t_nblist                    * gmx_restrict       nlist,
  62                      rvec                        * gmx_restrict          xx,
  63                      rvec                        * gmx_restrict          ff,
  64                      t_forcerec                  * gmx_restrict          fr,
  65                      t_mdatoms                   * gmx_restrict     mdatoms,
  66                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  67                      t_nrnb                      * gmx_restrict        nrnb)
  68 {
  69     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  70      * just 0 for non-waters.
  71      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  72      * jnr indices corresponding to data put in the four positions in the SIMD register.
  73      */
  74     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  75     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  76     int              jnrA,jnrB;
  77     int              j_coord_offsetA,j_coord_offsetB;
  78     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  79     real             rcutoff_scalar;
  80     real             *shiftvec,*fshift,*x,*f;
  81     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  82     int              vdwioffset0;
  83     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  84     int              vdwioffset1;
  85     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  86     int              vdwioffset2;
  87     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  88     int              vdwioffset3;
  89     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  90     int              vdwjidx0A,vdwjidx0B;
  91     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  92     int              vdwjidx1A,vdwjidx1B;
  93     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
  94     int              vdwjidx2A,vdwjidx2B;
  95     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
  96     int              vdwjidx3A,vdwjidx3B;
  97     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
  98     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  99     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
 100     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
 101     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
 102     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
 103     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
 104     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
 105     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
 106     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
 107     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
 108     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 109     real             *charge;
 110     int              nvdwtype;
 111     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 112     int              *vdwtype;
 113     real             *vdwparam;
 114     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 115     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 116     __m128d          dummy_mask,cutoff_mask;
 117     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 118     __m128d          one     = _mm_set1_pd(1.0);
 119     __m128d          two     = _mm_set1_pd(2.0);
 120     x                = xx[0];
 121     f                = ff[0];
 122
 123     nri              = nlist->nri;
 124     iinr             = nlist->iinr;
 125     jindex           = nlist->jindex;
 126     jjnr             = nlist->jjnr;
 127     shiftidx         = nlist->shift;
 128     gid              = nlist->gid;
 129     shiftvec         = fr->shift_vec[0];
 130     fshift           = fr->fshift[0];
 131     facel            = _mm_set1_pd(fr->epsfac);
 132     charge           = mdatoms->chargeA;
 133     krf              = _mm_set1_pd(fr->ic->k_rf);
 134     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 135     crf              = _mm_set1_pd(fr->ic->c_rf);
 136     nvdwtype         = fr->ntype;
 137     vdwparam         = fr->nbfp;
 138     vdwtype          = mdatoms->typeA;
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
 143     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
 144     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
 145     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 146
 147     jq1              = _mm_set1_pd(charge[inr+1]);
 148     jq2              = _mm_set1_pd(charge[inr+2]);
 149     jq3              = _mm_set1_pd(charge[inr+3]);
 150     vdwjidx0A        = 2*vdwtype[inr+0];
 151     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
 152     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
 153     qq11             = _mm_mul_pd(iq1,jq1);
 154     qq12             = _mm_mul_pd(iq1,jq2);
 155     qq13             = _mm_mul_pd(iq1,jq3);
 156     qq21             = _mm_mul_pd(iq2,jq1);
 157     qq22             = _mm_mul_pd(iq2,jq2);
 158     qq23             = _mm_mul_pd(iq2,jq3);
 159     qq31             = _mm_mul_pd(iq3,jq1);
 160     qq32             = _mm_mul_pd(iq3,jq2);
 161     qq33             = _mm_mul_pd(iq3,jq3);
 162
 163     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 164     rcutoff_scalar   = fr->rcoulomb;
 165     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 166     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 167
 168     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 169     rvdw             = _mm_set1_pd(fr->rvdw);
 170
 171     /* Avoid stupid compiler warnings */
 172     jnrA = jnrB = 0;
 173     j_coord_offsetA = 0;
 174     j_coord_offsetB = 0;
 175
 176     outeriter        = 0;
 177     inneriter        = 0;
 178
 179     /* Start outer loop over neighborlists */
 180     for(iidx=0; iidx<nri; iidx++)
 181     {
 182         /* Load shift vector for this list */
 183         i_shift_offset   = DIM*shiftidx[iidx];
 184
 185         /* Load limits for loop over neighbors */
 186         j_index_start    = jindex[iidx];
 187         j_index_end      = jindex[iidx+1];
 188
 189         /* Get outer coordinate index */
 190         inr              = iinr[iidx];
 191         i_coord_offset   = DIM*inr;
 192
 193         /* Load i particle coords and add shift vector */
 194         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
 195                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 196
 197         fix0             = _mm_setzero_pd();
 198         fiy0             = _mm_setzero_pd();
 199         fiz0             = _mm_setzero_pd();
 200         fix1             = _mm_setzero_pd();
 201         fiy1             = _mm_setzero_pd();
 202         fiz1             = _mm_setzero_pd();
 203         fix2             = _mm_setzero_pd();
 204         fiy2             = _mm_setzero_pd();
 205         fiz2             = _mm_setzero_pd();
 206         fix3             = _mm_setzero_pd();
 207         fiy3             = _mm_setzero_pd();
 208         fiz3             = _mm_setzero_pd();
 209
 210         /* Reset potential sums */
 211         velecsum         = _mm_setzero_pd();
 212         vvdwsum          = _mm_setzero_pd();
 213
 214         /* Start inner kernel loop */
 215         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 216         {
 217
 218             /* Get j neighbor index, and coordinate index */
 219             jnrA             = jjnr[jidx];
 220             jnrB             = jjnr[jidx+1];
 221             j_coord_offsetA  = DIM*jnrA;
 222             j_coord_offsetB  = DIM*jnrB;
 223
 224             /* load j atom coordinates */
 225             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 226                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 227                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 228
 229             /* Calculate displacement vector */
 230             dx00             = _mm_sub_pd(ix0,jx0);
 231             dy00             = _mm_sub_pd(iy0,jy0);
 232             dz00             = _mm_sub_pd(iz0,jz0);
 233             dx11             = _mm_sub_pd(ix1,jx1);
 234             dy11             = _mm_sub_pd(iy1,jy1);
 235             dz11             = _mm_sub_pd(iz1,jz1);
 236             dx12             = _mm_sub_pd(ix1,jx2);
 237             dy12             = _mm_sub_pd(iy1,jy2);
 238             dz12             = _mm_sub_pd(iz1,jz2);
 239             dx13             = _mm_sub_pd(ix1,jx3);
 240             dy13             = _mm_sub_pd(iy1,jy3);
 241             dz13             = _mm_sub_pd(iz1,jz3);
 242             dx21             = _mm_sub_pd(ix2,jx1);
 243             dy21             = _mm_sub_pd(iy2,jy1);
 244             dz21             = _mm_sub_pd(iz2,jz1);
 245             dx22             = _mm_sub_pd(ix2,jx2);
 246             dy22             = _mm_sub_pd(iy2,jy2);
 247             dz22             = _mm_sub_pd(iz2,jz2);
 248             dx23             = _mm_sub_pd(ix2,jx3);
 249             dy23             = _mm_sub_pd(iy2,jy3);
 250             dz23             = _mm_sub_pd(iz2,jz3);
 251             dx31             = _mm_sub_pd(ix3,jx1);
 252             dy31             = _mm_sub_pd(iy3,jy1);
 253             dz31             = _mm_sub_pd(iz3,jz1);
 254             dx32             = _mm_sub_pd(ix3,jx2);
 255             dy32             = _mm_sub_pd(iy3,jy2);
 256             dz32             = _mm_sub_pd(iz3,jz2);
 257             dx33             = _mm_sub_pd(ix3,jx3);
 258             dy33             = _mm_sub_pd(iy3,jy3);
 259             dz33             = _mm_sub_pd(iz3,jz3);
 260
 261             /* Calculate squared distance and things based on it */
 262             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 263             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 264             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 265             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 266             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 267             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 268             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 269             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 270             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 271             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 272
 273             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 274             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 275             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 276             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 277             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 278             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 279             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 280             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 281             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 282
 283             rinvsq00         = gmx_mm_inv_pd(rsq00);
 284             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 285             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 286             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 287             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 288             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 289             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 290             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 291             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 292             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 293
 294             fjx0             = _mm_setzero_pd();
 295             fjy0             = _mm_setzero_pd();
 296             fjz0             = _mm_setzero_pd();
 297             fjx1             = _mm_setzero_pd();
 298             fjy1             = _mm_setzero_pd();
 299             fjz1             = _mm_setzero_pd();
 300             fjx2             = _mm_setzero_pd();
 301             fjy2             = _mm_setzero_pd();
 302             fjz2             = _mm_setzero_pd();
 303             fjx3             = _mm_setzero_pd();
 304             fjy3             = _mm_setzero_pd();
 305             fjz3             = _mm_setzero_pd();
 306
 307             /**************************
 308              * CALCULATE INTERACTIONS *
 309              **************************/
 310
 311             if (gmx_mm_any_lt(rsq00,rcutoff2))
 312             {
 313
 314             /* LENNARD-JONES DISPERSION/REPULSION */
 315
 316             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 317             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 318             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 319             vvdw             = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 320                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 321             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 322
 323             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 324
 325             /* Update potential sum for this i atom from the interaction with this j atom. */
 326             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 327             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 328
 329             fscal            = fvdw;
 330
 331             fscal            = _mm_and_pd(fscal,cutoff_mask);
 332
 333             /* Calculate temporary vectorial force */
 334             tx               = _mm_mul_pd(fscal,dx00);
 335             ty               = _mm_mul_pd(fscal,dy00);
 336             tz               = _mm_mul_pd(fscal,dz00);
 337
 338             /* Update vectorial force */
 339             fix0             = _mm_add_pd(fix0,tx);
 340             fiy0             = _mm_add_pd(fiy0,ty);
 341             fiz0             = _mm_add_pd(fiz0,tz);
 342
 343             fjx0             = _mm_add_pd(fjx0,tx);
 344             fjy0             = _mm_add_pd(fjy0,ty);
 345             fjz0             = _mm_add_pd(fjz0,tz);
 346
 347             }
 348
 349             /**************************
 350              * CALCULATE INTERACTIONS *
 351              **************************/
 352
 353             if (gmx_mm_any_lt(rsq11,rcutoff2))
 354             {
 355
 356             /* REACTION-FIELD ELECTROSTATICS */
 357             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 358             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 359
 360             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
 361
 362             /* Update potential sum for this i atom from the interaction with this j atom. */
 363             velec            = _mm_and_pd(velec,cutoff_mask);
 364             velecsum         = _mm_add_pd(velecsum,velec);
 365
 366             fscal            = felec;
 367
 368             fscal            = _mm_and_pd(fscal,cutoff_mask);
 369
 370             /* Calculate temporary vectorial force */
 371             tx               = _mm_mul_pd(fscal,dx11);
 372             ty               = _mm_mul_pd(fscal,dy11);
 373             tz               = _mm_mul_pd(fscal,dz11);
 374
 375             /* Update vectorial force */
 376             fix1             = _mm_add_pd(fix1,tx);
 377             fiy1             = _mm_add_pd(fiy1,ty);
 378             fiz1             = _mm_add_pd(fiz1,tz);
 379
 380             fjx1             = _mm_add_pd(fjx1,tx);
 381             fjy1             = _mm_add_pd(fjy1,ty);
 382             fjz1             = _mm_add_pd(fjz1,tz);
 383
 384             }
 385
 386             /**************************
 387              * CALCULATE INTERACTIONS *
 388              **************************/
 389
 390             if (gmx_mm_any_lt(rsq12,rcutoff2))
 391             {
 392
 393             /* REACTION-FIELD ELECTROSTATICS */
 394             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 395             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 396
 397             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
 398
 399             /* Update potential sum for this i atom from the interaction with this j atom. */
 400             velec            = _mm_and_pd(velec,cutoff_mask);
 401             velecsum         = _mm_add_pd(velecsum,velec);
 402
 403             fscal            = felec;
 404
 405             fscal            = _mm_and_pd(fscal,cutoff_mask);
 406
 407             /* Calculate temporary vectorial force */
 408             tx               = _mm_mul_pd(fscal,dx12);
 409             ty               = _mm_mul_pd(fscal,dy12);
 410             tz               = _mm_mul_pd(fscal,dz12);
 411
 412             /* Update vectorial force */
 413             fix1             = _mm_add_pd(fix1,tx);
 414             fiy1             = _mm_add_pd(fiy1,ty);
 415             fiz1             = _mm_add_pd(fiz1,tz);
 416
 417             fjx2             = _mm_add_pd(fjx2,tx);
 418             fjy2             = _mm_add_pd(fjy2,ty);
 419             fjz2             = _mm_add_pd(fjz2,tz);
 420
 421             }
 422
 423             /**************************
 424              * CALCULATE INTERACTIONS *
 425              **************************/
 426
 427             if (gmx_mm_any_lt(rsq13,rcutoff2))
 428             {
 429
 430             /* REACTION-FIELD ELECTROSTATICS */
 431             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_add_pd(rinv13,_mm_mul_pd(krf,rsq13)),crf));
 432             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
 433
 434             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
 435
 436             /* Update potential sum for this i atom from the interaction with this j atom. */
 437             velec            = _mm_and_pd(velec,cutoff_mask);
 438             velecsum         = _mm_add_pd(velecsum,velec);
 439
 440             fscal            = felec;
 441
 442             fscal            = _mm_and_pd(fscal,cutoff_mask);
 443
 444             /* Calculate temporary vectorial force */
 445             tx               = _mm_mul_pd(fscal,dx13);
 446             ty               = _mm_mul_pd(fscal,dy13);
 447             tz               = _mm_mul_pd(fscal,dz13);
 448
 449             /* Update vectorial force */
 450             fix1             = _mm_add_pd(fix1,tx);
 451             fiy1             = _mm_add_pd(fiy1,ty);
 452             fiz1             = _mm_add_pd(fiz1,tz);
 453
 454             fjx3             = _mm_add_pd(fjx3,tx);
 455             fjy3             = _mm_add_pd(fjy3,ty);
 456             fjz3             = _mm_add_pd(fjz3,tz);
 457
 458             }
 459
 460             /**************************
 461              * CALCULATE INTERACTIONS *
 462              **************************/
 463
 464             if (gmx_mm_any_lt(rsq21,rcutoff2))
 465             {
 466
 467             /* REACTION-FIELD ELECTROSTATICS */
 468             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 469             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 470
 471             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
 472
 473             /* Update potential sum for this i atom from the interaction with this j atom. */
 474             velec            = _mm_and_pd(velec,cutoff_mask);
 475             velecsum         = _mm_add_pd(velecsum,velec);
 476
 477             fscal            = felec;
 478
 479             fscal            = _mm_and_pd(fscal,cutoff_mask);
 480
 481             /* Calculate temporary vectorial force */
 482             tx               = _mm_mul_pd(fscal,dx21);
 483             ty               = _mm_mul_pd(fscal,dy21);
 484             tz               = _mm_mul_pd(fscal,dz21);
 485
 486             /* Update vectorial force */
 487             fix2             = _mm_add_pd(fix2,tx);
 488             fiy2             = _mm_add_pd(fiy2,ty);
 489             fiz2             = _mm_add_pd(fiz2,tz);
 490
 491             fjx1             = _mm_add_pd(fjx1,tx);
 492             fjy1             = _mm_add_pd(fjy1,ty);
 493             fjz1             = _mm_add_pd(fjz1,tz);
 494
 495             }
 496
 497             /**************************
 498              * CALCULATE INTERACTIONS *
 499              **************************/
 500
 501             if (gmx_mm_any_lt(rsq22,rcutoff2))
 502             {
 503
 504             /* REACTION-FIELD ELECTROSTATICS */
 505             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 506             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 507
 508             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
 509
 510             /* Update potential sum for this i atom from the interaction with this j atom. */
 511             velec            = _mm_and_pd(velec,cutoff_mask);
 512             velecsum         = _mm_add_pd(velecsum,velec);
 513
 514             fscal            = felec;
 515
 516             fscal            = _mm_and_pd(fscal,cutoff_mask);
 517
 518             /* Calculate temporary vectorial force */
 519             tx               = _mm_mul_pd(fscal,dx22);
 520             ty               = _mm_mul_pd(fscal,dy22);
 521             tz               = _mm_mul_pd(fscal,dz22);
 522
 523             /* Update vectorial force */
 524             fix2             = _mm_add_pd(fix2,tx);
 525             fiy2             = _mm_add_pd(fiy2,ty);
 526             fiz2             = _mm_add_pd(fiz2,tz);
 527
 528             fjx2             = _mm_add_pd(fjx2,tx);
 529             fjy2             = _mm_add_pd(fjy2,ty);
 530             fjz2             = _mm_add_pd(fjz2,tz);
 531
 532             }
 533
 534             /**************************
 535              * CALCULATE INTERACTIONS *
 536              **************************/
 537
 538             if (gmx_mm_any_lt(rsq23,rcutoff2))
 539             {
 540
 541             /* REACTION-FIELD ELECTROSTATICS */
 542             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_add_pd(rinv23,_mm_mul_pd(krf,rsq23)),crf));
 543             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
 544
 545             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
 546
 547             /* Update potential sum for this i atom from the interaction with this j atom. */
 548             velec            = _mm_and_pd(velec,cutoff_mask);
 549             velecsum         = _mm_add_pd(velecsum,velec);
 550
 551             fscal            = felec;
 552
 553             fscal            = _mm_and_pd(fscal,cutoff_mask);
 554
 555             /* Calculate temporary vectorial force */
 556             tx               = _mm_mul_pd(fscal,dx23);
 557             ty               = _mm_mul_pd(fscal,dy23);
 558             tz               = _mm_mul_pd(fscal,dz23);
 559
 560             /* Update vectorial force */
 561             fix2             = _mm_add_pd(fix2,tx);
 562             fiy2             = _mm_add_pd(fiy2,ty);
 563             fiz2             = _mm_add_pd(fiz2,tz);
 564
 565             fjx3             = _mm_add_pd(fjx3,tx);
 566             fjy3             = _mm_add_pd(fjy3,ty);
 567             fjz3             = _mm_add_pd(fjz3,tz);
 568
 569             }
 570
 571             /**************************
 572              * CALCULATE INTERACTIONS *
 573              **************************/
 574
 575             if (gmx_mm_any_lt(rsq31,rcutoff2))
 576             {
 577
 578             /* REACTION-FIELD ELECTROSTATICS */
 579             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_add_pd(rinv31,_mm_mul_pd(krf,rsq31)),crf));
 580             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
 581
 582             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
 583
 584             /* Update potential sum for this i atom from the interaction with this j atom. */
 585             velec            = _mm_and_pd(velec,cutoff_mask);
 586             velecsum         = _mm_add_pd(velecsum,velec);
 587
 588             fscal            = felec;
 589
 590             fscal            = _mm_and_pd(fscal,cutoff_mask);
 591
 592             /* Calculate temporary vectorial force */
 593             tx               = _mm_mul_pd(fscal,dx31);
 594             ty               = _mm_mul_pd(fscal,dy31);
 595             tz               = _mm_mul_pd(fscal,dz31);
 596
 597             /* Update vectorial force */
 598             fix3             = _mm_add_pd(fix3,tx);
 599             fiy3             = _mm_add_pd(fiy3,ty);
 600             fiz3             = _mm_add_pd(fiz3,tz);
 601
 602             fjx1             = _mm_add_pd(fjx1,tx);
 603             fjy1             = _mm_add_pd(fjy1,ty);
 604             fjz1             = _mm_add_pd(fjz1,tz);
 605
 606             }
 607
 608             /**************************
 609              * CALCULATE INTERACTIONS *
 610              **************************/
 611
 612             if (gmx_mm_any_lt(rsq32,rcutoff2))
 613             {
 614
 615             /* REACTION-FIELD ELECTROSTATICS */
 616             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_add_pd(rinv32,_mm_mul_pd(krf,rsq32)),crf));
 617             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
 618
 619             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
 620
 621             /* Update potential sum for this i atom from the interaction with this j atom. */
 622             velec            = _mm_and_pd(velec,cutoff_mask);
 623             velecsum         = _mm_add_pd(velecsum,velec);
 624
 625             fscal            = felec;
 626
 627             fscal            = _mm_and_pd(fscal,cutoff_mask);
 628
 629             /* Calculate temporary vectorial force */
 630             tx               = _mm_mul_pd(fscal,dx32);
 631             ty               = _mm_mul_pd(fscal,dy32);
 632             tz               = _mm_mul_pd(fscal,dz32);
 633
 634             /* Update vectorial force */
 635             fix3             = _mm_add_pd(fix3,tx);
 636             fiy3             = _mm_add_pd(fiy3,ty);
 637             fiz3             = _mm_add_pd(fiz3,tz);
 638
 639             fjx2             = _mm_add_pd(fjx2,tx);
 640             fjy2             = _mm_add_pd(fjy2,ty);
 641             fjz2             = _mm_add_pd(fjz2,tz);
 642
 643             }
 644
 645             /**************************
 646              * CALCULATE INTERACTIONS *
 647              **************************/
 648
 649             if (gmx_mm_any_lt(rsq33,rcutoff2))
 650             {
 651
 652             /* REACTION-FIELD ELECTROSTATICS */
 653             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_add_pd(rinv33,_mm_mul_pd(krf,rsq33)),crf));
 654             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
 655
 656             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
 657
 658             /* Update potential sum for this i atom from the interaction with this j atom. */
 659             velec            = _mm_and_pd(velec,cutoff_mask);
 660             velecsum         = _mm_add_pd(velecsum,velec);
 661
 662             fscal            = felec;
 663
 664             fscal            = _mm_and_pd(fscal,cutoff_mask);
 665
 666             /* Calculate temporary vectorial force */
 667             tx               = _mm_mul_pd(fscal,dx33);
 668             ty               = _mm_mul_pd(fscal,dy33);
 669             tz               = _mm_mul_pd(fscal,dz33);
 670
 671             /* Update vectorial force */
 672             fix3             = _mm_add_pd(fix3,tx);
 673             fiy3             = _mm_add_pd(fiy3,ty);
 674             fiz3             = _mm_add_pd(fiz3,tz);
 675
 676             fjx3             = _mm_add_pd(fjx3,tx);
 677             fjy3             = _mm_add_pd(fjy3,ty);
 678             fjz3             = _mm_add_pd(fjz3,tz);
 679
 680             }
 681
 682             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
 683
 684             /* Inner loop uses 368 flops */
 685         }
 686
 687         if(jidx<j_index_end)
 688         {
 689
 690             jnrA             = jjnr[jidx];
 691             j_coord_offsetA  = DIM*jnrA;
 692
 693             /* load j atom coordinates */
 694             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 695                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
 696                                               &jy2,&jz2,&jx3,&jy3,&jz3);
 697
 698             /* Calculate displacement vector */
 699             dx00             = _mm_sub_pd(ix0,jx0);
 700             dy00             = _mm_sub_pd(iy0,jy0);
 701             dz00             = _mm_sub_pd(iz0,jz0);
 702             dx11             = _mm_sub_pd(ix1,jx1);
 703             dy11             = _mm_sub_pd(iy1,jy1);
 704             dz11             = _mm_sub_pd(iz1,jz1);
 705             dx12             = _mm_sub_pd(ix1,jx2);
 706             dy12             = _mm_sub_pd(iy1,jy2);
 707             dz12             = _mm_sub_pd(iz1,jz2);
 708             dx13             = _mm_sub_pd(ix1,jx3);
 709             dy13             = _mm_sub_pd(iy1,jy3);
 710             dz13             = _mm_sub_pd(iz1,jz3);
 711             dx21             = _mm_sub_pd(ix2,jx1);
 712             dy21             = _mm_sub_pd(iy2,jy1);
 713             dz21             = _mm_sub_pd(iz2,jz1);
 714             dx22             = _mm_sub_pd(ix2,jx2);
 715             dy22             = _mm_sub_pd(iy2,jy2);
 716             dz22             = _mm_sub_pd(iz2,jz2);
 717             dx23             = _mm_sub_pd(ix2,jx3);
 718             dy23             = _mm_sub_pd(iy2,jy3);
 719             dz23             = _mm_sub_pd(iz2,jz3);
 720             dx31             = _mm_sub_pd(ix3,jx1);
 721             dy31             = _mm_sub_pd(iy3,jy1);
 722             dz31             = _mm_sub_pd(iz3,jz1);
 723             dx32             = _mm_sub_pd(ix3,jx2);
 724             dy32             = _mm_sub_pd(iy3,jy2);
 725             dz32             = _mm_sub_pd(iz3,jz2);
 726             dx33             = _mm_sub_pd(ix3,jx3);
 727             dy33             = _mm_sub_pd(iy3,jy3);
 728             dz33             = _mm_sub_pd(iz3,jz3);
 729
 730             /* Calculate squared distance and things based on it */
 731             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 732             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
 733             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
 734             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
 735             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
 736             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
 737             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
 738             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
 739             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
 740             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
 741
 742             rinv11           = gmx_mm_invsqrt_pd(rsq11);
 743             rinv12           = gmx_mm_invsqrt_pd(rsq12);
 744             rinv13           = gmx_mm_invsqrt_pd(rsq13);
 745             rinv21           = gmx_mm_invsqrt_pd(rsq21);
 746             rinv22           = gmx_mm_invsqrt_pd(rsq22);
 747             rinv23           = gmx_mm_invsqrt_pd(rsq23);
 748             rinv31           = gmx_mm_invsqrt_pd(rsq31);
 749             rinv32           = gmx_mm_invsqrt_pd(rsq32);
 750             rinv33           = gmx_mm_invsqrt_pd(rsq33);
 751
 752             rinvsq00         = gmx_mm_inv_pd(rsq00);
 753             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
 754             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
 755             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
 756             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
 757             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
 758             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
 759             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
 760             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
 761             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
 762
 763             fjx0             = _mm_setzero_pd();
 764             fjy0             = _mm_setzero_pd();
 765             fjz0             = _mm_setzero_pd();
 766             fjx1             = _mm_setzero_pd();
 767             fjy1             = _mm_setzero_pd();
 768             fjz1             = _mm_setzero_pd();
 769             fjx2             = _mm_setzero_pd();
 770             fjy2             = _mm_setzero_pd();
 771             fjz2             = _mm_setzero_pd();
 772             fjx3             = _mm_setzero_pd();
 773             fjy3             = _mm_setzero_pd();
 774             fjz3             = _mm_setzero_pd();
 775
 776             /**************************
 777              * CALCULATE INTERACTIONS *
 778              **************************/
 779
 780             if (gmx_mm_any_lt(rsq00,rcutoff2))
 781             {
 782
 783             /* LENNARD-JONES DISPERSION/REPULSION */
 784
 785             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 786             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 787             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 788             vvdw             = _mm_sub_pd(_mm_mul_pd( _mm_sub_pd(vvdw12 , _mm_mul_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6))), one_twelfth) ,
 789                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 790             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 791
 792             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 793
 794             /* Update potential sum for this i atom from the interaction with this j atom. */
 795             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 796             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 797             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 798
 799             fscal            = fvdw;
 800
 801             fscal            = _mm_and_pd(fscal,cutoff_mask);
 802
 803             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 804
 805             /* Calculate temporary vectorial force */
 806             tx               = _mm_mul_pd(fscal,dx00);
 807             ty               = _mm_mul_pd(fscal,dy00);
 808             tz               = _mm_mul_pd(fscal,dz00);
 809
 810             /* Update vectorial force */
 811             fix0             = _mm_add_pd(fix0,tx);
 812             fiy0             = _mm_add_pd(fiy0,ty);
 813             fiz0             = _mm_add_pd(fiz0,tz);
 814
 815             fjx0             = _mm_add_pd(fjx0,tx);
 816             fjy0             = _mm_add_pd(fjy0,ty);
 817             fjz0             = _mm_add_pd(fjz0,tz);
 818
 819             }
 820
 821             /**************************
 822              * CALCULATE INTERACTIONS *
 823              **************************/
 824
 825             if (gmx_mm_any_lt(rsq11,rcutoff2))
 826             {
 827
 828             /* REACTION-FIELD ELECTROSTATICS */
 829             velec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_add_pd(rinv11,_mm_mul_pd(krf,rsq11)),crf));
 830             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
 831
 832             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
 833
 834             /* Update potential sum for this i atom from the interaction with this j atom. */
 835             velec            = _mm_and_pd(velec,cutoff_mask);
 836             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 837             velecsum         = _mm_add_pd(velecsum,velec);
 838
 839             fscal            = felec;
 840
 841             fscal            = _mm_and_pd(fscal,cutoff_mask);
 842
 843             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 844
 845             /* Calculate temporary vectorial force */
 846             tx               = _mm_mul_pd(fscal,dx11);
 847             ty               = _mm_mul_pd(fscal,dy11);
 848             tz               = _mm_mul_pd(fscal,dz11);
 849
 850             /* Update vectorial force */
 851             fix1             = _mm_add_pd(fix1,tx);
 852             fiy1             = _mm_add_pd(fiy1,ty);
 853             fiz1             = _mm_add_pd(fiz1,tz);
 854
 855             fjx1             = _mm_add_pd(fjx1,tx);
 856             fjy1             = _mm_add_pd(fjy1,ty);
 857             fjz1             = _mm_add_pd(fjz1,tz);
 858
 859             }
 860
 861             /**************************
 862              * CALCULATE INTERACTIONS *
 863              **************************/
 864
 865             if (gmx_mm_any_lt(rsq12,rcutoff2))
 866             {
 867
 868             /* REACTION-FIELD ELECTROSTATICS */
 869             velec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_add_pd(rinv12,_mm_mul_pd(krf,rsq12)),crf));
 870             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
 871
 872             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
 873
 874             /* Update potential sum for this i atom from the interaction with this j atom. */
 875             velec            = _mm_and_pd(velec,cutoff_mask);
 876             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 877             velecsum         = _mm_add_pd(velecsum,velec);
 878
 879             fscal            = felec;
 880
 881             fscal            = _mm_and_pd(fscal,cutoff_mask);
 882
 883             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 884
 885             /* Calculate temporary vectorial force */
 886             tx               = _mm_mul_pd(fscal,dx12);
 887             ty               = _mm_mul_pd(fscal,dy12);
 888             tz               = _mm_mul_pd(fscal,dz12);
 889
 890             /* Update vectorial force */
 891             fix1             = _mm_add_pd(fix1,tx);
 892             fiy1             = _mm_add_pd(fiy1,ty);
 893             fiz1             = _mm_add_pd(fiz1,tz);
 894
 895             fjx2             = _mm_add_pd(fjx2,tx);
 896             fjy2             = _mm_add_pd(fjy2,ty);
 897             fjz2             = _mm_add_pd(fjz2,tz);
 898
 899             }
 900
 901             /**************************
 902              * CALCULATE INTERACTIONS *
 903              **************************/
 904
 905             if (gmx_mm_any_lt(rsq13,rcutoff2))
 906             {
 907
 908             /* REACTION-FIELD ELECTROSTATICS */
 909             velec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_add_pd(rinv13,_mm_mul_pd(krf,rsq13)),crf));
 910             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
 911
 912             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
 913
 914             /* Update potential sum for this i atom from the interaction with this j atom. */
 915             velec            = _mm_and_pd(velec,cutoff_mask);
 916             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 917             velecsum         = _mm_add_pd(velecsum,velec);
 918
 919             fscal            = felec;
 920
 921             fscal            = _mm_and_pd(fscal,cutoff_mask);
 922
 923             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 924
 925             /* Calculate temporary vectorial force */
 926             tx               = _mm_mul_pd(fscal,dx13);
 927             ty               = _mm_mul_pd(fscal,dy13);
 928             tz               = _mm_mul_pd(fscal,dz13);
 929
 930             /* Update vectorial force */
 931             fix1             = _mm_add_pd(fix1,tx);
 932             fiy1             = _mm_add_pd(fiy1,ty);
 933             fiz1             = _mm_add_pd(fiz1,tz);
 934
 935             fjx3             = _mm_add_pd(fjx3,tx);
 936             fjy3             = _mm_add_pd(fjy3,ty);
 937             fjz3             = _mm_add_pd(fjz3,tz);
 938
 939             }
 940
 941             /**************************
 942              * CALCULATE INTERACTIONS *
 943              **************************/
 944
 945             if (gmx_mm_any_lt(rsq21,rcutoff2))
 946             {
 947
 948             /* REACTION-FIELD ELECTROSTATICS */
 949             velec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_add_pd(rinv21,_mm_mul_pd(krf,rsq21)),crf));
 950             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
 951
 952             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
 953
 954             /* Update potential sum for this i atom from the interaction with this j atom. */
 955             velec            = _mm_and_pd(velec,cutoff_mask);
 956             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 957             velecsum         = _mm_add_pd(velecsum,velec);
 958
 959             fscal            = felec;
 960
 961             fscal            = _mm_and_pd(fscal,cutoff_mask);
 962
 963             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 964
 965             /* Calculate temporary vectorial force */
 966             tx               = _mm_mul_pd(fscal,dx21);
 967             ty               = _mm_mul_pd(fscal,dy21);
 968             tz               = _mm_mul_pd(fscal,dz21);
 969
 970             /* Update vectorial force */
 971             fix2             = _mm_add_pd(fix2,tx);
 972             fiy2             = _mm_add_pd(fiy2,ty);
 973             fiz2             = _mm_add_pd(fiz2,tz);
 974
 975             fjx1             = _mm_add_pd(fjx1,tx);
 976             fjy1             = _mm_add_pd(fjy1,ty);
 977             fjz1             = _mm_add_pd(fjz1,tz);
 978
 979             }
 980
 981             /**************************
 982              * CALCULATE INTERACTIONS *
 983              **************************/
 984
 985             if (gmx_mm_any_lt(rsq22,rcutoff2))
 986             {
 987
 988             /* REACTION-FIELD ELECTROSTATICS */
 989             velec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_add_pd(rinv22,_mm_mul_pd(krf,rsq22)),crf));
 990             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
 991
 992             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
 993
 994             /* Update potential sum for this i atom from the interaction with this j atom. */
 995             velec            = _mm_and_pd(velec,cutoff_mask);
 996             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 997             velecsum         = _mm_add_pd(velecsum,velec);
 998
 999             fscal            = felec;
1000
1001             fscal            = _mm_and_pd(fscal,cutoff_mask);
1002
1003             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1004
1005             /* Calculate temporary vectorial force */
1006             tx               = _mm_mul_pd(fscal,dx22);
1007             ty               = _mm_mul_pd(fscal,dy22);
1008             tz               = _mm_mul_pd(fscal,dz22);
1009
1010             /* Update vectorial force */
1011             fix2             = _mm_add_pd(fix2,tx);
1012             fiy2             = _mm_add_pd(fiy2,ty);
1013             fiz2             = _mm_add_pd(fiz2,tz);
1014
1015             fjx2             = _mm_add_pd(fjx2,tx);
1016             fjy2             = _mm_add_pd(fjy2,ty);
1017             fjz2             = _mm_add_pd(fjz2,tz);
1018
1019             }
1020
1021             /**************************
1022              * CALCULATE INTERACTIONS *
1023              **************************/
1024
1025             if (gmx_mm_any_lt(rsq23,rcutoff2))
1026             {
1027
1028             /* REACTION-FIELD ELECTROSTATICS */
1029             velec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_add_pd(rinv23,_mm_mul_pd(krf,rsq23)),crf));
1030             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
1031
1032             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
1033
1034             /* Update potential sum for this i atom from the interaction with this j atom. */
1035             velec            = _mm_and_pd(velec,cutoff_mask);
1036             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1037             velecsum         = _mm_add_pd(velecsum,velec);
1038
1039             fscal            = felec;
1040
1041             fscal            = _mm_and_pd(fscal,cutoff_mask);
1042
1043             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1044
1045             /* Calculate temporary vectorial force */
1046             tx               = _mm_mul_pd(fscal,dx23);
1047             ty               = _mm_mul_pd(fscal,dy23);
1048             tz               = _mm_mul_pd(fscal,dz23);
1049
1050             /* Update vectorial force */
1051             fix2             = _mm_add_pd(fix2,tx);
1052             fiy2             = _mm_add_pd(fiy2,ty);
1053             fiz2             = _mm_add_pd(fiz2,tz);
1054
1055             fjx3             = _mm_add_pd(fjx3,tx);
1056             fjy3             = _mm_add_pd(fjy3,ty);
1057             fjz3             = _mm_add_pd(fjz3,tz);
1058
1059             }
1060
1061             /**************************
1062              * CALCULATE INTERACTIONS *
1063              **************************/
1064
1065             if (gmx_mm_any_lt(rsq31,rcutoff2))
1066             {
1067
1068             /* REACTION-FIELD ELECTROSTATICS */
1069             velec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_add_pd(rinv31,_mm_mul_pd(krf,rsq31)),crf));
1070             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
1071
1072             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
1073
1074             /* Update potential sum for this i atom from the interaction with this j atom. */
1075             velec            = _mm_and_pd(velec,cutoff_mask);
1076             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1077             velecsum         = _mm_add_pd(velecsum,velec);
1078
1079             fscal            = felec;
1080
1081             fscal            = _mm_and_pd(fscal,cutoff_mask);
1082
1083             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1084
1085             /* Calculate temporary vectorial force */
1086             tx               = _mm_mul_pd(fscal,dx31);
1087             ty               = _mm_mul_pd(fscal,dy31);
1088             tz               = _mm_mul_pd(fscal,dz31);
1089
1090             /* Update vectorial force */
1091             fix3             = _mm_add_pd(fix3,tx);
1092             fiy3             = _mm_add_pd(fiy3,ty);
1093             fiz3             = _mm_add_pd(fiz3,tz);
1094
1095             fjx1             = _mm_add_pd(fjx1,tx);
1096             fjy1             = _mm_add_pd(fjy1,ty);
1097             fjz1             = _mm_add_pd(fjz1,tz);
1098
1099             }
1100
1101             /**************************
1102              * CALCULATE INTERACTIONS *
1103              **************************/
1104
1105             if (gmx_mm_any_lt(rsq32,rcutoff2))
1106             {
1107
1108             /* REACTION-FIELD ELECTROSTATICS */
1109             velec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_add_pd(rinv32,_mm_mul_pd(krf,rsq32)),crf));
1110             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
1111
1112             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
1113
1114             /* Update potential sum for this i atom from the interaction with this j atom. */
1115             velec            = _mm_and_pd(velec,cutoff_mask);
1116             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1117             velecsum         = _mm_add_pd(velecsum,velec);
1118
1119             fscal            = felec;
1120
1121             fscal            = _mm_and_pd(fscal,cutoff_mask);
1122
1123             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1124
1125             /* Calculate temporary vectorial force */
1126             tx               = _mm_mul_pd(fscal,dx32);
1127             ty               = _mm_mul_pd(fscal,dy32);
1128             tz               = _mm_mul_pd(fscal,dz32);
1129
1130             /* Update vectorial force */
1131             fix3             = _mm_add_pd(fix3,tx);
1132             fiy3             = _mm_add_pd(fiy3,ty);
1133             fiz3             = _mm_add_pd(fiz3,tz);
1134
1135             fjx2             = _mm_add_pd(fjx2,tx);
1136             fjy2             = _mm_add_pd(fjy2,ty);
1137             fjz2             = _mm_add_pd(fjz2,tz);
1138
1139             }
1140
1141             /**************************
1142              * CALCULATE INTERACTIONS *
1143              **************************/
1144
1145             if (gmx_mm_any_lt(rsq33,rcutoff2))
1146             {
1147
1148             /* REACTION-FIELD ELECTROSTATICS */
1149             velec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_add_pd(rinv33,_mm_mul_pd(krf,rsq33)),crf));
1150             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
1151
1152             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
1153
1154             /* Update potential sum for this i atom from the interaction with this j atom. */
1155             velec            = _mm_and_pd(velec,cutoff_mask);
1156             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1157             velecsum         = _mm_add_pd(velecsum,velec);
1158
1159             fscal            = felec;
1160
1161             fscal            = _mm_and_pd(fscal,cutoff_mask);
1162
1163             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1164
1165             /* Calculate temporary vectorial force */
1166             tx               = _mm_mul_pd(fscal,dx33);
1167             ty               = _mm_mul_pd(fscal,dy33);
1168             tz               = _mm_mul_pd(fscal,dz33);
1169
1170             /* Update vectorial force */
1171             fix3             = _mm_add_pd(fix3,tx);
1172             fiy3             = _mm_add_pd(fiy3,ty);
1173             fiz3             = _mm_add_pd(fiz3,tz);
1174
1175             fjx3             = _mm_add_pd(fjx3,tx);
1176             fjy3             = _mm_add_pd(fjy3,ty);
1177             fjz3             = _mm_add_pd(fjz3,tz);
1178
1179             }
1180
1181             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1182
1183             /* Inner loop uses 368 flops */
1184         }
1185
1186         /* End of innermost loop */
1187
1188         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1189                                               f+i_coord_offset,fshift+i_shift_offset);
1190
1191         ggid                        = gid[iidx];
1192         /* Update potential energies */
1193         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1194         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1195
1196         /* Increment number of inner iterations */
1197         inneriter                  += j_index_end - j_index_start;
1198
1199         /* Outer loop uses 26 flops */
1200     }
1201
1202     /* Increment number of outer iterations */
1203     outeriter        += nri;
1204
1205     /* Update outer/inner flops */
1206
1207     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*368);
1208 }
1209 /*
1210  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomW4W4_F_sse4_1_double
1211  * Electrostatics interaction: ReactionField
1212  * VdW interaction:            LennardJones
1213  * Geometry:                   Water4-Water4
1214  * Calculate force/pot:        Force
1215  */
1216 void
1217 nb_kernel_ElecRFCut_VdwLJSh_GeomW4W4_F_sse4_1_double
1218                     (t_nblist                    * gmx_restrict       nlist,
1219                      rvec                        * gmx_restrict          xx,
1220                      rvec                        * gmx_restrict          ff,
1221                      t_forcerec                  * gmx_restrict          fr,
1222                      t_mdatoms                   * gmx_restrict     mdatoms,
1223                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1224                      t_nrnb                      * gmx_restrict        nrnb)
1225 {
1226     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1227      * just 0 for non-waters.
1228      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1229      * jnr indices corresponding to data put in the four positions in the SIMD register.
1230      */
1231     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
1232     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1233     int              jnrA,jnrB;
1234     int              j_coord_offsetA,j_coord_offsetB;
1235     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
1236     real             rcutoff_scalar;
1237     real             *shiftvec,*fshift,*x,*f;
1238     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1239     int              vdwioffset0;
1240     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1241     int              vdwioffset1;
1242     __m128d          ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1243     int              vdwioffset2;
1244     __m128d          ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1245     int              vdwioffset3;
1246     __m128d          ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1247     int              vdwjidx0A,vdwjidx0B;
1248     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1249     int              vdwjidx1A,vdwjidx1B;
1250     __m128d          jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1251     int              vdwjidx2A,vdwjidx2B;
1252     __m128d          jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1253     int              vdwjidx3A,vdwjidx3B;
1254     __m128d          jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1255     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1256     __m128d          dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1257     __m128d          dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1258     __m128d          dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1259     __m128d          dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1260     __m128d          dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1261     __m128d          dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1262     __m128d          dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1263     __m128d          dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1264     __m128d          dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1265     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
1266     real             *charge;
1267     int              nvdwtype;
1268     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1269     int              *vdwtype;
1270     real             *vdwparam;
1271     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
1272     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
1273     __m128d          dummy_mask,cutoff_mask;
1274     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1275     __m128d          one     = _mm_set1_pd(1.0);
1276     __m128d          two     = _mm_set1_pd(2.0);
1277     x                = xx[0];
1278     f                = ff[0];
1279
1280     nri              = nlist->nri;
1281     iinr             = nlist->iinr;
1282     jindex           = nlist->jindex;
1283     jjnr             = nlist->jjnr;
1284     shiftidx         = nlist->shift;
1285     gid              = nlist->gid;
1286     shiftvec         = fr->shift_vec[0];
1287     fshift           = fr->fshift[0];
1288     facel            = _mm_set1_pd(fr->epsfac);
1289     charge           = mdatoms->chargeA;
1290     krf              = _mm_set1_pd(fr->ic->k_rf);
1291     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
1292     crf              = _mm_set1_pd(fr->ic->c_rf);
1293     nvdwtype         = fr->ntype;
1294     vdwparam         = fr->nbfp;
1295     vdwtype          = mdatoms->typeA;
1296
1297     /* Setup water-specific parameters */
1298     inr              = nlist->iinr[0];
1299     iq1              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1300     iq2              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1301     iq3              = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1302     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
1303
1304     jq1              = _mm_set1_pd(charge[inr+1]);
1305     jq2              = _mm_set1_pd(charge[inr+2]);
1306     jq3              = _mm_set1_pd(charge[inr+3]);
1307     vdwjidx0A        = 2*vdwtype[inr+0];
1308     c6_00            = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1309     c12_00           = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1310     qq11             = _mm_mul_pd(iq1,jq1);
1311     qq12             = _mm_mul_pd(iq1,jq2);
1312     qq13             = _mm_mul_pd(iq1,jq3);
1313     qq21             = _mm_mul_pd(iq2,jq1);
1314     qq22             = _mm_mul_pd(iq2,jq2);
1315     qq23             = _mm_mul_pd(iq2,jq3);
1316     qq31             = _mm_mul_pd(iq3,jq1);
1317     qq32             = _mm_mul_pd(iq3,jq2);
1318     qq33             = _mm_mul_pd(iq3,jq3);
1319
1320     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
1321     rcutoff_scalar   = fr->rcoulomb;
1322     rcutoff          = _mm_set1_pd(rcutoff_scalar);
1323     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
1324
1325     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
1326     rvdw             = _mm_set1_pd(fr->rvdw);
1327
1328     /* Avoid stupid compiler warnings */
1329     jnrA = jnrB = 0;
1330     j_coord_offsetA = 0;
1331     j_coord_offsetB = 0;
1332
1333     outeriter        = 0;
1334     inneriter        = 0;
1335
1336     /* Start outer loop over neighborlists */
1337     for(iidx=0; iidx<nri; iidx++)
1338     {
1339         /* Load shift vector for this list */
1340         i_shift_offset   = DIM*shiftidx[iidx];
1341
1342         /* Load limits for loop over neighbors */
1343         j_index_start    = jindex[iidx];
1344         j_index_end      = jindex[iidx+1];
1345
1346         /* Get outer coordinate index */
1347         inr              = iinr[iidx];
1348         i_coord_offset   = DIM*inr;
1349
1350         /* Load i particle coords and add shift vector */
1351         gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1352                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1353
1354         fix0             = _mm_setzero_pd();
1355         fiy0             = _mm_setzero_pd();
1356         fiz0             = _mm_setzero_pd();
1357         fix1             = _mm_setzero_pd();
1358         fiy1             = _mm_setzero_pd();
1359         fiz1             = _mm_setzero_pd();
1360         fix2             = _mm_setzero_pd();
1361         fiy2             = _mm_setzero_pd();
1362         fiz2             = _mm_setzero_pd();
1363         fix3             = _mm_setzero_pd();
1364         fiy3             = _mm_setzero_pd();
1365         fiz3             = _mm_setzero_pd();
1366
1367         /* Start inner kernel loop */
1368         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1369         {
1370
1371             /* Get j neighbor index, and coordinate index */
1372             jnrA             = jjnr[jidx];
1373             jnrB             = jjnr[jidx+1];
1374             j_coord_offsetA  = DIM*jnrA;
1375             j_coord_offsetB  = DIM*jnrB;
1376
1377             /* load j atom coordinates */
1378             gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1379                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1380                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1381
1382             /* Calculate displacement vector */
1383             dx00             = _mm_sub_pd(ix0,jx0);
1384             dy00             = _mm_sub_pd(iy0,jy0);
1385             dz00             = _mm_sub_pd(iz0,jz0);
1386             dx11             = _mm_sub_pd(ix1,jx1);
1387             dy11             = _mm_sub_pd(iy1,jy1);
1388             dz11             = _mm_sub_pd(iz1,jz1);
1389             dx12             = _mm_sub_pd(ix1,jx2);
1390             dy12             = _mm_sub_pd(iy1,jy2);
1391             dz12             = _mm_sub_pd(iz1,jz2);
1392             dx13             = _mm_sub_pd(ix1,jx3);
1393             dy13             = _mm_sub_pd(iy1,jy3);
1394             dz13             = _mm_sub_pd(iz1,jz3);
1395             dx21             = _mm_sub_pd(ix2,jx1);
1396             dy21             = _mm_sub_pd(iy2,jy1);
1397             dz21             = _mm_sub_pd(iz2,jz1);
1398             dx22             = _mm_sub_pd(ix2,jx2);
1399             dy22             = _mm_sub_pd(iy2,jy2);
1400             dz22             = _mm_sub_pd(iz2,jz2);
1401             dx23             = _mm_sub_pd(ix2,jx3);
1402             dy23             = _mm_sub_pd(iy2,jy3);
1403             dz23             = _mm_sub_pd(iz2,jz3);
1404             dx31             = _mm_sub_pd(ix3,jx1);
1405             dy31             = _mm_sub_pd(iy3,jy1);
1406             dz31             = _mm_sub_pd(iz3,jz1);
1407             dx32             = _mm_sub_pd(ix3,jx2);
1408             dy32             = _mm_sub_pd(iy3,jy2);
1409             dz32             = _mm_sub_pd(iz3,jz2);
1410             dx33             = _mm_sub_pd(ix3,jx3);
1411             dy33             = _mm_sub_pd(iy3,jy3);
1412             dz33             = _mm_sub_pd(iz3,jz3);
1413
1414             /* Calculate squared distance and things based on it */
1415             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1416             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1417             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1418             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1419             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1420             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1421             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1422             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1423             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1424             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1425
1426             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1427             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1428             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1429             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1430             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1431             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1432             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1433             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1434             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1435
1436             rinvsq00         = gmx_mm_inv_pd(rsq00);
1437             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1438             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1439             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1440             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1441             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1442             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1443             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1444             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1445             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1446
1447             fjx0             = _mm_setzero_pd();
1448             fjy0             = _mm_setzero_pd();
1449             fjz0             = _mm_setzero_pd();
1450             fjx1             = _mm_setzero_pd();
1451             fjy1             = _mm_setzero_pd();
1452             fjz1             = _mm_setzero_pd();
1453             fjx2             = _mm_setzero_pd();
1454             fjy2             = _mm_setzero_pd();
1455             fjz2             = _mm_setzero_pd();
1456             fjx3             = _mm_setzero_pd();
1457             fjy3             = _mm_setzero_pd();
1458             fjz3             = _mm_setzero_pd();
1459
1460             /**************************
1461              * CALCULATE INTERACTIONS *
1462              **************************/
1463
1464             if (gmx_mm_any_lt(rsq00,rcutoff2))
1465             {
1466
1467             /* LENNARD-JONES DISPERSION/REPULSION */
1468
1469             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1470             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1471
1472             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1473
1474             fscal            = fvdw;
1475
1476             fscal            = _mm_and_pd(fscal,cutoff_mask);
1477
1478             /* Calculate temporary vectorial force */
1479             tx               = _mm_mul_pd(fscal,dx00);
1480             ty               = _mm_mul_pd(fscal,dy00);
1481             tz               = _mm_mul_pd(fscal,dz00);
1482
1483             /* Update vectorial force */
1484             fix0             = _mm_add_pd(fix0,tx);
1485             fiy0             = _mm_add_pd(fiy0,ty);
1486             fiz0             = _mm_add_pd(fiz0,tz);
1487
1488             fjx0             = _mm_add_pd(fjx0,tx);
1489             fjy0             = _mm_add_pd(fjy0,ty);
1490             fjz0             = _mm_add_pd(fjz0,tz);
1491
1492             }
1493
1494             /**************************
1495              * CALCULATE INTERACTIONS *
1496              **************************/
1497
1498             if (gmx_mm_any_lt(rsq11,rcutoff2))
1499             {
1500
1501             /* REACTION-FIELD ELECTROSTATICS */
1502             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
1503
1504             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1505
1506             fscal            = felec;
1507
1508             fscal            = _mm_and_pd(fscal,cutoff_mask);
1509
1510             /* Calculate temporary vectorial force */
1511             tx               = _mm_mul_pd(fscal,dx11);
1512             ty               = _mm_mul_pd(fscal,dy11);
1513             tz               = _mm_mul_pd(fscal,dz11);
1514
1515             /* Update vectorial force */
1516             fix1             = _mm_add_pd(fix1,tx);
1517             fiy1             = _mm_add_pd(fiy1,ty);
1518             fiz1             = _mm_add_pd(fiz1,tz);
1519
1520             fjx1             = _mm_add_pd(fjx1,tx);
1521             fjy1             = _mm_add_pd(fjy1,ty);
1522             fjz1             = _mm_add_pd(fjz1,tz);
1523
1524             }
1525
1526             /**************************
1527              * CALCULATE INTERACTIONS *
1528              **************************/
1529
1530             if (gmx_mm_any_lt(rsq12,rcutoff2))
1531             {
1532
1533             /* REACTION-FIELD ELECTROSTATICS */
1534             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1535
1536             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1537
1538             fscal            = felec;
1539
1540             fscal            = _mm_and_pd(fscal,cutoff_mask);
1541
1542             /* Calculate temporary vectorial force */
1543             tx               = _mm_mul_pd(fscal,dx12);
1544             ty               = _mm_mul_pd(fscal,dy12);
1545             tz               = _mm_mul_pd(fscal,dz12);
1546
1547             /* Update vectorial force */
1548             fix1             = _mm_add_pd(fix1,tx);
1549             fiy1             = _mm_add_pd(fiy1,ty);
1550             fiz1             = _mm_add_pd(fiz1,tz);
1551
1552             fjx2             = _mm_add_pd(fjx2,tx);
1553             fjy2             = _mm_add_pd(fjy2,ty);
1554             fjz2             = _mm_add_pd(fjz2,tz);
1555
1556             }
1557
1558             /**************************
1559              * CALCULATE INTERACTIONS *
1560              **************************/
1561
1562             if (gmx_mm_any_lt(rsq13,rcutoff2))
1563             {
1564
1565             /* REACTION-FIELD ELECTROSTATICS */
1566             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
1567
1568             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
1569
1570             fscal            = felec;
1571
1572             fscal            = _mm_and_pd(fscal,cutoff_mask);
1573
1574             /* Calculate temporary vectorial force */
1575             tx               = _mm_mul_pd(fscal,dx13);
1576             ty               = _mm_mul_pd(fscal,dy13);
1577             tz               = _mm_mul_pd(fscal,dz13);
1578
1579             /* Update vectorial force */
1580             fix1             = _mm_add_pd(fix1,tx);
1581             fiy1             = _mm_add_pd(fiy1,ty);
1582             fiz1             = _mm_add_pd(fiz1,tz);
1583
1584             fjx3             = _mm_add_pd(fjx3,tx);
1585             fjy3             = _mm_add_pd(fjy3,ty);
1586             fjz3             = _mm_add_pd(fjz3,tz);
1587
1588             }
1589
1590             /**************************
1591              * CALCULATE INTERACTIONS *
1592              **************************/
1593
1594             if (gmx_mm_any_lt(rsq21,rcutoff2))
1595             {
1596
1597             /* REACTION-FIELD ELECTROSTATICS */
1598             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
1599
1600             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
1601
1602             fscal            = felec;
1603
1604             fscal            = _mm_and_pd(fscal,cutoff_mask);
1605
1606             /* Calculate temporary vectorial force */
1607             tx               = _mm_mul_pd(fscal,dx21);
1608             ty               = _mm_mul_pd(fscal,dy21);
1609             tz               = _mm_mul_pd(fscal,dz21);
1610
1611             /* Update vectorial force */
1612             fix2             = _mm_add_pd(fix2,tx);
1613             fiy2             = _mm_add_pd(fiy2,ty);
1614             fiz2             = _mm_add_pd(fiz2,tz);
1615
1616             fjx1             = _mm_add_pd(fjx1,tx);
1617             fjy1             = _mm_add_pd(fjy1,ty);
1618             fjz1             = _mm_add_pd(fjz1,tz);
1619
1620             }
1621
1622             /**************************
1623              * CALCULATE INTERACTIONS *
1624              **************************/
1625
1626             if (gmx_mm_any_lt(rsq22,rcutoff2))
1627             {
1628
1629             /* REACTION-FIELD ELECTROSTATICS */
1630             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
1631
1632             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
1633
1634             fscal            = felec;
1635
1636             fscal            = _mm_and_pd(fscal,cutoff_mask);
1637
1638             /* Calculate temporary vectorial force */
1639             tx               = _mm_mul_pd(fscal,dx22);
1640             ty               = _mm_mul_pd(fscal,dy22);
1641             tz               = _mm_mul_pd(fscal,dz22);
1642
1643             /* Update vectorial force */
1644             fix2             = _mm_add_pd(fix2,tx);
1645             fiy2             = _mm_add_pd(fiy2,ty);
1646             fiz2             = _mm_add_pd(fiz2,tz);
1647
1648             fjx2             = _mm_add_pd(fjx2,tx);
1649             fjy2             = _mm_add_pd(fjy2,ty);
1650             fjz2             = _mm_add_pd(fjz2,tz);
1651
1652             }
1653
1654             /**************************
1655              * CALCULATE INTERACTIONS *
1656              **************************/
1657
1658             if (gmx_mm_any_lt(rsq23,rcutoff2))
1659             {
1660
1661             /* REACTION-FIELD ELECTROSTATICS */
1662             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
1663
1664             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
1665
1666             fscal            = felec;
1667
1668             fscal            = _mm_and_pd(fscal,cutoff_mask);
1669
1670             /* Calculate temporary vectorial force */
1671             tx               = _mm_mul_pd(fscal,dx23);
1672             ty               = _mm_mul_pd(fscal,dy23);
1673             tz               = _mm_mul_pd(fscal,dz23);
1674
1675             /* Update vectorial force */
1676             fix2             = _mm_add_pd(fix2,tx);
1677             fiy2             = _mm_add_pd(fiy2,ty);
1678             fiz2             = _mm_add_pd(fiz2,tz);
1679
1680             fjx3             = _mm_add_pd(fjx3,tx);
1681             fjy3             = _mm_add_pd(fjy3,ty);
1682             fjz3             = _mm_add_pd(fjz3,tz);
1683
1684             }
1685
1686             /**************************
1687              * CALCULATE INTERACTIONS *
1688              **************************/
1689
1690             if (gmx_mm_any_lt(rsq31,rcutoff2))
1691             {
1692
1693             /* REACTION-FIELD ELECTROSTATICS */
1694             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
1695
1696             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
1697
1698             fscal            = felec;
1699
1700             fscal            = _mm_and_pd(fscal,cutoff_mask);
1701
1702             /* Calculate temporary vectorial force */
1703             tx               = _mm_mul_pd(fscal,dx31);
1704             ty               = _mm_mul_pd(fscal,dy31);
1705             tz               = _mm_mul_pd(fscal,dz31);
1706
1707             /* Update vectorial force */
1708             fix3             = _mm_add_pd(fix3,tx);
1709             fiy3             = _mm_add_pd(fiy3,ty);
1710             fiz3             = _mm_add_pd(fiz3,tz);
1711
1712             fjx1             = _mm_add_pd(fjx1,tx);
1713             fjy1             = _mm_add_pd(fjy1,ty);
1714             fjz1             = _mm_add_pd(fjz1,tz);
1715
1716             }
1717
1718             /**************************
1719              * CALCULATE INTERACTIONS *
1720              **************************/
1721
1722             if (gmx_mm_any_lt(rsq32,rcutoff2))
1723             {
1724
1725             /* REACTION-FIELD ELECTROSTATICS */
1726             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
1727
1728             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
1729
1730             fscal            = felec;
1731
1732             fscal            = _mm_and_pd(fscal,cutoff_mask);
1733
1734             /* Calculate temporary vectorial force */
1735             tx               = _mm_mul_pd(fscal,dx32);
1736             ty               = _mm_mul_pd(fscal,dy32);
1737             tz               = _mm_mul_pd(fscal,dz32);
1738
1739             /* Update vectorial force */
1740             fix3             = _mm_add_pd(fix3,tx);
1741             fiy3             = _mm_add_pd(fiy3,ty);
1742             fiz3             = _mm_add_pd(fiz3,tz);
1743
1744             fjx2             = _mm_add_pd(fjx2,tx);
1745             fjy2             = _mm_add_pd(fjy2,ty);
1746             fjz2             = _mm_add_pd(fjz2,tz);
1747
1748             }
1749
1750             /**************************
1751              * CALCULATE INTERACTIONS *
1752              **************************/
1753
1754             if (gmx_mm_any_lt(rsq33,rcutoff2))
1755             {
1756
1757             /* REACTION-FIELD ELECTROSTATICS */
1758             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
1759
1760             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
1761
1762             fscal            = felec;
1763
1764             fscal            = _mm_and_pd(fscal,cutoff_mask);
1765
1766             /* Calculate temporary vectorial force */
1767             tx               = _mm_mul_pd(fscal,dx33);
1768             ty               = _mm_mul_pd(fscal,dy33);
1769             tz               = _mm_mul_pd(fscal,dz33);
1770
1771             /* Update vectorial force */
1772             fix3             = _mm_add_pd(fix3,tx);
1773             fiy3             = _mm_add_pd(fiy3,ty);
1774             fiz3             = _mm_add_pd(fiz3,tz);
1775
1776             fjx3             = _mm_add_pd(fjx3,tx);
1777             fjy3             = _mm_add_pd(fjy3,ty);
1778             fjz3             = _mm_add_pd(fjz3,tz);
1779
1780             }
1781
1782             gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1783
1784             /* Inner loop uses 303 flops */
1785         }
1786
1787         if(jidx<j_index_end)
1788         {
1789
1790             jnrA             = jjnr[jidx];
1791             j_coord_offsetA  = DIM*jnrA;
1792
1793             /* load j atom coordinates */
1794             gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1795                                               &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1796                                               &jy2,&jz2,&jx3,&jy3,&jz3);
1797
1798             /* Calculate displacement vector */
1799             dx00             = _mm_sub_pd(ix0,jx0);
1800             dy00             = _mm_sub_pd(iy0,jy0);
1801             dz00             = _mm_sub_pd(iz0,jz0);
1802             dx11             = _mm_sub_pd(ix1,jx1);
1803             dy11             = _mm_sub_pd(iy1,jy1);
1804             dz11             = _mm_sub_pd(iz1,jz1);
1805             dx12             = _mm_sub_pd(ix1,jx2);
1806             dy12             = _mm_sub_pd(iy1,jy2);
1807             dz12             = _mm_sub_pd(iz1,jz2);
1808             dx13             = _mm_sub_pd(ix1,jx3);
1809             dy13             = _mm_sub_pd(iy1,jy3);
1810             dz13             = _mm_sub_pd(iz1,jz3);
1811             dx21             = _mm_sub_pd(ix2,jx1);
1812             dy21             = _mm_sub_pd(iy2,jy1);
1813             dz21             = _mm_sub_pd(iz2,jz1);
1814             dx22             = _mm_sub_pd(ix2,jx2);
1815             dy22             = _mm_sub_pd(iy2,jy2);
1816             dz22             = _mm_sub_pd(iz2,jz2);
1817             dx23             = _mm_sub_pd(ix2,jx3);
1818             dy23             = _mm_sub_pd(iy2,jy3);
1819             dz23             = _mm_sub_pd(iz2,jz3);
1820             dx31             = _mm_sub_pd(ix3,jx1);
1821             dy31             = _mm_sub_pd(iy3,jy1);
1822             dz31             = _mm_sub_pd(iz3,jz1);
1823             dx32             = _mm_sub_pd(ix3,jx2);
1824             dy32             = _mm_sub_pd(iy3,jy2);
1825             dz32             = _mm_sub_pd(iz3,jz2);
1826             dx33             = _mm_sub_pd(ix3,jx3);
1827             dy33             = _mm_sub_pd(iy3,jy3);
1828             dz33             = _mm_sub_pd(iz3,jz3);
1829
1830             /* Calculate squared distance and things based on it */
1831             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1832             rsq11            = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1833             rsq12            = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1834             rsq13            = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1835             rsq21            = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1836             rsq22            = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1837             rsq23            = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1838             rsq31            = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1839             rsq32            = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1840             rsq33            = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1841
1842             rinv11           = gmx_mm_invsqrt_pd(rsq11);
1843             rinv12           = gmx_mm_invsqrt_pd(rsq12);
1844             rinv13           = gmx_mm_invsqrt_pd(rsq13);
1845             rinv21           = gmx_mm_invsqrt_pd(rsq21);
1846             rinv22           = gmx_mm_invsqrt_pd(rsq22);
1847             rinv23           = gmx_mm_invsqrt_pd(rsq23);
1848             rinv31           = gmx_mm_invsqrt_pd(rsq31);
1849             rinv32           = gmx_mm_invsqrt_pd(rsq32);
1850             rinv33           = gmx_mm_invsqrt_pd(rsq33);
1851
1852             rinvsq00         = gmx_mm_inv_pd(rsq00);
1853             rinvsq11         = _mm_mul_pd(rinv11,rinv11);
1854             rinvsq12         = _mm_mul_pd(rinv12,rinv12);
1855             rinvsq13         = _mm_mul_pd(rinv13,rinv13);
1856             rinvsq21         = _mm_mul_pd(rinv21,rinv21);
1857             rinvsq22         = _mm_mul_pd(rinv22,rinv22);
1858             rinvsq23         = _mm_mul_pd(rinv23,rinv23);
1859             rinvsq31         = _mm_mul_pd(rinv31,rinv31);
1860             rinvsq32         = _mm_mul_pd(rinv32,rinv32);
1861             rinvsq33         = _mm_mul_pd(rinv33,rinv33);
1862
1863             fjx0             = _mm_setzero_pd();
1864             fjy0             = _mm_setzero_pd();
1865             fjz0             = _mm_setzero_pd();
1866             fjx1             = _mm_setzero_pd();
1867             fjy1             = _mm_setzero_pd();
1868             fjz1             = _mm_setzero_pd();
1869             fjx2             = _mm_setzero_pd();
1870             fjy2             = _mm_setzero_pd();
1871             fjz2             = _mm_setzero_pd();
1872             fjx3             = _mm_setzero_pd();
1873             fjy3             = _mm_setzero_pd();
1874             fjz3             = _mm_setzero_pd();
1875
1876             /**************************
1877              * CALCULATE INTERACTIONS *
1878              **************************/
1879
1880             if (gmx_mm_any_lt(rsq00,rcutoff2))
1881             {
1882
1883             /* LENNARD-JONES DISPERSION/REPULSION */
1884
1885             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1886             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
1887
1888             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
1889
1890             fscal            = fvdw;
1891
1892             fscal            = _mm_and_pd(fscal,cutoff_mask);
1893
1894             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1895
1896             /* Calculate temporary vectorial force */
1897             tx               = _mm_mul_pd(fscal,dx00);
1898             ty               = _mm_mul_pd(fscal,dy00);
1899             tz               = _mm_mul_pd(fscal,dz00);
1900
1901             /* Update vectorial force */
1902             fix0             = _mm_add_pd(fix0,tx);
1903             fiy0             = _mm_add_pd(fiy0,ty);
1904             fiz0             = _mm_add_pd(fiz0,tz);
1905
1906             fjx0             = _mm_add_pd(fjx0,tx);
1907             fjy0             = _mm_add_pd(fjy0,ty);
1908             fjz0             = _mm_add_pd(fjz0,tz);
1909
1910             }
1911
1912             /**************************
1913              * CALCULATE INTERACTIONS *
1914              **************************/
1915
1916             if (gmx_mm_any_lt(rsq11,rcutoff2))
1917             {
1918
1919             /* REACTION-FIELD ELECTROSTATICS */
1920             felec            = _mm_mul_pd(qq11,_mm_sub_pd(_mm_mul_pd(rinv11,rinvsq11),krf2));
1921
1922             cutoff_mask      = _mm_cmplt_pd(rsq11,rcutoff2);
1923
1924             fscal            = felec;
1925
1926             fscal            = _mm_and_pd(fscal,cutoff_mask);
1927
1928             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1929
1930             /* Calculate temporary vectorial force */
1931             tx               = _mm_mul_pd(fscal,dx11);
1932             ty               = _mm_mul_pd(fscal,dy11);
1933             tz               = _mm_mul_pd(fscal,dz11);
1934
1935             /* Update vectorial force */
1936             fix1             = _mm_add_pd(fix1,tx);
1937             fiy1             = _mm_add_pd(fiy1,ty);
1938             fiz1             = _mm_add_pd(fiz1,tz);
1939
1940             fjx1             = _mm_add_pd(fjx1,tx);
1941             fjy1             = _mm_add_pd(fjy1,ty);
1942             fjz1             = _mm_add_pd(fjz1,tz);
1943
1944             }
1945
1946             /**************************
1947              * CALCULATE INTERACTIONS *
1948              **************************/
1949
1950             if (gmx_mm_any_lt(rsq12,rcutoff2))
1951             {
1952
1953             /* REACTION-FIELD ELECTROSTATICS */
1954             felec            = _mm_mul_pd(qq12,_mm_sub_pd(_mm_mul_pd(rinv12,rinvsq12),krf2));
1955
1956             cutoff_mask      = _mm_cmplt_pd(rsq12,rcutoff2);
1957
1958             fscal            = felec;
1959
1960             fscal            = _mm_and_pd(fscal,cutoff_mask);
1961
1962             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1963
1964             /* Calculate temporary vectorial force */
1965             tx               = _mm_mul_pd(fscal,dx12);
1966             ty               = _mm_mul_pd(fscal,dy12);
1967             tz               = _mm_mul_pd(fscal,dz12);
1968
1969             /* Update vectorial force */
1970             fix1             = _mm_add_pd(fix1,tx);
1971             fiy1             = _mm_add_pd(fiy1,ty);
1972             fiz1             = _mm_add_pd(fiz1,tz);
1973
1974             fjx2             = _mm_add_pd(fjx2,tx);
1975             fjy2             = _mm_add_pd(fjy2,ty);
1976             fjz2             = _mm_add_pd(fjz2,tz);
1977
1978             }
1979
1980             /**************************
1981              * CALCULATE INTERACTIONS *
1982              **************************/
1983
1984             if (gmx_mm_any_lt(rsq13,rcutoff2))
1985             {
1986
1987             /* REACTION-FIELD ELECTROSTATICS */
1988             felec            = _mm_mul_pd(qq13,_mm_sub_pd(_mm_mul_pd(rinv13,rinvsq13),krf2));
1989
1990             cutoff_mask      = _mm_cmplt_pd(rsq13,rcutoff2);
1991
1992             fscal            = felec;
1993
1994             fscal            = _mm_and_pd(fscal,cutoff_mask);
1995
1996             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1997
1998             /* Calculate temporary vectorial force */
1999             tx               = _mm_mul_pd(fscal,dx13);
2000             ty               = _mm_mul_pd(fscal,dy13);
2001             tz               = _mm_mul_pd(fscal,dz13);
2002
2003             /* Update vectorial force */
2004             fix1             = _mm_add_pd(fix1,tx);
2005             fiy1             = _mm_add_pd(fiy1,ty);
2006             fiz1             = _mm_add_pd(fiz1,tz);
2007
2008             fjx3             = _mm_add_pd(fjx3,tx);
2009             fjy3             = _mm_add_pd(fjy3,ty);
2010             fjz3             = _mm_add_pd(fjz3,tz);
2011
2012             }
2013
2014             /**************************
2015              * CALCULATE INTERACTIONS *
2016              **************************/
2017
2018             if (gmx_mm_any_lt(rsq21,rcutoff2))
2019             {
2020
2021             /* REACTION-FIELD ELECTROSTATICS */
2022             felec            = _mm_mul_pd(qq21,_mm_sub_pd(_mm_mul_pd(rinv21,rinvsq21),krf2));
2023
2024             cutoff_mask      = _mm_cmplt_pd(rsq21,rcutoff2);
2025
2026             fscal            = felec;
2027
2028             fscal            = _mm_and_pd(fscal,cutoff_mask);
2029
2030             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2031
2032             /* Calculate temporary vectorial force */
2033             tx               = _mm_mul_pd(fscal,dx21);
2034             ty               = _mm_mul_pd(fscal,dy21);
2035             tz               = _mm_mul_pd(fscal,dz21);
2036
2037             /* Update vectorial force */
2038             fix2             = _mm_add_pd(fix2,tx);
2039             fiy2             = _mm_add_pd(fiy2,ty);
2040             fiz2             = _mm_add_pd(fiz2,tz);
2041
2042             fjx1             = _mm_add_pd(fjx1,tx);
2043             fjy1             = _mm_add_pd(fjy1,ty);
2044             fjz1             = _mm_add_pd(fjz1,tz);
2045
2046             }
2047
2048             /**************************
2049              * CALCULATE INTERACTIONS *
2050              **************************/
2051
2052             if (gmx_mm_any_lt(rsq22,rcutoff2))
2053             {
2054
2055             /* REACTION-FIELD ELECTROSTATICS */
2056             felec            = _mm_mul_pd(qq22,_mm_sub_pd(_mm_mul_pd(rinv22,rinvsq22),krf2));
2057
2058             cutoff_mask      = _mm_cmplt_pd(rsq22,rcutoff2);
2059
2060             fscal            = felec;
2061
2062             fscal            = _mm_and_pd(fscal,cutoff_mask);
2063
2064             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2065
2066             /* Calculate temporary vectorial force */
2067             tx               = _mm_mul_pd(fscal,dx22);
2068             ty               = _mm_mul_pd(fscal,dy22);
2069             tz               = _mm_mul_pd(fscal,dz22);
2070
2071             /* Update vectorial force */
2072             fix2             = _mm_add_pd(fix2,tx);
2073             fiy2             = _mm_add_pd(fiy2,ty);
2074             fiz2             = _mm_add_pd(fiz2,tz);
2075
2076             fjx2             = _mm_add_pd(fjx2,tx);
2077             fjy2             = _mm_add_pd(fjy2,ty);
2078             fjz2             = _mm_add_pd(fjz2,tz);
2079
2080             }
2081
2082             /**************************
2083              * CALCULATE INTERACTIONS *
2084              **************************/
2085
2086             if (gmx_mm_any_lt(rsq23,rcutoff2))
2087             {
2088
2089             /* REACTION-FIELD ELECTROSTATICS */
2090             felec            = _mm_mul_pd(qq23,_mm_sub_pd(_mm_mul_pd(rinv23,rinvsq23),krf2));
2091
2092             cutoff_mask      = _mm_cmplt_pd(rsq23,rcutoff2);
2093
2094             fscal            = felec;
2095
2096             fscal            = _mm_and_pd(fscal,cutoff_mask);
2097
2098             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2099
2100             /* Calculate temporary vectorial force */
2101             tx               = _mm_mul_pd(fscal,dx23);
2102             ty               = _mm_mul_pd(fscal,dy23);
2103             tz               = _mm_mul_pd(fscal,dz23);
2104
2105             /* Update vectorial force */
2106             fix2             = _mm_add_pd(fix2,tx);
2107             fiy2             = _mm_add_pd(fiy2,ty);
2108             fiz2             = _mm_add_pd(fiz2,tz);
2109
2110             fjx3             = _mm_add_pd(fjx3,tx);
2111             fjy3             = _mm_add_pd(fjy3,ty);
2112             fjz3             = _mm_add_pd(fjz3,tz);
2113
2114             }
2115
2116             /**************************
2117              * CALCULATE INTERACTIONS *
2118              **************************/
2119
2120             if (gmx_mm_any_lt(rsq31,rcutoff2))
2121             {
2122
2123             /* REACTION-FIELD ELECTROSTATICS */
2124             felec            = _mm_mul_pd(qq31,_mm_sub_pd(_mm_mul_pd(rinv31,rinvsq31),krf2));
2125
2126             cutoff_mask      = _mm_cmplt_pd(rsq31,rcutoff2);
2127
2128             fscal            = felec;
2129
2130             fscal            = _mm_and_pd(fscal,cutoff_mask);
2131
2132             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2133
2134             /* Calculate temporary vectorial force */
2135             tx               = _mm_mul_pd(fscal,dx31);
2136             ty               = _mm_mul_pd(fscal,dy31);
2137             tz               = _mm_mul_pd(fscal,dz31);
2138
2139             /* Update vectorial force */
2140             fix3             = _mm_add_pd(fix3,tx);
2141             fiy3             = _mm_add_pd(fiy3,ty);
2142             fiz3             = _mm_add_pd(fiz3,tz);
2143
2144             fjx1             = _mm_add_pd(fjx1,tx);
2145             fjy1             = _mm_add_pd(fjy1,ty);
2146             fjz1             = _mm_add_pd(fjz1,tz);
2147
2148             }
2149
2150             /**************************
2151              * CALCULATE INTERACTIONS *
2152              **************************/
2153
2154             if (gmx_mm_any_lt(rsq32,rcutoff2))
2155             {
2156
2157             /* REACTION-FIELD ELECTROSTATICS */
2158             felec            = _mm_mul_pd(qq32,_mm_sub_pd(_mm_mul_pd(rinv32,rinvsq32),krf2));
2159
2160             cutoff_mask      = _mm_cmplt_pd(rsq32,rcutoff2);
2161
2162             fscal            = felec;
2163
2164             fscal            = _mm_and_pd(fscal,cutoff_mask);
2165
2166             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2167
2168             /* Calculate temporary vectorial force */
2169             tx               = _mm_mul_pd(fscal,dx32);
2170             ty               = _mm_mul_pd(fscal,dy32);
2171             tz               = _mm_mul_pd(fscal,dz32);
2172
2173             /* Update vectorial force */
2174             fix3             = _mm_add_pd(fix3,tx);
2175             fiy3             = _mm_add_pd(fiy3,ty);
2176             fiz3             = _mm_add_pd(fiz3,tz);
2177
2178             fjx2             = _mm_add_pd(fjx2,tx);
2179             fjy2             = _mm_add_pd(fjy2,ty);
2180             fjz2             = _mm_add_pd(fjz2,tz);
2181
2182             }
2183
2184             /**************************
2185              * CALCULATE INTERACTIONS *
2186              **************************/
2187
2188             if (gmx_mm_any_lt(rsq33,rcutoff2))
2189             {
2190
2191             /* REACTION-FIELD ELECTROSTATICS */
2192             felec            = _mm_mul_pd(qq33,_mm_sub_pd(_mm_mul_pd(rinv33,rinvsq33),krf2));
2193
2194             cutoff_mask      = _mm_cmplt_pd(rsq33,rcutoff2);
2195
2196             fscal            = felec;
2197
2198             fscal            = _mm_and_pd(fscal,cutoff_mask);
2199
2200             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2201
2202             /* Calculate temporary vectorial force */
2203             tx               = _mm_mul_pd(fscal,dx33);
2204             ty               = _mm_mul_pd(fscal,dy33);
2205             tz               = _mm_mul_pd(fscal,dz33);
2206
2207             /* Update vectorial force */
2208             fix3             = _mm_add_pd(fix3,tx);
2209             fiy3             = _mm_add_pd(fiy3,ty);
2210             fiz3             = _mm_add_pd(fiz3,tz);
2211
2212             fjx3             = _mm_add_pd(fjx3,tx);
2213             fjy3             = _mm_add_pd(fjy3,ty);
2214             fjz3             = _mm_add_pd(fjz3,tz);
2215
2216             }
2217
2218             gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2219
2220             /* Inner loop uses 303 flops */
2221         }
2222
2223         /* End of innermost loop */
2224
2225         gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2226                                               f+i_coord_offset,fshift+i_shift_offset);
2227
2228         /* Increment number of inner iterations */
2229         inneriter                  += j_index_end - j_index_start;
2230
2231         /* Outer loop uses 24 flops */
2232     }
2233
2234     /* Increment number of outer iterations */
2235     outeriter        += nri;
2236
2237     /* Update outer/inner flops */
2238
2239     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*303);
2240 }