src/gromacs/gmxlib/nonbonded/nb_kernel_sse4_1_double/nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_sse4_1_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015, 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 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/math/vec.h"
  46 #include "gromacs/legacyheaders/nrnb.h"
  47
  48 #include "gromacs/simd/math_x86_sse4_1_double.h"
  49 #include "kernelutil_x86_sse4_1_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sse4_1_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_VF_sse4_1_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwjidx0A,vdwjidx0B;
  84     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  85     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  86     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  87     real             *charge;
  88     int              nvdwtype;
  89     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  90     int              *vdwtype;
  91     real             *vdwparam;
  92     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  93     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  94     __m128d          dummy_mask,cutoff_mask;
  95     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  96     __m128d          one     = _mm_set1_pd(1.0);
  97     __m128d          two     = _mm_set1_pd(2.0);
  98     x                = xx[0];
  99     f                = ff[0];
 100
 101     nri              = nlist->nri;
 102     iinr             = nlist->iinr;
 103     jindex           = nlist->jindex;
 104     jjnr             = nlist->jjnr;
 105     shiftidx         = nlist->shift;
 106     gid              = nlist->gid;
 107     shiftvec         = fr->shift_vec[0];
 108     fshift           = fr->fshift[0];
 109     facel            = _mm_set1_pd(fr->epsfac);
 110     charge           = mdatoms->chargeA;
 111     krf              = _mm_set1_pd(fr->ic->k_rf);
 112     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 113     crf              = _mm_set1_pd(fr->ic->c_rf);
 114     nvdwtype         = fr->ntype;
 115     vdwparam         = fr->nbfp;
 116     vdwtype          = mdatoms->typeA;
 117
 118     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 119     rcutoff_scalar   = fr->rcoulomb;
 120     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 121     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 122
 123     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 124     rvdw             = _mm_set1_pd(fr->rvdw);
 125
 126     /* Avoid stupid compiler warnings */
 127     jnrA = jnrB = 0;
 128     j_coord_offsetA = 0;
 129     j_coord_offsetB = 0;
 130
 131     outeriter        = 0;
 132     inneriter        = 0;
 133
 134     /* Start outer loop over neighborlists */
 135     for(iidx=0; iidx<nri; iidx++)
 136     {
 137         /* Load shift vector for this list */
 138         i_shift_offset   = DIM*shiftidx[iidx];
 139
 140         /* Load limits for loop over neighbors */
 141         j_index_start    = jindex[iidx];
 142         j_index_end      = jindex[iidx+1];
 143
 144         /* Get outer coordinate index */
 145         inr              = iinr[iidx];
 146         i_coord_offset   = DIM*inr;
 147
 148         /* Load i particle coords and add shift vector */
 149         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 150
 151         fix0             = _mm_setzero_pd();
 152         fiy0             = _mm_setzero_pd();
 153         fiz0             = _mm_setzero_pd();
 154
 155         /* Load parameters for i particles */
 156         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 157         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 158
 159         /* Reset potential sums */
 160         velecsum         = _mm_setzero_pd();
 161         vvdwsum          = _mm_setzero_pd();
 162
 163         /* Start inner kernel loop */
 164         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 165         {
 166
 167             /* Get j neighbor index, and coordinate index */
 168             jnrA             = jjnr[jidx];
 169             jnrB             = jjnr[jidx+1];
 170             j_coord_offsetA  = DIM*jnrA;
 171             j_coord_offsetB  = DIM*jnrB;
 172
 173             /* load j atom coordinates */
 174             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 175                                               &jx0,&jy0,&jz0);
 176
 177             /* Calculate displacement vector */
 178             dx00             = _mm_sub_pd(ix0,jx0);
 179             dy00             = _mm_sub_pd(iy0,jy0);
 180             dz00             = _mm_sub_pd(iz0,jz0);
 181
 182             /* Calculate squared distance and things based on it */
 183             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 184
 185             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 186
 187             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 188
 189             /* Load parameters for j particles */
 190             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 191             vdwjidx0A        = 2*vdwtype[jnrA+0];
 192             vdwjidx0B        = 2*vdwtype[jnrB+0];
 193
 194             /**************************
 195              * CALCULATE INTERACTIONS *
 196              **************************/
 197
 198             if (gmx_mm_any_lt(rsq00,rcutoff2))
 199             {
 200
 201             /* Compute parameters for interactions between i and j atoms */
 202             qq00             = _mm_mul_pd(iq0,jq0);
 203             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 204                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 205
 206             /* REACTION-FIELD ELECTROSTATICS */
 207             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 208             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 209
 210             /* LENNARD-JONES DISPERSION/REPULSION */
 211
 212             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 213             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 214             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 215             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) ,
 216                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 217             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 218
 219             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 220
 221             /* Update potential sum for this i atom from the interaction with this j atom. */
 222             velec            = _mm_and_pd(velec,cutoff_mask);
 223             velecsum         = _mm_add_pd(velecsum,velec);
 224             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 225             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 226
 227             fscal            = _mm_add_pd(felec,fvdw);
 228
 229             fscal            = _mm_and_pd(fscal,cutoff_mask);
 230
 231             /* Calculate temporary vectorial force */
 232             tx               = _mm_mul_pd(fscal,dx00);
 233             ty               = _mm_mul_pd(fscal,dy00);
 234             tz               = _mm_mul_pd(fscal,dz00);
 235
 236             /* Update vectorial force */
 237             fix0             = _mm_add_pd(fix0,tx);
 238             fiy0             = _mm_add_pd(fiy0,ty);
 239             fiz0             = _mm_add_pd(fiz0,tz);
 240
 241             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 242
 243             }
 244
 245             /* Inner loop uses 54 flops */
 246         }
 247
 248         if(jidx<j_index_end)
 249         {
 250
 251             jnrA             = jjnr[jidx];
 252             j_coord_offsetA  = DIM*jnrA;
 253
 254             /* load j atom coordinates */
 255             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 256                                               &jx0,&jy0,&jz0);
 257
 258             /* Calculate displacement vector */
 259             dx00             = _mm_sub_pd(ix0,jx0);
 260             dy00             = _mm_sub_pd(iy0,jy0);
 261             dz00             = _mm_sub_pd(iz0,jz0);
 262
 263             /* Calculate squared distance and things based on it */
 264             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 265
 266             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 267
 268             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 269
 270             /* Load parameters for j particles */
 271             jq0              = _mm_load_sd(charge+jnrA+0);
 272             vdwjidx0A        = 2*vdwtype[jnrA+0];
 273
 274             /**************************
 275              * CALCULATE INTERACTIONS *
 276              **************************/
 277
 278             if (gmx_mm_any_lt(rsq00,rcutoff2))
 279             {
 280
 281             /* Compute parameters for interactions between i and j atoms */
 282             qq00             = _mm_mul_pd(iq0,jq0);
 283             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 284
 285             /* REACTION-FIELD ELECTROSTATICS */
 286             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
 287             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 288
 289             /* LENNARD-JONES DISPERSION/REPULSION */
 290
 291             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 292             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 293             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 294             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) ,
 295                                           _mm_mul_pd( _mm_sub_pd(vvdw6,_mm_mul_pd(c6_00,sh_vdw_invrcut6)),one_sixth));
 296             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 297
 298             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 299
 300             /* Update potential sum for this i atom from the interaction with this j atom. */
 301             velec            = _mm_and_pd(velec,cutoff_mask);
 302             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 303             velecsum         = _mm_add_pd(velecsum,velec);
 304             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 305             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 306             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 307
 308             fscal            = _mm_add_pd(felec,fvdw);
 309
 310             fscal            = _mm_and_pd(fscal,cutoff_mask);
 311
 312             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 313
 314             /* Calculate temporary vectorial force */
 315             tx               = _mm_mul_pd(fscal,dx00);
 316             ty               = _mm_mul_pd(fscal,dy00);
 317             tz               = _mm_mul_pd(fscal,dz00);
 318
 319             /* Update vectorial force */
 320             fix0             = _mm_add_pd(fix0,tx);
 321             fiy0             = _mm_add_pd(fiy0,ty);
 322             fiz0             = _mm_add_pd(fiz0,tz);
 323
 324             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 325
 326             }
 327
 328             /* Inner loop uses 54 flops */
 329         }
 330
 331         /* End of innermost loop */
 332
 333         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 334                                               f+i_coord_offset,fshift+i_shift_offset);
 335
 336         ggid                        = gid[iidx];
 337         /* Update potential energies */
 338         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 339         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 340
 341         /* Increment number of inner iterations */
 342         inneriter                  += j_index_end - j_index_start;
 343
 344         /* Outer loop uses 9 flops */
 345     }
 346
 347     /* Increment number of outer iterations */
 348     outeriter        += nri;
 349
 350     /* Update outer/inner flops */
 351
 352     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*54);
 353 }
 354 /*
 355  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sse4_1_double
 356  * Electrostatics interaction: ReactionField
 357  * VdW interaction:            LennardJones
 358  * Geometry:                   Particle-Particle
 359  * Calculate force/pot:        Force
 360  */
 361 void
 362 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_sse4_1_double
 363                     (t_nblist                    * gmx_restrict       nlist,
 364                      rvec                        * gmx_restrict          xx,
 365                      rvec                        * gmx_restrict          ff,
 366                      t_forcerec                  * gmx_restrict          fr,
 367                      t_mdatoms                   * gmx_restrict     mdatoms,
 368                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 369                      t_nrnb                      * gmx_restrict        nrnb)
 370 {
 371     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 372      * just 0 for non-waters.
 373      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 374      * jnr indices corresponding to data put in the four positions in the SIMD register.
 375      */
 376     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 377     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 378     int              jnrA,jnrB;
 379     int              j_coord_offsetA,j_coord_offsetB;
 380     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 381     real             rcutoff_scalar;
 382     real             *shiftvec,*fshift,*x,*f;
 383     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 384     int              vdwioffset0;
 385     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 386     int              vdwjidx0A,vdwjidx0B;
 387     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 388     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 389     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 390     real             *charge;
 391     int              nvdwtype;
 392     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 393     int              *vdwtype;
 394     real             *vdwparam;
 395     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 396     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 397     __m128d          dummy_mask,cutoff_mask;
 398     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 399     __m128d          one     = _mm_set1_pd(1.0);
 400     __m128d          two     = _mm_set1_pd(2.0);
 401     x                = xx[0];
 402     f                = ff[0];
 403
 404     nri              = nlist->nri;
 405     iinr             = nlist->iinr;
 406     jindex           = nlist->jindex;
 407     jjnr             = nlist->jjnr;
 408     shiftidx         = nlist->shift;
 409     gid              = nlist->gid;
 410     shiftvec         = fr->shift_vec[0];
 411     fshift           = fr->fshift[0];
 412     facel            = _mm_set1_pd(fr->epsfac);
 413     charge           = mdatoms->chargeA;
 414     krf              = _mm_set1_pd(fr->ic->k_rf);
 415     krf2             = _mm_set1_pd(fr->ic->k_rf*2.0);
 416     crf              = _mm_set1_pd(fr->ic->c_rf);
 417     nvdwtype         = fr->ntype;
 418     vdwparam         = fr->nbfp;
 419     vdwtype          = mdatoms->typeA;
 420
 421     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 422     rcutoff_scalar   = fr->rcoulomb;
 423     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 424     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 425
 426     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 427     rvdw             = _mm_set1_pd(fr->rvdw);
 428
 429     /* Avoid stupid compiler warnings */
 430     jnrA = jnrB = 0;
 431     j_coord_offsetA = 0;
 432     j_coord_offsetB = 0;
 433
 434     outeriter        = 0;
 435     inneriter        = 0;
 436
 437     /* Start outer loop over neighborlists */
 438     for(iidx=0; iidx<nri; iidx++)
 439     {
 440         /* Load shift vector for this list */
 441         i_shift_offset   = DIM*shiftidx[iidx];
 442
 443         /* Load limits for loop over neighbors */
 444         j_index_start    = jindex[iidx];
 445         j_index_end      = jindex[iidx+1];
 446
 447         /* Get outer coordinate index */
 448         inr              = iinr[iidx];
 449         i_coord_offset   = DIM*inr;
 450
 451         /* Load i particle coords and add shift vector */
 452         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 453
 454         fix0             = _mm_setzero_pd();
 455         fiy0             = _mm_setzero_pd();
 456         fiz0             = _mm_setzero_pd();
 457
 458         /* Load parameters for i particles */
 459         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 460         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 461
 462         /* Start inner kernel loop */
 463         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 464         {
 465
 466             /* Get j neighbor index, and coordinate index */
 467             jnrA             = jjnr[jidx];
 468             jnrB             = jjnr[jidx+1];
 469             j_coord_offsetA  = DIM*jnrA;
 470             j_coord_offsetB  = DIM*jnrB;
 471
 472             /* load j atom coordinates */
 473             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 474                                               &jx0,&jy0,&jz0);
 475
 476             /* Calculate displacement vector */
 477             dx00             = _mm_sub_pd(ix0,jx0);
 478             dy00             = _mm_sub_pd(iy0,jy0);
 479             dz00             = _mm_sub_pd(iz0,jz0);
 480
 481             /* Calculate squared distance and things based on it */
 482             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 483
 484             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 485
 486             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 487
 488             /* Load parameters for j particles */
 489             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 490             vdwjidx0A        = 2*vdwtype[jnrA+0];
 491             vdwjidx0B        = 2*vdwtype[jnrB+0];
 492
 493             /**************************
 494              * CALCULATE INTERACTIONS *
 495              **************************/
 496
 497             if (gmx_mm_any_lt(rsq00,rcutoff2))
 498             {
 499
 500             /* Compute parameters for interactions between i and j atoms */
 501             qq00             = _mm_mul_pd(iq0,jq0);
 502             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 503                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 504
 505             /* REACTION-FIELD ELECTROSTATICS */
 506             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 507
 508             /* LENNARD-JONES DISPERSION/REPULSION */
 509
 510             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 511             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 512
 513             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 514
 515             fscal            = _mm_add_pd(felec,fvdw);
 516
 517             fscal            = _mm_and_pd(fscal,cutoff_mask);
 518
 519             /* Calculate temporary vectorial force */
 520             tx               = _mm_mul_pd(fscal,dx00);
 521             ty               = _mm_mul_pd(fscal,dy00);
 522             tz               = _mm_mul_pd(fscal,dz00);
 523
 524             /* Update vectorial force */
 525             fix0             = _mm_add_pd(fix0,tx);
 526             fiy0             = _mm_add_pd(fiy0,ty);
 527             fiz0             = _mm_add_pd(fiz0,tz);
 528
 529             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 530
 531             }
 532
 533             /* Inner loop uses 37 flops */
 534         }
 535
 536         if(jidx<j_index_end)
 537         {
 538
 539             jnrA             = jjnr[jidx];
 540             j_coord_offsetA  = DIM*jnrA;
 541
 542             /* load j atom coordinates */
 543             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 544                                               &jx0,&jy0,&jz0);
 545
 546             /* Calculate displacement vector */
 547             dx00             = _mm_sub_pd(ix0,jx0);
 548             dy00             = _mm_sub_pd(iy0,jy0);
 549             dz00             = _mm_sub_pd(iz0,jz0);
 550
 551             /* Calculate squared distance and things based on it */
 552             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 553
 554             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 555
 556             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 557
 558             /* Load parameters for j particles */
 559             jq0              = _mm_load_sd(charge+jnrA+0);
 560             vdwjidx0A        = 2*vdwtype[jnrA+0];
 561
 562             /**************************
 563              * CALCULATE INTERACTIONS *
 564              **************************/
 565
 566             if (gmx_mm_any_lt(rsq00,rcutoff2))
 567             {
 568
 569             /* Compute parameters for interactions between i and j atoms */
 570             qq00             = _mm_mul_pd(iq0,jq0);
 571             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 572
 573             /* REACTION-FIELD ELECTROSTATICS */
 574             felec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
 575
 576             /* LENNARD-JONES DISPERSION/REPULSION */
 577
 578             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 579             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 580
 581             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 582
 583             fscal            = _mm_add_pd(felec,fvdw);
 584
 585             fscal            = _mm_and_pd(fscal,cutoff_mask);
 586
 587             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 588
 589             /* Calculate temporary vectorial force */
 590             tx               = _mm_mul_pd(fscal,dx00);
 591             ty               = _mm_mul_pd(fscal,dy00);
 592             tz               = _mm_mul_pd(fscal,dz00);
 593
 594             /* Update vectorial force */
 595             fix0             = _mm_add_pd(fix0,tx);
 596             fiy0             = _mm_add_pd(fiy0,ty);
 597             fiz0             = _mm_add_pd(fiz0,tz);
 598
 599             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 600
 601             }
 602
 603             /* Inner loop uses 37 flops */
 604         }
 605
 606         /* End of innermost loop */
 607
 608         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 609                                               f+i_coord_offset,fshift+i_shift_offset);
 610
 611         /* Increment number of inner iterations */
 612         inneriter                  += j_index_end - j_index_start;
 613
 614         /* Outer loop uses 7 flops */
 615     }
 616
 617     /* Increment number of outer iterations */
 618     outeriter        += nri;
 619
 620     /* Update outer/inner flops */
 621
 622     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*37);
 623 }