src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRF_VdwCSTab_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRF_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 120     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 121     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     vftab            = kernel_data->table_vdw->data;
 127     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 128
 129     /* Setup water-specific parameters */
 130     inr              = nlist->iinr[0];
 131     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 132     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 133     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 134     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 135
 136     /* Avoid stupid compiler warnings */
 137     jnrA = jnrB = 0;
 138     j_coord_offsetA = 0;
 139     j_coord_offsetB = 0;
 140
 141     outeriter        = 0;
 142     inneriter        = 0;
 143
 144     /* Start outer loop over neighborlists */
 145     for(iidx=0; iidx<nri; iidx++)
 146     {
 147         /* Load shift vector for this list */
 148         i_shift_offset   = DIM*shiftidx[iidx];
 149
 150         /* Load limits for loop over neighbors */
 151         j_index_start    = jindex[iidx];
 152         j_index_end      = jindex[iidx+1];
 153
 154         /* Get outer coordinate index */
 155         inr              = iinr[iidx];
 156         i_coord_offset   = DIM*inr;
 157
 158         /* Load i particle coords and add shift vector */
 159         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 160                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 161
 162         fix0             = _fjsp_setzero_v2r8();
 163         fiy0             = _fjsp_setzero_v2r8();
 164         fiz0             = _fjsp_setzero_v2r8();
 165         fix1             = _fjsp_setzero_v2r8();
 166         fiy1             = _fjsp_setzero_v2r8();
 167         fiz1             = _fjsp_setzero_v2r8();
 168         fix2             = _fjsp_setzero_v2r8();
 169         fiy2             = _fjsp_setzero_v2r8();
 170         fiz2             = _fjsp_setzero_v2r8();
 171
 172         /* Reset potential sums */
 173         velecsum         = _fjsp_setzero_v2r8();
 174         vvdwsum          = _fjsp_setzero_v2r8();
 175
 176         /* Start inner kernel loop */
 177         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 178         {
 179
 180             /* Get j neighbor index, and coordinate index */
 181             jnrA             = jjnr[jidx];
 182             jnrB             = jjnr[jidx+1];
 183             j_coord_offsetA  = DIM*jnrA;
 184             j_coord_offsetB  = DIM*jnrB;
 185
 186             /* load j atom coordinates */
 187             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 188                                               &jx0,&jy0,&jz0);
 189
 190             /* Calculate displacement vector */
 191             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 192             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 193             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 194             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 195             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 196             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 197             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 198             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 199             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 200
 201             /* Calculate squared distance and things based on it */
 202             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 203             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 204             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 205
 206             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 207             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 208             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 209
 210             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 211             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 212             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 213
 214             /* Load parameters for j particles */
 215             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 216             vdwjidx0A        = 2*vdwtype[jnrA+0];
 217             vdwjidx0B        = 2*vdwtype[jnrB+0];
 218
 219             fjx0             = _fjsp_setzero_v2r8();
 220             fjy0             = _fjsp_setzero_v2r8();
 221             fjz0             = _fjsp_setzero_v2r8();
 222
 223             /**************************
 224              * CALCULATE INTERACTIONS *
 225              **************************/
 226
 227             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 228
 229             /* Compute parameters for interactions between i and j atoms */
 230             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 231             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 232                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 233
 234             /* Calculate table index by multiplying r with table scale and truncate to integer */
 235             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 236             itab_tmp         = _fjsp_dtox_v2r8(rt);
 237             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 238             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 239             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 240
 241             vfconv.i[0]     *= 8;
 242             vfconv.i[1]     *= 8;
 243
 244             /* REACTION-FIELD ELECTROSTATICS */
 245             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 246             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 247
 248             /* CUBIC SPLINE TABLE DISPERSION */
 249             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 250             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 251             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 252             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 253             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 254             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 255             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 256             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 257             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 258             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 259             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 260
 261             /* CUBIC SPLINE TABLE REPULSION */
 262             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 263             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 264             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 265             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 266             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 267             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 268             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 269             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 270             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 271             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 272             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 273             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 274             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 275
 276             /* Update potential sum for this i atom from the interaction with this j atom. */
 277             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 278             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 279
 280             fscal            = _fjsp_add_v2r8(felec,fvdw);
 281
 282             /* Update vectorial force */
 283             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 284             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 285             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 286
 287             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 288             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 289             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 290
 291             /**************************
 292              * CALCULATE INTERACTIONS *
 293              **************************/
 294
 295             /* Compute parameters for interactions between i and j atoms */
 296             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 297
 298             /* REACTION-FIELD ELECTROSTATICS */
 299             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 300             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 301
 302             /* Update potential sum for this i atom from the interaction with this j atom. */
 303             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 304
 305             fscal            = felec;
 306
 307             /* Update vectorial force */
 308             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 309             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 310             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 311
 312             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 313             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 314             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 315
 316             /**************************
 317              * CALCULATE INTERACTIONS *
 318              **************************/
 319
 320             /* Compute parameters for interactions between i and j atoms */
 321             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 322
 323             /* REACTION-FIELD ELECTROSTATICS */
 324             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 325             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 326
 327             /* Update potential sum for this i atom from the interaction with this j atom. */
 328             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 329
 330             fscal            = felec;
 331
 332             /* Update vectorial force */
 333             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 334             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 335             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 336
 337             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 338             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 339             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 340
 341             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 342
 343             /* Inner loop uses 143 flops */
 344         }
 345
 346         if(jidx<j_index_end)
 347         {
 348
 349             jnrA             = jjnr[jidx];
 350             j_coord_offsetA  = DIM*jnrA;
 351
 352             /* load j atom coordinates */
 353             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 354                                               &jx0,&jy0,&jz0);
 355
 356             /* Calculate displacement vector */
 357             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 358             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 359             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 360             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 361             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 362             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 363             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 364             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 365             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 366
 367             /* Calculate squared distance and things based on it */
 368             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 369             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 370             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 371
 372             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 373             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 374             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 375
 376             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 377             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 378             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 379
 380             /* Load parameters for j particles */
 381             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 382             vdwjidx0A        = 2*vdwtype[jnrA+0];
 383
 384             fjx0             = _fjsp_setzero_v2r8();
 385             fjy0             = _fjsp_setzero_v2r8();
 386             fjz0             = _fjsp_setzero_v2r8();
 387
 388             /**************************
 389              * CALCULATE INTERACTIONS *
 390              **************************/
 391
 392             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 393
 394             /* Compute parameters for interactions between i and j atoms */
 395             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 396             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 397
 398             /* Calculate table index by multiplying r with table scale and truncate to integer */
 399             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 400             itab_tmp         = _fjsp_dtox_v2r8(rt);
 401             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 402             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 403             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 404
 405             vfconv.i[0]     *= 8;
 406             vfconv.i[1]     *= 8;
 407
 408             /* REACTION-FIELD ELECTROSTATICS */
 409             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 410             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 411
 412             /* CUBIC SPLINE TABLE DISPERSION */
 413             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 414             F                = _fjsp_setzero_v2r8();
 415             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 416             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 417             H                = _fjsp_setzero_v2r8();
 418             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 419             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 420             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 421             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 422             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 423             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 424
 425             /* CUBIC SPLINE TABLE REPULSION */
 426             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 427             F                = _fjsp_setzero_v2r8();
 428             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 429             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 430             H                = _fjsp_setzero_v2r8();
 431             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 432             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 433             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 434             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 435             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 436             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 437             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 438             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 439
 440             /* Update potential sum for this i atom from the interaction with this j atom. */
 441             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 442             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 443             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 444             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 445
 446             fscal            = _fjsp_add_v2r8(felec,fvdw);
 447
 448             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 449
 450             /* Update vectorial force */
 451             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 452             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 453             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 454
 455             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 456             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 457             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 458
 459             /**************************
 460              * CALCULATE INTERACTIONS *
 461              **************************/
 462
 463             /* Compute parameters for interactions between i and j atoms */
 464             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 465
 466             /* REACTION-FIELD ELECTROSTATICS */
 467             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 468             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 469
 470             /* Update potential sum for this i atom from the interaction with this j atom. */
 471             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 472             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 473
 474             fscal            = felec;
 475
 476             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 477
 478             /* Update vectorial force */
 479             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 480             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 481             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 482
 483             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 484             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 485             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 486
 487             /**************************
 488              * CALCULATE INTERACTIONS *
 489              **************************/
 490
 491             /* Compute parameters for interactions between i and j atoms */
 492             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 493
 494             /* REACTION-FIELD ELECTROSTATICS */
 495             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 496             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 497
 498             /* Update potential sum for this i atom from the interaction with this j atom. */
 499             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 500             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 501
 502             fscal            = felec;
 503
 504             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 505
 506             /* Update vectorial force */
 507             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 508             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 509             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 510
 511             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 512             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 513             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 514
 515             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 516
 517             /* Inner loop uses 143 flops */
 518         }
 519
 520         /* End of innermost loop */
 521
 522         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 523                                               f+i_coord_offset,fshift+i_shift_offset);
 524
 525         ggid                        = gid[iidx];
 526         /* Update potential energies */
 527         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 528         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 529
 530         /* Increment number of inner iterations */
 531         inneriter                  += j_index_end - j_index_start;
 532
 533         /* Outer loop uses 20 flops */
 534     }
 535
 536     /* Increment number of outer iterations */
 537     outeriter        += nri;
 538
 539     /* Update outer/inner flops */
 540
 541     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*143);
 542 }
 543 /*
 544  * Gromacs nonbonded kernel:   nb_kernel_ElecRF_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 545  * Electrostatics interaction: ReactionField
 546  * VdW interaction:            CubicSplineTable
 547  * Geometry:                   Water3-Particle
 548  * Calculate force/pot:        Force
 549  */
 550 void
 551 nb_kernel_ElecRF_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 552                     (t_nblist                    * gmx_restrict       nlist,
 553                      rvec                        * gmx_restrict          xx,
 554                      rvec                        * gmx_restrict          ff,
 555                      t_forcerec                  * gmx_restrict          fr,
 556                      t_mdatoms                   * gmx_restrict     mdatoms,
 557                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 558                      t_nrnb                      * gmx_restrict        nrnb)
 559 {
 560     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 561      * just 0 for non-waters.
 562      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 563      * jnr indices corresponding to data put in the four positions in the SIMD register.
 564      */
 565     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 566     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 567     int              jnrA,jnrB;
 568     int              j_coord_offsetA,j_coord_offsetB;
 569     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 570     real             rcutoff_scalar;
 571     real             *shiftvec,*fshift,*x,*f;
 572     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 573     int              vdwioffset0;
 574     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 575     int              vdwioffset1;
 576     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 577     int              vdwioffset2;
 578     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 579     int              vdwjidx0A,vdwjidx0B;
 580     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 581     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 582     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 583     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 584     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 585     real             *charge;
 586     int              nvdwtype;
 587     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 588     int              *vdwtype;
 589     real             *vdwparam;
 590     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 591     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 592     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 593     real             *vftab;
 594     _fjsp_v2r8       itab_tmp;
 595     _fjsp_v2r8       dummy_mask,cutoff_mask;
 596     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 597     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 598     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 599
 600     x                = xx[0];
 601     f                = ff[0];
 602
 603     nri              = nlist->nri;
 604     iinr             = nlist->iinr;
 605     jindex           = nlist->jindex;
 606     jjnr             = nlist->jjnr;
 607     shiftidx         = nlist->shift;
 608     gid              = nlist->gid;
 609     shiftvec         = fr->shift_vec[0];
 610     fshift           = fr->fshift[0];
 611     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 612     charge           = mdatoms->chargeA;
 613     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 614     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 615     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 616     nvdwtype         = fr->ntype;
 617     vdwparam         = fr->nbfp;
 618     vdwtype          = mdatoms->typeA;
 619
 620     vftab            = kernel_data->table_vdw->data;
 621     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 622
 623     /* Setup water-specific parameters */
 624     inr              = nlist->iinr[0];
 625     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 626     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 627     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 628     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 629
 630     /* Avoid stupid compiler warnings */
 631     jnrA = jnrB = 0;
 632     j_coord_offsetA = 0;
 633     j_coord_offsetB = 0;
 634
 635     outeriter        = 0;
 636     inneriter        = 0;
 637
 638     /* Start outer loop over neighborlists */
 639     for(iidx=0; iidx<nri; iidx++)
 640     {
 641         /* Load shift vector for this list */
 642         i_shift_offset   = DIM*shiftidx[iidx];
 643
 644         /* Load limits for loop over neighbors */
 645         j_index_start    = jindex[iidx];
 646         j_index_end      = jindex[iidx+1];
 647
 648         /* Get outer coordinate index */
 649         inr              = iinr[iidx];
 650         i_coord_offset   = DIM*inr;
 651
 652         /* Load i particle coords and add shift vector */
 653         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 654                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 655
 656         fix0             = _fjsp_setzero_v2r8();
 657         fiy0             = _fjsp_setzero_v2r8();
 658         fiz0             = _fjsp_setzero_v2r8();
 659         fix1             = _fjsp_setzero_v2r8();
 660         fiy1             = _fjsp_setzero_v2r8();
 661         fiz1             = _fjsp_setzero_v2r8();
 662         fix2             = _fjsp_setzero_v2r8();
 663         fiy2             = _fjsp_setzero_v2r8();
 664         fiz2             = _fjsp_setzero_v2r8();
 665
 666         /* Start inner kernel loop */
 667         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 668         {
 669
 670             /* Get j neighbor index, and coordinate index */
 671             jnrA             = jjnr[jidx];
 672             jnrB             = jjnr[jidx+1];
 673             j_coord_offsetA  = DIM*jnrA;
 674             j_coord_offsetB  = DIM*jnrB;
 675
 676             /* load j atom coordinates */
 677             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 678                                               &jx0,&jy0,&jz0);
 679
 680             /* Calculate displacement vector */
 681             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 682             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 683             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 684             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 685             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 686             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 687             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 688             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 689             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 690
 691             /* Calculate squared distance and things based on it */
 692             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 693             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 694             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 695
 696             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 697             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 698             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 699
 700             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 701             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 702             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 703
 704             /* Load parameters for j particles */
 705             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 706             vdwjidx0A        = 2*vdwtype[jnrA+0];
 707             vdwjidx0B        = 2*vdwtype[jnrB+0];
 708
 709             fjx0             = _fjsp_setzero_v2r8();
 710             fjy0             = _fjsp_setzero_v2r8();
 711             fjz0             = _fjsp_setzero_v2r8();
 712
 713             /**************************
 714              * CALCULATE INTERACTIONS *
 715              **************************/
 716
 717             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 718
 719             /* Compute parameters for interactions between i and j atoms */
 720             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 721             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 722                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 723
 724             /* Calculate table index by multiplying r with table scale and truncate to integer */
 725             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 726             itab_tmp         = _fjsp_dtox_v2r8(rt);
 727             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 728             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 729             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 730
 731             vfconv.i[0]     *= 8;
 732             vfconv.i[1]     *= 8;
 733
 734             /* REACTION-FIELD ELECTROSTATICS */
 735             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 736
 737             /* CUBIC SPLINE TABLE DISPERSION */
 738             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 739             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 740             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 741             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 742             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 743             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 744             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 745             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 746             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 747
 748             /* CUBIC SPLINE TABLE REPULSION */
 749             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 750             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 751             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 752             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 753             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 754             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 755             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 756             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 757             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 758             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 759
 760             fscal            = _fjsp_add_v2r8(felec,fvdw);
 761
 762             /* Update vectorial force */
 763             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 764             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 765             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 766
 767             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 768             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 769             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 770
 771             /**************************
 772              * CALCULATE INTERACTIONS *
 773              **************************/
 774
 775             /* Compute parameters for interactions between i and j atoms */
 776             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 777
 778             /* REACTION-FIELD ELECTROSTATICS */
 779             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 780
 781             fscal            = felec;
 782
 783             /* Update vectorial force */
 784             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 785             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 786             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 787
 788             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 789             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 790             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 791
 792             /**************************
 793              * CALCULATE INTERACTIONS *
 794              **************************/
 795
 796             /* Compute parameters for interactions between i and j atoms */
 797             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 798
 799             /* REACTION-FIELD ELECTROSTATICS */
 800             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 801
 802             fscal            = felec;
 803
 804             /* Update vectorial force */
 805             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 806             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 807             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 808
 809             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 810             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 811             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 812
 813             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 814
 815             /* Inner loop uses 120 flops */
 816         }
 817
 818         if(jidx<j_index_end)
 819         {
 820
 821             jnrA             = jjnr[jidx];
 822             j_coord_offsetA  = DIM*jnrA;
 823
 824             /* load j atom coordinates */
 825             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 826                                               &jx0,&jy0,&jz0);
 827
 828             /* Calculate displacement vector */
 829             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 830             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 831             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 832             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 833             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 834             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 835             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 836             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 837             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 838
 839             /* Calculate squared distance and things based on it */
 840             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 841             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 842             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 843
 844             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 845             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 846             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 847
 848             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 849             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 850             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 851
 852             /* Load parameters for j particles */
 853             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 854             vdwjidx0A        = 2*vdwtype[jnrA+0];
 855
 856             fjx0             = _fjsp_setzero_v2r8();
 857             fjy0             = _fjsp_setzero_v2r8();
 858             fjz0             = _fjsp_setzero_v2r8();
 859
 860             /**************************
 861              * CALCULATE INTERACTIONS *
 862              **************************/
 863
 864             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 865
 866             /* Compute parameters for interactions between i and j atoms */
 867             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 868             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 869
 870             /* Calculate table index by multiplying r with table scale and truncate to integer */
 871             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 872             itab_tmp         = _fjsp_dtox_v2r8(rt);
 873             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 874             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 875             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 876
 877             vfconv.i[0]     *= 8;
 878             vfconv.i[1]     *= 8;
 879
 880             /* REACTION-FIELD ELECTROSTATICS */
 881             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 882
 883             /* CUBIC SPLINE TABLE DISPERSION */
 884             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 885             F                = _fjsp_setzero_v2r8();
 886             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 887             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 888             H                = _fjsp_setzero_v2r8();
 889             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 890             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 891             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 892             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 893
 894             /* CUBIC SPLINE TABLE REPULSION */
 895             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 896             F                = _fjsp_setzero_v2r8();
 897             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 898             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 899             H                = _fjsp_setzero_v2r8();
 900             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 901             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 902             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 903             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 904             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 905
 906             fscal            = _fjsp_add_v2r8(felec,fvdw);
 907
 908             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 909
 910             /* Update vectorial force */
 911             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 912             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 913             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 914
 915             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 916             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 917             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 918
 919             /**************************
 920              * CALCULATE INTERACTIONS *
 921              **************************/
 922
 923             /* Compute parameters for interactions between i and j atoms */
 924             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 925
 926             /* REACTION-FIELD ELECTROSTATICS */
 927             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 928
 929             fscal            = felec;
 930
 931             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 932
 933             /* Update vectorial force */
 934             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 935             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 936             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 937
 938             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 939             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 940             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 941
 942             /**************************
 943              * CALCULATE INTERACTIONS *
 944              **************************/
 945
 946             /* Compute parameters for interactions between i and j atoms */
 947             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 948
 949             /* REACTION-FIELD ELECTROSTATICS */
 950             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 951
 952             fscal            = felec;
 953
 954             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 955
 956             /* Update vectorial force */
 957             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 958             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 959             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 960
 961             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 962             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 963             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 964
 965             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 966
 967             /* Inner loop uses 120 flops */
 968         }
 969
 970         /* End of innermost loop */
 971
 972         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 973                                               f+i_coord_offset,fshift+i_shift_offset);
 974
 975         /* Increment number of inner iterations */
 976         inneriter                  += j_index_end - j_index_start;
 977
 978         /* Outer loop uses 18 flops */
 979     }
 980
 981     /* Increment number of outer iterations */
 982     outeriter        += nri;
 983
 984     /* Update outer/inner flops */
 985
 986     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*120);
 987 }