src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_single/nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_avx_128_fma_single.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 avx_128_fma_single kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_single.h"
  48 #include "kernelutil_x86_avx_128_fma_single.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_128_fma_single
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Water4-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_VF_avx_128_fma_single
  59                     (t_nblist                    * gmx_restrict       nlist,
  60                      rvec                        * gmx_restrict          xx,
  61                      rvec                        * gmx_restrict          ff,
  62                      t_forcerec                  * gmx_restrict          fr,
  63                      t_mdatoms                   * gmx_restrict     mdatoms,
  64                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  65                      t_nrnb                      * gmx_restrict        nrnb)
  66 {
  67     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  68      * just 0 for non-waters.
  69      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
  70      * jnr indices corresponding to data put in the four positions in the SIMD register.
  71      */
  72     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  73     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  74     int              jnrA,jnrB,jnrC,jnrD;
  75     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
  76     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
  81     real             scratch[4*DIM];
  82     __m128           fscal,rcutoff,rcutoff2,jidxall;
  83     int              vdwioffset0;
  84     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  85     int              vdwioffset1;
  86     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  87     int              vdwioffset2;
  88     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  89     int              vdwioffset3;
  90     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  91     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  92     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  93     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  94     __m128           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  95     __m128           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  96     __m128           dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  97     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
  98     real             *charge;
  99     int              nvdwtype;
 100     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 101     int              *vdwtype;
 102     real             *vdwparam;
 103     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 104     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 105     __m128i          ewitab;
 106     __m128           ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 107     __m128           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 108     real             *ewtab;
 109     __m128           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 110     real             rswitch_scalar,d_scalar;
 111     __m128           dummy_mask,cutoff_mask;
 112     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 113     __m128           one     = _mm_set1_ps(1.0);
 114     __m128           two     = _mm_set1_ps(2.0);
 115     x                = xx[0];
 116     f                = ff[0];
 117
 118     nri              = nlist->nri;
 119     iinr             = nlist->iinr;
 120     jindex           = nlist->jindex;
 121     jjnr             = nlist->jjnr;
 122     shiftidx         = nlist->shift;
 123     gid              = nlist->gid;
 124     shiftvec         = fr->shift_vec[0];
 125     fshift           = fr->fshift[0];
 126     facel            = _mm_set1_ps(fr->epsfac);
 127     charge           = mdatoms->chargeA;
 128     nvdwtype         = fr->ntype;
 129     vdwparam         = fr->nbfp;
 130     vdwtype          = mdatoms->typeA;
 131
 132     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 133     beta             = _mm_set1_ps(fr->ic->ewaldcoeff_q);
 134     beta2            = _mm_mul_ps(beta,beta);
 135     beta3            = _mm_mul_ps(beta,beta2);
 136     ewtab            = fr->ic->tabq_coul_FDV0;
 137     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 138     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 139
 140     /* Setup water-specific parameters */
 141     inr              = nlist->iinr[0];
 142     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 143     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 144     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 145     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 146
 147     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 148     rcutoff_scalar   = fr->rcoulomb;
 149     rcutoff          = _mm_set1_ps(rcutoff_scalar);
 150     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
 151
 152     rswitch_scalar   = fr->rcoulomb_switch;
 153     rswitch          = _mm_set1_ps(rswitch_scalar);
 154     /* Setup switch parameters */
 155     d_scalar         = rcutoff_scalar-rswitch_scalar;
 156     d                = _mm_set1_ps(d_scalar);
 157     swV3             = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
 158     swV4             = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 159     swV5             = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 160     swF2             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
 161     swF3             = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 162     swF4             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 163
 164     /* Avoid stupid compiler warnings */
 165     jnrA = jnrB = jnrC = jnrD = 0;
 166     j_coord_offsetA = 0;
 167     j_coord_offsetB = 0;
 168     j_coord_offsetC = 0;
 169     j_coord_offsetD = 0;
 170
 171     outeriter        = 0;
 172     inneriter        = 0;
 173
 174     for(iidx=0;iidx<4*DIM;iidx++)
 175     {
 176         scratch[iidx] = 0.0;
 177     }
 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_ps(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_ps();
 198         fiy0             = _mm_setzero_ps();
 199         fiz0             = _mm_setzero_ps();
 200         fix1             = _mm_setzero_ps();
 201         fiy1             = _mm_setzero_ps();
 202         fiz1             = _mm_setzero_ps();
 203         fix2             = _mm_setzero_ps();
 204         fiy2             = _mm_setzero_ps();
 205         fiz2             = _mm_setzero_ps();
 206         fix3             = _mm_setzero_ps();
 207         fiy3             = _mm_setzero_ps();
 208         fiz3             = _mm_setzero_ps();
 209
 210         /* Reset potential sums */
 211         velecsum         = _mm_setzero_ps();
 212         vvdwsum          = _mm_setzero_ps();
 213
 214         /* Start inner kernel loop */
 215         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 216         {
 217
 218             /* Get j neighbor index, and coordinate index */
 219             jnrA             = jjnr[jidx];
 220             jnrB             = jjnr[jidx+1];
 221             jnrC             = jjnr[jidx+2];
 222             jnrD             = jjnr[jidx+3];
 223             j_coord_offsetA  = DIM*jnrA;
 224             j_coord_offsetB  = DIM*jnrB;
 225             j_coord_offsetC  = DIM*jnrC;
 226             j_coord_offsetD  = DIM*jnrD;
 227
 228             /* load j atom coordinates */
 229             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 230                                               x+j_coord_offsetC,x+j_coord_offsetD,
 231                                               &jx0,&jy0,&jz0);
 232
 233             /* Calculate displacement vector */
 234             dx00             = _mm_sub_ps(ix0,jx0);
 235             dy00             = _mm_sub_ps(iy0,jy0);
 236             dz00             = _mm_sub_ps(iz0,jz0);
 237             dx10             = _mm_sub_ps(ix1,jx0);
 238             dy10             = _mm_sub_ps(iy1,jy0);
 239             dz10             = _mm_sub_ps(iz1,jz0);
 240             dx20             = _mm_sub_ps(ix2,jx0);
 241             dy20             = _mm_sub_ps(iy2,jy0);
 242             dz20             = _mm_sub_ps(iz2,jz0);
 243             dx30             = _mm_sub_ps(ix3,jx0);
 244             dy30             = _mm_sub_ps(iy3,jy0);
 245             dz30             = _mm_sub_ps(iz3,jz0);
 246
 247             /* Calculate squared distance and things based on it */
 248             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 249             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
 250             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
 251             rsq30            = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
 252
 253             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 254             rinv10           = gmx_mm_invsqrt_ps(rsq10);
 255             rinv20           = gmx_mm_invsqrt_ps(rsq20);
 256             rinv30           = gmx_mm_invsqrt_ps(rsq30);
 257
 258             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 259             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
 260             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
 261             rinvsq30         = _mm_mul_ps(rinv30,rinv30);
 262
 263             /* Load parameters for j particles */
 264             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 265                                                               charge+jnrC+0,charge+jnrD+0);
 266             vdwjidx0A        = 2*vdwtype[jnrA+0];
 267             vdwjidx0B        = 2*vdwtype[jnrB+0];
 268             vdwjidx0C        = 2*vdwtype[jnrC+0];
 269             vdwjidx0D        = 2*vdwtype[jnrD+0];
 270
 271             fjx0             = _mm_setzero_ps();
 272             fjy0             = _mm_setzero_ps();
 273             fjz0             = _mm_setzero_ps();
 274
 275             /**************************
 276              * CALCULATE INTERACTIONS *
 277              **************************/
 278
 279             if (gmx_mm_any_lt(rsq00,rcutoff2))
 280             {
 281
 282             r00              = _mm_mul_ps(rsq00,rinv00);
 283
 284             /* Compute parameters for interactions between i and j atoms */
 285             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
 286                                          vdwparam+vdwioffset0+vdwjidx0B,
 287                                          vdwparam+vdwioffset0+vdwjidx0C,
 288                                          vdwparam+vdwioffset0+vdwjidx0D,
 289                                          &c6_00,&c12_00);
 290
 291             /* LENNARD-JONES DISPERSION/REPULSION */
 292
 293             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 294             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 295             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 296             vvdw             = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
 297             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 298
 299             d                = _mm_sub_ps(r00,rswitch);
 300             d                = _mm_max_ps(d,_mm_setzero_ps());
 301             d2               = _mm_mul_ps(d,d);
 302             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 303
 304             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 305
 306             /* Evaluate switch function */
 307             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 308             fvdw             = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
 309             vvdw             = _mm_mul_ps(vvdw,sw);
 310             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
 311
 312             /* Update potential sum for this i atom from the interaction with this j atom. */
 313             vvdw             = _mm_and_ps(vvdw,cutoff_mask);
 314             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 315
 316             fscal            = fvdw;
 317
 318             fscal            = _mm_and_ps(fscal,cutoff_mask);
 319
 320              /* Update vectorial force */
 321             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 322             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 323             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 324
 325             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 326             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 327             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 328
 329             }
 330
 331             /**************************
 332              * CALCULATE INTERACTIONS *
 333              **************************/
 334
 335             if (gmx_mm_any_lt(rsq10,rcutoff2))
 336             {
 337
 338             r10              = _mm_mul_ps(rsq10,rinv10);
 339
 340             /* Compute parameters for interactions between i and j atoms */
 341             qq10             = _mm_mul_ps(iq1,jq0);
 342
 343             /* EWALD ELECTROSTATICS */
 344
 345             /* Analytical PME correction */
 346             zeta2            = _mm_mul_ps(beta2,rsq10);
 347             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
 348             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 349             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 350             felec            = _mm_mul_ps(qq10,felec);
 351             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 352             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv10);
 353             velec            = _mm_mul_ps(qq10,velec);
 354
 355             d                = _mm_sub_ps(r10,rswitch);
 356             d                = _mm_max_ps(d,_mm_setzero_ps());
 357             d2               = _mm_mul_ps(d,d);
 358             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 359
 360             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 361
 362             /* Evaluate switch function */
 363             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 364             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
 365             velec            = _mm_mul_ps(velec,sw);
 366             cutoff_mask      = _mm_cmplt_ps(rsq10,rcutoff2);
 367
 368             /* Update potential sum for this i atom from the interaction with this j atom. */
 369             velec            = _mm_and_ps(velec,cutoff_mask);
 370             velecsum         = _mm_add_ps(velecsum,velec);
 371
 372             fscal            = felec;
 373
 374             fscal            = _mm_and_ps(fscal,cutoff_mask);
 375
 376              /* Update vectorial force */
 377             fix1             = _mm_macc_ps(dx10,fscal,fix1);
 378             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
 379             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
 380
 381             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
 382             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
 383             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
 384
 385             }
 386
 387             /**************************
 388              * CALCULATE INTERACTIONS *
 389              **************************/
 390
 391             if (gmx_mm_any_lt(rsq20,rcutoff2))
 392             {
 393
 394             r20              = _mm_mul_ps(rsq20,rinv20);
 395
 396             /* Compute parameters for interactions between i and j atoms */
 397             qq20             = _mm_mul_ps(iq2,jq0);
 398
 399             /* EWALD ELECTROSTATICS */
 400
 401             /* Analytical PME correction */
 402             zeta2            = _mm_mul_ps(beta2,rsq20);
 403             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
 404             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 405             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 406             felec            = _mm_mul_ps(qq20,felec);
 407             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 408             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv20);
 409             velec            = _mm_mul_ps(qq20,velec);
 410
 411             d                = _mm_sub_ps(r20,rswitch);
 412             d                = _mm_max_ps(d,_mm_setzero_ps());
 413             d2               = _mm_mul_ps(d,d);
 414             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 415
 416             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 417
 418             /* Evaluate switch function */
 419             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 420             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
 421             velec            = _mm_mul_ps(velec,sw);
 422             cutoff_mask      = _mm_cmplt_ps(rsq20,rcutoff2);
 423
 424             /* Update potential sum for this i atom from the interaction with this j atom. */
 425             velec            = _mm_and_ps(velec,cutoff_mask);
 426             velecsum         = _mm_add_ps(velecsum,velec);
 427
 428             fscal            = felec;
 429
 430             fscal            = _mm_and_ps(fscal,cutoff_mask);
 431
 432              /* Update vectorial force */
 433             fix2             = _mm_macc_ps(dx20,fscal,fix2);
 434             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
 435             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
 436
 437             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
 438             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
 439             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
 440
 441             }
 442
 443             /**************************
 444              * CALCULATE INTERACTIONS *
 445              **************************/
 446
 447             if (gmx_mm_any_lt(rsq30,rcutoff2))
 448             {
 449
 450             r30              = _mm_mul_ps(rsq30,rinv30);
 451
 452             /* Compute parameters for interactions between i and j atoms */
 453             qq30             = _mm_mul_ps(iq3,jq0);
 454
 455             /* EWALD ELECTROSTATICS */
 456
 457             /* Analytical PME correction */
 458             zeta2            = _mm_mul_ps(beta2,rsq30);
 459             rinv3            = _mm_mul_ps(rinvsq30,rinv30);
 460             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 461             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 462             felec            = _mm_mul_ps(qq30,felec);
 463             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 464             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv30);
 465             velec            = _mm_mul_ps(qq30,velec);
 466
 467             d                = _mm_sub_ps(r30,rswitch);
 468             d                = _mm_max_ps(d,_mm_setzero_ps());
 469             d2               = _mm_mul_ps(d,d);
 470             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 471
 472             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 473
 474             /* Evaluate switch function */
 475             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 476             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
 477             velec            = _mm_mul_ps(velec,sw);
 478             cutoff_mask      = _mm_cmplt_ps(rsq30,rcutoff2);
 479
 480             /* Update potential sum for this i atom from the interaction with this j atom. */
 481             velec            = _mm_and_ps(velec,cutoff_mask);
 482             velecsum         = _mm_add_ps(velecsum,velec);
 483
 484             fscal            = felec;
 485
 486             fscal            = _mm_and_ps(fscal,cutoff_mask);
 487
 488              /* Update vectorial force */
 489             fix3             = _mm_macc_ps(dx30,fscal,fix3);
 490             fiy3             = _mm_macc_ps(dy30,fscal,fiy3);
 491             fiz3             = _mm_macc_ps(dz30,fscal,fiz3);
 492
 493             fjx0             = _mm_macc_ps(dx30,fscal,fjx0);
 494             fjy0             = _mm_macc_ps(dy30,fscal,fjy0);
 495             fjz0             = _mm_macc_ps(dz30,fscal,fjz0);
 496
 497             }
 498
 499             fjptrA             = f+j_coord_offsetA;
 500             fjptrB             = f+j_coord_offsetB;
 501             fjptrC             = f+j_coord_offsetC;
 502             fjptrD             = f+j_coord_offsetD;
 503
 504             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 505
 506             /* Inner loop uses 221 flops */
 507         }
 508
 509         if(jidx<j_index_end)
 510         {
 511
 512             /* Get j neighbor index, and coordinate index */
 513             jnrlistA         = jjnr[jidx];
 514             jnrlistB         = jjnr[jidx+1];
 515             jnrlistC         = jjnr[jidx+2];
 516             jnrlistD         = jjnr[jidx+3];
 517             /* Sign of each element will be negative for non-real atoms.
 518              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 519              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 520              */
 521             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 522             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 523             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 524             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 525             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 526             j_coord_offsetA  = DIM*jnrA;
 527             j_coord_offsetB  = DIM*jnrB;
 528             j_coord_offsetC  = DIM*jnrC;
 529             j_coord_offsetD  = DIM*jnrD;
 530
 531             /* load j atom coordinates */
 532             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 533                                               x+j_coord_offsetC,x+j_coord_offsetD,
 534                                               &jx0,&jy0,&jz0);
 535
 536             /* Calculate displacement vector */
 537             dx00             = _mm_sub_ps(ix0,jx0);
 538             dy00             = _mm_sub_ps(iy0,jy0);
 539             dz00             = _mm_sub_ps(iz0,jz0);
 540             dx10             = _mm_sub_ps(ix1,jx0);
 541             dy10             = _mm_sub_ps(iy1,jy0);
 542             dz10             = _mm_sub_ps(iz1,jz0);
 543             dx20             = _mm_sub_ps(ix2,jx0);
 544             dy20             = _mm_sub_ps(iy2,jy0);
 545             dz20             = _mm_sub_ps(iz2,jz0);
 546             dx30             = _mm_sub_ps(ix3,jx0);
 547             dy30             = _mm_sub_ps(iy3,jy0);
 548             dz30             = _mm_sub_ps(iz3,jz0);
 549
 550             /* Calculate squared distance and things based on it */
 551             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 552             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
 553             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
 554             rsq30            = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
 555
 556             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 557             rinv10           = gmx_mm_invsqrt_ps(rsq10);
 558             rinv20           = gmx_mm_invsqrt_ps(rsq20);
 559             rinv30           = gmx_mm_invsqrt_ps(rsq30);
 560
 561             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 562             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
 563             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
 564             rinvsq30         = _mm_mul_ps(rinv30,rinv30);
 565
 566             /* Load parameters for j particles */
 567             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 568                                                               charge+jnrC+0,charge+jnrD+0);
 569             vdwjidx0A        = 2*vdwtype[jnrA+0];
 570             vdwjidx0B        = 2*vdwtype[jnrB+0];
 571             vdwjidx0C        = 2*vdwtype[jnrC+0];
 572             vdwjidx0D        = 2*vdwtype[jnrD+0];
 573
 574             fjx0             = _mm_setzero_ps();
 575             fjy0             = _mm_setzero_ps();
 576             fjz0             = _mm_setzero_ps();
 577
 578             /**************************
 579              * CALCULATE INTERACTIONS *
 580              **************************/
 581
 582             if (gmx_mm_any_lt(rsq00,rcutoff2))
 583             {
 584
 585             r00              = _mm_mul_ps(rsq00,rinv00);
 586             r00              = _mm_andnot_ps(dummy_mask,r00);
 587
 588             /* Compute parameters for interactions between i and j atoms */
 589             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
 590                                          vdwparam+vdwioffset0+vdwjidx0B,
 591                                          vdwparam+vdwioffset0+vdwjidx0C,
 592                                          vdwparam+vdwioffset0+vdwjidx0D,
 593                                          &c6_00,&c12_00);
 594
 595             /* LENNARD-JONES DISPERSION/REPULSION */
 596
 597             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 598             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 599             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 600             vvdw             = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
 601             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 602
 603             d                = _mm_sub_ps(r00,rswitch);
 604             d                = _mm_max_ps(d,_mm_setzero_ps());
 605             d2               = _mm_mul_ps(d,d);
 606             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 607
 608             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 609
 610             /* Evaluate switch function */
 611             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 612             fvdw             = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
 613             vvdw             = _mm_mul_ps(vvdw,sw);
 614             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
 615
 616             /* Update potential sum for this i atom from the interaction with this j atom. */
 617             vvdw             = _mm_and_ps(vvdw,cutoff_mask);
 618             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 619             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 620
 621             fscal            = fvdw;
 622
 623             fscal            = _mm_and_ps(fscal,cutoff_mask);
 624
 625             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 626
 627              /* Update vectorial force */
 628             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 629             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 630             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 631
 632             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
 633             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
 634             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
 635
 636             }
 637
 638             /**************************
 639              * CALCULATE INTERACTIONS *
 640              **************************/
 641
 642             if (gmx_mm_any_lt(rsq10,rcutoff2))
 643             {
 644
 645             r10              = _mm_mul_ps(rsq10,rinv10);
 646             r10              = _mm_andnot_ps(dummy_mask,r10);
 647
 648             /* Compute parameters for interactions between i and j atoms */
 649             qq10             = _mm_mul_ps(iq1,jq0);
 650
 651             /* EWALD ELECTROSTATICS */
 652
 653             /* Analytical PME correction */
 654             zeta2            = _mm_mul_ps(beta2,rsq10);
 655             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
 656             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 657             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 658             felec            = _mm_mul_ps(qq10,felec);
 659             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 660             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv10);
 661             velec            = _mm_mul_ps(qq10,velec);
 662
 663             d                = _mm_sub_ps(r10,rswitch);
 664             d                = _mm_max_ps(d,_mm_setzero_ps());
 665             d2               = _mm_mul_ps(d,d);
 666             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 667
 668             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 669
 670             /* Evaluate switch function */
 671             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 672             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
 673             velec            = _mm_mul_ps(velec,sw);
 674             cutoff_mask      = _mm_cmplt_ps(rsq10,rcutoff2);
 675
 676             /* Update potential sum for this i atom from the interaction with this j atom. */
 677             velec            = _mm_and_ps(velec,cutoff_mask);
 678             velec            = _mm_andnot_ps(dummy_mask,velec);
 679             velecsum         = _mm_add_ps(velecsum,velec);
 680
 681             fscal            = felec;
 682
 683             fscal            = _mm_and_ps(fscal,cutoff_mask);
 684
 685             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 686
 687              /* Update vectorial force */
 688             fix1             = _mm_macc_ps(dx10,fscal,fix1);
 689             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
 690             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
 691
 692             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
 693             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
 694             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
 695
 696             }
 697
 698             /**************************
 699              * CALCULATE INTERACTIONS *
 700              **************************/
 701
 702             if (gmx_mm_any_lt(rsq20,rcutoff2))
 703             {
 704
 705             r20              = _mm_mul_ps(rsq20,rinv20);
 706             r20              = _mm_andnot_ps(dummy_mask,r20);
 707
 708             /* Compute parameters for interactions between i and j atoms */
 709             qq20             = _mm_mul_ps(iq2,jq0);
 710
 711             /* EWALD ELECTROSTATICS */
 712
 713             /* Analytical PME correction */
 714             zeta2            = _mm_mul_ps(beta2,rsq20);
 715             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
 716             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 717             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 718             felec            = _mm_mul_ps(qq20,felec);
 719             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 720             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv20);
 721             velec            = _mm_mul_ps(qq20,velec);
 722
 723             d                = _mm_sub_ps(r20,rswitch);
 724             d                = _mm_max_ps(d,_mm_setzero_ps());
 725             d2               = _mm_mul_ps(d,d);
 726             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 727
 728             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 729
 730             /* Evaluate switch function */
 731             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 732             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
 733             velec            = _mm_mul_ps(velec,sw);
 734             cutoff_mask      = _mm_cmplt_ps(rsq20,rcutoff2);
 735
 736             /* Update potential sum for this i atom from the interaction with this j atom. */
 737             velec            = _mm_and_ps(velec,cutoff_mask);
 738             velec            = _mm_andnot_ps(dummy_mask,velec);
 739             velecsum         = _mm_add_ps(velecsum,velec);
 740
 741             fscal            = felec;
 742
 743             fscal            = _mm_and_ps(fscal,cutoff_mask);
 744
 745             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 746
 747              /* Update vectorial force */
 748             fix2             = _mm_macc_ps(dx20,fscal,fix2);
 749             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
 750             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
 751
 752             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
 753             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
 754             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
 755
 756             }
 757
 758             /**************************
 759              * CALCULATE INTERACTIONS *
 760              **************************/
 761
 762             if (gmx_mm_any_lt(rsq30,rcutoff2))
 763             {
 764
 765             r30              = _mm_mul_ps(rsq30,rinv30);
 766             r30              = _mm_andnot_ps(dummy_mask,r30);
 767
 768             /* Compute parameters for interactions between i and j atoms */
 769             qq30             = _mm_mul_ps(iq3,jq0);
 770
 771             /* EWALD ELECTROSTATICS */
 772
 773             /* Analytical PME correction */
 774             zeta2            = _mm_mul_ps(beta2,rsq30);
 775             rinv3            = _mm_mul_ps(rinvsq30,rinv30);
 776             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
 777             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
 778             felec            = _mm_mul_ps(qq30,felec);
 779             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
 780             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv30);
 781             velec            = _mm_mul_ps(qq30,velec);
 782
 783             d                = _mm_sub_ps(r30,rswitch);
 784             d                = _mm_max_ps(d,_mm_setzero_ps());
 785             d2               = _mm_mul_ps(d,d);
 786             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
 787
 788             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
 789
 790             /* Evaluate switch function */
 791             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 792             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
 793             velec            = _mm_mul_ps(velec,sw);
 794             cutoff_mask      = _mm_cmplt_ps(rsq30,rcutoff2);
 795
 796             /* Update potential sum for this i atom from the interaction with this j atom. */
 797             velec            = _mm_and_ps(velec,cutoff_mask);
 798             velec            = _mm_andnot_ps(dummy_mask,velec);
 799             velecsum         = _mm_add_ps(velecsum,velec);
 800
 801             fscal            = felec;
 802
 803             fscal            = _mm_and_ps(fscal,cutoff_mask);
 804
 805             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 806
 807              /* Update vectorial force */
 808             fix3             = _mm_macc_ps(dx30,fscal,fix3);
 809             fiy3             = _mm_macc_ps(dy30,fscal,fiy3);
 810             fiz3             = _mm_macc_ps(dz30,fscal,fiz3);
 811
 812             fjx0             = _mm_macc_ps(dx30,fscal,fjx0);
 813             fjy0             = _mm_macc_ps(dy30,fscal,fjy0);
 814             fjz0             = _mm_macc_ps(dz30,fscal,fjz0);
 815
 816             }
 817
 818             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 819             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 820             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 821             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 822
 823             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
 824
 825             /* Inner loop uses 225 flops */
 826         }
 827
 828         /* End of innermost loop */
 829
 830         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 831                                               f+i_coord_offset,fshift+i_shift_offset);
 832
 833         ggid                        = gid[iidx];
 834         /* Update potential energies */
 835         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 836         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 837
 838         /* Increment number of inner iterations */
 839         inneriter                  += j_index_end - j_index_start;
 840
 841         /* Outer loop uses 26 flops */
 842     }
 843
 844     /* Increment number of outer iterations */
 845     outeriter        += nri;
 846
 847     /* Update outer/inner flops */
 848
 849     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*225);
 850 }
 851 /*
 852  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_128_fma_single
 853  * Electrostatics interaction: Ewald
 854  * VdW interaction:            LennardJones
 855  * Geometry:                   Water4-Particle
 856  * Calculate force/pot:        Force
 857  */
 858 void
 859 nb_kernel_ElecEwSw_VdwLJSw_GeomW4P1_F_avx_128_fma_single
 860                     (t_nblist                    * gmx_restrict       nlist,
 861                      rvec                        * gmx_restrict          xx,
 862                      rvec                        * gmx_restrict          ff,
 863                      t_forcerec                  * gmx_restrict          fr,
 864                      t_mdatoms                   * gmx_restrict     mdatoms,
 865                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 866                      t_nrnb                      * gmx_restrict        nrnb)
 867 {
 868     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 869      * just 0 for non-waters.
 870      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
 871      * jnr indices corresponding to data put in the four positions in the SIMD register.
 872      */
 873     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 874     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 875     int              jnrA,jnrB,jnrC,jnrD;
 876     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 877     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 878     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 879     real             rcutoff_scalar;
 880     real             *shiftvec,*fshift,*x,*f;
 881     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 882     real             scratch[4*DIM];
 883     __m128           fscal,rcutoff,rcutoff2,jidxall;
 884     int              vdwioffset0;
 885     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 886     int              vdwioffset1;
 887     __m128           ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 888     int              vdwioffset2;
 889     __m128           ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 890     int              vdwioffset3;
 891     __m128           ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 892     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 893     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 894     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 895     __m128           dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 896     __m128           dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 897     __m128           dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 898     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 899     real             *charge;
 900     int              nvdwtype;
 901     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 902     int              *vdwtype;
 903     real             *vdwparam;
 904     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 905     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 906     __m128i          ewitab;
 907     __m128           ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 908     __m128           beta,beta2,beta3,zeta2,pmecorrF,pmecorrV,rinv3;
 909     real             *ewtab;
 910     __m128           rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 911     real             rswitch_scalar,d_scalar;
 912     __m128           dummy_mask,cutoff_mask;
 913     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 914     __m128           one     = _mm_set1_ps(1.0);
 915     __m128           two     = _mm_set1_ps(2.0);
 916     x                = xx[0];
 917     f                = ff[0];
 918
 919     nri              = nlist->nri;
 920     iinr             = nlist->iinr;
 921     jindex           = nlist->jindex;
 922     jjnr             = nlist->jjnr;
 923     shiftidx         = nlist->shift;
 924     gid              = nlist->gid;
 925     shiftvec         = fr->shift_vec[0];
 926     fshift           = fr->fshift[0];
 927     facel            = _mm_set1_ps(fr->epsfac);
 928     charge           = mdatoms->chargeA;
 929     nvdwtype         = fr->ntype;
 930     vdwparam         = fr->nbfp;
 931     vdwtype          = mdatoms->typeA;
 932
 933     sh_ewald         = _mm_set1_ps(fr->ic->sh_ewald);
 934     beta             = _mm_set1_ps(fr->ic->ewaldcoeff_q);
 935     beta2            = _mm_mul_ps(beta,beta);
 936     beta3            = _mm_mul_ps(beta,beta2);
 937     ewtab            = fr->ic->tabq_coul_FDV0;
 938     ewtabscale       = _mm_set1_ps(fr->ic->tabq_scale);
 939     ewtabhalfspace   = _mm_set1_ps(0.5/fr->ic->tabq_scale);
 940
 941     /* Setup water-specific parameters */
 942     inr              = nlist->iinr[0];
 943     iq1              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+1]));
 944     iq2              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+2]));
 945     iq3              = _mm_mul_ps(facel,_mm_set1_ps(charge[inr+3]));
 946     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 947
 948     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 949     rcutoff_scalar   = fr->rcoulomb;
 950     rcutoff          = _mm_set1_ps(rcutoff_scalar);
 951     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
 952
 953     rswitch_scalar   = fr->rcoulomb_switch;
 954     rswitch          = _mm_set1_ps(rswitch_scalar);
 955     /* Setup switch parameters */
 956     d_scalar         = rcutoff_scalar-rswitch_scalar;
 957     d                = _mm_set1_ps(d_scalar);
 958     swV3             = _mm_set1_ps(-10.0/(d_scalar*d_scalar*d_scalar));
 959     swV4             = _mm_set1_ps( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 960     swV5             = _mm_set1_ps( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 961     swF2             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar));
 962     swF3             = _mm_set1_ps( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 963     swF4             = _mm_set1_ps(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 964
 965     /* Avoid stupid compiler warnings */
 966     jnrA = jnrB = jnrC = jnrD = 0;
 967     j_coord_offsetA = 0;
 968     j_coord_offsetB = 0;
 969     j_coord_offsetC = 0;
 970     j_coord_offsetD = 0;
 971
 972     outeriter        = 0;
 973     inneriter        = 0;
 974
 975     for(iidx=0;iidx<4*DIM;iidx++)
 976     {
 977         scratch[iidx] = 0.0;
 978     }
 979
 980     /* Start outer loop over neighborlists */
 981     for(iidx=0; iidx<nri; iidx++)
 982     {
 983         /* Load shift vector for this list */
 984         i_shift_offset   = DIM*shiftidx[iidx];
 985
 986         /* Load limits for loop over neighbors */
 987         j_index_start    = jindex[iidx];
 988         j_index_end      = jindex[iidx+1];
 989
 990         /* Get outer coordinate index */
 991         inr              = iinr[iidx];
 992         i_coord_offset   = DIM*inr;
 993
 994         /* Load i particle coords and add shift vector */
 995         gmx_mm_load_shift_and_4rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,
 996                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 997
 998         fix0             = _mm_setzero_ps();
 999         fiy0             = _mm_setzero_ps();
1000         fiz0             = _mm_setzero_ps();
1001         fix1             = _mm_setzero_ps();
1002         fiy1             = _mm_setzero_ps();
1003         fiz1             = _mm_setzero_ps();
1004         fix2             = _mm_setzero_ps();
1005         fiy2             = _mm_setzero_ps();
1006         fiz2             = _mm_setzero_ps();
1007         fix3             = _mm_setzero_ps();
1008         fiy3             = _mm_setzero_ps();
1009         fiz3             = _mm_setzero_ps();
1010
1011         /* Start inner kernel loop */
1012         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
1013         {
1014
1015             /* Get j neighbor index, and coordinate index */
1016             jnrA             = jjnr[jidx];
1017             jnrB             = jjnr[jidx+1];
1018             jnrC             = jjnr[jidx+2];
1019             jnrD             = jjnr[jidx+3];
1020             j_coord_offsetA  = DIM*jnrA;
1021             j_coord_offsetB  = DIM*jnrB;
1022             j_coord_offsetC  = DIM*jnrC;
1023             j_coord_offsetD  = DIM*jnrD;
1024
1025             /* load j atom coordinates */
1026             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1027                                               x+j_coord_offsetC,x+j_coord_offsetD,
1028                                               &jx0,&jy0,&jz0);
1029
1030             /* Calculate displacement vector */
1031             dx00             = _mm_sub_ps(ix0,jx0);
1032             dy00             = _mm_sub_ps(iy0,jy0);
1033             dz00             = _mm_sub_ps(iz0,jz0);
1034             dx10             = _mm_sub_ps(ix1,jx0);
1035             dy10             = _mm_sub_ps(iy1,jy0);
1036             dz10             = _mm_sub_ps(iz1,jz0);
1037             dx20             = _mm_sub_ps(ix2,jx0);
1038             dy20             = _mm_sub_ps(iy2,jy0);
1039             dz20             = _mm_sub_ps(iz2,jz0);
1040             dx30             = _mm_sub_ps(ix3,jx0);
1041             dy30             = _mm_sub_ps(iy3,jy0);
1042             dz30             = _mm_sub_ps(iz3,jz0);
1043
1044             /* Calculate squared distance and things based on it */
1045             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1046             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1047             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1048             rsq30            = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
1049
1050             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1051             rinv10           = gmx_mm_invsqrt_ps(rsq10);
1052             rinv20           = gmx_mm_invsqrt_ps(rsq20);
1053             rinv30           = gmx_mm_invsqrt_ps(rsq30);
1054
1055             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
1056             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
1057             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
1058             rinvsq30         = _mm_mul_ps(rinv30,rinv30);
1059
1060             /* Load parameters for j particles */
1061             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1062                                                               charge+jnrC+0,charge+jnrD+0);
1063             vdwjidx0A        = 2*vdwtype[jnrA+0];
1064             vdwjidx0B        = 2*vdwtype[jnrB+0];
1065             vdwjidx0C        = 2*vdwtype[jnrC+0];
1066             vdwjidx0D        = 2*vdwtype[jnrD+0];
1067
1068             fjx0             = _mm_setzero_ps();
1069             fjy0             = _mm_setzero_ps();
1070             fjz0             = _mm_setzero_ps();
1071
1072             /**************************
1073              * CALCULATE INTERACTIONS *
1074              **************************/
1075
1076             if (gmx_mm_any_lt(rsq00,rcutoff2))
1077             {
1078
1079             r00              = _mm_mul_ps(rsq00,rinv00);
1080
1081             /* Compute parameters for interactions between i and j atoms */
1082             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
1083                                          vdwparam+vdwioffset0+vdwjidx0B,
1084                                          vdwparam+vdwioffset0+vdwjidx0C,
1085                                          vdwparam+vdwioffset0+vdwjidx0D,
1086                                          &c6_00,&c12_00);
1087
1088             /* LENNARD-JONES DISPERSION/REPULSION */
1089
1090             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1091             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
1092             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
1093             vvdw             = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
1094             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
1095
1096             d                = _mm_sub_ps(r00,rswitch);
1097             d                = _mm_max_ps(d,_mm_setzero_ps());
1098             d2               = _mm_mul_ps(d,d);
1099             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1100
1101             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1102
1103             /* Evaluate switch function */
1104             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1105             fvdw             = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
1106             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
1107
1108             fscal            = fvdw;
1109
1110             fscal            = _mm_and_ps(fscal,cutoff_mask);
1111
1112              /* Update vectorial force */
1113             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1114             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1115             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1116
1117             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1118             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1119             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1120
1121             }
1122
1123             /**************************
1124              * CALCULATE INTERACTIONS *
1125              **************************/
1126
1127             if (gmx_mm_any_lt(rsq10,rcutoff2))
1128             {
1129
1130             r10              = _mm_mul_ps(rsq10,rinv10);
1131
1132             /* Compute parameters for interactions between i and j atoms */
1133             qq10             = _mm_mul_ps(iq1,jq0);
1134
1135             /* EWALD ELECTROSTATICS */
1136
1137             /* Analytical PME correction */
1138             zeta2            = _mm_mul_ps(beta2,rsq10);
1139             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
1140             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1141             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1142             felec            = _mm_mul_ps(qq10,felec);
1143             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1144             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv10);
1145             velec            = _mm_mul_ps(qq10,velec);
1146
1147             d                = _mm_sub_ps(r10,rswitch);
1148             d                = _mm_max_ps(d,_mm_setzero_ps());
1149             d2               = _mm_mul_ps(d,d);
1150             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1151
1152             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1153
1154             /* Evaluate switch function */
1155             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1156             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
1157             cutoff_mask      = _mm_cmplt_ps(rsq10,rcutoff2);
1158
1159             fscal            = felec;
1160
1161             fscal            = _mm_and_ps(fscal,cutoff_mask);
1162
1163              /* Update vectorial force */
1164             fix1             = _mm_macc_ps(dx10,fscal,fix1);
1165             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
1166             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
1167
1168             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
1169             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
1170             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
1171
1172             }
1173
1174             /**************************
1175              * CALCULATE INTERACTIONS *
1176              **************************/
1177
1178             if (gmx_mm_any_lt(rsq20,rcutoff2))
1179             {
1180
1181             r20              = _mm_mul_ps(rsq20,rinv20);
1182
1183             /* Compute parameters for interactions between i and j atoms */
1184             qq20             = _mm_mul_ps(iq2,jq0);
1185
1186             /* EWALD ELECTROSTATICS */
1187
1188             /* Analytical PME correction */
1189             zeta2            = _mm_mul_ps(beta2,rsq20);
1190             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
1191             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1192             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1193             felec            = _mm_mul_ps(qq20,felec);
1194             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1195             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv20);
1196             velec            = _mm_mul_ps(qq20,velec);
1197
1198             d                = _mm_sub_ps(r20,rswitch);
1199             d                = _mm_max_ps(d,_mm_setzero_ps());
1200             d2               = _mm_mul_ps(d,d);
1201             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1202
1203             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1204
1205             /* Evaluate switch function */
1206             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1207             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
1208             cutoff_mask      = _mm_cmplt_ps(rsq20,rcutoff2);
1209
1210             fscal            = felec;
1211
1212             fscal            = _mm_and_ps(fscal,cutoff_mask);
1213
1214              /* Update vectorial force */
1215             fix2             = _mm_macc_ps(dx20,fscal,fix2);
1216             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
1217             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
1218
1219             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
1220             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
1221             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
1222
1223             }
1224
1225             /**************************
1226              * CALCULATE INTERACTIONS *
1227              **************************/
1228
1229             if (gmx_mm_any_lt(rsq30,rcutoff2))
1230             {
1231
1232             r30              = _mm_mul_ps(rsq30,rinv30);
1233
1234             /* Compute parameters for interactions between i and j atoms */
1235             qq30             = _mm_mul_ps(iq3,jq0);
1236
1237             /* EWALD ELECTROSTATICS */
1238
1239             /* Analytical PME correction */
1240             zeta2            = _mm_mul_ps(beta2,rsq30);
1241             rinv3            = _mm_mul_ps(rinvsq30,rinv30);
1242             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1243             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1244             felec            = _mm_mul_ps(qq30,felec);
1245             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1246             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv30);
1247             velec            = _mm_mul_ps(qq30,velec);
1248
1249             d                = _mm_sub_ps(r30,rswitch);
1250             d                = _mm_max_ps(d,_mm_setzero_ps());
1251             d2               = _mm_mul_ps(d,d);
1252             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1253
1254             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1255
1256             /* Evaluate switch function */
1257             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1258             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
1259             cutoff_mask      = _mm_cmplt_ps(rsq30,rcutoff2);
1260
1261             fscal            = felec;
1262
1263             fscal            = _mm_and_ps(fscal,cutoff_mask);
1264
1265              /* Update vectorial force */
1266             fix3             = _mm_macc_ps(dx30,fscal,fix3);
1267             fiy3             = _mm_macc_ps(dy30,fscal,fiy3);
1268             fiz3             = _mm_macc_ps(dz30,fscal,fiz3);
1269
1270             fjx0             = _mm_macc_ps(dx30,fscal,fjx0);
1271             fjy0             = _mm_macc_ps(dy30,fscal,fjy0);
1272             fjz0             = _mm_macc_ps(dz30,fscal,fjz0);
1273
1274             }
1275
1276             fjptrA             = f+j_coord_offsetA;
1277             fjptrB             = f+j_coord_offsetB;
1278             fjptrC             = f+j_coord_offsetC;
1279             fjptrD             = f+j_coord_offsetD;
1280
1281             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1282
1283             /* Inner loop uses 209 flops */
1284         }
1285
1286         if(jidx<j_index_end)
1287         {
1288
1289             /* Get j neighbor index, and coordinate index */
1290             jnrlistA         = jjnr[jidx];
1291             jnrlistB         = jjnr[jidx+1];
1292             jnrlistC         = jjnr[jidx+2];
1293             jnrlistD         = jjnr[jidx+3];
1294             /* Sign of each element will be negative for non-real atoms.
1295              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
1296              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
1297              */
1298             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
1299             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
1300             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
1301             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
1302             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
1303             j_coord_offsetA  = DIM*jnrA;
1304             j_coord_offsetB  = DIM*jnrB;
1305             j_coord_offsetC  = DIM*jnrC;
1306             j_coord_offsetD  = DIM*jnrD;
1307
1308             /* load j atom coordinates */
1309             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
1310                                               x+j_coord_offsetC,x+j_coord_offsetD,
1311                                               &jx0,&jy0,&jz0);
1312
1313             /* Calculate displacement vector */
1314             dx00             = _mm_sub_ps(ix0,jx0);
1315             dy00             = _mm_sub_ps(iy0,jy0);
1316             dz00             = _mm_sub_ps(iz0,jz0);
1317             dx10             = _mm_sub_ps(ix1,jx0);
1318             dy10             = _mm_sub_ps(iy1,jy0);
1319             dz10             = _mm_sub_ps(iz1,jz0);
1320             dx20             = _mm_sub_ps(ix2,jx0);
1321             dy20             = _mm_sub_ps(iy2,jy0);
1322             dz20             = _mm_sub_ps(iz2,jz0);
1323             dx30             = _mm_sub_ps(ix3,jx0);
1324             dy30             = _mm_sub_ps(iy3,jy0);
1325             dz30             = _mm_sub_ps(iz3,jz0);
1326
1327             /* Calculate squared distance and things based on it */
1328             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
1329             rsq10            = gmx_mm_calc_rsq_ps(dx10,dy10,dz10);
1330             rsq20            = gmx_mm_calc_rsq_ps(dx20,dy20,dz20);
1331             rsq30            = gmx_mm_calc_rsq_ps(dx30,dy30,dz30);
1332
1333             rinv00           = gmx_mm_invsqrt_ps(rsq00);
1334             rinv10           = gmx_mm_invsqrt_ps(rsq10);
1335             rinv20           = gmx_mm_invsqrt_ps(rsq20);
1336             rinv30           = gmx_mm_invsqrt_ps(rsq30);
1337
1338             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
1339             rinvsq10         = _mm_mul_ps(rinv10,rinv10);
1340             rinvsq20         = _mm_mul_ps(rinv20,rinv20);
1341             rinvsq30         = _mm_mul_ps(rinv30,rinv30);
1342
1343             /* Load parameters for j particles */
1344             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
1345                                                               charge+jnrC+0,charge+jnrD+0);
1346             vdwjidx0A        = 2*vdwtype[jnrA+0];
1347             vdwjidx0B        = 2*vdwtype[jnrB+0];
1348             vdwjidx0C        = 2*vdwtype[jnrC+0];
1349             vdwjidx0D        = 2*vdwtype[jnrD+0];
1350
1351             fjx0             = _mm_setzero_ps();
1352             fjy0             = _mm_setzero_ps();
1353             fjz0             = _mm_setzero_ps();
1354
1355             /**************************
1356              * CALCULATE INTERACTIONS *
1357              **************************/
1358
1359             if (gmx_mm_any_lt(rsq00,rcutoff2))
1360             {
1361
1362             r00              = _mm_mul_ps(rsq00,rinv00);
1363             r00              = _mm_andnot_ps(dummy_mask,r00);
1364
1365             /* Compute parameters for interactions between i and j atoms */
1366             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
1367                                          vdwparam+vdwioffset0+vdwjidx0B,
1368                                          vdwparam+vdwioffset0+vdwjidx0C,
1369                                          vdwparam+vdwioffset0+vdwjidx0D,
1370                                          &c6_00,&c12_00);
1371
1372             /* LENNARD-JONES DISPERSION/REPULSION */
1373
1374             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
1375             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
1376             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
1377             vvdw             = _mm_msub_ps(vvdw12,one_twelfth,_mm_mul_ps(vvdw6,one_sixth));
1378             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
1379
1380             d                = _mm_sub_ps(r00,rswitch);
1381             d                = _mm_max_ps(d,_mm_setzero_ps());
1382             d2               = _mm_mul_ps(d,d);
1383             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1384
1385             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1386
1387             /* Evaluate switch function */
1388             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1389             fvdw             = _mm_msub_ps( fvdw,sw , _mm_mul_ps(rinv00,_mm_mul_ps(vvdw,dsw)) );
1390             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
1391
1392             fscal            = fvdw;
1393
1394             fscal            = _mm_and_ps(fscal,cutoff_mask);
1395
1396             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1397
1398              /* Update vectorial force */
1399             fix0             = _mm_macc_ps(dx00,fscal,fix0);
1400             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
1401             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
1402
1403             fjx0             = _mm_macc_ps(dx00,fscal,fjx0);
1404             fjy0             = _mm_macc_ps(dy00,fscal,fjy0);
1405             fjz0             = _mm_macc_ps(dz00,fscal,fjz0);
1406
1407             }
1408
1409             /**************************
1410              * CALCULATE INTERACTIONS *
1411              **************************/
1412
1413             if (gmx_mm_any_lt(rsq10,rcutoff2))
1414             {
1415
1416             r10              = _mm_mul_ps(rsq10,rinv10);
1417             r10              = _mm_andnot_ps(dummy_mask,r10);
1418
1419             /* Compute parameters for interactions between i and j atoms */
1420             qq10             = _mm_mul_ps(iq1,jq0);
1421
1422             /* EWALD ELECTROSTATICS */
1423
1424             /* Analytical PME correction */
1425             zeta2            = _mm_mul_ps(beta2,rsq10);
1426             rinv3            = _mm_mul_ps(rinvsq10,rinv10);
1427             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1428             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1429             felec            = _mm_mul_ps(qq10,felec);
1430             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1431             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv10);
1432             velec            = _mm_mul_ps(qq10,velec);
1433
1434             d                = _mm_sub_ps(r10,rswitch);
1435             d                = _mm_max_ps(d,_mm_setzero_ps());
1436             d2               = _mm_mul_ps(d,d);
1437             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1438
1439             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1440
1441             /* Evaluate switch function */
1442             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1443             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv10,_mm_mul_ps(velec,dsw)) );
1444             cutoff_mask      = _mm_cmplt_ps(rsq10,rcutoff2);
1445
1446             fscal            = felec;
1447
1448             fscal            = _mm_and_ps(fscal,cutoff_mask);
1449
1450             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1451
1452              /* Update vectorial force */
1453             fix1             = _mm_macc_ps(dx10,fscal,fix1);
1454             fiy1             = _mm_macc_ps(dy10,fscal,fiy1);
1455             fiz1             = _mm_macc_ps(dz10,fscal,fiz1);
1456
1457             fjx0             = _mm_macc_ps(dx10,fscal,fjx0);
1458             fjy0             = _mm_macc_ps(dy10,fscal,fjy0);
1459             fjz0             = _mm_macc_ps(dz10,fscal,fjz0);
1460
1461             }
1462
1463             /**************************
1464              * CALCULATE INTERACTIONS *
1465              **************************/
1466
1467             if (gmx_mm_any_lt(rsq20,rcutoff2))
1468             {
1469
1470             r20              = _mm_mul_ps(rsq20,rinv20);
1471             r20              = _mm_andnot_ps(dummy_mask,r20);
1472
1473             /* Compute parameters for interactions between i and j atoms */
1474             qq20             = _mm_mul_ps(iq2,jq0);
1475
1476             /* EWALD ELECTROSTATICS */
1477
1478             /* Analytical PME correction */
1479             zeta2            = _mm_mul_ps(beta2,rsq20);
1480             rinv3            = _mm_mul_ps(rinvsq20,rinv20);
1481             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1482             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1483             felec            = _mm_mul_ps(qq20,felec);
1484             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1485             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv20);
1486             velec            = _mm_mul_ps(qq20,velec);
1487
1488             d                = _mm_sub_ps(r20,rswitch);
1489             d                = _mm_max_ps(d,_mm_setzero_ps());
1490             d2               = _mm_mul_ps(d,d);
1491             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1492
1493             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1494
1495             /* Evaluate switch function */
1496             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1497             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv20,_mm_mul_ps(velec,dsw)) );
1498             cutoff_mask      = _mm_cmplt_ps(rsq20,rcutoff2);
1499
1500             fscal            = felec;
1501
1502             fscal            = _mm_and_ps(fscal,cutoff_mask);
1503
1504             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1505
1506              /* Update vectorial force */
1507             fix2             = _mm_macc_ps(dx20,fscal,fix2);
1508             fiy2             = _mm_macc_ps(dy20,fscal,fiy2);
1509             fiz2             = _mm_macc_ps(dz20,fscal,fiz2);
1510
1511             fjx0             = _mm_macc_ps(dx20,fscal,fjx0);
1512             fjy0             = _mm_macc_ps(dy20,fscal,fjy0);
1513             fjz0             = _mm_macc_ps(dz20,fscal,fjz0);
1514
1515             }
1516
1517             /**************************
1518              * CALCULATE INTERACTIONS *
1519              **************************/
1520
1521             if (gmx_mm_any_lt(rsq30,rcutoff2))
1522             {
1523
1524             r30              = _mm_mul_ps(rsq30,rinv30);
1525             r30              = _mm_andnot_ps(dummy_mask,r30);
1526
1527             /* Compute parameters for interactions between i and j atoms */
1528             qq30             = _mm_mul_ps(iq3,jq0);
1529
1530             /* EWALD ELECTROSTATICS */
1531
1532             /* Analytical PME correction */
1533             zeta2            = _mm_mul_ps(beta2,rsq30);
1534             rinv3            = _mm_mul_ps(rinvsq30,rinv30);
1535             pmecorrF         = gmx_mm_pmecorrF_ps(zeta2);
1536             felec            = _mm_macc_ps(pmecorrF,beta3,rinv3);
1537             felec            = _mm_mul_ps(qq30,felec);
1538             pmecorrV         = gmx_mm_pmecorrV_ps(zeta2);
1539             velec            = _mm_nmacc_ps(pmecorrV,beta,rinv30);
1540             velec            = _mm_mul_ps(qq30,velec);
1541
1542             d                = _mm_sub_ps(r30,rswitch);
1543             d                = _mm_max_ps(d,_mm_setzero_ps());
1544             d2               = _mm_mul_ps(d,d);
1545             sw               = _mm_add_ps(one,_mm_mul_ps(d2,_mm_mul_ps(d,_mm_macc_ps(d,_mm_macc_ps(d,swV5,swV4),swV3))));
1546
1547             dsw              = _mm_mul_ps(d2,_mm_macc_ps(d,_mm_macc_ps(d,swF4,swF3),swF2));
1548
1549             /* Evaluate switch function */
1550             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1551             felec            = _mm_msub_ps( felec,sw , _mm_mul_ps(rinv30,_mm_mul_ps(velec,dsw)) );
1552             cutoff_mask      = _mm_cmplt_ps(rsq30,rcutoff2);
1553
1554             fscal            = felec;
1555
1556             fscal            = _mm_and_ps(fscal,cutoff_mask);
1557
1558             fscal            = _mm_andnot_ps(dummy_mask,fscal);
1559
1560              /* Update vectorial force */
1561             fix3             = _mm_macc_ps(dx30,fscal,fix3);
1562             fiy3             = _mm_macc_ps(dy30,fscal,fiy3);
1563             fiz3             = _mm_macc_ps(dz30,fscal,fiz3);
1564
1565             fjx0             = _mm_macc_ps(dx30,fscal,fjx0);
1566             fjy0             = _mm_macc_ps(dy30,fscal,fjy0);
1567             fjz0             = _mm_macc_ps(dz30,fscal,fjz0);
1568
1569             }
1570
1571             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
1572             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
1573             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
1574             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
1575
1576             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,fjx0,fjy0,fjz0);
1577
1578             /* Inner loop uses 213 flops */
1579         }
1580
1581         /* End of innermost loop */
1582
1583         gmx_mm_update_iforce_4atom_swizzle_ps(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1584                                               f+i_coord_offset,fshift+i_shift_offset);
1585
1586         /* Increment number of inner iterations */
1587         inneriter                  += j_index_end - j_index_start;
1588
1589         /* Outer loop uses 24 flops */
1590     }
1591
1592     /* Increment number of outer iterations */
1593     outeriter        += nri;
1594
1595     /* Update outer/inner flops */
1596
1597     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*213);
1598 }