src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_single/nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_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_ElecRFCut_VdwLJSh_GeomP1P1_VF_avx_128_fma_single
  52  * Electrostatics interaction: ReactionField
  53  * VdW interaction:            LennardJones
  54  * Geometry:                   Particle-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_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              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
  86     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
  89     real             *charge;
  90     int              nvdwtype;
  91     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  92     int              *vdwtype;
  93     real             *vdwparam;
  94     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
  95     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
  96     __m128           dummy_mask,cutoff_mask;
  97     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
  98     __m128           one     = _mm_set1_ps(1.0);
  99     __m128           two     = _mm_set1_ps(2.0);
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = _mm_set1_ps(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113     krf              = _mm_set1_ps(fr->ic->k_rf);
 114     krf2             = _mm_set1_ps(fr->ic->k_rf*2.0);
 115     crf              = _mm_set1_ps(fr->ic->c_rf);
 116     nvdwtype         = fr->ntype;
 117     vdwparam         = fr->nbfp;
 118     vdwtype          = mdatoms->typeA;
 119
 120     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 121     rcutoff_scalar   = fr->rcoulomb;
 122     rcutoff          = _mm_set1_ps(rcutoff_scalar);
 123     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
 124
 125     sh_vdw_invrcut6  = _mm_set1_ps(fr->ic->sh_invrc6);
 126     rvdw             = _mm_set1_ps(fr->rvdw);
 127
 128     /* Avoid stupid compiler warnings */
 129     jnrA = jnrB = jnrC = jnrD = 0;
 130     j_coord_offsetA = 0;
 131     j_coord_offsetB = 0;
 132     j_coord_offsetC = 0;
 133     j_coord_offsetD = 0;
 134
 135     outeriter        = 0;
 136     inneriter        = 0;
 137
 138     for(iidx=0;iidx<4*DIM;iidx++)
 139     {
 140         scratch[iidx] = 0.0;
 141     }
 142
 143     /* Start outer loop over neighborlists */
 144     for(iidx=0; iidx<nri; iidx++)
 145     {
 146         /* Load shift vector for this list */
 147         i_shift_offset   = DIM*shiftidx[iidx];
 148
 149         /* Load limits for loop over neighbors */
 150         j_index_start    = jindex[iidx];
 151         j_index_end      = jindex[iidx+1];
 152
 153         /* Get outer coordinate index */
 154         inr              = iinr[iidx];
 155         i_coord_offset   = DIM*inr;
 156
 157         /* Load i particle coords and add shift vector */
 158         gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 159
 160         fix0             = _mm_setzero_ps();
 161         fiy0             = _mm_setzero_ps();
 162         fiz0             = _mm_setzero_ps();
 163
 164         /* Load parameters for i particles */
 165         iq0              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
 166         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 167
 168         /* Reset potential sums */
 169         velecsum         = _mm_setzero_ps();
 170         vvdwsum          = _mm_setzero_ps();
 171
 172         /* Start inner kernel loop */
 173         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 174         {
 175
 176             /* Get j neighbor index, and coordinate index */
 177             jnrA             = jjnr[jidx];
 178             jnrB             = jjnr[jidx+1];
 179             jnrC             = jjnr[jidx+2];
 180             jnrD             = jjnr[jidx+3];
 181             j_coord_offsetA  = DIM*jnrA;
 182             j_coord_offsetB  = DIM*jnrB;
 183             j_coord_offsetC  = DIM*jnrC;
 184             j_coord_offsetD  = DIM*jnrD;
 185
 186             /* load j atom coordinates */
 187             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 188                                               x+j_coord_offsetC,x+j_coord_offsetD,
 189                                               &jx0,&jy0,&jz0);
 190
 191             /* Calculate displacement vector */
 192             dx00             = _mm_sub_ps(ix0,jx0);
 193             dy00             = _mm_sub_ps(iy0,jy0);
 194             dz00             = _mm_sub_ps(iz0,jz0);
 195
 196             /* Calculate squared distance and things based on it */
 197             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 198
 199             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 200
 201             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 202
 203             /* Load parameters for j particles */
 204             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 205                                                               charge+jnrC+0,charge+jnrD+0);
 206             vdwjidx0A        = 2*vdwtype[jnrA+0];
 207             vdwjidx0B        = 2*vdwtype[jnrB+0];
 208             vdwjidx0C        = 2*vdwtype[jnrC+0];
 209             vdwjidx0D        = 2*vdwtype[jnrD+0];
 210
 211             /**************************
 212              * CALCULATE INTERACTIONS *
 213              **************************/
 214
 215             if (gmx_mm_any_lt(rsq00,rcutoff2))
 216             {
 217
 218             /* Compute parameters for interactions between i and j atoms */
 219             qq00             = _mm_mul_ps(iq0,jq0);
 220             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
 221                                          vdwparam+vdwioffset0+vdwjidx0B,
 222                                          vdwparam+vdwioffset0+vdwjidx0C,
 223                                          vdwparam+vdwioffset0+vdwjidx0D,
 224                                          &c6_00,&c12_00);
 225
 226             /* REACTION-FIELD ELECTROSTATICS */
 227             velec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_macc_ps(krf,rsq00,rinv00),crf));
 228             felec            = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
 229
 230             /* LENNARD-JONES DISPERSION/REPULSION */
 231
 232             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 233             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 234             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 235             vvdw             = _mm_msub_ps(_mm_nmacc_ps(c12_00,_mm_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 236                                           _mm_mul_ps( _mm_nmacc_ps(c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 237             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 238
 239             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
 240
 241             /* Update potential sum for this i atom from the interaction with this j atom. */
 242             velec            = _mm_and_ps(velec,cutoff_mask);
 243             velecsum         = _mm_add_ps(velecsum,velec);
 244             vvdw             = _mm_and_ps(vvdw,cutoff_mask);
 245             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 246
 247             fscal            = _mm_add_ps(felec,fvdw);
 248
 249             fscal            = _mm_and_ps(fscal,cutoff_mask);
 250
 251              /* Update vectorial force */
 252             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 253             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 254             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 255
 256             fjptrA             = f+j_coord_offsetA;
 257             fjptrB             = f+j_coord_offsetB;
 258             fjptrC             = f+j_coord_offsetC;
 259             fjptrD             = f+j_coord_offsetD;
 260             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 261                                                    _mm_mul_ps(dx00,fscal),
 262                                                    _mm_mul_ps(dy00,fscal),
 263                                                    _mm_mul_ps(dz00,fscal));
 264
 265             }
 266
 267             /* Inner loop uses 57 flops */
 268         }
 269
 270         if(jidx<j_index_end)
 271         {
 272
 273             /* Get j neighbor index, and coordinate index */
 274             jnrlistA         = jjnr[jidx];
 275             jnrlistB         = jjnr[jidx+1];
 276             jnrlistC         = jjnr[jidx+2];
 277             jnrlistD         = jjnr[jidx+3];
 278             /* Sign of each element will be negative for non-real atoms.
 279              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 280              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 281              */
 282             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 283             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 284             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 285             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 286             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 287             j_coord_offsetA  = DIM*jnrA;
 288             j_coord_offsetB  = DIM*jnrB;
 289             j_coord_offsetC  = DIM*jnrC;
 290             j_coord_offsetD  = DIM*jnrD;
 291
 292             /* load j atom coordinates */
 293             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 294                                               x+j_coord_offsetC,x+j_coord_offsetD,
 295                                               &jx0,&jy0,&jz0);
 296
 297             /* Calculate displacement vector */
 298             dx00             = _mm_sub_ps(ix0,jx0);
 299             dy00             = _mm_sub_ps(iy0,jy0);
 300             dz00             = _mm_sub_ps(iz0,jz0);
 301
 302             /* Calculate squared distance and things based on it */
 303             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 304
 305             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 306
 307             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 308
 309             /* Load parameters for j particles */
 310             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 311                                                               charge+jnrC+0,charge+jnrD+0);
 312             vdwjidx0A        = 2*vdwtype[jnrA+0];
 313             vdwjidx0B        = 2*vdwtype[jnrB+0];
 314             vdwjidx0C        = 2*vdwtype[jnrC+0];
 315             vdwjidx0D        = 2*vdwtype[jnrD+0];
 316
 317             /**************************
 318              * CALCULATE INTERACTIONS *
 319              **************************/
 320
 321             if (gmx_mm_any_lt(rsq00,rcutoff2))
 322             {
 323
 324             /* Compute parameters for interactions between i and j atoms */
 325             qq00             = _mm_mul_ps(iq0,jq0);
 326             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
 327                                          vdwparam+vdwioffset0+vdwjidx0B,
 328                                          vdwparam+vdwioffset0+vdwjidx0C,
 329                                          vdwparam+vdwioffset0+vdwjidx0D,
 330                                          &c6_00,&c12_00);
 331
 332             /* REACTION-FIELD ELECTROSTATICS */
 333             velec            = _mm_mul_ps(qq00,_mm_sub_ps(_mm_macc_ps(krf,rsq00,rinv00),crf));
 334             felec            = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
 335
 336             /* LENNARD-JONES DISPERSION/REPULSION */
 337
 338             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 339             vvdw6            = _mm_mul_ps(c6_00,rinvsix);
 340             vvdw12           = _mm_mul_ps(c12_00,_mm_mul_ps(rinvsix,rinvsix));
 341             vvdw             = _mm_msub_ps(_mm_nmacc_ps(c12_00,_mm_mul_ps(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 342                                           _mm_mul_ps( _mm_nmacc_ps(c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 343             fvdw             = _mm_mul_ps(_mm_sub_ps(vvdw12,vvdw6),rinvsq00);
 344
 345             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
 346
 347             /* Update potential sum for this i atom from the interaction with this j atom. */
 348             velec            = _mm_and_ps(velec,cutoff_mask);
 349             velec            = _mm_andnot_ps(dummy_mask,velec);
 350             velecsum         = _mm_add_ps(velecsum,velec);
 351             vvdw             = _mm_and_ps(vvdw,cutoff_mask);
 352             vvdw             = _mm_andnot_ps(dummy_mask,vvdw);
 353             vvdwsum          = _mm_add_ps(vvdwsum,vvdw);
 354
 355             fscal            = _mm_add_ps(felec,fvdw);
 356
 357             fscal            = _mm_and_ps(fscal,cutoff_mask);
 358
 359             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 360
 361              /* Update vectorial force */
 362             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 363             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 364             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 365
 366             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 367             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 368             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 369             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 370             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 371                                                    _mm_mul_ps(dx00,fscal),
 372                                                    _mm_mul_ps(dy00,fscal),
 373                                                    _mm_mul_ps(dz00,fscal));
 374
 375             }
 376
 377             /* Inner loop uses 57 flops */
 378         }
 379
 380         /* End of innermost loop */
 381
 382         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 383                                               f+i_coord_offset,fshift+i_shift_offset);
 384
 385         ggid                        = gid[iidx];
 386         /* Update potential energies */
 387         gmx_mm_update_1pot_ps(velecsum,kernel_data->energygrp_elec+ggid);
 388         gmx_mm_update_1pot_ps(vvdwsum,kernel_data->energygrp_vdw+ggid);
 389
 390         /* Increment number of inner iterations */
 391         inneriter                  += j_index_end - j_index_start;
 392
 393         /* Outer loop uses 9 flops */
 394     }
 395
 396     /* Increment number of outer iterations */
 397     outeriter        += nri;
 398
 399     /* Update outer/inner flops */
 400
 401     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*57);
 402 }
 403 /*
 404  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_avx_128_fma_single
 405  * Electrostatics interaction: ReactionField
 406  * VdW interaction:            LennardJones
 407  * Geometry:                   Particle-Particle
 408  * Calculate force/pot:        Force
 409  */
 410 void
 411 nb_kernel_ElecRFCut_VdwLJSh_GeomP1P1_F_avx_128_fma_single
 412                     (t_nblist                    * gmx_restrict       nlist,
 413                      rvec                        * gmx_restrict          xx,
 414                      rvec                        * gmx_restrict          ff,
 415                      t_forcerec                  * gmx_restrict          fr,
 416                      t_mdatoms                   * gmx_restrict     mdatoms,
 417                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 418                      t_nrnb                      * gmx_restrict        nrnb)
 419 {
 420     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 421      * just 0 for non-waters.
 422      * Suffixes A,B,C,D refer to j loop unrolling done with AVX_128, e.g. for the four different
 423      * jnr indices corresponding to data put in the four positions in the SIMD register.
 424      */
 425     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 426     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 427     int              jnrA,jnrB,jnrC,jnrD;
 428     int              jnrlistA,jnrlistB,jnrlistC,jnrlistD;
 429     int              j_coord_offsetA,j_coord_offsetB,j_coord_offsetC,j_coord_offsetD;
 430     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 431     real             rcutoff_scalar;
 432     real             *shiftvec,*fshift,*x,*f;
 433     real             *fjptrA,*fjptrB,*fjptrC,*fjptrD;
 434     real             scratch[4*DIM];
 435     __m128           fscal,rcutoff,rcutoff2,jidxall;
 436     int              vdwioffset0;
 437     __m128           ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 438     int              vdwjidx0A,vdwjidx0B,vdwjidx0C,vdwjidx0D;
 439     __m128           jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 440     __m128           dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 441     __m128           velec,felec,velecsum,facel,crf,krf,krf2;
 442     real             *charge;
 443     int              nvdwtype;
 444     __m128           rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 445     int              *vdwtype;
 446     real             *vdwparam;
 447     __m128           one_sixth   = _mm_set1_ps(1.0/6.0);
 448     __m128           one_twelfth = _mm_set1_ps(1.0/12.0);
 449     __m128           dummy_mask,cutoff_mask;
 450     __m128           signbit = _mm_castsi128_ps( _mm_set1_epi32(0x80000000) );
 451     __m128           one     = _mm_set1_ps(1.0);
 452     __m128           two     = _mm_set1_ps(2.0);
 453     x                = xx[0];
 454     f                = ff[0];
 455
 456     nri              = nlist->nri;
 457     iinr             = nlist->iinr;
 458     jindex           = nlist->jindex;
 459     jjnr             = nlist->jjnr;
 460     shiftidx         = nlist->shift;
 461     gid              = nlist->gid;
 462     shiftvec         = fr->shift_vec[0];
 463     fshift           = fr->fshift[0];
 464     facel            = _mm_set1_ps(fr->epsfac);
 465     charge           = mdatoms->chargeA;
 466     krf              = _mm_set1_ps(fr->ic->k_rf);
 467     krf2             = _mm_set1_ps(fr->ic->k_rf*2.0);
 468     crf              = _mm_set1_ps(fr->ic->c_rf);
 469     nvdwtype         = fr->ntype;
 470     vdwparam         = fr->nbfp;
 471     vdwtype          = mdatoms->typeA;
 472
 473     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 474     rcutoff_scalar   = fr->rcoulomb;
 475     rcutoff          = _mm_set1_ps(rcutoff_scalar);
 476     rcutoff2         = _mm_mul_ps(rcutoff,rcutoff);
 477
 478     sh_vdw_invrcut6  = _mm_set1_ps(fr->ic->sh_invrc6);
 479     rvdw             = _mm_set1_ps(fr->rvdw);
 480
 481     /* Avoid stupid compiler warnings */
 482     jnrA = jnrB = jnrC = jnrD = 0;
 483     j_coord_offsetA = 0;
 484     j_coord_offsetB = 0;
 485     j_coord_offsetC = 0;
 486     j_coord_offsetD = 0;
 487
 488     outeriter        = 0;
 489     inneriter        = 0;
 490
 491     for(iidx=0;iidx<4*DIM;iidx++)
 492     {
 493         scratch[iidx] = 0.0;
 494     }
 495
 496     /* Start outer loop over neighborlists */
 497     for(iidx=0; iidx<nri; iidx++)
 498     {
 499         /* Load shift vector for this list */
 500         i_shift_offset   = DIM*shiftidx[iidx];
 501
 502         /* Load limits for loop over neighbors */
 503         j_index_start    = jindex[iidx];
 504         j_index_end      = jindex[iidx+1];
 505
 506         /* Get outer coordinate index */
 507         inr              = iinr[iidx];
 508         i_coord_offset   = DIM*inr;
 509
 510         /* Load i particle coords and add shift vector */
 511         gmx_mm_load_shift_and_1rvec_broadcast_ps(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 512
 513         fix0             = _mm_setzero_ps();
 514         fiy0             = _mm_setzero_ps();
 515         fiz0             = _mm_setzero_ps();
 516
 517         /* Load parameters for i particles */
 518         iq0              = _mm_mul_ps(facel,_mm_load1_ps(charge+inr+0));
 519         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 520
 521         /* Start inner kernel loop */
 522         for(jidx=j_index_start; jidx<j_index_end && jjnr[jidx+3]>=0; jidx+=4)
 523         {
 524
 525             /* Get j neighbor index, and coordinate index */
 526             jnrA             = jjnr[jidx];
 527             jnrB             = jjnr[jidx+1];
 528             jnrC             = jjnr[jidx+2];
 529             jnrD             = jjnr[jidx+3];
 530             j_coord_offsetA  = DIM*jnrA;
 531             j_coord_offsetB  = DIM*jnrB;
 532             j_coord_offsetC  = DIM*jnrC;
 533             j_coord_offsetD  = DIM*jnrD;
 534
 535             /* load j atom coordinates */
 536             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 537                                               x+j_coord_offsetC,x+j_coord_offsetD,
 538                                               &jx0,&jy0,&jz0);
 539
 540             /* Calculate displacement vector */
 541             dx00             = _mm_sub_ps(ix0,jx0);
 542             dy00             = _mm_sub_ps(iy0,jy0);
 543             dz00             = _mm_sub_ps(iz0,jz0);
 544
 545             /* Calculate squared distance and things based on it */
 546             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 547
 548             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 549
 550             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 551
 552             /* Load parameters for j particles */
 553             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 554                                                               charge+jnrC+0,charge+jnrD+0);
 555             vdwjidx0A        = 2*vdwtype[jnrA+0];
 556             vdwjidx0B        = 2*vdwtype[jnrB+0];
 557             vdwjidx0C        = 2*vdwtype[jnrC+0];
 558             vdwjidx0D        = 2*vdwtype[jnrD+0];
 559
 560             /**************************
 561              * CALCULATE INTERACTIONS *
 562              **************************/
 563
 564             if (gmx_mm_any_lt(rsq00,rcutoff2))
 565             {
 566
 567             /* Compute parameters for interactions between i and j atoms */
 568             qq00             = _mm_mul_ps(iq0,jq0);
 569             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
 570                                          vdwparam+vdwioffset0+vdwjidx0B,
 571                                          vdwparam+vdwioffset0+vdwjidx0C,
 572                                          vdwparam+vdwioffset0+vdwjidx0D,
 573                                          &c6_00,&c12_00);
 574
 575             /* REACTION-FIELD ELECTROSTATICS */
 576             felec            = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
 577
 578             /* LENNARD-JONES DISPERSION/REPULSION */
 579
 580             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 581             fvdw             = _mm_mul_ps(_mm_msub_ps(c12_00,rinvsix,c6_00),_mm_mul_ps(rinvsix,rinvsq00));
 582
 583             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
 584
 585             fscal            = _mm_add_ps(felec,fvdw);
 586
 587             fscal            = _mm_and_ps(fscal,cutoff_mask);
 588
 589              /* Update vectorial force */
 590             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 591             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 592             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 593
 594             fjptrA             = f+j_coord_offsetA;
 595             fjptrB             = f+j_coord_offsetB;
 596             fjptrC             = f+j_coord_offsetC;
 597             fjptrD             = f+j_coord_offsetD;
 598             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 599                                                    _mm_mul_ps(dx00,fscal),
 600                                                    _mm_mul_ps(dy00,fscal),
 601                                                    _mm_mul_ps(dz00,fscal));
 602
 603             }
 604
 605             /* Inner loop uses 40 flops */
 606         }
 607
 608         if(jidx<j_index_end)
 609         {
 610
 611             /* Get j neighbor index, and coordinate index */
 612             jnrlistA         = jjnr[jidx];
 613             jnrlistB         = jjnr[jidx+1];
 614             jnrlistC         = jjnr[jidx+2];
 615             jnrlistD         = jjnr[jidx+3];
 616             /* Sign of each element will be negative for non-real atoms.
 617              * This mask will be 0xFFFFFFFF for dummy entries and 0x0 for real ones,
 618              * so use it as val = _mm_andnot_ps(mask,val) to clear dummy entries.
 619              */
 620             dummy_mask = gmx_mm_castsi128_ps(_mm_cmplt_epi32(_mm_loadu_si128((const __m128i *)(jjnr+jidx)),_mm_setzero_si128()));
 621             jnrA       = (jnrlistA>=0) ? jnrlistA : 0;
 622             jnrB       = (jnrlistB>=0) ? jnrlistB : 0;
 623             jnrC       = (jnrlistC>=0) ? jnrlistC : 0;
 624             jnrD       = (jnrlistD>=0) ? jnrlistD : 0;
 625             j_coord_offsetA  = DIM*jnrA;
 626             j_coord_offsetB  = DIM*jnrB;
 627             j_coord_offsetC  = DIM*jnrC;
 628             j_coord_offsetD  = DIM*jnrD;
 629
 630             /* load j atom coordinates */
 631             gmx_mm_load_1rvec_4ptr_swizzle_ps(x+j_coord_offsetA,x+j_coord_offsetB,
 632                                               x+j_coord_offsetC,x+j_coord_offsetD,
 633                                               &jx0,&jy0,&jz0);
 634
 635             /* Calculate displacement vector */
 636             dx00             = _mm_sub_ps(ix0,jx0);
 637             dy00             = _mm_sub_ps(iy0,jy0);
 638             dz00             = _mm_sub_ps(iz0,jz0);
 639
 640             /* Calculate squared distance and things based on it */
 641             rsq00            = gmx_mm_calc_rsq_ps(dx00,dy00,dz00);
 642
 643             rinv00           = gmx_mm_invsqrt_ps(rsq00);
 644
 645             rinvsq00         = _mm_mul_ps(rinv00,rinv00);
 646
 647             /* Load parameters for j particles */
 648             jq0              = gmx_mm_load_4real_swizzle_ps(charge+jnrA+0,charge+jnrB+0,
 649                                                               charge+jnrC+0,charge+jnrD+0);
 650             vdwjidx0A        = 2*vdwtype[jnrA+0];
 651             vdwjidx0B        = 2*vdwtype[jnrB+0];
 652             vdwjidx0C        = 2*vdwtype[jnrC+0];
 653             vdwjidx0D        = 2*vdwtype[jnrD+0];
 654
 655             /**************************
 656              * CALCULATE INTERACTIONS *
 657              **************************/
 658
 659             if (gmx_mm_any_lt(rsq00,rcutoff2))
 660             {
 661
 662             /* Compute parameters for interactions between i and j atoms */
 663             qq00             = _mm_mul_ps(iq0,jq0);
 664             gmx_mm_load_4pair_swizzle_ps(vdwparam+vdwioffset0+vdwjidx0A,
 665                                          vdwparam+vdwioffset0+vdwjidx0B,
 666                                          vdwparam+vdwioffset0+vdwjidx0C,
 667                                          vdwparam+vdwioffset0+vdwjidx0D,
 668                                          &c6_00,&c12_00);
 669
 670             /* REACTION-FIELD ELECTROSTATICS */
 671             felec            = _mm_mul_ps(qq00,_mm_msub_ps(rinv00,rinvsq00,krf2));
 672
 673             /* LENNARD-JONES DISPERSION/REPULSION */
 674
 675             rinvsix          = _mm_mul_ps(_mm_mul_ps(rinvsq00,rinvsq00),rinvsq00);
 676             fvdw             = _mm_mul_ps(_mm_msub_ps(c12_00,rinvsix,c6_00),_mm_mul_ps(rinvsix,rinvsq00));
 677
 678             cutoff_mask      = _mm_cmplt_ps(rsq00,rcutoff2);
 679
 680             fscal            = _mm_add_ps(felec,fvdw);
 681
 682             fscal            = _mm_and_ps(fscal,cutoff_mask);
 683
 684             fscal            = _mm_andnot_ps(dummy_mask,fscal);
 685
 686              /* Update vectorial force */
 687             fix0             = _mm_macc_ps(dx00,fscal,fix0);
 688             fiy0             = _mm_macc_ps(dy00,fscal,fiy0);
 689             fiz0             = _mm_macc_ps(dz00,fscal,fiz0);
 690
 691             fjptrA             = (jnrlistA>=0) ? f+j_coord_offsetA : scratch;
 692             fjptrB             = (jnrlistB>=0) ? f+j_coord_offsetB : scratch;
 693             fjptrC             = (jnrlistC>=0) ? f+j_coord_offsetC : scratch;
 694             fjptrD             = (jnrlistD>=0) ? f+j_coord_offsetD : scratch;
 695             gmx_mm_decrement_1rvec_4ptr_swizzle_ps(fjptrA,fjptrB,fjptrC,fjptrD,
 696                                                    _mm_mul_ps(dx00,fscal),
 697                                                    _mm_mul_ps(dy00,fscal),
 698                                                    _mm_mul_ps(dz00,fscal));
 699
 700             }
 701
 702             /* Inner loop uses 40 flops */
 703         }
 704
 705         /* End of innermost loop */
 706
 707         gmx_mm_update_iforce_1atom_swizzle_ps(fix0,fiy0,fiz0,
 708                                               f+i_coord_offset,fshift+i_shift_offset);
 709
 710         /* Increment number of inner iterations */
 711         inneriter                  += j_index_end - j_index_start;
 712
 713         /* Outer loop uses 7 flops */
 714     }
 715
 716     /* Increment number of outer iterations */
 717     outeriter        += nri;
 718
 719     /* Update outer/inner flops */
 720
 721     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*40);
 722 }