src/gromacs/gmxlib/nonbonded/nb_kernel_avx_128_fma_double/nb_kernel_ElecEwSh_VdwLJEwSh_GeomP1P1_avx_128_fma_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS avx_128_fma_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "gromacs/simd/math_x86_avx_128_fma_double.h"
  48 #include "kernelutil_x86_avx_128_fma_double.h"
  49
  50 /*
  51  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomP1P1_VF_avx_128_fma_double
  52  * Electrostatics interaction: Ewald
  53  * VdW interaction:            LJEwald
  54  * Geometry:                   Particle-Particle
  55  * Calculate force/pot:        PotentialAndForce
  56  */
  57 void
  58 nb_kernel_ElecEwSh_VdwLJEwSh_GeomP1P1_VF_avx_128_fma_double
  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 refer to j loop unrolling done with SSE double precision, e.g. for the two 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;
  75     int              j_coord_offsetA,j_coord_offsetB;
  76     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  77     real             rcutoff_scalar;
  78     real             *shiftvec,*fshift,*x,*f;
  79     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  80     int              vdwioffset0;
  81     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  82     int              vdwjidx0A,vdwjidx0B;
  83     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  84     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  85     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  86     real             *charge;
  87     int              nvdwtype;
  88     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  89     int              *vdwtype;
  90     real             *vdwparam;
  91     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  92     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  93     __m128d           c6grid_00;
  94     real             *vdwgridparam;
  95     __m128d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
  96     __m128d           one_half  = _mm_set1_pd(0.5);
  97     __m128d           minus_one = _mm_set1_pd(-1.0);
  98     __m128i          ewitab;
  99     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 100     real             *ewtab;
 101     __m128d          dummy_mask,cutoff_mask;
 102     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 103     __m128d          one     = _mm_set1_pd(1.0);
 104     __m128d          two     = _mm_set1_pd(2.0);
 105     x                = xx[0];
 106     f                = ff[0];
 107
 108     nri              = nlist->nri;
 109     iinr             = nlist->iinr;
 110     jindex           = nlist->jindex;
 111     jjnr             = nlist->jjnr;
 112     shiftidx         = nlist->shift;
 113     gid              = nlist->gid;
 114     shiftvec         = fr->shift_vec[0];
 115     fshift           = fr->fshift[0];
 116     facel            = _mm_set1_pd(fr->epsfac);
 117     charge           = mdatoms->chargeA;
 118     nvdwtype         = fr->ntype;
 119     vdwparam         = fr->nbfp;
 120     vdwtype          = mdatoms->typeA;
 121     vdwgridparam     = fr->ljpme_c6grid;
 122     sh_lj_ewald      = _mm_set1_pd(fr->ic->sh_lj_ewald);
 123     ewclj            = _mm_set1_pd(fr->ewaldcoeff_lj);
 124     ewclj2           = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
 125
 126     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 127     ewtab            = fr->ic->tabq_coul_FDV0;
 128     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 129     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 130
 131     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 132     rcutoff_scalar   = fr->rcoulomb;
 133     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 134     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 135
 136     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 137     rvdw             = _mm_set1_pd(fr->rvdw);
 138
 139     /* Avoid stupid compiler warnings */
 140     jnrA = jnrB = 0;
 141     j_coord_offsetA = 0;
 142     j_coord_offsetB = 0;
 143
 144     outeriter        = 0;
 145     inneriter        = 0;
 146
 147     /* Start outer loop over neighborlists */
 148     for(iidx=0; iidx<nri; iidx++)
 149     {
 150         /* Load shift vector for this list */
 151         i_shift_offset   = DIM*shiftidx[iidx];
 152
 153         /* Load limits for loop over neighbors */
 154         j_index_start    = jindex[iidx];
 155         j_index_end      = jindex[iidx+1];
 156
 157         /* Get outer coordinate index */
 158         inr              = iinr[iidx];
 159         i_coord_offset   = DIM*inr;
 160
 161         /* Load i particle coords and add shift vector */
 162         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 163
 164         fix0             = _mm_setzero_pd();
 165         fiy0             = _mm_setzero_pd();
 166         fiz0             = _mm_setzero_pd();
 167
 168         /* Load parameters for i particles */
 169         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 170         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 171
 172         /* Reset potential sums */
 173         velecsum         = _mm_setzero_pd();
 174         vvdwsum          = _mm_setzero_pd();
 175
 176         /* Start inner kernel loop */
 177         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 178         {
 179
 180             /* Get j neighbor index, and coordinate index */
 181             jnrA             = jjnr[jidx];
 182             jnrB             = jjnr[jidx+1];
 183             j_coord_offsetA  = DIM*jnrA;
 184             j_coord_offsetB  = DIM*jnrB;
 185
 186             /* load j atom coordinates */
 187             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 188                                               &jx0,&jy0,&jz0);
 189
 190             /* Calculate displacement vector */
 191             dx00             = _mm_sub_pd(ix0,jx0);
 192             dy00             = _mm_sub_pd(iy0,jy0);
 193             dz00             = _mm_sub_pd(iz0,jz0);
 194
 195             /* Calculate squared distance and things based on it */
 196             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 197
 198             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 199
 200             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 201
 202             /* Load parameters for j particles */
 203             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 204             vdwjidx0A        = 2*vdwtype[jnrA+0];
 205             vdwjidx0B        = 2*vdwtype[jnrB+0];
 206
 207             /**************************
 208              * CALCULATE INTERACTIONS *
 209              **************************/
 210
 211             if (gmx_mm_any_lt(rsq00,rcutoff2))
 212             {
 213
 214             r00              = _mm_mul_pd(rsq00,rinv00);
 215
 216             /* Compute parameters for interactions between i and j atoms */
 217             qq00             = _mm_mul_pd(iq0,jq0);
 218             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 219                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 220             c6grid_00       = gmx_mm_load_2real_swizzle_pd(vdwgridparam+vdwioffset0+vdwjidx0A,
 221                                                                vdwgridparam+vdwioffset0+vdwjidx0B);
 222
 223             /* EWALD ELECTROSTATICS */
 224
 225             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 226             ewrt             = _mm_mul_pd(r00,ewtabscale);
 227             ewitab           = _mm_cvttpd_epi32(ewrt);
 228 #ifdef __XOP__
 229             eweps            = _mm_frcz_pd(ewrt);
 230 #else
 231             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 232 #endif
 233             twoeweps         = _mm_add_pd(eweps,eweps);
 234             ewitab           = _mm_slli_epi32(ewitab,2);
 235             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 236             ewtabD           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,1) );
 237             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 238             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 239             ewtabFn          = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,1) +2);
 240             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 241             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 242             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 243             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
 244             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 245
 246             /* Analytical LJ-PME */
 247             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 248             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 249             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 250             exponent         = gmx_simd_exp_d(ewcljrsq);
 251             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 252             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
 253             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 254             vvdw6            = _mm_mul_pd(_mm_macc_pd(-c6grid_00,_mm_sub_pd(one,poly),c6_00),rinvsix);
 255             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 256             vvdw             = _mm_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 257                                _mm_mul_pd(_mm_sub_pd(vvdw6,_mm_macc_pd(c6grid_00,sh_lj_ewald,_mm_mul_pd(c6_00,sh_vdw_invrcut6))),one_sixth));
 258             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 259             fvdw             = _mm_mul_pd(_mm_add_pd(vvdw12,_mm_msub_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6),vvdw6)),rinvsq00);
 260
 261             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 262
 263             /* Update potential sum for this i atom from the interaction with this j atom. */
 264             velec            = _mm_and_pd(velec,cutoff_mask);
 265             velecsum         = _mm_add_pd(velecsum,velec);
 266             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 267             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 268
 269             fscal            = _mm_add_pd(felec,fvdw);
 270
 271             fscal            = _mm_and_pd(fscal,cutoff_mask);
 272
 273             /* Update vectorial force */
 274             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 275             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 276             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 277
 278             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 279                                                    _mm_mul_pd(dx00,fscal),
 280                                                    _mm_mul_pd(dy00,fscal),
 281                                                    _mm_mul_pd(dz00,fscal));
 282
 283             }
 284
 285             /* Inner loop uses 78 flops */
 286         }
 287
 288         if(jidx<j_index_end)
 289         {
 290
 291             jnrA             = jjnr[jidx];
 292             j_coord_offsetA  = DIM*jnrA;
 293
 294             /* load j atom coordinates */
 295             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 296                                               &jx0,&jy0,&jz0);
 297
 298             /* Calculate displacement vector */
 299             dx00             = _mm_sub_pd(ix0,jx0);
 300             dy00             = _mm_sub_pd(iy0,jy0);
 301             dz00             = _mm_sub_pd(iz0,jz0);
 302
 303             /* Calculate squared distance and things based on it */
 304             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 305
 306             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 307
 308             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 309
 310             /* Load parameters for j particles */
 311             jq0              = _mm_load_sd(charge+jnrA+0);
 312             vdwjidx0A        = 2*vdwtype[jnrA+0];
 313
 314             /**************************
 315              * CALCULATE INTERACTIONS *
 316              **************************/
 317
 318             if (gmx_mm_any_lt(rsq00,rcutoff2))
 319             {
 320
 321             r00              = _mm_mul_pd(rsq00,rinv00);
 322
 323             /* Compute parameters for interactions between i and j atoms */
 324             qq00             = _mm_mul_pd(iq0,jq0);
 325             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 326             c6grid_00       = gmx_mm_load_1real_pd(vdwgridparam+vdwioffset0+vdwjidx0A);
 327
 328             /* EWALD ELECTROSTATICS */
 329
 330             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 331             ewrt             = _mm_mul_pd(r00,ewtabscale);
 332             ewitab           = _mm_cvttpd_epi32(ewrt);
 333 #ifdef __XOP__
 334             eweps            = _mm_frcz_pd(ewrt);
 335 #else
 336             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 337 #endif
 338             twoeweps         = _mm_add_pd(eweps,eweps);
 339             ewitab           = _mm_slli_epi32(ewitab,2);
 340             ewtabF           = _mm_load_pd( ewtab + _mm_extract_epi32(ewitab,0) );
 341             ewtabD           = _mm_setzero_pd();
 342             GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
 343             ewtabV           = _mm_load_sd( ewtab + _mm_extract_epi32(ewitab,0) +2);
 344             ewtabFn          = _mm_setzero_pd();
 345             GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
 346             felec            = _mm_macc_pd(eweps,ewtabD,ewtabF);
 347             velec            = _mm_nmacc_pd(_mm_mul_pd(ewtabhalfspace,eweps) ,_mm_add_pd(ewtabF,felec), ewtabV);
 348             velec            = _mm_mul_pd(qq00,_mm_sub_pd(_mm_sub_pd(rinv00,sh_ewald),velec));
 349             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 350
 351             /* Analytical LJ-PME */
 352             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 353             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 354             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 355             exponent         = gmx_simd_exp_d(ewcljrsq);
 356             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 357             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
 358             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 359             vvdw6            = _mm_mul_pd(_mm_macc_pd(-c6grid_00,_mm_sub_pd(one,poly),c6_00),rinvsix);
 360             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 361             vvdw             = _mm_msub_pd(_mm_nmacc_pd(c12_00,_mm_mul_pd(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 362                                _mm_mul_pd(_mm_sub_pd(vvdw6,_mm_macc_pd(c6grid_00,sh_lj_ewald,_mm_mul_pd(c6_00,sh_vdw_invrcut6))),one_sixth));
 363             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 364             fvdw             = _mm_mul_pd(_mm_add_pd(vvdw12,_mm_msub_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6),vvdw6)),rinvsq00);
 365
 366             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 367
 368             /* Update potential sum for this i atom from the interaction with this j atom. */
 369             velec            = _mm_and_pd(velec,cutoff_mask);
 370             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 371             velecsum         = _mm_add_pd(velecsum,velec);
 372             vvdw             = _mm_and_pd(vvdw,cutoff_mask);
 373             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 374             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 375
 376             fscal            = _mm_add_pd(felec,fvdw);
 377
 378             fscal            = _mm_and_pd(fscal,cutoff_mask);
 379
 380             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 381
 382             /* Update vectorial force */
 383             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 384             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 385             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 386
 387             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 388                                                    _mm_mul_pd(dx00,fscal),
 389                                                    _mm_mul_pd(dy00,fscal),
 390                                                    _mm_mul_pd(dz00,fscal));
 391
 392             }
 393
 394             /* Inner loop uses 78 flops */
 395         }
 396
 397         /* End of innermost loop */
 398
 399         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 400                                               f+i_coord_offset,fshift+i_shift_offset);
 401
 402         ggid                        = gid[iidx];
 403         /* Update potential energies */
 404         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 405         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 406
 407         /* Increment number of inner iterations */
 408         inneriter                  += j_index_end - j_index_start;
 409
 410         /* Outer loop uses 9 flops */
 411     }
 412
 413     /* Increment number of outer iterations */
 414     outeriter        += nri;
 415
 416     /* Update outer/inner flops */
 417
 418     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*78);
 419 }
 420 /*
 421  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomP1P1_F_avx_128_fma_double
 422  * Electrostatics interaction: Ewald
 423  * VdW interaction:            LJEwald
 424  * Geometry:                   Particle-Particle
 425  * Calculate force/pot:        Force
 426  */
 427 void
 428 nb_kernel_ElecEwSh_VdwLJEwSh_GeomP1P1_F_avx_128_fma_double
 429                     (t_nblist                    * gmx_restrict       nlist,
 430                      rvec                        * gmx_restrict          xx,
 431                      rvec                        * gmx_restrict          ff,
 432                      t_forcerec                  * gmx_restrict          fr,
 433                      t_mdatoms                   * gmx_restrict     mdatoms,
 434                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 435                      t_nrnb                      * gmx_restrict        nrnb)
 436 {
 437     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 438      * just 0 for non-waters.
 439      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 440      * jnr indices corresponding to data put in the four positions in the SIMD register.
 441      */
 442     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 443     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 444     int              jnrA,jnrB;
 445     int              j_coord_offsetA,j_coord_offsetB;
 446     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 447     real             rcutoff_scalar;
 448     real             *shiftvec,*fshift,*x,*f;
 449     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 450     int              vdwioffset0;
 451     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 452     int              vdwjidx0A,vdwjidx0B;
 453     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 454     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 455     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 456     real             *charge;
 457     int              nvdwtype;
 458     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 459     int              *vdwtype;
 460     real             *vdwparam;
 461     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 462     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 463     __m128d           c6grid_00;
 464     real             *vdwgridparam;
 465     __m128d           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 466     __m128d           one_half  = _mm_set1_pd(0.5);
 467     __m128d           minus_one = _mm_set1_pd(-1.0);
 468     __m128i          ewitab;
 469     __m128d          ewtabscale,eweps,twoeweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 470     real             *ewtab;
 471     __m128d          dummy_mask,cutoff_mask;
 472     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 473     __m128d          one     = _mm_set1_pd(1.0);
 474     __m128d          two     = _mm_set1_pd(2.0);
 475     x                = xx[0];
 476     f                = ff[0];
 477
 478     nri              = nlist->nri;
 479     iinr             = nlist->iinr;
 480     jindex           = nlist->jindex;
 481     jjnr             = nlist->jjnr;
 482     shiftidx         = nlist->shift;
 483     gid              = nlist->gid;
 484     shiftvec         = fr->shift_vec[0];
 485     fshift           = fr->fshift[0];
 486     facel            = _mm_set1_pd(fr->epsfac);
 487     charge           = mdatoms->chargeA;
 488     nvdwtype         = fr->ntype;
 489     vdwparam         = fr->nbfp;
 490     vdwtype          = mdatoms->typeA;
 491     vdwgridparam     = fr->ljpme_c6grid;
 492     sh_lj_ewald      = _mm_set1_pd(fr->ic->sh_lj_ewald);
 493     ewclj            = _mm_set1_pd(fr->ewaldcoeff_lj);
 494     ewclj2           = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
 495
 496     sh_ewald         = _mm_set1_pd(fr->ic->sh_ewald);
 497     ewtab            = fr->ic->tabq_coul_F;
 498     ewtabscale       = _mm_set1_pd(fr->ic->tabq_scale);
 499     ewtabhalfspace   = _mm_set1_pd(0.5/fr->ic->tabq_scale);
 500
 501     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 502     rcutoff_scalar   = fr->rcoulomb;
 503     rcutoff          = _mm_set1_pd(rcutoff_scalar);
 504     rcutoff2         = _mm_mul_pd(rcutoff,rcutoff);
 505
 506     sh_vdw_invrcut6  = _mm_set1_pd(fr->ic->sh_invrc6);
 507     rvdw             = _mm_set1_pd(fr->rvdw);
 508
 509     /* Avoid stupid compiler warnings */
 510     jnrA = jnrB = 0;
 511     j_coord_offsetA = 0;
 512     j_coord_offsetB = 0;
 513
 514     outeriter        = 0;
 515     inneriter        = 0;
 516
 517     /* Start outer loop over neighborlists */
 518     for(iidx=0; iidx<nri; iidx++)
 519     {
 520         /* Load shift vector for this list */
 521         i_shift_offset   = DIM*shiftidx[iidx];
 522
 523         /* Load limits for loop over neighbors */
 524         j_index_start    = jindex[iidx];
 525         j_index_end      = jindex[iidx+1];
 526
 527         /* Get outer coordinate index */
 528         inr              = iinr[iidx];
 529         i_coord_offset   = DIM*inr;
 530
 531         /* Load i particle coords and add shift vector */
 532         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 533
 534         fix0             = _mm_setzero_pd();
 535         fiy0             = _mm_setzero_pd();
 536         fiz0             = _mm_setzero_pd();
 537
 538         /* Load parameters for i particles */
 539         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 540         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 541
 542         /* Start inner kernel loop */
 543         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 544         {
 545
 546             /* Get j neighbor index, and coordinate index */
 547             jnrA             = jjnr[jidx];
 548             jnrB             = jjnr[jidx+1];
 549             j_coord_offsetA  = DIM*jnrA;
 550             j_coord_offsetB  = DIM*jnrB;
 551
 552             /* load j atom coordinates */
 553             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 554                                               &jx0,&jy0,&jz0);
 555
 556             /* Calculate displacement vector */
 557             dx00             = _mm_sub_pd(ix0,jx0);
 558             dy00             = _mm_sub_pd(iy0,jy0);
 559             dz00             = _mm_sub_pd(iz0,jz0);
 560
 561             /* Calculate squared distance and things based on it */
 562             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 563
 564             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 565
 566             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 567
 568             /* Load parameters for j particles */
 569             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 570             vdwjidx0A        = 2*vdwtype[jnrA+0];
 571             vdwjidx0B        = 2*vdwtype[jnrB+0];
 572
 573             /**************************
 574              * CALCULATE INTERACTIONS *
 575              **************************/
 576
 577             if (gmx_mm_any_lt(rsq00,rcutoff2))
 578             {
 579
 580             r00              = _mm_mul_pd(rsq00,rinv00);
 581
 582             /* Compute parameters for interactions between i and j atoms */
 583             qq00             = _mm_mul_pd(iq0,jq0);
 584             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 585                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 586             c6grid_00       = gmx_mm_load_2real_swizzle_pd(vdwgridparam+vdwioffset0+vdwjidx0A,
 587                                                                vdwgridparam+vdwioffset0+vdwjidx0B);
 588
 589             /* EWALD ELECTROSTATICS */
 590
 591             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 592             ewrt             = _mm_mul_pd(r00,ewtabscale);
 593             ewitab           = _mm_cvttpd_epi32(ewrt);
 594 #ifdef __XOP__
 595             eweps            = _mm_frcz_pd(ewrt);
 596 #else
 597             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 598 #endif
 599             twoeweps         = _mm_add_pd(eweps,eweps);
 600             gmx_mm_load_2pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),ewtab+_mm_extract_epi32(ewitab,1),
 601                                          &ewtabF,&ewtabFn);
 602             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
 603             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 604
 605             /* Analytical LJ-PME */
 606             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 607             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 608             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 609             exponent         = gmx_simd_exp_d(ewcljrsq);
 610             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 611             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
 612             /* f6A = 6 * C6grid * (1 - poly) */
 613             f6A              = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
 614             /* f6B = C6grid * exponent * beta^6 */
 615             f6B              = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
 616             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 617             fvdw              = _mm_mul_pd(_mm_macc_pd(_mm_msub_pd(c12_00,rinvsix,_mm_sub_pd(c6_00,f6A)),rinvsix,f6B),rinvsq00);
 618
 619             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 620
 621             fscal            = _mm_add_pd(felec,fvdw);
 622
 623             fscal            = _mm_and_pd(fscal,cutoff_mask);
 624
 625             /* Update vectorial force */
 626             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 627             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 628             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 629
 630             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,
 631                                                    _mm_mul_pd(dx00,fscal),
 632                                                    _mm_mul_pd(dy00,fscal),
 633                                                    _mm_mul_pd(dz00,fscal));
 634
 635             }
 636
 637             /* Inner loop uses 63 flops */
 638         }
 639
 640         if(jidx<j_index_end)
 641         {
 642
 643             jnrA             = jjnr[jidx];
 644             j_coord_offsetA  = DIM*jnrA;
 645
 646             /* load j atom coordinates */
 647             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 648                                               &jx0,&jy0,&jz0);
 649
 650             /* Calculate displacement vector */
 651             dx00             = _mm_sub_pd(ix0,jx0);
 652             dy00             = _mm_sub_pd(iy0,jy0);
 653             dz00             = _mm_sub_pd(iz0,jz0);
 654
 655             /* Calculate squared distance and things based on it */
 656             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 657
 658             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 659
 660             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 661
 662             /* Load parameters for j particles */
 663             jq0              = _mm_load_sd(charge+jnrA+0);
 664             vdwjidx0A        = 2*vdwtype[jnrA+0];
 665
 666             /**************************
 667              * CALCULATE INTERACTIONS *
 668              **************************/
 669
 670             if (gmx_mm_any_lt(rsq00,rcutoff2))
 671             {
 672
 673             r00              = _mm_mul_pd(rsq00,rinv00);
 674
 675             /* Compute parameters for interactions between i and j atoms */
 676             qq00             = _mm_mul_pd(iq0,jq0);
 677             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 678             c6grid_00       = gmx_mm_load_1real_pd(vdwgridparam+vdwioffset0+vdwjidx0A);
 679
 680             /* EWALD ELECTROSTATICS */
 681
 682             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 683             ewrt             = _mm_mul_pd(r00,ewtabscale);
 684             ewitab           = _mm_cvttpd_epi32(ewrt);
 685 #ifdef __XOP__
 686             eweps            = _mm_frcz_pd(ewrt);
 687 #else
 688             eweps            = _mm_sub_pd(ewrt,_mm_round_pd(ewrt, _MM_FROUND_FLOOR));
 689 #endif
 690             twoeweps         = _mm_add_pd(eweps,eweps);
 691             gmx_mm_load_1pair_swizzle_pd(ewtab+_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
 692             felec            = _mm_macc_pd(eweps,ewtabFn,_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF));
 693             felec            = _mm_mul_pd(_mm_mul_pd(qq00,rinv00),_mm_sub_pd(rinvsq00,felec));
 694
 695             /* Analytical LJ-PME */
 696             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 697             ewcljrsq         = _mm_mul_pd(ewclj2,rsq00);
 698             ewclj6           = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
 699             exponent         = gmx_simd_exp_d(ewcljrsq);
 700             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 701             poly             = _mm_mul_pd(exponent,_mm_macc_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half,_mm_sub_pd(one,ewcljrsq)));
 702             /* f6A = 6 * C6grid * (1 - poly) */
 703             f6A              = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
 704             /* f6B = C6grid * exponent * beta^6 */
 705             f6B              = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
 706             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 707             fvdw              = _mm_mul_pd(_mm_macc_pd(_mm_msub_pd(c12_00,rinvsix,_mm_sub_pd(c6_00,f6A)),rinvsix,f6B),rinvsq00);
 708
 709             cutoff_mask      = _mm_cmplt_pd(rsq00,rcutoff2);
 710
 711             fscal            = _mm_add_pd(felec,fvdw);
 712
 713             fscal            = _mm_and_pd(fscal,cutoff_mask);
 714
 715             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 716
 717             /* Update vectorial force */
 718             fix0             = _mm_macc_pd(dx00,fscal,fix0);
 719             fiy0             = _mm_macc_pd(dy00,fscal,fiy0);
 720             fiz0             = _mm_macc_pd(dz00,fscal,fiz0);
 721
 722             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,
 723                                                    _mm_mul_pd(dx00,fscal),
 724                                                    _mm_mul_pd(dy00,fscal),
 725                                                    _mm_mul_pd(dz00,fscal));
 726
 727             }
 728
 729             /* Inner loop uses 63 flops */
 730         }
 731
 732         /* End of innermost loop */
 733
 734         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 735                                               f+i_coord_offset,fshift+i_shift_offset);
 736
 737         /* Increment number of inner iterations */
 738         inneriter                  += j_index_end - j_index_start;
 739
 740         /* Outer loop uses 7 flops */
 741     }
 742
 743     /* Increment number of outer iterations */
 744     outeriter        += nri;
 745
 746     /* Update outer/inner flops */
 747
 748     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*63);
 749 }