src/gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_outer.h

   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
  37 /* Half-width SIMD operations are required here.
  38  * As the 4xn kernels are the "standard" kernels and some special operations
  39  * are required only here, we define those in nbnxn_kernel_simd_utils_...
  40  *
  41  * Half-width SIMD real type:
  42  * gmx_mm_hpr
  43  *
  44  * Half-width SIMD operations
  45  * Load reals at half-width aligned pointer b into half-width SIMD register a:
  46  * gmx_load_hpr(a, b)
  47  * Set all entries in half-width SIMD register *a to b:
  48  * gmx_set1_hpr(a, b)
  49  * Load one real at b and one real at b+1 into halves of a, respectively:
  50  * gmx_load1p1_pr(a, b)
  51  * Load reals at half-width aligned pointer b into two halves of a:
  52  * gmx_loaddh_pr(a, b)
  53  * Store half-width SIMD register b into half width aligned memory a:
  54  * gmx_store_hpr(a, b)
  55  * gmx_add_hpr(a, b)
  56  * gmx_sub_hpr(a, b)
  57  * Sum over 4 half SIMD registers:
  58  * gmx_sum4_hpr(a, b)
  59  * Sum the elements of halfs of each input register and store sums in out:
  60  * gmx_mm_transpose_sum4h_pr(a, b)
  61  * Extract two half-width registers *b, *c from a full width register a:
  62  * gmx_pr_to_2hpr(a, b, c)
  63  */
  64
  65
  66 {
  67     const nbnxn_ci_t   *nbln;
  68     const nbnxn_cj_t   *l_cj;
  69     const int          *type;
  70     const real         *q;
  71     const real         *shiftvec;
  72     const real         *x;
  73     const real         *nbfp0, *nbfp1, *nbfp2 = NULL, *nbfp3 = NULL;
  74     real                facel;
  75     real               *nbfp_ptr;
  76     int                 n, ci, ci_sh;
  77     int                 ish, ish3;
  78     gmx_bool            do_LJ, half_LJ, do_coul, do_self;
  79     int                 sci, scix, sciy, sciz, sci2;
  80     int                 cjind0, cjind1, cjind;
  81     int                 ip, jp;
  82
  83 #ifdef ENERGY_GROUPS
  84     int         Vstride_i;
  85     int         egps_ishift, egps_imask;
  86     int         egps_jshift, egps_jmask, egps_jstride;
  87     int         egps_i;
  88     real       *vvdwtp[UNROLLI];
  89     real       *vctp[UNROLLI];
  90 #endif
  91
  92     gmx_simd_real_t  shX_S;
  93     gmx_simd_real_t  shY_S;
  94     gmx_simd_real_t  shZ_S;
  95     gmx_simd_real_t  ix_S0, iy_S0, iz_S0;
  96     gmx_simd_real_t  ix_S2, iy_S2, iz_S2;
  97     gmx_simd_real_t  fix_S0, fiy_S0, fiz_S0;
  98     gmx_simd_real_t  fix_S2, fiy_S2, fiz_S2;
  99     /* We use an i-force SIMD register width of 4 */
 100     /* The simd4 stuff might be defined in nbnxn_kernel_simd_utils.h */
 101     gmx_simd4_real_t fix_S, fiy_S, fiz_S;
 102
 103     gmx_simd_real_t  diagonal_jmi_S;
 104 #if UNROLLI == UNROLLJ
 105     gmx_simd_bool_t  diagonal_mask_S0, diagonal_mask_S2;
 106 #else
 107     gmx_simd_bool_t  diagonal_mask0_S0, diagonal_mask0_S2;
 108     gmx_simd_bool_t  diagonal_mask1_S0, diagonal_mask1_S2;
 109 #endif
 110
 111     unsigned            *exclusion_filter;
 112     gmx_exclfilter       filter_S0, filter_S2;
 113
 114     gmx_simd_real_t      zero_S = gmx_simd_set1_r(0.0);
 115
 116     gmx_simd_real_t      one_S = gmx_simd_set1_r(1.0);
 117     gmx_simd_real_t      iq_S0 = gmx_simd_setzero_r();
 118     gmx_simd_real_t      iq_S2 = gmx_simd_setzero_r();
 119
 120 #ifdef CALC_COUL_RF
 121     gmx_simd_real_t      mrc_3_S;
 122 #ifdef CALC_ENERGIES
 123     gmx_simd_real_t      hrc_3_S, moh_rc_S;
 124 #endif
 125 #endif
 126
 127 #ifdef CALC_COUL_TAB
 128     /* Coulomb table variables */
 129     gmx_simd_real_t   invtsp_S;
 130     const real       *tab_coul_F;
 131 #ifndef TAB_FDV0
 132     const real       *tab_coul_V;
 133 #endif
 134     /* Thread-local working buffers for force and potential lookups */
 135     int               ti0_array[2*GMX_SIMD_REAL_WIDTH], *ti0 = NULL;
 136     int               ti2_array[2*GMX_SIMD_REAL_WIDTH], *ti2 = NULL;
 137 #ifdef CALC_ENERGIES
 138     gmx_simd_real_t   mhalfsp_S;
 139 #endif
 140 #endif
 141
 142 #ifdef CALC_COUL_EWALD
 143     gmx_simd_real_t beta2_S, beta_S;
 144 #endif
 145
 146 #if defined CALC_ENERGIES && (defined CALC_COUL_EWALD || defined CALC_COUL_TAB)
 147     gmx_simd_real_t  sh_ewald_S;
 148 #endif
 149
 150 #if defined LJ_CUT && defined CALC_ENERGIES
 151     gmx_simd_real_t   p6_cpot_S, p12_cpot_S;
 152 #endif
 153 #ifdef LJ_POT_SWITCH
 154     gmx_simd_real_t   rswitch_S;
 155     gmx_simd_real_t   swV3_S, swV4_S, swV5_S;
 156     gmx_simd_real_t   swF2_S, swF3_S, swF4_S;
 157 #endif
 158 #ifdef LJ_FORCE_SWITCH
 159     gmx_simd_real_t   rswitch_S;
 160     gmx_simd_real_t   p6_fc2_S, p6_fc3_S;
 161     gmx_simd_real_t   p12_fc2_S, p12_fc3_S;
 162 #ifdef CALC_ENERGIES
 163     gmx_simd_real_t   p6_vc3_S, p6_vc4_S;
 164     gmx_simd_real_t   p12_vc3_S, p12_vc4_S;
 165     gmx_simd_real_t   p6_6cpot_S, p12_12cpot_S;
 166 #endif
 167 #endif
 168 #ifdef LJ_EWALD_GEOM
 169     real              lj_ewaldcoeff2, lj_ewaldcoeff6_6;
 170     gmx_simd_real_t   mone_S, half_S, lje_c2_S, lje_c6_6_S, lje_vc_S;
 171 #endif
 172
 173 #ifdef LJ_COMB_LB
 174     const real       *ljc;
 175
 176     gmx_simd_real_t   hsig_i_S0, seps_i_S0;
 177     gmx_simd_real_t   hsig_i_S2, seps_i_S2;
 178 #else
 179 #ifdef FIX_LJ_C
 180     real              pvdw_array[2*UNROLLI*UNROLLJ+GMX_SIMD_REAL_WIDTH];
 181     real             *pvdw_c6, *pvdw_c12;
 182     gmx_simd_real_t   c6_S0, c12_S0;
 183     gmx_simd_real_t   c6_S2, c12_S2;
 184 #endif
 185
 186 #if defined LJ_COMB_GEOM || defined LJ_EWALD_GEOM
 187     const real       *ljc;
 188
 189     gmx_simd_real_t   c6s_S0, c12s_S0;
 190     gmx_simd_real_t   c6s_S2  = gmx_simd_setzero_r();
 191     gmx_simd_real_t   c12s_S2 = gmx_simd_setzero_r();
 192 #endif
 193 #endif /* LJ_COMB_LB */
 194
 195     gmx_simd_real_t  vctot_S, Vvdwtot_S;
 196     gmx_simd_real_t  sixth_S, twelveth_S;
 197
 198     gmx_simd_real_t  avoid_sing_S;
 199     gmx_simd_real_t  rc2_S;
 200 #ifdef VDW_CUTOFF_CHECK
 201     gmx_simd_real_t  rcvdw2_S;
 202 #endif
 203
 204     int ninner;
 205
 206 #ifdef COUNT_PAIRS
 207     int npair = 0;
 208 #endif
 209
 210 #if defined LJ_COMB_GEOM || defined LJ_COMB_LB || defined LJ_EWALD_GEOM
 211     ljc = nbat->lj_comb;
 212 #endif
 213 #if !(defined LJ_COMB_GEOM || defined LJ_COMB_LB)
 214     /* No combination rule used */
 215     nbfp_ptr    = (4 == nbfp_stride) ? nbat->nbfp_s4 : nbat->nbfp;
 216 #endif
 217
 218     /* Load j-i for the first i */
 219     diagonal_jmi_S    = gmx_simd_load_r(nbat->simd_2xnn_diagonal_j_minus_i);
 220     /* Generate all the diagonal masks as comparison results */
 221 #if UNROLLI == UNROLLJ
 222     diagonal_mask_S0  = gmx_simd_cmplt_r(zero_S, diagonal_jmi_S);
 223     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 224     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 225     diagonal_mask_S2  = gmx_simd_cmplt_r(zero_S, diagonal_jmi_S);
 226 #else
 227 #if 2*UNROLLI == UNROLLJ
 228     diagonal_mask0_S0 = gmx_simd_cmplt_r(zero_S, diagonal_jmi_S);
 229     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 230     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 231     diagonal_mask0_S2 = gmx_simd_cmplt_r(zero_S, diagonal_jmi_S);
 232     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 233     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 234     diagonal_mask1_S0 = gmx_simd_cmplt_r(zero_S, diagonal_jmi_S);
 235     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 236     diagonal_jmi_S    = gmx_simd_sub_r(diagonal_jmi_S, one_S);
 237     diagonal_mask1_S2 = gmx_simd_cmplt_r(zero_S, diagonal_jmi_S);
 238 #endif
 239 #endif
 240
 241     /* Load masks for topology exclusion masking. filter_stride is
 242        static const, so the conditional will be optimized away. */
 243     if (1 == filter_stride)
 244     {
 245         exclusion_filter = nbat->simd_exclusion_filter1;
 246     }
 247     else /* (2 == filter_stride) */
 248     {
 249         exclusion_filter = nbat->simd_exclusion_filter2;
 250     }
 251
 252     /* Here we cast the exclusion filters from unsigned * to int * or real *.
 253      * Since we only check bits, the actual value they represent does not
 254      * matter, as long as both filter and mask data are treated the same way.
 255      */
 256     filter_S0 = gmx_load_exclusion_filter(exclusion_filter + 0*2*UNROLLJ*filter_stride);
 257     filter_S2 = gmx_load_exclusion_filter(exclusion_filter + 1*2*UNROLLJ*filter_stride);
 258
 259 #ifdef CALC_COUL_RF
 260     /* Reaction-field constants */
 261     mrc_3_S  = gmx_simd_set1_r(-2*ic->k_rf);
 262 #ifdef CALC_ENERGIES
 263     hrc_3_S  = gmx_simd_set1_r(ic->k_rf);
 264     moh_rc_S = gmx_simd_set1_r(-ic->c_rf);
 265 #endif
 266 #endif
 267
 268 #ifdef CALC_COUL_TAB
 269     /* Generate aligned table index pointers */
 270     ti0 = prepare_table_load_buffer(ti0_array);
 271     ti2 = prepare_table_load_buffer(ti2_array);
 272
 273     invtsp_S  = gmx_simd_set1_r(ic->tabq_scale);
 274 #ifdef CALC_ENERGIES
 275     mhalfsp_S = gmx_simd_set1_r(-0.5/ic->tabq_scale);
 276 #endif
 277
 278 #ifdef TAB_FDV0
 279     tab_coul_F = ic->tabq_coul_FDV0;
 280 #else
 281     tab_coul_F = ic->tabq_coul_F;
 282     tab_coul_V = ic->tabq_coul_V;
 283 #endif
 284 #endif /* CALC_COUL_TAB */
 285
 286 #ifdef CALC_COUL_EWALD
 287     beta2_S = gmx_simd_set1_r(ic->ewaldcoeff_q*ic->ewaldcoeff_q);
 288     beta_S  = gmx_simd_set1_r(ic->ewaldcoeff_q);
 289 #endif
 290
 291 #if (defined CALC_COUL_TAB || defined CALC_COUL_EWALD) && defined CALC_ENERGIES
 292     sh_ewald_S = gmx_simd_set1_r(ic->sh_ewald);
 293 #endif
 294
 295     /* LJ function constants */
 296 #if defined CALC_ENERGIES || defined LJ_POT_SWITCH
 297     sixth_S      = gmx_simd_set1_r(1.0/6.0);
 298     twelveth_S   = gmx_simd_set1_r(1.0/12.0);
 299 #endif
 300
 301 #if defined LJ_CUT && defined CALC_ENERGIES
 302     /* We shift the potential by cpot, which can be zero */
 303     p6_cpot_S    = gmx_simd_set1_r(ic->dispersion_shift.cpot);
 304     p12_cpot_S   = gmx_simd_set1_r(ic->repulsion_shift.cpot);
 305 #endif
 306 #ifdef LJ_POT_SWITCH
 307     rswitch_S = gmx_simd_set1_r(ic->rvdw_switch);
 308     swV3_S    = gmx_simd_set1_r(ic->vdw_switch.c3);
 309     swV4_S    = gmx_simd_set1_r(ic->vdw_switch.c4);
 310     swV5_S    = gmx_simd_set1_r(ic->vdw_switch.c5);
 311     swF2_S    = gmx_simd_set1_r(3*ic->vdw_switch.c3);
 312     swF3_S    = gmx_simd_set1_r(4*ic->vdw_switch.c4);
 313     swF4_S    = gmx_simd_set1_r(5*ic->vdw_switch.c5);
 314 #endif
 315 #ifdef LJ_FORCE_SWITCH
 316     rswitch_S = gmx_simd_set1_r(ic->rvdw_switch);
 317     p6_fc2_S  = gmx_simd_set1_r(ic->dispersion_shift.c2);
 318     p6_fc3_S  = gmx_simd_set1_r(ic->dispersion_shift.c3);
 319     p12_fc2_S = gmx_simd_set1_r(ic->repulsion_shift.c2);
 320     p12_fc3_S = gmx_simd_set1_r(ic->repulsion_shift.c3);
 321 #ifdef CALC_ENERGIES
 322     {
 323         gmx_simd_real_t mthird_S  = gmx_simd_set1_r(-1.0/3.0);
 324         gmx_simd_real_t mfourth_S = gmx_simd_set1_r(-1.0/4.0);
 325
 326         p6_vc3_S     = gmx_simd_mul_r(mthird_S,  p6_fc2_S);
 327         p6_vc4_S     = gmx_simd_mul_r(mfourth_S, p6_fc3_S);
 328         p6_6cpot_S   = gmx_simd_set1_r(ic->dispersion_shift.cpot/6);
 329         p12_vc3_S    = gmx_simd_mul_r(mthird_S,  p12_fc2_S);
 330         p12_vc4_S    = gmx_simd_mul_r(mfourth_S, p12_fc3_S);
 331         p12_12cpot_S = gmx_simd_set1_r(ic->repulsion_shift.cpot/12);
 332     }
 333 #endif
 334 #endif
 335 #ifdef LJ_EWALD_GEOM
 336     mone_S           = gmx_simd_set1_r(-1.0);
 337     half_S           = gmx_simd_set1_r(0.5);
 338     lj_ewaldcoeff2   = ic->ewaldcoeff_lj*ic->ewaldcoeff_lj;
 339     lj_ewaldcoeff6_6 = lj_ewaldcoeff2*lj_ewaldcoeff2*lj_ewaldcoeff2/6;
 340     lje_c2_S         = gmx_simd_set1_r(lj_ewaldcoeff2);
 341     lje_c6_6_S       = gmx_simd_set1_r(lj_ewaldcoeff6_6);
 342     /* Determine the grid potential at the cut-off */
 343     lje_vc_S         = gmx_simd_set1_r(ic->sh_lj_ewald);
 344 #endif
 345
 346     /* The kernel either supports rcoulomb = rvdw or rcoulomb >= rvdw */
 347     rc2_S    = gmx_simd_set1_r(ic->rcoulomb*ic->rcoulomb);
 348 #ifdef VDW_CUTOFF_CHECK
 349     rcvdw2_S = gmx_simd_set1_r(ic->rvdw*ic->rvdw);
 350 #endif
 351
 352     avoid_sing_S = gmx_simd_set1_r(NBNXN_AVOID_SING_R2_INC);
 353
 354     q                   = nbat->q;
 355     type                = nbat->type;
 356     facel               = ic->epsfac;
 357     shiftvec            = shift_vec[0];
 358     x                   = nbat->x;
 359
 360 #ifdef FIX_LJ_C
 361     pvdw_c6  = gmx_simd_align_r(pvdw_array);
 362     pvdw_c12 = pvdw_c6 + UNROLLI*UNROLLJ;
 363
 364     for (jp = 0; jp < UNROLLJ; jp++)
 365     {
 366         pvdw_c6 [0*UNROLLJ+jp] = nbat->nbfp[0*2];
 367         pvdw_c6 [1*UNROLLJ+jp] = nbat->nbfp[0*2];
 368         pvdw_c6 [2*UNROLLJ+jp] = nbat->nbfp[0*2];
 369         pvdw_c6 [3*UNROLLJ+jp] = nbat->nbfp[0*2];
 370
 371         pvdw_c12[0*UNROLLJ+jp] = nbat->nbfp[0*2+1];
 372         pvdw_c12[1*UNROLLJ+jp] = nbat->nbfp[0*2+1];
 373         pvdw_c12[2*UNROLLJ+jp] = nbat->nbfp[0*2+1];
 374         pvdw_c12[3*UNROLLJ+jp] = nbat->nbfp[0*2+1];
 375     }
 376     c6_S0            = gmx_simd_load_r(pvdw_c6 +0*UNROLLJ);
 377     c6_S1            = gmx_simd_load_r(pvdw_c6 +1*UNROLLJ);
 378     c6_S2            = gmx_simd_load_r(pvdw_c6 +2*UNROLLJ);
 379     c6_S3            = gmx_simd_load_r(pvdw_c6 +3*UNROLLJ);
 380
 381     c12_S0           = gmx_simd_load_r(pvdw_c12+0*UNROLLJ);
 382     c12_S1           = gmx_simd_load_r(pvdw_c12+1*UNROLLJ);
 383     c12_S2           = gmx_simd_load_r(pvdw_c12+2*UNROLLJ);
 384     c12_S3           = gmx_simd_load_r(pvdw_c12+3*UNROLLJ);
 385 #endif /* FIX_LJ_C */
 386
 387 #ifdef ENERGY_GROUPS
 388     egps_ishift  = nbat->neg_2log;
 389     egps_imask   = (1<<egps_ishift) - 1;
 390     egps_jshift  = 2*nbat->neg_2log;
 391     egps_jmask   = (1<<egps_jshift) - 1;
 392     egps_jstride = (UNROLLJ>>1)*UNROLLJ;
 393     /* Major division is over i-particle energy groups, determine the stride */
 394     Vstride_i    = nbat->nenergrp*(1<<nbat->neg_2log)*egps_jstride;
 395 #endif
 396
 397     l_cj = nbl->cj;
 398
 399     ninner = 0;
 400     for (n = 0; n < nbl->nci; n++)
 401     {
 402         nbln = &nbl->ci[n];
 403
 404         ish              = (nbln->shift & NBNXN_CI_SHIFT);
 405         ish3             = ish*3;
 406         cjind0           = nbln->cj_ind_start;
 407         cjind1           = nbln->cj_ind_end;
 408         ci               = nbln->ci;
 409         ci_sh            = (ish == CENTRAL ? ci : -1);
 410
 411         shX_S = gmx_simd_load1_r(shiftvec+ish3);
 412         shY_S = gmx_simd_load1_r(shiftvec+ish3+1);
 413         shZ_S = gmx_simd_load1_r(shiftvec+ish3+2);
 414
 415 #if UNROLLJ <= 4
 416         sci              = ci*STRIDE;
 417         scix             = sci*DIM;
 418         sci2             = sci*2;
 419 #else
 420         sci              = (ci>>1)*STRIDE;
 421         scix             = sci*DIM + (ci & 1)*(STRIDE>>1);
 422         sci2             = sci*2 + (ci & 1)*(STRIDE>>1);
 423         sci             += (ci & 1)*(STRIDE>>1);
 424 #endif
 425
 426         /* We have 5 LJ/C combinations, but use only three inner loops,
 427          * as the other combinations are unlikely and/or not much faster:
 428          * inner half-LJ + C for half-LJ + C / no-LJ + C
 429          * inner LJ + C      for full-LJ + C
 430          * inner LJ          for full-LJ + no-C / half-LJ + no-C
 431          */
 432         do_LJ   = (nbln->shift & NBNXN_CI_DO_LJ(0));
 433         do_coul = (nbln->shift & NBNXN_CI_DO_COUL(0));
 434         half_LJ = ((nbln->shift & NBNXN_CI_HALF_LJ(0)) || !do_LJ) && do_coul;
 435 #ifdef LJ_EWALD_GEOM
 436         do_self = TRUE;
 437 #else
 438         do_self = do_coul;
 439 #endif
 440
 441 #ifdef ENERGY_GROUPS
 442         egps_i = nbat->energrp[ci];
 443         {
 444             int ia, egp_ia;
 445
 446             for (ia = 0; ia < UNROLLI; ia++)
 447             {
 448                 egp_ia     = (egps_i >> (ia*egps_ishift)) & egps_imask;
 449                 vvdwtp[ia] = Vvdw + egp_ia*Vstride_i;
 450                 vctp[ia]   = Vc   + egp_ia*Vstride_i;
 451             }
 452         }
 453 #endif
 454
 455 #ifdef CALC_ENERGIES
 456 #if UNROLLJ == 4
 457         if (do_self && l_cj[nbln->cj_ind_start].cj == ci_sh)
 458 #endif
 459 #if UNROLLJ == 8
 460         if (do_self && l_cj[nbln->cj_ind_start].cj == (ci_sh>>1))
 461 #endif
 462         {
 463             if (do_coul)
 464             {
 465                 real Vc_sub_self;
 466                 int  ia;
 467
 468 #ifdef CALC_COUL_RF
 469                 Vc_sub_self = 0.5*ic->c_rf;
 470 #endif
 471 #ifdef CALC_COUL_TAB
 472 #ifdef TAB_FDV0
 473                 Vc_sub_self = 0.5*tab_coul_F[2];
 474 #else
 475                 Vc_sub_self = 0.5*tab_coul_V[0];
 476 #endif
 477 #endif
 478 #ifdef CALC_COUL_EWALD
 479                 /* beta/sqrt(pi) */
 480                 Vc_sub_self = 0.5*ic->ewaldcoeff_q*M_2_SQRTPI;
 481 #endif
 482
 483                 for (ia = 0; ia < UNROLLI; ia++)
 484                 {
 485                     real qi;
 486
 487                     qi = q[sci+ia];
 488 #ifdef ENERGY_GROUPS
 489                     vctp[ia][((egps_i>>(ia*egps_ishift)) & egps_imask)*egps_jstride]
 490 #else
 491                     Vc[0]
 492 #endif
 493                         -= facel*qi*qi*Vc_sub_self;
 494                 }
 495             }
 496
 497 #ifdef LJ_EWALD_GEOM
 498             {
 499                 int  ia;
 500
 501                 for (ia = 0; ia < UNROLLI; ia++)
 502                 {
 503                     real c6_i;
 504
 505                     c6_i = nbat->nbfp[nbat->type[sci+ia]*(nbat->ntype + 1)*2]/6;
 506 #ifdef ENERGY_GROUPS
 507                     vvdwtp[ia][((egps_i>>(ia*egps_ishift)) & egps_imask)*egps_jstride]
 508 #else
 509                     Vvdw[0]
 510 #endif
 511                         += 0.5*c6_i*lj_ewaldcoeff6_6;
 512                 }
 513             }
 514 #endif      /* LJ_EWALD */
 515         }
 516 #endif
 517
 518         /* Load i atom data */
 519         sciy             = scix + STRIDE;
 520         sciz             = sciy + STRIDE;
 521         gmx_load1p1_pr(&ix_S0, x+scix);
 522         gmx_load1p1_pr(&ix_S2, x+scix+2);
 523         gmx_load1p1_pr(&iy_S0, x+sciy);
 524         gmx_load1p1_pr(&iy_S2, x+sciy+2);
 525         gmx_load1p1_pr(&iz_S0, x+sciz);
 526         gmx_load1p1_pr(&iz_S2, x+sciz+2);
 527         ix_S0          = gmx_simd_add_r(ix_S0, shX_S);
 528         ix_S2          = gmx_simd_add_r(ix_S2, shX_S);
 529         iy_S0          = gmx_simd_add_r(iy_S0, shY_S);
 530         iy_S2          = gmx_simd_add_r(iy_S2, shY_S);
 531         iz_S0          = gmx_simd_add_r(iz_S0, shZ_S);
 532         iz_S2          = gmx_simd_add_r(iz_S2, shZ_S);
 533
 534         if (do_coul)
 535         {
 536             gmx_simd_real_t facel_S;
 537
 538             facel_S    = gmx_simd_set1_r(facel);
 539
 540             gmx_load1p1_pr(&iq_S0, q+sci);
 541             gmx_load1p1_pr(&iq_S2, q+sci+2);
 542             iq_S0      = gmx_simd_mul_r(facel_S, iq_S0);
 543             iq_S2      = gmx_simd_mul_r(facel_S, iq_S2);
 544         }
 545
 546 #ifdef LJ_COMB_LB
 547         gmx_load1p1_pr(&hsig_i_S0, ljc+sci2+0);
 548         gmx_load1p1_pr(&hsig_i_S2, ljc+sci2+2);
 549         gmx_load1p1_pr(&seps_i_S0, ljc+sci2+STRIDE+0);
 550         gmx_load1p1_pr(&seps_i_S2, ljc+sci2+STRIDE+2);
 551 #else
 552 #ifdef LJ_COMB_GEOM
 553         gmx_load1p1_pr(&c6s_S0, ljc+sci2+0);
 554         if (!half_LJ)
 555         {
 556             gmx_load1p1_pr(&c6s_S2, ljc+sci2+2);
 557         }
 558         gmx_load1p1_pr(&c12s_S0, ljc+sci2+STRIDE+0);
 559         if (!half_LJ)
 560         {
 561             gmx_load1p1_pr(&c12s_S2, ljc+sci2+STRIDE+2);
 562         }
 563 #else
 564         nbfp0     = nbfp_ptr + type[sci  ]*nbat->ntype*nbfp_stride;
 565         nbfp1     = nbfp_ptr + type[sci+1]*nbat->ntype*nbfp_stride;
 566         if (!half_LJ)
 567         {
 568             nbfp2 = nbfp_ptr + type[sci+2]*nbat->ntype*nbfp_stride;
 569             nbfp3 = nbfp_ptr + type[sci+3]*nbat->ntype*nbfp_stride;
 570         }
 571 #endif
 572 #endif
 573 #ifdef LJ_EWALD_GEOM
 574         /* We need the geometrically combined C6 for the PME grid correction */
 575         gmx_load1p1_pr(&c6s_S0, ljc+sci2+0);
 576         if (!half_LJ)
 577         {
 578             gmx_load1p1_pr(&c6s_S2, ljc+sci2+2);
 579         }
 580 #endif
 581
 582         /* Zero the potential energy for this list */
 583         Vvdwtot_S        = gmx_simd_setzero_r();
 584         vctot_S          = gmx_simd_setzero_r();
 585
 586         /* Clear i atom forces */
 587         fix_S0           = gmx_simd_setzero_r();
 588         fix_S2           = gmx_simd_setzero_r();
 589         fiy_S0           = gmx_simd_setzero_r();
 590         fiy_S2           = gmx_simd_setzero_r();
 591         fiz_S0           = gmx_simd_setzero_r();
 592         fiz_S2           = gmx_simd_setzero_r();
 593
 594         cjind = cjind0;
 595
 596         /* Currently all kernels use (at least half) LJ */
 597 #define CALC_LJ
 598         if (half_LJ)
 599         {
 600             /* Coulomb: all i-atoms, LJ: first half i-atoms */
 601 #define CALC_COULOMB
 602 #define HALF_LJ
 603 #define CHECK_EXCLS
 604             while (cjind < cjind1 && nbl->cj[cjind].excl != NBNXN_INTERACTION_MASK_ALL)
 605             {
 606 #include "nbnxn_kernel_simd_2xnn_inner.h"
 607                 cjind++;
 608             }
 609 #undef CHECK_EXCLS
 610             for (; (cjind < cjind1); cjind++)
 611             {
 612 #include "nbnxn_kernel_simd_2xnn_inner.h"
 613             }
 614 #undef HALF_LJ
 615 #undef CALC_COULOMB
 616         }
 617         else if (do_coul)
 618         {
 619             /* Coulomb: all i-atoms, LJ: all i-atoms */
 620 #define CALC_COULOMB
 621 #define CHECK_EXCLS
 622             while (cjind < cjind1 && nbl->cj[cjind].excl != NBNXN_INTERACTION_MASK_ALL)
 623             {
 624 #include "nbnxn_kernel_simd_2xnn_inner.h"
 625                 cjind++;
 626             }
 627 #undef CHECK_EXCLS
 628             for (; (cjind < cjind1); cjind++)
 629             {
 630 #include "nbnxn_kernel_simd_2xnn_inner.h"
 631             }
 632 #undef CALC_COULOMB
 633         }
 634         else
 635         {
 636             /* Coulomb: none, LJ: all i-atoms */
 637 #define CHECK_EXCLS
 638             while (cjind < cjind1 && nbl->cj[cjind].excl != NBNXN_INTERACTION_MASK_ALL)
 639             {
 640 #include "nbnxn_kernel_simd_2xnn_inner.h"
 641                 cjind++;
 642             }
 643 #undef CHECK_EXCLS
 644             for (; (cjind < cjind1); cjind++)
 645             {
 646 #include "nbnxn_kernel_simd_2xnn_inner.h"
 647             }
 648         }
 649 #undef CALC_LJ
 650         ninner += cjind1 - cjind0;
 651
 652         /* Add accumulated i-forces to the force array */
 653         fix_S = gmx_mm_transpose_sum4h_pr(fix_S0, fix_S2);
 654         gmx_simd4_store_r(f+scix, gmx_simd4_add_r(fix_S, gmx_simd4_load_r(f+scix)));
 655
 656         fiy_S = gmx_mm_transpose_sum4h_pr(fiy_S0, fiy_S2);
 657         gmx_simd4_store_r(f+sciy, gmx_simd4_add_r(fiy_S, gmx_simd4_load_r(f+sciy)));
 658
 659         fiz_S = gmx_mm_transpose_sum4h_pr(fiz_S0, fiz_S2);
 660         gmx_simd4_store_r(f+sciz, gmx_simd4_add_r(fiz_S, gmx_simd4_load_r(f+sciz)));
 661
 662 #ifdef CALC_SHIFTFORCES
 663         fshift[ish3+0] += gmx_simd4_reduce_r(fix_S);
 664         fshift[ish3+1] += gmx_simd4_reduce_r(fiy_S);
 665         fshift[ish3+2] += gmx_simd4_reduce_r(fiz_S);
 666 #endif
 667
 668 #ifdef CALC_ENERGIES
 669         if (do_coul)
 670         {
 671             *Vc += gmx_simd_reduce_r(vctot_S);
 672         }
 673         *Vvdw += gmx_simd_reduce_r(Vvdwtot_S);
 674 #endif
 675
 676         /* Outer loop uses 6 flops/iteration */
 677     }
 678
 679 #ifdef COUNT_PAIRS
 680     printf("atom pairs %d\n", npair);
 681 #endif
 682 }