2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015, 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.
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.
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.
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.
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.
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.
36 * Note: this file was generated by the Verlet kernel generator for
44 #include "gromacs/math/vectypes.h"
45 #include "gromacs/mdlib/nb_verlet.h"
46 #include "gromacs/mdlib/nbnxn_simd.h"
47 #include "gromacs/mdtypes/interaction_const.h"
48 #include "gromacs/mdtypes/md_enums.h"
50 #ifdef GMX_NBNXN_SIMD_2XNN
52 /* Include the full-width SIMD macros */
53 #include "gromacs/simd/vector_operations.h"
55 #if !(GMX_SIMD_REAL_WIDTH == 8 || GMX_SIMD_REAL_WIDTH == 16)
56 #error "unsupported SIMD width"
59 #define GMX_SIMD_J_UNROLL_SIZE 2
60 #include "nbnxn_kernel_simd_2xnn.h"
62 #include "gromacs/mdlib/force_flags.h"
63 #include "gromacs/mdlib/gmx_omp_nthreads.h"
64 #include "gromacs/mdlib/nbnxn_kernels/nbnxn_kernel_common.h"
65 #include "gromacs/simd/simd.h"
66 #include "gromacs/utility/fatalerror.h"
67 #include "gromacs/utility/real.h"
69 /*! \brief Kinds of electrostatic treatments in SIMD Verlet kernels
72 coulktRF, coulktTAB, coulktTAB_TWIN, coulktEWALD, coulktEWALD_TWIN, coulktNR
75 /*! \brief Kinds of Van der Waals treatments in SIMD Verlet kernels
78 vdwktLJCUT_COMBGEOM, vdwktLJCUT_COMBLB, vdwktLJCUT_COMBNONE, vdwktLJFORCESWITCH, vdwktLJPOTSWITCH, vdwktLJEWALDCOMBGEOM, vdwktNR
81 /* Declare and define the kernel function pointer lookup tables.
82 * The minor index of the array goes over both the LJ combination rules,
83 * which is only supported by plain cut-off, and the LJ switch/PME functions.
85 static p_nbk_func_noener p_nbk_noener[coulktNR][vdwktNR] =
88 nbnxn_kernel_ElecRF_VdwLJCombGeom_F_2xnn,
89 nbnxn_kernel_ElecRF_VdwLJCombLB_F_2xnn,
90 nbnxn_kernel_ElecRF_VdwLJ_F_2xnn,
91 nbnxn_kernel_ElecRF_VdwLJFSw_F_2xnn,
92 nbnxn_kernel_ElecRF_VdwLJPSw_F_2xnn,
93 nbnxn_kernel_ElecRF_VdwLJEwCombGeom_F_2xnn,
96 nbnxn_kernel_ElecQSTab_VdwLJCombGeom_F_2xnn,
97 nbnxn_kernel_ElecQSTab_VdwLJCombLB_F_2xnn,
98 nbnxn_kernel_ElecQSTab_VdwLJ_F_2xnn,
99 nbnxn_kernel_ElecQSTab_VdwLJFSw_F_2xnn,
100 nbnxn_kernel_ElecQSTab_VdwLJPSw_F_2xnn,
101 nbnxn_kernel_ElecQSTab_VdwLJEwCombGeom_F_2xnn,
104 nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombGeom_F_2xnn,
105 nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombLB_F_2xnn,
106 nbnxn_kernel_ElecQSTabTwinCut_VdwLJ_F_2xnn,
107 nbnxn_kernel_ElecQSTabTwinCut_VdwLJFSw_F_2xnn,
108 nbnxn_kernel_ElecQSTabTwinCut_VdwLJPSw_F_2xnn,
109 nbnxn_kernel_ElecQSTabTwinCut_VdwLJEwCombGeom_F_2xnn,
112 nbnxn_kernel_ElecEw_VdwLJCombGeom_F_2xnn,
113 nbnxn_kernel_ElecEw_VdwLJCombLB_F_2xnn,
114 nbnxn_kernel_ElecEw_VdwLJ_F_2xnn,
115 nbnxn_kernel_ElecEw_VdwLJFSw_F_2xnn,
116 nbnxn_kernel_ElecEw_VdwLJPSw_F_2xnn,
117 nbnxn_kernel_ElecEw_VdwLJEwCombGeom_F_2xnn,
120 nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_F_2xnn,
121 nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_F_2xnn,
122 nbnxn_kernel_ElecEwTwinCut_VdwLJ_F_2xnn,
123 nbnxn_kernel_ElecEwTwinCut_VdwLJFSw_F_2xnn,
124 nbnxn_kernel_ElecEwTwinCut_VdwLJPSw_F_2xnn,
125 nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_F_2xnn,
129 static p_nbk_func_ener p_nbk_ener[coulktNR][vdwktNR] =
132 nbnxn_kernel_ElecRF_VdwLJCombGeom_VF_2xnn,
133 nbnxn_kernel_ElecRF_VdwLJCombLB_VF_2xnn,
134 nbnxn_kernel_ElecRF_VdwLJ_VF_2xnn,
135 nbnxn_kernel_ElecRF_VdwLJFSw_VF_2xnn,
136 nbnxn_kernel_ElecRF_VdwLJPSw_VF_2xnn,
137 nbnxn_kernel_ElecRF_VdwLJEwCombGeom_VF_2xnn,
140 nbnxn_kernel_ElecQSTab_VdwLJCombGeom_VF_2xnn,
141 nbnxn_kernel_ElecQSTab_VdwLJCombLB_VF_2xnn,
142 nbnxn_kernel_ElecQSTab_VdwLJ_VF_2xnn,
143 nbnxn_kernel_ElecQSTab_VdwLJFSw_VF_2xnn,
144 nbnxn_kernel_ElecQSTab_VdwLJPSw_VF_2xnn,
145 nbnxn_kernel_ElecQSTab_VdwLJEwCombGeom_VF_2xnn,
148 nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombGeom_VF_2xnn,
149 nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombLB_VF_2xnn,
150 nbnxn_kernel_ElecQSTabTwinCut_VdwLJ_VF_2xnn,
151 nbnxn_kernel_ElecQSTabTwinCut_VdwLJFSw_VF_2xnn,
152 nbnxn_kernel_ElecQSTabTwinCut_VdwLJPSw_VF_2xnn,
153 nbnxn_kernel_ElecQSTabTwinCut_VdwLJEwCombGeom_VF_2xnn,
156 nbnxn_kernel_ElecEw_VdwLJCombGeom_VF_2xnn,
157 nbnxn_kernel_ElecEw_VdwLJCombLB_VF_2xnn,
158 nbnxn_kernel_ElecEw_VdwLJ_VF_2xnn,
159 nbnxn_kernel_ElecEw_VdwLJFSw_VF_2xnn,
160 nbnxn_kernel_ElecEw_VdwLJPSw_VF_2xnn,
161 nbnxn_kernel_ElecEw_VdwLJEwCombGeom_VF_2xnn,
164 nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_VF_2xnn,
165 nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_VF_2xnn,
166 nbnxn_kernel_ElecEwTwinCut_VdwLJ_VF_2xnn,
167 nbnxn_kernel_ElecEwTwinCut_VdwLJFSw_VF_2xnn,
168 nbnxn_kernel_ElecEwTwinCut_VdwLJPSw_VF_2xnn,
169 nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_VF_2xnn,
173 static p_nbk_func_ener p_nbk_energrp[coulktNR][vdwktNR] =
176 nbnxn_kernel_ElecRF_VdwLJCombGeom_VgrpF_2xnn,
177 nbnxn_kernel_ElecRF_VdwLJCombLB_VgrpF_2xnn,
178 nbnxn_kernel_ElecRF_VdwLJ_VgrpF_2xnn,
179 nbnxn_kernel_ElecRF_VdwLJFSw_VgrpF_2xnn,
180 nbnxn_kernel_ElecRF_VdwLJPSw_VgrpF_2xnn,
181 nbnxn_kernel_ElecRF_VdwLJEwCombGeom_VgrpF_2xnn,
184 nbnxn_kernel_ElecQSTab_VdwLJCombGeom_VgrpF_2xnn,
185 nbnxn_kernel_ElecQSTab_VdwLJCombLB_VgrpF_2xnn,
186 nbnxn_kernel_ElecQSTab_VdwLJ_VgrpF_2xnn,
187 nbnxn_kernel_ElecQSTab_VdwLJFSw_VgrpF_2xnn,
188 nbnxn_kernel_ElecQSTab_VdwLJPSw_VgrpF_2xnn,
189 nbnxn_kernel_ElecQSTab_VdwLJEwCombGeom_VgrpF_2xnn,
192 nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombGeom_VgrpF_2xnn,
193 nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombLB_VgrpF_2xnn,
194 nbnxn_kernel_ElecQSTabTwinCut_VdwLJ_VgrpF_2xnn,
195 nbnxn_kernel_ElecQSTabTwinCut_VdwLJFSw_VgrpF_2xnn,
196 nbnxn_kernel_ElecQSTabTwinCut_VdwLJPSw_VgrpF_2xnn,
197 nbnxn_kernel_ElecQSTabTwinCut_VdwLJEwCombGeom_VgrpF_2xnn,
200 nbnxn_kernel_ElecEw_VdwLJCombGeom_VgrpF_2xnn,
201 nbnxn_kernel_ElecEw_VdwLJCombLB_VgrpF_2xnn,
202 nbnxn_kernel_ElecEw_VdwLJ_VgrpF_2xnn,
203 nbnxn_kernel_ElecEw_VdwLJFSw_VgrpF_2xnn,
204 nbnxn_kernel_ElecEw_VdwLJPSw_VgrpF_2xnn,
205 nbnxn_kernel_ElecEw_VdwLJEwCombGeom_VgrpF_2xnn,
208 nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_VgrpF_2xnn,
209 nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_VgrpF_2xnn,
210 nbnxn_kernel_ElecEwTwinCut_VdwLJ_VgrpF_2xnn,
211 nbnxn_kernel_ElecEwTwinCut_VdwLJFSw_VgrpF_2xnn,
212 nbnxn_kernel_ElecEwTwinCut_VdwLJPSw_VgrpF_2xnn,
213 nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_VgrpF_2xnn,
219 reduce_group_energies(int ng, int ng_2log,
220 const real *VSvdw, const real *VSc,
221 real *Vvdw, real *Vc)
223 const int unrollj = GMX_SIMD_REAL_WIDTH/GMX_SIMD_J_UNROLL_SIZE;
224 const int unrollj_half = unrollj/2;
225 int ng_p2, i, j, j0, j1, c, s;
227 ng_p2 = (1<<ng_2log);
229 /* The size of the x86 SIMD energy group buffer array is:
230 * ng*ng*ng_p2*unrollj_half*simd_width
232 for (i = 0; i < ng; i++)
234 for (j = 0; j < ng; j++)
240 for (j1 = 0; j1 < ng; j1++)
242 for (j0 = 0; j0 < ng; j0++)
244 c = ((i*ng + j1)*ng_p2 + j0)*unrollj_half*unrollj;
245 for (s = 0; s < unrollj_half; s++)
247 Vvdw[i*ng+j0] += VSvdw[c+0];
248 Vvdw[i*ng+j1] += VSvdw[c+1];
249 Vc [i*ng+j0] += VSc [c+0];
250 Vc [i*ng+j1] += VSc [c+1];
258 #else /* GMX_NBNXN_SIMD_2XNN */
260 #include "gromacs/utility/fatalerror.h"
262 #endif /* GMX_NBNXN_SIMD_2XNN */
265 nbnxn_kernel_simd_2xnn(nbnxn_pairlist_set_t gmx_unused *nbl_list,
266 const nbnxn_atomdata_t gmx_unused *nbat,
267 const interaction_const_t gmx_unused *ic,
268 int gmx_unused ewald_excl,
269 rvec gmx_unused *shift_vec,
270 int gmx_unused force_flags,
271 int gmx_unused clearF,
272 real gmx_unused *fshift,
274 real gmx_unused *Vvdw)
275 #ifdef GMX_NBNXN_SIMD_2XNN
278 nbnxn_pairlist_t **nbl;
279 int coulkt, vdwkt = 0;
282 nnbl = nbl_list->nnbl;
285 if (EEL_RF(ic->eeltype) || ic->eeltype == eelCUT)
291 if (ewald_excl == ewaldexclTable)
293 if (ic->rcoulomb == ic->rvdw)
299 coulkt = coulktTAB_TWIN;
304 if (ic->rcoulomb == ic->rvdw)
306 coulkt = coulktEWALD;
310 coulkt = coulktEWALD_TWIN;
315 if (ic->vdwtype == evdwCUT)
317 switch (ic->vdw_modifier)
320 case eintmodPOTSHIFT:
321 switch (nbat->comb_rule)
323 case ljcrGEOM: vdwkt = vdwktLJCUT_COMBGEOM; break;
324 case ljcrLB: vdwkt = vdwktLJCUT_COMBLB; break;
325 case ljcrNONE: vdwkt = vdwktLJCUT_COMBNONE; break;
326 default: gmx_incons("Unknown combination rule");
329 case eintmodFORCESWITCH:
330 vdwkt = vdwktLJFORCESWITCH;
332 case eintmodPOTSWITCH:
333 vdwkt = vdwktLJPOTSWITCH;
336 gmx_incons("Unsupported VdW interaction modifier");
339 else if (ic->vdwtype == evdwPME)
341 if (ic->ljpme_comb_rule == eljpmeLB)
343 gmx_incons("The nbnxn SIMD kernels don't suport LJ-PME with LB");
345 vdwkt = vdwktLJEWALDCOMBGEOM;
349 gmx_incons("Unsupported VdW interaction type");
351 // cppcheck-suppress unreadVariable
352 nthreads = gmx_omp_nthreads_get(emntNonbonded);
353 #pragma omp parallel for schedule(static) num_threads(nthreads)
354 for (nb = 0; nb < nnbl; nb++)
356 // Presently, the kernels do not call C++ code that can throw, so
357 // no need for a try/catch pair in this OpenMP region.
358 nbnxn_atomdata_output_t *out;
361 out = &nbat->out[nb];
363 if (clearF == enbvClearFYes)
365 clear_f(nbat, nb, out->f);
368 if ((force_flags & GMX_FORCE_VIRIAL) && nnbl == 1)
374 fshift_p = out->fshift;
376 if (clearF == enbvClearFYes)
378 clear_fshift(fshift_p);
382 if (!(force_flags & GMX_FORCE_ENERGY))
384 /* Don't calculate energies */
385 p_nbk_noener[coulkt][vdwkt](nbl[nb], nbat,
391 else if (out->nV == 1)
393 /* No energy groups */
397 p_nbk_ener[coulkt][vdwkt](nbl[nb], nbat,
407 /* Calculate energy group contributions */
410 for (i = 0; i < out->nVS; i++)
414 for (i = 0; i < out->nVS; i++)
419 p_nbk_energrp[coulkt][vdwkt](nbl[nb], nbat,
427 reduce_group_energies(nbat->nenergrp, nbat->neg_2log,
428 out->VSvdw, out->VSc,
433 if (force_flags & GMX_FORCE_ENERGY)
435 reduce_energies_over_lists(nbat, nnbl, Vvdw, Vc);
440 gmx_incons("nbnxn_kernel_simd_2xnn called when such kernels "
441 " are not enabled.");
444 #undef GMX_SIMD_J_UNROLL_SIZE