1 // clang-format off
2 /* ----------------------------------------------------------------------
3    LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
4    https://www.lammps.org/, Sandia National Laboratories
5    Steve Plimpton, sjplimp@sandia.gov
6 
7    This software is distributed under the GNU General Public License.
8 
9    See the README file in the top-level LAMMPS directory.
10 ------------------------------------------------------------------------- */
11 
12 /* ----------------------------------------------------------------------
13    Contributing author: Axel Kohlmeyer (Temple U)
14 ------------------------------------------------------------------------- */
15 
16 #include "pair_zbl_omp.h"
17 
18 #include "atom.h"
19 #include "comm.h"
20 #include "force.h"
21 #include "neigh_list.h"
22 #include "suffix.h"
23 
24 #include <cmath>
25 
26 #include "omp_compat.h"
27 using namespace LAMMPS_NS;
28 
29 /* ---------------------------------------------------------------------- */
30 
PairZBLOMP(LAMMPS * lmp)31 PairZBLOMP::PairZBLOMP(LAMMPS *lmp) :
32   PairZBL(lmp), ThrOMP(lmp, THR_PAIR)
33 {
34   suffix_flag |= Suffix::OMP;
35   respa_enable = 0;
36 }
37 
38 /* ---------------------------------------------------------------------- */
39 
compute(int eflag,int vflag)40 void PairZBLOMP::compute(int eflag, int vflag)
41 {
42   ev_init(eflag,vflag);
43 
44   const int nall = atom->nlocal + atom->nghost;
45   const int nthreads = comm->nthreads;
46   const int inum = list->inum;
47 
48 #if defined(_OPENMP)
49 #pragma omp parallel LMP_DEFAULT_NONE LMP_SHARED(eflag,vflag)
50 #endif
51   {
52     int ifrom, ito, tid;
53 
54     loop_setup_thr(ifrom, ito, tid, inum, nthreads);
55     ThrData *thr = fix->get_thr(tid);
56     thr->timer(Timer::START);
57     ev_setup_thr(eflag, vflag, nall, eatom, vatom, nullptr, thr);
58 
59     if (evflag) {
60       if (eflag) {
61         if (force->newton_pair) eval<1,1,1>(ifrom, ito, thr);
62         else eval<1,1,0>(ifrom, ito, thr);
63       } else {
64         if (force->newton_pair) eval<1,0,1>(ifrom, ito, thr);
65         else eval<1,0,0>(ifrom, ito, thr);
66       }
67     } else {
68       if (force->newton_pair) eval<0,0,1>(ifrom, ito, thr);
69       else eval<0,0,0>(ifrom, ito, thr);
70     }
71     thr->timer(Timer::PAIR);
72     reduce_thr(this, eflag, vflag, thr);
73   } // end of omp parallel region
74 }
75 
76 template <int EVFLAG, int EFLAG, int NEWTON_PAIR>
eval(int iifrom,int iito,ThrData * const thr)77 void PairZBLOMP::eval(int iifrom, int iito, ThrData * const thr)
78 {
79   const dbl3_t * _noalias const x = (dbl3_t *) atom->x[0];
80   dbl3_t * _noalias const f = (dbl3_t *) thr->get_f()[0];
81   const int * _noalias const type = atom->type;
82   const int * _noalias const ilist = list->ilist;
83   const int * _noalias const numneigh = list->numneigh;
84   const int * const * const firstneigh = list->firstneigh;
85 
86   double xtmp,ytmp,ztmp,delx,dely,delz,fxtmp,fytmp,fztmp;
87   double rsq,t,fswitch,eswitch,evdwl,fpair;
88 
89   const int nlocal = atom->nlocal;
90   int j,jj,jnum,jtype;
91 
92   evdwl = 0.0;
93 
94   // loop over neighbors of my atoms
95 
96   for (int ii = iifrom; ii < iito; ++ii) {
97     const int i = ilist[ii];
98     const int itype = type[i];
99     const int    * _noalias const jlist = firstneigh[i];
100     const double * _noalias const sw1i = sw1[itype];
101     const double * _noalias const sw2i = sw2[itype];
102     const double * _noalias const sw3i = sw3[itype];
103     const double * _noalias const sw4i = sw4[itype];
104     const double * _noalias const sw5i = sw5[itype];
105 
106     xtmp = x[i].x;
107     ytmp = x[i].y;
108     ztmp = x[i].z;
109     jnum = numneigh[i];
110     fxtmp=fytmp=fztmp=0.0;
111 
112     for (jj = 0; jj < jnum; jj++) {
113       j = jlist[jj];
114       j &= NEIGHMASK;
115 
116       delx = xtmp - x[j].x;
117       dely = ytmp - x[j].y;
118       delz = ztmp - x[j].z;
119       rsq = delx*delx + dely*dely + delz*delz;
120       jtype = type[j];
121 
122       if (rsq < cut_globalsq) {
123         const double r = sqrt(rsq);
124         fpair = dzbldr(r, itype, jtype);
125 
126         if (r > cut_inner) {
127           t = r - cut_inner;
128           fswitch = t*t *
129             (sw1i[jtype] + sw2i[jtype]*t);
130           fpair += fswitch;
131         }
132 
133         fpair *= -1.0/r;
134         fxtmp += delx*fpair;
135         fytmp += dely*fpair;
136         fztmp += delz*fpair;
137 
138         if (NEWTON_PAIR || j < nlocal) {
139           f[j].x -= delx*fpair;
140           f[j].y -= dely*fpair;
141           f[j].z -= delz*fpair;
142         }
143 
144         if (EFLAG) {
145           evdwl = e_zbl(r, itype, jtype);
146           evdwl += sw5i[jtype];
147           if (r > cut_inner) {
148             eswitch = t*t*t *
149               (sw3i[jtype] + sw4i[jtype]*t);
150             evdwl += eswitch;
151           }
152         }
153 
154         if (EVFLAG) ev_tally_thr(this,i,j,nlocal,NEWTON_PAIR,
155                                  evdwl,0.0,fpair,delx,dely,delz,thr);
156       }
157     }
158     f[i].x += fxtmp;
159     f[i].y += fytmp;
160     f[i].z += fztmp;
161   }
162 }
163 
164 /* ---------------------------------------------------------------------- */
165 
memory_usage()166 double PairZBLOMP::memory_usage()
167 {
168   double bytes = memory_usage_thr();
169   bytes += PairZBL::memory_usage();
170 
171   return bytes;
172 }
173