1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
|
/***************************************************************************
yukawa.cpp
-------------------
Trung Dac Nguyen (ORNL)
Class for acceleration of the yukawa pair style.
__________________________________________________________________________
This file is part of the LAMMPS Accelerator Library (LAMMPS_AL)
__________________________________________________________________________
begin :
email : nguyentd@ornl.gov
***************************************************************************/
#if defined(USE_OPENCL)
#include "yukawa_cl.h"
#elif defined(USE_CUDART)
const char *yukawa=0;
#else
#include "yukawa_cubin.h"
#endif
#include "lal_yukawa.h"
#include <cassert>
namespace LAMMPS_AL {
#define YukawaT Yukawa<numtyp, acctyp>
extern Device<PRECISION,ACC_PRECISION> device;
template <class numtyp, class acctyp>
YukawaT::Yukawa() : BaseAtomic<numtyp,acctyp>(), _allocated(false) {
}
template <class numtyp, class acctyp>
YukawaT::~Yukawa() {
clear();
}
template <class numtyp, class acctyp>
int YukawaT::bytes_per_atom(const int max_nbors) const {
return this->bytes_per_atom_atomic(max_nbors);
}
template <class numtyp, class acctyp>
int YukawaT::init(const int ntypes,
double **host_cutsq, double kappa,
double **host_a, double **host_offset,
double *host_special_lj, const int nlocal,
const int nall, const int max_nbors,
const int maxspecial, const double cell_size,
const double gpu_split, FILE *_screen) {
int success;
success=this->init_atomic(nlocal,nall,max_nbors,maxspecial,cell_size,gpu_split,
_screen,yukawa,"k_yukawa");
if (success!=0)
return success;
// If atom type constants fit in shared memory use fast kernel
int lj_types=ntypes;
shared_types=false;
int max_shared_types=this->device->max_shared_types();
if (lj_types<=max_shared_types && this->_block_size>=max_shared_types) {
lj_types=max_shared_types;
shared_types=true;
}
_lj_types=lj_types;
// Allocate a host write buffer for data initialization
UCL_H_Vec<numtyp> host_write(lj_types*lj_types*32,*(this->ucl_device),
UCL_WRITE_ONLY);
for (int i=0; i<lj_types*lj_types; i++)
host_write[i]=0.0;
coeff.alloc(lj_types*lj_types,*(this->ucl_device),UCL_READ_ONLY);
this->atom->type_pack4(ntypes,lj_types,coeff,host_write,host_a,host_offset,
host_cutsq);
UCL_H_Vec<double> dview;
sp_lj.alloc(4,*(this->ucl_device),UCL_READ_ONLY);
dview.view(host_special_lj,4,*(this->ucl_device));
ucl_copy(sp_lj,dview,false);
_kappa = kappa;
_allocated=true;
this->_max_bytes=coeff.row_bytes()+sp_lj.row_bytes();
return 0;
}
template <class numtyp, class acctyp>
void YukawaT::clear() {
if (!_allocated)
return;
_allocated=false;
coeff.clear();
sp_lj.clear();
this->clear_atomic();
}
template <class numtyp, class acctyp>
double YukawaT::host_memory_usage() const {
return this->host_memory_usage_atomic()+sizeof(Yukawa<numtyp,acctyp>);
}
// ---------------------------------------------------------------------------
// Calculate energies, forces, and torques
// ---------------------------------------------------------------------------
template <class numtyp, class acctyp>
int YukawaT::loop(const int eflag, const int vflag) {
// Compute the block size and grid size to keep all cores busy
const int BX=this->block_size();
int GX=static_cast<int>(ceil(static_cast<double>(this->ans->inum())/
(BX/this->_threads_per_atom)));
int ainum=this->ans->inum();
int nbor_pitch=this->nbor->nbor_pitch();
this->time_pair.start();
if (shared_types) {
this->k_pair_sel->set_size(GX,BX);
this->k_pair_sel->run(&this->atom->x, &coeff, &_kappa, &sp_lj,
&this->nbor->dev_nbor, &this->_nbor_data->begin(),
&this->ans->force, &this->ans->engv, &eflag,
&vflag, &ainum, &nbor_pitch,
&this->_threads_per_atom);
} else {
this->k_pair.set_size(GX,BX);
this->k_pair.run(&this->atom->x, &coeff, &_kappa, &_lj_types, &sp_lj,
&this->nbor->dev_nbor, &this->_nbor_data->begin(),
&this->ans->force, &this->ans->engv, &eflag, &vflag,
&ainum, &nbor_pitch, &this->_threads_per_atom);
}
this->time_pair.stop();
return GX;
}
template class Yukawa<PRECISION,ACC_PRECISION>;
}
|
