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
|
/***************************************************************************
born_coul_wolf_cs.cpp
-------------------
Trung Dac Nguyen (Northwestern)
Class for acceleration of the born/coul/wolf/cs pair style.
__________________________________________________________________________
This file is part of the LAMMPS Accelerator Library (LAMMPS_AL)
__________________________________________________________________________
begin :
email : ndactrung@gmail.com
***************************************************************************/
#ifdef USE_OPENCL
#include "born_coul_wolf_cs_cl.h"
#elif defined(USE_CUDART)
const char *born_coul_wolf_cs=0;
#else
#include "born_coul_wolf_cs_cubin.h"
#endif
#include "lal_born_coul_wolf_cs.h"
#include <cassert>
namespace LAMMPS_AL {
#define BornCoulWolfCST BornCoulWolfCS<numtyp, acctyp>
extern Device<PRECISION,ACC_PRECISION> device;
template <class numtyp, class acctyp>
int BornCoulWolfCST::init(const int ntypes, double **host_cutsq, double **host_rhoinv,
double **host_born1, double **host_born2, double **host_born3,
double **host_a, double **host_c, double **host_d,
double **host_sigma, 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,
double **host_cut_ljsq, const double host_cut_coulsq,
double *host_special_coul, const double qqrd2e,
const double alf, const double e_shift, const double f_shift) {
int success;
success=this->init_atomic(nlocal,nall,max_nbors,maxspecial,cell_size,gpu_split,
_screen,born_coul_wolf_cs,"k_born_coul_wolf_cs",1);
if (success!=0)
return success;
// If atom type constants fit in shared memory use fast kernel
int lj_types=ntypes;
this->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;
this->shared_types=true;
}
this->_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;
this->coeff1.alloc(lj_types*lj_types,*(this->ucl_device),UCL_READ_ONLY);
this->atom->type_pack4(ntypes,lj_types,this->coeff1,host_write,host_rhoinv,
host_born1,host_born2,host_born3);
this->coeff2.alloc(lj_types*lj_types,*(this->ucl_device),UCL_READ_ONLY);
this->atom->type_pack4(ntypes,lj_types,this->coeff2,host_write,host_a,host_c,
host_d,host_offset);
this->cutsq_sigma.alloc(lj_types*lj_types,*(this->ucl_device),UCL_READ_ONLY);
this->atom->type_pack4(ntypes,lj_types,this->cutsq_sigma,host_write,host_cutsq,
host_cut_ljsq,host_sigma);
this->sp_lj.alloc(8,*(this->ucl_device),UCL_READ_ONLY);
for (int i=0; i<4; i++) {
host_write[i]=host_special_lj[i];
host_write[i+4]=host_special_coul[i];
}
ucl_copy(this->sp_lj,host_write,8,false);
this->_cut_coulsq=host_cut_coulsq;
this->_qqrd2e=qqrd2e;
this->_alf=alf;
this->_e_shift=e_shift;
this->_f_shift=f_shift;
this->_allocated=true;
this->_max_bytes=this->coeff1.row_bytes()+this->coeff2.row_bytes()
+this->cutsq_sigma.row_bytes()+this->sp_lj.row_bytes();
return 0;
}
template class BornCoulWolfCS<PRECISION,ACC_PRECISION>;
}
|
