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
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
|
// **************************************************************************
// coul.cu
// -------------------
// Trung Dac Nguyen
//
// Device code for acceleration of the coul/cut pair style
//
// __________________________________________________________________________
// This file is part of the LAMMPS Accelerator Library (LAMMPS_AL)
// __________________________________________________________________________
//
// begin :
// email : ndtrung@umich.edu
// ***************************************************************************
#if defined(NV_KERNEL) || defined(USE_HIP)
#include "lal_aux_fun1.h"
#ifndef _DOUBLE_DOUBLE
_texture( pos_tex,float4);
_texture( q_tex,float);
#else
_texture_2d( pos_tex,int4);
_texture( q_tex,int2);
#endif
#else
#define pos_tex x_
#define q_tex q_
#endif
__kernel void k_coul(const __global numtyp4 *restrict x_,
const __global numtyp *restrict scale,
const int lj_types,
const __global numtyp *restrict sp_cl_in,
const __global int *dev_nbor,
const __global int *dev_packed,
__global acctyp3 *restrict ans,
__global acctyp *restrict engv,
const int eflag, const int vflag, const int inum,
const int nbor_pitch,
const __global numtyp *restrict q_,
const __global numtyp *restrict cutsq,
const numtyp qqrd2e, const int t_per_atom) {
int tid, ii, offset;
atom_info(t_per_atom,ii,tid,offset);
__local numtyp sp_cl[8];
int n_stride;
local_allocate_store_charge();
sp_cl[0]=sp_cl_in[0];
sp_cl[1]=sp_cl_in[1];
sp_cl[2]=sp_cl_in[2];
sp_cl[3]=sp_cl_in[3];
acctyp3 f;
f.x=(acctyp)0; f.y=(acctyp)0; f.z=(acctyp)0;
acctyp energy, e_coul, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
e_coul=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
if (ii<inum) {
int i, numj, nbor, nbor_end;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,nbor_end,nbor);
numtyp4 ix; fetch4(ix,i,pos_tex); //x_[i];
numtyp qtmp; fetch(qtmp,i,q_tex);
int itype=ix.w;
numtyp factor_coul;
for ( ; nbor<nbor_end; nbor+=n_stride) {
ucl_prefetch(dev_packed+nbor+n_stride);
int j=dev_packed[nbor];
factor_coul = sp_cl[sbmask(j)];
j &= NEIGHMASK;
numtyp4 jx; fetch4(jx,j,pos_tex); //x_[j];
int jtype=jx.w;
// Compute r12
numtyp delx = ix.x-jx.x;
numtyp dely = ix.y-jx.y;
numtyp delz = ix.z-jx.z;
numtyp rsq = delx*delx+dely*dely+delz*delz;
int mtype=itype*lj_types+jtype;
if (rsq<cutsq[mtype]) {
numtyp r2inv=ucl_recip(rsq);
numtyp forcecoul, force;
fetch(forcecoul,j,q_tex);
forcecoul *= qqrd2e*scale[mtype]*qtmp*ucl_rsqrt(rsq)*factor_coul;
force = forcecoul * r2inv;
f.x+=delx*force;
f.y+=dely*force;
f.z+=delz*force;
if (EVFLAG && eflag) {
e_coul += forcecoul;
}
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
virial[3] += delx*dely*force;
virial[4] += delx*delz*force;
virial[5] += dely*delz*force;
}
}
} // for nbor
} // if ii
store_answers_q(f,energy,e_coul,virial,ii,inum,tid,t_per_atom,offset,eflag,
vflag,ans,engv);
}
__kernel void k_coul_fast(const __global numtyp4 *restrict x_,
const __global numtyp *restrict scale,
const __global numtyp *restrict sp_cl_in,
const __global int *dev_nbor,
const __global int *dev_packed,
__global acctyp3 *restrict ans,
__global acctyp *restrict engv,
const int eflag, const int vflag, const int inum,
const int nbor_pitch,
const __global numtyp *restrict q_,
const __global numtyp *restrict _cutsq,
const numtyp qqrd2e, const int t_per_atom) {
int tid, ii, offset;
atom_info(t_per_atom,ii,tid,offset);
__local numtyp cutsq[MAX_SHARED_TYPES*MAX_SHARED_TYPES];
__local numtyp sp_cl[4];
int n_stride;
local_allocate_store_charge();
if (tid<4)
sp_cl[tid]=sp_cl_in[tid];
if (tid<MAX_SHARED_TYPES*MAX_SHARED_TYPES) {
cutsq[tid]=_cutsq[tid];
}
acctyp3 f;
f.x=(acctyp)0; f.y=(acctyp)0; f.z=(acctyp)0;
acctyp energy, e_coul, virial[6];
if (EVFLAG) {
energy=(acctyp)0;
e_coul=(acctyp)0;
for (int i=0; i<6; i++) virial[i]=(acctyp)0;
}
__syncthreads();
if (ii<inum) {
int i, numj, nbor, nbor_end;
nbor_info(dev_nbor,dev_packed,nbor_pitch,t_per_atom,ii,offset,i,numj,
n_stride,nbor_end,nbor);
numtyp4 ix; fetch4(ix,i,pos_tex); //x_[i];
numtyp qtmp; fetch(qtmp,i,q_tex);
int iw=ix.w;
int itype=fast_mul((int)MAX_SHARED_TYPES,iw);
numtyp factor_coul;
for ( ; nbor<nbor_end; nbor+=n_stride) {
ucl_prefetch(dev_packed+nbor+n_stride);
int j=dev_packed[nbor];
factor_coul = sp_cl[sbmask(j)];
j &= NEIGHMASK;
numtyp4 jx; fetch4(jx,j,pos_tex); //x_[j];
int mtype=itype+jx.w;
// Compute r12
numtyp delx = ix.x-jx.x;
numtyp dely = ix.y-jx.y;
numtyp delz = ix.z-jx.z;
numtyp rsq = delx*delx+dely*dely+delz*delz;
if (rsq<cutsq[mtype]) {
numtyp r2inv=ucl_recip(rsq);
numtyp forcecoul, force;
fetch(forcecoul,j,q_tex);
forcecoul *= qqrd2e*scale[mtype]*qtmp*ucl_rsqrt(rsq)*factor_coul;
force = forcecoul * r2inv;
f.x+=delx*force;
f.y+=dely*force;
f.z+=delz*force;
if (EVFLAG && eflag) {
e_coul += forcecoul;
}
if (EVFLAG && vflag) {
virial[0] += delx*delx*force;
virial[1] += dely*dely*force;
virial[2] += delz*delz*force;
virial[3] += delx*dely*force;
virial[4] += delx*delz*force;
virial[5] += dely*delz*force;
}
}
} // for nbor
} // if ii
store_answers_q(f,energy,e_coul,virial,ii,inum,tid,t_per_atom,offset,eflag,
vflag,ans,engv);
}
|
