-
Notifications
You must be signed in to change notification settings - Fork 0
/
kernel.c
173 lines (148 loc) · 3.89 KB
/
kernel.c
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
#include "kernel.h"
extern "C" struct sparse_matrix **create_sparse_matrix()
{
struct sparse_matrix *mx = (sparse_matrix*) malloc(sizeof(sparse_matrix))
mx->m = 0;
mx->d = 0;
mx->nnz = 0;
mx->val = 0;
mx->rowptr = 0;
mx->colind = 0;
return &mx;
}
extern "C" void delete_sparse_matrix(struct sparse_matrix* matrix)
{
free(matrix->val);
free(matrix->rowptr);
free(matrix->colind);
free(matrix);
}
extern "C" struct sparse_matrix **compute_rbf(const struct sparse_matrix *X, const struct sparse_matrix *Z, float sigma)
{
unsigned long int i, j, k, l;
unsigned long int aa, ab, ba, bb;
double l2, tmp;
double_vec_t val;
int_vec_t rowptr;
int_vec_t colind;
//#pragma omp parallel for private(j, l2, tmp, Q, aa, ab, ba, bb, k, l) shared(X, Z, labels)
for (i = 0; i < Z->m; ++i)
{
aa = Z->rowptr[i];
ab = Z->rowptr[i+1];
// tmp_data = (double*) malloc(sizeof(double)*(ab - aa));
// memcpy(tmp_data, Z->val+aa, sizeof(double)*(ab - aa));
// #pragma omp parallel for private(l2, tmp, ba, bb, k, l) shared(X, tmp_data, labels, aa, ab, Q)
unsigned long int offset = 0;
rowptr.push_back(offset);
for(j = 0; j < X->m; ++j)
{
ba = X->rowptr[j];
bb = X->rowptr[j+1];
k = aa;
l = ba;
l2 = 0;
while (k < ab && l < bb)
{
if (Z->colind[k] == X->colind[l])
{
tmp = (Z->val[k] - X->val[l]);
l2 += tmp*tmp;
k++;
l++;
}
else if (Z->colind[k] > X->colind[l])
{
tmp = X->val[l];
l2 += tmp*tmp;
l++;
}
else
{
tmp = Z->val[k];
l2 += tmp*tmp;
k++;
}
}
if (l2 > 0)
{
val.push_back(exp(-l2/(2.0*sigma)));
colind.push_back(j);
offset++;
}
}
rowptr.push_back(offset);
}
// re-format the kernel result into a sparse matrix
struct sparse_matrix *Z = (sparse_matrix*) malloc(sizeof(sparse_matrix));
Z->m = rowptr.size()-1;
Z->d = rowptr.size()-1;
Z->nnz = val.size();
Z->val = (double*) malloc(sizeof(double)*val.size());
Z->rowptr = (unsigned long int*) malloc(sizeof(unsigned long int)*rowptr.size());
Z->colind = (int*) malloc(sizeof(int)*colind.size());
std:copy(val.begin(), val.end(), Z->val);
std:copy(colind.begin(), colind.end(), Z->colind);
std:copy(rowptr.begin(), rowptr.end(), Z->rowptr);
return &Z;
}
extern "C" void compute_linear(const struct sparse_matrix *X, const struct sparse_matrix *Z)
{
unsigned long int i, j, k, l;
unsigned long int aa, ab, ba, bb;
double l2, tmp;
double_vec_t val;
int_vec_t rowptr;
int_vec_t colind;
//#pragma omp parallel for private(j, l2, tmp, Q, aa, ab, ba, bb, k, l) shared(X, Z, labels)
for (i = 0; i < Z->m; ++i)
{
aa = Z->rowptr[i];
ab = Z->rowptr[i+1];
// tmp_data = (double*) malloc(sizeof(double)*(ab - aa));
// memcpy(tmp_data, Z->val+aa, sizeof(double)*(ab - aa));
// #pragma omp parallel for private(l2, tmp, ba, bb, k, l) shared(X, tmp_data, labels, aa, ab, Q)
unsigned long int offset = 0;
rowptr.push_back(offset);
for(j = 0; j < X->m; ++j)
{
ba = X->rowptr[j];
bb = X->rowptr[j+1];
k = aa;
l = ba;
l2 = 0;
while (k < ab && l < bb)
{
if (Z->colind[k] == X->colind[l])
{
l2 += Z->val[k] * X->val[l];
k++;
l++;
}
else if (Z->colind[k] > X->colind[l])
l++;
else
k++;
}
if (l2 > 0)
{
val.push_back(l2);
colind.push_back(j);
offset++;
}
}
rowptr.push_back(offset);
}
// re-format the kernel result into a sparse matrix
struct sparse_matrix *Z = (sparse_matrix*) malloc(sizeof(sparse_matrix));
Z->m = rowptr.size()-1;
Z->d = rowptr.size()-1;
Z->nnz = val.size();
Z->val = (double*) malloc(sizeof(double)*val.size());
Z->rowptr = (unsigned long int*) malloc(sizeof(unsigned long int)*rowptr.size());
Z->colind = (int*) malloc(sizeof(int)*colind.size());
std:copy(val.begin(), val.end(), Z->val);
std:copy(colind.begin(), colind.end(), Z->colind);
std:copy(rowptr.begin(), rowptr.end(), Z->rowptr);
return &Z;
}