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
|
#ifndef CAFFE2_UTILS_MATH_DETAIL_H_
#define CAFFE2_UTILS_MATH_DETAIL_H_
namespace caffe2 {
class CPUContext;
namespace math {
namespace detail {
// proxy to a class because of partial specialization limitations for functions
template<typename T, class Context, int FixedSize>
struct ScaleImpl {
inline void operator()(
const int N,
const float alpha,
const T* x,
T* y,
Context* context) {
Scale(N, alpha, x, y, context);
}
};
// Put light-weight implementations in .h file to enable inlining
template<typename T>
struct ScaleImpl<T, CPUContext, 1> {
inline void operator()(
const int N,
const float alpha,
const T* x,
T* y,
CPUContext* /*context*/) {
TORCH_DCHECK_EQ(N, 1);
*y = *x * alpha;
}
};
template<typename T, class Context, int FixedSize>
struct AxpyImpl {
inline void operator()(
const int N,
const float alpha,
const T* x,
T* y,
Context* context) {
Axpy(N, alpha, x, y, context);
}
};
// Put light-weight implementations in .h file to enable inlining
template<typename T>
struct AxpyImpl<T, CPUContext, 1> {
inline void operator()(
const int N,
const float alpha,
const T* x,
T* y,
CPUContext* /*context*/) {
TORCH_DCHECK_EQ(N, 1);
*y += *x * alpha;
}
};
} // namespace detail
template <typename T, class Context, int FixedSize>
inline void ScaleFixedSize(
const int N,
const float alpha,
const T* x,
T* y,
Context* context) {
detail::ScaleImpl<T, Context, FixedSize>()(N, alpha, x, y, context);
}
template <typename T, class Context, int FixedSize>
inline void AxpyFixedSize(
const int N,
const float alpha,
const T* x,
T* y,
Context* context) {
detail::AxpyImpl<T, Context, FixedSize>()(N, alpha, x, y, context);
}
} // namespace math
} // namespace caffe2
#endif // CAFFE2_UTILS_MATH_DETAIL_H_
|
