shortcut.hpp
2.34 KB
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
// This file is part of OpenCV project.
// It is subject to the license terms in the LICENSE file found in the top-level directory
// of this distribution and at http://opencv.org/license.html.
#ifndef OPENCV_DNN_SRC_CUDA4DNN_PRIMITIVES_SHORTCUT_HPP
#define OPENCV_DNN_SRC_CUDA4DNN_PRIMITIVES_SHORTCUT_HPP
#include "../../op_cuda.hpp"
#include "../csl/stream.hpp"
#include "../csl/tensor.hpp"
#include "../csl/tensor_ops.hpp"
#include "../kernels/shortcut.hpp"
#include <opencv2/core.hpp>
#include <utility>
namespace cv { namespace dnn { namespace cuda4dnn {
template <class T>
class ShortcutOp final : public CUDABackendNode {
public:
using wrapper_type = GetCUDABackendWrapperType<T>;
ShortcutOp(csl::Stream stream_) : stream(std::move(stream_)) { }
void forward(
const std::vector<cv::Ptr<BackendWrapper>>& inputs,
const std::vector<cv::Ptr<BackendWrapper>>& outputs,
csl::Workspace& workspace) override
{
CV_Assert(outputs.size() == 1);
auto output_wrapper = outputs[0].dynamicCast<wrapper_type>();
auto output = output_wrapper->getSpan();
auto input_wrapper = inputs[0].dynamicCast<wrapper_type>();
auto input = input_wrapper->getView();
/* output shape is determined by the input shape */
CV_Assert(is_shape_same(output, input));
for (int i = 1; i < inputs.size(); i++)
{
auto from_wrapper = inputs[i].dynamicCast<wrapper_type>();
auto from = from_wrapper->getView();
CV_Assert(output.rank() == from.rank());
for (int i = 0; i < output.rank(); i++) {
if (i != 1) {
CV_Assert(from.get_axis_size(i) == output.get_axis_size(i));
}
}
if (i == 1)
{
/* optimized path for first two inputs */
kernels::input_shortcut<T>(stream, output, input, from);
}
else
{
kernels::input_shortcut<T>(stream, output, output, from);
}
}
}
private:
csl::Stream stream;
};
}}} /* namespace cv::dnn::cuda4dnn */
#endif /* OPENCV_DNN_SRC_CUDA4DNN_PRIMITIVES_SHORTCUT_HPP */