megenginelite-sys 1.8.2

A safe megenginelite wrapper in Rust
Documentation 
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

/**
 * \file dnn/src/cuda/local_share/opr_impl.h
 * MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
 *
 * Copyright (c) 2014-2021 Megvii Inc. All rights reserved.
 *
 * Unless required by applicable law or agreed to in writing,
 * software distributed under the License is distributed on an
 * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 */
#pragma once
#include "megdnn/oprs.h"
#include "src/cuda/utils.h"

namespace megdnn {
namespace cuda {

class LocalShareForwardImpl : public LocalShareForward {
public:
    using LocalShareForward::LocalShareForward;
    void exec(
            _megdnn_tensor_in src, _megdnn_tensor_in filter, _megdnn_tensor_out dst,
            _megdnn_workspace workspace) override;
    size_t get_workspace_in_bytes(
            const TensorLayout& src, const TensorLayout& filter,
            const TensorLayout& dst) override;
    const char* get_algorithm_set_name() const override;

    class AlgoBase;
    class AlgoCHWNBatchSizeAware;
    class AlgoCHWNBatchSizeAwareSmallImage;
    class AlgoBatchedMatMul;

    class AlgoPack;

    static const AlgoPack& algo_pack() { return sm_algo_pack; }
    Algorithm* get_algorithm_from_desc(const AlgorithmDesc& desc) override;

protected:
    std::vector<Algorithm*> get_all_algorithms(
            const TensorLayout& filter, const TensorLayout& diff,
            const TensorLayout& grad) override;
    std::vector<Algorithm*> get_all_algorithms_safe(
            const TensorLayout& src, const TensorLayout& filter,
            const TensorLayout& dst) override;
    Algorithm* get_algorithm_heuristic(
            const TensorLayout& src, const TensorLayout& filter,
            const TensorLayout& dst, size_t workspace_limit_in_bytes,
            const AlgoAttribute& positive_attr,
            const AlgoAttribute& negative_attr) override;

private:
    static AlgoPack sm_algo_pack;
};

class LocalShareBackwardDataImpl : public LocalShareBackwardData {
public:
    using LocalShareBackwardData::LocalShareBackwardData;
    void exec(
            _megdnn_tensor_in filter, _megdnn_tensor_in diff, _megdnn_tensor_out grad,
            _megdnn_workspace workspace) override;
    size_t get_workspace_in_bytes(
            const TensorLayout& filter, const TensorLayout& diff,
            const TensorLayout& grad) override;
    const char* get_algorithm_set_name() const override;

    class AlgoBase;
    class AlgoImplicitGemm;
    class AlgoBatchedMatMul;

    class AlgoPack;

    static const AlgoPack& algo_pack() { return sm_algo_pack; }
    Algorithm* get_algorithm_from_desc(const AlgorithmDesc& desc) override;

protected:
    std::vector<Algorithm*> get_all_algorithms(
            const TensorLayout& filter, const TensorLayout& diff,
            const TensorLayout& grad) override;
    std::vector<Algorithm*> get_all_algorithms_safe(
            const TensorLayout& filter, const TensorLayout& diff,
            const TensorLayout& grad) override;
    Algorithm* get_algorithm_heuristic(
            const TensorLayout& filter, const TensorLayout& diff,
            const TensorLayout& grad, size_t workspace_limit_in_bytes,
            const AlgoAttribute& positive_attr,
            const AlgoAttribute& negative_attr) override;

private:
    static AlgoPack sm_algo_pack;
};

class LocalShareBackwardFilterImpl : public LocalShareBackwardFilter {
public:
    using LocalShareBackwardFilter::LocalShareBackwardFilter;
    void exec(
            _megdnn_tensor_in src, _megdnn_tensor_in diff, _megdnn_tensor_out grad,
            _megdnn_workspace workspace) override;
    size_t get_workspace_in_bytes(
            const TensorLayout& src, const TensorLayout& diff,
            const TensorLayout& grad) override;
    const char* get_algorithm_set_name() const override;

    class AlgoBase;
    class AlgoImplicitGemm;
    class AlgoBatchedMatMul;

    class AlgoPack;

    static const AlgoPack& algo_pack() { return sm_algo_pack; }
    Algorithm* get_algorithm_from_desc(const AlgorithmDesc& desc) override;

protected:
    std::vector<Algorithm*> get_all_algorithms(
            const TensorLayout& src, const TensorLayout& diff,
            const TensorLayout& grad) override;
    std::vector<Algorithm*> get_all_algorithms_safe(
            const TensorLayout& src, const TensorLayout& diff,
            const TensorLayout& grad) override;
    Algorithm* get_algorithm_heuristic(
            const TensorLayout& src, const TensorLayout& diff, const TensorLayout& grad,
            size_t workspace_limit_in_bytes, const AlgoAttribute& positive_attr,
            const AlgoAttribute& negative_attr) override;

private:
    static AlgoPack sm_algo_pack;
};

}  // namespace cuda
}  // namespace megdnn

// vim: syntax=cpp.doxygen