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

/**
 * \file dnn/src/cuda/roi_align/opr_impl.cpp
 * 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.
 */
#include "src/cuda/roi_align/opr_impl.h"

#include "src/common/roi_align_helper.h"
#include "src/cuda/roi_align/roi_align.cuh"
#include "src/cuda/utils.h"

namespace megdnn {
namespace cuda {

void ROIAlignForwardImpl::exec(
        _megdnn_tensor_in src, _megdnn_tensor_in rois, _megdnn_tensor_out dst,
        _megdnn_tensor_out index, _megdnn_workspace workspace) {
    check_exec(src.layout, rois.layout, dst.layout, index.layout, workspace.size);
    auto stream = cuda_stream(handle());
    int nthreads = dst.layout.total_nr_elems();
    float spatial_scale = param().spatial_scale;
    float offset = param().offset;
    int sample_height = param().sample_height;
    int sample_width = param().sample_width;
    int channels = src.layout.shape[1];
    int height = src.layout.shape[2];
    int width = src.layout.shape[3];
    int pooled_height = dst.layout.shape[2];
    int pooled_width = dst.layout.shape[3];
    using namespace ::megdnn::roi_align;
    using namespace ::megdnn::cuda::roi_align;
#define cb(DType)                                                                  \
    if (src.layout.dtype == DType()) {                                             \
        using T = typename DTypeTrait<DType>::ctype;                               \
        switch (param().mode) {                                                    \
            case param::ROIAlign::Mode::MAX:                                       \
                forward_proxy<T, MaxPooler<T>>(                                    \
                        nthreads, src.ptr<T>(), spatial_scale, offset, channels,   \
                        height, width, pooled_height, pooled_width, sample_height, \
                        sample_width, rois.ptr<T>(), dst.ptr<T>(),                 \
                        index.ptr<dt_int32>(), stream);                            \
                break;                                                             \
            case param::ROIAlign::Mode::AVERAGE:                                   \
                forward_proxy<T, AveragePooler<T>>(                                \
                        nthreads, src.ptr<T>(), spatial_scale, offset, channels,   \
                        height, width, pooled_height, pooled_width, sample_height, \
                        sample_width, rois.ptr<T>(), dst.ptr<T>(),                 \
                        index.ptr<dt_int32>(), stream);                            \
                break;                                                             \
            default:                                                               \
                megdnn_assert_internal(false);                                     \
        }                                                                          \
    }
    MEGDNN_FOREACH_COMPUTING_DTYPE_FLOAT(cb)
#undef cb
}

void ROIAlignBackwardImpl::exec(
        _megdnn_tensor_in diff, _megdnn_tensor_in rois, _megdnn_tensor_in index,
        _megdnn_tensor_out grad, _megdnn_workspace workspace) {
    check_exec(diff.layout, rois.layout, index.layout, grad.layout, workspace.size);
    auto stream = cuda_stream(handle());
    int nthreads = diff.layout.total_nr_elems();
    float spatial_scale = param().spatial_scale;
    float offset = param().offset;
    int sample_height = param().sample_height;
    int sample_width = param().sample_width;
    int channels = grad.layout.shape[1];
    int height = grad.layout.shape[2];
    int width = grad.layout.shape[3];
    int pooled_height = diff.layout.shape[2];
    int pooled_width = diff.layout.shape[3];
    using namespace ::megdnn::roi_align;
    using namespace ::megdnn::cuda::roi_align;
    cuda_check(cudaMemsetAsync(
            grad.raw_ptr(), 0, grad.layout.total_nr_elems() * grad.layout.dtype.size(),
            stream));
#define cb(DType)                                                                      \
    if (diff.layout.dtype == DType()) {                                                \
        using T = typename DTypeTrait<DType>::ctype;                                   \
        switch (param().mode) {                                                        \
            case param::ROIAlign::Mode::MAX:                                           \
                roi_align::backward_proxy<T, BwdMaxPooler<T>>(                         \
                        nthreads, diff.ptr<T>(), index.ptr<dt_int32>(), spatial_scale, \
                        offset, channels, height, width, pooled_height, pooled_width,  \
                        sample_height, sample_width, rois.ptr<T>(), grad.ptr<T>(),     \
                        stream);                                                       \
                break;                                                                 \
            case param::ROIAlign::Mode::AVERAGE:                                       \
                roi_align::backward_proxy<T, BwdAveragePooler<T>>(                     \
                        nthreads, diff.ptr<T>(), index.ptr<dt_int32>(), spatial_scale, \
                        offset, channels, height, width, pooled_height, pooled_width,  \
                        sample_height, sample_width, rois.ptr<T>(), grad.ptr<T>(),     \
                        stream);                                                       \
                break;                                                                 \
            default:                                                                   \
                megdnn_assert_internal(false);                                         \
        }                                                                              \
    }
    MEGDNN_FOREACH_COMPUTING_DTYPE_FLOAT(cb)
#undef cb
}

}  // namespace cuda
}  // namespace megdnn
// vim: syntax=cpp.doxygen