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
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
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197

/**
 * \file dnn/src/naive/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/naive/roi_align/opr_impl.h"

#include "src/common/roi_align_helper.h"
#include "src/common/utils.h"
#include "src/naive/handle.h"

using namespace megdnn;
using namespace roi_align;

namespace {

using Param = megdnn::ROIAlign::Param;

template <typename T, typename Pooler>
void forward_impl(
        _megdnn_tensor_in src, _megdnn_tensor_in rois, _megdnn_tensor_in dst,
        _megdnn_tensor_out index, float spatial_scale, float offset,
        const int sample_height, const int sample_width) {
    size_t channels = src.layout[1], hi = src.layout[2], wi = src.layout[3];
    size_t pooled_height = dst.layout[2], pooled_width = dst.layout[3];

    size_t total_nr_elems = dst.layout.total_nr_elems();
    int height = hi, width = wi;
    for (size_t idx = 0; idx < total_nr_elems; ++idx) {
        int pw = idx % pooled_width;
        int ph = (idx / pooled_width) % pooled_height;
        int c = (idx / pooled_width / pooled_height) % channels;
        int n = idx / pooled_width / pooled_height / channels;

        auto rois_ptr = rois.ptr<T>() + n * 5;
        int roi_batch_ind = rois_ptr[0];
        float roi_start_w = rois_ptr[1] * spatial_scale - offset;
        float roi_start_h = rois_ptr[2] * spatial_scale - offset;
        float roi_end_w = rois_ptr[3] * spatial_scale - offset;
        float roi_end_h = rois_ptr[4] * spatial_scale - offset;

        float roi_width = std::max(roi_end_w - roi_start_w, ((float)(0.0)));
        float roi_height = std::max(roi_end_h - roi_start_h, ((float)(0.0)));
        float bin_size_h =
                static_cast<float>(roi_height) / static_cast<float>(pooled_height);
        float bin_size_w =
                static_cast<float>(roi_width) / static_cast<float>(pooled_width);

        auto feat_map_ptr =
                src.ptr<T>() + (roi_batch_ind * channels + c) * height * width;
        float sample_h_rate = 1.0f / float(sample_height);
        float sample_w_rate = 1.0f / float(sample_width);
        float hcenter;
        float wcenter;

        Pooler pooler;
        for (int h_iter = 0; h_iter < sample_height; ++h_iter) {
            for (int w_iter = 0; w_iter < sample_width; ++w_iter) {
                hcenter = roi_start_h +
                          bin_size_h * (ph + sample_h_rate * (h_iter + 0.5f));
                wcenter = roi_start_w +
                          bin_size_w * (pw + sample_w_rate * (w_iter + 0.5f));
                T val = bilinear_interp(feat_map_ptr, hcenter, wcenter, height, width);
                int idx = h_iter * sample_width + w_iter;
                pooler.feed(val, idx);
            }
        }
        pooler.writeback_val(dst.ptr<T>()[idx]);
        pooler.writeback_idx(index.ptr<dt_int32>()[idx]);
    }
}

template <typename T>
void forward(
        _megdnn_tensor_in src, _megdnn_tensor_in rois, _megdnn_tensor_out dst,
        _megdnn_tensor_out index, const Param& param) {
    using namespace ::megdnn::roi_align;
    switch (param.mode) {
        case param::ROIAlign::Mode::MAX:
            forward_impl<T, MaxPooler<T>>(
                    src, rois, dst, index, param.spatial_scale, param.offset,
                    param.sample_height, param.sample_width);
            break;
        case param::ROIAlign::Mode::AVERAGE:
            forward_impl<T, AveragePooler<T>>(
                    src, rois, dst, index, param.spatial_scale, param.offset,
                    param.sample_height, param.sample_width);
            break;
        default:
            megdnn_assert_internal(false);
    }
}

template <typename T, typename BwdPooler>
void backward_impl(
        _megdnn_tensor_in diff, _megdnn_tensor_in rois, _megdnn_tensor_in index,
        _megdnn_tensor_out grad, float spatial_scale, float offset,
        const int sample_height, const int sample_width) {
    size_t channels = grad.layout[1], hi = grad.layout[2], wi = grad.layout[3];
    size_t pooled_height = diff.layout[2], pooled_width = diff.layout[3];

    size_t total_nr_elems = diff.layout.total_nr_elems();
    int height = hi, width = wi;

    for (size_t idx = 0; idx < total_nr_elems; ++idx) {
        int pw = idx % pooled_width;
        int ph = (idx / pooled_width) % pooled_height;
        int c = (idx / pooled_width / pooled_height) % channels;
        int n = idx / pooled_width / pooled_height / channels;

        auto rois_ptr = rois.ptr<T>() + n * 5;
        int roi_batch_ind = rois_ptr[0];
        float roi_start_w = rois_ptr[1] * spatial_scale - offset;
        float roi_start_h = rois_ptr[2] * spatial_scale - offset;
        float roi_end_w = rois_ptr[3] * spatial_scale - offset;
        float roi_end_h = rois_ptr[4] * spatial_scale - offset;

        float roi_width = std::max(roi_end_w - roi_start_w, ((float)(0.0)));
        float roi_height = std::max(roi_end_h - roi_start_h, ((float)(0.0)));
        float bin_size_h =
                static_cast<float>(roi_height) / static_cast<float>(pooled_height);
        float bin_size_w =
                static_cast<float>(roi_width) / static_cast<float>(pooled_width);

        // regularly sample from a sample_height * sample_width grid
        auto grad_ptr = grad.ptr<T>() + (roi_batch_ind * channels + c) * height * width;
        BwdPooler pooler{ph,    pw,          sample_height, sample_width, height,
                         width, roi_start_h, roi_start_w,   bin_size_h,   bin_size_w};
        pooler.update(
                static_cast<int>(idx), diff.ptr<T>(), index.ptr<dt_int32>(), grad_ptr);
    }
}

template <typename T>
void backward(
        _megdnn_tensor_in diff, _megdnn_tensor_in rois, _megdnn_tensor_in index,
        _megdnn_tensor_out grad, const Param& param) {
    using namespace ::megdnn::roi_align;
    switch (param.mode) {
        case param::ROIAlign::Mode::MAX:
            backward_impl<T, BwdMaxPooler<T>>(
                    diff, rois, index, grad, param.spatial_scale, param.offset,
                    param.sample_height, param.sample_width);
            break;
        case param::ROIAlign::Mode::AVERAGE:
            backward_impl<T, BwdAveragePooler<T>>(
                    diff, rois, index, grad, param.spatial_scale, param.offset,
                    param.sample_height, param.sample_width);
            break;
        default:
            megdnn_assert_internal(false);
    }
}

}  // namespace

namespace megdnn {
namespace naive {

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);
#define cb(DType)                                                                \
    if (src.layout.dtype == DType()) {                                           \
        MEGDNN_DISPATCH_CPU_KERN_OPR(forward<typename DTypeTrait<DType>::ctype>( \
                src, rois, dst, index, param()));                                \
        return;                                                                  \
    }
    MEGDNN_FOREACH_COMPUTING_DTYPE_FLOAT(cb)
#undef cb
    megdnn_throw("bad dtype");
}

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);
#define cb(DType)                                                                 \
    if (diff.layout.dtype == DType()) {                                           \
        MEGDNN_DISPATCH_CPU_KERN_OPR(backward<typename DTypeTrait<DType>::ctype>( \
                diff, rois, index, grad, param()));                               \
        return;                                                                   \
    }
    MEGDNN_FOREACH_COMPUTING_DTYPE_FLOAT(cb)
#undef cb
    megdnn_throw("bad dtype");
}

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