ncnn/src/layer/slice.cpp

// Tencent is pleased to support the open source community by making ncnn available.
//
// Copyright (C) 2017 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.

#include "slice.h"

namespace ncnn {

Slice::Slice()
{
}

int Slice::load_param(const ParamDict& pd)
{
    slices = pd.get(0, Mat());
    axis = pd.get(1, 0);

    return 0;
}

int Slice::forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt) const
{
    const Mat& bottom_blob = bottom_blobs[0];
    int dims = bottom_blob.dims;
    size_t elemsize = bottom_blob.elemsize;
    const int* slices_ptr = slices;

    if (dims == 1) // axis == 0
    {
        int w = bottom_blob.w;

        int q = 0;
        for (size_t i = 0; i < top_blobs.size(); i++)
        {
            int slice = slices_ptr[i];
            if (slice == -233)
            {
                slice = static_cast<int>((w - q) / (top_blobs.size() - i));
            }

            Mat& top_blob = top_blobs[i];
            top_blob.create(slice, elemsize, opt.blob_allocator);
            if (top_blob.empty())
                return -100;

            const unsigned char* ptr = (const unsigned char*)bottom_blob + q * elemsize;
            unsigned char* outptr = top_blob;
            memcpy(outptr, ptr, slice * elemsize);

            q += slice;
        }

        return 0;
    }

    if (dims == 2 && axis == 0)
    {
        int w = bottom_blob.w;
        int h = bottom_blob.h;

        int q = 0;
        for (size_t i = 0; i < top_blobs.size(); i++)
        {
            int slice = slices_ptr[i];
            if (slice == -233)
            {
                slice = static_cast<int>((h - q) / (top_blobs.size() - i));
            }

            Mat& top_blob = top_blobs[i];
            top_blob.create(w, slice, elemsize, opt.blob_allocator);
            if (top_blob.empty())
                return -100;

            int size = w * slice;

            const unsigned char* ptr = bottom_blob.row<const unsigned char>(q);
            unsigned char* outptr = top_blob;
            memcpy(outptr, ptr, size * elemsize);

            q += slice;
        }

        return 0;
    }

    if (dims == 2 && axis == 1)
    {
        int w = bottom_blob.w;
        int h = bottom_blob.h;

        int q = 0;
        for (size_t i = 0; i < top_blobs.size(); i++)
        {
            int slice = slices_ptr[i];
            if (slice == -233)
            {
                slice = static_cast<int>((w - q) / (top_blobs.size() - i));
            }

            Mat& top_blob = top_blobs[i];
            top_blob.create(slice, h, elemsize, opt.blob_allocator);
            if (top_blob.empty())
                return -100;

            #pragma omp parallel for num_threads(opt.num_threads)
            for (int j = 0; j < h; j++)
            {
                unsigned char* outptr = top_blob.row<unsigned char>(j);
                const unsigned char* ptr = bottom_blob.row<const unsigned char>(j) + q * elemsize;
                memcpy(outptr, ptr, slice * elemsize);
            }

            q += slice;
        }

        return 0;
    }

    if (dims == 3 && axis == 0)
    {
        int w = bottom_blob.w;
        int h = bottom_blob.h;
        int channels = bottom_blob.c;

        int q = 0;
        for (size_t i = 0; i < top_blobs.size(); i++)
        {
            int slice = slices_ptr[i];
            if (slice == -233)
            {
                slice = static_cast<int>((channels - q) / (top_blobs.size() - i));
            }

            Mat& top_blob = top_blobs[i];
            top_blob.create(w, h, slice, elemsize, opt.blob_allocator);
            if (top_blob.empty())
                return -100;

            int size = static_cast<int>(bottom_blob.cstep * slice);

            const unsigned char* ptr = bottom_blob.channel(q);
            unsigned char* outptr = top_blob;
            memcpy(outptr, ptr, size * elemsize);

            q += slice;
        }

        return 0;
    }

    if (dims == 3 && axis == 1)
    {
        int w = bottom_blob.w;
        int h = bottom_blob.h;
        int channels = bottom_blob.c;

        int q = 0;
        for (size_t i = 0; i < top_blobs.size(); i++)
        {
            int slice = slices_ptr[i];
            if (slice == -233)
            {
                slice = static_cast<int>((h - q) / (top_blobs.size() - i));
            }

            Mat& top_blob = top_blobs[i];
            top_blob.create(w, slice, channels, elemsize, opt.blob_allocator);
            if (top_blob.empty())
                return -100;

            #pragma omp parallel for num_threads(opt.num_threads)
            for (int p = 0; p < channels; p++)
            {
                int size = w * slice;

                unsigned char* outptr = top_blob.channel(p);
                const unsigned char* ptr = bottom_blob.channel(p).row<const unsigned char>(q);
                memcpy(outptr, ptr, size * elemsize);
            }

            q += slice;
        }

        return 0;
    }

    if (dims == 3 && axis == 2)
    {
        int w = bottom_blob.w;
        int h = bottom_blob.h;
        int channels = bottom_blob.c;

        int q = 0;
        for (size_t i = 0; i < top_blobs.size(); i++)
        {
            int slice = slices_ptr[i];
            if (slice == -233)
            {
                slice = static_cast<int>((w - q) / (top_blobs.size() - i));
            }

            Mat& top_blob = top_blobs[i];
            top_blob.create(slice, h, channels, elemsize, opt.blob_allocator);
            if (top_blob.empty())
                return -100;

            #pragma omp parallel for num_threads(opt.num_threads)
            for (int p = 0; p < channels; p++)
            {
                unsigned char* outptr = top_blob.channel(p);
                const Mat m = bottom_blob.channel(p);

                for (int j = 0; j < h; j++)
                {
                    const unsigned char* ptr = m.row<const unsigned char>(j) + q * elemsize;
                    memcpy(outptr, ptr, slice * elemsize);

                    outptr += slice * elemsize;
                }
            }

            q += slice;
        }

        return 0;
    }

    return 0;
}

} // namespace ncnn