Method run

src/core/NEON/kernels/NEReorgLayerKernel.cpp:115–158

}

void NEReorgLayerKernel::run(const Window &window, const ThreadInfo &info)
{
    ARM_COMPUTE_UNUSED(info);
    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICPPKernel::window(), window);

    const DataLayout data_layout = _input->info()->data_layout();
    const size_t     idx_w       = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
    const size_t     idx_h       = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
    const size_t     idx_c       = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);

    const unsigned int stride = _stride;
    const unsigned int out_c  = _output->info()->tensor_shape()[idx_c] / (stride * stride);
    const uint8_t     *in_ptr = _input->buffer();

    // Collapse
    Window collapsed_window = window.collapse_if_possible(window, 4);

    // Create Iterator
    Iterator out(_output, collapsed_window);

    // Perform reorg
    execute_window_loop(
        collapsed_window,
        [&](const Coordinates &id)
        {
            // Get spatial coords and channels
            const unsigned int w = id[idx_w];
            const unsigned int h = id[idx_h];
            const unsigned int c = id[idx_c];

            // Calculate mapping
            const unsigned int offset     = c / out_c;
            Coordinates        map_coords = id;
            map_coords.set(idx_w, w * stride + offset % stride);
            map_coords.set(idx_h, h * stride + offset / stride);
            map_coords.set(idx_c, c % out_c);

            // Perform mapping
            std::memcpy(out.ptr(), in_ptr + _input->info()->offset_element_in_bytes(map_coords),
                        _input->info()->element_size());
        },
        out);
}
} // namespace arm_compute