Function memcpy_stack

src/core/NEON/kernels/NEStackLayerKernel.cpp:116–146

}

void memcpy_stack(const std::vector<ITensor *> &input, ITensor *output, uint32_t axis, const Window &window)
{
    const int32_t element_size   = input[0]->info()->element_size();
    const int32_t chunk_size     = input[0]->info()->tensor_shape().total_size_lower(axis) * element_size;
    const int32_t num_tensors    = input.size();
    const int32_t out_chunk_step = chunk_size * num_tensors;

    const int32_t start_x = window.x().start();
    const int32_t end_x   = window.x().end();
    const int32_t start_y = window.y().start();
    const int32_t end_y   = window.y().end();

    uint8_t *out_ptr_base = output->buffer() + output->info()->offset_first_element_in_bytes() + start_x * chunk_size;

    for (int32_t x = start_x; x < end_x; ++x)
    {
        const uint8_t *in_ptr =
            input[x]->buffer() + input[x]->info()->offset_first_element_in_bytes() + start_y * chunk_size;
        uint8_t *out_ptr = out_ptr_base + start_y * out_chunk_step;

        for (int32_t y = start_y; y < end_y; ++y)
        {
            std::memcpy(out_ptr, in_ptr, chunk_size);

            in_ptr += chunk_size;
            out_ptr += out_chunk_step;
        }

        out_ptr_base += chunk_size;
    }
}

} // namespace