https://hgpu.org/?p=1536
A middleware for efficient stream processing in CUDA