Explain the architecture of a video decoding pipeline. Where would you offload processing to hardware accelerators?
