The eventual goal is to have a multithreaded implementation of wined3d_cs_ops
and offload some processing to a different thread.