We think the design of the language will support that in the future, but we need to write a GPU-based runtime that lives inside of the runtime and get it to dec

We think the design of the language will support that in the future, but we need to write a GPU-based runtime that lives inside of the runtime and get it to decide when to offload work to GPU first, and that will take time.