Could CloudHaskell's API also apply to OpenCL so the code same code could run on the GPU? A parallel algorithms library https://github.com/thrust/thrust would be nice to have it for CloudHaskell and OpenCL.