I do not quite understand the usefulness of such a project. So you are compiling whatever comes from a LLVM frontend to OpenCL code and run it on a GPU. For which reason? So you can transform code...