The programming model for GPUs is pretty different from CPUs. (Branching is discouraged on GPUs, simpler hardware, highly parallel, etc.). I think the right approach is to have separate languages for the two.
Yes, and the question was whether it makes sense for Rust to target the GPU, not whether it's possible. I think most people would agree that in an ideal world you would have a different programming language for the GPU than you have for the CPU because the hardware is so different. CUDA uses C++ (and C and Fortran) because there is already lots of code written in these languages and it makes it easy to port code to run on the GPU.
