NVIDIA CUDA Code In A JIT Interpreted Manner Via Cling
Based on Clang's CUDA support, university students have been working on adding CUDA support to Cling so it can be interpreted at run-time -- basically JIT'ed for the GPU being used. This is believed to be the first interpreter out there for the CUDA runtime API. They are using it for some interesting use-cases including working on CUDA-based code off notebooks.
Those curious about this experimental support can see this slide deck, which was sent into us this morning. Those unfamiliar with the Cling C++ interpreter can find it on GitHub.