Very nice! I look forward to trying this out. There a few issues which come to mid though: 1. How do these benchmark results compare to writing x64_64 impleme

Very nice! I look forward to trying this out. There a few issues which come to mid though: 1. How do these benchmark results compare to writing x64_64 implementations directly? That is, is it enough to reach OpenCL-level performance? 2. What about if I want to use both a CPU and and a GPU? It seems that would not be supported, as this library produces a `libcuda.so`. 3. What about AMD chips? 4. The README says: > Authors of CUDA benchmarks used CUDA functions atomicInc > and atomicDec which have direct hardware support on NVIDIA > cards, but no hardware support on Intel cards. They have > to be emulated in software, which limits performance is there really no support, not even in the near future, for atomic increment on Intel chips? 5. This is all written in Rust. It's a bit fishy to me for a C library to be implemented in Rust, though maybe I'm just a little prejudiced. 6. Where is the documentation of the semantic differences between proper CUDA and ZLUDA? e.g. - what do streams do? How do events work? etc.