> The result is very fast native code with no GC, no syscalls, and no memory allocations at runtime.

I guess, that's only achievable for certain kind of code, already designed with hard real-time in mind. It would be good to have some information about the limitations of this approach.