If you look at the LLVM documentation there is still a lot to do to be able to support most performant GC algorithms.