I think re-targeting bitcode for a different architecture with the same pointer size and endinness is a pretty light weight operation, but I might be wrong. I'

I think re-targeting bitcode for a different architecture with the same pointer size and endinness is a pretty light weight operation, but I might be wrong. I'm pretty sure it's orders of magnitude lower overhead and less complicated than dynamically recompiling native code.