I'm speeding up the JVM interop code in SCI (and hence bb) with 3x-5x (as a prep for more FFI stuff to come)