It was slower than JITing, but the kinds of J2ME feature phones that ran it weren't using JITing runtimes anyway. And it was for sure faster than interpreter loop they otherwise would have.