we won't get much faster than that (though I expect that a parse that can directly emit Bytecode should be a little bit faster still because more compact representation), but the Rust rewrite still needs to be at least as fast as now and that's no small feat