Can't we merge those two exhaustive lists (the first is only used in
the non-threaded case and the other only in the threaded case).
If we turn the first set of #defines into an enum, we should be able to
bring the syntax of the two declarations close enough that the
difference can be abstracted out into a few macros.
> FWIW I suspect a bigger win would be had by applying this treatment --
> or even better, a full threaded interpreter -- to the regexp matcher.
> Back when I did some profiling of Emacs, the regexp matcher was higher
> in my profiles than the bytecode interpreter.
There's a lot of room for improvement in the regexp-matcher as well,
yes. If we could generate the code via something like vmgen it would be
even better.