This is the mail archive of the
mailing list for the Guile project.
Re: Is guile byte-code compiled?
Miroslav Silovic <firstname.lastname@example.org> writes:
> Well, I managed to push SART through hobbit (the only thing that
> needed to change was #i(.....) -> something else - I'm not sure if
> hobbit can work with literals now). The result was under 50% in all
> the cases I tried - SART spends REALLY lots of time in C.
There's not really much hobbit can do there... still, ~50% isn't bad, either.
> In my opinion, the only other way to measurably improve SART speed
> would be gengc - the C code still conses a lot.
Yeah... I'm planning on getting back to it RSN(tm); the problem is, I
got caught in a sort of recursive tasklist: move stuff from scheme->c;
get annoyed by how easy it is to make mistakes swapping longs with SCM
values (and not testing predicate values correctly); working on
changing SCM to an undefined pointer, and making SCM_BOOL_F== c's
false; run into the grossness of the current numbers implementation;
start working on a new numbers implementation, which is where I'm at
right now. Once I can unwind this (pretty soon, hopefully... the
numbers stuff is mostly finished, 'cept for the glue), I'm getting
back to the gc :)