Re: OK, what about some resolution (Re: GUILE's GC - why we struggling to solve already solved problems?)

[Michael Livshin <mlivshin at bigfoot dot com>]

Clark McGrew <mcgrew@ale.physics.sunysb.edu> writes:

> >>>>> "Michael" == Michael Livshin <mlivshin@bigfoot.com> writes:
> 
>     Michael> * current Guile GC sucks.
> 
> Really?  That's news to me.

OK, so much for using relative terms and not saying that I do.

> The current evidence is that GGC is 50% faster than BGC.  Obviously
> BGC can be tuned and made more precise, it will always need to scan
> more than GGC.

my hunch is that, properly integrated, BGC will be faster then the
current GC, due to various algorithmic reasons (lazy sweep,
"generational" collection, explicit markstack, ...).

> I guess it's worth pointing out that BGC uses a VM barrier to
> implement generational GC.  VM barriers were considered for
> Generational GC in guile and discarded because they were far too
> inefficient.

they are better than nothing.  but yes, I *believe* that SW write
barrier would be even better.  apparently, BGC has some hooks to
implement that -- the adventurous people who want to try integrating
BGC with Guile should look into it.

-- 
programmer, n:
        A red eyed, mumbling mammal capable of conversing with inanimate
        monsters.