Speeding up ggc-simple on stage1
Jeffrey A Law
law@cygnus.com
Fri Jan 14 10:10:00 GMT 2000
In message < orbt6oykn5.fsf@garnize.lsd.ic.unicamp.br >you write:
> > Presumably that host can't use gcc-page?
>
> Yep. configure claims the platform doesn't support mmap nor valloc :-(
Actually, I've found valloc is not a good thing either. It effectively
doubles the amount of memory we need if I remember correctly. I just never
got around to writing up a report on that and asking for suggestions about
what to do :-)
> > Yes, ggc-simple is painfully slow. It took several days to
> > bootstrap an m68k box I've got access to using gcc-simple.
>
> Well, the unoptimized stage1 cc1 has been trying to build tree.c for
> the past 24 hours or so :-(
>
> Nevertheless, `ps' says it's used just 54 minutes of CPU. So it looks
> like the system is thrashing, which is not that surprising, given that
> it's got just 16MB of memory :-( :-(
Yea, you're thrashing. The only aspect of this change that might help that
is a smaller call chain within the garbage collector.
Another way to test it would be to use ggc-simple on a more beefy box
(there's a configure option to tweak the garbage collection selection).
> > Not without at least knowing it's an improvement. I'm not talking about
> > a massive study of its performance. Build cc1 with and without this
> > change and run a substantial file through the compiler.
>
> Should it be an optimized or unoptimized cc1? Can it be cc1 from
> stage1, built with gcc 2.95.2?
Your choice. If it's an improvement I'd expect it to show up regardless of
those issues.
jeff
More information about the Gcc-patches
mailing list