This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: Advice requested: how big can we be?



On Tue, 16 Sep 2003, Paul Koning wrote:

> >>>>> "S" == S Bosscher <S.Bosscher@student.tudelft.nl> writes:
>
>  >> The current libgcj build requires a process size > 256M bytes.
>
>  S> Actually, its more than 370MB for mainline -- I had a look at this
>  S> yesterday and I was shocked.  Why is it so big???
>
>  >> Is this unreasonable?  Should we be able to build on boxes with
>  >> less swap than that?  Or shall we just say "512M swap or don't
>  >> bother building libgcj" ?
>
>  S> Please no! It's bad enough as it is.
>
>  S> There still are people who try to make their contribution to GCC
>  S> but who don't have the latest and greatest from the hardware
>  S> stores.
>
>  S> I have an Athlon XP2000 with 265MB and 512MB swap, and I like to
>  S> think this is a decent machine.  Certainly that should be enough
>  S> to build GCC without trouble.  But building libjava more than
>  S> doubles my bootstrap time and my computer is swapping like mad
>  S> when it's building the lib (making it impossible to bootstrap in
>  S> the background, for example).  A bootstrap without Java doesn't
>  S> cause any difficulty at all.
>
> Same here.  Now I understand why bootstrapping takes such an insane
> amount of time.
The answer is to try to keep some of the trees/structures we need on disk.
We obviously aren't hitting all these trees at once, even if we have them
allin memory.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]